FUNDAMENTAI~ OF INORGANIC MEMBRANE SCIENCE AND TECHNOLOGY
This Page Intentionally Left Blank
M e m b r a n e Science and Technology Series, 4
F U N D A M E N T A L S OF INORGANIC MEMBRANE SCIENCE AND TECHNOLOGY Edited by
A.J. Burggraaf Laboratory of Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands and
L. Cot Laboratory des Materiaux et Proc#des Membranes, (UMR 9987 CNRS-ENSCM-4411), Ecole Nationale Sup#rieure de Chimie 8 Rue de I'Ecole de Chimie, 34053 Montpellier, France
1996 ELSEVIER Amsterdam ~
Lausanne -- New York ~
Oxford ~
Shannon ~
Tokyo
ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211, 1000 AE Amsterdam, The Netherlands
ISBN 0-444-81877-4
9 1996 ELSEVIER SCIENCE B.V. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher, Elsevier Science B.V., Copyright & Permissions Department, P.O. Box 521, 1000 AM Amsterdam, The Netherlands. Special regulations for readers in the U.S.A.: This publication has been registered with the Copyright Clearance Center Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923. Information can be obtained from the CCC about conditions under which photocopies of parts of this publication may be made in the U.S.A. All other copyright questions, including photocopying outside the U.S.A., should be referred to the copyright owner, Elsevier Science B.V., unless otherwise specified. No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a m a t t e r of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. This book is printed on acid-free paper. Printed in The Netherlands
Preface Inorganic membrane science and technology is a new field of membrane separation technology which until recently was dominated by the earlier field of polymer membranes. Currently the subject is undergoing rapid development and innovation. Asymmetric inorganic membranes for microfiltration of liquids were first commercialized in the early 1980s and today form the largest application in the agro-food, biotechnology, pharmaceutical and other industries. Other types of inorganic membranes and separation processes (i.e., gas separation, nanofiltration, membrane reactors) have also been developed on a laboratory scale. The commercial exploitation of some of these developments is feasible although it has not yet been realized. This situation is hardly surprising when the history of polymer membrane development is considered. Economically competitive, large-scale industrial applications were only achieved 20 to 25 years after the initial "break-through" development of asymmetric polymer membranes in the early 1950s. It was only recently that the first International Conference on Inorganic Membranes (ICIM-1) was organised in Montpellier (France) in 1989. This Conference has subsequently become a bi-annual event, thus demonstrating the rapid growth in the field of inorganic membranes. The first text on the science and technology of inorganic membranes was the book edited by R.R. Bhave entitled "Inorganic Membranes". This was published in 1991 and contained contributions from several well known experts in the field. This was followed in 1995 by the book by R.D. Noble and S.A. Stern "Membrane Separation Technology" which was devoted to separation properties with the main focus on polymer membranes. The present book is intended to describe the fundamental principles of both synthesis of inorganic membranes and membrane supports and also the associated phenomena of transport and separation in a semi-quantitative form. It is hoped that this part of the text will form a basis that will not require major revision in the foreseeable future. Examples are given which illustrate the state of the art in the synthesis and properties of membranes with controlled properties. Where possible, future possibilities and limitations are discussed and potential areas for future innovation indicated. Because the field of membrane science and technology is broad and multidisciplinary in character it has been necessary to give (semi) quantitative descriptions of different aspects in a concise form and provide the reader with references to more extended treatments in the literature. It is nevertheless not the intention to provide a complete overview of all the relevant literature but rather
vi to focus on those types of membranes and developments which are foreseen to have the most important impact in future applications. Examples which illustrate this aspect for future development have therefore been emphasized. The various chapters differ significantly in length, emphasis and depth and, in order to retain the perspectives of the different contributors, no uniformity has been sought in the presentation of the subject matter. For those chapters where the authors have used English as a second language, the editing has been the responsibilty of the publisher. By combining aspects of both the science and technology of inorganic membranes it is hoped that this book will serve as a useful source of information equally for scientists and engineers working in this field. Finally the book will also provide some of the insights of important investigators who have contributed to the development of this subject. It should be noted that there has truly been a great renewal in the knowledge of the synthesis of ceramic membranes and the mechanisms of the separation process.
Acknowledgements A.J. Burggraaf would like to acknowledge A. Nijmeijer, Mrs C. Twilt-van Giessen and Dr. G.M.H. van der Velde for help with many editorial questions, and also a number of colleagues and coworkers especially H. Verweij and B.N. Nair for their support in preparing particular text sections.
A.J. Burggraaf L. Cot
vii
Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v xvii
Chapter 1. GENERAL OVERVIEW, TRENDS AND PROSPECTS A.J. Burggraaf and L. Cot 1.1 1.2 1.3
Introduction ....................................... Market Situation and Prospects ............................ M a i n Barriers to T e c h n o l o g i c a l D e v e l o p m e n t a n d A c c e p t a n c e . . . . . . . . . . 1.3.1 R e q u i r e m e n t s a n d Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1.1 C h a r a c t e r i s t i c s of c e r a m i c fabrication . . . . . . . . . . . . . . 1.3.2 State of t h e A r t a n d N e e d s . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2.1 A v a i l a b i l i t y a n d cost . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2.2 Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2.3 L o n g - t e r m stability . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2.4 Surface area to v o l u m e ratio . . . . . . . . . . . . . . . . . . . 1.3.2.5 Specific c o m b i n a t i o n s of h i g h s e p a r a t i o n factors a n d high permeation .......................... 1.4 T r e n d s , T e c h n o l o g i c a l a n d Scientific P r o s p e c t s . . . . . . . . . . . . . . . . . . . 1.4.1 I n f r a s t r u c t u r e for F u t u r e W o r k . . . . . . . . . . . . . . . . . . . . . . . 1.4.2 Some Trends ................................. 1.4.3 P r o s p e c t s for I n t e r e s t i n g M e m b r a n e A p p l i c a t i o n s . . . . . . . . . . . 1.4.3.1 Industrial production processes ................ 1.4.3.2 Energy-related applications .................... 1.4.3.3 Environmental applications ................... 1.4.3.4 Others ............................... 1.4.4 I n t e r e s t i n g Fields for F u t u r e R & D . . . . . . . . . . . . . . . . . . . . . 1.4.4.1 L o n g - t e r m c h e m i c a l stability . . . . . . . . . . . . . . . . . . 1.4.4.2 Thin-layer deposition technology ............... 1.4.4.3 Support technology ....................... 1.4.4.4 M i c r o p o r o u s m e m b r a n e s for g a s / v a p o u r s e p a r a t i o n . . . . 1.4.4.5 Nanofiltration membranes ................... 1.4.4.6 D e n s e ( n o n - p o r o u s ) m e m b r a n e s a n d surface r e a c t i o n limitation ............................. 1.4.4.7 Mixed (hybrid) processes and materials . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 2 3 3 4 5 5 6 6 7 7 9 9 10 12 12 13 13 14 14 15 15 15 16 17 18 19 19
Chapter 2. IMPORTANT CHARACTERISTICS OF INORGANIC MEMBRANES A.J. Burggraaf 2.1 2.2
Introduction ...................................... T y p e s of I n o r g a n i c M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . . . .
21 21
viii 2.3
2.4 2.5
Microstructural Pore and Pore Network Characteristics 2.3.1 Modified Structures .............................
............
2.3.2 Supports ................................... A r c h i t e c t u r e of M e m b r a n e S y s t e m s . . . . . . . . . . . . . . . . . . . . . . . .
27 27
Some General Characteristics ............................ 2.5.1
Commercially Available Inorganic Membranes
23 26
31 .............
2.6 Considerations on Chemical Resistance ...................... References ..........................................
31 33 34
Chapter 3. ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS Yi Hua Ma 3.1
Introduction ......................................
35
3.2
Adsorption Isotherms
36
3.3
3.4
3.5
................................
3.2.1
T y p e s of I s o t h e r m s
3.2.2
The L a n g m u i r Isotherm . . . . . . . . . . . . . . . . . . . . . . . . . . .
.............................
37
3.2.3
T h e BET I s o t h e r m s
40
.............................
36
3.2.4
I s o t h e r m s D e r i v e d f r o m t h e E q u a t i o n of State . . . . . . . . . . . . . .
41
3.2.5
The Potential Theory
42
Experimental Techniques
............................
..............................
D e t e r m i n a t i o n of A d s o r p t i o n I s o t h e r m s
3.3.2
Surface Area Determinations . . . . . . . . . . . . . . . . . . . . . . . .
46
3.3.3
P o r e Size D i s t r i b u t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
Adsorption on Membranes
.................
43
3.3.1
.............................
3.4.1
A d s o r p t i o n of G a s e s o n M i c r o p o r o u s Silica M e m b r a n e s a n d
3.4.2
Interrelation between Adsorption and Permeation ........... A d s o r p t i o n on Sol-Gel Derived Ceramic M e m b r a n e s . . . . . . . . . .
3.4.3 Liquid Adsorption on Membranes Summary .......................................
References
....................
..........................................
43
57 57 60 61 64 64
Chapter 4. METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE IN MEMBRANE MATERIALS A. Julbe and J.D.F. Ramsay 4.1 4.2
4.3
General Introduction .................................
67
D e s c r i p t i o n of P o r o u s M a t e r i a l s - - D e f i n i t i o n s . . . . . . . . . . . . . . . . . .
69
4.2.1
O r i g i n of P o r e S t r u c t u r e
..........................
4.2.2 4.2.3
Q u a n t i t a t i v e D e s c r i p t i o n of P o r e S t r u c t u r e s . . . . . . . . . . . . . . . M o d e l s for P o r o u s S t r u c t u r e s . . . . . . . . . . . . . . . . . . . . . . .
69 70 72
Static C h a r a c t e r i s a t i o n T e c h n i q u e s . . . . . . . . . . . . . . . . . . . . . . . . .
74
4.3.1
Stereology
74
4.3.2
Intrusive Methods ..............................
4.3.3
..................................
4.3.2.1 4.3.2.2
Mercury porosimetry ...................... Gas adsorption/desorption isotherms (physisorption)
4.3.2.3 4.3.2.4
Calorimetric determinations Nuclear magnetic resonance
.................. ..................
Non-intrusive Methods ...........................
78 78 . . . 78 84 87 91
ix 4.3.3.1 R a d i a t i o n scattering . . . . . . . . . . . . . . . . . . . . . . . 4.3.3.2 Wave propagation ........................ 4.3.3.3 I o n - b e a m analysis . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3.4 Positron lifetime s p e c t r o s c o p y . . . . . . . . . . . . . . . . . 4.4 D y n a m i c Characterisation T e c h n i q u e s . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Rejection M e a s u r e m e n t s . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Liquid D i s p l a c e m e n t T e c h n i q u e s . . . . . . . . . . . . . . . . . . . . . 4.4.2.2 L i q u i d / g a s m e t h o d s (bubble point, liquid e x p u l s i o n permporometry) .. ....................... 4.4.2.3 Liquid-liquid d i s p l a c e m e n t p o r o s i m e t r y (or biliquid permporometry) ........................ 4.4.3 Fluid F l o w M e a s u r e m e n t s . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.1 Liquid p e r m e a b i l i t y . . . . . . . . . . . . . . . . . . . . . . 4.4.3.2 Gas p e r m e a b i l i t y . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3.3 Permporometry ........................ 4.5 Conclusion and Recommendations ....................... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91 94 96 97 98 98 99 99 101 102 102 103 104 106 110
Chapter 5. CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS A. Larbot 5.1 5.2
Introduction ..................................... Extrusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 C e r a m i c Paste P r e p a r a t i o n . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 T u b e Shaping, D r y i n g a n d Firing . . . . . . . . . . . . . . . . . . . . 5.2.3 A n E x a m p l e of P r e p a r a t i o n . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Tape C a s t i n g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Slurry P r e p a r a t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 S h a p i n g a n d Flat Ceramics . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Specific C h a r a c t e r i z a t i o n M e t h o d s for S u p p o r t s . . . . . . . . . . . . . . . . 5.4.1 Bubble Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Mechanical Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2.1 Burst p r e s s u r e (test for tubes) . . . . . . . . . . . . . . . . . 5.4.2.2 B e n d i n g s t r e n g t h (test for cylindrical specimen) . . . . . . 5.5 Conclusion ..................................... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
119 121 121 124 127 130 131 133 134 134 136 136 136 137 138
Chapter 6. PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING B.C. Bonekamp 6.1 6.2
Introduction ..................................... S u p p o r t s for C e r a m i c M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 The Multilayer S u p p o r t S y s t e m . . . . . . . . . . . . . . . . . . . . . 6.2.2 Support Requirements .......................... 6.2.3 Layer F o r m a t i o n on P o r o u s Substrates . . . . . . . . . . . . . . . . . 6.2.4 S u s p e n s i o n s a n d Sols . . . . . . . . . . . . . . . . . . . . . . . . . . .
141 143 143 146 150 159
6.2.4.1 Van der Waals attraction . . . . . . . . . . . . . . . . . . . 6.2.4.2 Electrostatic interaction . . . . . . . . . . . . . . . . . . . . 6.2.4.3 Polymeric interaction . . . . . . . . . . . . . . . . . . . . . 6.2.4.4 Rheology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.5 Drying a n d Sintering of Particulate Coatings . . . . . . . . . . . . . 6.2.6 Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Dip-coating with Porous Substrates . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Capillary Colloidal Filtration . . . . . . . . . . . . . . . . . . . . . . . 6.3.1.1 C o n t i n u u m description . . . . . . . . . . . . . . . . . . . . 6.3.2 Film-coating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2.1 Coating flow dynamics . . . . . . . . . . . . . . . . . . . . 6.3.2.2 Closer examination . . . . . . . . . . . . . . . . . . . . . . . 6.3.2.3 Substrate wetting and dewetting . . . . . . . . . . . . . . . 6.3.2.4 Stability of liquid coatings . . . . . . . . . . . . . . . . . . . 6.3.3 Macromolecular Thickeners and Binders . . . . . . . . . . . . . . . . 6.3.4 C o m p a c t (Cake) Structure . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements .................................... List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
163 164 166 171 175 178 183 183 184 189 190 191 195 200 203 207 209 218 218 219 221
Chapter 7. SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING Christian Guizard 7.1 7.2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pore Formation in Sol-Gel Derived Ceramic M e m b r a n e s . . . . . . . . . . . 7.2.1 Packing of Colloidal Particles . . . . . . . . . . . . . . . . . . . . . . 7.2.2 Aggregation of Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.3 Utilization of Template Agents . . . . . . . . . . . . . . . . . . . . . 7.3 Colloidal Suspensions to Prepare Mesoporous M e m b r a n e s . . . . . . . . . 7.3.1 C h e m i s t r y of Colloidal Sols . . . . . . . . . . . . . . . . . . . . . . . 7.3.2 Examples of M e m b r a n e Preparation . . . . . . . . . . . . . . . . . . 7.4 Inorganic Polymers to Prepare Microporous M e m b r a n e s . . . . . . . . . . . 7.4.1 Formation and Aggregation of Clusters . . . . . . . . . . . . . . . . 7.4.2 Examples of M e m b r a n e Preparation . . . . . . . . . . . . . . . . . . 7.5 The Concept of N a n o p h a s e Ceramics Applied to the Preparation of Microporous M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5.1 Formation and Coating of A q u e o u s Nanoparticulate Sols . . . . . . 7.5.2 Formation and Coating of Organic Nanoparticulate Sols . . . . . . 7.6 Tailor-made Porous M e m b r a n e s via Templates Containing Systems . . . . 7.6.1 Utilization of Amphiphilic Media . . . . . . . . . . . . . . . . . . . . 7.6.2 Insertion of Organic and Inorganic Entities or P o l y m e r Particles in Gel Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
227 229 229 230 231 232 232 233 237 237 239 240 240 243 245 246 251 254 255
xi
Chapter 8. FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING A.J. Burggraaf 8.1
Synthesis a n d Processing of S u p p o r t e d M e s o p o r o u s M e m b r a n e s . . . . . . 259 8.1.1 Introduction ................................ 259 8.1.2 Film F o r m a t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 8.1.2.1 Initial layer f o r m a t i o n . . . . . . . . . . . . . . . . . . . . . 260 8.1.2.2 M e s o p o r o u s film f o r m a t i o n . . . . . . . . . . . . . . . . . . 261 8.1.2.3 Theoretical aspects of the d r y i n g process from lyogel to xerogel film . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 8.1.2.4 C o n s o l i d a t i o n to the final m e m b r a n e s t r u c t u r e b y heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 8.1.3 Illustrative E x p e r i m e n t a l O b s e r v a t i o n s of Stress a n d C r a c k i n g in Membranes ................................. 283 8.1.3.1 Stress m e a s u r e m e n t s in s u p p o r t e d p o r o u s m e m b r a n e s . . 283 8.1.3.2 D r y i n g characteristics of m e m b r a n e s . . . . . . . . . . . . 287 8.1.3.3 Stress a n d cracking in m e m b r a n e s d u r i n g d r y i n g . . . . . 287 8.1.3.4 Stress f o r m a t i o n in m e m b r a n e s d u r i n g calcination . . . . 291 8.1.3.5 A m o d e l discussion of stress a n d a v o i d i n g cracking . . . . 292 8.1.4 T h e r m a l Stability of Ceramic M e m b r a n e s . . . . . . . . . . . . . . . 297 8.2 Synthesis a n d Processing of S u p p o r t e d M i c r o p o r o u s M e m b r a n e s . . . . . . 298 8.2.1 M i c r o p o r o u s M e m b r a n e s O b t a i n e d b y S o l - G el Processes . . . . . . 298 8.2.1.1 I n t r o d u c t i o n a n d o v e r v i e w of film f o r m a t i o n . . . . . . . . 298 8.2.1.2 I m p o r t a n t p a r a m e t e r s in p r e c u r s o r synthesis . . . . . . . . 300 8.2.1.3 Illustrative e x a m p l e s of m e m b r a n e synthesis an d microstructure development . . . . . . . . . . . . . . . . . 306 8.2.2 M i c r o p o r o u s M e m b r a n e s O b t a i n e d by CVD . . . . . . . . . . . . . . 310 8.2.2.1 CVD m e t h o d s . . . . . . . . . . . . . . . . . . . . . . . . . . 310 8.2.2.2 O t h e r m e t h o d s a n d m i c r o p o r o u s m e m b r a n e s y s t e m s . . . 312 8.2.3 Zeolite M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312 8.2.3.1 O v e r v i e w a n d i n t r o d u c t i o n to zeolite c h e m i s t r y . . . . . . 312 8.2.3.2 Illustrative e x a m p l e s of zeolite m e m b r a n e synthesis and processing . . . . . . . . . . . . . . . . . . . . . . . . . 317 8.3 Conclusions and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
Chapter 9. TRANSPORTAND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS A.J. Burggraaf 9.1
9.2
Introduction ..................................... 9.1.1 C h a p t e r O u t li n e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1.2 O v e r v i e w of I m p o r t a n t Points . . . . . . . . . . . . . . . . . . . . . 9. Gas T r a n s p o r t in Simple M e m b r a n e Structures . . . . . . . . . . . . . . . . . 9.2.1 I m p o r t a n t Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.2 Pore Characteristics a n d M e m b r a n e Architecture . . . . . . . . . . . 9.2.3 Single Gas P e r m e a t i o n in M a c r o p o r o u s a n d M e s o p o r o u s Systems . 9.2.3.1 Viscous flow . . . . . . . . . . . . . . . . . . . . . . . . . .
331 331 332 333 333 335 337 337
xii 9.2.3.2 K n u d s e n diffusion a n d the transition region . . . . . . . . 338 9.2.3.3 Surface diffusion a n d capillary c o n d e n s a t i o n . . . . . . . . 345 9.2.4 P e r m e a t i o n in binary gas mixtures in m a c r o p o r o u s a n d mesoporous membranes ......................... 355 9.2.4.1 General considerations . . . . . . . . . . . . . . . . . . . . 355 9.2.4.2 K n u d s e n diffusion . . . . . . . . . . . . . . . . . . . . . . . 357 9.2.4.3 Viscous Flow a n d the Transition Region . . . . . . . . . . 357 9.3 Separation of Binary Mixtures in Simple M e s o p o r o u s M e m b r a n e s . . . . . 364 9.3.1 I m p o r t a n t Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364 9.3.2. Separation in the K n u d s e n and Transition Regions . . . . . . . . . . 365 9.3.3 Separation with Surface Diffusion and Capillary C o n d e n s a t i o n . . 368 9.4 P e r m e a t i o n a n d Separation in M i c r o p o r o u s M e m b r a n e s . . . . . . . . . . . 374 9.4.1 Introduction a n d I m p o r t a n t Concepts . . . . . . . . . . . . . . . . . 374 9.4.2 P h e n o m e n o l o g i c a l Description of Single Gas P e r m e a t i o n . . . . . . 377 9.4.2.1 Qualitative description of gas p e r m e a t i o n . . . . . . . . . 378 9.4.2.2 Quantitative description of gas p e r m e a t i o n a n d separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382 9.4.2.3 P e r m e a t i o n a n d separation in binary (ternary) gas mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386 9.4.2.4 Illustrative examples of p e r m e a t i o n a n d separation with m i c r o p o r o u s m e m b r a n e s . . . . . . . . . . . . . . . . 391 9.4.4 Surface Effects on P e r m e a t i o n in Microporous M e m b r a n e s . . . . . 411 9.5 P e r m e a t i o n a n d Separation in More Complicated Systems . . . . . . . . . . 413 9.5.1 H o l l o w Fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413 9.5.2 Multilayered, A s y m m e t r i c S u p p o r t e d Systems . . . . . . . . . . . . 414 9.6 O v e r v i e w of I m p o r t a n t Results . . . . . . . . . . . . . . . . . . . . . . . . . . 416 9.6.1 I n t r o d u c t o r y Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 417 9.6.2 Typical P e r m e a t i o n a n d Separation Data for Porous M e m b r a n e s . . 418 9.6.3 C o m p a r i s o n of P e r m e a t i o n a n d Separation Data of Porous a n d Dense M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422 9.7 Conclusions a n d Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . 424 List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Chapter 10. DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION H.J.M. Bouwmeester and A.J. Burggraaf 10.1
Introduction
10.2
General Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2.1 Major M e m b r a n e Concepts . . . . . . . . . . . . . . . . . . . . . . . . 10.2.2 Data: O x y g e n Permeability of Solid Oxide M e m b r a n e s 10.2.3 Factors Controlling O x y g e n P e r m e a t i o n . . . . . . . . . . . . . . . . Fundamentals .................................... 10.3.1 Bulk T r a n s p o r t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.3.1.1 W a g n e r equation . . . . . . . . . . . . . . . . . . . . . . . . 10.3.1.2 Chemical diffusion coefficient . . . . . . . . . . . . . . . .
10.3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.......
435
436 436 440 448 449 449 449 451
xiii 10.3.1.3 T r a p p i n g of electronic a n d ionic defects . . . . . . . . . . . 10.3.1.4 E m p i r i c a l e q u a t i o n s . . . . . . . . . . . . . . . . . . . . . . 10.3.2 Surface O x y g e n E x c h a n g e . . . . . . . . . . . . . . . . . . . . . . . . 10.3.2.1 Characteristic m e m b r a n e t h i c k n e s s Lc . . . . . . . . . . . . 10.3.2.2 M e a s u r i n g Lc . . . . . . . . . . . . . . . . . . . . . . . . . . 10.3.2.3 T h e effect of surface r o u g h n e s s a n d p o r o s i t y . . . . . . . . 10.4 Solid O x i d e Electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4.1 I n t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4.2.1 Diffusion of electronic c h a r g e carriers . . . . . . . . . . . . 10.4.2 O x y g e n S e m i - p e r m e a b i l i t y of O x i d e Electrolytes . . . . . . . . . . . 10.4.2.2 M o d e l l i n g e q u a t i o n s . . . . . . . . . . . . . . . . . . . . . . 10.4.2.3 E x a m p l e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4.3 E l e c t r o c h e m i c a l O x y g e n S e p a r a t i o n . . . . . . . . . . . . . . . . . . . 10.4.3.1 O x y g e n p u m p . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4.3.2 D u a l - p h a s e c o m p o s i t e s . . . . . . . . . . . . . . . . . . . . 10.5 I n t r o d u c i n g Electronic C o n d u c t i o n in F l u o r i t e - t y p e O x y g e n I o n Conductors ...................................... 10.5.1 I n t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.5.2 Defect C h e m i s t r y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.5.3 E x a m p l e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6 A c c c e p t o r - d o p e d P e r o v s k i t e a n d P e r o v s k i t e - r e l a t e d O x i d e s . . . . . . . . . 10.6.1 I n t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6.2 S t r u c t u r e a n d Defect C h e m i s t r y . . . . . . . . . . . . . . . . . . . . . 10.6.2.1 P e r o v s k i t e S t r u c t u r e . . . . . . . . . . . . . . . . . . . . . . 10.6.2.2 N o n s t o i c h i o m e t r y . . . . . . . . . . . . . . . . . . . . . . . 10.6.2.3 L o c a l i z e d v e r s u s D e l o c a l i z e d Electrons . . . . . . . . . . . 10.6.3 O x y g e n D e s o r p t i o n a n d P e r o v s k i t e Stability . . . . . . . . . . . . . . 10.6.4 E q u a t i o n s for O x y g e n T r a n s p o r t . . . . . . . . . . . . . . . . . . . . . 10.6.5 Electronic C o n d u c t i v i t y . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6.6 E x t e n d e d Defects a n d V a c a n c y O r d e r i n g . . . . . . . . . . . . . . . . 10.6.6.1 Static Lattice S i m u l a t i o n . . . . . . . . . . . . . . . . . . . . 10.6.6.2 V a c a n c y O r d e r i n g . . . . . . . . . . . . . . . . . . . . . . . 10.6.6.3 M i c r o d o m a i n F o r m a t i o n . . . . . . . . . . . . . . . . . . . 10.6.6.4 B r o w n m i l l e r i t e S t r u c t u r e . . . . . . . . . . . . . . . . . . . 10.6.6.5 H i g h T e m p e r a t u r e N M R . . . . . . . . . . . . . . . . . . . 10.6.7 O b s e r v a t i o n s f r o m P e r m e a b i l i t y M e a s u r e m e n t s . . . . . . . . . . . . 10.6.7.1 SrCo0.8Fe0.203-6 . . . . . . . . . . . . . . . . . . . . . . . . . 10.6.7.2 E x p e r i m e n t a l difficulties . . . . . . . . . . . . . . . . . . . 10.6.7.3 Surface e x c h a n g e kinetics . . . . . . . . . . . . . . . . . . . 10.6.7.4 B e h a v i o u r in l a r g e Po2-gradients . . . . . . . . . . . . . . . 10.6.7.5 G r a i n b o u n d a r y diffusivity . . . . . . . . . . . . . . . . . . 10.7 Final R e m a r k s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements .................................... List of A b b r e v i a t i o n s a n d S y m b o l s . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
453 454 455 456 459 461 462 462 463 463 464 465 469 469 470 472 472 472 475 479 479 482 482 483 486 488 489 492 495 495 497 498 499 500 502 502 503 506 507 508 510 513 513 515
xiv
Chapter 1 I. CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS Jose Sanchez and Theodore T. Tsotsis 11.1 11.2
Introduction ..................................... D e n s e M e t a l M e m b r a n e Reactors . . . . . . . . . . . . . . . . . . . . . . . . . 11.2.1 C o s t a n d A v a i l a b i l i t y . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2.2 M e c h a n i c a l a n d T h e r m a l Stability . . . . . . . . . . . . . . . . . . . . 11.2.3 P o i s o n i n g a n d C a r b o n D e p o s i t i o n P r o b l e m s . . . . . . . . . . . . . . 11.3 P o r o u s I n o r g a n i c M e m b r a n e Reactors . . . . . . . . . . . . . . . . . . . . . . 11.4 Solid O x i d e M e m b r a n e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.5 T h e o r e t i c a l C o n s i d e r a t i o n s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.6 E m e r g i n g A p p l i c a t i o n s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.7 C o n c l u d i n g R e m a r k s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements .................................... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
529 532 534 534 535 537 546 549 555 560 561 561
Chapter 12. TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION WITH INORGANIC AND HYBRID MEMBRANES Christian Guizard and Gilbert Rios 12.1
Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.2
Basic P h e n o m e n a in P r e s s u r e D r i v e n P r o c e s s e s . . . . . . . . . . . . . . . . 12.2.1 M o d e l l i n g of H y d r o d y n a m i c s a n d M a s s T r a n s p o r t . . . . . . . . . . 12.2.2 F o u l i n g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.2.3 Specific A s p e c t s A t t a c h e d to C e r a m i c M e m b r a n e s . . . . . . . . . . 12.3 R e c e n t D e v e l o p m e n t s in Microfiltration a n d Ultrafiltration w i t h C e r a m i c Membranes ..................................... 12.3.1 H y d r o d y n a m i c s of Micro- a n d Ultrafiltration S y s t e m s . . . . . . . . 12.3.2 I n f l u e n c e of M e m b r a n e M a t e r i a l o n P e r m e a b i l i t y a n d Solute Rejection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.4 N a n o f i l t r a t i o n w i t h C e r a m i c M e m b r a n e s . . . . . . . . . . . . . . . . . . . . 12.4.1 S e p a r a t i o n of N e u t r a l Solutes in A b s e n c e of Electrolytes . . . . . . . 12.4.2 Salt Rejection of Electrolyte S o l u t i o n s . . . . . . . . . . . . . . . . . . 12.4.3 S e p a r a t i o n of A q u e o u s I o n i z e d M o l e c u l e - S a l t S o l u t i o n s . . . . . . . 12.5 P r o s p e c t i v e A s p e c t s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.5.1 O r g a n i c - I n o r g a n i c H y b r i d M e m b r a n e s a n d R e l a t e d P r o c e s s e s . . . 12.5.2 C o u p l e d M e m b r a n e Processes . . . . . . . . . . . . . . . . . . . . . . 12.6 C o n c l u s i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
569
570 570 575 581 590 590 593 595 596 598 604 606 606 608 613 614
Chapter 13. APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION C.A.M. Siskens 13.1
Introduction
13.2
T r e a t m e n t of W a s t e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.2.1 W a s t e s of Oily E m u l s i o n s . . . . . . . . . . . . . . . . . . . . . . . . 13.2.1.1 C o m p r e s s o r - c o n d e n s a t e . . . . . . . . . . . . . . . . . . . . 13.2.1.2 C e n t r a l i s e d t r e a t m e n t of i n d u s t r i a l e m u l s i o n s
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.......
619
620 620 620 621
XV
13.2.2
13.2.1.3 Bilge w a t e r t r e a t m e n t . . . . . . . . . . . . . . . . . . . . . 13.2.1.4 V e g e t a b l e w a s t e w a t e r . . . . . . . . . . . . . . . . . . . . . Wastes Based on Semi-solids ......................
13.2.2.1 F i s h f a c t o r y e f f l u e n t . . . . . . . . . . . . . . . . . . . . . . 13.2.2.2 M a n u r e ..... . ....................... 13.3 R e g e n e r a t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.3.1 R e c y c l i n g of Solids f r o m S u s p e n s i o n s . . . . . . . . . . . . . . . . . . 13.3.1.1 C e r a m i c s i n d u s t r y . . . . . . . . . . . . . . . . . . . . . . . 13.3.1.2 P a i n t a n d i n k . . . . . . . . . . . . . . . . . . . . . . . . . . 13.3.2 L i f e t i m e E x t e n s i o n of C l e a n i n g B a t h s . . . . . . . . . . . . . . . . . . 13.3.2.1 A l k a l i n e d e g r e a s i n g b a t h s . . . . . . . . . . . . . . . . . . . 13.3.2.2 I n d u s t r i a l w a s h i n g o p e r a t i o n s . . . . . . . . . . . . . . . . 13.3.3 R e c y c l i n g in C h e m i c a l P r o c e s s e s . . . . . . . . . . . . . . . . . . . . 13.3.3.1 C l e a n i n g of o r g a n i c a n d i n o r g a n i c r e a g e n t s 13.3.3.2 G a l v a n i c b a t h s . . . . . . . . . . . . . . . . . . . . . . . . . 13.4 P r o c e s s i n g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.1 T r e a t m e n t of L i q u i d P r o d u c t s . . . . . . . . . . . . . . . . . . . . . . 13.4.1.1 F r u i t juices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.1.2 Beer b r e w i n g . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.1.3 Beer a n d w i n e c l a r i f i c a t i o n . . . . . . . . . . . . . . . . . . 13.4.1.4 P o t a b l e w a t e r . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.2 T r e a t m e n t of S e m i - s o l i d P r o d u c t s . . . . . . . . . . . . . . . . . . . . 13.4.2.1 P r o t e i n s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.2.2 W h e y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.2.3 S u g a r s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.2.4 P a p e r a n d p u l p . . . . . . . . . . . . . . . . . . . . . . . . . 13.4.3 B i o t e c h n o l o g y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements .................................... References .........................................
621 622 622
........
622 623 623 623 623 624 625 625 625 626 626 627 627 627 627 627 628 629 630 630 630 631 632 632 634 634
Chapter 14. FEASIBILITYOF THE APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES IN SOME LARGE-SCALE CHEMICAL PROCESSES Henk M. van Veen, Maarten Bracht, Edwin Hamoen and Peter T. Alderliesten 14.1 14.2
14.3
Introduction ..................................... Background Information ............................. 14.2.1 M a t e r i a l s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.2.2 M e m b r a n e R e a c t o r s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.2.3 M e m b r a n e P r o c e s s M o d e l l i n g . . . . . . . . . . . . . . . . . . . . . . G a s S e p a r a t i o n A p p l i c a t i o n s for I n o r g a n i c M e m b r a n e s . . . . . . . . . . . . 14.3.1 D e h y d r o g e n a t i o n of P r o p a n e ...................... 14.3.1.1 I n t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.3.1.2 T h e r m o d y n a m i c s of p r o p a n e d e h y d r o g e n a t i o n . . . . . . 14.3.1.3 A d i a b a t i c r e a c t o r c o n c e p t s ; r e a c t o r m o d e l l i n g e v a l u a t i o n 14.3.1.4 I s o t h e r m a l r e a c t o r c o n c e p t s ; e c o n o m i c e v a l u a t i o n . . . . . 14.3.1.5 G e n e r a l c o n c l u s i o n s p r o p a n e d e h y d r o g e n a t i o n . . . . . .
641 643 643 645 646 648 648 648 649 650 654 657
xvi 14.3.2
14.3.3
D e h y d r o g e n a t i o n of E t h y l b e n z e n e to S t y r e n e . . . . . . . . . . . . . 14.3.2.1 I n t r o d u c t i o n . . . . . . . . . . . . . . . . . . . . . . . . . . .
657 657
14.3.2.2
Conventional process description
658
14.3.2.3 14.3.2.4
I m p l e m e n t a t i o n of m e m b r a n e s . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14.3.2.5
Discussion ............................
664
Conclusions ...........................
665
W a t e r - - G a s Shift M e m b r a n e R e a c t o r
..................
Introduction ...........................
14.3.3.2
W G S m e m b r a n e r e a c t o r for CO2 e m i s s i o n c o n t r o l . . . . .
667
Full-scale process considerations . . . . . . . . . . . . . . .
672
14.3.3.5
Conclusion
Acknowledgements
...........................
....................................
List of S y m b o l s a n d A b b r e v i a t i o n s Appendix
...........................
..........................................
Subject Index
665 665
14.3.3.4
Conclusions .....................................
References
659 661
14.3.2.6 14.3.3.1
14.4
..............
......................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
672 673 674 675 676 676 681
xvii
List of contributors Peter T. Alderliesten Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands B.C. Bonekamp Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands H.J.M. Bouwmeester Laboratory for Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands Maarten Bracht Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands A.J. Burggraaf Laboratory of Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands L. Cot Laboratory des Materiaux et Proc6des Membranaires, (UMR 9987 CNRSENSCM-4411), Ecole Nationale Supdrieure de Chimie, 8 Rue de l'Ecole de Chimie, 34053 Montpellier, France Christian Guizard Laboratoire des Mat6riaux et Proc6d6s Membranaires (UMR 5635 CNRSENSC-UMII), Ecole Nationale Sup6rieure de Chimie, 8, rue de l'Ecole Normale, 34053 Montpellier, France Edwin Hamoen Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands A. Julbe Laboratoire des Mat6riaux et Proc6d6s Membranaires (UMR 5635 CNRSENSC-UMII), Ecole Nationale Sup6rieure de Chimie de Montpellier, 8 Rue de l'Ecole Normale, 34053 Montpellier, France A. Larbot Laboratoire des Mat6riaux et Proc6d6s Membrainaires, Ecole Nationale Sup6rieure de Chemie, 8 rue de l'Ecole Normale, 34053 Montpellier, France
xviii Yi Hua Ma
Department of Chemical Engineering, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 1609, USA J.D.F. Ramsay
Laboratoire des Mat6riaux et Proc6d6s Membranaires (UMR 9987 CNRSENSC-UMII), Ecole Nationale Sup6rieure de Chimie de Montpellier, 8 Rue de l'Ecole Normale, 34053 Montpellier, France Gilbert Rios
Laboratoire des Mat6riaux et Proc6d6s Membranaires (UMR 5635 CNRSENSCM-UMII), Ecole Nationale Sup6rieure de Chimie, 8, rue de l'Ecole Normale, 34053 Montpellier, France Jose Sanchez
Laboratoire des Materiaux et Proc6d6s Membranaires, UMR 5635 CNRS ENSCM UMII, 2, Place E. Bataillon, cc 024, 34095 Montpellier Cedex 5, France C.A.M. Siskens Ministry of Transport, Public Works and Water Management, Road and Hydraulic Engineering Division, P.O. Box 5044, 2600 GA Delft, The Netherlands Theodore T. Tsotsis Department of Chemical Engineering, University of Southern California, Los Angeles, CA 90089-1211, USA Henk M. van Veen
Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 1
General overview, trends and prospects A.J. Burggraaf I a n d L. CoF 1Laboratory of Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands 2Laboratory des Materiaux et Proc6des Membranes, UMR 9987 CNRS-ENSCM-4411, Ecole Nationale Sup6rieure de Chimie, 8, rue de I'Ecole de Chimie, 34053 Montpellier, France
1.1 INTRODUCTION The aim of this chapter is threefold: (i) to account for some important focal points in the book, (ii) to identify main barriers to technological development of membranes and their commercialisation in potentially important areas and (iii) to summarise some interesting trends, developments and R&D areas. In recent years the volume of research and development of inorganic membranes has grown considerably and a large diversity of new ideas, development directions and potential applications have emerged. Today, research funding is more and more coupled with concepts and developments which give promise for at least medium-term commercialisation. As will be shown below, most of the interesting long-term concepts and developments depend on progress in the field of ceramic membranes and on successful commercialisation in liquid filtration as a basis for R&D in other directions. Focus is therefore given to macro-, meso-, and microporous ceramic based (composite) systems. A separate chapter treats dense (non-porous) oxide membranes because of their importance for oxygen or hydrogen permeation properties in gas (air) separation and chemical reactors. The planned chapter on dense metal membranes and bioseparations could not be produced due to problems with the scheduled authors. Properties of metal membranes are treated in two chapters (Chapters 8 and 10).
2
1 -- GENERALOVERVIEW,TRENDS AND PROSPECTS
1.2 MARKET S I T U A T I O N A N D PROSPECTS
The market situation and prospects have been described in several documents. Although the reported figures are not always consistent, the trend is clear. Crull [1] and Charpin et al. [2] predict total sales of inorganic membranes in 1999 of US$ 432 million, of which ceramic membranes make up 80%. According to Crull, carbon membranes and metal membranes will make up 11.5 and 5.5% respectively of this market (see Table 1.1). TABLE 1.1 Actual and projected sales for inorganic membrane materials. From Crull [1] Material
1986
1989
1994
1999
in US$ million
Growth
(%)
Ceramics
6
18
75
345
34
Carbon
0
3
9
50
32
Metals and other
5
8
13
25
12
Glass
0
0
1
3
>100
Other
1
2
4
9
16
Total
12
31
102
432
30
In a more recent study by the Business Communication Company [3] the total sales of inorganic membranes is estimated to be US$ 228 in the year 2003 (about 40 million in 1993), of which 70% are ceramic membranes. These reports predict growth rates of about 30% or more. The total market for membrane sales including polymer membranes is much larger and it is stated that 15% of it will be inorganic in 2003. In a recent study [3,4] these total sales are estimated to be US$1000 million in the year 2000. Larger figures for this total market are given by the Freedonia Group [5] which estimates the total membrane sales in the USA at US$1300 million in 1998 with an annual growth rate of 5%. Studies by Frost and Sullivan [6] indicate growth rates of about 10% for liquid filtration which make up 70% of the total market to 16.5% for use in industrial production. The market for gas separation applications is considered to be potentially very important. In 1993 the total membrane (polymeric) sales for gas separation are about US$ 75 million, which is expected to grow by a factor of three in the year 2000 [4]. Many gas membrane applications are envisaged (Table 13.1 in Ref. [7]). The market for N 2 o r 0 2 production especially is expected to grow for non-cryogenic and membrane applications (in 1993 this was 3-5%). According to Thorogood [9], membrane applications will particularly be found in medium
1 - - G E N E R A L O V E R V I E W , T R E N D S A N D PROSPECTS
3
and small capacity processes. So far, commercialised membrane applications have been strongly (liquid separation) or exclusively (gas separation) dominated by polymer membranes. Inorganic membranes will have their share of the future growth if use can be made of their following strong points: (a) a relatively high thermal stability; (b) a relatively high chemical stability and biocompatability of specific materials; and (c) good erosion resistance and non-compactability under high pressure. This holds especially for ceramic membranes and to some extent for carbon membranes and ceramic-metal composite membranes. The weak points of inorganic membranes should be minimised (see next section). Ceramic membrane production is intrinsically more expensive and complicated than polymer. Inorganic membrane applications should therefore preferably be found in fields where polymer membranes cannot or do not perform well.
1.3 M A I N BARRIERS TO T E C H N O L O G I C A L D E V E L O P M E N T A N D
ACCEPTANCE 1.3.1 Requirements and Issues
A membrane system is built from components and assembled into modules which, in turn, form the complete system (see Chapter 2)vThe-single components consist of a (usually ceramic) supporting system and the final (usually ceramic) separation layer. The supporting system can be a single plate, tube, hollow fibre or monolithic multichannel or honeycomb structure. The final separation layer can be porous or dense and single phase or composite. A hierarchic system can be built, as discussed in Chapter 2, from a sandwich of macro-, meso-, microporous layers, which can be tailor-made by changing the chemical or physical nature of the pore system. Each step (product) in the manufacturing process can be used for specific applications. The quality of the underlying support (system) determines, to a high degree, the properties and quality of the final top layer and the number of steps necessary in a multi-step coating process to obtain a defect-free final separation layer. The support system must also fulfil strict quality standards and requirements and must be compatible with other components of the membrane module and system. Before a membrane system is accepted by users in applications on a commercial scale, many requirements must be fulfilled. The main requirements are related to a large number of technological problems to be solved and/or a variety of possibilities for realisation. A brief overview of important aspects is given in Table 1.2.
4
1-- GENERALOVERVIEW,TRENDSANDPROSPECTS
TABLE 1.2 Requirements for commercial application of membrane systems 1.
Low cost production of separation units (modules and/or installations) Relation with: -easy scaling up from laboratory to production installations - reproducible fabrication processes availability of not too expensive, high quality supports -
2.
Reliability of components in: 2.1. ambient conditions 2.2. high-temperature applications (T >_'200~ Relation with: reliability of ceramic components (2.1 and 2.2) - availability of reliable sealing technologies (2.2) -
3.
Long-term stability of pore (material) structure Relation with: - thermal or chemical properties - separation and/or permeation properties mechanical stability of support and separation layer under cyclic temperature and/or pressure regimes -
4.
Reasonable to good surface area to volume ratio Relation with: - module architecture Specific conditions of (high) separation and (high) permeation Relation with: - intrinsic membrane properties and limitations process conditions and membrane architecture fabrication technology of thin separation layers on large surfaceareas of asymmetric graded support systems - knowledge of permeation limiting surface processes -
-
1.3.1.1 Characteristics of ceramic fabrication P r e p a r a t o r y to discussing Table 1.2, it is useful to s u m m a r i s e briefly the m a i n characteristics of ceramic materials and their production technology. The fabrication process of ceramic materials always involves stages where particle assemblies ( p o w d e r suspensions) are formed which, during processing, form in a statistical w a y s e c o n d a r y particles called a g g l o m e r a t e s and aggregates. In the next stage p o r o u s c o m p a c t s are f o r m e d b y the p a c k i n g of the a b o v e - m e n t i o n e d particle assemblies into "green" compacts w i t h a certain shape (plate, tube, h o n e y c o m b , etc.). The particle organisation in the c o m p a c t is again a statistical process a n d results in a distribution of pore size and shape w h i c h is related to the distribution of the particle size and shape and of local statistical fluctuations of the p o r o s i t y and of the pore size (so a certain i n h o m o g e n e i t y occurs). In the last stage this "green" c o m p a c t is consolidated by heat t r e a t m e n t at high temperature
1 - - GENERAL OVERVIEW, TRENDS A N D PROSPECTS
5
(sintering process) during which considerable shrinkage occurs. In this process the porosity and pore size distribution as well as the product dimensions change again to their final values. Due to the statistical nature of the initial powder and the subsequent processing, reproducibility of the final product properties is not easy to obtain and requires many precautions. The last step in the production process is machining to obtain final dimensions and surface quality. Ceramic materials are intrinsically brittle. This means that tensile stresses surpassing a certain critical value easily give rise to catastrophic failure (fracture). This is caused by the presence of defects in the material which act as stress concentrators. Larger defects give rise to easier fracture (at lower stress value). Because defect number and size are also statistically distributed throughout the ceramic product, the strength of the material is not a unique material property. It shows a statistical fluctuation which depends on, for example, the fabrication process and the chance of failure increases with the size of the ceramic product. This chance of failure is expressed by a reliability factor for which usually the Weibull modulus m is used. For common ceramics m < 3, for high-tech products values up to 10 can be obtained (note: for metals m >_20 (hard metals) to 30). Despite these drawbacks, ceramic materials are in widespread use. Even hightech components with very precise specifications such as turbine components have been developed, although at great cost. The conclusion must be that the fabrication of ceramic products is relatively expensive. Prices rise substantially with increasing demands on such product properties as porosity, pore size (distribution), reproducibility and reliability. 1.3.2 State of the Art and Needs
In this section the requirements given in Table 1.2 will be used as a starting point for the discussion.
1.3.2.1 Availability and cost The commercial availability of high quality support (systems) is a critical issue in the further development of membrane separation units. To meet compatibility requirements with other components during assemblage, supports must fulfil strict requirements of (a) dimensional uniformity (i.e. roundness, flatness), and (b) thermal expansion coefficient and chemical inertness in hightemperature application. To make them suitable as support for thin layers the surface roughness should not be too great. Finally the pore size distribution of the support and/or support system (including layers) should be reasonably sharp and larger defects, or relatively large pores, should be absent. Commercially available support systems are usually developed earlier for non-mem-
6
1 - - GENERAL OVERVIEW, TRENDS A N D PROSPECTS
brane applications and are adapted only to some extent for micro- and ultrafiltration applications, these being the largest market today. For other, more demanding, applications this coupling to microfiltration-based supports is not a favourable situation. This is especially true for the further development of microporous membranes which are not yet commercially available. For the development of zeolite membranes, porous stainless steel might be a partial alternative support. Scaling up of the processes to large surface areas (i.e. to obtain asymmetric membrane systems with several layers) as is necessary for large-scale operations has been successfully demonstrated for micro/ultrafiltration and bioseparation processes, but not for other applications such as gas/vapour separation and membrane reactors, for which only small-scale laboratory equipment is available. The cost of inorganic membranes per unit area is reported to be much higher than for organic membranes. As argued by, e.g., Fain [10], it is not appropriate to price organic membranes by the unit area. To be comparable with polymer membranes the module cost should be reduced by an estimated factor of about three. This factor can be lower for complete installations. Nevertheless ceramic membrane systems will always be more expensive than polymer-based ones.
1.3.2.2 Reliability Reliability problems in the sense of avoiding fracture of components resulting in breakdown of installations is especially important in large units such as, for example, membrane reactors. The problems cannot be solved in a satisfactory way by improving the material properties only. By appropriately designing modules and processes, satisfactory solutions might be obtained as has been shown for industrial processes with related problems, e.g., ethylene oxide production. Reliable sealing technologies for use at temperatures up to 800~ are available for, e.g., alumina-based tubular membrane systems [11] but need further development for other shapes and materials.
1.3.2.3 Long-term stability The reliability of separation/permeation performance is coupled with (a) fouling problems and (b) the stability of the micro(pore) structure of the membrane system. Fouling is a problem in almost all liquid separation applications. Strategies have been developed to cope with this problem, usually in a satisfactory way. The microstructure of inorganic membranes is very stable against compressive forces. This means that they can withstand large pressure differences
1 ~ G E N E R A L OVERVIEW, TRENDS A N D PROSPECTS
7
without compaction. This is an advantage compared with polymer membranes. In liquid separations under not too harsh corrosive environments long-term stability has been proven and results in long lifetimes which are usually much longer than those obtained with polymer membranes. Remarkably enough, statements of high chemical stability under harsh corrosive environments such as pH < 3 or pH > 9 are not substantiated in the literature by reliable measurements on membrane systems and much more work is needed here. Problems that occur with the bursting pressure of support tubes after long-term usage might indicate local corrosion at the contact points between the ceramic particles making up the microstructure. Long-term stability at high temperature of mesoporous systems can be good when appropriate materials are used, but is unknown for the new emerging microporous membranes.
1.3.2.4 Surface area to volume ratio In polymer membrane systems large surface area to volume ratios can be obtained. For a given module size this area can be larger by a factor of 1000 times the amount obtainable with ceramic membrane systems. This is due to a membrane architecture such as spiral-wound systems (see Chapter 2) which are not available for ceramic systems; but it is also true that for many applications inorganic membranes can be produced having much greater permeance than those of polymer membranes. This means that for many applications the size needed to produce a given volume of product is about the same for inorganic and polymer membranes [10]. Nevertheless, th6producfionof large quantities of products in industrial installations or in waste water treatments requires large membrane (reactor) volumes. One development to increase the surface area to volume ratio in inorganic membrane systems has been the use of monolithic mulfichannel a n d / o r honeycomb structures. An interesting possibility is the transfer of the architecture of flat-stack Solid Oxygen Fuel Cells (SOFCs) to the membrane field. The architecture of these SOFCs is very similar to those required for membrane systems with a relatively large surface area to volume ratio. The production of reliable ceramic hollow-fibre systems is problematic and it is doubtful whether reliable systems with a large surface area can be developed. Carbon membranes have the best chance here.
1.3.2.5 Specific combinations of high separation factors and high permeation Scientists and developers are usually confronted with a demand for the largest possible separation factor as a first requirement. Later, it becomes obvious that for the realisation of commercial applications a high permeance (permeation) is also needed to reduce the size of the membrane separation (reactor) installation.
8
1 - - GENERAL OVERVIEW, TRENI~ AND PROSPECTS
1000
_
100 10
!
0.1
t
1
t,
10
I-
100
,
L
1000
permeance (a.u.) Fig. 1.1. Schematic picture of separation factor 0c versus p e r m e a n c e (permeation).
Here we confront an intrinsic materials problem which is schematically shown in Fig. 1.1. Larger separation factors (z are obtained at the cost of smaller permeation values. There is some degree of freedom within the indicated band width, which is also affected by process conditions. Focusing only on higher membrane selectivity is economically not always justified if it is at the cost of strongly decreasing permeance. In every application there is an economic optimum at a given combination of selectivity and permeance [7,12]. Membrane staging can provide a dramatic increase in separation performance in many situations compared with single-stage membrane systems due to higher product recoveries. Despite an increase in the cost of multiple-stage installations, staging will often be the most cost-effective design [7]. In addition, process conditions, e.g. sweep rates and hydrodynamic conditions, are important in avoiding e.g. concentration polarisation effects which decrease permeance and separation. The use of large amounts of sweep gases or liquids is economically unfavourable and should be minimised. The flux with a given membrane material(s) and structure can be increased by decreasing the membrane thickness. The thinner the separation layer, however, the larger the risk of forming defects which decrease the separation factor. Mesoporous separation layers of good quality with layer thicknesses down to 5-10 ~tm on macroporous supports has been realised with reasonably large surface areas. For microporous layers this has been shown only on small plates for silica (layer thickness 0.1 ~tm) and zeolites (layer thickness 5-10 ~tm). For dense (non-porous) membranes used for oxygen separation the flux becomes insensitive for a decrease of the layer thickness for a critical thickness which is of the order of 0.1-0.3 mm depending on the permeant-membrane
1 - - G E N E R A L OVERVIEW, TRENDS A N D PROSPECTS
9
combination. This is due to a kinetic limitation of the permeance by surface processes. This phenomenon probably also plays a role in hydrogen permeation through metals such as Pd alloys and in the functioning of (catalytic) membrane reactors. A combination of dense membranes with adsorptive porous coatings or metal-oxide composite membranes can relieve this problem. Moreover the synthesis of defect-free, dense (non-porous) layers with a thickness smaller than a few ~tm on a porous support has not yet been definitely solved. Combinations of high separation factors (> 50) and reasonable permeance interesting for practical applications have been realised for gas separation on small surface areas with microporous membranes. These are discussed in Chapter 9 on transport properties and particularly concern the separation of hydrogen and CO2 from each other and from hydrocarbons as well as some hydrocarbon separations. Nanofiltration with reasonable-to-good rejection values for small molecules are reported for small membranes at the boundary of the meso- and micropore region. These fields of microporous materials are in their infancy and much more work is necessary to delineate their potential for practical purposes (see Chapter 11). An important barrier to rapid development of supported microporous membranes is also the lack of direct measurement methods to determine porosity and pore size (distribution) of the (supported) separation layer.
1.4 TRENDS, TECHNOLOGICAL AND SCIENTIFIC PROSPECTS
1.4.1 Infrastructure for Future Work The commercial availability of inorganic membranes is currently limited to a few applications in the micro- and ultrafiltration and bioseparation fields. The commercial development of new inorganic (ceramic) membranes is slow and production costs are high, showing no tendency to decrease. This process is not unusual in the development of new, highly sophisticated products as has been discussed in Section 1.1. Nevertheless, it has led some experts to believe that useful inorganic membranes are not viable, as cited in Ref. [10]. It is therefore important to achieve a practical and cost-effective inorganic membrane in some new applications. This must be done by bringing together solutions to most of the aspects mentioned in Table 1.2. This is not a task for universities only; the best option is to formulate funding for the cooperation of: (a) national laboratories, for the large amounts of necessary practical developments; (b) industries, to articulate market needs and to cooperate in the transfer of knowledge and manufacturing methods; and (c) universities, to produce the basic knowledge needed to interest national laboratories
10
1 - - G E N E R A L OVERVIEW, T R E N D S A N D PROSPECTS
and industries and to create new opportunities and possibilities, i.e. fundamental research, research with a strategic character. 1.4.2 Some Trends
The trend for market penetration will probably follow a path as shown schematically in Fig. 1.2. The figure does not pretend to give quantitative information but merely shows the relative importance of different application fields in time and illustrates the increasing complexity. Gas separation with microporous membranes will probably only start on a commercial scale if membrane business for liquid filtration has become sufficiently profitable to bear the developments necessary to produce commercial gas separation membranes. Commercial availability should therefore be improved for applications not directly making use of liquid filtration membranes. Prospects for commercial applications as described in different reports differ significantly in their conclusions. Fain [10] describes the potential for successful implementation of inorganic membranes for hydrogen separation from coal gas, from C H 4 / C O 2 mixtures and from catalytic reactors as excellent. Sealy [13] concludes that the hydrogen membrane separation market for existing refinery/petrochemical applications is small and difficult to access while high temperature (>100~ should not be an advantage. This last statement is a remarkable one, which conflicts with most of the technical reports in literature. Alderliesten et al. [4] report possibilities for high-temperature applications in the same field.
,?
' o5
=
o.
+'o /
/
/
/
/
/
_,,o~
_#~..,oO//_# . ~,L+,.-//,@--
_#
#"
.
.00.'0~,, .~_,~+//. ~,,oO .,,oO'#,
~o,-/.,
oo'~, ,
^| o.~O-
....-:__ 1980
1990
2000
Fig. 1.2. Qualitative scheme of expected market penetration as a function of time for different groups of membrane applications.
I m G E N E R A L O V E R V I E W , T R E N D S A N D PROSPECTS
11
The temperature requirements of membranes sometimes conflict with each other. For gas separation with microporous membranes, some authors promote finding membranes stable at temperatures higher than 500~ while others require temperatures less than 500~ [4] for dense membranes with very high separation factors, mainly due to compatibility problems of the membrane with other components of the membrane module. The conclusion must be that reliable prospects of commercial applications can be made only for specific applications and specifications, and strongly depend on particular assumptions relating to membrane performance, process conditions and design and on particular future prospects which easily invalidate economic comparisons made today. The acceptance of ceramic membranes in commercial applications is more difficult in fields in which other already well-developed solutions for separation problems are accepted (e.g. PSA, polymeric membranes) and where the reliability and reproducibility is very important as in, e.g., membrane reactors and gas separation in high-temperature applications. All reports agree that acceptance of catalytic membrane reactors on a commercial scale is at least 10 years away. Here, more experimental performance data for particular processes and process conditions are required to stimulate further development. This needs the commercial availability of a larger assortment of microporous as well as dense membranes with a variety of combinations of good separation factors and good performance values. The use of membrane reactors allows process conditi0ns which cannot be obtained with more conventional processes (see Chapter 10 and overviews [13]) and which allow improved yields and selectivities, the use of two simultaneously occurring reactions (e.g. the main reaction and a decoking reaction to eliminate carbon deposits), controlled supply of reactant, etc. By appropriately designing the membrane reactor, the possibility of decreasing the reactor volume to a given, required capacity with respect to that of a conventional unit or conversely increasing the capacity given the reactor volume is equally important. In addition, the energy balance can be improved considerably using membrane reactors, as reported by several authors. The best strategy for acceptance in high-temperature gas/vapour separation (T > 200~ and catalytic membrane reactors is probably its introduction in small-scale processes and/or hybrid installations which exist in two types: (a) a combination of a membrane separation unit with a conventional process [4,7,12]; and (b) interstage removal of component(s) by a membrane unit in between two reactors in series [15]. This type of solution reduces the risk of accepting membrane reactors or separators because in the event of failure of the membrane unit these can be switched off without catastrophic consequences. Furthermore, process economics can be improved in this way.
12
I m GENERAL OVERVIEW, TRENDS A N D PROSPECTS
In all applications there will be a drive to decrease the size of the membrane unit by increasing the surface area to volume ratio. This implies a trend from tubular to multichannel (honeycomb) structures and, further, to stacked plate designs.
1.4.3 Prospects for Interesting Membrane Applications Several groups of application fields can be distinguished: (a) industrial (production) processes; (b) energy-related applications; (c) environmental applications; and (d) others
1.4.3.1 Industrial production processes In gas or vapour separation and membrane reactor applications the most obvious field is related to a future hydrogen economy a n d / o r processes in which hydrogen is important. These particularly concern re-forming of natural gas and/or syn-gas production and/or syn-gas ratio adjustment. Examples are hydrogen separation from coal gas and from CH4/CO2 [10]. As stated by Alderliesten et al. [4] the re-use of H 2 by separation from reaction products instead of using it as fuel is economically interesting due to the 8-fold higher value of hydrogen as reactant. Processes related to (de)hydrogenation reactions and isomer separation might also be interesting. Examples are (see also Chapters 10 and 13): naphtha reforming to provide high value (>95%) H2. alkane dehydrogenation to alkene(s) examples: propane ~ propene, isobutane ~ isobutene - Isomer separation of e.g. p-xylene from o-m-xylene Methanol synthesis or methanol conversion Air separation by membranes to produce N 2 or 0 2 is interesting especially for small to medium capacities and medium (up to 99%) purity. The highest purities of oxygen (>99%) are to be obtained with dense (non-porous) membranes. The economics of gas separation processes is discussed in detail by Spilman [7] together with a number of potentially interesting applications. In liquid separation applications interesting fields with micro- and ultrafiltration are: - in the food and beverage industry: concentration of whole or skimmed milk, removal of bacteria, phospholipids and casein fines, production of fresh cream cheese, extraction of serum proteins from whey, clarification and sterilisation of fruit juices, wine and beer, etc. in the biotechnology and pharmaceutical industry: cell culture, separation of fermentation broths and of microorganisms, extraction of vitamins, separation and concentration of proteins, separation of metabolic products from blood, etc. -
-
-
-
-
1 - - GENERAL OVERVIEW, TRENDS A N D PROSPECTS
13
Nanofiltration. This new process involves a low rejection for salts (monovalent ions) and ionised organics (MW < 100) in combination with a high rejection of salts (multivalent ions) and organics with MW > 300 which should be separated from the earlier ones at low operating pressures compared with reverse osmosis. Examples are given in Chapter 11 and zirconia or titania membranes seem to be especially suitable due to their relatively good chemical stability.
1.4.3.2 Energy-related applications This group is in most cases a subgroup of industrial applications but has a main goal of enhancing the energy efficiency of processes by incorporating membrane processes in the total design. Energy savings of up to 25% of the total energy costs have been reported by, e.g., manipulation of hydrogen streams in (de)hydrogenation processes or using pervaporation membranes which can reduce energy costs in distillation operations. High-temperature membrane separation operations which avoid cooling down/heating up cycles can be beneficial. A quite different example is the use of oxygen-enriched air in diesel engines to reduce fuel consumption and waste in the exhaust gas.
1.4.3.3 Environmental applications Two main groups of application can be distinguished: (a) non-waste applications, and (b) waste removal related applications
Non-waste applications It is expected that this group has great potential and will become more and more important in the future. Examples of potential applications are given below and are partly discussed by Spilman [7]. Applications on a smaller scale can be found here and may promote widespread acceptance of membrane systems: - clean/pure liquid or gases in electronic industry (e.g. clean room atmospheric control, ultrapure water in microelectronics) pretreating water prior to reverse osmosis processes - recovery of homogeneous catalysts or enzymes from production processes aeration or de-aeration of liquids air dehumidification - atmosphere control in buildings -
-
-
Waste-related applications Here the main problem is to reduce the environmental load by removal or recovery of unwanted products from the exhaust streams of production processes,
14
1 - - GENERAL OVERVIEW, TRENDS A N D PROSPECTS
energy generating systems or engines, or to design closed production processes with closed-cycle processing. Important strategies in waste management are volume reduction of the waste a n d / o r recycling, and membrane processes can be helpful here. Many examples can be found in petrochemical, food processing, pulp and paper and biotechnology industries. A new example is the removal or conversion of CO2 from or in process streams in relation to the "greenhouse effect". Given the massive amounts of CO2 involved it is doubtful whether membrane processes can play a role here. As previously discussed, applications on a small scale might be of interest to promote acceptance of gas/vapour separation systems. Examples are: - organic vapour recovery systems, e.g. in domestic applications (cooking/frying/fat vapours) or hydrocarbon removal in industry gas filtration in buildings diesel engine filtration Large-scale applications might also be possible as indicated by Bishop et al. [16] using a honeycomb dead-end type microfiltration membrane structure with a relatively large surface area to volume ratio of 155 ft2/ft 3. Particulates are filtered from air streams and the filter cake can be easily removed by gas back-pulsing. The system can be loaded with catalysts and is said to be suitable for NOx reductions and VOC (volatile organic compounds) oxidation with efficiencies of >95% and >99% respectively. -
-
1.4.3.4 Others
The combination of membranes and sensors has been mentioned in the literature as a potentially interesting field. To date, however, very few examples of applications have been realised. The measurement of water content in soils and in underground rock formations is potentially interesting as has been proved by field tests [11]. The use of membranes in some consumer products has been discussed in the preceding section and can be extended with, e.g., oxygen enrichment of air for medical purposes. Photocatalytic membranes are mentioned in the literature as potentially interesting for a number of applications. Here, the membrane acts in the first place as a transparent storage medium for reactants, i.e., as a transparent microreactor system. 1.4.4 Interesting Fields for Future R & D
In this section some important technological and scientific R&D fields will be summarised. The authors do not pretend to give a complete overview but indicate fields which in their opinion certainly deserve further attention. The list of items given below is not given in order of priority.
I m G E N E R A L O V E R V I E W , T R E N D S A N D PROSPECTS
15
1.4.4.1 Long-term chemical stability Reliable data as well as reliable, standardised methods for the measurement of chemical stability reflecting operational conditions in practice do not exist, neither for separation layers, nor for support systems (see Chapter 6). A measuring method has been proposed by Z~iter et al. [17]. Methods to improve the chemical resistance of the pore structure against strong acids by coating techniques might also be interesting. Z~iter [17] reports improvement of y-alumina membranes against acids with pH < 3 by partial coating with zirconica, Bhave [18] indicates an increase of the resistance of alumina against phosphoric acid by coating with titania. More work is urgently needed if membranes are to be used in conditions of very large or very small pH values or in steam. This holds even more for the thermal stability for microporous membranes and for both chemical and thermal stability of membranes for nanofiltration. In catalytic membrane reactors the compatibility of catalysts and membrane material requires attention.
1.4.4.2 Thin-layer deposition technology Preparation of defect-free or defect-poor supported thin separation layers both porous and dense ~ with a thickness of <1 ~tm (or preferably, for microporous gas separation membranes, <0.1 ~tm) is necessary to obtain high fluxes. Technologies capable of easy scaling-up to large surface areas would be preferred. An important strategy might be the use of modification technologies which result in the deposition of thin (<1 ~tm) plugs within the pore entrance of precursor membranes. Some results are reported in literature by sol-gel as well as CVD techniques but results indicate no good control of the thickness and structure of the plugs. With similar modification techniques, plugging of defects (cracks or pinholes) is also possible. Optimalisation and a better understanding of these processes is necessary especially if separation layers have to be deposited on more complex precursor structures (e.g. monolithic, multichannel structures).
1.4.4.3 Support technology Further development of compact support systems with high surface area to volume ratios is very important. This should be done in combination with the requirement of low flow resistance. Development of membranes built up of densely packed flat membranes (flat stacks) of a type as used in solid oxygen fuel cells (SOFC) seems to be an interesting option. Some results with a rather simple structure obtained with tapecasting are reported by Simon et al. [19].
16
1 w GENERAL OVERVIEW, T R E N I ~ A N D PROSPECTS
For some applications non-oxide, stainless steel supports appear interesting. This support does not suffer from the brittleness of ceramic supports and can be quite easily connected to other module components, but does not allow strongly oxidising conditions or high temperatures.
1.4.4.4 Microporous membranesfor gas/vapour separation This new field is still in its infancy and offers a variety of useful and interesting R&D opportunities. For hydrogen or carbon dioxide separations from other gases good selectivities and good permeation values are reported with silica microporous membranes. For other mixtures and membrane types, however, selectivities can be very good in most cases but permeation is too low. Furthermore, microporous membranes suffer from the problem shown in Fig. 1. Research is urgently needed to resolve this problem by making use of the materials depending band width sketched in Fig. 1.1 and by optimising other factors (small thickness, large porosity) taking into account long-term stability requirements (high temperature, steam). For further considerations it is useful to distinguish between systems with (a) small micropores (pore diameter dp< 0.5 rim) and (b) wide micropores (dp = 1.0-2.0 nm) with a transition region in between for intermediate pore diameters. Small pore systems are important when separation of gas mixtures with the size exclusion mechanism is necessary. This is the case for separations at high temperature under conditions when adsorption selectivity does not play an important role and where selectivity will sharply drop for other mechanisms. Wide pore systems have been little investigated but are expected to have interesting properties for separation of mixtures of larger molecules which differ strongly in their interaction energy with the pore walls. Here, large permeation values in combination with good selectivities can be expected. The theory of transport in multicomponent gas mixtures in micropore systems should be further developed and more data on competitive adsorption of multicomponent gas mixtures on the membrane material should become available together with adequate characterisation methods of porosity and pore size of supported systems. Three groups of materials can be distinguished: (a) non-crystalline (X-ray amorphous); (b) zeolite type; and (c) crystalline, non zeolite type.
(a) Non-crystalline (X-ray amorphous) This group is mainly formed by silica or carbon membranes. For silica in the small pore and intermediate pore region very good combinations of selectivity and fluxes are reported. The porosity of the membrane seems to be too low however (note: good measurement methods for supported microporous membranes do not exist). Porosities of at least 20% of theoretical density should give considerable improvement in the permeance. A strategy to overcome this is,
1 ~ GENERAL OVERVIEW, TRENDS AND PROSPECTS
17
e.g., the use of template molecules in a variety of ways during synthesis followed by pyrolysis (see Chapter 7). Further research to fine tune the pore size and distribution is necessary, particularly in the small pore size range where the use of template molecules might be useful in addition to "engineering" of size and packing of inorganic polymeric precursor molecules. Finally, research into thermal and chemical stability problems is important. As has already been discussed, controlled formation of very thin high porosity microporous plugs within the pore entrance of the supporting system should be very interesting. Trials with silica, carbon and SiC deposits have been reported however without giving high selectivities and permeances.
(b) zeolite type: Defect-free zeolite membranes have so far only been produced for membranes of the MFI (silicalite type) with thicknesses of about 50 ~tm on stainless steel supports and 3-10 ~tm on alumina and carbon supports. They are produced by in situ methods of zeolite crystals grown directly on the support system. There are some reports of formation of defective membranes with, e.g., zeolite A. Much more research is needed to widen the range of available zeolite membrane types especially small and wide pore systems. The permeance values of the defect-free membranes is lower than that of the amorphous membranes (see Chapter 6) and to improve this the layer thickness must be decreased together with improving the crystal quality (no impurities, no surface layers, high crystallinity, crystal orientation) and microstructure (grain boundary engineering). (c) Crystalline, non-zeolite type: Very wide pore systems with good stability are especially difficult to make with zeolites. Packing of very small particles (diameter 1-2 nm) is reported to give membranes with very wide pores (1.5-2.5 nm) but again with low stability. Alternative strategies might be the use of "pillared-clay systems" as investigated in catalysis and of systems which form subunits in solution and have wide pores when these subunits are packed in a film to form membranes. Micro-emulsion techniques with self organising (surfactant) molecules or the use of imogolite-related materials [20] are interesting (see also Chapter 7 on sol-gel chemistry and its application to porous membranes). The problem with these alternatives is that they are usually strongly anisotropic and so need to be oriented on the support. Research to select available possibilities seems interesting. 1.4.4.5 Nanofiltration membranes Good rejection values are reported for substances with intermediate molecular weights in liquid filtration with mesoporous membranes. To achieve optimal
18
1 -- GENERAL OVERVIEW, TRENDS AND PROSPECTS
values for low molecular weight compounds (MW < 100-300) or multivalent cations, nanofiltration membranes with pore diameter in the wide-pore range (1.0-2.0 nm) are necessary. Similar problems as discussed for wide-pore systems for gas separation also obtain here and are discussed in Chapter 7. In addition to liquid filtration membranes, extensive research has been reported to modify the chemical nature of the internal pore surface either by coating the constituting particles before making the membrane or by grafting organic coatings with functional groups relevant for the intended separation purposes onto the internal surface. Rejection values are increased and fouling can be reduced (Chapter 11, [24]). Extension of these methods to wide-pore microporous systems probably will yield interesting results. Finally functional groups can be introduced also in a membrane system with inorganic-organic composites. Care should be taken in this case to select combinations which add advantages of both components and avoid too many of the disadvantages.
1.4.4.6 Dense (non-porous) membranes and surface reaction limitation These types of membranes are currently only suitable for oxygen separation mainly with oxidic layers, or hydrogen separation m mainly with Pd(Ag) alloys. As discussed in Chapter 6, reasonable fluxes (permeation values) can be obtained at high (for oxygen) to intermediate (for hydrogen) temperatures. The production of defect-free thin layers of these materials remains problematic. The best prospects seem to be present for deposition of Pd(Ag) plugs into the pores of a porous support material as e.g. reported by Morooka et al. [21]. Further research to confirm the reproducible synthesis of this system and of its long-term stability for hydrogen separation is worthwhile. In the case of oxygen transport the best prospects at this moment are the use of metal-oxide composites with high electronic conductivity, or separation with perovskite-derived membranes as reported by Balachandral et al. [22]. These latter membranes are thick (0.5-1.0 mm) and have long-term stability at high temperature. The use of thinner membranes of this type increases the permeance but to a lesser extent the thinner the membrane. With a thickness around 0.3 mm (depending on the precise system) surface reactions which transfer oxygen from the gas phase to solid material completely become rate determining. This phenomenon limits the much higher permeances which are potentially based on the very high bulk permeation (see Chapters 6 and 8). Strategies to solve this problem involve application of adsorptive porous layers on the dense membrane and of metal-oxide composites where the exchange reaction is catalysed by the metal. The study of these surface reactions and of ways to decrease their effect is important.
1 - - GENERAL OVERVIEW, TRENDS A N D PROSPECTS
19
1.4.4.7 Mixed (hybrid) processes and materials The combination of an existing, accepted process w i t h a special m e m b r a n e process m i g h t result in both attractive prospects and acceptance. A n example m i g h t be the combination of pressure swing absorption (PSA) for gas (e.g. air) separation w i t h m e m b r a n e s in the form of a rotating v o l u m e filled w i t h small spheres consisting of a highly adsorptive core and a selective (microporous) m e m b r a n e coating. If the core consists of a relatively w i d e - p o r e m i c r o p o r o u s material, equilibrium a d s o r p t i o n conditions can be approached. The result m i g h t be a rapid, highly selective h y b r i d (PSA-membrane) process w i t h a large surface area to v o l u m e ratio. REFERENCES
1. A. Crull, Prospects for the inorganic membrane business key. Eng. Mater., 61/62 (1991) 279-288. 2. J. Charpin, A.J. Burggraaf and L. Cot, A survey of ceramic membranes for separation in liquid and gaseous media. Ind. Ceram., 11 (1991) 83-90. 3. InorganicMembranes: Markets, Technologies, Players. Business Communication Company, 21 February, 1994. 4. P.T. Alderliesten, C.A.M. Siskens and C.J. Sealy, Gun H2 een tweede ronde. Potytechn. Tijdschr., 10 (1993) 30-31. 5. Membrane Separation Technologies. Freedonia Group, Cleveland, OH, USA. 6. Market in Membrane Technology 1994. Frost and Sullivan, 2525 Charleston Road, Mountain View, CA 94043, USA. 7. R. Spillman, Economics of gas separation membrane processes, in: R.D. Noble and S.A. Stern, (Eds.), Membrane Separation Technology. Elsevier, Amsterdam, 1995, Chap. 13, pp. 589-667. 8. N. McMullen and M. Hogsak, Reconsider non cryogenic systems for on site nitrogen generation. Chem. Eng. Progr., Sept. (1993) 58-61. 9. R.M. Thorogood, Developments in air separation. Gas Sep. Purif., 5 (1991) 83-94. 10. D.E. Fain, Inorganic membranes: the new industrial revolution, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 1.0-14 July 1994, Worcester, MA, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 365-380. 11. Velterop, Product Information Brochure. Ceramic Membrane Technology, P.O. Box 545, 7500 AE Enschede, The Netherlands. 12. A. Sengupta and K.K. Sirkar, Analysis and design of membrane permeation for gas separation, in: R.D. Noble and S.A. Stern (Eds.), Membrane Separation Technology. Elsevier, Amsterdam, 1994, pp. 499-553. 13. C.J. Sealy and A.D. Little, Report on Membranes in Hydrogen Separation, March, 1995. 14. J. Armor, Membrane catalysis, where is it now, what needs to be done. Catal. Today, 25 (1995) 199-207. 15. M.E.Rezac, S.J. Miller and W.J. Koros, Membrane assisted dehydrogenation of n-butane using polymer-ceramic composite membranes. The International Congress on Membranes and Membrane Processes, 30 August-3 September 1993, Heidelberg, Germany.
20
1 u G E N E R A L O V E R V I E W , T R E N D S A N D PRO SPE CT S
16. B.A. Bishop, R.J. Higgins, R.F. Abrams and R.L. Goldsmith, Compact ceramic membrane filters for advanced air pollution control, in: Yi Hua Ma (Ed.), Proceedings of the
Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 355-364. 17. (a) J.M. Hoffmann-Z/iter, Chemical and thermal stability of (modified) mesoporous ceramic membranes. PhD Thesis, University of Twente, Enschede, The Netherlands, 1995. (b) J.M. Z/iter, W. Boer, K. Keizer, H. Verweij and A.J. Burggraaf, Thermal and chemical stability of classical and modified mesoporous membranes, in: Yi Hua Ma (Ed.), Pro-
18. 19.
20.
21.
22.
23.
24.
ceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 381-390. R.R. Bhave, private communication. C. Simon, A. Solheim and R. Bredesen, Testing and evaluation of flat ceramic membrane modules, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 637-640. R. Segal, J.C. Huling and C.J. Brinker, Use of silica sols in inorganic molecular sieving membranes, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 85-94. S. Yan, H. Maeda, K. Kusakabe and S. Morooka, Thin palladium membrane formed in support pores by metal organic chemical vapor deposition and application to hydrogen separation. J. E.C. Research, 33 (1994) 616-652. (a) U. Balachandran, M. Kleefisch, Th. Kobylinski and S.L. Morisetti, Oxygen ion conducting dense membranes. Patent Publ. no WO94/24065 dd. 27 Oct 1994, Appl. no. PCT/US94/03704. (b) U. Balachandran, J.T. Dusek, A.C. Bose, Dense ceramic membranes for partial oxygenation of methane, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 229-237. V.T. Zaspalis and A.J. Burggraaf, Inorganic membrane reactors to enhance the productivity of chemical processes, in: R.R. Bhave (Ed.), Inorganic Membranes. van Nostrand Reinhold, New York, 1991, pp. 177-207. J. Randon, H. de Lucena Lira and R. Paterson, Improved separations using surface modification of ceramic membranes, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, USA. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 429--435.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraafand L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 2
Important characteristics of inorganic membranes A.J. Burggraaf Department of Chemical Technology, Laboratory of Inorganic Materials Science, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
2.1 INTRODUCTION The transport properties (i.e. permeation and separation efficiency) of inorganic membrane systems depend, to a large extent, on the microstructural features of the membrane and the architecture of membranes and modules. The microstructural features, such as pore shape and morphology, pore size (distribution), interconnectivity/tortuosity, as well as the architecture of the membrane and membrane-support combinations will be briefly described. Here, architecture means the way the different parts of the membrane system or module are shaped and combined. The aim of this chapter is to serve as a background and guide to the variety of membrane systems and module types which are used in subsequent chapters. Very briefly, the state of the art concerning some materials' properties will be treated. Economically the most important group is that of ceramic membranes and therefore a certain focus is given to it.
2.2 TYPES OF I N O R G A N I C M E M B R A N E S
A membrane can be described as a semipermeable barrier between two phases which prevents intimate contact. The barrier must be permselective.
22
2 ~ IMPORTANTCHARACTERISTICSOFINORGANICMEMBRANES
U s u a l l y a n u m b e r of m e m b r a n e s are c o m b i n e d in a m o d u l e , w h i c h is the smallest practical unit containing a set of m e m b r a n e s a n d a n y s u p p o r t i n g structures. T w o m a i n classes of m e m b r a n e can be d i s t i n g u i s h e d : d e n s e (nonp o r o u s ) a n d p o r o u s ones. Dense m e m b r a n e s are m a d e f r o m solid layers of metals (e.g. P d alloys) for h y d r o g e n separation, or of m i x e d (electronic, ionic) c o n d u c t i n g oxides for o x y g e n separation. A special f o r m are the LIMs (liquid i m m o b i l i s e d m e m branes) w h i c h consist of a p o r o u s s u p p o r t filled w i t h a liquid or m o l t e n salt w h i c h is s e m i p e r m e a b l e . Porous m e m b r a n e s consist of a p o r o u s wall or p o r o u s top layers (metal, oxide, glass) on a p o r o u s (metal-oxide) s u p p o r t . A v a r i e t y of p o r e s h a p e s a n d architectures exist, as s h o w n in Table 2.1 The m o s t simple form is a single, u n i f o r m l y s t r u c t u r e d wall of a certain material, the so-called s y m m e t r i c , s t a n d - a l o n e m e m b r a n e s . E x a m p l e s are d e n s e metal or oxide tubes a n d p o r o u s hollow fibres. To obtain sufficient m e c h a n i c a l strength, single-walled symmetric systems usually have a considerable thickness. TABLE 2.1 Types of inorganic membranes Type ( c l a s s ) Shlgle wall (symmetric)
Material/process
Architecture
dense oxide or metal -LIM porous glass or carbon - track etch porous alumina, anodic oxidation of A1
tubes, plate
-
-
-
Single wall (asymmetric)
Supported, multilayered (asymmetric)
tube, hollow fibre (thin) sheets
- porous alumina (anodic oxidation) oxide or metal - porous ceramic membranes: alumina, zirconia, titania, carbon composite ceramic-metal, ceramic-ceramic
layers on porous support tube, disk multflayers on porous support plate, disk, tube, monolith
- ceramic
(partially) plugged pores, intra pore deposits intra pore coating
-
d
e
n
s
e
-
Modified structures
-
Support
ceramic-organic
porous alumina - porous carbon porous metal -
-
plate, disc, tube, monolith plate, tube, hollow fibre woven structures, disc, tube
2 -- IMPORTANT
CHARACTERISTICS OF INORGANIC
23
MEMBRANES
This is a disadvantage for obtaining large fluxes, which require thin separation layers. The solution to this problem is asymmetric structures. Usually these consist of a supporting system with large pores (low flow resistance) of sufficient mechanical strength on top of which are layers with gradually decreasing pore size. This is shown in Fig. 2.6 for asymmetric ceramic membranes. The top layer contacting the liquid or gaseous phase to be separated or reacted has the smallest pore diameter. The pore diameter as well as the physico-chemical nature of the pore walls can be changed by additional treatments. In subsequent sections a brief discussion will be given of pore characteristics and architectural aspects of modules as well as details of some systems.
2.3 MICROSTRUCTURAL
PORE
AND
PORE
NETWORK
CHARACTERISTICS
Porous materials have a very complex structure. Many studies have been devoted to describing and characterising them (see Chapter 9, Refs. [1-3,6,18]). Typical pore shapes are schematically represented in Fig. 2.1 and s u m m a rised in Table 2.2 The most simple pore morphologies (Figs. 2.1A and B) are those of more or less straight cylindrical or conical-shaped pores. This type of pore is formed in so-called track etch and in "anopore" membranes. The latter is obtained by anodic oxidation of A1 metal foils and results in porous (amorphous) alumina (rnesoporous) membranes. A detailed discussion is given by Burggraaf and Keizer in Ref. [1]. These types of m e m b r a n e are useful for fundamental trans-
TABLE 2.2 Microstructural pore (network) characteristics Pore morphology
Process, material
Straight channels -cylindrical, conical, (a)symmetric
Anodic oxidation (alumina) track etch
Interconnected voids/pores - Spongy structures - Packing of particles - spherical particles plates - fibrilles, inorganic polymers - zeolite structures Modified structures
porous glass, carbon ceramic membranes -macro-/mesoporous (x-A1203,zirconia, titania, ~'-A1203microporous silica
-
-
- MFI (ZSM5, Silicalite) Sol-gel, CVD e.g. silica, metals, many catalytic materials
24
2 ~ IMPORTANTCHARACTERISTICSOF INORGANICMEMBRANES
J
1
G
Fig. 2.1. Schematic picture of pore shapes. A and B are single wall, symmetric and asymmetric membranes respectively with straight cylindrical (a) or conical (b) pore shape; (c) represents a ceramic asymmetric multilayered membrane with interconnected pores.
port studies but are not important for practical process applications due to low porosity (track etch) and mechanical weakness and vulnerability. Pore systems found in useful membranes are characterised by interconnected voids (pores). Spongy microstructures are observed by phase decomposition and leaching [1] or in porous carbon, obtained by controlled pyrolysis of polymeric precursors [1]. The economically most important ceramic membranes are characterised by a pore network obtained by a packing of particles in each layer of the multilayered system (Fig. 2.1c). In the case of a packing of spherical particles there is a relation between the particle diameter, the pore diameter and the packing symmetry (coordination number). The sharper the particle size distribution and the more ideal the packing, the sharper the pore size distribution will be. In this type of packing there exist two characteristic pore sizes: (a) the pore diameter of the large inscribed sphere determined by the cavities in the packing and (b) the inscribed diameter of the "windows" between these cavities. The first determines the porosity, the second represents the narrow passages in the network and is thus important for the separation properties. Packings can also be obtained by a packing of plates as shown in Fig 2.2 [2]. Note that here the pores have a slit-shaped structure with a limiting pore diameter in only one direction. Because thermostable particles with diameters below 5-6 nm are very difficult to make, microporous membranes with a pore diameter below 2 nm cannot be produced by packings of spherical or plateshaped particles. Packings of fibrillous particles can result in microporous membranes as observed by de Lange et al. [3] with polymeric silica particles (see also the Chapter 8). Finally, zeolite membranes are formed by intergrown
2
-
IMPORTANT CHARACTERISTICS OF INORGANIC MEMBRANES
-
25
m
~
,
m
o
..d | I
7 /
p
I
9
I
9
,a O~lm~4....
!
!
s
..--
S
A
I
.J
_~
S
9
I- S
r
s
..
j -
-
_j
SJ
Fig. 2.2. Idealised model of a packing of plates as formed in the boehmite membrane structure; d is the distance between two boehmite crystals A and B, ~ is the thickness of the boehmite plates (Leenaars and Burggraaf [2]).
)
@
',-
__
Fig. 2.3. Schematic picture of pore types in a porous membrane, a: Isolated pore; b,f: dead end pore; c,d: tortuous and/or rough pores (d) with constrictions (c); e: conical pore.
zeolite particles with interparticle pores filled with another material. The intracrystalline pores are part of the crystallographic structure and have a very uniform diameter. After the initial packing of the particles into layers from s u s p e n s i o n s or colloidal sols, the layers are dried and heat-treated to produce the stable structure. During heat treatment, sintering occurs resulting in loss of porosity, increase of density and changes in pore size and yielding the final pore network structure. A g a i n several types of pore can be distinguished as s h o w n in Fig. 2.3.
26
2 m IMPORTANT CHARACTERISTICS OF INORGANIC
MEMBRANES
The interconnectivity and the related tortuosity as s h o w n by pore d in Fig. 2.3 is important. Systems with interconnected pore systems usually have tortuous pore networks with m a n y constrictions and dead-end pores. This is not favourable for their transport properties and becomes more serious the lower is their porosity.
2.3.1 Modified Structures Pore structures of membranes formed in a primary process can be modified by subsequent modification processesto change properties and to obtain tailorm a d e systems for specific applications. Figure 2.4 shows schematically some typical microstructures. Figure 2.4a shows a mono- or multi-layer coating on the internal surface of the pore network. This type of coating can be used to decrease the pore size or to change the chemical nature of the pore structure. An example of a metal-organic coating on the inside of a porous glass m e m b r a n e is given in Fig. 2.5. This makes the m e m b r a n e useful for desalting applications in reverse osmosis processes [4]. Figure 2.4b shows deposits of particles within the porous structure. This is useful in catalytic m e m b r a n e reactors [5]. If particles are concentrated on certain sites, they can partially plug the pore system (Fig. 2.4d). If the plugs are concentrated near the pore entrance this is usually accompanied by the formation of a thin layer and vice versa. Samples of such a structure (Fig. 2.4c) are given by de Lange et al. [6] for microporous silica m e m b r a n e s m a d e by a sol-gel process with a layer thickness of about 50 n m and plugs penetrating the mesopores of the 7-A1203 support by about 50 nm. Another example (Fig. 2.4d) has been reported by Tsapatsis and Gavalas [7] for silica membranes m a d e with
A
B
G
D
Fig. 2.4. Schematic representation of modified membrane top-layer structures. (A) monolayer or multilayer deposit; (B) nanoparticles within pores; (C) thin film on top of the membrane and plugs penetrating the pore entrance; (C) plugs or constrictions at a certain site in the top layer.
2 - - IMPORTANT CHARACTERISTICS OF I N O R G A N I C MEMBRANES
H
H
1.1
c-c-cSChH pc ~I
I~ H
H
H H
H
27
.03
C-C-C- S0)H Fig. 2.5. Metal-organic coating on a modified porous glass m e m b r a n e [4].
a CVD process with porous plugs of 10 ~tm just below the pore entrances of a porous glass support.
2.3.2 Supports The quality of the support (or supporting layers, see Figs. 2.1 and 2.6) beneath the separation layer is critical for the quality of the membrane itself. Defects and irregularities in the support usually produce defects in the layer applied on it. Defects are, for example, pores much larger than the average pore diameter of the support as well as grains broken out of the support surface. They will give rise to so-called "pinholes" in the layer on top of the support. Surface irregularities causing rough surfaces exclude the formation of defect-free thin and smooth layers in a single step. Finattythe wetting behaviour of the surface is ~mportant in layer formation processes. Severe local changes in wettability result in pinhole formation. Consequently, high quality supports should be smooth, have constant and homogeneous surface characteristics (wettability) and preferably have a relatively narrow pore size distribution. They should have sufficient mechanical strength which does not age with time.
2.4 ARCHITECTURE OF M E M B R A N E SYSTEMS
Membrane systems and modules are produced in a variety of architectural types, as shown in Table 2.3 Multilayered asymmetric (ceramic) membranes with an architecture as shown schematically in Fig. 2.6 are economically the most important. The basic concept behind such asymmetric (composite) structures is to minimise the overall hydraulic resistance of the permeate flow through the membrane structure. This requires a defect-free separation layer with small pores and as thin as possible on top of a support with large pores.
28
2 ~ IMPORTANT CHARACTERISTICS OF INORGANIC MEMBRANES
TABLE 2.3 Architecture of membrane systems and modules 1 2
Multilayered asymmetric membranes tubes, disc, plate Hollow fibre modules porous glass ensemble of parallel fibres 3 Multichannel(ceramic) monolithic element module of elements with 6-18 channels. a-alumina, cordierite 4a Multilayered asymmetric membranes (tubes) modules alumina, alumina-zirconia ensemble of tube bundles 4b Multilayered asymmetric membranes modules stack of flat sheets (SOFC type) sheets of porous plates
4 3
2 I
Fig. 2.6. Schematic representation of an asymmetric (composite) ceramic membrane. 1. Porous support (1-15 lxmpores); 2. intermediate layer(s) (pore diameter dp = 100-1500nm); 3. mesoporous separation layer (dp = 3-100nrn); 4. Modificationof 3 to microporousseparation layer ( d p = 0.5-2 nm). In most synthesis processes it is not possible to produce the thin separation layer directly on top of a support with large pores because the precursor system from which the separation layer is m a d e will significantly penetrate the supporting pores (e.g. small particles from which small-pore m e m b r a n e s are m a d e will penetrate m u c h larger pores). This will result in a strongly increasing flow resistance. Furthermore, thin layers covering wide pores are mechanically unstable and will crack or peel off easily. A practical solution is to produce a graded system by adding one or more intermediate layers with gradually decreasing layer thickness and pore size between bulk support and separation layer. The larger the difference is between the pore size of the support and of the separation layer, the larger is the n u m b e r of intermediate layers. Sometimes the separation layer a n d / o r its supporting layer is produced not in a single step but in two or more steps, which means that it is a bi- or multilayer system itself. This is done to enhance the layer quality by decreasing the n u m b e r and size of defects by covering or filling the defects in the underlying layer with the final (top) layer.
2 m IMPORTANTCHARACTERISTICSOF INORGANICMEMBRANES
29
A
-
= 99 "~ r -
"N.
Membrane layer
! First
e> 97-
\
layer
Sec/nd
9er~
95
~
.,y=' . . . . . .
10
support ! I -1
&~.~
I 100
,
,
, Pore
l. . . . . 1,000 dlameter
I 10,000
~
,I ....-~ ,,
, 100,000
(A)
Fig. 2.7. Pore size distribution of a four-layered alumina membrane (Hsieh et al. [8]).
An example of the pore characteristics of a four-layer alumina system is shown in Fig. 2.7 in which the top layer is a mesoporous (alumina) membrane with an average pore diameter of about 6 nm and a thickness of 5--6 ~tm. The intermediate layers have thicknesses of 15-20 ~tm and pore diameters of 0.5-0.7 ~tm, respectively. A microporous system has been reported by de Lange et al. [6] based on a n 0c-A1203 support (pore diameter =0.2 mm) with t w o 7 - A ] 2 0 3 intermediate layers (total thickness =7 ~tm, pore diameter ~4 nm) and a final silica separation layer with a thickness of about 100 nm (see Fig. 2.4c) and a pore diameter of about 0.5 nm. The intermediate T-A1203 layers are very smooth the average roughness is about 40 nm B which is a necessary requirement to obtain high-quality silica top layers. The two-step synthesis of the intermediate layers improves the defect quality of the system. A two-step synthesis of the microporous silica (titania) layer further improves the quality of the microporous system (as determined by its gas separation properties). This type of multilayered system is produced today in disc, plate and tube form. In the case of tubes these can be assembled in a module containing a number of tubes connected to a single manifold system. To increase the mechanical robustness and the surface area-to-volume ratio, which gives more filtration area per unit volume of membrane element, alumina multichannel monolithic elements have been developed, as shown in Fig. 2.8. These monolithic elements can again be combined into modules. Surface area-to-volume ratios of 30-250 m 3 / m 2 for tubes, 130-400 m3/m 2 for multichannel monolithics and up to 800 m 3 / m 2 for honeycombmultichannel monolithics are reported by Hsieh [9]. A further increase of the packing density (surface area-to-volume ratio) is obtained with experimental hollow-fibre modules, as shown in Fig. 2.9. They consist of long, hollow fibres with an internal diameter ranging from 40 to 300 ~tm and wall thicknesses of 10-100 ~tm. Surface area-to-volume ratios of more
30
2 -- IMPORTANTCHARACTERISTICSOFINORGANICMEMBRANES
ulk a , m n n d
S
Channel
" / Permeate
/ /
m hlyer
'/
~
~ate hsy~
/
Fig. 2.8. Schematic picture of a porous mulfichannel monolithic membrane element.
,UE FIBER B1
PLU HOLLOW
PERMEA~
MODUL SHELL H IGH-PRESSU]~ GAS M IXTUR
PERMEATE
OUTLET
Fig. 2.9. Hollow-fibre module.
t h a n 1000 m 3 / m 2 h a v e b e e n r e p o r t e d [9]. The small size of the internal bore can often p r e s e n t a p r o b l e m w h e n large quantities of gas are p e r m e a t i n g (high p r e s s u r e d r o p d o w n the fibre bore). H o l l o w fibres are m a d e of h i g h - t e m p e r a ture p o l y m e r s , p o r o u s glass or carbon, a l t h o u g h s o m e e x p e r i m e n t a l w o r k on ceramic fibres is r e p o r t e d in literature.
2 -- IMPORTANT
CHARACTERISTICS OF INORGANIC
31
MEMBRANES
PERMEATI E
FEED
Fig. 2.10. Flow path in a stack of flat membranes. A very interesting development would be the production of flat stacks of composite sheets (plates) similar to those developed in solid oxygen fuel cell (SOFC) technology. In this way a very high packing density can be obtained in a robust module configuration with modest pressure drops. The principle is shown in Fig. 2.10 where the feed gas flows through channels in a stack of semipermeable porous sheets, gas permeates through the walls of the sheet and the permeate flows out of the system through a separator space. 2.5 SOME GENERAL CHARACTERISTICS
Porosities of membrane components vary widely and values are reported ranging from 20 to 60%. Commonly, values of 30-40% are used. Pore sizes range from macropores (>500 nm) via mesopores (20-500 nm) to micropores (<2 nm). A great problem is the lack of reliable measurement methods to measure the porosity and pore size distribution of supported membranes (see Chapter 4). The thickness of the separation membrane layer and of other layers in asymmetric membranesrepresents a trade-off between high flux requirements (requiring thin layers) on the one hand, and physical integrity and defect requirements (requiring thick layers) on the other hand. Current commercial products generally show layer thicknesses of the separation layer in the 10-20 ~tm range, but values of approximately 5 ~tm have also been reported. For microporous layers, thinner layers should be developed. The bulk support and intermediate layers have thickness values ranging from I to 2 mm, to provide sufficient mechanical strength, and 10-100 ~tm, respectively. In the case of single wall, stand-alone systems thickness values are similar to that of supports in the case of, e.g., porous glass and are in the range of 50-100 ~tm for hollow fibres.
2.5.1 Commercially available inorganic membranes A variety of membrane materials has been investigated and reported [1,9] and an overview of commercially available systems has been given by Hsieh [9]. Alumina, zirconia and, more recently, titania membranes are used in largescale applications. The more complex shapes, i.e. monolith and honeycomb, are almost exclusively based on (z-alumina or cordierite.
32
2 ~ IMPORTANT CHARACTERISTICS OF INORGANIC
MEMBRANES
Some examples of different systems and new developments are given below with their trade name and producer. Carbosep membranes (Tech-Sep, France) are made of a zirconia layer attached to a porous carbon supporting tube assembled into modules containing up to 252 tubes. The same company produces Kerasep membranes of alumina or titania on a monolithic alumina-titania support containing 7-19 channels. Membralox membranes produced by US Filter/SCT (USA) is the name of a group of tubular and monolithic (multichannel) alumina membranes. The supporting system is formed by high-purity c~-alumina multilayers with a final coating of alumina or zirconia. This system has now been developed further to obtain smaller pore diameters for nanofiltration and gas separation. Ceramem membranes (CeraMem, USA) produces honeycomb-shaped monolithic supports of cordierite, the channels of which are coated with zirconia, silica, y-alumina or c~-alumina separation layers. Le Carbone Lorraine Company (France) produces all carbon asymmetric membranes. The support consists of a porous fibre composite tube which is coated on the inner face with one or more porous carbon films obtained by hydrolysis of polymer precursor films. Dedest Corporation (USA) has announced the production of porous inox steel membrane systems (ultrafiltration applications). Mott Corporation (USA) produces porous metal filters (inox, nickel, monel, inconel, silver, platinum) and has announced the development of porous metallic membranes in disc and tube form. Membrane systems with pore diameters in the micropore range (gas separation, nanofiltration) are not yet commercially available but are produced for development and marketing purposes by, e.g., Velterop B.V. (Enschede, Netherlands) and Media and Process Technology Inc. (Pittsburgh, USA). These systems have an c~-alumina support combined with multilayered y-alumina (mesoporous) layers and a silica (microporous) separation layer. Experimental zeolite membranes are reported to be grown directly on c~-alumina or stainless steel (disc or tubular) supports. It should be noted that most of the systems discussed have been developed for use in liquid separations at low ambient temperatures. For high-temperature applications, such as in gas phase processing and membrane reactors, the sealing of the membrane system elements in modules is critical. Several solutions have been tried to connect ceramic membranes to housing or header plates of the module, such as local cooling, the use of high temperature polymers (up to 230~ the use of graphite/carbon filament and graphite packings (up to 300~ in oxidizing and to 1000~ in reducing environments) and ceramic-tometal connections. Further development to obtain reliable and less expensive solutions for high-temperature applications is necessary.
2 m I M P O R T A N T CHARACTERISTICS OF I N O R G A N I C MEMBRANES
33
2.6 C O N S I D E R A T I O N S O N CHEMICAL RESISTANCE
Inorganic membranes are inherently more stable than organic membranes at temperatures over 200~ and with various chemicals such as aggressive organic compounds and liquids with extreme pH values. Nevertheless, for applications in environments which call for long-term contact between corrosive chemicals (e.g. strong acids or bases, hot gases) the corrosion behaviour of the membrane components should be quantitatively known. It is surprising that there are very limited data on the chemical resistance of various oxide materials. Most of the data are obtained on solid, non-porous materials and with simple dissolution test methods. In the few cases where porous materials are used, hydrodynamic conditions are not or are inadequately taken into account and flow of the aggressive media through the pore network does not occur during the tests. The problem has been analyzed by Z~iter and Burggraaf [10,11] and a strategy and test method has been proposed to measure adequately corrosion behaviour of supported membranes in liquid media under conditions approaching those in applications. Their conclusion was that reliable, quantitative data on chemical resistance of ceramic membranes has not appeared in the published literature. The procedure proposed by Z~iter et al. is a combination of (acid) corrosion under conditions where the liquid flows through the membrane, as in applications, measurement of the change in pore-size distribution with permporometry and of the water permeation as a function of corrosion time, and measurement of the amount of corroded material in the aggressive liquid. Some of the more interesting observations made on corrosion of y-alumina and composite materials of alumina-zirco~a inHNO3 solutions are: (a) Corrosion of mesoporous membranes might differ from bulk material due to small particle size (and consequently severe curvature of the pore surface). (b) Due to differences in curvature, corrosion is sometimes strongly localised. Without observable changes in average pore diameter quite severe changes in water permeation rate occur, indicating changes in the internal microstructure of the pore network. (c) Impurities have a great effect on the corrosion rate which can initially be high and then after some time levels off. (d) Even o~-A1203supports were slightly affected by nitric acid solutions of pH = 1 or 2. In a control experiment with high purity (99.99%) Sumitomo powders (grain size 0.6-0.8 ~tm) a similar result was found indicating that indeed small particles (high surface area) a n d / o r very small residual impurity levels concentrated on grain boundaries ~ necks between grains and surface regions ~ affect the corrosion behaviour. These results might explain the experimental observation of a slow deterioration of mechanical properties (decrease of burst pressure) which sometimes
34
2 -- IMPORTANTCHARACTERISTICSOFINORGANICMEMBRANES
occurs after p r o l o n g e d use of the m e m b r a n e s y s t e m in liquid filtration applications. P r o b a b l y local corrosion of the necks b e t w e e n grains in the p o r o u s ceramic structure (formed d u r i n g sintering in the p r o d u c t i o n process) w e a k e n s the necks a n d so the s t r e n g t h of the m e m b r a n e .
REFERENCES
1. A.J. Burggraaf and K. Keizer, Synthesis of Inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characterisation and Applications. van Nostrand Reinhold, New York, 1991, pp. 10-63. 2. A.F.M. Leenaars, K. Keier and A.J. Burggraaf, The preparation and characterisation of alumina membranes with ultrafine pores. ]. Coll. Interface Sci., 105 (1985) 27-40. 3. (a) R.S.A. de Lange (1994). Microporous sol-gel derived ceramic membranes for gas separation. PhD Thesis, University of Twente, Enschede, The Netherlands. (b) R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Polymeric silica based sols for membrane modification applications. J. Non-Cryst. Solids, 191 (1995) 1-16. 4. R. Schnabel and W. Vaulont, High pressure techniques with porous glass membranes. Desalination, 24 (1978) 249-272. 5. V.T. Zaspalis and A.J. Burggraaf, Inorganic membrane reactors to enhance the productivity of chemical processes, in: R.R. Bhave (Ed.), Inorganic Membranes Characterisation and Applications. von Nostrand Reinhold, New York, 1991, pp. 177-208. 6. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterisation of supported microporous ceramic membranes prepared by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57-75. 7. M. Tsapatsis and G. Gavalas, Structure and aging characteristics of the permselective SiO2-Vycor membranes. J. Membr. Sci., 87 (1994) 281-296. 8. H.P. Hsieh, R.R. Bhave and H.L. Flemming, Microporous alumina membranes. J. Membr. Sci., 39 (1988) 221-241. 9. H.P. Hsieh, General characteristics of inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes Characterisation and Applications. van Nostrand Reinhold, New York, 1991, pp. 64-94. 10. J. Z6ter, Chemical and thermal stability of (modified) mesoporous ceramic membranes. PhD Thesis, University of Twente, Enschede, The Netherlands, 1995. 11. J. Z6ter, W. Boer, K. Keizer, H. Verweij and A.J. Burggraaf, The thermal and chemical stability of classical and modified mesoporous membranes. In: Y.H. Ma (Ed.), Proc. of the 3rd. Int. Conference on Inorganic Membranes (ICIM 94), July 10-14, 1994, Worcester, Mass, USA, pp. 381-390.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.]. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 3
Adsorption phenomena in membrane systems Yi Hua Ma Department of Chemical Engineering, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 1609, USA
3.1 INTRODUCTION In a gas solid system, some adsorbate species are generally distributed between a solid surface and a gas phase. This is due to the fact that below critical temperature, all gases tend to be adsorbed on a solid as a result of the van der Waals interactions w i t h the solid surface. This t y p e o f a d s o r p t i o n is ~called physical adsorption (physisorption). In this case, the important factors affecting adsorption include the magnitude and nature of adsorbent-adsorbate and adsorbate-adsorbate interactions. The degree of surface heterogeneity and the translational and internal degrees of freedom which the adsorbed molecules possess can also be important. A certain distribution of the adsorbate between the gas and solid phases exists in a gas-solid system. The distribution depends on temperature and pressure and is customarily expressed as the moles or weight adsorbed per unit weight of solid either as a function of pressure at a constant temperature (isotherm) or as a function of temperature at a constant pressure (isobar). Thermodynamic treatments can be used to develop adsorption models to describe the distribution of the species between the gas and solid phases. Some of the more commonly used adsorption models will be discussed in the following sections. Experimentally, adsorption isotherms are usually determined to describe the a m o u n t adsorbed, n, as a function of pressure, p. The measurements are normally carried out for several temperatures. The data can be alternatively plotted
36
3 - - A D S O R P T I O N P H E N O M E N A IN M E M B R A N E SYSTEMS
as n versus temperature, T, at constant pressure (isobars) or as p versus T at constant n (isosteres). Adsorption models are then used to fit the experimental data. However, it should be cautioned that such a fit is not sufficient for the theoretical verification of the adsorption model. A consideration of the variation of the energy and entropy of adsorption is a more reliable way of testing the models. A second type of adsorption is called chemisorption. In this case, the adsorption energy is comparable to the chemical bond energies and adsorbate molecules have the tendency to be localized at particular sites even though surface diffusion or some molecular mobility may still occur. Due to the chemical nature of the interactions between the gas and the solid surface, the equilibrium gas pressure in the adsorption system can be extremely low. This enables one to study the adsorbent-adsorbate system under high vacuum using diffraction and spectroscopic techniques for the identification of the actual species presented on the surface and the determination of their packing and chemical state. Although there exists a large amount of adsorption data in the literature on adsorbents such as activated alumina and silica, adsorption data on inorganic membranes relating adsorption phenomena to membrane permeation are scarce. The importance of having a better understanding of the adsorption phenomena in interpreting permeation data has been demonstrated by the recent work of Ma and his co-workers [1-4] on the theoretical analysis and experimental investigation of adsorption and diffusion in silica membranes. The objective of this chapter is to present the fundamental theories of adsorption followed by the description and discussion of experimental techniques for the measurements of adsorption isotherms and for the determination of surface area and pore size distribution. The adsorption of gases on microporous membranes and the inter-relation between adsorption and permeation are then discussed. The adsorption in liquid phase is briefly presented. The chapter concludes with a brief summary.
3.2 A D S O R P T I O N I S O T H E R M S
3.2.1 Types of Isotherms As discussed in the previous section, the distribution between the adsorbate phase and the adsorbed phase can be described by an adsorption model, known as adsorption isotherms. Based on experimental data reported in the literature, Brunauer et al. [5] divided adsorption isotherms into five different types which are shown in Fig. 3.1 (BDDT classification). The first two types are by far the most frequently encountered in adsorption systems. The Type I isotherm is the well-known Langmuir isotherm which will be discussed in the next section. The
37
3 - - ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS
l,,,
o r,~ "0 r = o
I
III
f 1 1.0
0
,,
]
0
1
1.0
0
1.0
Relative pressure, p/pO /i
IV
V
/
o
<
0
I
~0
,,
, ]
1.0
Relative pressure, p/pO Fig. 3.1. The BDDT classification of the five types of adsorption isotherm.
Langmuir isotherm assumes a monolayer coverage while the Type II isotherm deals with multilayer adsorption followed by capillary condensation. Types II and III are closely related-to Types IV and V. The only significant difference between Types II, III and Types IV, V is that a maximum adsorption is reached for the latter case while for the former case, the adsorption increases as the adsorbate gas approaches its vapour pressure.
3.2.2 The Langmuir Isotherm The Langmuir isotherm was first developed by Langmuir in 1915 [6] to describe monolayer adsorption. Since then, a number of different approaches have appeared in the literature for the derivation of the Langmuir adsorption isotherm. For example, the statistical thermodynamic derivation was given by Adamson [7]. The following kinetic derivation is essentially similar to that given by Langmuir [8]. It is assumed that the surface consists of a certain number of adsorption sites, So. The number of sites already occupied by the adsorbate molecules is designated as $2. The number of unoccupied sites is equal to $1 = So- $2. If we assume that the rate of adsorption is proportional to the number of unoccupied sites and the gas pressure, p, while the rate of desorption is proportional to the number of sites already occupied by the
38
3 -- ADSORPTION PHENOMENA IN MEMBRANESYSTEMS
adsorbate molecules and at equilibrium, the rate of adsorption is equal to that of desorption, then kl $1 p = kl p(So- $2) = k2 S 1
(3.1)
Dividing Eq. (3.1) by So and recognizing that the fraction of surface covered, 0, equals $2/So, then we have the conventional Langmuir equation 0=
bp l+pb
(3.2)
where b=~
kl
(3.3)
k2
and is called the Langmuir constant. 0 can also be expressed as n/no, where n denotes the kg-moles adsorbed per kg of adsorbent and no is the saturation capacity or the amount adsorbed at the saturation monolayer coverage. The Langmuir isotherm then becomes n-
nobp l+bp
(3.4)
Typical Langmuir isotherms are shown in Fig. 3.2(a) for several gases on a microporous silica membrane [1]. It is interesting to note that at low pressures, Eq. (3.4) reduces to a linear isotherm (3.5)
n = nobp
while at high pressures, n approaches the saturation value of no. Equation (3.5) is sometimes called Henry's law or Henry's equation [9]. One conventional way to test the experimental data for the Langmuir isotherm is to rearrange Eq. (3.4) in the following linear form p 1 -=~+~ n bno
p no
(3.6)
A plot of p / n versus p gives a straight line (shown in Fig. 3.2(b)) with the slope of l/n0 and the intercept of 1/bn0, from which no and b can be determined. The rate constants kl and k2 can be related to the concepts of adsorption time, which is the average time an adsorbed molecule spent on the surface, and the Langmuir constant b can then be expressed as b = bo e-Q/RT
(3.7)
and bo =
No o ~ ~o
~/2nMRT
(3.8)
3 -- ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS
(a)
39
2.s T=30 ~ 2.0
.... Dubinin-Radushkevich CO 2
1.5 O
C2H 4 1.0
0.5 CH 4 N2 0.0
I
0.0
0.4
0.8
1.2
1.6
p, [MPa]
(b)
1.2 &
T=30 ~
0.8
g~
0.4
CO 2
_
0.0 0.0
0.4
0.8
1.2
1.6
p, [MPa] Fig. 3.2. A d s o r p t i o n isotherms of gases on the m i c r o p o r o u s silica membranes: (a) the isotherms; (b) the plot of Eq. (3.6). The solid lines in (a) are also the L a n g m u i r isotherm fits.
40
3 - - ADSORPTION P H E N O M E N A IN M E M B R A N E SYSTEMS
where No is the number of elementary sites each with a site area of G~ Q is the energy of adsorption and t0 can be approximated by the molecular vibration time. From the assumption of one molecule per site, the total number of adsorbed molecules cannot exceed No. In fact, at monolayer coverage, No is equal to the saturation value at high pressures. The Langmuir isotherm can be extended to multicomponent adsorption systems by the following expression nobiPi Yli =
(3.9)
l
where J is the total number of species in the multicomponent system. 3.2.3 The B E T I s o t h e r m s
Brunauer et al. [10] derived an isotherm equation for multilayer adsorption of gases which includes both Type I and Type II isotherm. A more generalized form of the isotherm was later derived by Brunauer et al. [5] to include all five types of the isotherm. Assuming that the Langmuir equation applies to each layer, Brunauer et al. [5] made the following additional assumptions: (1) The heat of adsorption for the first layer has a distinct value while that of the second and succeeding layers is equal to the heat of condensation of the liquid adsorbate. (2) Adsorption and desorption can only occur at the exposed layer. Based on these assumptions and applying the similar approach used in the derivation of the Langmuir equation, the following equation known as BET isotherm is obtained Yl
no
CX
-
(1 - x) [1 + ( c - 1) x]
(3.10)
where x is the relative pressure equal to p/pO with p being the pressure of the adsorbate and p0 being the vapour pressure of the adsorbate and c ~ e a Q / R T with AQ equal to the difference between the heat of adsorption and the latent heat of condensation. Physically, c may be interpreted as the ratio of the adsorption time of the molecules in the first layer and the adsorption time of the molecules in the second and subsequent layers. Like the Langmuir isotherm, Eq. (3.10) can be rearranged to give x 1 ~ = ~ + n ( 1 - x ) cno
(c - 1)x ~ cno
(3.11)
Again, a plot of x/[n(1 - x)] versus x gives a straight line and c and no can be obtained from the slope and the intercept of the line.
3 - - A D S O R P T I O N PHENOMENA IN MEMBRANE SYSTEMS
41
The BET equation has been used as the general method for the determination of surface area from adsorption data because experimentally it is relatively easy to apply. This will be discussed in detail in Section 3.3.2. In addition to Types I and II isotherm, the BET equation also appears to cover the Type III isotherm. For large values of c, i.e., large AQ which is equivalent to cases where the heat of adsorption is much larger than the latent heat of condensation, the BET equation reduces to the Langmuir isotherm. For small values of c, the BET equation follows Type III behaviour. For the intermediate values of c (e.g., --100 for the adsorption of permanent gases such as nitrogen and argon on polar surface), the BET equation corresponds to the Type II isotherm. In this case, the following approximate form of the BET equation can be used n no
1 1- x
(3.12)
Equation (3.12) can fit well in the usual region of the BET equation and has been used to estimate surface area with a single point [11]. This one point method has been incorporated into commercial equipment for rapid surface area determinations. 3.2.4 Isotherms Derived from the Equation of State
The use of an equation of state to derive isotherms is based on the assumption that the adsorbed layer can be treated as a two-dimensional phase~ Inthis case, the fundamental equations in classical thermodynamics can be applied. Thus, at constant temperature, the Gibbs adsorption isotherm becomes Ad~ = nd~t
(3.13)
where A is the surface area, ~ is the spreading pressure, n is the number of moles adsorbed per unit mass of adsorbent and ~t is the chemical potential of the adsorbate. If we assume that the gas phase can be treated as an ideal gas, then, at thermodynamic equilibrium and constant temperature d~t = RTd lnp
(3.14)
The combination of Eqs. (3.13) and (3.14) gives d~ n RT dlnp A
(3.15)
Equation (3.15) forms the basis for the derivation of adsorption isotherms from equations of state. Alternatively, it can also be used to obtain equations of state from adsorption isotherms. For example, for a linear isotherm, n / p is constant and the integration of Eq. (3.15) gives
42
3 - - ADSORPTION P H E N O M E N A IN M E M B R A N E SYSTEMS
(3.16)
~ = RT
where o is the area occupied by an adsorbed molecule. A simple modification of Eq. (3.16) is to include a co-volume term to give (3.17)
~(c~- b)= RT
Substitution of Eq. (3.17) into Eq. (3.15) and integration give the isotherm kp = ~ 0
e~176
(3.18)
1-0
A large number of equations of state has been used to derive adsorption isotherms. Some of the isotherms and their corresponding equations of state have been presented by Adamson [7]. 3.2.5 The Potential Theory
Polanyi [12] took a somewhat different approach to multilayer adsorption by assuming that dispersion forces play the determining role in adsorption, resulting in the existence of a potential field in the vicinity of the adsorbent surface. The adsorbed layer has the highest density at the solid surface and its density decreases as the distance from the surface increases. Thus, it is possible to draw equipotential surfaces as shown in Fig. 3.3. The space between each adjacent potential surface represents a definite adsorption volume which is a function of the potential field. Mathematically, it can be represented as W =fie)
(3.19)
where W is the adsorbed volume above the surface with potential energy field ~. Equation (3.19) is called the characteristic curve which is characteristic of a Gas
I1211.................... i
Fig. 3.3. Equipotential contours according to the Polanyi potential theory.
3 - - A D S O R P T I O N P H E N O M E N A IN MEMBRANE SYSTEMS
43
particular gas-solid system although its form is unspecified. Since the characteristic curve is temperature independent, for a given adsorption system all isotherms should result in the same characteristic curve. In theory, this allows one to predict adsorption isotherms at any temperature from measurements carried out at one single temperature. The isotherm derived from the potential theory is particularly useful in interpreting capillary condensation or pore filling. Thus, it is most appropriate to be used to describe the adsorption in microporous solids. One way of plotting the characteristic curve can be obtained by considering the work required to transport molecules from the solid surface to the gas phase. Since at the adsorption equilibrium, the change of free energy of the system must be zero. The removal of gas molecules from the surface must, therefore, be compensated by the compression work on the surface. Thus 0
P -
0
] v @ - RTh
(3.20) P
P
where p0 is the vapour pressure of the adsorbate. Based on the theory of dispersion interaction, the ratio of the forces of attraction of different molecules is equal to that of polarizability of the molecules of the vapours. This ratio is called an affinity coefficient ~ and is introduced into the potential function [13]. Furthermore, a parameter K, reflecting the function of the size distribution of volume of the pores, is also included in the equation. Thus, the form of the characteristic curve can be written as [13] -C 2 - K ~2
W = W0 e
(3.21)
where W0 is the limiting volume of the adsorbent sites that represent the volume of the micropores of the adsorbent. Equation (3.21) is called the DubininRadushkevich or DR equation. From Eq. (3.21), a plot of In W v e r s u s 1~2 should yield a straight line. A plot for CO2 and H 2 0 on microporous silica membranes is shown in Fig. 3.4 [1].
3.3 EXPERIMENTAL TECHNIQUES
3.3.1 Determination of Adsorption Isotherms Both volumetric and gravimetric techniques have been used for the determination of adsorption isotherms of gases on solids. The volumetric technique is one of the widely used methods for the adsorption isotherm determination and is based on the measurement of the pressure-volume relation to determine the
44
3 ~ ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS -2
W o = 0 . 1 1 cm3/g CO 2 T=30~ -.3
i
A
CO 2 T=30~
A
CO 2 T=70~ CO 2 T=70~ H20 T=70~
-5 N
-6
7i
Glass M e m b r a n e
J,i ~ i , ~ , l , , 5
J" ~ ' " l ~ ' " ' l " ~ l 10
'''lt''''l''' 15
''""
20
25
e 2 / 10 6 , [cal 2 / m o l 2] Fig. 3.4. D u b i n i n - R a d u s h k e v i c h plot for the m i c r o p o r o u s silica m e m b r a n e s .
amount of adsorbate gas on a adsorbent at different relative pressures (p/p~ It requires an accurate calibration of the dead volume in the system. The gravimetric method was first developed by McBain and Bakr [14]. The early version of the gravimetric technique makes use of a delicate spiral quartz spring. The adsorbent weight gained during adsorption is determined by measuring the extension of the spring which is pre-calibrated with known weights. The modern gravimetric unit, however, uses electromicrobalances or transducers. The progress of adsorption can be followed by continuously recording the weight gain as a function of time. This makes it possible for the technique to be used for the measurement of the kinetics of adsorption and the determination of diffusion coefficients. Typical gravimetric units for both sub-atmosphere and high pressure measurements of equilibrium adsorption isotherms are shown schematically in Fig. 3.5 [1,15]. The all-glass sub-atmosphere ulxit is housed in an insulated constant temperature box free from any kind of disturbance, either mechanical or thermal. The unit contains a Cahn 2000 electrobalance with sensitivity of 10 -7 g for
45
3 m A D S O R P T I O N P H E N O M E N A I N M E M B R A N E SYSTEMS
(a) 7Data acquistion
2:
4j
-
[
7----
'
11
Roughing pump
(b)
Oma ~ q u i ~ k ~ Sy.tom ,
..' ".......
Insuiated
.......
Box
i,i:.-.~..
..."
Transducer
..." ..
8,dence mmmbly ...."
l
Cold
Trap
Antivilxation Stamcl
-... ... "..
'--V
V__ G~ Vacuum Cytk~m' Pump
Tom;~rturo Controller
Fig. 3.5. Schematic of the gravimetric units. (a) Sub-atmosphere unit (0-100 kPa): 1.Balance assembly. 2. Controlled heaters. 3. Sample. 4. Thermocouple. 5. Gas reservoir. 6. Pressure sensors. 7. Injection assembly. 8. Zeolite trap. 9. Cold trap. 10. Turbomolecular pump. 11. Insulated box. (b) High pressure unit (0-5 MPa)
w e i g h t m e a s u r e m e n t s . The s y s t e m can be e v a c u a t e d to h i g h v a c u u m b y an oil-free t u r b o - m o l e c u l a r p u m p in conjunction w i t h a r o u g h i n g p u m p . The h i g h v a c u u m is n e e d e d for s a m p l e activation at elevated t e m p e r a t u r e s before ad-
46
3 -- ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS ,e--high vocuum line
,~--nitrogen
F
Fig. 3.6. Schematic of the volumetric unit. (A) Gas burette filled with mercury; (B) diffusion pump; (C) adsorption vessel; (D) vessel containing some pure condensed nitrogen; (E) capillary differential manometer; (F,G) manometer; (H) vacuum and pressure chamber [16].
sorption measurements. The use of an oil-free turbo-molecular pump instead of the conventional oil diffusion pump is to avoid the possible contamination of the sample by the oil vapour. During the adsorption experiment, the pressure is measured by capacitance pressure gauges. Temperature control and measurements are accomplished by digital controllers. The weight change during the adsorption measurement is recorded by a data acquisition system equipped with a personal computer. The high pressure unit shown in Fig. 3.5(b) is similar to the sub-atmosphere unit except it is constructed of 316 stainless steel capable of operation up to 3 MPa. The volumetric unit shown in Fig. 3.6 is typically used for the determination of the BET surface area. The major parts of the unit are shown in the figure. Detailed description of the operation of the volumetric unit can be found in Lippens et al. [16].
3.3.2 Surface Area Determinations A detailed physical chemical interpretation of permeation mechanisms in microporous membranes is possible if the specific surface of the membrane is determined. Such a determination is mostly done by measuring gas adsorption in the solid although microscopic techniques such as scanning electron micros-
3 w ADSORI~ION PHENOMENA IN MEMBRANE SYSTEMS
47
copy can also be used. We shall restrict our discussion to the adsorption techniques for the surface area determinations. Since the definition of surface area can sometimes be ambiguous, it is important to recognize that the macroscopic determination of surface area generally involves the measurement of a certain property of the solid (e.g., equilibrium adsorption capacity) which can be a qualitative measure of the development of the surface. This property can then be related to the actual surface area through an appropriate theory (e.g., BET isotherm). Therefore, one should not be surprised that the results from different models may give different values of the surface area. The BET equation (Eq. (3.11)) can be written in terms of volume adsorbed x Va(1 - x)
-
1 VmC
( c - 1)x +~
Vmc
(3.22)
where Va is the volume of gas adsorbed and Vm is the volume of gas at the monolayer coverage. As previously discussed, a plot of x~ {Va(1 - x)} versus x will, in most cases, give a straight line between a relative pressure of 0.05 and 0.25. Vm can be calculated from the slope and intercept of the straight line and can be used for the determination of the specific surface area when the area occupied by a molecule in the monolayer is known. In addition to the assumption that the liquid structure could be modeled as the closest packing of spheres, Emmet and Brunauer [17] also assumed that the density of the multilayer adsorbed molecules is equal to that of the liquid at the same temperature. With these assumptions, the area of a nitrogen molecule is found to be 16.27 A2. Using this value, the BET surface area of a solid substance can be determined from SBEw-- 4.371 Vm
(3.23)
where SBETis in m2/g and Vm is in cm 3 (at STP) per gram of adsorbent. By assuming the thickness of a monolayer coverage being equal to the diameter of a nitrogen molecule, Shull [18] demonstrated that by plotting V a / V m versus the relative pressure, a number of non-porous solids could be represented by a single curve. The value of 4.3 A for a nitrogen molecule used by Shull corresponds to the closest packing of spheres, which appears to be inconsistent with the assumption that each molecule of the subsequent layer in a multilayer adsorption is simply situated on top of a nitrogen molecule of the previous layer. In order to calculate the t-value, it is, therefore, necessary to assume that the density of the adsorbed layer is the same as the density of the normal liquid nitrogen. Thus, for nitrogen, we have t = 15.47
Va
(3.24)
SBET
where t is the statistical thickness of the adsorbed layer and is in A. SBETiS the
48
3 E ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS
BET surface area determined by Eq. (3.23). Substitution of Eq. (3.23) into Eq. (3.24) gives t = 3.54
Wa
Vm
(3.25)
Eq. (3.25) gives t - 3.54 ~ for a monolayer coverage which is considerably different from the value of 4.3 ~ used by Shull. Linsen [19] showed that a number of solids could be represented by a single t-curve up to a relative pressure of 0.75. Above this value, deviations from the single curve were observed due to capillary condensation. The thickness of the adsorbed layer is not very much affected by the nature of the solid surface for most adsorbents. Deviations from a single t-curve, however, have been observed for certain adsorbents at a relative pressure considerably less than the value of 0.75 although the effect of the nature of the solid surface is still negligible. For example, for Aerosil (SiO2) , t-values for relative pressures less than 0.15 are essentially identical to the common t-curve [19]. For certain adsorbents, notably graphitized carbon blacks, the t-values calculated from Eq. (3.25) using the Vm values determined from the BET isotherm (Eq. (3.22)) do not fall on the common t-curve. On the other hand, a plot of Va versus t for these materials gives a straight line whose slope gives the surface area St (Eq. (3.24)) which is greater than the BET surface area SBET.For a detailed discussion of common t-curves, the reader is referred to Linsen [19]. As previously discussed, SBET c a n be replaced by St to give Wa
S t -- 15.47 t
(3.26)
In some cases, St obtained from Eq. (3.26) is greater than SBET.However, in the range of low relative pressures, the Va versus t plot gives a straight line passing through the origin in almost all the cases. St obtained from the slope of the straight line in this range is generally in good agreement with the BET surface area, SBET. At high relative pressures, deviations from the straight line may exist. The three possible cases of the Va versus t plot are shown in Fig. 3.7 [16]. The shape of these curves can provide considerable insight concerning the shape and dimension of the pore. Curve I shows a straight line passing through the origin for the entire range of the relative pressures. This is an indication that the surface is freely accessible to the adsorbate molecules up to high relative pressures. Multilayer formation is unhindered on all parts of the surface. Curve 2 indicates that above a certain relative pressure, negative deviation from the straight line occurs. The straight line portion of the curve is an indication of unhindered multilayer adsorption. As the adsorption continues,
3 m ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS
49
o
3
S
1
u
o u
o
al
/ J f f
-
,
t , Layer thickness
Fig. 3.7. T o t a l a d s o r b e d v o l u m e as a f u n c t i o n of t h e l a y e r t h i c k n e s s .
the space in the pores available for adsorption decreases due to the formation of adsorbed layers. In the large holes of the solid, pore filling can only be accomplished by capillary condensation at relative pressures close to unity. On the other hand, for a slit-shaped pore, adsorption takes place on both parallel walls of the slit until they are completely filled at a certain relative pressure, above which the surface area becomes inaccessible to the adsorbate molecules. With no capillary condensation taking place in a slit-shaped pore, the negative deviation from the straight line corresponds to a smaller slope-which represents the surface area still accessible to adsorbate molecules. The positive deviation from the straight line shown by curve 3 signifies larger adsorption than simple multilayer adsorption due, primarily, to capillary condensation, which can take place in pores with a certain shape and dimension. The increased slope of the Va versus t curve represents this increased adsorption. An important modification of the de Boer t-plot has been proposed by Sing and his co-workers [20], who introduced the concept of "standard isotherm" for each adsorbent system. The standard isotherm is defined for a non-porous adsorbent with a similar composition to that of the porous one being investigated. He further introduced a quantity, C~s= (n/nx)s, where nx is the amount adsorbed on the non-porous reference material, to be used for the correction of pore radii for multilayer adsorption. 3.3.3 Pore Size Distribution
Historically, the calculation of pore size distributions in porous materials has been primarily based on various forms and modifications of capillary theory
50
3 - - A D S O R P T I O N P H E N O M E N A IN M E M B R A N E SYSTEMS
using carefully measured complete nitrogen adsorption isotherms. The most general characteristics of adsorbents having a wide range of pore size distribution is the presence of a hysteresis loop, i.e., the adsorption branch does not coincide with the desorption branch in a certain range of relative pressures. Although no single mechanism has been put forward to explain all the experimentally observed hysteresis phenomena, capillary condensation remains the most commonly cited explanation for this complicated phenomenon. In this case, the pores are generally modelled as a bundle of capillaries with different sizes and the radius of the pores is related to the relative pressure through the Kelvin equation RT ln pp~
yV r
(3.27)
where y is the surface tension and r is the radius of the pore. Equation (3.27) can be applied to both adsorption and desorption branches of the isotherm. For the model of a bundle of capillary tubes, it is more appropriate to use the desorption branch of the isotherm for the determination of the pore size distribution. The basic idea is that the effective meniscus radius is the difference between the capillary radius and the thickness of the multilayer adsorption at p/pO, which can be obtained from de Boer's t-plot. In practice, at each desorption pressure, Pd, the capillary radius can be calculated from Eq. (3.27). The actual pore radius is then the sum of the calculated capillary radius and the estimated thickness of the multilayer. The exposed pore volume and surface area can be obtained from the volume desorbed at that specific desorption pressure. This step can be repeated at different desorption pressures. Except for the first desorption step, the desorbed volume should be corrected for the multilayer thinning on the sum of the area of the previously exposed pores. The pore size distribution can then be determined from the slope of the cumulative volume versus r curve. A severe limitation of the bundle of capillaries model is that it can give erroneous readings for materials with "ink-bottle" pores. In this case, the pores are emptied at the capillary pressure of the neck followed by the discharge of the large cavity, resulting in a large reading of the desorbed volume at the capillary pressure of the "ink-bottle" neck. Brunauer and co-workers [21,22] used the following general thermodynamic relation [23] to obtain the pore size distribution (referred to as the "modelless" method) 7dS = A~tdn
(3.28)
where S is the surface area covered by the adsorbate during pore filling due to capillary condensation and A~t = RT ln(p/p ~ is the change in chemical potential during adsorption. Equation (3.28) can be integrated to give
3 m ADSORI~ION PHENOMENA
I N M E M B R A N E SYSTEMS
51
n s
S - - ~ RT ~ ln Pp dn
(3.29)
Yl h
A hydraulic radius can be defined as V rh - S
(3.30)
where V is the volume in the pores and S is the corresponding surface area. The procedure for the determination of the pore size distribution is the same as before. In this case, both the volume and the surface area (determined by Eq. (3.29)) are determined by the change in n through the desorption branch. A way to check the pore size distribution is to compare the area determined by Eq. (3.29) with the BET area determined by the adsorption branch before the hysteresis loop. For meso- and macro-pore materials, the Laplace [24] equation has also been applied for the determination of pore size distribution with the assumption that the pores are cylindrical, resulting in the equality of the two radii of curvature in the Laplace equation. In practice, the penetration of a non-wetting liquid such as mercury into the pores at a specific pressure is related to the pore radius through the following equation, with the assumption that all pores are equally accessible r>
27 I cos0c I
(3.31)
P where r is the radius of the pore, 0c is the contact angle, y is the surface tension and p is the pressure. By incrementing the pressure, the volume of mercury penetrating into the pores is also increased. This relation between the applied pressure and the penetrating mercury volume can be used to determine the pore size distribution from the slope of the V versus p plot through the following equation [7] P(r) - p dV r dp
(3.32)
where P(r) is the pore size distribution and V is the cumulating volume of the penetrating mercury into pores with radius smaller than r. This technique has the same limitation as that of adsorption isotherm described previously. For "ink-bottle" pores, a large pressure is required for the mercury to penetrate the small pore neck, but once this pressure is reached, mercury will be able to penetrate into the wide opening of the pore with ease, giving erroneously large pore volume reading for the size of the pore neck. Mercury porosimeters are commercially available and a typical V versus P plot and the pore size distribution determined from a commercial instrument are shown in Fig. 3.8.
52
3 -- ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS 0.30 -
0.25
A
0.20
O) 0 "0
0.15 0.10
0.05
0.00 10000
1000
100
10
1
Radius (R#), nm
Fig. 3.8. Pore size distribution from mercury porosimeter (courtesy of Quantachrome).
The mercury porosimetry data can also be used to calculate the surface area through the following thermodynamic relation Vt
S --
1---L ~ pdV
7cos0
(3.33)
0
where Vt is the total penetrated volume. It has been shown that agreement between the surface area determined from Eq. (3.33) and that from the nitrogen adsorption isotherm is reasonably good
[25]. For microporous materials (pore size smaller than 2 nm), no capillary condensation occurs in the pores. During the mulfilayer adsorption, the fining of the pores is achieved by the meeting of the adsorbed layers from the opposing walls [1]. For this type of pore fining, the adsorption isotherm may show a steep increase in the low pressure region exhibiting the characteristics of a high c-value BET isotherm followed by a levelling off displaying the property of a Langmuir isotherm. Although the BET equation for n layers adsorption can be used to fit this type of adsorption, the Langmuir equation has also been used to fit experimental data of this type of adsorption behaviour (e.g., adsorption on zeolites). Most recently, the Dubinin-Raduschkevich (DR) equation has also been frequently used to describe the adsorption in microporous materials. For this case, the DR equation is placed in the following form W0 - exp - B
log 2
(3.34)
3 m ADSORPTION
PHENOMENA
IN MEMBRANE SYSTEMS
53
where ~ is the similarity coefficient characteristic of the adsorbate, B is an empirical constant characteristic of the adsorbent, W is the adsorbed volume at p/p0 and W0 is the adsorbed volume when the pores are completely filled. Just like meso- and macro-pore materials, microporous materials generally consist of a range of pore sizes. As expected, there were many attempts on using adsorption isotherms for the determination of the micropore size distribution. These include the inclusion of a Gaussian distribution [26] or Gamma distribution [27] in the empirical constant B in the DR equation. Because this approach introduces artificial physical constraints into the DR equation, its usefulness has been debated. An entirely different approach, generally known as the MP method, uses an extension of the t-curve method for obtaining surface area [28]. However, the method has been criticized for the fact that in the low p/pOrange where the micropore filling takes place, the assumption of the validity of the t-plot in the MP method is least viable. Everett and Powl [29] used intermolecular potential functions to describe the adsorption in the Henry's law region in slit-like and cylindrical micropores. The 12-6 Lennard-Jones potential and other forms of potential functions derived from it were used to describe the interactions between single atomic or molecular species, and the interactions between single molecule and solid surfaces of different configurations. The values of the resulting potential functions were computed to illustrate the change of the shape of the potential energy curve and the adsorption potential minimum as functions of the widths of the slit or radius of the capillary and compared with those for infinite separation or infinite radius. They further proposed several methods to compute the effective pore radius, surface area and pore volume from experimental adsorption data of Ar, Kr and Xe on different activated carbons. Although in most cases, the agreement between the results calculated from different methods was generally reasonably good, some variations did exist due to the uncertainty in the values of the parameters used in the calculations. For more detailed discussion of the models and methods of the data analysis, the readers are referred to the original work of Everett and Powl. Using the concept of Gibbs free energy for adsorption, Horvath and Kawazoe [30] extended Everett and Powl's potential function approach to describe the adsorption in microporous molecular sieve carbons. According to Horvath and Kawazoe, the Gibbs free energy for adsorption, RTln(p/p~ is the sum of the energy for adsorbate-adsorbate interactions and adsorbate-adsorbate-adsorbent interactions. The potential energy of interaction, r used by Horvath and Kawazoe can be expressed as r
= Kr I - / ~ / 4 +/~/1~
(3.35)
54
3 ~ ADSORPTION PHENOMENA IN MEMBRANE SYSTEMS
where r is the distance between an adsorbate molecule and an atom in the surface layer, a* is the potential energy minimum and according to Everett and Powl [29], K and c~ can be expressed as m
K=innm)ln~-m
(3.36)
and 1
r~ =
do
(3.37)
where do is the arithmetic mean of the diameters of adsorbent atoms in the wall and the adsorbate atoms. As reported by Horvath and Kawazoe [30], with n = 10 and m = 4, K - 3.07 and r~ = 0.858 do. The potential energy of one molecule between two parallel layers of distance L apart is given by
~,(r)-K~,*[-(-~14+i-~)l~176
(3.38)
Since the free energy for adsorption is expressed as the sum of the adsorbateadsorbate and adsorbate-adsorbate-adsorbent interactions, the potential energy minimum, r corresponding to these two interactions can be expressed in terms of two dispersion constants, Aa_ a and As-a, respectively, the number density of the adsorbate molecules (Na) and adsorbent atoms (Ns) per unit area [32]. Thus, the potential function becomes
f,(r)-NaAa-a+NsAa-s[aa 9/;)1~ (, 4
-
(~yy/4
~_
c~ r
)1~1
(3.39)
where according to Kirkwood-Miiller 6 m e c2 ~ s (Xa
(3.40)
As_a = (Xs
(Xa
Zs
Za
and a a - a -"
3 m e c2 (Xa ~a 2
(3.41)
where me is the mass of an electron, c is the speed of light and ~ and Z are the polarizability and magnetic susceptibility respectively with subscripts a and s indicating adsorbate and adsorbent respectively.
3m ADSORPTIONPHENOMENAINMEMBRANESYSTEMS
55
Applying Eq. (3.39) and integration over the slit space [30] or applying an energy balance on the adsorbent system [31] gives the final form of the equation for the calculation of the pore size distribution
RT In Pp0=
(3.42)
Ns Aa-s + Na Aa_a Av
(34
(310 3 --
(34(L - d)
3(L-d /
(34 9 -- ~
9(L-d /
q-
3(d /
(310 9 9(d)
where d = ds + da, with ds and da being the diameter of the adsorbent atom and the adsorbate molecule, respectively and Av is Avogadro's number. Since Eq. (3.42) was derived for a slit-like pore, its application to other geometries, such as cylindrical pores, requires further consideration. Saito and Foley [31] followed the same procedure as that used by Horvath and Kawazoe to derive an equation for cylindrical pores with specific applications to the determination of pore size distribution in zeolites. In addition to using a cylindrical potential energy function, they also made the following assumptions: (1) a perfect cylindrical pore with infinite length; (2) The formation of the inside wall of the cylinder by a single layer of atoms (oxide ions in the case of zeolites); and (3) adsorption taking place only on the inside wall of the cylinder and due, only, to the adsorbate and adsorbent interactions. The final equations derived by Saito and Foley are oo
p
RT In pO -
12~Av(NaAa_a + Ns As-a)
d
~-~
(3.43)
o
[ 2k+1 1 ( 1 - ~ d;k{_~O~k 21 (~pp/10for the line-averaged potential energy and oo
p
RT In pO =
12/l:Av(NaAa_a + N s As-a)
d
~-~
(3.44)
0 Ikll
(1-d~k{-~220~k~d~O ~pp)~ppJ
d 4
-~k(~ppl }1
for the area-averaged potential energy and dp is the diameter of the cylindrical pore and (x0.5 =
F(-4.5) F(-4.5-k) F(k + 1)
(3.45)
56
3 -- ADSORPTION PHENOMENA
IN MEMBRANE
SYSTEMS
TABLE 3.1 Physical parameters for micropore size distribution calculation Parameter
Adsorbent Carbon [30]
Adsorbate Nitrogen [30]
Adsorbent Oxide [ 3 1 ]
Adsorbate Argon [31]
Diameter (d, nm) Polarizability (0~,cm 3) Magnetic susceptibility (X, cm3) Density (N, moles/cm 2)
0.34 1.02x10 -24 13.5• -29 3.845x1015
0.3 1.46x10 -24 2x10-29 6.7x1014
0.276 2.5x10-24 1.3x10-29 1.31x1015
0.336 1.63x10-24 3.25x10-29 1.31x1015
130.5 =
F(-4.5) F ( - 1 . 5 - k) F(k + 1)
(3.46)
where F() is the gamma function. To use Eqs. (3.42), (3.43) or (3.44) for the determination of pore size distributions, the physical parameters in these equations have to be estimated. The parameters used by Horvath and Kawazoe for the adsorption of nitrogen on carbons and by Saito and Foley for the adsorption of argon on zeolites are summarized in Table 3.1. Simple sensitivity analyses on the parameters performed by Saito and Foley [31] show that as expected, the diameter of the oxide ion has a large effect on the pore size distribution calculation due to it being raised to the higher power in the equations. The effect of the number density of the adsorbent atom per unit area and that of the adsorbate and the magnet susceptibility of the adsorbate (argon) is moderate while the effect of adsorbent ion (oxide ion) is observed to be negligible on the calculation of the pore size distribution.
Or
f
j
~o
0
u.4
0.5
0.6
0.7
0.8
0.9
~iameter (rim)
Fig. 3.9. Micropore size distribution of fresh FCC catalyst [31].
3 -- ADSORPTION
PHENOMENA
IN MEMBRANE SYSTEMS
57
The procedure for using Eqs. (3.42), (3.43) and (3.44) for the determination of the pore size distribution is quite straightforward. Nitrogen is the most commonly used gas for the experiment although argon may be preferred here and was used by Saito and Fogey [31] in their work. Once the nitrogen (or argon) adsorption isotherm is measured, appropriate values of L can be chosen to be substituted into these equations to determine the values of p/pO. The values of V/Vo corresponding to the calculated values of p/pO can then be determined from the measured adsorption isotherm. A plot of V~ Voversus (L - ds) gives the pore size distribution. Micropore size distributions for fresh FCC catalysts determined by Saito and Foley [31] using different models are shown in Fig. 3.9.
3.4 A D S O R P T I O N O N M E M B R A N E S
There are relatively few studies dealing with adsorption on microporous inorganic membranes. Except the work by Ma and his co-workers and Burggraaf and his co-workers, few studies on the interrelation between adsorption and permeation have been reported. The extremely thin membrane layer on a relatively thick membrane support makes the adsorption measurement rather difficult. Neither gravimetric nor volumetric technique will provide sufficient accuracy for the measurement due to the extremely small fraction of the membrane layer in a supported membrane. Nevertheless, adsorption measurements can give important information on pore sizes and permeation mechanisms in microporous membranes. This section will examine the adsorption of gases on microporous membranes and of liquids on mesoporous and macroporous membranes.
3.4.1 Adsorption of Gases on Microporous Silica Membranes and Interrelation between Adsorption and Permeation In a series of papers, Ma and his co-workers [1-4] systematically examined the interrelationship between adsorption, permeation and diffusion in microporous silica membranes. Both equilibrium and nonequilibrium properties of the microporous inorganic gas separation membranes were studied. Both high pressure and low pressure gravimetric units were used in their adsorption measurements. The adsorption equilibrium isotherms of several gases (CO2, H20, C2H4, N 2, C2H5OH, and CH2C12) on the silica membrane were determined. The data could best be fitted with the Dubinin-Radushkevich (DR) equation although the Langmuir isotherm could also fit the data quite well (see Figs. 3.2 and 3.4). Such a good fit of the equilibrium adsorption data has important theoretical implications on the pore size of the membrane.
58
3 - - A D S O R P T I O N P H E N O M E N A IN M E M B R A N E SYSTEMS
One of the most interesting aspects of their adsorption study is the determination of the upper and lower bounds of the pore size of the hollow fibre silica membranes. Even though very different assumptions were made in the derivation of the DR equation and the Langmuir isotherm, the physical meaning of the constants W0 in the DR equation and no in the Langmuir isotherm is quite similar. If the pores are sufficiently small and homogeneous, there will not be enough space to provide conditions for multilayer adsorption. In this case, essentially there will be no difference between the values of W0 and no. On the other hand, with increasing pore size, the value of W0 will gradually increase while no will remain the same value as-the value for monolayer coverage. When the pores are large enough, capillary condensation will take place and deviations from the DR equation will be observed. For the hollow fibre microporous silica membrane, the values of no and W0 reported by Bhandarkar et al. [1] are very close, implying that the pores in the membrane are small enough so that only micropore filling takes place. This assertion of micropore filling was further verified by the fact that no hysteresis was observed for the adsorption of ethanol on the hollow fibre glass membrane. Since no adsorption/desorpfion hysteresis was observed, this implied that the pores were small enough so that the phenomenon of capillary condensation did not take place in the membrane pores. Therefore, the upper limit of the size of the pores in the membrane could be estimated to be 20 ~. The DR equation can be used to estimate the lower limit of the pore size. From the fitting of the experimental data, it is possible to obtain the limiting micropore volume filling W0. The W0 values obtained for different adsorbates in the microporous hollow fibre silica glass membrane are quite similar ranging from 0.08 to 0.15 cm3/g. The small deviations can be attributed to the error involved in the determination of the gas molar volume since the true volume of the gas adsorbed on the porous adsorbent is difficult to determine. On the other hand, for the condensible gases below their critical points, excellent agreement between the W0 values was achieved (see Fig. 3.4). Based on the fact that the values of W0 obtained for various gases with different molecular diameters were almost the same, it was concluded that the micropore volume in the membrane was accessible to all the gases regardless of their molecular diameter. Therefore, the lower limit for the diameter of the micropore in the hollow fibres is at least the molecular diameter of the largest gas used in the study (Kinetic diameter of CH2C12 = 4.8 A). Therefore, the pore widths in the microporous hollow fibre glass membrane are within the limits of 5 A < d < 20 ~. The effect of adsorption on the separation of gaseous mixtures can be further demonstrated by the experimental results presented in Fig. 3.10 [33] which shows the effect of adsorption on the selectivity coefficient. The temperature and pressure dependency of the experimentally determined selectivity coefficients are shown in Fig. 3.10. As shown in Fig. 3.2(a), the equilibrium adsorption
3 - - ADSORI~ION PHENOMENA IN MEMBRANE SYSTEMS
59
30 Purge Gas: He Feed: CO2(10.4%)+H 2
Ak
T-343 K T=363 K
Smail-OD Fiber
25 om
E o e~
M
20--
15 1.0
1.5
2.0
2.5
3.0
Feed Gas Pressure, [MPa] Fig. 3.10. Selectivity coefficient of N2--CO2 m i x t u r e s as a ftmction of t e m p e r a t u r e a n d pressure.
capacity for C O 2 o n the microporous hollow fibre membrane is considerably higher than that of N2-At any temperature, an increase in pressure will result in an increase in the adsorption of CO2, which will make it more difficult for N 2 to permeate. This will result in an increase in the selectivity coefficient of C O 2 / N 2 a s shown in the figure. On the other hand, at any pressure, an increase in temperature will reduce the CO2 adsorption and thereby, causing the selectivity coefficient to decrease. The effect of adsorption on separation is only significant at low temperatures. At high temperatures, physical adsorption is negligible. The permeation through microporous membranes will probably be primarily controlled by diffusion. In this case, the size of the diffusing molecule relative to that of the pore will play an important role. Shelekhin et al. [2] show the dependence of the permeability on the kinetic diameters of several gases (see Fig. 3.11). The kinetic diameters were calculated from the minimum equilibrium cross-sectional diameter from the Lennard-Jones potential. A dramatic decrease in the permeability coefficients is observed with an increase of the penetrant kinetic diameter. Therefore, one of the most important factors controlling the permeation through microporous membranes is the restriction imposed by the molecular size of the penetrant.
60
3 u A D S O R P T I O N P H E N O M E N A IN M E M B R A N E SYSTEMS
1000.00 _
T=30~ H
100.00-=_
2
CO 2OO
10.00-=_
. m
1.00---=
gh 0.10--=
CH
_
4
i11t,llllli,lt,I111111~11~illtIl~li~'''
0.01 2.4
2~
3.2
3.6
4.0
Kinetic diameter, d, [A] Fig. 3.11. Permeability coefficientsas a function of the gas kinetic diameter.
3.4.2 Adsorption on Sol-Gel Derived Ceramic Membranes
Adsorption studies were performed on non-supported SiO 2 and SiO2/TiO2 (30 mol% TiO2) membrane top-layer materials by de Lange et al. [34]. The SiO 2 membrane top layer was prepared by acid catalysed hydrolysis of tetra-ethylortho-silicate (TEOS) in ethanol while the SiO2/TiO2 top layer was prepared by prehydrolysis of TEOS in ethanol with an acid catalyst followed by the addition of titanium precursor Ti-(OnBu)4 in ethanol. The details of the synthesis conditions can be found in de Lange [35] and de Lange et al. [36]. Both volumetric and gravimetric units were used for the determination of the equilibrium isotherms. Measurements were made at both sub-atmospheric and high pressures over a wide range of temperatures. The equilibrium adsorption isotherms for CO2 at several temperatures are shown in Fig. 3.12. At low temperatures, the isotherms showed the Langmuir behaviour while at high temperatures, Henry's law could be applied. Other gases, such as H2, CH4 and iso-C4Hlo showed similar behaviour although some deviation from the Langmuir isotherm for iso-C4Hlo was observed. The adsorption capacity for CO2 and iso-
C4H10on silica is relatively high at the ambient temperature while that of H2 and
61
3 - - A D S O R P T I O N PHENOMENA IN MEMBRANE SYSTEMS
6ol A
I
I--
-~ 273 K
g
~.o 3!'
,U~."
2 E
>
~ 305 K
-~- 323 K
:~;
-E- 348 K
~ 00
--~ 373 K " 25 " "50
" 75
100
125
473K
Pressure (kPa) Fig. 3.12. C02 adsorption isotherms on Si02 [34]. CH 4 is extremely low. This is consistent with the low isosteric heats of adsorption for H 2 and CH 4 calculated from the adsorption isotherms. De Lange et al. [34] also reported that the variation of the calculated isosteric heat of adsorption as a function of surface coverage was relatively small for all the four gases they studied, indicating low adsorbate-adsorbate interactions. The slight decrease in the isosteric heat of adsorption for iso-C4H10 as the coverage increases was attributed to be caused either by the non-Henry behaviour of isobutane at low temperatures or the stronger interaction between the surface and the adsorbed molecules at low coverages. Finally, the effect of TiO2 on adsorption was reported to be small although only limited investigation was carried out. It is interesting to note that their high pressure adsorption experiments for CO2 and CH 4 showed that Henry's law could be applied to pressures of about 15 and 8 bar, respectively for temperatures above 373 K. An interesting implication of this linearity is that if the gas permeation is indeed linearly proportional to the amount adsorbed, then the Henry's law constant can be used to calculate the gas fluxes through microporous membranes at high pressures.
3.4.3 Liquid Adsorption on Membranes One of the factors causing fouling in ultrafiltration membranes is the adsorption of solutes in the membrane pores. Since fouling, in general, has been discussed in the previous chapter, the discussion presented here will be restricted to the adsorption phenomenon. Clark et al. [37] studied the relationship between membrane fouling and protein adsorption on alumina ultrafiltration membranes. Equilibrium adsorption of bovine serum albumin (BSA) was measured by the standard static method at 7~ Their study covered the concentration range between 1 and 10 g/l, pH values between 2 and 10 and NaC1
62
"1
3 - - A D S O R P T I O N P H E N O M E N A IN M E M B R A N E SYSTEMS
4
t-
O
O
J3
pH ,,, 4.9
m
pH,,
4
pH -
s
pH '- 10
_ 0
5 Concentration
i
pH= .
I0 BSA ( g / l )
.
.
2 .
15
Fig. 3.13. BSA adsorption isotherms on 40/~ alumina membrane as a function of pH (7~ [37].
concentrations of 0, 0.1 and 0.2 M. The equilibrium adsorption isotherms are shown in Fig. 3.13 for different pH values. The adsorption of BSA showed a maximum near the isoelectric point of the protein (pH = 4.9) due, in part, to the increasing tendency for the protein to come out of the solution at the point where its net charge is zero. The similar phenomenon has been observed by others in polymeric systems [38,39]. The maximum adsorption at the isoelectric point is consistent with the minimum flux observed during filtration experiment when the pH of the filtrate is around 4.9. They also reported that increasing NaC1 concentration caused the BSA adsorption to decrease and that the increased adsorption near the isoelectric point was essentially eliminated in the presence of NaC1. The presence of salt either interferes with the electrostatic interaction between the membrane and the protein or increases the solubility of the protein at its isoelectric point. The adsorption of tetracycline and hemoglobin on alumina membranes was also reported by Ma et al. [40] and Bansal et al. [41,42]. Both bioproducts showed the same adsorption behaviour as that of BSA. They also developed a technique, believed to be the first time, to quantitatively determine the extent of fouling (adsorption) of inorganic membranes by proteins. The technique involves the staining of the protein on the membrane with phosphotungstic acid and the use of the energy dispersive capability of an electron microscope to determine the amount of tungsten present. The calibration can be obtained by measuring known amounts of adsorbed protein from equilibrium adsorption studies. The technique has been applied to the determination of the location and amount of foulant (protein) within the membrane pores from a digital X-ray map showing element constituents stained on the foulant. Typical four
3 -- ADSORPTIONPHENOMENA IN MEMBRANE SYSTEMS
63
(a)
(b)
Fig. 3.14. (a) 250x magnification X-ray map of clean 0.2 llm pore size m e m b r a n e showing cross section near the m e m b r a n e surface. (b) 250x magnification X-ray m a p s h o w i n g cross section near the m e m b r a n e surface of a 0.2 llm pore size m e m b r a n e which h a d been subjected to equilibrium adsorotion in 1.6 ~/1 hemoglobin solution at o H 6.9 [41 l.
64
3 -- ADSORPTION PHENOMENA
IN MEMBRANE
SYSTEMS
quadrant X-ray maps of a clean membrane and a membrane surface which had been subjected to adsorption in hemoglobin solution are shown in Fig. 3.14. Figure 3.14(a) represents a control and indicates that no tungsten or phosphorous can be seen when no protein is present. On the other hand, the presence of phosphorous and tungsten in the lower left and upper right quadrants, respectively shown in Figure 3.14(b), indicates protein adsorbed throughout the membrane pores. 3.5 S U M M A R Y
Basic adsorption isotherms have been described in this chapter. For microporous membranes, the use of the DR equation to describe micropore filling has been shown to be quite adequate. Techniques for the determination of surface area and pore size distribution have ben presented. The use of potential functions for the determination of pore size distribution in microporous materials has been described. Although the potential function techniques give consistent and satisfactory results, caution must be exerted in using these techniques for the calculation of the pore size distribution, due to the uncertainty involved in the values of the parameters used in the calculation and the simplifying assumptions employed in the derivation of the model equations. Adsorption plays an important role in the separation of gaseous mixtures by microporous membranes and of liquids in ultra- and microfiltration. Adsorption can either enhance or reduce the selectivity coefficient, depending, in part, on the affinity of the individual gases. Adsorption can cause membrane fouling in ultra- and microfiltration. A thorough understanding of the interrelation between adsorption and separation in microporous membranes can provide information for improvement of membrane synthesis.
REFERENCES
1. 2. 3.
4.
M. Bhandarkar, A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. I. Adsorption of gases on microporous glass membranes. J. Membr. Sci., 75 (1992) 221. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. II. Permeation of gases in microporous glass membranes. J. Membr. Sci., 75 (1992) 233. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. III. Application of percolation theory to interpretation of porosity, tortuosity, and surface area in microporous glass membranes. J. Membr. Sci., 83 (1993) 181. A.B.Shelekhin, A.G. Dixon and Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular sieve membranes. AIChE J., 41 (1995) 58.
3 m ADSORPTIONPHENOMENAIN MEMBRANESYSTEMS
65
5. S. Brunauer, L.S. Deming, W.E. Deming and E.J. Teller On a theory of the van der Waals adsorption of gases. J. Am. Chem. Soc., 62 (1940) 1732. 6. I. Langmuir, Chemical reactions at low pressures. J. Am. Chem. Soc., 27 (1915) 1139. 7. A.W. Adamson, Physical Chemistry of Surfaces, 5th Edn. Wiley, New York, 1990, p. 598. 8. I. Langmuir, The adsorption of gases on plane surfaces of glass, mica and platinum. J. Am. Chem. Soc., 40 (1918) 1361. 9. S.J. Gregg and K.S.W. Sing, Adsorption, Surface Area and Porosity. Academic Press, London and New York, 1967. pp. 197, 221. 10. S. Brunauer, P.H. Emmett and E. Teller, Adsorption of gases in multimolecular layers. J. Am. Chem. Soc., 60 (1938) 309. 11. K.J. Katz, An explicit function for specific surface area. Anal. Chem., 26 (1954) 734. 12. M. Polanyi, Verh. Dt. Phys. Ges., 15 (1916) 55. 13. M.M. Dubinin, The potential theory of adsorption of gases and vapours for adsorbents with energetically nonuniform surfaces. Chem. Rev., 60 (1960) 235. 14. J.W. McBain and A.M. Bakr, A new sorption balance. J. Am. Chem. Soc., 48 (1926) 690. 15. Y.H. Ma, Weiruo Sun, M. Bhandarkar, Jinqu Wang and G.W. Miller, Adsorption and diffusion of nitrogen, oxygen, and methane in molecular sieve carbon at elevated pressures. Sep. Techn., 1 (1991) 90. 16. B.G. Linsen, Ed., J.M.H. Fortuin, C. Okkerse and J.J. Steggerda (Eds.), Physical and Chemical Aspects of Adsorbents and Catalysts. Academic Press, London and New York, 1970. 17. P.H. Emmet and S. Brunauer, Accumulation of alkali promoters on surfaces of iron synthetic ammonia catalysts. J. Am. Chem. Soc., 59 (1937) 310. 18. C.G. Shull, The determination of pore size distribution from gas adsorption data. J. Am. Chem. Soc., 70 (1948) 1405. 19. B.G. Lisen (Ed.), Physical and Chemical Aspects of Adsorbents and Catalysts. Academic Press, New York, 1970. 20. G.D. Parfitt, K.S. Sing and D.Urwin, The analysis of the nitrogen adsorption isotherms of microporous materials. J. Colloid Interface Sci., 53 (1975) 187. 21. S. Brunauer, R.Sh. Mikhail and E.E. Bodor, Pore structure analysis without a pore shape model. J. Colloid Interface Sci., 24 (1967) 451. 22. J. Hagymassy, Jr., I. Odler, M. Yudenfreund, J. Skalny and S. Brunauer, Pore structure analysis by water vapour adsorption III. Analysis of hydrated calcium silicates and Portland cement. J. Colloid Interface Sci., 38 (1972) 20. 23. A.V. Kiselev, Usp. Khim., 14 (1945) 367. 24. P.S. de Laplace, Mechanique Celeste. Supplement to Book 10, 1806. 25. L.C. Drake and H.L. Rittrer, Ind. Eng. Chem., Anal. Ed., 17 (1945) 787. 26. H.F. Stoceckli, Generalization of the Dubinin-Radushkevich equation for the filling of heterogeneous microporous systems. ]. Colloid Interface Sci., 59 (1977) 184. 27. M. jaroniec, R. Madey, J. Choma, B. McEnaney and T.J. Mays, Comparison of adsorption methods for characterizing the microporosity of activated carbons. Carbon, 27 (1989) 77. 28. R.Sh. Mikhail, S. Brunauer and E.E. Bodor, Investigation of a complete pore structure analysis I. Analysis of micropores. J. Colloid Interface Sci., 26 (1968) 45. 29. D.H. Everett and J.C. Powl, Adsorption in slit-like and cylindrical micropores in the Henry's law region. ]. Chem. Soc., Faraday Trans., I, 72 (1976) 619. 30. G. Horvath and K. Kawazoe, Method of calculation of effectiveness pore size distribution in molecular sieve carbons. J. Chem. Eng., Jpn., 16 (1983) 470.
66
3 -- ADSORPTIONPHENOMENAIN MEMBRANESYSTEMS
31. A. Saito and H.C. Foley, Curvature and parametric sensitivity in models for adsorption in micropores. AIChE J., 37 (1991) 429. 32. P.L. Walker, Chemistry and Physics of Carbon, 2. Marcel Dekker, New York, 1996. 33. M. Bhandarkar and Y.H. Ma, unpublished work. 34. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer, A.J. Burggraaf and Y.H. Ma, Sorption studies of microporous sol-gel modified ceramic membranes. J. Porous Materials, 2 (1995) 141. 35. R.S.A. de Lange, Microporous sol-gel derived ceramic membranes for gas separation, synthesis, gas transport and separation properties, PhD Thesis, University of Twente, Enschede, The Netherlands, 1994. 36. R.S.A. de Lange, K-N.P. Kumar, J.H.A. Hekkink, G.M.H. van de Velde, K. Keizer, A.J. Burggraaf, W.H. Dokter, H.F. van Garderen and T.P.M. Beelen, J. Sol-Gel Sci. Tech., 2 (1994) 489. 37. W.M. Clark, A. Bansal, M. Sontakke "and Y.H. Ma, Protein adsorption and fouling in ceramic ultrafiltration membranes. J. Membr. Sci., 55 (1991) 21. 38. M. Tucker and J. Hubble, Membrane fouling in a constant-flux ultrafiltration cell. J. Membr. Sci., 34 (1987) 267. 39. W.J. Dillman and I.F. Miller, On the adsorption of serum proteins on polymeric membrane surfaces. J. Colloid Interface Sci., 44 (1973) 221. 40. Y.H. Ma, A. Bansal and W.M. Clark, Adsorption of proteins and antibiotics on porous alumina membranes, in: M. Suzuki (Ed.). Proceedings of the IVth International Conference on Fundamentals of Adsorption. Kodansha, Japan, 1993, p. 389. 41. A. Bansal, R.R. Biederman, Y.H. Ma and W.M. Clark, Protein adsorption and fouling of ceramic membranes as measured by scanning electron microscopy with digital X-ray mapping. Chem. Eng. Comm., 108 (1991) 365. 42. A. Bansal, Y.H. Ma and W.M. Clark, A quantitative investigation of membrane fouling by proteins using energy dispersive spectroscopy, in: A.J. Burggraaf, J. Charpin and L. Cot, (Eds.), Proceedings of the 2nd International Conference on Inorganic Membranes. Trans Tech Publications, 1991, p. 505.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.I. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 4
Methods for the characterisation of porous structure in m e m b r a n e materials A. Julbe and J.D.F. Ramsay Laboratoire des Mat6riaux et Proc6d6s Membranaires (UMR 5635 CNRSENSC-UMII), Ecole Nationale Sup6rieure de Chimie de Montpellier, 8 Rue de I'Ecole Normale, 34053 Montpellier cedex 1, France
4.1 GENERAL I N T R O D U C T I O N
With the development of more complex and sophisticated inorganic membranes there is a need for a better understanding of membrane structures and their influence on the mechanisms of separation processes. This requirement for a better insight into the relationships between (a) the membrane synthesis route, (b) the membrane microstructure or morphological properties and (c) the permeation properties, has been widely emphasised in the literature. Information on membrane characteristics is essential for membrane users, manufacturers and scientists to choose an appropriate membrane for a specific application, controlling membrane quality and preparation process parameters or understanding transport mechanisms. In this section these relationships will be explored in more detail with particular emphasis on the porous properties of membranes and their characterisation. Firstly we will present the general definitions and terminology used to describe porous media. The origin of porosity in inorganic materials will also be outlined and related to a quantitative description of pore structures in
68
4 -- METHODS
FOR THE CHARACTERISATION
OF POROUS
STRUCTURE
~
[ MEMBRANE PREPARATION I . . . . , ,, [ C H A R A C T E R I S A T I O N OF T H E P O R O U S S T R u c T U R E I
Morphology related parameters
Morphology and Permation related parameters
(active and inactive pores)
(active pores only) -
Pore size distribution Specific surface area Porosity (Pore shape information ?)...
.
! ,
O,
ONOF,E
-
O
Active pore size distribution Cut-offvalue Bubble point Fluid flow. hydraulic pore radius
ANCE
I
"
Fig. 4.1. Methodologyfor membrane characterisafion. Listingmethods and related parameters. idealised model systems. In such model systems, the pore geometry can be defined precisely in terms of pore size, shape, connectivity, etc. This has provided the basis for recent theoretical developments describing diffusion and transport processes in such porous materials. This link between the concept of a model porous structure and the theoretical prediction of diffusivity and mass transport is also crucial in the characterisation of porous materials. Thus each characterisafion technique yields experimental parameters which are related to the pore structure of a material; these parameters are then generally used to define the porous properties on the basis of an assumed model pore structure. This aspect will be emphasised in the description of the different characterisation methods described. In the characterisafion of membrane materials, it is important to distinguish "static" characterisation techniques leading to morphology related parameters and "dynamic" techniques in which membrane permeability is involved, leading to permeation related parameters and in some cases to morphology related parameters concerning active pores only [1]. Figure 4.1 shows a list of static and dynamic characterisation methods and the corresponding characteristic parameters which can be obtained. There are already in the literature several book chapters and reviews which list a range of characterisation techniques for membranes [1-4] but the selection of the appropriate characterisation techniques is not always obvious. A recent IUPAC report entitled "Recommendations for the characterisation of porous solids" [5] has stressed conclusions of particular significance, which include the following:
4 - - M E T H O D S FOR THE C H A R A C T E R I S A T I O N OF POROUS STRUCTURE
69
1. The selection of a method of characterisation must start from the material and from its intended use. 2. The method chosen must assess a parameter "related as directly as possible" to phenomena involved in the application for which the material is used. 3. The complexity of the porous texture of materials is such that, even on theoretical grounds, the concepts which can be used to describe the texture usually entail the introduction of simplifying assumptions. 4. No experimental method provides the absolute value of parameters such as porosity, surface area, pore size, etc.; each gives a characteristic value which depends on the principles involved and the probe used. 5. As a consequence, "perfect agreement" between parameters provided by different methods should not be sought. Instead, there must be an awareness of the specific, limited and complimentary significance of the information delivered by each method of characterisation of a porous solid. The choice of a characterisation technique is also dictated by the particular sample characteristics such as the nature of the material, whether it is supported or not, its size, shape, isotropy and mechanical resistance as well as the range of pore size. The destructive nature of the technique may also require attention. Thus careful preparation of the sample is often needed (drying, outgassing, cleaning) to eliminate adsorbed species like water or hydrocarbons especially in the case of microporous materials. Finally it should be noted that the characterisation of membranes is more demanding than most other porous materials. Firstly, the membranes separation layer is generally thin and supported, which requires a sensitive technique capable of analysing a sample in such a form. The characterisation of a powder "equivalent" to the membrane cannot in all cases be considered as ~representative of the membrane texture. Secondly, the structure is frequently anisotropic and moreover often microporous. Assessment of the microporosity is much less advanced compared to meso- and macro-porosity, despite emphasis given to this in the recent IUPAC symposia [6--8]. The current and widespread interest in the characterisation of microporous materials is well illustrated by the numerous and varied publications found in these symposia proceedings. These highlight recent developments in characterisation techniques, their applications and limitations. The particular features of importance in membrane studies will be considered in the light of the characterisation techniques to be described. 4.2 D E S C R I P T I O N
OF POROUS
MATERIALS
-- DEFINITIONS
4.2.1 Origin of Pore Structure Porous materials can be formed in several ways, although the following three are most important in the synthesis of membrane structures:
70
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
(1) In the first, the pores may be an inherent feature of crystalline structures (e.g. zeolites, clay minerals). Such intracrystalline pores are generally of molecular dimensions and result in very regular networks often described as "structural" porosity. (2) Secondly, the pores may be formed by the packing and subsequent consolidation of small particles as may occur in some inorganic gels and ceramics. These processes have been defined (IUPAC) as 'constitutive', because the final structure depends on the original arrangement of the primary particles and on their size. (3) The third route is defined as"substractive' (IUPAC), in that certain elements of an original structure are selectively removed to create pores. Examples include the formation of porous metal oxides by thermal decomposition of hydroxides, of porous glasses by chemical etching, of activated carbons by controlled pyrolysis, of ceramic foam membranes by burning off a polymer (e.g. polyurethane), of alumina by anodic oxidation of aluminium to give oriented cylindrical pores with a narrow size distribution.
4.2.2 Quantitative Description of Pore Structures It is useful at this stage to define pore structure and terminology. Here again we m a i ~ y follow the recommendations for the characterisation of porous solids recently proposed by IUPAC [5,9]. Porosity is defined as the fraction r of the total volume of the sample which is attributed to the pores detected by the method used
~= Vp/V
(4.1)
The value of this fraction depends on the method used to determine the apparent volume V (geometrical, fluid displacement) and on that used to assess the pore volume Vp. We note some methods (e.g. methods using a gas or a fluid) only have access to "open pores" (through pores a n d / o r blind pores) whereas others may have also access to "closed pores" (e.g. methods using radiation scattering). Evidently for membrane characterisation, methods which distinguish open pores are of particular importance. Furthermore, for any given method, ~ will depend on the size of the molecular probe (fluid displacement, adsorption) or of the scale of measurement (stereology). This latter concept is developed extensively in the theory and methods involving fractal analysis. We must also note the concept of inaccessible or "latent pores" [10] corresponding to open pores with a pore width smaller than the probe molecular size. The determination of true density, apparent density and He-replacement density can be used to determine both open and closed porosity of bulk crystalline materials [10] but in the case of supported layers it is not as simple. Image analysis is commonly used to estimate the porosity of unsupported thin films,
4 - - M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S STRUCTURE
71
but it must be appreciated that bulk membrane porosity and surface porosity of a porous medium differ markedly. The specific surface area is defined as the accessible area of solid surface per unit mass of material. It is also dependant on the method employed and the size of the probe used (e.g. adsorbate, molecular probe, wavelength of radiation, etc.). The recorded value will also depend on the assumptions inherent in the simplified models applied to interpret experimental data. This consideration is of particular significance for materials containing micropores. The pore size, or more precisely pore width, is the distance between two opposite walls of the pore. Pore size which is a property of the utmost importance in membrane applications, is even less susceptible to precise definition as already discussed by IUPAC [5]. The problems encountered for the determination of the surface area are further complicated by the fact that the pore shape is generally highly irregular, leading to a variety of definitions of "the size". Moreover, porous systems generally consist of interconnected networks, and results will depend on the sequence in which pores are accessed by the method. It is partly for this reason that quantitative descriptions of pore systems are generally based on model structures. The following classification of pore sizes based on the average width of the pores has been recommended by IUPAC [9]: Macropores: width > 50 nm (0.05 ~tm) Mesopores: width between 50 and 2 nm Micropores: width < 2 nm (supermicropores 0.7 nm, ultramicropores < 0.7 nm) This definition is based on different physical adsorption phenomena of gases in pores of different size. Adsorption interactions of adsorbates are stronger in micropores and modify the bulk properties (density, surface tension) of the adsorbed fluids. The maximum size of ultramicropores corresponds to the bilayer thickness of nitrogen molecules adsorbed on a solid surface (2 x 0.354 nm). Usually the pores in a material do not have the same size but exist as a distribution of size which can be wide or sharp. We can characterise a film by a nominal or an absolute pore size. In fact this definition rather characterises the size of the particles or molecules retained by the layer. Pore size distribution is classically represented by the derivatives dSp/drp or d Vp/drp as a function of rp (pore radius) where Sp and Vp are respectively the wall area and volume of the pores. The size in question is here the radius, which implies that the pores are known to be, or assumed to be, cylindrical. In other cases, rp should be replaced by the width. The mean pore hydraulic radius rH for a porous solid is obtained through the relationship rH = Vp/S
(4.2)
where Vp is the pore volume determined at saturation and S the surface area, e.g. determined by the BET method, rH can be theoretically related to the sample
72
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
mean pore width w by means of the relation w = 2rl r H
(4.3)
in which 11is a pore shape factor (11 =1, 1.4, 2, 3 respectively for slit shaped pores, voids between randomly-packed spheres, cylindrical pores and spherical pores) [11]. Tortuosityfactor or tortuosity ~ was first introduced by Carman in 1937 [12,13] by reference to a direction that corresponds to a given macroscopic flow. It was defined as the square of the ratio of the "effective average path length" in the porous medium Le to the shortest distance L measured along the direction of macroscopic flow
"r (ae/a) 2
(4.4)
For cylindrical perpendicular pores (case of Anotec | membranes), the tortuosity is equal to unity. For more complex porous structures the tortuosity is usually higher [14]. The Kozeny-Carman equation describing the laminar flow through porous media can be used to calculate the tortuosity of the membrane, by introducing in the equation the membrane thickness, permeability, internal surface area, pore volume fraction and pore shape. Due to variabilities of the parameters involved, this parameter is usually not used to describe a membrane pore structure, but has been mainly introduced to correct the calculated permeability and obtain agreement with the experimental value. This parameter, reflecting the complexity of porous media (random orientation of the pores in the material), is not clearly defined in the IUPAC reports. Other definitions of tortuosity can be given such as those found in [13] for geometric tortuosity or in [15] relating tortuosity to diffusion coefficients and porosity ~. In this latter case tortuosity is expressed by "r, = EDp/ Def f
(4.5)
where Dp is the diffusion coefficient in a pore (cm2/s) and Deff is the effective diffusion coefficient (cm2/s). In some specific cases other parameters can be considered as important in the characterisation of membrane morphology like the surface roughness, pore anisotropy and porous network connectivity [16,17]. Concepts of percolation and fractal geometry are also of interest to better describe the statistical and random structures of many porous solids [14,18,19].
4.2.3 Models for Porous Structures One important, but often not clearly defined variable in the characterisation of porous layers, is the shape (or geometry) of the pores. In order to interpret the characterisation results and relate pore size to physical models, it is often
73
4 -- METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
al
,,2.:'..
9
~7 . ; ~ : ~..~ 9" ~
a2
:.'.: : .. ~i .
~
~
,
a3
r,. ,...
,
,
~ . .~,.~ ; ..... . ~,,~:,
b
c
"":';~?~~":'~ !~";':~~:'"": "
Fig. 4.2. Some idealised pore structures: (a) cylindrical pores (al,a2: parallel non-intersecthlg capillaries; a~: z = 1; a2:1:> 1; a3: non-parallel intersecting capillaries; (b) slit-shaped pores; (c) voids between packed spheres; (d) closed pores; (e) blind pores; (f) open pores; (g) funnel; (h) ink bottle; (i) pores with constrictions. essential to make assumptions about the pore geometry (Fig. 4.2). Classical simple descriptions of pores involve cylindrical, slit or spherical shapes. Simple pore shape geometry can be considered close to reality in some specific cases such as, for example, Anotec | alumina membranes (cylinders), some zeolites (prisms, cavities, windows, etc.), clays and activated carbons (slits). However in most ceramic membranes resulting from oxide sintering pores are voids left between packed particles, they have neither a regular shape, nor regular size and contain constrictions. This kind of material can contain closed pores, blind pores (open only at one end) or through pores (open at two ends). The concept of stochastic geometry of pores has been applied in [20] to simulate by statistical models, the structure of an Anotec | cellular ceramic membrane, a sintered ceramic membrane and a foam membrane. The reported agreement between theory and experiments (laminar flow of pure solvent and during fouling) is very good. A corrugated random pore model has been developed in [21] to predict hysteresis loops and Hg entrapment in mercury porosimetry measurements. Many other elaborate models of membrane structure are described in the literature and a major part is reported in [22].
74
4 - - M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S S T R U C T U R E
In addition, it is not the average pore size which is the determining factor in membrane performance, but the smallest constriction in the porous medium. Indeed some characterisation techniques determine the dimension of the pore entrance rather than the pore size. Such techniques often provide better information about permeation related characteristics, provided that through pores are concerned.
4.3 STATIC CHARACTERISATION TECHNIQUES
4.3.1 Stereology Microscopy can provide visual details of the membrane surface and crosssection morphology. Image analysis of micrographs can give quantitative data from these direct observations. The resolution of optical microscopy, usually performed at a magnification of 500-1000 is often sufficient to observe large defects at the membrane surface but higher resolution is needed for the observation of fine texture. Scanning electron microscope (SEM) generates electron beams and forms an image from the emitted electrons as a result of interaction between the bombarding electrons and the atoms of the specimen. Since electrons have a much shorter wavelength than light photons, higher resolution information can be obtained from SEM than from optical microscopy. SEM is now routinely used to obtain magnifications of 105; the resolution can reach 5 nm. Field emission scanning electron microscopy (FESEM) enable the SEM resolution to be decreased to around 1.5 nm. This improvement is mainly due to the electron source (cold cathode instead of thermoionic source).With the latter technique, samples can be observed at low accelerating voltage and with a lighter m e t a l l i z a t i o n - which is important in the case of fragile membrane structures. Figure 4.3 compares the SEM and FESEM micrographs of mesoporous and microporous silica membranes. The best transmission electron microscopes (TEM) have about I nm resolution. However, because of limitations in sample preparation, frequently the interpretation is restricted to about 10 nm. In the TEM, the electrons that form the image must go through the specimen, which limits the thickness of the sample. The technique is thus only able to analyse unsupported thin membranes or stripped membrane layers. High resolution transmission electron microscopy (HRTEM), reaching a resolution of 0.3 nm, enables the organisation of crystalline planes of atoms to be observed. A new approach to HRTEM is the Z-contrast scanning transmission electron microscopy (STEM) which provides a direct image of material atomic structure and composition. The intensity or brightness of the image is proportional to the square of the atomic number (Z) [23]. Atomic force and scanning tunnelling microscopieshave recently revolutionised
4 ~
METHODS
FOR THE
CHARACTERISATION
OF POROUS
75
STRUCTURE
FESEM
SEM
~ " ~ : :"-" ~ ~-: ,T~r*~-7".~ ~ . ~ , * ~ : ~ ~ ; ~ .~ . ~ , . ~
..
,
:
'
.. . . . .
9 '"~
I
r
~
~
~
~
. ~.~;,~;,~: . .,,. ~
,4.~:,.,, ~,
,.
.
.
.
.
.
""
"'4!,
.
,,
,
...................
'
.
9
~
"
i
,
~y~,;:-~,'.~.~,.~.,,,
,~:,~'~-~,~
-.,~
~
]~ ,.~
,
,
,
................................................................................................................
Fig. 4.3.Imagesof silica membrane surfacesobtained by SEMand FESEM:(a) mesoporous membrane; (b) microporous membrane.
the study of surface structures. These techniques, collectively called scanning probe microscopies, are capable of imaging angstrom to micron sized surface features. They are non-destructive and require little sample pre-treatment. The basic physical phenomenon of scanning tunnelling electron microscopy (STEM) is the "tunnelling effect of electrons in vacuum". A very fine metallic tip-probe is placed at a few atomic distances (d - 5-10/~) of the surface studied, in order to allow a slight recovering of the electronic state functions of the probe and of the sample surface (Fig. 4.4a). A potential difference V (a few mV to a few V) is applied between these electrodes and an electronic current can then go from the probe to the surface (or in reverse, depending of V polarity). These electrons, by the tunnelling effect, get over the potential barrier due to the existing vacuum between the tip and the analysed surface (Fig. 4.4b). By similarity with the tunnelling effect between two flat electrodes, it is possible to estimate the tunnel current intensity I and to predict its variation with the distance p r o b e / s a m p l e [24]. At low polarisation tension limit V and low temperature
76
4 - - M E T H O D S FOR THE C H A R A C r E R I S A T I O N OF P O R O U S STRUCTURE
o v !
a,
....::... . . .
d
b)
ENERGY
EF
SAMPLE TIP Fig. 4.4. Physical principle of tunnelling microscopy [24].
I = V exp(-2h -1 q2m~ d)
(4.6)
where ~) is the potential barrier due to vacuum. For a typical value ~) = 4eV, Eq. (4.6) establishes that I decreases with one order of magnitude for an increase of only I A in the distance probe/sample d. From this extreme sensitivity, the high resolution of the microscope perpendicular to the studied surface is obtained. Lateral resolution depends on the dimensions of the probe used. The samples used must be smooth on a microscopic scale and, up to now, have a regular structure to allow interpretation of the results. In addition, STEM samples must be conductive. Pores on a nanometer scale can be observed, although the interpretation of the results remain the most difficult part. Atomic force microscopy (AFM), developed in 1986, allows the direct observation of non-conductive materials and is attractive for membrane surface examination in air or even under liquids. There is no need for any specific sample preparation before examination. A diamond or tungsten or quartz tip placed at the end of a flexible lever (cantilever), applies a constant force (<< to typical bonding energy) on the atoms of the surface studied. The surface is displaced under the tip and, as the interaction force between tip and surface varies from place to place, a lever deflection is observed. This extremely weak deflection is measured by STEM or by laser interferometry. An image of the interacting forces between the tip and the surface is obtained by measuring the cantilever deflection as a function of
4 - - M E T H O D S F O R THE C H A R A C T E R I S A T I O N OF POROUS STRUCTURE
77
the lateral tip position. The application of this technique to the membrane field has mainly been confined to the study of polymeric membranes [25,26] but the technique has also been adapted to the characterisation (particle size, shape and surface roughness) of ultrafiltration and microfiltration inorganic membranes [27]. Several other techniques referred to as microscopy and based on several different phenomena can be found in the literature. These include acoustic microscopy based on the interactions of acoustic waves with materials [28] the projection microscopy which is still under development and gives a hologram image of the sample illuminated by a beam of low energy electrons [29]. For membrane applications a scanning electrochemical microscope has been developed based on the measurement of the local flux of electroactive ions across the membrane. The ability to detect ~ 1 ~tm radius pores separated by 50-100 ~tm has been demonstrated with mica membranes [30]. Image analysis can be performed on sample micrographs to obtain pore density, pore size, pore area and porosity. The analyser used is generally a video camera linked with a computer, allowing enhancement of the image by adjustment of the contrast and colour overlays of the features (such as pores) prior to measurement. Because some subjectivity is involved in adjusting the contrast and extent of overlay, a statistical analysis of replicated measurements is required. Mean and standard deviations of the populations are necessary to calculate the confidence interval and adequately estimate the sample surface characteristics. Quantitative image analysis is a useful technique which enables reliable numerical data to be obtained from sample surface micrographs [31-33] or digitised negatives [34]. A fractal analysis, studying the behaviour of a geometric parameter over a range of scales of observations has been proposed [35] from SEM images of silica membranes. The principal advantage of quantitative microscopy as a tool for characterising macropores in porous solids is the ability to measure both open and closed porosity and the shape, location and orientation of pores but only two dimensional measurements can be performed [36]. Image analysis is well adapted to estimate the porosity of a supported membrane, this parameter is often difficult to obtain by other techniques. Up to now this method has been applied to a large extent to study the porous texture of organic membranes. Nuclear magnetic resonance imaging has become a powerful tool in medical diagnostics and can also be applied to a variety of ceramic problems, including the imaging of liquids in pores. The simplest information obtained is the spatial variation of porosity which can be obtained by careful selection of the pulse sequence and delays in the imaging experiment. This technique needs significant expertise to obtain reliable data [37]. The dispersion of paramagnetic tracers in porous media can also be studied by NMR imaging and give access to the 3D tracer distribution [38].
78
4 m M E T H O D S FOR THE C H A R A C T E R I S A T I O N OF P O R O U S STRUCTURE
4.3.2 Intrusive Methods
4.3.2.1 Mercury porosimetry In this method, mercury (which is a non-wetting liquid) is forced into the pores of a dry sample. For each applied pressure, the volume of mercury entering the sample porous structure is determined very accurately (e.g. by measuring the variation of capacity induced by the reduction in height of the Hg column connected to the measuring cell). The relationship between pressure P and pore radius rp is given by the modified Laplace equation (Washburn equation) rp = -
27 cos0/P
(4.7)
where r p is the radius of a capillary shaped pore and 7 the surface tension at the liquid/air interface. As mercury does not wet the membrane, the contact angle 0 is greater than 90 ~ and cos0 has a negative value. A widely accepted value for oxides is 0 H g / o x i d e = 140 ~ and ~rlg/air = 0.48 N / m . These values can vary with the experimental conditions (sample material, temperature) and may affect the results. Very high pressures are needed for pores in the nanometer range. Indeed a pore radius of 1.5 nm corresponds to a pressure of about 450 MPa, which may damage the ceramic layer. Typical results for a commercial tubular 0 ~ A 1 2 0 3 asymmetric support (SCT-US Filter) are given in Fig. 4.5. With thin supported ceramic membranes, the pore volume due to the membrane is relatively small and better results are obtained if a major part of the support is scraped off. Specific preparation of samples (e.g. support embedded in a resin) can change the results [39]. If the membrane weight is known and if its pore size can be well differentiated from that of the support the method can be used to determine the porosity of a supported layer. In parallel with mercury porosimetry in which a non wetting liquid is used, we can mention the suction porosimetry in which a wetting liquid like water (0 _<0 < ~/2) is held within the porous solid [5]. In this case the Laplace equation predicts that it will experience a reduced hydrostatic pressure, inversely proportional to the radius of pores in which menisci are formed. The lower limit of pore size accessible to this technique is around a few tens of microns.
4.3.2.2 Gas adsorption/desorption isotherms (physisorption) This technique is one of the most important and extensively used methods in the characterisation (porous volume, specific surface area and pore size distribution) of porous inorganic materials [40,41]. Nevertheless, real solid/gas interfaces are complex, leading to uncertainties in the assumptions made, and different mechanisms may contribute to physisorption (e.g. monolayer-multilayer ad-
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
0.
.. .. .. .. .. .. .. .. :::: : : : =.;-.: +-....
----.~
060
:
....
a
0.05.5
i~i;
-
ii:~:
~:
.'."
.
~
79
.
i
~.~i~
"-~-.
:
.... 9 9 9
.~
~'i~
i
~
..":
9
......
9
Y:
O. 0 5 0
.......
0. 0 4 5
;...... : . . . . . . . . . . . .
:..._..._2._2 . . . .
..
0. 0 4 0
.
.
.
.
.
.
.
.
.
.
: .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . ~ .
. . . . . . . . . . . .
0. 0 3 5 0. 0 3 0 ~
' ~
I
0.
025
-
0.
020
.
. . . . . . . . . . . .
.
.
.
~
.
.
.
.
,.,.
.
::
9
0.015. . . .
..
. . . .
,,,
.
-.
.
. . . .
-"
"
-~"-':
. . . .
.
:':i
.
,, .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . .
.
9. . . . . . . . . . . . . .
"-
-
-
" m
0. 0 0 5
,
. "
0.010
9
'
....
"
......
M
-
0. 0 0 0
-
..
100
10
1
0. 1
0.01
Diameter (/an) 0,1
1
10 (MPa)
Pressure
0.07
~
~
.
.
.
.
.
.
.
.
.
.
.
.
iiiiiii! ! i!
!!!!!
iiiiiii
0.05
-
0.04
-
0.03
-
0.02
-
0 . 0 1
-
0 . 0 0
.
100
i
.
:
......
iiiiii
iiiiiii iiiiiiiiiii!i !iili iii
i
~
100
10
I
Diameter (/an)
0. I
0.01
Fig. 4.5. M e r c u r y p o r o s i m e t r y analysis of a c o m m e r c i a l (SCT-US Filter) t u b u l a r 0~A1203 a s y m m e t r i c m e m b r a n e s u p p o r t (Micromeritics ASAP 2000). (a) C u m u l a t i v e intrusion v o l u m e as a function of the applied p r e s s u r e / p o r e diameter; (b) differential intrusion v o l u m e as a function of the p o r e diameter.
sorption, capillary condensation or micropore filling). Consequently attention has to be paid to the problems and ambiguities arising in connection with the reporting and interpretation of physisorption data. Details of definitions and methodology on this particular technique is given in Ref. [9].
80
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
The adsorption and desorption isotherms of an inert gas (classically N 2 at 77 K) on an outgassed sample are determined as a function of the relative pressure (Prel = P/Po, i.e. the ratio between the applied pressure and the saturation pressure. The adsorption isotherm is determined by measuring the quantity of gas adsorbed for each value of p/po by a gravimetric or a volumetric method (less accurate but simpler). A surface acoustic wave device can also be used as a mass sensor or microbalance in order to determine the adsorption isotherms of small thin films samples (only 0.2 c m 2 of sample are required in the cell) [42,43]. The adsorption isotherm starts at a low relative pressure. At a certain minimum pressure, the smallest pores Will be filled with liquid nitrogen. As the pressure is increased still further, larger pores will be filled and near the saturation pressure, all the pores are filled. The total pore volume is determined by the quantity of gas adsorbed near the saturation pressure. Desorption occurs when the pressure is decreased from the saturation pressure. The majority of physisorption isotherms may be grouped into six types [9]. Due to capillary condensation, many mesoporous systems exhibit a distinct adsorption--desorption behaviour which leads to characteristic hysteresis loops (Type IV and V isotherms) whose shape is related to pore shape. Type I isotherms, characterised by a plateau at high partial pressure, are characteristic of microporous samples. A typical isotherm, representative of a mesoporous sample is given in Fig. 4.6, with a schematic representation of the adsorption steps. Mesopore size calculations are usually made with the aid of the Kelvin equation in the form ln(p/po) = -fyV/rKRT
(4.8)
w i t h f i s the geometrical factor depending on the shape of the meniscus formed by the liquid in the capillary (f = 1 for slit-shaped pores, f = 2 for cylindrical pores); I is the surface tension of the liquid condensate at the absolute temperature T; V is the molar volume of the liquid at the absolute temperature T; and rK the Kelvin radius, dimension characteristic of the capillary (radius of a cylinder or slit width). The pore radius rp of a cylindrical pore may be calculated from rp = r K + t
(4.9)
where t is the thickness of the adsorbed layer of vapour in the pores, before capillary condensation occurs (t is estimated from calibration curves with similar non-porous solids). For a parallel-sided slit, the slit width Wp is given by Wp = r K +
2t
(4.10)
The Kelvin equation is useful to calculate the distributions of pore volume and area as a function of pore diameter in the mesoporous range (isotherms of Type IV). This can be performed by the BJH (Barret, Joyner and Hallenda)
81
4 -- METHODS FOR THE CHARACTERISATIONOF POROUS STRUCTURE
IV DESORPTION
C
II and III ,...
Relative pressure (p/p o)
"'['"
I
.1_.2rp
1
. . . . . v . v :. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
II
IV I l l ........................................................................................................ ii[i2rK [.:+:.:.:.:.:.:.: :.:.:.:.:,-.:.-.-.:.-,--.-.:.-.:.:.:.:.:.:.:.:.:,:.:.:.!.k.:,:.:.:.:.:.:.:.:':.:,: |
Fig. 4.6. A typical adsorption-desorption isotherm for a mesoporous sample showing corresponding steps of adsorption [1]" rp: pore radius; rK: Kelvin radius; t: t-layer thickness.
method which considers opened cylindrical pores and may be applied to the desorption branch. Nevertheless, the Kelvin equation is based on thermodynamic considerations which are not valid for micropores and are only valid for larger mesopores [41,44]. Thus it is claimed that Kelvin type adsorption models overestimate micropore filling pressures and are unreliable for pore size distribution determinations below 75 A [45]. Nevertheless the method has recently been applied to considerably lower limits (1.3-1.7 nm) [10] which highlights current uncertainty and the complexity which still exists with this problem. The oversimplified BET (Brunauer, Emmet and Teller) theory, valid for relative pressures between 0.05 and 0.35, allows the calculation of the specific surface area of solids and the estimation of the interactions between the solid and the vapour (from the value of the constant c). The BET equation is mainly applicable for Type II and IV isotherms. The specific surface area deduced from Type I isotherms has no physical meaning because the notion of a monolayer is not applicable in the case of micropores [9,41,46,47].
82
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
Case of microporous systems The classical methods for interpreting adsorption data rely on equations that are of uncertain validity for micropores and small mesopores mainly because of increased adsorbate-adsorbate and adsorbate-adsorbent interactions. Many papers in the literature deal with this problem of characterisation of micropores from adsorption experiments [6-8,10]. Several methods have been proposed to determine the microporous volume accessible to a given gas, namely the t-plot, the nonane preadsorption technique and the Dubinin-Radushkevich equation. The simplest technique to determine a sample microporous volume is the t-method of de Boer and its extensions. This method compares the isotherms of porous solids with standard isotherms obtained with non-porous reference solids of similar composition. In the original t-method, the amount adsorbed is plotted against t which is the multilayer thickness calculated from the standard isotherm. Any deviation in shape of the given isotherm from that of the standard is detected as a deviation of the "t-plot" from linearity. For the assessment of microporosity, the thickness of the multilayer is irrelevant and Sing proposed that this is preferably replaced by c~s = na/n~ where n asis the amount adsorbed at a fixed P/Po = s. [9,48]. Usually s is set at 0.4 (which is the limit between meso and micropores). Figure 4.7 shows typical types of t-plot and ms plots [10]. The nonane preadsorption technique consists in measuring the adsorption before and after micropore blocking with nonane which is held inside the pores even after degassing at room temperature [49]. The Dubinin methods are based on the theory of micropore volume filling [50,51]. The total microporous volume accessible to a given adsorbate Wo can be obtained from the DR equation
v
Micr~176176
(cm~/g)
9
Microporous volume ' "':'i....... :'i"i; 9
"
"
..~...,..' !
"
:
t (A)
.
..'""
9 i
20
Fig. 4.7.Typicalt-plot obtained from adsorption isotherms on mesoporous, microporous and nonporous samples.
4 m M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S S T R U C T U R E
W = Wo exp [-(A / ~Eo)2]
83
(4.11)
where A = RT ln(po/p) is the adsorption potential and Eo and [~ are specific parameters of the system studied (viz energy of adsorption and affinity coefficient) and which need to be known. In order to estimate the pore size distributions in microporous materials several methods have been developed, which are all controversial. Brunauer has developed the MP method [52] using the de Boer t-curve. This pore shape modelless method gives a pore hydraulic radius r H, which represents the ratio porous volume/surface (it should be realised that the BET specific surface area used in this method has no meaning for the case of micropores!). Other methods like the Dubinin-Radushkevich or Dubinin-Astakov equations (involving slitshaped pores) continue to attract extensive attention and discussion concerning their validity. This method is essentially empirical in nature and supposes a Gaussian pore size distribution. An alternative method proposed by Horvath and Kawazoe for slit shaped pores [53] and recently extended to cylindrical pores [54], is to calculate an average potential function inside the micropores, relating the average fluidfluid and solid-fluid interaction energy of an adsorbed molecule to its free energy change upon adsorption. This method is an improvement over the Kelvin approach in that it acknowledges the strong fluid-solid attractive forces in micropores, but does not take into account pore wetting [45]. Another method, based upon an improved molecular description of adsorption (modern statistical mechanics) using a nonlocal Density Functional Theory (DFT) [55], has also been applied to derive pore size distributions in activated carbons (slitshaped pores). The method leading to pore size distributions in the whole range of micro to mesopores is being extended to cylindrical pore geometries for modelling silica and zeolites. It must be concluded that the quantitative determination of micropore size is still an ambiguous problem: new theories, models, mechanisms and simulations are still under study [56-58]. Therefore isotherm interpretations must be used carefully and can be considered as useful mainly for qualitative studies. No reliable method has been developed for the determination of the micropore size distribution. At present the most promising approach appears to be that of pre-adsorption linked with the use of various probe molecules of known size and shape [59--61]. For example, this approach has been applied successfully for silica compacts characterisation in [61] using spherical symmetrical inert molecules, such as neopentane and trimethylsiloxysilane [(CH3)3SiO]4Siwith diameters of-~6.5 and 11.5 A respectively. In general the limited availability of volatile probe molecules with diameters extending above 10 ~ puts a restriction on the applicability of this method. Furthermore effective pore sizes determined by this technique depend on the kinetic and thermodynamic properties of the
84
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
adsorbate in the porous material (rate of diffusion, total porous volume, electrostatic interactions, etc.). To make further progress it will be necessary to employ well defined microporous and small pore mesoporous adsorbents and have available non-porous reference solids of the same surface structure. This will be possible with the current progress in material synthesis procedures (e.g. organised amorphous structures obtained in the presence of templating agents [62,63]. We have to note that zeolites are often not suitable reference materials because the strong electrostatic fields within the crystalline cavities can polarise the gas molecules. Therefore the adsorption equilibrium in these materials is not solely a function of the size of the adsorbate. The analysis of thin supported films by N2 adsorption is often difficult due to the very small percentage of pore volume contributed by the thin layer relative to that of the support. Usually it is necessary to scrape off most of the bulk support layer to increase the pore volume percentage of the thin film. Recent technical improvements in pressure sensors on commercial apparatus (reaching now a sensitivity of 5.10-s mmHg) or new sophisticated detection techniques using surface acoustic waves may in some cases solve this problem.
4.3.2.3 Calorimetric Determinations Immersion calorimetry The measurement of the heat of immersion of a "dry" material in different liquids can permit a rapid and accurate determination of the surface area and pore size distribution below 10 A. The enthalpy change is related to the extent of the solid surface, to the presence of micropores and to the chemical and structural nature of the surface. The technique has been mainly applied to carbons [64]. The immersion liquid is usually water for hydrophilic oxides like mineral oxides, or an organic liquid (benzene, n-hexane) for hydrophobic solids like carbons. One of the limitations of this technique is that the specific enthalpy of immersion of the open surface must be determined with a non-porous standard material of surface composition similar to the porous solid studied. The non-microporous part of the surface area can be determined by prefilling the micropores with an absorbate prior to immersion. Information on the size of micropores can be obtained from the kinetics and enthalpy of immersion into a set of liquids with increasing molecular size [5].
Thermoporometry Thermoporometry is a thermal method which is based on the thermal analysis of the liquid-solid phase transformation of a capillary condensate held inside the porous body under study. The technique was developed by Brunet
4 - - METHODS FOR THE CHARACTERISATIONOF POROUS STRUCTURE
85
al. in the 1970s [65] and is based on the principle that the equilibrium state of the solid, liquid and gaseous phases of a highly dispersed pure substance is determined by the curvature of the interfaces. In a liquid filled porous material, the solidification temperature of the liquid depends only on the liquid/solid interface (which is always almost spherical) in the pore [66,67]. The smallest size of a stable crystal (R~) is inversely proportional to the solidification temperature depression AT = T - To
(4.12)
where To is the normal phase transition temperature of the liquid and T is the temperature where the phase transition is actually observed when this liquid is contained in the pores. In finely porous materials, the liquid therefore crystallises or melts at the temperature where the pore radius r equals Re + t, where t is the thickness of a liquid-like film adhering to the solid matrix of the porous material and which does not undergo a change of state [68]. The melting or solidification thermogram can be monitored in a differential scanning calorimeter (DSC). The exothermic heat effect measured from the solidification of a liquid in a porous medium is schematically shown in Fig. 4.8. From the solidification thermogram it is possible to determine: - the pore radius distribution (between 1.5 and 150 nm), from the depression of the solidification temperature, AT, due to the Gibbs-Thompson effect, - the pore volume, from the energy Wa evolved in the phase transformation. This determination must take into account that the heat of the phase transformation i s a function of freezing point depression, the pore surface, from the simultaneous measurements of AT and Wa. -
1
'
........ i
............. I ! i I I
. . . . . . . . . . . . . . . .
i/ .
~
..........
.................
0
T(~
Fig. 4.8. Schematic illustration of the exothermic heat effect from the solidification of a liquid in a p o r o u s m a t e r i a l as a function of t e m p e r a t u r e . (A) N a r r o w p o r e size distribution; (B) b r o a d pore size distribution [1].
86
4 w METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
The relation between pore size r(nm) and the extent of temperature depression AT (Fig. 4.8) is obtained from the equations derived by Brun from GibbsDuhem and Laplace equations. For cylindrical pores with water inside (0 < AT < -40 and t = 0.8 nm), it leads to during solidification: r(nm) = (-64.67/AT) + 0.57
(4.13)
during melting: r(nm) = (-32.33/AT) + 0.68
(4.14)
In the same way, the apparent energy of solidification Wa can be measured and related to AT by two other equations. The corresponding pore volume d V between T and T + AT is given by
dV=dW/PWa
(4.15)
where p is the density of water corresponding to AT. By combining the various equations, a pore volume distribution as a function of pore radius can then be derived from the thermograms [65]. The difference between the solidification and melting temperatures in cylindrical pores is due to the fact that the shapes of the interfaces present during these transitions are different. In spherical shaped pores however, there is no difference and the same thermodynamic equation can be used to describe both solid -+ liquid and liquid --> solid transitions. Consequently by analysing both the melting and solidification curves, one determines a pore shape factor. In thermoporometry the shape factor for a porous material [68,69] can vary generally between I (spherical pores) and 2 (cylindrical pores). In membrane filtration, water-filled pores are frequently encountered and consequently the liquid-solid transition of water is often used for membrane pore size analysis. Other condensates can however also be used such as benzene, hexane, decane or potassium nitrate [68]. Due to the marked curvature of the solid-liquid interface within pores, a freezing (or melting) point depression of the water (or ice) occurs. Figure 4.9a illustrates schematically the freezing of a liquid (water) in a porous medium as a function of the pore size. Solidification within a capillary pore can occur either by a mechanism of nucleation or by a progressive penetration of the liquid-solid meniscus formed at the entrance of the pore (Figure 4.9b). Thermoporometry is a method which measures cavity sizes and not inlet sizes. It has been mainly used for the characterisation of organic mesoporous membrane texture [70-73] but has been also applied to inorganic alumina symmetric membranes [73] with a good reliability. However the solidification of water in small pores may sometimes damage the membrane structure due to the expansion of the condensate and consequently different results can be obtained after several runs [74].
87
4 m METHODS FOR THE CHARACTERISATIONOF POROUS STRUCTURE
•• ,~
r3
rl
r2
BULK PHASE
... :.......
/~/U D-LI,KELAYER,t ~-.- .~. : ' r
:-'-:'" :":.'-"::",":-7.:..
PENETRATION
t
NUCLEATION
Fig. 4.9. Schematic d r a w i n g of the extent of depression of the solidification temperature: (a) in relation to the pore diameter; L = liquid (water); S = Solid (ice); r = pore radius (rl > r2 > r3) [2]. (b) detail of the solidification processes [68].
4.3.2.4 Nuclear magnetic resonance Nuclear magnetic resonance, which is sensitive to short-range order, has been recently used to obtain information on the structure of pores. Two main techniques can be found in the literature [75]" one is based on the study of NMR relaxation times of a fluid inside pores and the other on the chemical shift of 129Xe trapped in the material.
N M R spin-lattice relaxation measurements The use of the low-field NMR spin-lattice relaxation technique has recently been successfully demonstrated [75-80] as a pore structure tool for porous materials saturated with a solvent (usually water). The basic principle is that the portion of pore fluid near a pore wall undergoes spin-lattice (T1) and spin-spin (T2) relaxation in a magnetic field at a faster rate than the bulk fluid. This, coupled with the fast diffusional exchange of fluid between regions within
88
4 m METHODS FOR THE CHARACTERISATIONOF POROUS STRUCTURE
"--'- Surface affected phase 4----
Bulk fluid phase
Pore wall
Fig. 4.10. Schematic diagram of pore fluid during an NMR experiment [80].
a pore, yields a unique average relaxation time constant for each different geometrical environment (i.e. pore size) within a porous material. This relaxation time constant is dependent on the ratio of surface fluid to non-surface (or bulk) fluid in the pore, which is a measure of the pore hydraulic radius. From the two-fraction fast-exchange model, the measured T 1 can be related to the pore hydraulic radius rH by 1 / T 1 = o~ + ~ / r H
(4.16)
where c~is 1/Tlbulk is determined from the spin-lattice experiment on the fluid only and ~ is a surface interaction parameter determined with partially saturated samples with different fluid contents. The physical model associated with this equation is illustrated in Fig. 4.10. The thickness of the surface affected phase is typically 0.3 + 0.1 nm. A saturated porous medium, such as a membrane, should relax as a combination of its different geometrical regions, i.e. pores. A pore size distribution can be determined via NMR relaxation experiments if the observed relaxation curve can be deconvoluted into its component parts, and if the relationship between pore size and relaxation time constants can be determined. The technique can be applied to macro, meso and microporous samples if the pore geometry is known or assumed. Examples of pore structure characterisation of flat organic and inorganic membranes are given respectively in [77] and [76], NMR results tend to compare favourably with those of other more classical techniques The maximum size of the membrane samples to be analysed by this experimental method is limited by the homogeneity of the magnetic field (around 10 cm). 129Xe N M R
Another adaptation of the NMR technique involves the measurement of the chemical shift of 129Xe adsorbed in a sample. The recent development of this last technique has mainly been concerned with the study of the porous structure of microporous materials such as zeolites [81,82], mesoporous silica [11,83], chlatrates [84], organic polymers and supported metal catalysts [82]. 129Xe is an inert
4 -- METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
89
spherical molecule with a diameter of 0.44 nm and a 1/2 spin nucleus. Polarisation of its spherical electronic shell, during adsorption or due to collision with other Xe atoms or with the surface of a solid, affects the NMR chemical shift. Consequently by studying the resonance line (shape and position) as a function of the quantity of adsorbed xenon, information can be obtained on the porous structure of sorbents as well as on the rate of exchange between Xe atoms in the gas phase and those adsorbed on the sorbent surface. Nevertheless in most cases, the interpretation of the results is not easy and depends also on the chemical nature of the materials, homogeneity, crystallinity and on the ease of diffusion of Xe molecules in the pore structure. In the case of zeolites, the interpretation of 129Xe NMR chemical shift ~Xeis usually described using the following relationship [82] PXe ~Xe = ~o + ~s + ~E 4- y ~(Xe-Xe) ' d P x e 0
(4.17)
where 8o is the reference (chemical shift of xenon gas at zero pressure), 8s is the contribution due to collisions of xenon atoms with the zeolite walls, 8E arises from the electric field created by cations and PXe is the density of Xe adsorbed in the cavities. The dependence of 8 upon the quantity of Xe is usually explained in terms of the contribution due to collisions between Xe atoms in the micropores, resulting from the increase in the local Xe density compared to the equilibrium one. The slope of the straight line for the relationship 8Xe = flPXe) is an important parameter which indicates the range of pore sizes. This slope is usually large for microporous materials or zeolites and near zero for mesoporous silica. From a qualitative point of view, the shape of the NMR peaks can, in tortuous materials, be considered as an indication of the homogeneity of the material pore structure. Indeed two distinct signals can be obtained [85] by mixing two samples of sol-gel derived microporous silicas prepared from different TEOS sols [86]. Figure 4.11 compares the pore hydraulic radius distributions (obtained by the MP method from N2 adsorption isotherms) of isolated and mixed samples and the corresponding NMR peaks. For a quantitative exploitation of the technique, several fundamental problems relate to the extent and rate of motion which the Xe atoms undergo in the porous solid and the interpretation of the average chemical shift [81]. Even empirical correlations which have been proposed in the literature [87,11] relating the chemical shift (extrapolated at PXe"-O) to pore size cannot be considered as general and often give inconsistent results from one type of material to another [75,83]. NMR techniques are powerful but not yet entirely general. They have probably scope for more development, are non destructive and can provide microscopic information on pore networks. These techniques have future scope in
4 -- METHODS FOR THE CHARACTERISATIONOF POROUS STRUCTURE
90
1..Nz":ADSORPTION
129Xe
]
NMR
i
SAMPLE A ] 0,3
~
t
0,2
t
-~0,1
o
o 4 s t2 16 Pore hydraulic radius r. (A)
9"150 " ' '100
50.6xe(ppm)
fSAMPLE B i "~ 0,3
=~ 0,2
~,
0
4
8
12
-16
Pore hydraulic radius rH (/x.)
' 1sO " " 1 o 0 '
SAMPLEA+B
"SO3xe(ppm)
[
i
" ~ 0,2
0,1
0
4
8
12
16
Pore hydraulic radius r. (A)
"~1~ . . . . 160 . . . . 506xe(ppm
Fig. 4.11. Comparison of the hydraulic radius distributions of pores (obtained by the MP method from N2 adsorption isotherms) of separate and mixed silica samples and the corresponding 129Xe
NMR spectra [85]. particular for a better characterisation of microporous materials. A better understanding of the physical chemical phenomena is required for the technique to be fully exploited. In this respect, measurements at below ambient temperature [75] and at progressive degrees of pore filling are informative.
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
91
4.3.3 Non-intrusive Methods
4.3.3.1 Radiation scattering Radiation scattering from solids can arise from variations of scattering length density (see below) which occur over distances exceeding the normal interatomic spacings. Such variations occur when solids contain pores, and details of the porosity and surface area can be obtained from measurements of the angular distribution of scattered intensity. The appropriate angular range (see below) where this information is contained is given by d ~ ~/20
(4.18)
where d is the pore size and ~,the wavelength of radiation, which may be X-rays, neutrons or light. Since the theory and analysis of radiation scattering from porous solids has been published extensively [88-90], it will consequently not be detailed here. In a scattering experiment, a monochromatic beam of electromagnetic radiation (light, X-rays) or neutrons, of wavelength ~o, intensity Io and wavevector Ko, is directed on a sample and the scattered intensity I(Q) is measured as a function of angle, 20, to the incident direction (Fig. 4.12). Here Q is the momentum transfer (Q = 4~zsin0/~,). Typical wavelengths for different types of radiation and the corresponding spatial resolution of density fluctuations or inhomogeneities such as pores, which may be determined experimentally are given in Table 4.1. The form of the scattering curve obtained from radiation scattering measurements depends on the complexity of the system. Arelatively simplecase which is often encountered is that of a two-phase system, composed of particles (or pores) having a homogeneous scattering density, dispersed in another continuous medium, such as a liquid (or solid for the case of pores). Here an analysis of the scattering can provide details of the spatial and orientational correlations of the particles (or pores) [91]. A simplified analysis of small angle scattering frequently involves the Guinier approximation which is valid for randomly distributed (uncorrelated) systems of particles (or pores) with a relatively
Is,k~ Io, ko
~~.."~
20
\\
r
SAMPLE Fig. 4.12. Schematic r e p r e s e n t a t i o n of the incident a n d scattered b e a m s in a SAS experiments.
92
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
TABLE 4.1 Typical wavelengths for different types of radiation and corresponding spatial resolution of inhomogeneities Type of radiation
Typicalwavelength Inhomogeneity (nm)
Light X-rays Neutrons
400-600 0.1-0.4 0.1-2.5
Spatial resolution (nm)
Refractive index 200-20 000 Electron density 5-50 Scattering length density 50-500
narrow distribution in size. The Guinier equation (4.19) describes the form of the scattering I(Q) in the initial region of the curve (see Fig. 4.13)
I(Q) -- e x p ( - Q2. R~)/3
(4.19)
where Rg is the radius of gyration of the particle or pore (NB: for spheres of radius R, R = 1.29 Rg). This relationship is valid when Q.Rg < 1 and applies to inhomogeneities of any shape. At higher Q, (Q.Rg > 4), I(Q) decays approximately as Q-4, which is predicted by the Porod law [92] viz. "" al
I~ I
b)
,
rZGuinier region
~ -'.\
a) I i I _ ~ It ~
~1 I
\ I
\l
I ~l~i~..i.. [
t --"- ................ ~ 1
\\~....~j!
,, 1
!-
\
l p-
Mass ffactal \ range \ ,Porod ~ \ reg!on.'.
/
[-/
/
~
I
I
I log Q
\(Surface fractai \ range)
I~
J
Fig. 4.13. Schematic representation of a particle aggregate (a) having a range of self-similarity between approximately al and a2 (mass fractal structure). The form of the scattering expected is depicted in (b) [91].
4 - - M E T H O D S FOR THE CHARACTERISATION OF POROUS STRUCTURE
I(Q) ~ ( S / V ) . Q-4
93 (4.20)
where ( S / V ) is the surface to volume ratio of the particle (or pore). This law is obeyed when the interface between the two phases is smooth in the scale of Q-space measured. A more general expression which applies to surfaces which are irregular or have a curvature on a scale smaller than the reciprocal Q space is given by I(Q) ~
S
.
Q-(6-D~)
(4.21)
where Ds is the surface fractal dimension. For a 'smooth' surface Ds is 2 (the Euclidean 2D dimension) whereas for fractally rough surface Ds may approach 3. This leads to power law exponents which may range between 3 and 4. Such situations frequently arise for systems containing very small pores, such as amorphous microporous oxide gels. Scattering techniques are non intrusive, non-destructive and have particular advantages in the characterisation of the surface and porous properties of materials containing either closed or molecular sized pores and where outgassing pre-treatment may result in irreversible changes in microstructure. Measurements may be carried out, indeed, in the presence of a gas or with hydrated materials and there is no need to evacuate the sample. Both small angle X-ray (SAXS) and neutron scattering (SANS) are established techniques and their experimental application is similar. However, limitations on sample size, thickness and containment are much more restricted with X-rays because of absorption of radiation. One problem which can arise with neutrons is the subtraction of the flat incoherent contribution which can be quite large in the case of hydrogenous materials. This disadvantage can be partially offset by the possibility of using isotopic substitution. SANS is particularly powerful because the penetrating power of neutrons makes it possible to study material microstructure in the wet state. Instrumentally, both SAXS and SANS require a source of radiation, collimation system, sample containment and a detection system. Important information on the contribution of porosity contained in closed and open pores, pore morphology, pore size distribution and/or surface texture can be derived from X-ray and neutron scattering. This is achieved by condensing probe molecules having a similar scattering cross section to the solid, in the open pores ~ a technique known as contrast matching [93,94]. This approach has been developed in detail using SANS where isotopic effects have been exploited (e.g. H20/D20) to study oxide gels and porous carbons [94,95]. Small angle scattering has been used extensively to analyse a wide range of porous materials (ceramics, carbons, oxide gels, cement, bone materials, etc.) [96-99]. An important and recent development concerns the investigation of materials which contains an oriented porous texture, such a fibres and layer-
94
4 - - M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S S T R U C T U R E
~~6
b
! 04
10:(Y ~'
FI
ACF Neutron beam
.-. .0
I--
% 0
ii
0
,02
1
10-3
J
~
10-2
~
QIj, -~
10'1
Fig. 4.14. SANS of oriented activated carbon fibres. (a) Schematic representation of a sample of carbon fibres oriented with their axes parallel to the incident neutron beam; (b) corresponding anisotropic scattering along the two axes of the 2D detector. (i) SANSalong the vertical axis, (ii) SANS along the horizontal axis of the detector [101]. like materials. Frequently the pores are highly anisotropic and aligned with respect to a specific particle orientation forming the porous texture. For these materials unique microstructural information can be derived from small angle scattering measurements with both neutrons and X-rays. These details are not obtainable from bulk measurements, such as adsorption isotherms. This application of SANS has recently been demonstrated with ceramic alumina fibres [100] and microporous carbon fibres [101]. Such measurements on materials having anisotropic pore structures require the detailed analysis of the equivalent anisotropic scattered intensity, measured on a two-dimensional detector, as shown for a sample of aligned activated fibres (ACF) in Fig. 4.14. The application of small angle scattering to determine the structure of membrane material in situ is illustrated by recent investigations of Anotec | alumina membranes containing oriented cylindrical pores [102] and those on oriented porous sol-gel layers deposited on metal substrates [103]. In future such applications will increase the growing availability of SAS facilities at synchrotron and neutron sources.
4.3.3.2 Wave propagation Ellipsometry If light is reflected at the boundary of two different optical media, the polarisation of the electromagnetic vibration is changed according to the Fresnel equations [104]. The change of the status of polarisation is characterised by
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
95
two ellipsometric angles representing the phase shift and the change of the amplitude. When analysing a thin layer supported on a substrate with a knowr~ index (e.g. Si), the measurement of these two angles at extinction gives information on the layer thickness and refractive index n. This latter parameter is sensitive to the film porosity. Knowing the index of refraction for the constituents of the film, various optical models can be used to calculate the volume fraction of solid [105]. The effective medium theories, i.e. theories of Bruggeman, Garnett and Lorentz-Lorenz, for a two phase system are based on the following equation (n 2 - n~) /(n 2 + 2n 2) = Va[(n 2 - n~) /(n 2 + 2n2)] +
(4.22)
Vb [(n~- n 2)/(n~ + 2n~)]
where Va and Vb are the volume fractions of phase a and b and na and nb are the refractive indices of these phases respectively. The measured refractive index of the film is n and nh is a host refractive index which is assigned different values according to the model used. In the Bruggeman effective medium approximation, which can be successfully applied to films with a random configuration of two phases, n h is set equal to n. The Garnett theory (in which n h = na o r nb) is generally applied to a film consisting of one phase completely surrounded by the other. The Lorentz-Lorenz theory is obtained by setting nh equal to 1, i.e. empty space and is the most frequently applied to estimate thin film porosity [106,107]. The volume fraction of solid Vs is obtained through the simplified equation V s = [(y/2 - 1 ) (y/2 +
2)]/[(n 2 + 2) (n 2 - 1)]
(4.23)
in which n a is the refractive index of the solid skeleton. Other relationships can be found in literature between refractive index n and density p like the Gladstone-Dale equation [108,109] p = Kx(n- 1)
(4.24)
Ultrasonic methods
The attenuation and velocity propagation of ultrasound in porous solids depends on the pore size and porosity. The theory of the method is very complex and has been developed mainly for two phase systems containing heterogeneities (e.g. solid/voids). Ultrasonic methods have applications in the area of non-destructive testing of materials, particularly metals and ceramics, where such heterogeneities may be interfaces, pores, inclusions, grain boundaries, or compositional variations [110]. Experimentally measured quantities are the backscattered ultrasonic signal or characteristics of the forward-scattered wave, such as propagation velocity or attenuation. Ultrasonic techniques can be applied to assess the width of voids, d, and porosity, 8, from variations in the
96
4 - - METHODS FOR THE C H A RA CT E RI SA T I O N OF POROUS STRUCTURE
ultrasonic velocity, v, and attenuation 0ca [111]. The method is appropriate when o.d > 0.2, where r~ is the longitudinal wave number in the matrix given by r~ = ~ / v
(4.25)
co(= 2~:v) being the circular frequency of the sound of frequency r~. Typically, when v > 107Hz, the method is applicable for the non-destructive examination of voids in solids, when d > 1 ~tm [5]. Recent developments have extended the ultrasonic techniques to the characterisation of thin layers of metals, and polymers deposited on substances to obtain measurements of the thickness/density product. Using techniques where the film are immersed in a fluid, such as water, measurements have been made, by the low frequency normal incidence double through-transmission method, with film thickness ranging from 20 to 200 ~tm [112] a range which is of particular relevance to membrane systems. An example of acoustic microscopy investigations (by microechography, acoustic signature V(z) and acoustic imaging) of porous silicon layers (290 ~tm) is given in [113]. Information which can be obtained concerns the elastic properties of the material, the layer thickness and density as well as a mapping of the surface and subsurface of the material, revealing roughness and defects. Surface acoustic waves (SAW), which are sensitive to surface changes, are especially sensitive to mass loading and theoretically orders of magnitude more sensitive than bulk acoustic waves [43]. Adsorption of gas onto the device surface causes a perturbation in the propagation velocity of the surface acoustic wave, this effect can be used to observe very small changes in mass density of 10 -12 g / c m 3 (the film has to be deposited on a piezoelectric substrate). SAW device can be useful as sensors for vapour or solution species and as monitors for thin film properties such as diffusivity. They can be used for example as a mass sensor or microbalance to determine the adsorption isotherms of small thin film samples (only 0.2 c m 2 of sample are required in the cell) [42]. 4.3.3.3 Ion-beam analysis
The application of ion beam analysis techniques to determine pore size and pore volume or density of thin silica gel layers was first described by Armitage and co-workers [114]. These techniques are non-destructive, sensitive and ideally suited for the analysis of thin porous films such as membrane layers (dense support is needed for backscattering). However, apart from a more recent report on ion-beam analysis of sol-gel films [115] using Rutherford backscattering and forward recoil spectrometry, ion beam techniques have not been developed further despite their potential for membrane characterisation. This is probably due to the limited availability of ion beam sources, such as charged particles accelerators.
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
97
,
~ lOOOO-
.............. ........
PEAK
rO
E,~ = 3.3 MeV I~
sooo
~".~i
"-" ,
OXYGEN
.~ ;
/
szuco~
:
4,
EDGE
9k
~-v,l '~"- " . " "
0
"~
.........
.~
1
2
E N E R G Y OF BACKSCATTERED P A R T I C L E S (MelO
Fig. 4.15.Experimentalarrangement and RBSspectrum of porous ( ~ ) and non-porous (.....) silica thin films [114]. The work of Armitage and co-workers was carried out using both tandem Van der Graaff and linear accelerators as sources for focused proton and o~-particle beams. The method involves observing the energy distribution of ions elastically backscattered after exciting a resonance in one of the nuclei of the sample. In this work the resonances for protons (2.66 MeV) and ~-particles (3.05 MeV) scattered from 160 were employed. An illustration of the experimental arrangement with an example of a spectrum obtained from a porous S i O 2 sample with 3.3 MeV ~-particles is s h o w n i n Fig. _4.!5, together with the spectrum from a sample of non-porous S i O 2 glass. It will be noted that the resonance peak for the porous sample is broader. This peak is associated with the 3.05 MeV resonance arising from the elastic scattering of o~-particles from 160. A description is given in [114] of the methods of calculation of pore size from such backscattering data, with various assumptions for the pore and interpore path length distributions. This method, using a microfocused beam, has unique advantages over other techniques which could be very useful in membrane characterisation. Thus in the above work examples of measurements made on gel layers as a function of sampling depth (from -3 ~tm to 100 ~tm) and as a function of distance across the sample were illustrated. It will also be noted that the technique is equally appropriate for measurements in the micro and mesoporous ranges. 4.3.3.4 Positron lifetime spectroscopy When a positron (generated e.g. by a 22Na source) enters a condensed medium, it may be annihilated directly with an electron, or it may capture an
98
4 B METHODS FOR THE CH A RA CT E RI SA T I O N OF POROUS STRUCTURE
120 100 80
@
%,)
~
40
.
20
o
so
,00
,so
Pore diameters, Nz adsorption results (~
Fig. 4.16. Positronium lifetime dependence on pore diameter. Solid line is the theoretical relation found from the SchrOdingerequation for spherical pores [118]. electron to form a hydrogen-like atom, called a positronium. Ortho- or parapositroniums decay spontaneously by direct annihilation but with very different time scales (0.1 ns for the para- and 110 ns for the ortho-positronium in air). The lifetime of orthopositronium (measured with a scintillation counter detecting the I radiation) depends on its interaction with surrounding molecules (pick-off quenching caused by interactions with unpaired molecular electrons). Consequently, a positronium decays faster inside a pore than in air. The various measured lifetimes can be attributed to several kinds of trapping sites: dislocations, intersections of two or three crystallite interfaces and internal surface of larger voids. The experimental set-up consists of a positron source (22Na), a scintillation counter, to detect the I radiation from the positronium decay, and electronic peripheral equipment to analyse the time spectrum of the positron annihilation. Positronium lifetime spectroscopy is particularly well suited for studying defects in crystals and structural fluctuations in amorphous materials and can give an estimate of free volumes in condensed matter [116]. It is a useful technique to estimate the free volume of polymeric membranes [117]. In a study on silica gels, the decay lifetime has been found (Fig. 4.16) to be proportional to the pore diameters (measured by N2 adsorption) between 30 and 100 ~ [118]. Information on pore size distribution and surface area may also be obtained by means of calibration curves. 4.4 DYNAMIC CHARACTERISATION TECHNIQUES 4.4.1 Rejection measurements
Rejection measurements with reference molecules like dextrans, proteins or polyglycols are often used by membrane manufacturers. A parameter extensively used for membranes characterisation is the "cut-off" value, which is
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
99
defined as the lower limit of solute molecular weight for which the rejection is at least 90%. It is argued that these rejection measurements have the closest parallel to operating conditions. But it must be realised that rejection measurements always depend on the type of solute (shape and flexibility of the macromolecular solute, charge of the solute), the membrane (its interaction with the solute) and the process parameters used (pressure, cross-flow, velocity, geometry of the test cell, concentration and type of solute, pH). In particular charge effects, concentration/polarisation, pore blocking and fouling phenomena will affect rejection measurements considerably. Recent studies based on comparison between gel permeation chromatography and ultra/micro-filtration [119] have shown that whatever the chemical nature and shape of the model macromolecule used, it is possible to predict the cut-off value of a membrane by considering the hydrodynamic volume of the macromolecule. This parameter provides an appropriate definition of the effective solute size to be considered in hydrodynamic models.
4.4.2 Liquid Displacement Techniques 4.4.2.2 Liquid~gas methods (bubble point, liquid expulsion permporometry) This very simple and established method has become a standard technique used by suppliers to measure the largest active pores (as well as cracks or pinholes) in a membrane. The principle is to measure the pressure needed to force air through a liquid-filled membrane. The bottom of the filter is in contact with air and, as the air pressure is gradually increased, air bubbles penetrate through the membrane at a certain pressure. The pressure and pore radius are related by the Laplace equation rp =
2y cos0/AP
(4.26)
where rp is the radius of a capillary shaped pore and y the surface tension at the liquid/air interface (~/(water/air)---- 72.3 10-3 N/m; ~/(t-butanol/air)- 2 0 . 0 1 0 -3 N/m; "~(fluorocarbon/air)= 16.0 10-3 N/m). An air bubble will pass through the pore when its radius is equal to that of the pore (Fig. 4.17), assuming that the contact angle is zero. Penetration will first occur through the largest pores and since the pressure is known, the pore radius can be calculated from Laplace equation. With water as the wetting medium, the water / air surface tension is relatively high and it is necessary to apply a high pressure if small pores are present (145 bars for a pore radius of 0.01 ~tm); water can then be replaced by another liquid (e.g. alcohols or hydrocarbons). Nevertheless, as the method is dependent on the type of liquid used (different wetting effects), i-propanol is often used as a
100
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
P2
bubble ,~ ,,~..~, ,,.
.....
9
, ".
"
.
,
".;.
i.
9 ,.
,o
,,
"
L.
Wetting liquid i
.
_
i
Porous Pl
membrane
Fig. 4.17. Principle of liquid expulsion permporometry.
D.now /
Gas flow
/ Bubble point
~,~
(
/
, 4/
"Wet" I flow c u r v e
l
~ff Mean flow p o r e ...1 Pressure Pore d i a m e t e r s
Fig. 4.18. Maximum and mean flow pore size determinations by bubble pressure point test [3].
standard liquid. The rate at which the pressure is increased and the pore length can also influence the measurements. The mean pore size and pore size distribution can be evaluated by performing this measurement by a stepwise increase of the pressure. In this case the gas flow across the wet defect-free membrane is recorded (Fig. 4.18) as a function of the applied pressure difference across the sample ("wet curve"). The point of first flow is identified as the "bubble point". This continues until the smallest detectable pore is reached. Then the flow rate response corresponds to the situation in a completely dry sample. The measurement of gas flow through the same membrane in a dry state gives a linear function of the applied pressure difference ("dry curve"). The pressure at which the "half-dry" curve intersects with the "wet" curve can be used to calculate an average pore diameter. Pore number distributions can also be derived from flow distribution curves. This method has been approved as a ASTM procedure [120-122] and used in the commercial computer controlled Coulter Porometer (Coulter Electronic Ltd) for pore sizes much larger than 0.44 lxm. However the theoretical basis used for the evaluation of the accumulated data neglects the specificity of gas flow in pores and incorrectly considers the flowing gas as an incompressible fluid. The assumption of gas flow dependency only on AP distorts the resulting
101
4 - - M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S STRUCTURE
dimensionless distributions [39]. A more realistic description of the situation, in which progressively smaller pores contribute to inert flow due to the applied pressure difference, is given in [123].
4.4.2.3 Liquid-liquid displacement porosimetry (or biliquid permporometry) [124-128] In this technique, which is a combination of bubble pressure and solvent permeability methods, a liquid A wetting the membrane is displaced by a fluid B (non miscible liquid less wetting than A). The principle is based upon the Laplace equation determining the mechanical equilibrium at the interface AP = 2ycos0/r
(4.27)
where AP is the pressure drop through the membrane, r is the pore radius, 0 is the contact angle between the two liquids and the solid and t~ is the interfacial tension between the two liquids. The liquid A will be displaced by B only when the pressure applied to B has reached the value
AP = 2y/r (Cantor's equation)
(4.28)
The technique consists in measuring the B (e.g. water) flow rate (J) through a membrane impregnated with A (e.g. isobutanol or mixtures of alcohols and water) as a function of the pressure difference AP. We have to note that it is possible to modify the method from "pressure controlled" to "flow controlled" in order to reduce the test time and increase its flexibility [126]. At a certain minimum pressure the largest pores become permeable, while the smaller pores still remain impermeable. This minimum pressure depends mainly on the type of membrane material (contact angle), type of permeate (surface tension) and pore size. When all pores are filled with B, the liquid flux J through the membrane becomes directly proportional to the pressure. A typical flux versus pressure curve is shown in Fig. 4.19. The recorded Pi and Ji values introduced in the Laplace equation directly give the pore equivalent pore radius ri and the distribution of permeability (Ji/Pi) vs. pore radius.
r(max)
r(mi.)
Fig. 4.19. Flux versus pressure for a membrane having a distribution of pore sizes.
102
4 - - METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
Pore number N i, a r e a A i ( A i = ~ N i r2) and their distributions are indirectly obtained by using the Hagen-Poiseuille relationship [126] Ji = [lr,Ni r4 Pi] /[8~Ax]
(4.29)
where ~t is the absolute viscosity of the testing liquid and Ax is the equivalent length of the capillary pores. If pores are assumed to be voids between packed spheres, the Kozeny-Carman equation has to be used to express the flux as a function of pressure. This technique can be applied to ultra and micro-filtration membranes and is very close to the situation that can take place in filtration as only the through pores are considered and the membranes are studied in wet conditions. 4.4.3 Fluid Flow Measurements
These techniques involving the measurement of membrane permeability to a fluid (liquid or gas) lead to a mean pore radius (usually the effective hydraulic radius rH) whose quantitative value is often highly ambiguous. The flux of a fluid through a porous material is sensitive to all structural aspects of the material [129]. Thus, in spite of the simplicity of the method, the interpretation of flux data, even for the simplest case of steady state, is subject to uncertainties and depends on the models and approximations used. 4.4.3.1 Liquid permeability
The steady state volume flux Jv of an incompressible fluid though a porous medium of cross-sectional area Ac, thickness I and porosity ~ under a pressure differential AP, can be expressed in terms of the mean velocity in the pores Up Jv = Ace Up = A c Kc e r~HAP/2rll
(4.30)
If ~ and Kc are known or can be estimated, rH and hence A the specific surface area of the porous medium can be determined. According to the well known Kozeny-Carman treatment K c = ~c/,i: 2
(4.31)
where [5cis a pore shape factor (varying between I for cylindrical pores and 2/3 for pores in the form of slits) and z > I is the tortuosity factor that takes account that the fluid follows tortuous paths through the porous solid. If the porous medium is pictured as a bundle of N cylindrical capillaries of radius re -- 2rH and length le = 1:l, and if we ignore the effect of tortuosity on the flow pattern, we obtain ~, = N ~ r e2 le /Acl = Nlr,~ T,/A c
and then
(4.32)
4 --METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
Jv = Nnr2e AP/811 ]e =
a c t~ r 2
[3c AP/2n~21
103
(4.33)
The first part of the relation (4.33) is the well known Poiseuille equation for cylindrical capillaries. The main problem is the lack of theoretical basis for assigning a value to "r a priori (an empirical value of 1.5 is suggested by Kozeny). This obviously precludes the use of these equations as an independent method for determining r H or A. An example of experimental determination of Kc for y alumina membranes is given in Ref. [130]. The corresponding tortuosity is very high in agreement with the observed microstructure of this type of membrane made of plate-shaped crystallites. 4.4.3.2 Gas permeability
The situation is rather more complex in the case of a gas (assumed to be non adsorbed by the membrane material) because compressibility and molecular effects, which predominate at low pressure, introduce a pressure dependence. Nevertheless the interpretation of results can yield more information than obtained with liquid media. The measurement of the permeability of non adsorbed gases is classically used to determine the range of pore size in membranes (macro, meso or micropores). Indeed by plotting the permeability as a function of gas pressure, a straight line is usually obtained whose slope gives an indication of the gas transport mechanism in the membrane. A quantitative description of pore structur e can be attempted from the results. One method which is known under the name of permeametry [131] or Poiseuille-Knudsen method [124] is based on the law of gas permeability in a porous media in the two flow regimes" molecular flow (Knudsen) and laminar or viscous flow (Poiseuille). According to Darcy's law, the gas flux through a membrane with a thickness I can be written as J = KAP/I, where K is the permeability coefficient and AP (AP = P1 - P2) the pressure difference across the membrane. If the membrane pore diameter is comparable to the mean free path of the permeating gas, K can be expressed as a sum of a viscous and a non-viscous term K = Ko + BoP/n
(4.34)
where P = (P1 + P 2 ) / 2 is the mean pressure, 11 is the gas viscosity and Ko and Bo are the characteristics of the membrane and the gas. Using the theoretical equations suggested by Carman and the average molecular velocity v of the gas with a molecular weight M v = ( 8 R T / ~ M ) 1/2
the mean hydraulic pore radius of the membrane can be expressed by
(4.35)
104
4 u METHODS FOR THE CH A RA CT E RI SA T I O N OF POROUS STRUCTURE
rH = (16/3)(Bo/Ko)(2RT/gM) 1/2
(4.36)
Ko can be obtained by extrapolation of K versus P at P = 0 and Bo is obtained by multiplying the slope of the same function by the gas viscosity rl. The experiment involves the measurement of the membrane permeability as a function of AP, in a pressure range which involves successively molecular flow and viscous flow. By plotting the membrane permeability coefficient as a function of the mean pressure, a straight line can be obtained whose slope and origin can be related to the mean pore hydraulic radius of the membrane. A special apparatus has to be used which is appropriate for the measured pore range, membrane size, thickness and geometry. This technique can be used to evaluate pores ranging from between several A to several gm. The modelling of gas permeation has been applied by several authors in the qualitative characterisation of porous structures of ceramic membranes [132138]. Concerning the difficult case of gas transport analysis in microporous membranes, we have to notice the extensive works of A.B. Shelekhin et al. on glass membranes [139,14] as well as those more recent of R.S.A. de Lange et al. on sol-gel derived molecular sieve membranes [137,138]. The influence of errors in measured variables on the reliability of membrane structural parameters have been discussed in [136]. The accuracy of experimental data and the mutual relation between the resistance to gas flow of the separation layer and of the support are the limitations for the application of the permeation method. The interpretation of flux data must be further considered in heterogeneous media due to the effects of pore size distribution and pore connectivity. This can be conveniently done in terms of structure factors [5]. Furthermore the adsorption of gas is often considered as negligible in simple kinetic theories. Application of flow methods should always be critically examined with this in mind. 4.4.3.3 Permporometry [140-143]
This technique, developed by Eyraud [140] modified by Katz et al. [143] and recently by Cuperus et al. [141], is based on the controlled blocking of pores by capillary condensation of a vapour (e.g. CC14, methanol, ethanol, cyclohexane), present as a component of a gas mixture, and the simultaneous measurement of the gas flux through the remaining open pores of the membrane. The capillary condensation process is related to the relative vapour pressure by the Kelvin equation. Thus for a cylindrical pore model and during desorption we have In P/Po = -27V cos0/rK RT
(4.37)
with r = r K + t (t has to be determined, t = 0.5 nm for cyclohexane [141]). Careful control of the relative vapour pressure permits the stepwise blocking of pores. Starting from a relative pressure equal to 1, all the pores of the
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
105
membrane are filled, hence unhindered gas transport through the membrane is not possible. When the vapour pressure is reduced, pores with a size larger than that corresponding to the vapour pressure are emptied and become available for gas transport. By measuring the gas transport through the membrane upon decreasing the relative vapour pressure, the size distribution of the active pores can be found, in the limit of validity of the Kelvin equation (mesopores). The experiments are usually conducted during the desorption process because the equilibrium time is much longer during adsorption [142]. The calculation of the size distribution of the active pores requires a well defined transport regime. In the absence of an overall mechanical pressure gradient and by using the principle of counter diffusion of two different gases (e.g. 02 and N2), a diffusion of Knudsen type can be usually assumed and the corresponding equation applied [141,142] 2
with
Jk = N~rp D k AP /RT xlA
D k = -~
rp ~/8RT /~M w
(4.38)
with Jk the Knudsen gas flux (mol/m2/s), N the number of pores, rp the pore radius (m), Dk the Knudsen diffusion coefficient, AP the pressure difference across the membrane (Pa), Mw the gas molecular weight (g/mol), A the membrane surface (m2), 1: the tortuosity, and l the layer thickness (m).
N2+ Ethanol
+ !
0~+
!
Ethanol
v
i'",
mbrane
I/lll|l|l 02
~
o| *
ImI 11i*
a
Oz
1
mlnmi
Oz
[~_~__Emptypore: Filled pores Membrane matrix
O
_
I I Ill-" Tranport in all pores
Fig. 4.20.Experimentalarrangementof permporometry(a) and principle of technique (b) [2].
106
4 - - M E T H O D S FOR THE C H A R A C T E R I S A T I O N OF P O R O U S STRUCTURE
A stepwise reduction of the relative pressure means that a range of pores will become accessible and contribute to the gas flux. Therefore number of pores as well as flow averages distributions can be obtained as a function of pore sizes. A schematic diagram of the experimental equipment employed [141] is given in Fig. 4.20 together with the principle of the analysis. During the experiments, there is no difference in hydrostatic pressure across the membrane and gas transport proceeds only by diffusion, the flow of one of the two non-condensable gases being measured (for example that of oxygen can be measured with an oxygen selective electrode). The technique has been applied in the characterisation of active pores in alumina membranes with a good reliability [142]. The method does not require a high mechanical stability of the membrane as most of the other techniques and can be applied in the limit of validity of the Kelvin equation, i.e. for pore sizes above 1.5 nm.
4.5 C O N C L U S I O N A N D R E C O M M E N D A T I O N S
The features of the more important characterisation techniques described here are summarised in Table 4.2. As is evident from this review, there is no technique which is universally applicable for the characterisation of the porous properties of all materials. The choice is made on the basis of many criteria, such as the range of pore size, the nature of the material and its form, together with the application envisaged. Frequently, more than one technique is required in a detailed examination. In the case of membranes, particular problems are encountered because of their form and the small quantity of active material involved. Furthermore, other complexities arise in the case of microporous thin films, although, as we have noted here, currently this is an area of active progress. This involves the development of new techniques and advances in phenomenological theories to describe the properties of such nanostructured supported materials. The techniques given in Table 4.2 are well established and have been sub-divided into those which are described as either static or dynamic. We feel this distinction is of particular importance in the characterisation of the porous structure of membranes. Here the performance is determined by the complex link between the structural texture and transport behaviour. An insight into this cOmplexity is frequently provided by dynamic techniques, which are not restricted by the limited quantity of membrane material and are sensitive to the active pathways through the porous structure. Further developments are required in this area both in the improvement of existing techniques and introduction of new techniques. Progress will also come from advances in the theory and modelling of flow behaviour in such porous media, which involve percolation theory and fractal geometry for example. With the refinement of such
4 -- METHODS
TABLE 4.2 Detailed summary of the main characterisation techniques for the determination of the texture of porous membranes Technique
Pore shape assump tion
Stereology SEM, E M Image analysis STEM, AFM
Remarks
>1.5 nm
2D image. Porosity Pore size distribution
Preparation technique can influence the real Porous texture
Macropores 2D image, Pore size ?, to atomic scale Surface rugosity ...?
Smooth samples needed. Interpretation of results is difficult
Laplace (Washburn)
Cylindrical
5 nm-15 Fm
Pore size distribution (including dead-end pores) Porosity
Outgassed (dry) samples. Measurement of pore entrance. Destructive method. For small pore sizes damage of the porous structure may occur. Network percolation effects derived.
Kelvin (B.E.T. B.J.H.)
Cylindrical or slits
2-50 nm
Pore size distribution (including dead-end pores). Pore shape information. Specific surface area, Porosity
Dry samples. Main problem: relationship between the pore geometry and a model which allows the pore sizes and pore size distribution to be determined from the isotherms. Network effect.
<2 nm
Microporous volume. Pore size distribution?
Ibid. Scraped support. Validity of quantitative results still ambiguous.
De Boer, Cylindrical/ Brunauer slits Dubinin, HorvathKawazoe, DFT
STRUCTURE
Gas adsorption desorption
Main characteristic parameters obtained
OF POROUS
Static methods Mercury intrusion
Pore size range
FOR THE CHARACTERISATION
Theoretical basis
107
(continued)
108
TABLE 4.2 (continuation) Technique
Theoretical basis
Pore shape assumption
Remarks
Thermoporometry
Laplace/ Cylindrical/ Gibbs-Duhem spherical
2-30 nm
Pore size distribution (including dead-end pores). Pore shape information. Porosity
Simple technique. Possible deformation of pore structure during the solidification process. Dry samples not necessary.
micro to macropores
"Cut-off" value
Very simple. Quantitative predictions of membrane performance cannot be obtained.
Cylindrical
0.44-20 pm
Largest active pores (bubble point). Mean pore sue. Pore size distribution (stepwise increase of pressure)
Dry samples. Different results with different liquids. Rate of pressure increase and pore length may influence the results.
Laplace (Cantor) HagenCylindrical Poiseuille
2 nm-5 pm
Pore size distribution (only active pores)
Wet samples. Combination of bubble pressure and solvent permeability methods.
Dynamic methods Rejection measurements
Liquid displacement Liquid/gas Laplace
Liquid/liquid
KozenyCaI7lMIl
Voids between spheres
OF POROUS STRUCTURE
Main characteristic parameters obtained
4 -- METHODS FOR THE CHARACTERISATION
Pore size range
Pore size range
Main characteristic parameters obtained
Remarks
Liquid permeabil- Hagenity Poiseuille KozenyCarman
Cylindrical
0.1-10 p
Pore hydraulic radius
Experimental simplicity. Assumptions: laminar flow in HP equation, zero wetting angle, no pre-existing agent on the surface Great influence of pore geometry and tortuosity on the interpretation of results. Network effects.
Gas permeability Poiseuille/ Knudsen
Cylindrical
A few 8, to several pm
Pore hydraulic radius
Dry samples. Specific apparatus for each pressure range. Tedious technique
Permporometry
Cylindrical
2-20 nm
Pore size distribution (only active pores)
Wet samples. Experimental difficulties (the same vapour pressure has to be maintained on both sides of the membrane so that some time is necessary to attain thermodynamic equilibrium). No mechanical pressure applied.
Theoretical basis
Kelvin Knudsen
Voids between spheres
4 m M E T H O D S FOR THE C H A R A C T E R I S A T I O N OF P O R O U S S T R U C T U R E
Pore shape assump tion
Technique
109
110
4 - - M E T H O D S F O R T H E C H A R A C T E R I S A T I O N OF P O R O U S S T R U C T U R E
theories they m a y be used directly in the interpretation of experimental data. Of the established static techniques, which we have considered here, that involving gas adsorption isotherm measurements remains one of the most powerful and widely applicable. It is indeed very accessible with the availability of automated commercial equipment and the variety of data treatment facilities available. Nevertheless, it is still circumscribed by the assumptions implicit in the choice of a pore shape model in the case of mesoporous materials. Its application to microporous structures has recently advanced considerably, although there are here certain reservations which still exist concerning the general application of theories to describe adsorption in such small pores in ill defined structures. A variety of other static characterisation methods have been described in this chapter which are not listed in Table 4.2. M a n y of these are new and in a state of rapid evolution, as for example those involving N M R and radiation scattering. Whilst appropriate for research investigations they do not seem yet to be appropriate as a means of general characterisation. H o w e v e r with the rapid progress under w a y in these areas, some of these techniques we feel m a y in the future be ideally suited to m e m b r a n e characterisation. REFERENCES
1.
2.
3.
4. 5.
6.
7.
8.
(a) F.P. Cuperus and C.A. Smolders, Characterization of UF membranes, Membrane characteristics and characterization techniques. Adv. Colloid Interface Sci., 34 (1991) 135. (b) F.P. Cuperus, Characterization of ultrafiltration membranes. PhD report, Twente University, The Netherlands, 14 Sept., 1990. M. Mulder, Characterization of membranes, in: M. Mulder (Ed.), Basic Principles of Membrane Technology. Kluwer Academic Publishers, The Netherlands, 1991, pp. 110133. H.P. Hsieh, General characteristics of inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, NY, 1991, pp. 64-94. S.I. Nakao, Determination of pore size and pore size distribution 3. Filtration membranes, ]. Membr. Sci. 96 (1994) 131. J. Rouquerol, D. Avnir, C.W. Fairbridge, D.H. Everett, J.H. Haynes, N. Pernicone, J.D.F. Ramsay, K.S.W. Sing and K.K. Unger, Recommendations for the characterisation of porous solids, Pure Appl. Chem., 66(8) (1994) 1739. K.K.Unger, J. Rouquerol, K.S.W. Sing and H. Kral (Eds.), Characterization of Porous Solids I, Studies in Surface Science and Catalysis Vol. 39, Proc. of the IUPAC Symposium (COPS I), Bad Soden, Germany, April 1988, Elsevier, Amsterdam. F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991. J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994.
4 -- METHODSFORTHE CHARACTERISATIONOF POROUSSTRUCTURE
111
9. K.S.W. Sing, D.H. Everett, R.A.W. Haul, L. Moscou, R.A. Pierotti, J. Rouquerol and T. Siemieniewska, Reporting physisorption data for gas/solid systems with special reference to the determination of surface area and porosity. Pure Appl. Chem., 57 (1985) 603. 10. K. Kaneko, Determination of pore size and pore size distribution 1. Adsorbents and catalysts, J. Membr. Sci., 96 (1994) 59. 11. V.V. Terskikh, I.L. Mudrakovskii and V.M. Mastikhin, 129XeNMR Magnetic resonance studies of the porous structure of silica gels, J. Chem. Soc. Faraday Trans., 89 (23) (1993) 4239. 12. F.A. Dullien, Porous Media, Fluid Transport and Pore Structure. Academic Press, London, 1979. 13. P.M. Adler, Porous Media. Series in Chemical Engineering, Butterworth-Heineman, Boston, MA, 1992. 14. A.B. Shelekin, A.G. Dixon and Y.H. Ma, Adsorption, diffusion and permeation of gases in microporous membranes III. Application of percolation theory to interpretation of porosity, tortuosity and surface area in microporous glass membranes, J. Membr. Sci., 83 (1993) 181. 15. C.N. Satterfield and T.K. Sherwood, The Role of Diffusion in Catalysis, Addison-Wesley Publishing Company, Inc., Reading, MA, 1963. 16. N.A. Seaton, Determination of the connectivity of porous solids from nitrogen sorption measurements, Chem. Eng. Sci., 46 (8) (1991) 1895. 17. H. Liu, L. Zhang and N.A. Seaton, Characterisation of mesoporous solids using sorption hysteresis measurements, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, pp. 129-140. 18. D. Stauffer, Introduction to Percolation Theory. Taylor & Francis, London, 1985. 19. A.V. Neimark, New theoretical approaches to characterisation of mesoporous solids, in: C.A.C. Sequeira and M.J. Hudson (Eds.), Multifunctional Mesoporous Inorganic Solids, NATO Asi Ser., Ser. C, Vol. 400, Kluwer, The Netherlands, 1993, pp. 27-35. 20. J.H.F.Lim, X. Jia, R. Jafferali and G.A. Davies, Statistical models to describe the structure of porous ceramic membranes. Sep. Sci. Technol., 28 (1-3) (1993) 821. 21. A. Tsetsekou and G. Androutsopoulos, Mercury porosimetry hysteresis and entrapment predictions based on a corrugated random pore model, Chem. Eng. Comm., 110 (1991) 1. 22. T.T. Tsotsis, M. Sahimi, and G. Gavalas, Membrane and Membrane Processes, in: Statistical Modeling of Fluid Solid Reaction Systems. Oxford Press, 1997, in press. 23. J.S. Pennycook, Powerful new analytical technique directly images a material's atomic structure and composition, Materials & Processing report 6 (1) (1991) 1. 24. A. Humbert, La microscopie par effet Tunnel, in: Michel Ammou (Ed.), Microcaractdrisation des solides, Diff. CRAM-LPSES-CNRS, Sophia-Antipolis-Valbonne (France), 1989, pp. 253-289. 25. A. Bessiere, M. Meireles, P. Aimar, V. Sanchez, R. Coratger and J. Beauvillain, Surface analysis of ultrafiltration membranes by atomic force microscopy, in: P. Aimar and P. Aptel (Eds.), Proc. Euromembrane 92 ~ Paris, R6cents Progr6s en G6nie des Proc6d6s Vol. 6 (22), 1992, pp. 111-116. 26. A.K. Fitzche, A.R. Arevalo, M.D. Moore and C. O'Hara, The surface structure and morphology of polyacrylonitrile membranes by atomic force microscopy. J. Membr. Sci., 81 (1993) 109.
112
4 m METHODSFORTHE CHARACTERISATIONOF POROUSSTRUCTURE
27. A. Bottino, G. Capannelli, A. Grosso, O. Monticelli, O. Cavalleri, R. Rolandi and R. Soria, Surface characterization of ceramic membranes by atomic force microscopy. J. Membr. Sci., 95 (1994) 289. 28. R.J.M. Da Fonseca, J.M. Saurel, A. Foucaran, E. Massone, T. Taliercio and J. Camassel, Acoustic microscopy investigation of porous silicon. Thin Solid Films,255 (1995) 155. 29. Microscopies, J. CNRS (Meudon-France), 65 (1995) 11. 30. E.R. Scott and H.S. White, Scanning electrochemical microscopy of a porous membrane, J. Membr. Sci., 58 (1991) 71. 31. K.J. Kim, P.V. Stevens and A.G. Fane, Porosity dependence of pore entry shape in track-etched membranes by image analysis. J. Membrane Sci., 93 (1994) 79. 32. L. Zeman and L. Denault, Characterization of microfiltration membranes by image analysis of electron micrographs Part I. Method develoment. J. Membr. Sci., 71 (1992) 221. 33. L. Zeman, Characterization of microfiltration membranes by image analysis of electron micrographs Part II. Functional and morphological parameters. J. Membr. Sci., 71 (1992) 233. 34. H. Vivier, M.N. Pons and J.F. Portala, Study of microporous membrane structure by image analysis. J. Membr. Sci., 46 (1989) 81. 35. J.P. Valley and C.A. Brown, Fractal analysis of SEM images of silica membrane morphology, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Inorganic Membranes ICIM2-91, Key Engineering Materials 61 & 62, Trans Tech Pub. Zurich, 1991, pp. 83-88. 36. B. McEnaney and T.J. Mays, Characterization of macropores using quantitative microscopy, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994. pp. 327-338. 37. B. Ewing, P.J. Davis, P.D. Majors, G.Drobny, M. Smith and W.L. Earl, Determination of spatially resolved pore size information, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 709-715. 38. G. Guillot, G. Kassab, J.P. Hulin and P. Rigord, Monitoring of tracer dispersion in porous media by NMR imaging, J. Phys. D: Appl. Phys., 24 (1991) 763. 39. J. Rocek and P. Uchytil, Evaluation of selected methods for the characterisation of ceramic membranes. J. Membr. Sci., 89 (1994) 119. 40. S. Lowell and J.E. Shields, Powder Surface Area and Porosity, 2nd edn. Powder Technology Series, Chapman & Hall, London, 1984. 41. S.J. Gregg and K.S.W. Sing, Adsorption, Surface Area and Porosity. Academic Press, London, 1982. 42. G.C. Frye, A.J. Ricco, S.J Martin and C.J. Brinker, Characterization of the surface area and porosity of sol-gel films using SAW devices. Mater. Res. Soc Symp. Proc., 121 (1988) 349. 43. S.L. Hietala, D.M. Smith, V.M. Hietala, G.C. Frye and S.J. Martin, Pore structure characterization of thin films using a surface acoustic wave/volumetric adsorption technique. Langmuir, 9 (1993) 249. 44. R.G. Avery and J.D.F. Ramsay, The sorption of nitrogen in porous compacts of silica and zirconia powders. J. Colloid Interface Sci., 42 (1973) 597. 45. C. Lastoskie, K.E. Gubbins and N. Quirke, Pore size distribution analysis and networking: studies of microporous sorbents, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science
4 -- METHODSFORTHECHARACTERISATIONOFPOROUSSTRUCTURE
46.
47. 48. 49. 50. 51.
52. 53. 54. 55.
56.
57.
58.
59.
113
and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 51-60. K. Kaneko, C. Ishii and T. Rybolt, Superhigh surface area determination of microporous carbons, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 583-592. J.D.F. Ramsay, Nitrogen adsorption in microporous silica compacts. J. Collloid Interface Sci., 51 (1975) 205. K.S.W. Sing, Utilization of adsorption data in the BET region, in: D.H. Everett and R.H. Ottewill (Eds.), Surface Area Determination. Butterworths, London, 1970, pp. 25-42. S.J. Gregg and J.F. Langford, Evaluation of microporosity with special reference to a carbon black. Trans. Faraday Soc., 65 (1969) 1394. M.M. Dubinin, The potential theory of adsortion of gases and vapors for adsorbents with energitically nonuniform surfaces. Chem. Rev., 60 (1960) 235. M.M. Dubinin, Physical adsorption of gases and vapors in micropores, in: D.A. Cahenhead (Ed.) Progress in Surface and Membrane Sience Vol. 9. Academic Press, New York, NY, 1975, pp. 1-70. R.S. Mikhail, S. Brunauer and E.E. Bodor, Investigations of a complete pore structure analysis. I. Analysis of micropores. J. Colloid Interface Sci., 26 (1) (1968) 45. G. Horvath and K. Kawazoe, Method for calculation of effective pore size distribution in molecular sieve carbon. J. Chem. Eng. Jpn., 16 (1983) 470. A. Saito and H.C. Foley, Curvature and parametric sensitivity in models for adsorption in micropores. AIChE J., 37 (3) (1991) 429. J.P. Olivier, W.B. Conklin and M.V. Szombathely, Determination of pore size distribution from density functional theory: a comparison of nitrogen and argon results, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier~Amsterdam, 1994, pp. 81-90. M.J. Bojan and W. Steele, Simulation of sorption in pores with constrictions, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 11-20. M. Molina-Sabio, F. Rodriguez-Reinoso, D. Valladares and G. Zgrablich, A lattice-gas model for adsorption in microporous solids, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 573-582. D.D. Do and H.D. Do, Characterisation of microporous carbon using sorption data of multiple sorbates at vaious temperature, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseilte, France, May 1993, Elsevier, Amsterdam, 1994, pp. 641-650. D.H.T. Spencer, The use of molecular probes in the characterization of carbonaceous materials, in: R.L. Bond (Ed.), Porous Carbon Solids. Academic Press, New York, NY, 1967, pp. 87-154.
114
4 m METHODS FOR THE CHARACTERISATION OF POROUS STRUCTURE
60. E. Grosgogeat, J.R. Fried, R.G. Jenkins and S.T. Hwang, A method for the determination of the pore size distribution of molecular sieve materials and its application to the characterization of partially pyrolyzed polysilastyrene/porous glass composite membranes. ]. Membr. Sci., 57 (1991) 237. 61. J.D.F. Ramsay and R.G. Avery, Adsorption in silica compacts containing pores of molecular size, in: Pore Structure and Properties of Materials, Part I, Proc. Int. Symp. IUPAC, Academia Prague, 1973, pp. B37-B45. 62. C.T. Kresge, M.E. Leonowicz, W.J. Roth, J.C. Vartuli and J.S. Beck, Ordered mesoporous molecular sieve synthesized by a liquid-crystal template mechanism. Nature, 359 (1992) 710. 63. T. Dabadie, A. Ayral, C. Guizard, L. Cot, J.C. Robert and O. Poncelet, Liquid crystal templating effects on silica gels synthesized using quaternary ammonium surfactants. Mater. Res. Soc. Syrup. Proc., 346 (1994) 849. 64. H.F. Stoekli and F. Kraehenbuel, The external surface of microporous carbons, derived from adsorption and immersion studies. Carbon, 22 (1984) 297. 65. M. Brun, A. Lallemand, J.F. Quinson and C. Eyraud, A new method for the simultaneous determination of the size and shape of pores: the thermoporometry. Thermochim. Acta, 21 (1977) 59. 66. M. Brun, A. Lallemand, G. Lorette, J.F. Quinson, M. Richard, L. Eyraud and C. Eyraud, Changement d'6tat liquide-solide dans les milieux poreux I. J. Chimie Phys., 6 (1973) 973 67. M. Brun, A. Lallemand, J.F. Quinson and C. Eyraud, Changement d'6tat liquide-solide dans les milieux poreux II. J. Chimie Phys. (1973) 979. 68. J.F. Quinson, M. Brun, Progress in thermoporometry, in: K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral (Eds.), Characterization of Porous Solids I, Studies in Surface Science and Catalysis Vol. 39, Proc. of the IUPAC Symposium (COPS I), Bad Soden, Germany, April 1988, Elsevier, Amsterdam, pp. 307-315. 69. C. Jallut, J. Lenoir, C. Bardot and C. Eyraud, Thermoporometry. Modelling and simulation of a mesoporous solid. J. Membr. Sci., 68 (1992) 271. 70. L. Zeman, G. Tkacik and P. le Parlouer, Characterisation of porous sublayers in UF membranes by thermoporometry. J. Membr. Sci., 32 (1987) 329. 71. T. Kobayashi, M. Todoki, M. Fujii, T. Takeyama and H. Tanzawa, Permeability and structure of PMMA stereocomplex hollow fiber membrane for hemodialysis, in: E. Drioli and M. Nakagaki (Eds.), Proc. Eur.-Jpn Cong. Membr. Membr. Processes, Plenum, New York, NY, 1986, pp. 507-513. 72. J.F. Quinson, N. Mameri, L. Guihard and B. Bariou, The study of the swelling of an ultrafiltration membrane under the influence of solvents by thermoporometry and measurement of permeability. J. Membr. Sci., 58 (1991) 191. 73. C. Eyraud, J.F. Quinson and M. Brun, The role of thermoporometry in the study of porous solids, in: K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral (Eds.), Characterization of Porous Solids I, Studies in Surface Science and Catalysis Vol. 39, Proc. of the IUPAC Symposium (COPS I), Bad Soden, Germany, April 1988, Elsevier, Amsterdam, pp. 295--305. 74. Cuperus, D. Bargeman and C.A. Smolders, Critical points in the analysis of membrane pore structures by thermoporometry. ]. Membr. Sci., 66 (1992) 45. 75. W.L. Earl, Y.W. Kim and D.M. Smith, NMR measurement of pore structure, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp.
4 --METHODSFORTHE CHARACTERISATIONOF POROUSSTRUCTURE
115
301-309. 76. C.L. Glaves, P.J. Davis, K.A. Moore, D.M. Smith and H.P. Hsieh, Pore structure characterization of composite membranes. J. Colloid Interface Sci., 133 (2) (1989) 377. 77. C.L. Glaves and D.M. Smith, Membrane pore structure analysis via NMR-spin-lattice relaxation experiments. J. Membr. Sci., 46 (1989)167. 78. D.P. Gallegos, D. M. Smith and C.J. Brinker, A NMR technique for the analysis of pore structure: application to mesopores and micropores. J. ColloidInterface Sci., 124 (1) (1988) 186. 79. D.M. Smith and P.J. Davis, Pore analysis of wet materials via low-field NMR, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 301-310. 80. D.M. Smith, R. Deshpande, C.J. Brinker, W.L. Earl, B. Ewing and P.J. Davis, In-situ pore structure characterisation during sol-gel synthesis of controlled porosity materials. Proc. Symp. on Catalyst Supports: Chemistry, Forming and Characterization. American Chemical Soc., New York City meeting, August 25-30, 1991, pp. 489-495. 81. J.A. Ripmeester and C.I. Ratcliffe, The anisotropic chemical shift of 129Xe in the molecular sieve ALPO-11: a dynamic averaging model. J. Phys. Chem., 99 (1995) 619. 82. J. Fraissard, T. Ito and L.C. De M6norval, Nuclear Magnetic Resonance study of Xenon adsorbed on zeolites and metal-zeolites, Application to catalyst research, in: Proceedings 8th International Conference on Catalysis, Berlin, 1984, pp. 25-35. 83. W.C. Corner, E.L. Weist, T. Ito and J. Fraissard, Characterization of the porous structure of agglomerated microspheres by 129Xe NMR spectroscopy. J. Phys. Chem., 93 (1989) 4138. 84. J.A. Ripmeester and D.W. Davidson, Xe-129 NMR in the clathrate hydrate of xenon. J. Mol. Struct., 75 (1981) 67. ~ 85. L.C. de M6norval, A. Julbe, H. Jobic, J.A. Dalmon and C. Guizard, Potentialities of an innovative technique like 129XeNMR and of SAXS for the characterization of microporous sol-gel derived SiO2. MRS Symp. Proc. Ser., Vol. 431 (Microporous anc~ Macroporous Materials, 1996. In press. 86. A. Julbe, C. Balzer, J.M. Barthez, C. Guizard, A. Larbot and L. Cot, Effect of non-ionic surface active agents on TEOS derived sols, gels and materials. J. Sol-Gel Sci. Technot., 4 (1995) 89. 87. J. Demarquay and J. Fraissard, Xe-129 NMR of xenon adsorbed on zeolites. Relationship between the chemical shift and the void space. J. Chem. Phys. Lett., 136 (1987) 314. 88. A. Guinier and G. Fournet, Small Angle Scattering of X-rays. Wiley, New York, 1955. 89. B.O. Booth and J.D.F. Ramsay, Characterisation of porous oxides by SANS and N2 adsorption isotherms measurements, in: M.J. Haynes and P.R. Doria (Edsi), Proc. of
RILEM/CNR International Symp. on Principles and Applications of Pore Structural Characterisation. J.W. Arrowsmith, Bristol, 1985, pp. 97-116. 90. J.D.F. Ramsay, Applications of neutron scattering in investigations of adsorption processes in porous materials. Pure Appl. Chem., 65(1993) 2164. ~ 91. J.D.F. Ramsay, Recent developments in the characterization of oxide sols using small angle neutron scattering techniques. Chem. Soc. Revl, 15 (1986) 335. 92. G. Porod, General theory, in: O. Glatter and O. Kratky (Eds,)~ ,~Small Angle X-ray Scattering. Academic Press, London, 1982, Chap. 2, pp. 17-51. 93. D.W. Hua, J.V. D'Souza, P.W. Schmidt and D.M: Smith, Poi'e structure analysis Via small angle X-ray scattering, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and
116
94.
95.
96.
97.
98.
99. 100.
101.
102.
103.
104.
4 -- METHODSFORTHECHARACTERISATIONOFPOROUSSTRUCTURE K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 255-262. J.D.F. Ramsay, Neutron scattering investigations of adsorption in microporous adsorbents having controlled pore geometry, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 235-245. E. Hoinkis and J.A. Allen, A small angle neutron scattering study of porous graphitic materials before and after adsorption and condensation of C6D6 within the accessible pores. J. Colloid Interface Sci., 142 (1991) 540. J.D.F. Ramsay and R.G. Avery, Neutron scattering investigation of adsorption processes in model porous systems, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 235-244. J.D.F. Ramsay, P.J. Russell and S.W. Swanton, Gel precipitated oxide gels with controlled porosity-Determination of structure by small angle neutron scattering and adsorption isotherms measurements, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 257-265. J.C. Dore and A.N. North, Small angle and ultrasmall angle scattering techniques for characterization of porous materials, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 245-255. J.D.F. Ramsay and G. Wing, Small angle neutron scattering investigations of water sorption in porous silica and ceria gels. ]. Colloid Interface Sci., 141 (1991) 475. M.H. Stacey, Evolution of porosity during conversion of rl-alumina to a novel porous (~-alumina fibre, in: F. Rodriguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids II, Studies in Surface Science and Catalysis Vol. 62, Proc. of the IUPAC Symposium (COPS II), Alicante, Spain, May 1990, Elsevier, Amsterdam, 1991, pp. 615-624. A. Matsumoto, K. Kaneko and J.D.F. Ramsay, Neutron scattering investigations of the structure and adsorption properties of activated carbon fibers, in: M. Suzuki (Ed.), Fundamentals of Adsorption, Studies in Surface Science and Catalysis Vol. 80, Elsevier, Amsterdam, 1993, pp. 405-412. J.S. Rigden, J.C. Dore and A.N. North, Determination of anisotropic features in porous materials by small angle X-ray Scattering, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III, Studies in Surface Science and Catalysis Vol. 87, Proc. of the IUPAC Symposium (COPS III), Marseille, France, May 1993, Elsevier, Amsterdam, 1994, pp. 263-272. L. Auvray, A. Ayral, T. Dabadie, L. Cot, C. Guizard and J.D.F. Ramsay, Sol-gel systems with controlled structure formed in surfactant media, Faraday Discuss. Chem. Soc.,, 101 (1995) 235. R.M.A. Azzam and N.M. Bashara, Ellipsometry and Polarized Light. North Holland, Amsterdam, 1977.
4 n METHODSFORTHECHARACTERISATIONOF POROUSSTRUCTURE
117
105. C.F. Bohren and D.R. Huffman, Absorption and Scattering of Light by Small Particles, Wiley-Interscience, New York, NY, 1983, Chap. 8. 106. E. Daugy, B. Pointu and G. Villela, D6p6t par pulv6risation et assistance ionique d'oxydes et d'argent pour r6alisation de miroirs m6talliques prot6g6s. Le Vide, les Couches Minces, 245 (1989) 59. 107. R. Sehgal, J.C. Huling and C.J. Brinker, Use of silica sols in inorganic sieving membranes, in: Y.H. Ma (Ed), Proc. 3rd International Conference on Inorganic Membranes, Worcester, MA, USA, July 1994, pp. 85-94. 108. J.H. Gladstone and T.P. Dale, Trans. Roy. Soc. (Lond.), 153 (1964) 317. 109. W.A. Plinski, Comparison of properties of dielectric films deposited by various methods. J. Vac. Sci. Technol. 14 (5) (1977) 1064. 110. K.E. Amin, Radiograph and ultrasound detect defects early. Am. Ceram. Soc. Bull., 74 (1) (1995) 7. 111. D.C. Kunerth, K.L. Telschow and J.B. Walter, Characterization of porosity distributions in advanced ceramics: a comparison of ultrasonic methods. Materials Eval., 47 (1989) 571. 112. W. Huang and S.I. Rokhlin, Low-frequency normal incidence ultrasonic method for thin layer characterization. Materials Eval., 11 (1993) 1279. 113. R.J.M. Da Fonseca, J.M. Saurel, A. Foucaran, E. Massone, T. Taliercio and J. Camassel, Acoustic microscopy investigation of porous silicon. Thin Solid Films, 255 (1995) 155. 114. B.H. Armitage, F.P. Bradey and J.D.F. Ramsay, The measurement of pore size in porous and microporous materials using resonant ion beam backscattering. Nucl. Instrum. Meth., 149 (1978) 329. 115. J.L. Keddie and E.P. Giannelis, Ion-beam analysis of sol-gel films: structural and compositional evolution. J. Am. Ceram. Soc., 73 (10) (1990) 3106. 116. H.E. Shaefer, R. Wurschum, R. Birringer and H. Gleiter, Structure of nanometer-sized polycrystalline iron investigated by positron lifetime spectroscopy. Phys. Rev. B, 38 (14) (1988) 9545. 117. J.Y. Dolveck, G.H. Dai, P. Moser, M. Pineri and M. Escoubes, Free volume measurements in polymides by positronium anihilation. Materials Sci. Forum, 105-110 (1992) 1549. 118. T.W. Zerda, G. Hoang, B. Miller, C.A. Quarles and G. Orcei, P0sitronium decay study of pore sizes in silica sol-gels. Mater. Res. Symp. Proc., 121 (1988) 653. 119. M. Meireles, A. Beissieres, I. Rogissart, P. Aimar and V. Sanchez, An appropriate molecular size parameter for porous membrane calibration. J. Membr. Sci., 103 (1/2) (1995) 105. 120. ASTM F 316-86, Standard test method for pore size characterisation of membrane filters by bubble point and mean flow pore test, Annual Book of ASTM Standards, Vol. 10.05, 1990 (originally published as F316-70, last previous edition F316-80). 121. ASTM E 1294-89, Standard test method for pore size characteristics of membrane filters using automated liquid porosimeter. 122. ASTM F316 Test method for pore size characteristics of membrane filter for use in aerospace fluids, Annual Book of ASTM Standards, Vol. 11.01. 123. P. Sneider and P. Uchytil, Liquid expulsion permporometry for characterization of porous membranes. I. Membr. Sci., 95 (1994) 29. 124. A. Bottino, G. Capannelli, P. Petit-Bon, N. Cao, M. Pegoraro and G. Zoia, Pore size and pore size distribution in microfiltration membranes. Sep. Sci. Technot., 26 (10/11) (1991) 1315. 125. M.G. Liu, R. Ben Aim and M. Mietton Peuchot, Characterization of inorganic membranes by permporometry method: importance of non equilibrium phenomena, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Inorganic Membranes. Key Engineering Materials 61 & 62, Trans Tech Publications, Zurich, 1991, pp. 603--605.
118
4 - - M E T H O D S FOR THE C H A R A C T E R I S A T I O N OF P O R O U S STRUCTURE
126. G. Capannelli, I. Becchi, A. Bottino, P. Moretti and S. Munari, Computer driven porosimeter for ultrafiltration membranes, in: K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral (Eds.), Characterization of Porous Solids I, Studies in Surface Science and Catalysis Vol. 39, Proc. of the IUPAC Symposium (COPS I), Bad Soden, Germany, April 1988, Elsevier, Amsterdam, pp. 283-294. 127. S. Munari, A. Bottino, P. Moretti, G. Capannelli and I. Becchi, Permporometric study on ultrafiltration membranes. J. Membr. Sci., 41(1989) 69. 128. S. Munari, A. Bottino, G. Capannelli and P. Moretti, Membrane morphology and transport properties. Desalination 53 (1985) 11. 129. J.H. Petropoulos and J.K. Petrou, Structural characterization of porous solids. Sep. Technol., 2 (10) (1992) 162. 130. A.F.M. Leenaars and A.J. Burggraaf, The preparation and characterization of alumina membranes with ultra-fine pores: Part 3. The permeability for pure liquids. J. Membr. Sci., 24 (1985) 245. 131. J. Charpin and B. Rasneur, Caracterisation de la texture poreuse des mat6riaux, in: Techniques de l'Ing6nieur (Ed.), Caractdrisation. Paris, P-1050-1, 1987. 132. R.W. Schofield, A.G. Fane and C.J.D. Fell, Gas and vapour transport through microporous membranes. I. Knudsen-Poiseuille transition. J. Membr. Sci., 53 (1990) 159. 133. F.W. Altena, H.A.M. Knoef, H. Heskamp, D. Bargeman and C.A. Smolders, Some comments on the applicability of gas permeation methods to characterise porous membranes based on improved experimental accuracy and data handling. J. Membr. Sci., 12 (1983) 313. 134. P. Uchytil, Gas permeation in ceramic membranes. Part I. Theory and testing of ceramic membranes. J. Membr. Sci., 97 (1994) 139. 135. P. Uchytil and Z. Broz, Gas permeation in ceramic membranes. Part II. Modeling of gas permeation through ceramic membranes with one separation layer. J. Membr. Sci., 97 (1994) 145. 136. P. Uchytil, Z. Wagner, J. Rocek and Z. Broz, Possibility of pore size determination in separation layer of ceramic membrane using permeation method. J. Membr. Sci., 103 (1995) 151. 137. R.S.A. de Lange, K. Keiser and A.J. Burggraaf, Analysis and theory of gas transport in microporous sol-gel derived ceramic membranes. J. Membr. Sci., 104 (1995) 81. 138. R.S.A. de Lange, J.H.A. Hekk-ink, K. Keiser and A.J. Burggraaf, Permeation and separation studies on microporous sol-gel modified ceramic membranes. Microporous Materials, 4 (1995) 169. 139. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. II.Permeation of gases in microporous glass membranes. ]. Membr. Sci., 75 (1992) 232. 140. C. Eyraud, Application of gas-liquid permporometry to characterization of inorganic ultrafilters, in: E. Dridi and M. Nakagaki (Eds.), Proc. Eur.-JPN Cong. Memb. Processes, 1984, pp. 629-634. 141. F.P. Cuperus, D. Bargeman and C.A. Smolders, Permporometry. The determination of the size distribution of active pores in UF membranes. J. Membr. Sci., 71 (1992) 57. 142. G.Z. Cao, J. Meijerink, H.W. Brinkman and A.J. Burggraaf, Permporomerty study on the size distribution of active pores in porous ceramic membranes. J. Membr. Sci., 83(1993) 221. 143. M. Katz and G. Barush, New insights into the structure of microporous membranes obtained using a new pore size evaluation method. Desalination, 58 (1986) 199.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 5
Ceramic processing techniques of support systems
for membranes synthesis A. Larbot
Laboratoire des Mat6riaux et Proc6d4s Membranaires, Ecole Nationale Sup~rieure de Chimie, 8 rue de I'Ecole Normale, 34053 Montpellier, France
INTRODUCTION
5.1
An inorganic membrane can be described as an asymmetric porous ceramic formed by a macroporous support with successive thin layers deposited on it. The support provides mechanical resistance to the medium. The successive layers are active in microfiltration (MF), ultrafiltration (UF) or nanofiltration (NF), depending on their pore diameters. Typically, the support thickness is about 1-2 mm, the MF layer is 10-30 ~tm thick, UF membranes are a few microns thick and NF membranes are less than 1 ~tm thick. The support and the MF layer are elaborated by classical ceramic techniques; the top layers (UF or NF) are formed using the sol-gel process (Fig. 5.1). A ceramic support is formed by shaping a powder and then consolidation of the green body by sintering. The fabrication process consists of four main stages: the choice of inorganic material, paste preparation, shaping, and firing (Fig. 5.2). There are two main methods of elaborating ceramic supports: extrusion; tape casting (or doctor blade). The choice of the shaping process must be adapted to the geometry required for the final product: -
-
120
5 m CERAMICPROCESSINGTECHNIQUESOF SUPPORTSYSTEMSFOR MEMBRANESSYNTHESIS Ultrafine layer Intermediate
layer
Support
Fig. 5.1. Classical inorganic filtration medium.
RAW MATERIALS I I I
[CONDITIONING]
SLURRY
PLASTIC PASTE
SLIP CASTING TAPE CASTING
EXTRUSION
I I I I
Fig. 5.2. Classical techniques for preparing ceramic supports.
- extrusion for tubular configurations (mono- or multichannels); - tape casting for flat supports. The elaboration of the green bodies is a critical step because it affects the properties and characteristics of the final material. The preparation of green pieces requires the use of temporary additives, usually organic binders, which are eliminated during the thermal treatment. The choice of these organics depends on the ease of shaping and on their decomposition. The organic part to inorganic part ratio is an important parameter which varies with the process. In practice, optimal formulation is characteristic of each material related to the shaping method and can only be determined after several experiments.
5 m CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
5 . 2
E
X
T
R
U
S
I
O
121
N
Extrusion can be defined as paste flow and extrudate formation from ceramic paste [1]. A membrane support must provide a high mechanical resistance and permeability. Tubular configurations correspond to this criterion and are adapted to the tangential filtration. Shaping is possible only if the paste, formed from the powder, has rheological properties close to those of clay, so it is necessary to add organic compounds to the inorganic powder. The most important parameters are the size of the ceramic grains and the nature and proportion of organic additives. Each powder reacts differently and the preparation of a paste with good rheological characteristics is very empirical. The various steps involved in the preparation of ceramic supports are as follows: choice of inorganic powder (nature and grain size), - choice of organic compounds, mixing, pugging to obtain a paste, ageing of the paste, shaping by extrusion, - drying and sintering. -
-
-
-
-
5.2.1 Ceramic Paste Preparation
Choice of inorganic powder The particle morphology is an important parameter because it affects the support porosity (or the density) and the pore size of the ceramic. The density increases when the grain size decreases and the pore size varies in the same way as the grain size. The ratio of grain size to pore size is equal to about 2.5, but this ratio is strongly dependent on the shape of particles. In order to facilitate the coating of a homogeneous thin layer on the support, the pore size must be adapted to the grain size of the layer which is to be deposited. The presence of large pores on the internal surface of channels could lead to penetration of the grains into the support and thus to defects in the membrane. The density of the support must be sufficient to ensure an excellent mechanical resistance. However, a low density shows a resistance to the flux through the support, so a compromise must be found for the choice of grain size. In terms of extrusion, the particle size can vary from 0.1 to 100 ~tm. In general, the ratio of ceramic wall dimension to particle size must be higher than 10 to ensure good mechanical resistance. A mixture of grain sizes can be used to facilitate the sintering.
122
5 m CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
Choice of organic compounds The plastic properties of clay are mainly influenced by its capability to retain water and by the possibility of water migration into the macroscopic structure. A ceramic powder is therefore usually mixed with the solvent (water) and organic additives [2-5], giving the paste the required plastic properties for shaping without losing its cohesion. Another effect of organic additives is to increase the unfired material strength, during shaping and drying, to avoid the formation of cracks. To be suitable for ceramic processes, organic compounds must be burnt out without leaving ash and tar. The different organics can be classified as a function of their properties. Binders bind the particles together by organic bonds and allow good strength after shaping and during drying. They give density to the material. They are polymers, usually linear and soluble in water. However, these bonds are the weaker parts of the unfired material and cracks appear on them. The binder distribution throughout the body is therefore very important [3] and the best location for the binders is at the neck regions between the particles (Fig. 5.3). Poor bonding between particles is the result of the use of a non-wetting binder solution. If the liquid is too viscous, the capillary attraction will be too low and the particles will be completely covered by the solution, reducing the bonds between particles. A wetting binder solution will be located in the neck regions by capillary attraction. Cellulose derivatives (methyl cellulose, hydroxyethyl cellulose, etc.) and starch derivatives are more useful binders. Plasticizers give plastic properties to the paste. Viscous and wetting polymers are used in this case. They also have the capability of holding water in the paste. The plastic property can be characterized by the saturation degree Sd:
Sd = volume of liquid solution / volume of empty space
nowettingliquid
wettingliquid
viscousliquid
Fig. 5.3. Locationof organic solution round the particles.
5 m CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
123
The most plastic state is reached if 0.9 < Sd < 1. Polyethylene glycols with high molecular weight and polyvinyl alcohols are used as plasticizers. Lubricants, used in small quantities, help the paste to slide in the extrusion apparatus. Glycerine is essentially added as lubricant. Deflocculants avoid powder agglomeration by steric effect or/and by controlling the surface charge of the particles. Acrylic acid polymers or fatty acids are very convenient. Other types of organic additives worthy of mention are anti-foaming agents, promotors of porosity, water retention agents which avoid water migration during shaping [6], antistatic, chelating and bactericide agents. The choice and quantity of organic additives are very important for the elaboration of porous ceramic bodies. Generally, 15-20% of the total weight of mixed powders is sufficient; the density and strength are not ameliorated if the quantity added exceeds this limit.
Mixing Mixing is an essential step in obtaining good dispersion and perfect homogeneity by the even distribution of constituents [7]. The order of adding the constituents (inorganic and organic parts and solvent) has an important effect on the characteristics of the paste [8]. If several inorganic powders are used, they are mixed first. Then, the organic compounds are added to the inorganic powders, beginning with the most hygroscopic to aid good dispersion in the mixture. At this stage, a small quantity of water can be added to obtain a wet powder. Water is immediately absorbed by the organics, therefore it must be added slowly to preserve the homogeneity of the mixture and avoid agglomeration. Rapid introduction of water into the mixture is conducive to agglomeration of organics which produces holes in the structure of the ceramic after firing.
Pugging After mixing, pugging is necessary to obtain a paste. During pugging, the progressive addition of water to the powder mixture leads to a very viscous paste. The quantity of water introduced is an important parameter as it gives the paste the necessary plasticity for shaping. It should be noted that the quantity of water that gives a good extrusion is the same as that corresponding to the best strength for wet supports. If more water is added, the paste becomes too soft for shaping. The pugging time must be carefully measured. If it is too long after a homogeneous paste has been obtained, the viscosity decreases and the paste grows too soft for shaping. As a consequence, tube deformation is observed during extrusion. A paste with plastic behaviour is able to be deformed without losing its cohesion and without returning to its initial shape.
124
5-
CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
Press
f
granulation
ageing
Fig. 5.4. Evolution of extrusion pressure versus ageing.
Ageing of the paste After pugging, the paste is kept in a closed box under high humidity to avoid premature drying and to ensure complete diffusion of the water and organic additives. The effect of the ageing step can be seen by measuring the extrusion pressure as a function of time (Fig. 5.4). The increase of viscosity during the ageing step (region I) is due to the fixation of water molecules by the polymeric chains and the polymeric network is blocked. The granulation breaks down the bridges formed between the chains and allows one to find again the extrusion pressure (region II) measured before ageing.
5.2.2 Tube shaping, drying and firing Shaping by extrusion Shaping is performed by extrusion. During this step, the paste is kept under vacuum to avoid the presence of bubbles. The paste is forced through the opening of a die with the help of an endless screw (in industry, Fig. 5.5) or a piston (in the laboratory). The extrusion speed can be chosen to elaborate homogeneous tubes. The geometry of the extrusion cone and die ensure the shaping and also the density of the wet tubes. Several types of dies for the shaping of mono- and multichannels tubes are shown in Fig. 5.6. Different shapes are obtained by changing the geometry of the die (e.g., the number of channels, diameter of channels, external diameter of the tubes). The influence of the presence or absence of lubricants on the extrusion pressure, which decreases when the percentage of lubricant increases, and on the compaction of the ceramic can be seen in Fig. 5.7. After extrusion, the tubes expand due to the relaxation of the elastic paste after compression. This relaxation must be taken into consideration for the drying shrinkage.
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
.
~/1
/1
/3
125
/31
Fig. 5.5. Schematic view of industrial extrusion apparatus. (1) Endless screw; (2) paste inlet; (3) compression; (4) vacuum; (5) pressure gauge; (6) vacuum chamber; (7) die.
1
1
I ...... Itubular 0 support
1 ~
~
multichannel support
Fig. 5.6. Different geometries ofdies.
Drying The wet tubes are set on rollers to ensure homogeneous drying and to avoid twisting and bending. The drying temperature is always less than 100~ The organic additives have a temperature of decomposition higher than this value, so only water is eliminated during the drying stage. The first step corresponds to the elimination of interstitial water which fills up the space between particles; shrinkage occurs at this stage. A low energy is necessary for this elimination. The second step of water elimination is due to the removal of adsorbed water (Van der Waals bonds) without shrinkage but with formation of pores.
126
5 ~
CERAMIC
PROCESSING
TECHNIQUES
OF SUPPORT
~ ....
,9 ~,~, 9~ ~ " ~ / ~
~.. ~.~,~. ~" ~ ' - ~ , ~ , ~
;~
SYSTEMS
FOR
' ~L..~.
MEMBRANES
~.~r162
~~~
SYNTHESIS
-..~
_~, ~ , ~.:;
~,,~ ~ "
~
~
~
i
~ ~.~
with lubricant
without lubricant
Fig. 5.7. Compaction versus presence or absence of lubricant.
Shrinkage (%) 10
1"0
15
Weight loss (%)
Fig. 5.8. Shrinkage during drying.
The speed of water migration in the structure determines the rate of temperature increase. Nevertheless, the speed can be optimized by controlling the humidity of the drying atmosphere. Figure 5.8 shows the evolution of shrinkage versus weight loss. This curve is interesting for establishing the drying temperature treatment: when shrinkage occurs the drying kinetics must be very low to avoid the formation of defects and when shrinkage is finished the drying rate can increase without damage to the ceramic.
Firing The firing treatment strengthens the ceramic. This treatment must be established on the basis of a thermal study to determine the elimination temperature of bonded water and organic additives and, eventually, the temperature of
5 m CERAMICPROCESSINGTECHNIQUESOF SUPPORTSYSTEMSFOR MEMBRANESSYNTHESIS
127
allotropic transitions. The firing treatment can be described in two stages: the first corresponds to the combustion of the organics, and the second allows the ceramic to sinter by densification and grain growth. The density, porous volume, pore diameters and mechanical resistance depend on the temperature and the time of sintering. The firing treatment is easier in the case of small particles [9]. Several workers [10,11] describe the sintering mechanism in oxide powders as a diffusion of vacancies in the ceramic. The grain growth corresponds to a reduction of surface free energy.
5.2.3 An Example of Preparation The elaboration values correspond piston extruder. There are about wt% of organics is
of cordierite tubular supports is described in Fig. 5.9. The to 3 kg paste preparation and shaping with a laboratory 4 parts of cordierite powder for 1 part of organics. The 18 composed as follows: 4 wt% of gel of corn starch as binder,
ORGANIC ADDITIVES cellulose derivates + starch
CORDIERITE POWDER 20 to 100 lam 82 wt %
[
!
18 w t %
mixing (15 min) + water (30 wt % of powders) pugging (30 min)
PASTE
I I ageing (3 days)
I I extrusion I rate = 1 cm / s I I pressure = 15 bar WET TUBES
I
I drying at room temperature for, at least, 2 days
I I sintering (> 1260 ~
i [ CERAMIC SUPPORTS ]
Fig. 5.9. Preparation scheme of cordierite supports.
128
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
1260 ~
1300 ~
1340 ~
1380 ~
Fig. 5.10. M i c r o g r a p h s of the ceramic at various sintering t e m p e r a t u r e s .
4 wt% of methyl-hydroxypropyl cellulose as binder-plasticizer and 10 wt% of starch as promotor of porosity. The quantity of water is 30 wt% related to the total weight of organic-inorganic mixture. The apparent viscosity measured after 2 days of ageing is equal to 8.105 Pa.s. The lowest firing temperature allowing a good mechanical resistance is 1260~ The evolution of the structure during the sintering can be observed in Fig. 5.10 (experimental results) and Fig. 5.11 (theoretical evolution). A good correlation is observed. A study of the influence of the sintering temperature was conducted on the porous characteristics and the mechanical resistance (fracture stress and Young modulus). Figure 5.12 gives some results of this study [12]
5 m CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
Sintering temperature (~
Textural evolution
1220
129
Mechanical resistance
no sintering
,
I no sinterin 8 ....
1260
slight mechanical resistance
formation of grain boundaries
1300
growth of grain bdundaries
increase of mechanical resistance relative deformation
1340 decrease of elasticity monolythic cerarmc
1380
increase of resistance
Fig. 5.11. Theoretical evolution of the structure during the sintering.
130
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS ~Q
,u
45 4Q 35' 3g'
25
2~ . ~ 4~e
l a 6 e~
.... 1 2 aao
a3~aQ
l ~ k e'
J . ~ 4~e
1 ~ "6 e
1 3 s: g
14ee
A 50 45 4@ §
35 311 25 21t, 15, 10,
@
124Q
:
126g
'
I'
128g
-
i
13@@
i
132@
"
13
i
GI
I
1360
t
138g
,
J.-40g
B
Fig. 5.12. (A) Evolution of the porosity (%) and (B) evolution of Young modulus (+ = E in GPa) and fracture stress (m = r~in MPa) as a function of the firing temperature. 5.3 TAPE CASTING Tape casting is a well k n o w n method for making thin, flat and dense ceramics [13-15]. This technique is limited by the thickness (a few millimetres) of the films obtained. The most important applications are in the electronics industry. Simon et al. [16] have described the preparation of flat ceramic membranes. This type of geometry is receiving interest because of the possibility of obtaining a high module compactness (compactness is defined by the ratio of membrane surface area to module volume).
5 - - C E R A M I C PROCESSING T E C H N I Q U E S OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
131
~SOLVEN'T]
l' INORGANICPOWDER ! I
I
deflocculant milling U.S. treatment binder + plasticizer
[STABLE SLURRY I Fig. 5.13.Flow diagram of slurry preparation. The tape casting process is composed of two different steps: the slip (or slurry) preparation, and the shaping [17]. The process needs a slip of inorganic powder dispersed in a liquid. The slip also contains binders and plasticizers which are long-chain polymers giving good flexibility to the green tape. The slip is then spread on a flat support. After drying, the tape is removed from the support and handles very easily because of its plastic characteristics.
5.3.1 Slurry Preparation The process for preparing a stable slurry is given in Fig. 5.13.
Solvent Water or organic liquids can used as the solvent. Water requires a higher drying temperature or a longer drying time than an organic solvent. In aqueous media, hydrogen bonds can be established and lead to a flocculation so, in order to accelerate drying, organic liquids such as alcohols and aromatic hydrocarbons are commonly used. It is convenient to use a mixture of solvents to control the tape drying rate. After evaporation, it leaves a dense organic-ceramic composite. In all cases, the solvent must dissolve the organic compounds and distribute them uniformly throughout the slurry [18]. The most volatile solvents are suitable for very thin films and lower volatile solvents for thicker films.
Deflocculants Deflocculants are indispensable for keeping the ceramic particles in a stable suspension. They act through charge repulsion or steric hindrance. In both
132
5 ~ CERAMIC PROCESSINGTECHNIQUESOF SUPPORT SYSTEMSFOR MEMBRANESSYNTHESIS
cases, the deflocculants are adsorbed on the surface of the particles. The influence of charge repulsion on the stability of suspensions is an important parameter [19-21]. This influence increases when the particle size decreases. Menhaden fish-oil (fatty unsaturated acid), phosphate ester and glyceryl trioleate are commonly noted in the literature as acting efficiently as deflocculants.
Milling and ultrasonic treatment Milling and ultrasonic treatment are used for homogenizing the slurry. Ball milling can occur in two stages: after the addition of deflocculants and after the addition of binders and plasticizers if the viscosity is not too high. The parameters influencing milling and ultrasonic treatment are the viscosity of the slip, the quantity of deflocculants and, of course, the time of the applied treatment. Ultrasonic treatment is a better technique than ball milling to avoid impurities. Binders and plasticizers must have the same properties as those necessary for extrusion. They must decompose without leaving residues. Binders ensure the mechanical resistance of the green tape because they form a film round the particles; they are long-chain polymers allowing good flexibility to be achieved. The choice of binders depends on the solvent type and the viscosity necessary for the thickness desired. Many polymers can be used as binders; the more commonly employed are: polyvinyl acetate, polyvinyl butyral, acrylic compounds, polystyrene, etc., and polyvinyl alcohol in aqueous media. Plasticizers make the tapes flexible for easy handling; alkyl phthalate and polyethylene glycol (PEG) are used as plasticizers. To avoid changing the properties of the binder when adding the plasticizer, couples of binder-plasticizer such as polyvinyl butyral with PEG and acrylic compounds with alkyl phthalate are used. The quantities of binders and plasticizers are optimized to give the dried tape good mechanical properties. TABLE 5.1 Tape casting compositions in wt% Material
Solvent
Deflocculant
Binder
Plasticizer
Viscosity (mPa.s)
Ref.
A1203 (74.6)
water (18.4)
polyacrylic resin 0.3/A120 3
styrene acrylic latex 9.1/A120 3
-
1000
22
A1N (75)
MEK*/ethanol phosphate 66/34 vol ester (17.1) (0.3)
acrylic resin (3.4)
butyl benzyl phthalate (4.2)
1000
23
* MEK = butanone 2.
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
133
Each composition has to be adjusted as a function of the composition and the desired tape thickness. Some rules must be taken into account, e.g.: the ratio of plasticizer to binder (less than 2); the deflocculant q u a n t i t y - not more than the sufficient quantity for complete adsorption (to determine deflocculant adsorption quantity, ceramic powder is immersed in deflocculant solution, and after filtering and evaporation of the solution, the unadsorbed deflocculant quantity is determined by the residual weight); the solvent amount (just necessary for dissolving the organic additives); and the weight ratio of organics to inorganic powder as low as possible, in the range 0.05 to 0.15. The slurry is characterized by its viscosity. Usually, the viscosity varies in the range 1000 to 5000 mPa.s. Two examples of tape casting composition are given in Table 5.1.
5.3.2 Shaping and Flat Ceramics Tape casting results from the relative movement between the doctor blade and the support or carrier (Fig. 5.14). There are two possibilities: the doctor blade moves and the carrier is fixed, or the doctor blade is fixed and the carrier moves [24]. In laboratory operations, batch casting with a mobile doctor blade and stationary carrier are used. Plate glass is usually used as a carrier. Tape drying can take place on the support by electric heaters placed under the support. If only partial drying occurs, the tape, with or without the support, is placed in an oven to complete the drying. Tapes of 0.5 m wide, 2 m long and 1-2 mm thick can be obtained. Large productions are based on continuous casting machines where the doctor blade is stationary. A moving carrier, such as an endless stainless-steel or plastic-film belt, moves under the fixed doctor blade. In this case, the carrier width is I m and the length can reach 40 m. The tape is dried by passing through a tunnel where solvents are evaporated. Dried tape can be removed from the carrier and rolled for storage at the end of the carrier. drvino
11
carrier
doctor blades
slip
/
support
heater elements
Fig. 5.14. Doctorblade apparatus.
carrier storage roll
134
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
The speed of casting depends on the length of the carrier, the drying time, the thickness of the tape, and the type of casting machine - - continuous or batch. In general, casting speeds vary from 0.1 to 1.5 m/min. The thickness of the tape is a function of the inorganic powder content of the slurry, the viscosity, the casting speed and the doctor blade height. All these parameters must be controlled to get reproducible tapes. The drying stage comprises three steps: solvent diffuses through the slurry to the surface, solvent evaporates at the surface, solvent is removed from the surface by a counter-flow of air. The second step is slow, because of the need for heat to evaporate the solvent. The third step is necessary to avoid a high concentration of solvent at the surface of the slurry and the formation of a skin on the upper surface of the tape. The drying temperature is limited by the boiling point of the solvent if the formation of bubbles is to be avoided. The same procedure of firing treatment is applied to this technique as is applied to extrusion. The ratio of the freshly deposited layer thickness to the dried tape thickness is about 2. The final tape thickness depends on the shrinkage during firing.
5.4 SPECIFIC
CHARACTERIZATION
METHODS
FOR
SUPPORTS
Several determination methods of a porous texture can be used for characterizing the supports. Certain of these methods (i.e., scanning electron microscopy, mercury porosimetry) are described in other chapters of this book. We shall only report here the specific methods of support characterization. 5.4.1 Bubble Point [25]
The bubble point measurement is recognized as an ASTM procedure (F31680 [26] and E128-61 [27]). This technique allows the determination of pore diameters and the presence of defects in the membrane. Bubble point is based on Jurin's law. If a porous membrane is impregnated with a liquid (e.g., water, alcohol) each pore has a meniscus of condensate at the gas-liquid interface which opposes the flux of gas. To unblock the pores a pressure Ap must be applied. According to the Jurin's law, the smaller the pores, the higher the pressure required r. Ap = 2(~cos 0
where r is the pore radius, Ap is the over-pressure with respect to the atmospheric pressure, c~is the surface tension, and 0 is the contact angle.
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
135
In this method a wetting liquid is chosen to obtain a cos 0 = 1. The requisite properties of the liquid used are that: it must not attack the material, it can be easily evaporated, it allows pressures compatible with the mechanical resistance. The method consists of observing the bubbles leaving the m e m b r a n e w h e n the pressure increases, or measuring the quantity of gas crossing the membrane. Jurin's equation indicates that larger pores bubble first. The surface tension of the liquids are in the range 16-105 N cm -1 to 30x105 N cm -1, except for water (72x105 N cm-1). Ethanol or water are therefore most c o m m o n l y used. Jurin's equation takes the following values: -
-
-
r-
0.446 1.44 for ethanol and r = ~ for water Ap Ap
where r is in m m and Ap in 10 5 Pa. It is easier to use water or ethanol to study macropores or micropores respectively with reasonable pressures. Figure 5.15 gives the typical curve obtained by this method. Point A corresponds to the bubble point; the linear part, after Point B, corresponds to the gas flow measured with a support not immersed in the liquid; Point C, at the inflection point of the curve, represents the mean value of pore diameters. The bubble point technique is thus a very convenient test to visualize the location of possible defects. In the presence of cracks, for instance, the first bubbles are located on the defects and the bubbles do not appear uniformly on the surface of the support.
GAS V O L U M E
B
/G
/ / PRESSURE Fig. 5.15. Bubble p o i n t d e t e r m i n a t i o n .
136
5 - - CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMS FOR MEMBRANES SYNTHESIS
5.4.2 Mechanical Resistance
We have already noted that the support provides mechanical resistance to the filtration medium, so this characteristic must be determined.
5.4.2.1 Burst pressure (test for tubes) Figure 5.16 shows an example of one type of apparatus for burst pressure measurement. The tube is placed on the outside of the deformable rubber membrane in the housing which is a protector wall. Pressure is then applied to the rubber membrane and measured. The value of burst pressure is an important parameter for assessing the limit of mechanical resistance during the filtration process. It depends on the firing temperature and can be compared with the results of bending strength.
5.4.2.2 Bending strength (test for cylindrical specimen) [28] The three-point bending strength test is generally used to control the quality of a material. This test is not really a test for membrane characterization but only
3
1 : housing rubber membrane
2 :
ceramic support pressure gauge 5 : air inlet
3 :
1
4 :
I
I\LJ I
Fig. 5.16. Burst pressure apparatus.
.
.
.
.
.
.
.
.
t
5 -- CERAMIC PROCESSING TECHNIQUES OF SUPPORT SYSTEMSFOR MEMBRANES SYNTHESIS
137
y \:-I.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
L Fig. 5.17. Three-point bending strength.
for the sintered material because of the correlation between the mechanical resistance and the sintering temperature; this latter is directly linked with the porosity characteristics of the material. The principle consists in setting the sample on two supports and applying a stress until the sample fractures, as is shown in Fig. 5.17. The stress and the strain are measured and allow determination of the mechanical resistance of the material as a function of the firing temperature. It is easier to prepare cylindrical samples by extrusion. The length-to-diameter ratio is always taken higher than 10. The fracture stress is defined by the equation:
L.P ~3max = 8
D3
(5.1)
with r~ in Pa, L = length between the supports in m, D = sample diameter in m, and P = applied stress in N. The Young m o d u l u s E is:
4L3.p =
3~D4. y
(5.2)
with E in Pa, and Y - strain at the fracture in m . From Eqs. (5.1) and (5.2), a relation between t~maxand E can be found: r~ = k E with k = a.D.Y/L 2 5.5 CONCLUSION In conclusion, it can be stated that the extrusion technique is the most important method of preparing ceramic supports, because the tubular configuration is the most common configuration. In the literature, there are an increasing
138
5 m CERAMICPROCESSINGTECHNIQUESOFSUPPORTSYSTEMSFORMEMBRANESSYNTHESIS
n u m b e r of publications related to the tape casting t e c h n i q u e and, in the future, this c o n v e n i e n t m e t h o d will be a p p l i e d to the p r e p a r a t i o n of flat configuration. In these t w o techniques m u c h e x p e r i m e n t s a n d t h o r o u g h analysis of each p r e p a r a t i o n step are necessary to obtain r e p r o d u c i b l e ceramic s u p p o r t s .
REFERENCES
1. J. Benbow and J. Bridgwater, Paste Flow and Extrusion. Clarendon Press, Oxford, 1993. 2. T.A. Smith, Organic binders and other additives for glazes and engobes. Br. Ceram. Trans. J., 61 (1962) 523. 3. G.Y. Onoda, Theoretical strength of dried green bodies with organic binders. J. Am. Ceram. Soc., 58 (1976) 1975. 4. C.A. Bruch, Die-pressing submicron size A1203 powder. Ceram. Age, 92 (1967) 44. 5. A.G. Pincus and L.E. Shipley, Utilization of microstructure in processing of ceramics. Ceram. Ind., 92 (1969) 106. 6. A.R. Teter, Binders for machinable ceramics, Ceram. Age, 82 (1966) 30. 7. A.G. Evans, Considerations of inhomogeneity effects in sintering. J. Am. Ceram. Soc., 65 (1982) 497. 8. A.L. Salamone and J.S. Reed, Preparation and microscopic analysis of cellulose binder solutions. Am. Ceram. Soc. Bull., 58 (1979) 618. 9. D.Ph. Hasselman, Strength behavior of polycrystalline alumina subjected to thermal shock. J. Am. Ceram. Soc., 53 (1970) 490. 10. G.C. Kuczynski, L. Abemethy and J. Allan, Sintering of alumina, in: W.D. Kingery (ed.), Kinetics of High Temperature Processes. MIT Press, Cambridge, MA, 1959, pp. 163-172. 11. D.L. Johnson and I.B. Cutler, Diffusion sintering, h initial stage models and their applications to shrinkage of powder compacts, II: intial sintering kinetics of alumina. J. Am. Ceram. Soc., 46 (1963) 541. 12. V. Thoraval, Elaboration et caract6risation d'un 616ment de microfiltration et d'ultrafiltration en c6ramique. Thesis, Montpellier, 1990. 13. E.P. Hyatt, Making thin, flat ceramics. Ceramic Bull., 65 (1986) 637. 14. R.E. Mistler, D.J. Shanefield and R.B. Runk, Tape casting of ceramics, in: G.Y. Onoda and L.L. Hench (ed.), Ceramic Fabrication Processing Before Firing. Wiley Interscience, New York, NY, 1978, pp. 411-448. 15. P. Boch, T. Chartier and M. Huttepain, Tape-casting of alumina/zirconia laminated composites. J. Am. Ceram. Soc., 69 (1986) C191. 16. C. Simon, R. Bredesen, H. Raeder, M. Seiersten, A. Julbe, C. Monteil, I. Laaziz, J. Etienne and L. Cot, Tape casting of flat ceramic membranes. Key Eng. Materials, 61/62 (1991) 425. 17: T. Chartier, Tape casting, in: D. Bloor, R.J. Brook, M.C. Flemings and S. Mahajan (eds.), The Encyclopedia of Advanced Materials. Pergamon, Oxford, 1994, p. 2763. 18. H. Burrell, Solubility parameters, Interchim. Rev., 14 (1955) 3. 19. G.D. Parfitt, Dispersion of Powders in Liquids. Elsevier, New York, NY, 1969, p. 315. 20. V.T. Crowl and M.A. Malati, Adsorption of polymers and the stability of pigment dispersions. Discuss. Faraday Soc., 42 (1966) 301. 21. D.H. Napper and A. Netschey, Steric stabilization of colloidal particles. J. Colloid Interface Sci., 37 (1971) 528. 22. A. Kristoffersson and E. Carlstr6m, Tape casting of alumina in water with an acrylic
5 -- CERAMICPROCESSINGTECHNIQUESOF SUPPORTSYSTEMSFORMEMBRANESSYNTHESIS
23.
24. 25. 26. 27. 28.
139
latex binder. Extended abstract of International Conference on Shaping of Advanced Ceramics, April 25-27, 1995, Mol, Belgium. M. Descamps, B. Thierry, A. Leriche, Powder characteristics influence on the properties of tape casted alumina and aluminium nitride materials. Extended abstract of International Conference on Shaping of Advanced Ceramics, April 25-27, 1995, Mol, Belgium. R.B. Runk and M.J. Andrejco, A precision tape casting machine for fabricating thin, organically suspended, ceramic tapes. Am. Ceram. Soc. Bull., 54 (1975) 199. J. Charpin and B. Rasneur, Caract6risation de la texture poreuse des mat6riaux. Techniques de l'Ingdnieur, P 1050 (1987) 17. Fiche A.S.T.M. F 316-80, Standard test method for pore size characteristics of membrane filters for use with aerospace fluids. Fiche A.S.T.M. E 128-61, Standard test method for maximum pore diameter permeability of rigid porous filters for laboratory use. J.M. Dorlot, J.P. Bailon and J. Masounave, Des Mat&iaux. Ecole Polytechnique de Montreal, (ed.) Canada, 1986, p 11.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 6
Preparation of asymmetric ceramic membrane supports
by dip-coating
B.C. Bonekamp Netherlands Energy Research Foundation ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands
6.1 INTRODUCTION Ceramic membranes are asymmetric layered structures composed of a separation layer which fulfils the actual m e m b r a n e function, and a ceramic support structure comprising I t o 5 layers. The support structure serves as a substrate, it is needed for general mechanical stability and must have larger pores than the separation layer. The support structure usually consists of a single or multi-hole support tube, layer 1, to which as m a n y ceramic layers are applied as are necessary to support the separation layer with the required selectivity. A typical single-hole support tube thickness is 3 m m (outer/inner diameter 14/8 mm) and typical pore diameter 5 ~tm. Each subsequent layer has smaller pores and a thickness of 100-1000 times the pore diameter. The higher the m e m b r a n e selectivity required, the more support layers are needed. Single-hole and multihole tubular support structures are shown in Fig. 6.1. Ceramic m e m b r a n e s m a y generally be classified as follows: Membrane for Structure Typicalaverage pore diameter Separationlayer Microfiltration I layer 5 Bm macroporous 2 layers 0.25 Bm macroporous 3 layers 0.10 Bin macroporous Ultrafiltration 4 layers 5 nm mesoporous Hyperfiltration/gas sep5 layers 10 A microporous aration / p ervaporation
142
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
Fig. 6.1. R a n g e of single-hole a n d m u l t i - h o l e t u b u l a r s u p p o r t structures.
It will be clear to any ceramic engineer that the support structure requirements for a 20 ~tm thick microfiltration membrane with 200 nm pores and for a support structure for a 100 nm thick hyperfiltration membrane with 0.7 nm pores are very different. These differences relate to the final properties of the support structure needed as well as the processing route to prepare them. A discussion of these differences is included in this chapter. Porous ceramic coatings on a porous substrate can be prepared by dipping the substrate into a ceramic dispersion and by subsequently withdrawing it from that dispersion. This 'dip-coating' or 'withdrawal coating' method, as a method to prepare one or more porous ceramic layers on extruded porous tubes, is the main subject of this chapter. The emphasis is largely on the wet stage in the layer forming process rather than on the pore structure formation during drying and sintering. Layer material properties are only briefly discussed. The discussion is limited to dip-coating with tubular systems and systems where the pore space of the coating is determined by particle packing principles. Although sol-gel coatings with particulate sols are in principle included, the main attention is focused on coatings obtained with suspensions of submicrometer powders. In addition to the dip-coating method, ceramic layers on a porous substrate can also be formed by filtration of a sol or by filtration of a suspension using the substrate as a filter and applying an external pressure. These methods are not discussed in this chapter. The reader is referred to Ref. [1] for more information. Section 6.2 gives a rather general and largely qualitative description of forming support coatings by dip-coating as a form of colloidal processing. The preparation of multi-layered supports for thin microporous membranes is
6 --PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
143
discussed including the preparation of simpler supports, such as for ultrafiltration membranes. Section 6.3 treats dip-coating in more depth to give the reader a better insight into the operative mechanisms and to show how coating theory can assist in the design and analyses of actual coating conditions. In particular, the importance of wetting and dewetting in withdrawal coating is explained and the role of macromolecules to control the layer forming process is discussed. The chapter ends with an example and exercise section to illustrate how the content of Section 6.3 can be used in the preparation of experiments in a development path of ceramic support coatings. The extrusion of porous ceramic tubes, which are the starting point of ceramic membranes, is treated in Chapter 5.
6.2 SUPPORTS FOR CERAMIC MEMBRANES
6.2.1 The Multilayer Support System A schematic drawing of a multilayer membrane support tube is shown in Fig. 6.2. The enlarged part reveals the layered structure of the material. The extruded support tube is layer 1.
9
o.
Fig. 6.2. Schematicmultilayer membrane support.
144
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
Fig. 6.3. Fracture surface of an all alumina multflayer membrane support (SEMpicture). Figure 6.3 is a SEM photograph of a cross-section of an all alumina multilayer tubular support as developed at ECN. It shows that the material is built up of particulate materials. The gradual decrease in grain size from layer I onwards is clearly visible. The granular structure of layer 4, a mesoporous y-alumina layer is not perceptible on this scale. Each support layer, except layer 1, is prepared according to the processing flow sheet shown in Fig. 6.4. The support shown in Fig. 6.3 serves as a support for a microporous membrane, for example, a silica membrane. Figure 6.5 is a SEM photograph of such a membrane layer. The silica layer is only 200 nm thick and is supported by the system shown in Fig. 6.3. The substrate for the silica membrane is the very smooth mesoporous y-alumina layer 4. Only a multilayer system can provide a substrate which is sufficiently smooth and flawless to serve as a support on which an almost defect-free microporous membrane can be made. In the following sections it will become
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
+
v
Suspension / preparation ,,
145
/ /Substra prepatioy
~
Withdrawal/ coating
Calcining+/ sintering
,,
Support
Fig. 6.4. Processflow sheet. clear why such a multilayer support system is needed. Ultrafiltration (UF) and microfiltration (MF) membranes can be made on less sophisticated supports. The simplest MF tubular membrane consists of an extruded porous tube (layer 1) as a support coated on the inside or outside with a macroporous layer (layer 2) which serves as the functional filtration layer. The support system shown in Fig. 6.3 is in fact a sophisticated UF or Knudsen gas separation membrane. For less demanding applications a 2-layer support could also be used. Most ceramic membranes are made from oxide powders. ZrO 2 and A1203 are the most common compounds for microfiltration membrane materials. Mesoporous UF membranes usually also consist of oxides. A1203, ZrO2, TiO2, CeO2 are most commonly used. Membranes from non-oxide compounds such as C, are also noted in literature (see Ref. [2] for an example).
146
6 ~ P R E P A R A T I O N OF ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
9
-....
9
.
.
.
.
.
..
~-
.
,.
"4,
Fig. 6.5. Fracture surface of supported silica membrane (SEMpicture). Alumina layer 3 supports alumina layer 4, which supports the silica layer 5. The following section will show that support requirements are partly generic and partly specific in character. The specific aspects are closely connected to the chemical composition of the membrane material.
6.2.2 Support Requirements Figure 6.6 shows SEM photographs of the surfaces of two different coarsegrained macroporous alumina extruded tubes. An important difference between support a and support b is the sinter temperature, about 1800~ for a and 1550~ for b. This means that production costs for b will be considerably lower than for a. The median pore size of material a is approximately 10 ~tm and about 4 ~tm for material b. The mean open porosity is 30 to 35% in both cases. The surface roughness is measured with a mechanical probe. It is about 2 ~tm in average (Ra) and the m a x i m u m roughness (Rmax) is approximately 20 ~tm for substrate a. The corresponding values are 3 to 4 ~tm and 30 ~tm for substrate b. The m i n i m u m surface area which has the same average properties with respect to pore size, porosity and roughness as the average properties over a large surface is somewhat smaller for substrate a than for substrate b. In other words, surface a is more homogeneous than surface b. Despite their different properties, these two materials are both suitable bases for the successful preparation of mesoporous membranes. However, the pore size of these supports and their roughness are too large (even more so for b than for a) to enable direct application of a mesoporous membrane with a thickness of a few ~tm on the tube surface. But, it is impractical to make similar support tubes with a smaller pore size and surface roughness, because the support structure would suppress the
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
147
(a)
9
~ "~ " " ~ ~
..... ' ,~'~
' W-I,~,;~, -i:i
(b)
~ d '''! ..... ~
;;,,, ~ . . . .
.,.,...: ..4F~ .
'
~
~,<,. . . . . . .
~< . . . .
:"
,.=.,,
,~
!: i.! .... ~,
,
,
,
~
Fig. 6.6. S u r f a c e of t w o d i f f e r e n t m a c r o p o r o u s e x t r u d e d a l u m i n a s u p p o r t t u b e s (SEM p i c t u r e ) . N o t e the difference in e n l a r g e m e n t . See also text.
permeability of the m e m b r a n e system too much. This means that there are generic reasons for the application of one or more layers: to decrease the average pore size of the substrate to decrease the surface roughness of the substrate - to decrease the void defect density. The average layer thickness of the first coating (layer 2) on the tube structures s h o w n in Fig. 6.6 must be at least 30 to 40 ~tm to cover the substrate completely, i.e. to ensure the absence of substrate surface roughness effects on the defect -
-
148
6 ~ P R E P A R A T I O N O F ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
density in layer 2. These substrate related defects involve (micro)cracks and voids. The permeability of the support system must be higher than that of the membrane layer (at least a factor 10). The layer thickness of each layer should therefore be as small as possible. The 'building bricks' for the coating layers are ceramic particles. Generally, the particle size must not be much smaller than the pore size in the substrate to enable successful coating. The essential prerequisites for a support structure have been discussed above. But there are more factors to be taken into account. In considering support requirements, it is useful to distinguish between support requirements for the dip-coating and drying steps and the support requirements for the subsequent calcining and sintering steps. For the latter processes the most relevant factors are: thermal expansion behaviour of the coating in comparison with the substrate substrate/membrane interaction chemical compatibility between substrate material and membrane material. There must be sufficient sintering or chemical bonding or interlocking of the membrane material with the substrate to ensure proper adhesion of the membrane to the support during application. Possible interaction of the membrane material with the substrate should not lead to decreased permeability or defects in the membrane system such as micro-cracks. However, in practice, the most important factor is that the support thermal expansion coefficient should match that of the membrane layer in order to avoid cracking during the sintering step. If the chemical composition of the membrane differs from that of the substrate, reasons for applying intermediate layers between the first tube material and the actual membrane layer may be: matching thermal expansion - buffer zone in case of chemical incompatibility during processing. These items will be discussed further in Section 6.2.5. During the dip-coating step and the subsequent drying, the following support properties are of primary importance: - pore size and pore size distribution (bulk and surface) - porosity - surface roughness - surface homogeneity - bulk homogeneity - wetting behaviour - surface chemistry. Unavoidable deviations from the mean values of the above parameters determine details of the layer formation process and properties of the coating to be obtained. -
-
-
-
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
149
(a) '<
,
.,
,
,
,:.
9
+
,.
,
~
.
(b)
i
+:,"
' [
Fig. 6.7. Substrate surfaces for ultrafiltration coating. Surface (a) is more ordered and smooth than surface (b). The 'smooth' spots in (a) are artefacts due to preparation. (SEM picture)
Figure 6.7 shows SEM photographs of the surface of supports for ultrafiltration membranes. Two differences between support a and b are immediately visible: (1) substrate surface a is much more ordered than b, while the average size of the 'building bricks' is not very different (2) surface a shows less spatial pore size variations than surface b, while the average pore size of both substrates is between 100 and 200 nm. It seems to be more difficult to achieve complete covering of the surface by a m i n i m u m layer thickness for substrate b than for substrate a, when the layer
150
6 - - P R E P A R A T I O N O F ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
forming mechanism is capillary filtration (see Sections 6.2.3 and 6.3.1). This is due to the larger spatial fluctuations in local porosity and pore size in b. In other words, the surface homogeneity of material b is less than that of a. Researchers at Twente University (The Netherlands) maintain that a sol-gel boehmite layer can only be applied on a substrate if the average roughness is smaller than 500 nm [3]. Our experience is, however, that on surfaces with an average roughness of about I ~tm virtually defect-free sol-gel layers can also be made. In fact, not only the average roughness R a matters, but other roughness characteristics such as roughness 'wavelength' are equally important. The minimum wavelength and amplitude which can be achieved is about the size of the grains in the substrate surface. The most demanding support requirements are those for ultra thin microporous gas separation membranes, which are currently being developed in several research organisations worldwide including ECN (Petten, the Netherlands). In principle, a mesoporous Knudsen or UF membrane can serve as support for these membranes if the defect density in the substrate surface, i.e. the mesoporous layer, is low enough. Indeed, the quality of the Knudsen or UF membrane as support for a microporous gas separation membrane should be higher than is usually needed for the UF or Knudsen function [4]. This means that not every mesoporous ceramic membrane is a suitable support for microporous or dense amorphous gas separation membranes. 6.2.3 Layer Formation on Porous Substrates
A ceramic coating technique suitable for the preparation of layered membrane (support) structures should cover a substrate surface completely with a ceramic layer in a controlled manner with a well-defined thickness, structure and texture. Also, the adhesion between substrate and coating should be such that delamination during application does not occur. When a tubular (or flat) substrate is withdrawn from a suspension or colloidal solution (sol) in a welldefined manner, a wet and more or less dense dispersion layer of well-defined thickness covering the substrate surface can be obtained. A prerequisite is that the substrate is wetted by the dispersion liquid. After drying, calcining or sintering of this layer (see Fig. 6.4) a consolidated coating can be achieved. The microstructure, texture and macrostructure of the coating can be controlled using the principles of powder and dispersion technology and colloidal ceramic processing. A great deal of information on these subjects can be found in Ref. [5]. A multilayer structure can be obtained by repeating the coating procedure, usually including the sintering step and using adapted conditions for each coating. Dip dispersion coating is a colloid processing technique used as a method for the preparation of coatings. There are two possible compaction modes when a porous substrate is withdrawn from a dispersion:
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
151
Suspension
Important parameters
....,... :.:.:.: ,....., :"-ii ...
Q
Suspension 9 Volume fraction of solids 9 Agglomerate size distribution (time dependent) 9 Viscosity continuous phase
Q
Support 9 Porosity 9 Pore size distribution 9 Capillary action
Filtration cake
Q
Contact time
(coating)
@ Particle size/pore size ratio
A,/X
Fig. 6.8. Summary of prhaciple and important parameters in the capillary colloidal filtration mode of dip-coating.
- capillary colloidal filtration, - film-coating. Capillary colloidal filtration occurs when the dry substrate comes into contact with a dispersion and the pore surface is wetted by the dispersion liquid. The capillary suction of the substrate which occurs, in effect drives particles to the interface. If the surface is not permeable for the particles, the particles concentrate at the substrate-dispersion boundary and a compact layer is formed. The thickness of the compact layer grows with time according to the well-known square root time law (see Section 6.3.1), until the substrate is saturated with dispersion liquid or, in case of a dispersion of small colloidal particles, a stationary state due to back-diffusion occurs. Figure 6.8 summarises the capillary filtration mode of dip-coating. Film-coating is the formation of an adhering dispersion layer as a result of the drag force exerted by the substrate during withdrawal from a dispersion ~. To achieve pure film-coating, the capillary suction of the porous substrate must be suppressed. The thickness of the entrained dispersion layer increases with increasing withdrawal speed and increasing dispersion viscosity 2. The contact 1 Draining a fluid from a tube is included here. When the frame of reference is in the fluid it is considered dip coating. 2 In practice coating fluids are often non-Newtonian dispersions. The complete rheological behaviour is then of importance.
152
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
Vw
Important parameters: Q
Suspension 9 Viscosity 9 Yield value 9 Density 9 Surface tension Support 9 Porosity 9 Surface roughness 9 Radius 9 Wetting
Q
Withdrawal speed Vw
Tube wall Fig. 6.9. Summary of principle and important parameters in the film-coating mode of dip-coating.
time is not relevant here. The thickness of the layer after drying is determined by the wet thickness, the volume fraction of solids in the dispersion and the packing density of the particles in the dry coating. The film-coating mechanism is summarised in Fig. 6.9. Figure 6.10 shows the layer thickness as a function of withdrawal speed for film-coating and capillary filtration as the main coating mechanisms. The decrease for capillary filtration is caused by the decreasing contact time with increasing withdrawal speed. At constant withdrawal speed and constant contact time the coating layer thickness can be varied by varying the volume fraction of solids in the dispersion. When film-coating is the prevailing mechanism, increasing volume fraction means increasing viscosity and hence increasing wet layer thickness. In the case of capillary filtration, increasing volume fraction solids causes a faster growth of the layer. In both cases the actual layer thickness obtained depends on the volume fraction (or porosity) of the particle packing obtained on the substrate. A particle packing process is a process which causes an increase in the volume fraction of a particulate material, i.e. the dispersion. The densification takes place stepwise in the colloidal filtration process and becomes more gradual
6 - - PREPARATION OF ASYMMETRICCERAMIC MEMBRANE SUPPORTS BY DIP-COATING ~
153
~O
withdrawal speed Vw Fig. 6.10. Layer thickness as a function of w i t h d r a w a l (drainage) speed (schematically) for filmcoating m o d e and capillary filtration mode.
and restricted in the drying step which follows. In the film-coating process the particle packing process takes place during the drying of the entrained wet coating. The porosity of the dry compact layer depends primarily on the shape and normalised size distribution of the kinetic units (freely moving particles) in the dispersion. The main factor determining the pore size of the particle packing is the size and shape of the kinetic (freely moving) particles which are present in the dispersion. The size distribution is here of secondary importance. Details of the dry coating pore structure depend on the pore properties of the wet packing and the effect of capillary drying forces, which in turn depend on the strength of particle packing. The pore structure of the wet packing is very sensitive to the interaction between the particles in the dispersion. This is where colloid science and colloidal processing play a role. When the particles have a high sticking probability upon collision (agglomerate forming), the porosity of the particle packing will be high (low packing density). Low sticking probability leads to a dense wet particle packing with a narrower pore size distribution. Figure 6.11 shows diagrams of typical packing structures in concentrated suspensions and / or wet particle packings. Colloidal stable dispersions without agglomerates and aggregates are generally a prerequisite for the preparation of defect-free, homogeneous substrate coatings. Agglomerated suspensions usually give rise to compressible compacts in capillary filtration which result in a more dense compact near the substrate surface [7]. When the agglomerates are (purposely) very weak the packing process in capillary filtration can still cause homogeneous close packing of the particles [8]. There is no principal difference in particle packing from suspensions and particulate sols. In both cases the densification of the dispersion may be accompanied by agglomeration of the particles and the formation of a continuous
154
6 - - P R E P A R A T I O N OF ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
a
b
c
Fig. 6.11. Typical packing structures in concentrated suspensions: (a) simple network of chains; (b) denser flocs, connectedby chains; (c) dense packing of spheres surrounded by a thick immobile layer of solvent or absorbed species (redrawn from Sonntag and Strenge [6]). network or a gel structure. However, in ceramics literature the term 'gel' is usually reserved for the gelation of a sol. Also, very highly viscous sols are often incorrectly called a gel. For example, an alumina coating with a median pore size of typically 100 n m can be prepared from a suspension (in water) of commercially available submicron alumina p o w d e r with a mass based median diameter of 500 nm. In such a suspension colloidal interactions determine to a large extent the properties of the suspension. The particle packing properties are disturbed by the presence of a fraction of aggregates which always exist in such commercial powders. This fraction can be removed from a colloidally stable suspension by means of sedimentation fractionation (see Ref. [5] for an example). In practice, both capillary filtration and film-coating often occur simultaneously. The contact time as well as the withdrawal speed must then be controlled. This can be done with a set-up as shown in Fig. 6.12. A similar set-up can be used for inside tube coating, at least, for not too small inner diameters. The w i t h d r a w a l speed can be controlled as indicated. The contact time can be controlled by adjusting the withdrawal speed a n d / o r adjusting the length of the coating bead. End effects are minimised by using tube holders as indicated. The coating of the whole hole surface of multi-hole tubes is more difficult and requires other skills. It will be shown later in this chapter how it could be advantageous to prepare ceramic m e m b r a n e substrate coatings with film-coating only as the withdrawal layer forming mechanism or with mainly colloidal filtration as the forming mechanism. In the former case, layer forming by capillary action of the substrate must be suppressed. In the latter case substrate wetting and pore properties are important, and the dispersion concentration should be high enough to ensure cake filtration conditions but low enough to prevent film-coating. When macromolecular binders are added to the dispersion the film coat contribution can also increase due to the thickening effect of the binder. Dip-coating in the film-coating m o d e with high volume fraction suspensions requires complete suppression of the capillary action of the support. This can
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
155
~l~:'--~ ~, ..~
tube holder
coating /
suspension
/
? outlet suspens.!on 9
! /I
I/iiiiill
l~!iiil, II
I I
l~i~i~il
sus.pension inlet :
flexible diaphragm suspension container
porou s ceramic tube
adjustable speed, Vw tube holder
~,. s
Fig. 6.12. Experimental set-up for dip-coating of tubular substrates (outside coating).
be achieved by hydrophobizing the support by treating it with a silane coupling agent (see Fig. 6.13). For coating purposes the contact angle of the substrate for water needs to be larger than 90 ~ However, a contact angle larger than 120 ~ is undesirable, because the dynamic contact angle then does not become small enough for the initial forming of a coating flow. The dynamic coating also becomes too unstable, i.e. the probability of dewetting becomes high. These instabilities can be suppressed by adding surfactant to the dispersion. Section 6.3.2 will further discuss this subject. In principle, both colloidal filtration and film-coating can be used to prepare coatings with a thickness adjustable between 100 nm and 100 pm and pore sizes covering the micropore, mesopore and part of the macropore range. Much depends on the properties of the substrate as has been shown.
156
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING RSi(OCH3)3 Hydrolysis 3H20
3CH3OH (Methanol)
~
RSi(OH)3 Condensation ~
2RSi(OH)3
R I
R I
2H20
R I
HO-Si-O-Si-O-Si-OH I
I
OH OH
I
OH
OH OH OH I I I
R ]
R I
Substrate
R I
HO-Si-O-Si-O-Si-OH Hydrogen Bonding
I
I
I
I
I
I
O O O / \/ \/ \ H FIH HH \ /\ /\ / O O O
A
2.20
R
Bond Formation
Substrate
R
R
I I I HO-Si-O-Si-O-Si-OH I
I
I
0 0 0 I [ I Substrate Fig. 6.13. Hydrophobising an oxide surface by treatment with a silane coupling agent (redrawn from Witucki [9]). M a c r o m o l e c u l e s are v e r y i m p o r t a n t in the p r o c e s s i n g of p o w d e r s into p o r o u s s u p p o r t layers. T h e y a r e u s e d to e n s u r e t h e c o l l o i d a l s t a b i l i t y of c e r a m i c s u s p e n s i o n s , to c o n t r o l t h e r h e o l o g y of s u s p e n s i o n s a n d p a s t e s , to a v o i d c r a c k i n g
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
157
during drying, to control the mechanical properties of the green ceramic, and to control the pore space structure. There is usually one main purpose for the use of a polymer additive which produces one or more side effects. Ceramic suspensions generally contain more than one polymeric compound, surfactants, lubricants, and plasticisers. The interaction between all these compounds determines the suspension behaviour and the microstructural development during compaction, drying and calcining. For example, the thickening effect of a polymer could be increased by the interaction with a wetting agent (see for polymer-surfactant interactions for example Ref. [10]. Competitive adsorption effects [11,12] are important for the order in which components must be added in suspension preparations. Polymer solutions prepared from a polymer powder often contain aggregates of non-dissolved polymer molecules or microgels which can cause voids and surface defects in the ceramic part in the micrometer range. A heterogeneous distribution of polymer in the green layer may be one of the causes of macroscopic heterogeneity in the fired porous ceramic support. The flow behaviour of polymer through the compact during formation a n d / o r drying determines this distribution. Wetting phenomena play an important role in almost every step of the processing of ceramic powders. Dip-coating is no exception to this. The suction power of a substrate depends on pore size and the wettability of the pore space. The drying shrinkage of particulate layers depends on the wetting and hence capillary properties. Because the porous layer is formed on a substrate, the initial wetting and possible break-up of the coating play a role and should be controlled. In coating operations, the most important reason for adding a wetting agent is the improvement of substrate w e t t i n g b y means of the coating liquid. In choosing a surfactant for this purpose, attention should be paid to possible undesirable interference with the stability of the suspension or sol. Unwanted induction of agglomeration or phase separation in coating suspensions may also occur when polymeric thickeners or mixtures of polymers with different chemical composition are introduced into the dispersion to enhance the viscosity of the dispersing liquid. The presence of thickening polymers in the dispersion often also causes practical problems: some polymers are soluble at low temperature and others at high temperature. Although the solubility may be high, the solution process itself can be slow. The formation of non-wetted polymer powder agglomerates enveloped by a gelatinous dissolved polymer layer should be avoided. The order in which formulation components are added can give different results. Mixing a highly viscous polymer solution with a low viscous particle dispersion may be cumbersome. High shear during mixing can degrade the polymer, especially, when the molecular mass is large. In practice, the development of an optimum preparation route for a coating dispersion, for example, consisting of oxide powder, surfactant, dispersant and
158
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
thickener is often a combination of empirism and application of colloid science and technology. As an alternative, experimental design methods could be used to find the optimum preparation path (see for example Ref. [13]). Rheological properties of ceramic dispersions and green structures are related to the interaction forces between the particles and with the microstructure of the material. Unfortunately, detailed quantitative interpretation in terms of microstructure is only possible for model systems of monodisperse particles of which the interaction energies are precisely known [14]. Rheological properties of sols and suspensions are very important in withdrawal coating processes. In the colloidal filtration mode, rheological measurements on the coating suspensions provide a clue for the presence of agglomerates in the dispersion, so that information about the degree of dispersion can be obtained. For the film-coating mode, however, the constitutive equation of the dispersion is of primary importance. Rheology will feature later in Section 6.2.4. However, for an introduction to rheology the reader is referred to a handbook such as Ref. [15]. An important motivation for continuing to develop theories or models for colloidal filtration and film-coating is to find possible answers to vital questions in the technological design of dip-coating as a unit operation in ceramic processing (see Ref. [16]). However, there is still a long way to go before actual applications will be feasible with concentrated suspensions of irregular and anisometric particles in connection with rough (on the particle size scale) substrates. Nevertheless, the models can at least be qualitatively used to predict trends for layer forming kinetics and to predict the relationship between dispersion properties and support properties on the one hand and structure of the coating on the other hand. Section 6.3 will further elaborate on this. Theories for drying and sintering thin films on supports are still being developed. Only a few aspects will be discussed in Section 6.2.5 on drying and sintering. The theoretical description of capillary filtration is much more difficult when the continuous phase is a polymer solution which behaves in a non-Newtonian way. For a full appreciation of this problem, Darcy's law [17], which is the basis for filtration theories, must be considered. For non-Newtonian liquids the viscosity is shear-rate dependent. For a shear thinning liquid flowing through a porous medium, the shear rate is dependent on the local pore size and local pressure gradient. Hence it follows that in order to describe the filtration kinetics, a model is needed for the pore structure. This is not necessary for Newtonian liquids. Relevant literature on non-Newtonian filtration can be found in Refs. [18,19]. When the pore size becomes very small the compact starts to behave as a filter for the macromolecules themselves. The use of macromolecular binders can then change the pore size in the green compact (see Ref. [20]). Adsorbing as well as non-adsorbing polymers influence the interaction en-
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
159
ergies between particles in suspension and can thus affect the mode of particle packing. The wet packing mode can be influenced by the structure in the suspension: agglomeration, gelation, and phase separation can occur depending on the polymer concentration and on polymer adsorption or depletion from the particle surfaces. The balance between the interaction forces and the effective stresses in a compact due to hydrodynamic drag forces during compaction determine, together with geometric factors (packing of agglomerates or phase domains; possibilities for particle rearrangements), the pore structure of the wet compact (see for example Ref. [5] or [21]).
6.2.4 Suspensions and Sols A suspension is a dispersion of solid particles in a liquid. A colloidal suspension is a sol. Colloidal properties become significant when the size of the particles is of the order of a few micrometer or less. In suspensions of large particles, for example, of some 10 ~tm or higher, hydrodynamic interactions dominate the suspension flow properties and particle packing behaviour. In colloidal suspensions interaction forces between the particles as well as hydrodynamic interactions play a role in determining the flow and particle packing properties. Solid dispersions are the vehicles in dip-coating from which the coating is formed. In the filtration coating mode, there is a transition from low to high volume fraction solids (i.e. the coating) near the substrate boundary. In this process, the flow behaviour of the suspension and that of the particulate flow units are important. In the film-coating mode, the flow behaviour of the suspension is most important. During the drying process, the suspension properties change. This affects the final microstructure of the particle packing (i.e. coating) obtained. A homogeneous particle packing can be obtained when the suspension flow units do not change on a time scale of that of the coating process. In industrial practice it is often also desirable that the suspension properties are stable on a longer time scale (shelf life) for economic reasons. This means that sedimentation and ageing cannot be permitted and must accordingly be prevented. A dispersion of solids can be prepared by dispersing a powder in a liquid or by synthesising the solid particles in situ in a liquid. Examples of the first method are suspensions of submicrometer alumina or zirconia powders in water. Examples of the second method are boehmite sols and titania sols prepared from organo-metallic precursors. 3 3 The dispersion of powders in liquids is a unit operation in colloidal processing which is very important and deservesa separate treatment. Thisis, however, beyond the scopeof this chapter. (The reader is referred to Ref. [5] or [21]for more information.) The sameis valid for the synthesis of sols (see for example Ref. [41]).
160
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
Solid particles 4 dispersed in a liquid have a two-particle collision frequency d u e to Brownian m o v e m e n t or to a shear flow field in the s u s p e n s i o n after stirring, mixing, p u m p i n g or otherwise. The initial Brownian collision rate 5 is given by R B r = 8/1;
D a n2
(6.1)
w h e r e D is the m u t u a l particle diffusion coefficient, a the particle radius a n d no the initial particle n u m b e r concentration. Note that Eq. (6.1) corresponds to second order kinetics for particles which stick u p o n collision. The collision rate d u e to shear is given by 4. Rsh - -- ]I q) n o 7~
(6.2)
w h e r e T is the shear rate and qo the solids v o l u m e fraction. Equation (6.2) gives rise to first order kinetics w h e n sticking u p o n collision is considered. For low particle concentrations (unhindered diffusion of single particles) the diffusion coefficient D is given by the Stokes-Einstein diffusion coefficient
kT
D=~ = Do 6nTI0a
(6.3)
w h e r e kT is Boltzmann's constant times absolute t e m p e r a t u r e a n d ~0 is the viscosity of the continuous phase (dispersant liquid). C o m b i n i n g Eq. (6.1), Eq. (6.2) and (p = 4 / 3 xa3no, the ratio Rsh/RBr of the two processes can be written as Rsh
4 110 a 3 -
aBr
~
(6.4)
kT
In water, this ratio is 1.2x10-4 T for 50 n m particles and 0.12 Tfor 500 n m particles. This m e a n s that shear rates d u e to p o u r i n g a suspension are already sufficient for shear rate collisions to d o m i n a t e in 500 n m particle suspensions, while in heavily stirred sols of 50 n m particles Brownian collisions b e t w e e n particles still dominate. Suppose each collision b e t w e e n two particles leads to a doublet (two-particle agglomerate). From the collision frequency follows a characteristic agglomeration time tl/2 for these sticky spheres given by Br
3TI0
~ rl~
(6.5)
tl/2 = 4kT n-------~- (p kT
Ill the treatment below suspensions with monodisperse spherical particles are considered unless specific reference is made to anisometric particles or particle size distributions. 5 More elaborate treatments of the theory in this chapter can be found in Hunter [22], Hiemenz [23] or other textbooks on colloid science.
4
161
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
TABLE 6.1 Characteristic agglomeration times TI0 = 1 mPa.s
00 = 1000 kg.m -3
T = 300 K
q) = 0.02 a = 25 nm 6X10-4 S
Br tl/2
250 nm 0.6 S
sh tl/2
rl0 = 1000 mPa.s
q0 = 0.2 2500 nm 592 S
25 nm 6X10-5 S
~/= 1 s-1
;/= 100 s -1
39 s
0.04 s
P0 = 1000 kg.m -3
a = 25 nm 0.6 S
250 nm 600 S
qo= 0.2 2500 nm ~7 days
25 nm 0.06 s
= 1 s-1 sh tl/2
agglomeration t~2-
250 nm 60 s
2500 nm 16 h
~ = 100 s-1
39 s
for Brownian
2500 nm 595 S
T = 300 K
qo= 0.02
tl/2Br
250 nm 0.065 S
0.04 s
and by
n
(6.6)
49q~
f o r s h e a r i n d u c e d a g g l o m e r a t i o n 6. T h e c h a r a c t e r i s t i c a g g l o m e r a t i o n
(tl/2)
times
f o r s e v e r a l c a s e s a r e s h o w n i n T a b l e 6.1. The characteristic time-scale for agglomeration characteristic times for the process to be performed
must be compared
with the
with the suspension
(coat-
i n g i n t h i s case). T h i s c a n b e d o n e b y u s i n g t h e s o - c a l l e d D e b o r a h n u m b e r , d e f i n e d as:
De =
De,
characteristic process time / 'observation time'
6 These formulas are in principle only valid for dilute suspensions. The forming of trimers and higher agglomerates are neglected. This later stage of the Brownian agglomeration or diffusion limited agglomeration (DLA) process gives rise to tenuous agglomerate or get structures with fractal properties. Other approximations are: diffusion coefficient independent from distance of approach, no hydrodynamic interactions; sticking force only operative at contact between particles; additivity of the shear and Brownian agglomeration process (Smoluchowski approximation). All these effects are discussed by van de Ven [24]. The purpose of the formulas shown here is to show the relative importance of the parameters which play a role.
162
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
If De >> 1, the agglomeration process does not need to be considered. The observation time in this case, for example, is the time necessary for forming a coating by colloidal filtration which may last any time from a few seconds to, say, 30 min. The reciprocal shear rates in the film-coating process are in the region of 1-100 s. Table 6.1 shows that the characteristic agglomeration time is of the order of microseconds for practical suspensions, hence De > 0.01 in most cases. Sticking spheres prove to be undesirable in coating dispersions. Because attractive interactions are omnipresent in solid dispersions (see below) repulsive interactions have to be deliberately introduced to counteract the agglomeration process. An energy barrier must be introduced reducing the sticking probability between suspended particles w h e n they collide. In this w a y kinetic m not thermodynamic m stability is usually introduced. The effectiveness of such a potential barrier in opposing agglomeration can be expressed in terms of the so-called stability ratio W defined as (see for example Refs. [22-24]). Rf W=~ Rs
(6.7)
Rf is the fast agglomeration rate in the absence of a repulsive potential. Thus, Rf equals RBr, Rsh or a combination of both depending on the conditions. Rs is the rate of slow agglomeration in the presence of a barrier. The characteristic time constants can now be found by multiplying by W. It has been shown that W can be approximately expressed as (see Ref. [25]) W = Woo + 0.25(e aG/kT) - 1)
(6.8)
where Woo is the stability ratio in the absence of repulsive interactions and AG the height of the energy barrier. In practice, the value of Woo deviates from 1 because attractive interactions and hydrodynamic interactions are still present. To protect an aqueous alumina suspension with a volume fraction of 0.2 and a particle radius of 100 nm from Brownian agglomeration in water, for a period of 2 weeks, AG should be larger than 20 kT. In real suspensions and sols several interaction forces (or interaction free energies) play a role. Some are attractive and some are repulsive. The total interaction free energy 7 is determined by the sum of all these contributions. The microstructure of the dispersion depends on the sum of the contributions. Because this sum is the total of large positive (repulsive) and large negative (attractive) contributions, the final result is a delicate balance. Prediction of the overall force between the particles requires that the separate contributions are accurately k n o w n which is often rather difficult in practice. Nevertheless, it is 7 The total interaction free enthalpy can also be expressed as an excess pressure compared to the
bulk. When the interaction free enthalpy is repulsive this pressure is called the disjoining pressure, or conjoining pressure when the interaction free energy is attractive (see Ref. [23]).
6 w P R E P A R A T I O N OF ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
163
important for practical suspension formulation to recognise the various contributions to the total interaction force a n d i f possible to estimate their relative contributions. For practical ceramic colloidal processing it is important to understand the relative importance of the interactions discussed below, how mutual combinations and combinations with other forces which occur during processing evolve. The relevant interaction Gibbs free energies are AGD, van der Waals attraction or dispersion AGH, hydrogen bond attraction AGs, solvation interaction AGel, electrostatic interaction AGpo1,polymeric interaction This results in a total interaction Gibbs free energy AGT = AGD + AGH + AGs + AGel + AGpo1
(6.8)
The attractive van der Waals interaction is always present and can be modelled. The electrostatic force and the polymeric force, primarily the most important forces in suspension formulation, can be controlled in order to give the suspension the required properties; they can also be modelled. Three cases for the total Gibbs free energy will now be dealt with. (1) AGT = AGD +--h~ + ~ ; s + AGel +~pr~l This case has been evaluated in the well known DLVO theory named after the Russians Derjaguin and Landau and the Dutch Verwey and Overbeek. (2) AGT = AGD + ~ T I + ~ s + ~ e l + AGpol (3) AGT = AGD +-~oTi + z~Crs + AGel + AGpol Solvation forces from the solvent structure near the particle surface can be attractive (hydrophobic attraction) or repulsive (hydration repulsion). Treatises on these forces can be found in Refs. [26,27] and [22]. Hydrogen bonding may be important but its effect is not yet well-known. For the time being the effect is not taken into account. If the suspension behaviour can not be predicted satisfactorily, at least not qualitatively, we suggest that the effect of solvation forces should be investigated. The origin of these forces and their effects are still being studied and possible theories are being discussed (see for example Ref. [27]).
6.2.4.1 Van der Waalsattraction The van der Waals free energy between two particles is approximately given by
AGD2- a 2A[ --+ 2a2 + ln tH-(H+ 4aH)] 6 H2+4all (H+2a)2 +2a) 2
(6.9)
164
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBY DIP-COATING
TABLE 6.2 H a m a k e r constants (derived from Tables A9.4 and A9.5, Lyklema, 1991 [29]) Substance
Harnaker A131 constant (1=substance, 3=water) in units kT at r o o m temperature (1 kT = 4.1x10 -21 J)
SiO2 (fused)
2.09
MgO
4.32
A120 3 C (diamond)
10.2 34.4
Polystyrene
2.33
A is the Hamaker constant for material (1) in medium (3) and H the separation distance between the particles (H = R - 2a, with R the centre to centre distance). The Hamaker constant depends on the dielectric and refractive properties of the particle material (1) and the dispersant medium (3) and is given by Ref. [26] 2
3kT Ir A - A131= ~ r
- r +r
] +
Shy (n2 - n2) 2 32~q~- (n 2 + n2)3/2
(6.10)
r and r are the static dielectric constants for the particle and the dispersant medium respectively; nl and n3 are the refractive indices of particle material and dispersion medium respectively; h is Planck's constant and v is the electromagnetic absorption frequency in the UV region (see also Refs. [26,28]). A131 is given in Table 6.2 for several materials in various dispersion media. We see that A and thus the attractive force is fairly large for ceramic materials i n m o s t solvents. However, it is sometimes possible almost to eliminate the van der Waals attraction by specific choice of a dispersion fluid. Equation (6.14) or more complex formulas can be used to obtain A131for new ceramic compound formulations. 6.2.4.2 Electrostatic Interaction
An oxide particle dispersed in an electrolyte solution will usually acquire a surface charge by the dissociation or protonation of surface (S) OH groups" S-MOH + H20 ~- S-MO-+H3 O+ S-MOH + H20 ~-- S-MOH 2 +OHM can be a metal cation such as A13+, Ti 4+ or Zr 4+. The surface charge depends on the pH and the electrolyte concentration. The p H where the surface charge is zero is the point of zero charge (p.z.c.) of the oxide in the medium. Because of the electroneutrality requirement the surface charge must be compensated
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
165
by a counter charge which surrounds the particles in a diffuse cloud. The charge cloud is diffuse due to thermal motion. The surface charge layer and counter charge layer form the so-called electrical double layer (ED). The effective thickness of the charge cloud is given by the Debye length ~1: K-1
r176162 11/2 ~e2 ~ ni Zi2
=!
(6.11)
where r and er are the dielectric constant of vacuum and the relative dielectric constant of the medium, respectively, ni is the number density of ion i and zi the valency of ion i in the dispersion medium. When two equally charged particles approach each other the ion clouds start to overlap and hence the osmotic pressure between the particles rises compared to the osmotic pressure of the bulk solution. This effectively amounts to a repulsion force between the particles. When the double layer overlap is small and the particles are dispersed in 1-1 electrolyte the repulsion free energy is given by AGel =
64kT na3[2e-~H K2
(6.12)
where n l = ~ ni, ~[ = tanh(e~o/4kT) and ~0 the surface potential (see for example Ref. [23]). The so-called ~ potential can be taken as a first estimate for the surface potential. The ~ potential is the electrostatic potential at the hydrodynamic shear plane close to the particle surface. It can be determined from electrophoretic mobility measurements of the particles in an electric field (see for example Ref. [23]). The ~ potential is zero when the charge within the shear plane is zero. This is the case as the surface charge plus the charge due to adsorbed ions other than hydrogen (for example A1OH2 in the case of alumina suspensions) is zero. This point is the iso-electro-point (i.e.p.) of the material in the dispersion medium. The suspension pH with respect to the i.e.p, is an important criterion for a first judgement of possible electrostatic stability. Figure 6.14 schematically depicts the van der Waals interaction free energy, the electrostatic repulsion and the combination of these two. As mentioned earlier in this discussion, the height of the energy barrier is important for stability. When the particles cross this barrier they become trapped in the primary minimum. The depth of this minimum can in principle be of the order of 100 kT or more, but in practice it is often restricted by a minimum distance of approach between the particles. This minimum distance can be determined by adsorbed (hydrated) ions, by the solvent structure near the particle surface or by surface roughness effects (see Ref. [30]). Due to these effects the minimum depth may become only a few kT making agglomerationreversible. The secondary minimum shown in Fig. 6.14 is only present in the case of larger particles. The
166
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
! 1 i o
_, energybarrier
~
o
,
~=
o "" _=
/
I
\
I
~
Total
-
secondary minimum(AGs)
,,..' m.ar, minimum (AGp)
~ \
L
I II
a
~. ~ .,-, I 'l "~., atl I
I
~
-
5 "" " " . . . .
10
distance (nm)
..""
t
I ~ van der Waals l / I t attracti~
"
r ~ S /
/
ing salt, decreasing surface potential
Fig. 6.14. Interaction energy versus distance. (a) Surfaces repel strongly: small colloidal particles remain 'stable'. (b) Surfaces come into stable equilibrium at secondary minimum if it is deep enough: colloids remain 'stable'. (c) Surfaces come into secondary minimum: colloids coagulate slowly. (d) Surfaces may remain in secondary minimum: colloids coagulate rapidly. (e) Surfaces and colloids coalesce rapidly. (After Israelachvili [26].) d i s p e r s i o n force then b e c o m e s stronger at longer distances. In the absence of shear such s u s p e n s i o n s are a l w a y s w e a k l y agglomerated. The h e i g h t of the e n e r g y barrier can be m a n i p u l a t e d in a q u e o u s s u s p e n s i o n s of oxides b y a d j u s t i n g the p H , ionic strength, the a d s o r p t i o n of c o m p l e x ions or c h a r g e d surfactants, polyelectrolytes, etc. M o r e i n f o r m a t i o n can be f o u n d in t e x t b o o k s o n colloid science s u c h as Ref. [22] a n d 'state-of-the-art' books.
6.2.4.3 Polymeric Interaction The i n t e r a c t i o n e n e r g y b e t w e e n particles in a d i s p e r s i o n can also be m a n i p u lated b y the a d d i t i o n of a d s o r b i n g or n o n - a d s o r b i n g n o n - i o n i c p o l y m e r s 8. A t l o w c o n c e n t r a t i o n , a d s o r b i n g p o l y m e r s can i n d u c e b r i d g i n g a g g l o m e r a t i o n . At h i g h e r c o n c e n t r a t i o n , w h e r e the surfaces are c o m p l e t e l y covered, there can be 8 The effect of charged polymers (polyelectrolytes) is mainly electrostatic at low electrolyte concentration, although in some cases they show charge effects and steric repulsion (see text).
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
~G
1
2
167
3
H
Fig. 6.15. Interaction free energy AGT = AGpol,st+ AGD for several layer thicknesses 15in a good solvent (X < 0.5); ~il < ~2 < ~53. a repulsion of the particle u p o n approach or an attraction depending on the properties of the adsorbed polymer layer. An approximate expression first derived by Fisher (see Ref. [32] p. 460) for the adsorbed polymer steric interaction energy AGpol,stbetween two spheres is9:
AGpol,st-43~vRT[1-X](p2[~-H)2(3a+2r HI
(6.13)
where X is the Flory-Huggins polymer interaction parameter, (P2 is the m e a n volume fraction polymer in the adsorbed layer, ~i is the thickness of the adsorbed layer and v3 is the molar volume of the solvent. This equation assumes a step function for the segment density of the adsorbed polymer and neglects elastic contributions to the interaction free energy which become important for H < 8. Despite its simplicity, it shows the significant parameters for the interaction. In a good solvent for the polymer X < 0.5 and the interaction is repulsive. In a bad solvent X > 0.5 which results in attraction. In a solvent with X = 0.5 (0 solvent) there is no mixing contribution to the free energy of interaction. When particles approach each other a repulsion occurs (in a good solvent) which is osmotic in origin, rather similar to the repulsion due to ED overlap. In Fig. 6.15 the interaction free energy distance curves are shown for several layer thicknesses adsorbed polymer in a good solvent. The weak m i n i m u m due to the van der Waals force decreases with increasing layer thickness. In Fig. 6.16, a typical interaction free energy curve is shown in the presence of van der Waals attraction, electrostatic repulsion and steric repulsion due to adsorbed polymer. Note the absence of a primary minimum. 9
See also Fleer et al. [32]. This actually is Helmholtz and not Gibbs free energy.
168
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBY DIP-COATING !
IIAGeI+AGD+AGpol,st
I I !
AG ~'~
AGpol,st
t',~~
+
AGD + AGpol,st
o
/-/
"9" ~ AGD
AGel + AGD
Fig. 6.16. Interaction energies versus the interparticle distances H for sterically stabilized particles: (a) without electrical double layer repulsion (AGT = AGD + AGpol,st); (b) with electrical double layer repulsion (AGT = AGe! AGD + AGpol,st). For comparison the curves in the absence of AGel are also plotted 9(After Pugh, Chap. 4 in Ref. [5].)
Suspensions of particles in non-adsorbing polymer solutions can weakly agglomerate due to the fact that the zone around the particles free of polymer (thickness of the order of the gyration radius of the polymer) can be reduced by agglomeration, hence decreasing the overall negative adsorption of the polymer leading to attraction (depletion agglomeration, see for example Ref. [31]). Extensive discussions of polymers at interfaces and their effect upon particle interaction can be found in Ref. [32]. The methods as discussed above for controlling the stability of colloidal ceramic dispersions are summarised in Fig. 6.17. By dimensional analysis the following dimensionless groups can be identified [25] for electrocratic (only electrostatic and dispersion forces) systems: td = 3~:Tla3 tp kT tp Pe =
(6.14)
3~rla3-------~V
11ua2
kT
6~:rla3 ~ Nf=
A G0 gr
Xr -aK
a
kT
(6.15)
A V kT
lt~/2a
V
~:0Cr ~02a kT
(6.16) (6.17) (6.18)
6-- PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING Electrostatic stabilization
169
Steric stabilization
i
s Stabilization by hydration forces
Electrosteric stabilization
""0 o
9 o
.9~ . . . ~
9O
9 o 9
9 Ojt~ 0~.
9 9
9
0
O0
~." 9 ~
~
o-
.-
O O
O-
e O
9
9 o e~/- 9 9
~
oo
o 9 OOoO 9
Stabilization
by masking van der Waalsforces
Depletion stabilization
Fig. 6.17. Methods of stabilizing colloidal ceramicparticles in liquids. (Redrawn from Pugh, Chap. 4 in Ref. [5].) Changes in microstructure of t h e suspension become important w h e n the diffusion time t d becomes long compared to the characteristic time of the process, tp. This n u m b e r has been discussed earlier as the De number. The importance of convection relative to diffusion is compared in the Peclet n u m b e r Pe (in which u is the fluid velocity). The importance of convection forces relative to the dispersion force is compared in Nf just as the dispersion force c o m p a r e d to the Brownian force. The electrical force compared to the dispersion or Brownian force is given by N r. The particle size compared to the range of the electrical force is compared in a~:. Table 6.3 gives magnitudes for these numbers for alumina particles of 100 n m and 1000 n m in water. It must be emphasised that an analysis in terms of dimensionless n u m b e r s as above neglects the different distance dependence of the forces [25]. The actual resulting force depends on the distance between the particles as has been shown. Nevertheless, such an analysis demonstrates w h a t is important in the processing of colloidal suspensions. When the average distance between colloidal particles in a suspension decreases the colloidal interactions become more important. The average distance decreases with the solids volume fraction in the suspension as [15]:
170
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
TABLE 6.3 M a g n i t u d e of d i m e n s i o n l e s s n u m b e r s for a l u m i n a particles in water. T = 300 K a = 50 n m
t d / t p (flOW at 4/= 100/s)
2.8x10 -2
Pe (filtration w i t h q = 5 ~ m / s )
3•
Nf
2.4
Nr (~l/0 = + 50 mV)
21
a ~c (electrolyte conc. = 10 --4 M)
1.6
10 4
~
10 3
.~ ~ ..
.
28
-3
0.3 2.4 214 16
Sphere diameter d / ~ n 0.05 . . . . . . 0.10 0.50 -1.00
10 5 =
a = 500 n m
10 2 101 I .,..q
10 0 10 -1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Solids volume fraction Fig. 6.18. A v e r a g e d i s t a n c e b e t w e e n particles in a r a n d o m particle d i s p e r s i o n as a f u n c t i o n of the solids v o l u m e fraction. The horizontal lines indicate the r a n g e of various particle interaction energies. ( R e d r a w n f r o m Barnes [15].)
H = 2a
1
+
/2
-1
(6.19)
In Fig. 6.18 this distance dependence is plotted for several sphere diameters. The effective ranges of several interaction forces is also indicated. We see that the mean distance becomes of the order of the interaction range at a volume fraction of about 0.2 for particles of 50 n m and about 0.5-0.6 for particles of 1000 nm. Below these volume fractions the suspension can be considered as diluted or semi-diluted and above these volume fractions as concentrated. It appears that the smaller the particle size, the lower the volume fraction for the transition to the concentrated regime. The flow and particle packing behaviour of a suspension is determined by
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
171
the effective volume fraction q0eff, (6.20) 5 is the thickness of an adsorbed polymer layer, the electrical double layer thickness, a hydration layer or a combination of these effects. Not included in this relationship is the increase of the effective volume fraction due to agglomeration. The effective volume fraction increases due to immobilisation of dispersant between the agglomerated particles. 6.2.4.4 Rheology
The viscosity of a very diluted suspension of non-interacting particles is given by the well-known Einstein equation: TI = TI0(1 + 2.5q0)
(6.21)
where 1] is the suspension viscosity and 1]0 the viscosity of the dispersant medium. The factor 2.5 is the value of the so-called intrinsic viscosity, [rl], for spheres. Its value becomes higher for anisotropic particles. Dilute dispersions of particles in low molecular mass liquids still behave as Newtonian. This means that the relationship between the shear rate y and the shear stress ~ is linear: z = T1?
(6.22)
Semi-empirical models such as Dougherty-Krieger's are useful for higher volume fractions of practical importance: Tir = rl
no-
(1
q~ /-[n]q~"~
(6.23)
where q0mis the maximum effective particle packing volume fraction. It appears that q)m corresponds to loose random packing density in the limit Pe ---> 0 and dense random packing in the limit Pe -> o,, (see Table 6.4). Van Houten [33] has shown that the maximum packing fraction obtained from the rheology of concentrated alumina suspensions is predictive for the maximum wet packing fraction that can be obtained in colloidal ceramic processing. With increasing shear, P e - the relative viscosity of s u s p e n s i o n s - usually decreases (see Fig. 6.19). This shear thinning effect is quite moderate in colloidally stable suspensions, which actually can behave as nearly Newtonian up to
172
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
TABLE 6.4 Packing density of spherical particles Type of packing
Volume fraction
Hexagonal Body centred cubic Simple cubic Tetrahedral
0.741 0.686 0.524 0.340
Random, dense Random, loose
0.64 0.59
log Tlr shear thinning
/
~176176176176 ~176176 log Pe Fig. 6.19. Typical viscosity vs shear rate curve for a concentrated suspension.
high volume fraction, but it is quite strong in agglomerated suspensions where the viscosity decreases because of shear induced disruption of the agglomerates. This difference in behaviour can be used to distinguish between colloidally stable and unstable suspensions (see for example Ref. [14]). The rheological behaviour of agglomerated suspensions depends strongly on the height of the repulsive maximum, if present, and on the depth of the (primary) minimum where the particles are trapped in the absence of a shear stress. In some cases a minimum shear stress seems to be necessary to cause flow. However, this behaviour can also be influenced by the time-scale (Denumber). The (apparent) minimum stress required is called the yield value of the suspension. This behaviour is illustrated in Fig. 6.20. At high shear rate concentrated suspensions can become shear thickening (a sort of crowding effect). This behaviour can make pumping or mixing the suspension impossible, but it does not occur in low shear dip-coating operations.
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
I
I shearstress "c(Pa)
173
2
B ----~
shear rate ~,(s-!) Fig. 6.20. Flow curves (stress-strain rate) for concentrated suspensions. In curve (1) pseudoplastic behaviour without a yield value is shown. Only an extrapolated so-called Bingham yield value can be seen (~B).Curve (2) shows non-linear plastic behaviour. An apparent yield value "r is present. Curve (3) shows the 'almost' Newtonian behaviour of a stable concentrated suspension.
shear stress 1
.
.
.
.
shear rate Fig. 6.21. Up and down (ha shear rate) in a non-thixotropic (1) and a thixotropic (2) suspension show different shear stress paths.
W h e n diffusional relaxation of a s u s p e n s i o n b r o u g h t out of e q u i l i b r i u m by shearing is slow w i t h respect to the time-scale of the process (De n u m b e r ) , the s u s p e n s i o n is said to be thixotropic. This b e h a v i o u r is illustrated in Fig. 6.21. Thixotropy is usually u n w a n t e d in ceramic m e m b r a n e s u p p o r t coatings, b u t does occur for some suspension formulations. The layer thickness obtained in film-coating with the same suspension but with a different shear history can then differ. M a n y (semi-)empirical relationships h a v e been p r o p o s e d to describe n o n N e w t o n i a n s u s p e n s i o n behaviour. For m o r e i n f o r m a t i o n the reader is referred to Ref. [15] or other textbooks on s u s p e n s i o n rheology. The particle c o m p a c t (or concentrated suspension) f o r m e d on the substrate
174
6 - - PREPARATION OF ASYMMETRICCERAMIC MEMBRANE SUPPORTS BY DIP-COATING
0.7
0.6-
.0
-
0.5
-
0.4
-
t
"
O
E
0.3
I_ 0.1 101
l, ~
102
103
_/
104
105
106
Compressive yield stress (Pa) Fig. 6.22. V o l u m e fraction v e r s u s c o m p r e s s i v e yield stress of a flocculated p o l y d i s p e r s e a l u m i n a suspension in decalin with various m a g n i t u d e s of the attractive particle interaction energy. (From Bergstr6m et al., I. Am. Ceram. Soc., 75 (1992) 3305.)
during withdrawal coating experiences a compressive stress due to the fluid flow through the compact in the colloidal filtration mode or due to capillary action during drying. This means that the compression rheology of the compact is also of importance. This behaviour has been studied by Bergstr6m [28] for agglomerated alumina suspensions in decalin for several magnitudes of the attractive particle interaction energy (see Fig. 6.22). Bergstr6m showed that agglomerated suspensions can also give dense packing if the agglomerate strength determined by the attractive minimum depth is small enough. A suggestion in this direction had also been made by Philipse et al. [34]. Until now the discussion has mainly been on the properties of monodisperse dilute dispersions. In coating dispersions used in practice this is usually not the case. For example, alumina and zirconia coating suspensions for macroporous support coatings consist of irregular particles having a log normal size distribution. This has a profound effect on the interactions between the particles and the flow behaviour of concentrated suspensions. The principles discussed above are still relevant but the consequences are much more complicated. Surface roughness
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
175
reduces the effect of the van der Waals force between the particles. The interaction becomes dependent on particle orientation; the stability ratio for smaUsmall (say 10-100 nm) and small-large particle interaction is less than for two large (say 1000 nm) particles. The reader is referred to Strauss [35,36] for more information. In concentrated suspensions many body interactions between the colloidal particles determine the effective colloid-colloid interaction. Beresford-Smith and Chan (1983) [37] showed that in that case the effective colloid-colloid interaction can nevertheless be described by an effective pair interaction energy to characterise the electrical double layer interaction. This pair interaction energy also has a screened Coulomb form just as in the classical DLVO theory but the Debye screening parameter ~cnow depends on the intrinsic counterion concentration and the concentration of added electrolyte in the system. This makes the effective pair energy dependent on the volume fraction of the particles (see general discussion of the paper of Beresford-Smith and Chan [38].
6.2.5 Drying and Sintering of Particulate Coatings Consolidation of the wet coating takes place during drying and sintering. These steps must be performed in such a way that defect formation in the form of cracks and voids is avoided. A quick drying of the wet film during film-coating can also be important in order to prevent too much drainage of the wet coating which can give rise to thickness gradients along the substrate. However, thickness gradients may also be caused by convective drying itself as was shown by Chiu et al. [39]. The main topic of this section is the phenomenology of the drying process, although the formation of the desired microstructure is, of course, also of great importance. The green microstructure is largely determined by properties of the coating dispersion such as particle size, particle shape and colloidal interactions as previously discussed. Consolidation of the microstructure by neck-formation is the goal in the sintering step. Microstructural changes here are minor compared to the changes in pore size and porosity occurring in the wet-dry layer transition. Usually the drying step starts simultaneously with the dip-coating step. In effect the withdrawn part of a substrate tube starts to dry if the atmosphere has a relative humidity below 100%. So the upper part of a tube may be dry (unsaturated) already when the last part of a tube is withdrawn from the coating suspension. So there usually exists a drying front along the tube. This front must be distinguished from that which can occur in the unsaturated part of the coating normal to the substrate due to liquid-air menisci penetrating the pore space of the coating. However, such a front may be too diffuse on the length scale of the coating thickness to be observed in practice (see Chiu et al., 1993). Dry in the
176
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
context above means that the moisture content is the equilibrium level corresponding to the bulk relative humidity of the drying air. The green structure then still contains capillary and physisorbed water. This water is removed in the low temperature part of the heat treatment applied in the sintering step. A good general treatment of drying of ceramic bodies (especially gels) is given by Scherer [40] and Brinker and Scherer [41]. The reader is referred to this for an introduction to drying processes. The drying of particulate coatings of submicrometer particles was recently investigated by Chiu et al. [39]. Some of their findings will be discussed later in this section. In part II of his drying article series Scherer develops the theory of drying of thin films [42]. Although he discusses sol-gel films in particular, the theory is also of relevance to submicrometer particulate films. The drying and sintering of sol-gel films (mesoporous and microporous coatings) is discussed in Chapter 8. The drying of sol-gel films during dip-coating is discussed by Hurd and Brinker [43]. The drying of a particulate layer on a substrate occurs in several stages and this can also occur in the drying of bulk particulate materials. The first drying stage is the so-called constant drying rate period (CRP). During this period the drying rate is close to that of the pure dispersion liquid under comparable conditions of partial vapour pressure (relative humidity), temperature and vapour velocity along the evaporating surface. During this stage drying stresses in the layer start to build up as soon as the particles in the layer form a (visco)elastic microstructure. This microstructure can be that of strongly repulsive colloid particles in a stable concentrated dispersion or that of a coagulation or flocculation structure (gel network). If the particulate coating is still a diluted dispersion, as can be the case with a filmcoat layer, the shrinkage of the layer normal to the substrate surface is rather trivial the consequence of the evaporation of the dispersion liquid. But as soon as the particle microstructure becomes visco-elastic the liquid surface cannot any longer be flat on the scale of the particles and a tensile stress in the liquid develops. The tensile stress in the liquid is compensated by a compressive stress on the particle microstructure in the case of a free film, which therefore starts to shrink as much as is necessary to ensure complete wetting of the solid particle structure. The tensile stress is adapted to the compliancy of the particle microstructure. In the case of a coating on a substrate with a high friction between the wet coating and the substrate (strong adherence) there can be no strain in the substrate-coating interface and consequentially the stress in the substrate is compressive and tensile in the film parallel to the substrate. Shrinkage can only take place in the direction normal to the substrate. The shrinkage of particulate coatings which are already close packed in the wet stage amounts to 2-10%. This is much less than the shrinkage observed for sol-gel films or particulate coatings formed from flocculated suspensions. The CRP comes to an end and the FRP1 starts if the radius of
6 - - P R E P A R A T I O N OF A S Y M M E T R I C C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
177
curvature of the menisci becomes equal to the pore radius. Then menisci start to penetrate into the layer which then also ceases to shrink. As previously mentioned Chiu et al. [39] argue that on the scale of the coating thickness the menisci front must be very diffuse. The tensile stress in the coating stays the same in the direction normal to the substrate. The transition between the CRP and the FRP1 is called the critical point. The second falling rate period, FRP2, where the decrease of the moisture content of the film is determined by vapour diffusion to the surface is probably not important for the microstructure development or defect formation in particulate coatings. The tensile stress developing in the solid microstructure of the coating can cause cracking of the film. Because this stress increases with the square of the pore size of the microstructure it becomes clear that support coatings become increasingly more susceptible to cracking with decreasing pore sizes. It is our experience that particulate coatings from submicron powders applied by film-coating do not crack upon drying in air (relative humidity 20-70%). The reason for this is that the optimum film thickness is below the critical value. The reason for the existence of a critical cracking thickness (CCT) is that cracking only occurs as the energy required to form a crack is less than the energy release upon relieving the strain in the film (see Refs. [39,44,46]). In this respect the drying process is very similar to the sintering of the coating where tensile stresses due to sinter shrinkage can cause cracking of the coating. The magnitude of the tensile stress in the coating and the fracture resistance Kic of the coating material determines the CCT (hc) as [39]:
hc
[1.4r~)
(6.24)
The CCT can be increased by increasing the Kic or by decreasing the tensile stress ~ in the film. Chiu et al. [39] found experimentally that CCT increases with increasing particle size, or eliminating the electrostatic double layer between the particles in the coating before drying. The first decreases the drying stress and the second increases the fracture strength of the coating. Coatings from materials with higher Hamaker constants but otherwise similar also appear to have a higher CCT because the fracture resistance is higher. Aging of alumina coatings at low pH (2) before drying enhances recrystallisation of oxide material in the particle necks and increases the fracture strength and thus the CCT. Chiu et al. [39] found that the CCT increased linearly with the PVA binder concentration in the coating. They attribute this to an increase in the fracture resistance of the coating. However, stress measurements on boehmite sol-gel coatings on an alumina substrate showed that the decrease in cracking tendency is due to stress relaxation due to the PVA added in that case [3]. It seems likely that this is also the case with ~-alumina coatings. For a further discussion of the strength of granular films the reader is referred to Ref. [39b].
178
6 w P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
In practice, the sintering of dried macroporous coatings for the purpose of support preparation is not difficult compared to the sintering of sol-gel coatings (see Chapter 8). Due to the tensile stress in the constrained sinter process there is a tendency to more open structures compared to the sintering of bulk material at least at higher sinter temperatures [45]. Sinter stress related damage does not occur in ~-alumina coatings on 0~-alumina substrates at the moderate sinter temperatures used (1100-1250~ for consolidation. Of course, stress concentrations on inhomogeneities should be omitted. The reader is referred to Refs. [46-49] for further reading on the subject of constrained sintering.
6.2.6
Defects
A defect in a porous layer on a porous support is a microstructural or textural feature which hampers application of a defect-free functional membrane layer. Defects are cracks or micro-cracks in the substrate layer, irregularities in surface roughness, pinholes or voids percolating the layer or large percolating pores as a result of the particle packing process. These last defects are not really defects because they are an unavoidable result of the particle size distribution in the dispersion and random packing. The size of defects considered to be significant depends on the function of the membrane layer. In a substrate for a UF membrane larger defects and a greater density of defects can be tolerated than in a support for a gas separation membrane [4]. As a rule, defects in a support layer which are of the same size or thickness as the next layer 'transfer' to the next support layer or to the membrane coating. Smaller defects can often be repaired by applying another layer on top with the same or a somewhat smaller pore size distribution. Without special precautions, several types of defects may appear in dispersion coatings on porous substrates. Insufficient de-aeration of dispersions for film-coating can lead to voids as a result of bubbles in the coating (see Fig. 6.23). These bubbles can cause pinholes in the layer depending on the layer thickness to bubble diameter ratio and the wetting properties of the substrate. Figure 6.24 shows a SEM photograph of a pinhole in a second layer which was probably produced by a large surface void in the rather heterogeneous commercial substrate used. There are also cracks in this coating. These cracks are correlated with the pinholes. Stress build-up due to the difference in expansion coefficient between the alumina coating and the substrate also plays a role here. Figure 6.25 shows a SEM photograph of cracks in a filmcoat layer due to differences in thermal expansion. Coatings which are too thick can have crack patterns due to drying shrinkage tensile stresses (see Section 6.2.5). Although the crack width is often very small, application of a mesoporous alumina coating with a thickness of a few ~tm by colloidal filtration of a boehmite sol proves to be cumbersome
179
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
'?~
~,.~
~"
,
'~
,
~.~
,
~" ~ . ' ' ~ "
.'..'~
.
,,,,.
.....
,r
~
..
~
,'~
~
~
,
",.~
.~'
~.'~~
'
..
~.-
',
Fig. 6.23. Void due to poor de-aeration of coating suspension. SEM picture of fracture surface of layer 2 coating.
Fig. 6.24. Two pinholes. SEM picture of coating surface layer 2 coating obtained by film-coating of an electrostatically stabilised alumina suspension.
as shown in Fig. 6.26. Where cracks exist, no boehmite layer is formed. Moreover, the layer shows pinholes. Pinholes in layer 3/4 sol-gel coatings may be caused by large voids in the support underneath and by dewetting problems or by both. The example in Fig. 6.27 indicates that the pinholes are probably caused by large voids in the substrate, where no slip casting of the sol has taken place. These voids in turn are caused by the inhomogeneous irregular tubes which were used causing defects in layer 2. The 'transferred' defects largely disappear
180
6 ~
PREPARATION
OF ASYMMETRIC
CERAMIC
MEMBRANE
SUPPORTS
BY DIP-COATING
Fig. 6.25. Cracks due to the difference in thermal expansion coefficient of substrate and coating. SEM picture of layer 2 coating.
. :~,,:v.:~
.
,,.
~, 9... , ~ @ . ~ . .
".
..~
9. ~.....
e q~
Fig. 6.26. Attempt to apply a mesoporous ~-A1203 coating on a layer 2 substrate with cracks by capillary colloidal filtration of a boehmite sol without macromolecular additives. In the layer 2 crack regions no boehmite coating develops. The layer 3 coating shows pinholes (SEM picture).
6 ~ PREPARATION OF ASYMMETRICCERAMIC MEMBRANE SUPPORTS BY DIP-COATING
181
Fig. 6.27. Pinholes in ~,-A1203 coating (layer 4) on a 3-layer substrate. The pinholes are related to voids in the layer 3 substrate surface (SEM picture).
when more homogeneous full alumina support tubes are applied. Suitable tubes are a prerequisite for successful development of layered support systems for Knudsen/UF and microporous gas separation membrane tubes. When the thickness of the layers to be applied decreases below 5 ~tm and the pore size below 100 nm, dust and other foreign particulates in the air or coating fluid may cause defects. However, it is difficult to prove which defects are the result and what is the resulting defect density. Preventive measures should at least be taken such as substrate cleaning and filtration of coating liquids. Defects which may occur in sol-gel coatings (support layer 4) applied by colloidal filtration on an alumina substrate surface with a mean pore size of 100-200 nm will be discussed in more detail. Firstly, the filtration process on a surface as in Fig. 6.7b is considered. The process is performed by using an alkoxide boehmite sol without any polymeric or surfactant additives. On such a substrate the minimum layer thickness (optimum permeability) is limited by the inhomogeneities in the substrate as shown in Fig. 6.28. On patches where the substrate pore size is larger than average no cake filtration takes place for short filtration times. When the layer becomes thicker, the layer still shows pinholes which could be remnants of the earlier bare patches centred around the largest hole in the patch. The pinholes and dry patches in a similar coating as shown in Fig. 6.29 are probably correlated with dewetting of the wet sol-gel coating before the layer solidifies due to drying. Addition of a thickener to the sol slows the dewetting kinetics (see Section 6.3.2.4) and prevents pinhole forming. Figure 6.30 shows an example of a largely defect-free coating. The surface roughness of the substrate (layer 3) is smoothed by layer 4, the sol-gel
182
6 ~ PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
Fig. 6.28. Layer 3 substrate surface (shown in Fig. 6.7) partially covered with a thin ~-A1203coating. Coating process: capillary filtration with a boehmite sol without surfactants or macromolecular additives (SEM picture).
Fig. 6.29. Dewetting problem during drying after capillary filtration causes the uncoated regions. SEM picture of surface of layer 4 ~,-A1203coating on layer 3 (~-A1203substrate.
layer. W h e n the s u p p o r t s h o w n in Fig. 6.7a is u s e d as a basis, there are m a n y f e w e r p r o b l e m s w i t h pinholes (in fact such a slip cast layer can be a p p l i e d w i t h o u t u s i n g thickeners in that case) a n d the surface is m u c h s m o o t h e r . The surface r o u g h n e s s is t h e n c o m p l e t e l y d e t e r m i n e d b y the g r a i n size of the coating ( a p p r o x i m a t e l y 100 nm).
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
nnn
- -
9
1~,
_
_
183
.
Fig. 6.30. Nucleation and growth of pinholes during drying is suppressed by the addition of polymer thickener to the boehmite coating sol. SEM picture of the surface of a largely defect-free ~'-A1203 coating (thickness ~ 3 ~tm) on the substrate shown in Fig. 6.7b (3 layers).
6.3 DIP-COATING WITH POROUS SUBSTRATES
6.3.1 Capillary Colloidal Filtration Particles f r o m a d i s p e r s i o n can be c o n v e c t e d to the i n n e r or o u t e r s u r f a c e 1~ of a p o r o u s s u b s t r a t e in c o n t a c t w i t h the d i s p e r s i o n d u e to fluid flow t h r o u g h the p o r o u s s u p p o r t . Also b o d y forces d u e to c e n t r i f u g a l or electric fields can, in p r i n c i p l e , be u s e d to assist the particle t r a n s p o r t t o w a r d s the s u b s t r a t e . W h e n the s u p p o r t is n o t p e r m e a b l e for the particles in the d i s p e r s i o n , the p a r t i c l e t r a n s p o r t r e s u l t s in a m o r e or less d e n s e particle c o m p a c t 11. T h e g r a v i t a t i o n a l force on the particles can also c o n t r i b u t e to the particle p a c k i n g p r o c e s s w h e n the g r a v i t a t i o n a l force is in the s a m e d i r e c t i o n as the fluid flow. The d r i v i n g force b e h i n d the fluid flow can be the c a p i l l a r y s u c t i o n p r e s s u r e of the s u p p o r t or a n e x t e r n a l a p p l i e d p r e s s u r e . In the f o r m e r case the p r o c e s s is e q u i v a l e n t to the slip c a s t i n g p r o c e s s in p l a s t e r m o u l d s w e l l - k n o w n in c e r a m i c s , 10 From a topological point of view an open porous medium has only one very complicated surface. Here we mean by outer or inner surface the geometrical surfaces on a macroscale of the outside and inside of single or multihole tubes, respectively. The definition of these surfaces on a microscale is arbitrary to some extent and depends on the yardstick used. 11 We use the term compact in the broad sense. Both a concerftrated dispersion near close packing but with overall repulsion between the particles is called a wet compact as well as the case were the particles are trapped in an energy minimum. In the latter case the relaxation time for particle breakup from the compact, i.e. due to diffusion over an energy barrier (activated diffusion), can be much longer than the time-scale of practical interest and the particle compact is in fact consolidated.
184
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
in the latter case the process is equivalent to pressure filtration, a well-known operation in chemical engineering. When the particles in the dispersion show appreciable Brownian movement, diffusion of the particles starts to play a role in the filtration process. If in that case the state of the compact is in effect that of a very concentrated dispersion (particles not in a (primary) minimum), the osmotic pressure difference between compact and dispersion will lead to back diffusion of the particles. The compaction process can then best be modelled starting from the convective diffusion equation (see Ref. [25]). However, the following sections will mainly consider classical capillary filtration (slipcasting) theory as treated amongst others by Leenaars [50,51] and Tiller [52,53]. The role of diffusion will be only briefly mentioned. Leenaars was the first to apply slip cast theory to describe the forming of mesoporous ceramic ultrafiltration membranes by dip-coating of porous supports in a sol. Firstly, the geometry of the support must be considered. The radial flow through a cylindrical support is quite different from the unidirectional flow encountered in flat supports. Fortunately, in most cases the compact layer thickness is small compared to the curvature radius of the support surface. The filtration process can then approximately be described as one-dimensional filtration [54].
6.3.1.1 Continuum description Secondly, the unidirectional compact growth geometry must be considered as shown in Fig. 6.31. The velocity v of the compact/suspension boundary must be defined as dLc v - dt
(6.25)
where Lc is the compact thickness. For an observer on this boundary the particle flux in the compact is stationary 12 and given by particle flux =-Vq0c
(6.26)
q0cis the volume fraction in the cake, which is considered as incompressible. In the suspension particle flux = - ( v + q)q00
(6.27)
where q is the superficial fluid velocity and q00the solids volume fraction in the suspension. So, -Vq0c = - ( v + q)q00
12
1am grateful to Professor A.P. Philipse for drawing my attention to this approach.
(6.28)
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
185
x o
o
o
o o
o
o
o
suspension 9o
o
o
o
0 0
0
o
0
0
0 o
0
x= Lc 0o o~ o 0 ,,, o .., 0 o c %00o00000 compact(cake) ~0o~ 0 0 0 P _o _ o (Pc Oo o^o o ,OoO o , OoO 9
NNNNNNNNNNN~I::.'..e'..'...'.~!~4~~
~ ~
"~'.-?-~.~
~
:
~ , . . . , . : ~~ ~9~ ~ , . _ .~_.~ : z, ~.~:~ . ~ :
~~i~:r
',
Fig. 6.31. Unidirectional compact growth in capillary colloidal filtration. Hence, (I)0
v- ~ . q q0c- 90
(6.29)
(Pc - q)0 dLc q=~ . % dt
(6.30)
and
Darcy's law for the flow through a porous medium reads, in differential form (see for example Ref.[55]): q-q
k dP: dx-
k dPs 1] dx
(6.31)
where P1 and Ps are the local liquid and solid pressure in the compact, respectively (see Fig. 6.32), and k the local permeability (m 2) of the compact. Integrating over the compact thickness gives: 0
APc
= -
=
AP c
=
(6.32)
0
where Kc is the average permeability (m 2) of the compact. The superficial velocity q is the same in the support and the compact, hence K1
Kr
qq=-~l AVl=-~cAVc
(6.33)
186
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
4-Ll
...... x ........ ' ~
Lc
Fig. 6.32. C a p i l l a r y c o l l o i d a l f i l t r a t i o n (after Tiller a n d T s a i [52]).
K1 is the permeability of the substrate and L1 is the penetration depth of the liquid in the substrate and AP1 the corresponding pressure difference. Further the capillary suction pressure of the support AP is AP = AP ~ + AP c
(6.34)
Solving for APc gives: APc =
AP _ 1 + KL1/K1 Lc
(6.35)
From a volume balance it follows that L1 _ (Pc/%- 1 Lc el
(6.36)
where r is the porosity of the substrate. Substituting Eq. (6.36) in Eq. (6.35) gives AP
-
(6.37)
1 + Kle~
So the pressure drop across the compact (cake) is constant in time, in contrast to external pressure filtration cases. This was previously observed by Leenaars [50,51]. This expression can now be used for APc in Eq. (6.33) and Eq. (6.30) can be used for q to obtain dLc Lc dt
_
%
Kc
AP
(Pc (P0 11 1 + Kit;1
(;01/
(6.38)
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
187
After integration this equation yields L2 =
2AP t 1]
-1
+
=2Ct -1
(6.39)
/~1K1
See also Ref. [52]. A similar equation was derived by Leenaars [50,51]. It appears that for the capillary filtration process the thickness of the compact is proportional to the square root of the contact time substrate/suspension. As shown, the constant C depends on the properties of the support and compact: C = f (porosity, permeability, capillary pressure, suspension solids volume fraction, compact solids volume fraction) Prediction of the layer growth kinetics of support layers to be prepared from (sub)micrometer suspensions and sols is, in principle, possible on the basis of Eq. (6.39). Usually, the average pore properties of the substrate, ~1 and K1 are known or can be rather easily determined. Estimation of the effective volume fraction of sols, however, may be more difficult as the particles become smaller (see Section 6.2.4). The permeability of a porous medium can be obtained with the CarmanKozeny equation given by: E3
K=
(1 - ~)2 k0 k~ S20
(6.40)
where So is the specific surface of the porous medium per volume solids, k0 is a particle shape factor and k~ accounts for the tortuosity of the porous medium. For many particle packings such as exist in practice kok, appears to be approximately 5. A useful discussion of particle packings and the Carman-Kozeny equation can be found in [56]. Estimation of the mean permeability of the wet compact on the basis of the Carman-Kozeny equation is not always straightforward, because the porosity of the packing is not a priori known and also the specific surface area experienced by the flow is not necessarily the same as that determined by gas adsorption measurements. Despite these difficulties, it is worthwhile making first estimates of the growth kinetics using the above theory. This will be shown in Section 6.4. Equation (6.39) is only valid when the wetting front in the substrate is smooth, which depends on the (pore) homogeneity of the substrate. Tiller and Tsai [52] showed that there is an optimum pore size of the substrate which produces the maximum pressure drop across the cake. This is shown by Eqn. (6.37). A smaller pore size gives a larger capillary pressure AP, but also a smaller substrate permeability K1. As a consequence, local differences in growth kinetics may arise, which limit the minimum layer thickness and are a source of defects.
188
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
The smaller the particles size the more important the particle diffusion becomes. In this respect the convective Pe is important (see Section 6.2.4):
Pe=
6~'rla2u k---T--
(6.41)
u is given by u=
K
APe 9~ n(1-(Pc) Lc
(6.42)
For Pe >> 1, Brownian movement is not important. Back diffusion from the particle compact plays a role for Pe << 1. This case will eventually result in a stationary state compact thickness, but only if the cake stays mobile, i.e. the particles are not trapped in an attractive minimum. Now consider the presence of a DLVO interaction energy between the particles. Due to the compressive pressure on the particles in the compact during filtration (see Fig. 6.32), the repulsion barrier can be crossed if the compressive force becomes high enough. If a DLVO interaction energy is present between the particles, the repulsion barrier can be crossed due to the compressive pressure on the particles in the compact during filtration, if the compressive force becomes high enough. The particles then become trapped in the primary minimum. The compressive force increases in the direction of the substrate. Suppose the cake is to be built of layers with thickness 2a. The ratio n between the electrical force and the force between two particles due to hydrodynamic drag will be a measure for the position in the compact where the transition between mobile and immobile compact occurs. This transition point is given by Lira = Lc- 2 a n
(6.43)
and n is given by [57]: n=
~o ~ ~2o 6~ rlu f(q0c)
(6.44)
f(q0c) is a function of the pore properties of the compact, which can be described by the Carman-Kozeny equation. The immobile part of the cake grows also according to the qt- law. The mobile part of the cake is unstable [57]. Due to Brownian diffusion or shear induced diffusion the mobile part can 'dissolve' when convection to the surface stops. The extent of the mobile part of the cake depends on the stability ratio W. If W is too high this can result in bad coating behaviour. For example, a large part of the coating can 'disappear' after draining the coating suspension from a substrate tube.
6 - - P R E P A R A T I O N O F ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
189
6.3.2 Film-coating The film-coating mode of dip-coating is defined by Rushak [58] as "a fluid flow that is useful for covering a large surface area with one or more thin, uniform liquid layers". It is important to distinguish between the wet coating thickness and the thickness of the coating after drying. Film-coating by dipcoating is a so-called self-metering coating flow. This means that the wet coating thickness is a dependent variable [58]. In pre-metered coating flows, the coating thickness can be varied as an independent variable within certain limits [58]. The coating thickness is then the feed rate suspension per unit width divided by the substrate speed [59]. The wet coating thickness is here independent of the rheological properties of the coating dispersion. The difference between self-metered film-coating and pre-metered film-coating is shown in Fig. 6.33. Although the pre-metered situation is obviously preferable, it is difficult to accomplish in ceramic membrane support manufacturing, because the precision needed for slot coating with seal cannot be achieved due to irregularities in the support, which are absent, for example, during the preparation of magnetic tapes with smooth polymeric substrates. The following sections will therefore only deal with self-metering coating flows. The dry thickness is directly related to the wet thickness and always depends on suspension properties such as volume fraction solids and colloidal stability. In this respect self-metered and pre-metered coating flows do not differ. The self-metered process is much more complex, because the wet thickness also depends on the suspension properties. The coating of substrate tubes is a batch process. This means that there is a start and finish to the film-coating of each tube. However, for tubes of sufficient
-~:
a. Slotcoating
~:-~.~.e ~.~.~, ~::u .:
b. Slotcoating with' seal'
--.;-
__
==
. . - . . .. .. . . . . . . . . . . . . . . . . . . . . . . . ~
_=_ -=__
d.
Fig.6.33.Pre-meteredfilm-coatingflows(a) and (b) and self-meteredcoatingflows(afterRushak [58]).
190
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
length (> 50 cm), a large part of the tube experiences a steady state during the film-coating process. When withdrawal is stopped, there is no further lifting of the liquid while gravity draining of the coating is still proceeding. This can cause a gradient in film thickness along the tube, if the coating is not already sufficiently solidified due to drying or high viscosity at the low shear stresses than present. The coating thickness depends on six competing forces which operate in the fluid entrainment region (see also Ref. [59]): 1. the upward viscous drag force on the liquid 2. the downward gravity force on the liquid 3. the Laplace pressure due to the curved meniscus 4. the inertial force of the boundary layer liquid arriving at the deposition region 5. the force due to surface tension gradients 6. the disjoining or conjoining pressure of the thin film. Inertial forces can usually be neglected (lubrication approximation). The disjoining pressure becomes important for films with a thickness smaller than 1000 nm. This is unusual for membrane substrate coatings. The disjoining pressure, however, is also important for thicker films as it determines the wet film stability (see the end of this section). The force due to surface tension gradients along the entrained film are often important in practice but difficult to quantify. A surface tension gradient makes in effect the liquid air interface 'rigid', thus slowing down liquid drainage in the film which causes thicker coatings under otherwise similar conditions. Surface tension gradients occur in the case of mixed dispersion liquids if differential evaporation occurs or if surfactants or polymers are present in the suspension. The surface tension of the coating then increases due to surface depletion of adsorbed species caused by the surface expansion occurring [60].
6.3.2.1 Coating Flow Dynamics In the simplest case the stationary film thickness is a function of the withdrawal speed Vw, the dynamic viscosity 11, specific gravity pg and the surface tension 7. So,
-fi= h(vw,rl,pg,~D
(6.45)
To express these five variables only three fundamental units are needed. According to Buckingham's ~ theorem these variables can be combined in 5-3 = 2 dimensionless parameters (see for example Ref. [61]) as follows
hl~vwT pg )
=f~~)
(6.46)
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
191
with I~V w Nst =
(6.47)
pgh 2
the Stokes number comparing the viscous and gravity force and Ca-
TlVw
(6.48)
T
the capillary number comparing the viscous and surface tension force. At high withdrawal speed (Ca >> 1) the coating film is independent from the nature of the static meniscus, hence - I q Vw~l/2 = constant
pgj
(6.49)
h = k ~ Pg
(6.50)
or
Derjaguin showed the constant to be unity. For an arbitrary withdrawal speed expression (6.46) can be rewritten [62] as h = t--~-J
f
(6.51)
The function f can be obtained by a detailed analysis of the fluid dynamics of the coating flow.
6.3.2.2 Closer Examination
Figure 6.34 schematically depicts a diagram of the steady state film-coating process, vw is the vertical withdrawal speed from the suspension, which is here considered to be Newtonian. The liquid entrained by the substrate forms a hydrodynamic boundary layer splitting in two parts: one forming the coating and one returning to the suspension pool. According to Scriven [59] the coating thickness obtained is related to the location of the dividing streamline between the two flows. The fluid meniscus in Fig. 6.35 can be considered as consisting of two parts: an undisturbed static meniscus far from the substrate and a dynamic part in the fluid flow region. There must be a smooth transition between the two. The matching of the two gives an important boundary condition for solving fluid flow equations for the film-coating problem. This point will return later in the discussion.
192
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
Vw h
1 Fig. 6.34. Streamlines in film-coating flow of a Newtonian liquid (after Scriven [59]).
X
Region 1" Lubrication film
T k'w
Region 2: Dynamic meniscus Region 3: Static meniscus
Fig. 6.35. Steady-state film-coating of a flat substrate. In region I the flow is nearly unidirectional. Region 2 is the transition region to the static meniscus region 3. (After van Rossum [61].)
T h e s t a r t i n g p o i n t for the d e s c r i p t i o n of t h e fluid d y n a m i c s of t h e film-coati n g p r o c e s s is the N a v i e r - S t o k e s e q u a t i o n a n d the c o n t i n u i t y e q u a t i o n . T h e N a v i e r - S t o k e s e q u a t i o n reads:
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
p [~3 - ~~+ ~ . V
~ , //= - V P + r
IV2 ~
193
(6.52)
where ~ i s the fluid velocity vector, p the fluid density, 11the fluid viscosity and p the dynamic pressure defined as p=P +~
(6.53)
and 9 is given by pg = - V ~ (see for example Ref.[25]). The continuity equation is 3p ~+ 3t
V. (p ~ = 0
(6.54)
When the density is constant this condition reduces to (6.55)
v . u--*-o The Reynolds number Re is defined as
(6.56)
R e - p uL q
u and L are the characteristic velocity and length scales for the process under consideration. In withdrawal coating u = Vwand L = h. Under conditions of low Reynolds number (slow viscous flow), the NavierStokes equations are reduced to the Stokes or creeping-flow equations given by: V. ~ = 0;
V p - TI V2 ~
(6.57)
TO proceed, a specific coating geometry and the appropriate boundary conditions must first be considered. It is not possible to solve the equations for the whole coating flow. Solutions for the two regions indicated in Fig. 6.35 have to be found separately. The coating of cylindrical surfaces can be approximated by an infinite flat plate and the film as a (nearly) parallel liquid layer if the film thickness is small compared to the radius of the substrate. In this respect the Goucher number, Go, is important. tube radius tpg/1/2 Go = capillary length - R 21,
(6.58)
For Go > 3 the flat plate approximation is valid [63] and needs further consideration. In the lubrication film region the flow is nearly unidirectional and hence Uz = Uy = 0 m/s. Further, in the stationary state 3Ux ~<< 3x
3Ux .... 3y
(6.59)
194
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
It then follows that the Stokes equations reduce to:
3P
~=rl r)x
32Ux
(6.60)
0y 2
3P -0 3y ~U x
3x
(6.61)
-0
(6.62)
The pressure in the film is constant and m u s t be equal to the capillary pressure:
Pcap = 7
+ ~ rmax
(6.63)
a n d r m i n are the radii of curvature of a curved surface, can be a p p r o x i m a t e d by [64]:
rmax
1
d2h ~~ rmin dx 2
r m a x = oo 13
and
Ymin
(6.64)
Then 3Pcap _ ~ d3 h 0----~ - - ~/dx 3
(6.65)
Hence d3h 32 Ux + rl + pg = 0 ~/ dx 3 ~)y2
(6.66)
This differential equation describes the coating flow in the lubrication regime. Solutions can be obtained by taking into account the correct b o u n d a r y conditions (4). Matching the lubrication region with the static meniscus region is the most difficult of these conditions (see for example Refs. [62] or [64]). L a n d a u and Levich solved Eq. (6.66) without the gravity term pg. The solution is then a competition b e t w e e n viscous drag and surface tension forces (see for example Ref. [62] or [59]). They obtained h = 0.944 11 Vw /6
1,I Vw
(6.67)
L Pg This shows that the thickness of the film is proportional to Vgw/3a n d 112/3. The coating thickness is only a w e a k function of the surface tension. Equation (6.67) 13
For a small cylinder radius this approximation is not valid, resulting in a different expression for the pressure.
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
195
appears to be valid for Ca < 0.01. The most important physical aspects of the problem have now been considered. Several attempts have been made to derive analytical formulas for the intermediate Ca region where surface tension, gravity and viscous forces are important. For further information, the reader is referred to Refs. [63] and [64]. Although the coating of long tubes is a stationary process, the coating process stops at a given time. From then, no further viscous lift occurs but drainage of the coating continues. To prevent this, suspensions can be used with a yield value high enough to stop drainage (in the order of I Pa), or else drying of the coating during the coating operation must be enhanced to ensure a consolidated green coating when the withdrawal speed drops to zero. Many attempts have been made to obtain (semi-)analytical descriptions for non-Newtonian coating flows. These are necessarily approximate and the approximations made to obtain tractable mathematics are sometimes non-physical [58]. These models do not predict the coating behaviour very well from the rheological parameters. The thickness is usually considerably overestimated. It seems more advantageous to simulate non-Newtonian coating flows by computational fluid dynamic methods (see also Ref. [58]). A problem especially present with coarse suspensions is that due to the shear stress gradient across the film during coating shear induced diffusion [65] of particles occurs away from the substrate surface. This results in a volume fraction gradient which in turn gives rise to differences in rheological properties across the film. The relatively low viscosity at the substrate surface could result in a lower film thickness than that expected on the basis of the rheology of the homogeneous suspension [66].
6.3.2.3 S ubstrate Wetting and Dewetting Wetting phenomena are especially important in film-coating when the capillary action of the substrate is suppressed by changing the high energy (solid surface tension higher than that of the coating liquid) oxide surface into a low energy hydrophobic surface by the action, for example, of an adsorbed surfactant or a silane coupling agent. The discussion below is restricted to the behaviour of low energy substrates. Much of the information in this section stems from Refs. [67] and [68]. The chapter by Blake 'Dynamic Contact Angles and Wetting Kinetics' and that of Kistler: 'Hydrodynamics of Wetting' in Berg's book [67] are especially relevant for a more complete understanding of wetting aspects in film-coating and the interested reader is referred to these. The equilibrium situation which exists after a liquid drop is brought into contact with a smooth homogeneous substrate depends on the balance between adhesion and cohesion free energies involved. Three cases can be identified:
196
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
.
.
.
.
;~
.
.
.
.
.
C
partial wetting
complete non-wetting
a complete wetting Fig. 6.36. W e t t i n g , n o n - w e t t i n g , p a r t i a l w e t t i n g .
(a) partial wetting as depicted in Fig. 6.36; the fluid drop makes a contact angle 0 with the surface; the contact or wetting line is the circle on the substrate where the solid, liquid and vapour phases meet; (b) complete non-wetting (Fig. 6.36): the contact angle is 180~ (c) complete wetting (Fig. 6.36)" the contact angle is 0 ~ The contact angle 0 above is the static or equilibrium contact angle and is a thermodynamically defined quantity as we will se e below. The static contact angle is given by the Young ('Gibbs) equation (see for example Ref. [69])" cos 0 =
/1;SL -- ?~SV
(6.68)
7LV where 7LV is the liquid/vapour surface tension, nsv and/I~SL are the spreading pressure of the vapour on the solid and liquid on the solid given by nsv = 7 s - 7sv
(6.69)
/1;SL ----ItS -- ]tSL
(6.70)
and
where Ts is the surface tension of the solid in vacuum and ]r and Tsv the solid/liquid and solid/vapour interfacial free energy per unit surface area, respectively. We see that the contact angle is determined by the interaction of the liquid with the solid (riSK- ~SV).Only the difference between Tsv and ]r is of importance.
197
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
The difference between Ys and ~SLis due to vapour adsorption on the surface: P0
~sv = RT
~
Fsv dP
(6.71)
0
where Fsv is the adsorbed amount of vapour. Similarly, in the case of wetting by a surfactant solution instead of pure liquid ~SL is given by c
/~SL = RT ~ FSL d In c
(6.72)
0
where FSL is the adsorbed amount of surfactant and c the surfactant concentration. The reversible work of adhesion Wad between the solid and liquid is defined by: W a d = '~/SV + '~LV -- ~/SL
(6.73)
It is the work necessary to separate a unit area of liquid from a solid. Hence Wad = -
AGad
(6.74)
With Eq. (6.68) we can also write Wad - 7LV(1 + COS0)
(6.75)
We see that Wad i8 only zero when 0 is 180 ~ The reversible work of cohesion Wcohis: Wcoh = 27Lv
(6.76)
N o w we can write the contact angle in terms of Wad and Wcohas cos 0 -
2 Wad Wcoh
- 1
(6.77)
Another important wetting parameter is the 'work of wetting' [70] or adhesion tension Ww defined as: (6.78)
W w = ~ S L -- ~ S V -- ~/LV COS 0
Ww corresponds to the free energy change connected with the withdrawal of a substrate surface from a liquid pool (see Fig. 6.37 or [70]) or by the expulsion of liquid from a pore. The third wetting parameter is defined as W s "- ~/sv - '~LV - ~SL "- ~/LV (COS 0 --
1)
(6.79)
Ws is the free energy of dewetting (retraction) of a liquid thin film from a solid surface. Usually Ws is called the spreading parameter.
198
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
Wad -- TSV+TLV-TSL
ILl
]c I ilii
Iii!!iiii!l
Ww = ~ V - ~ L
.K L Ws - 7SV-~V-TSL
I
S
I "-~ I
S
1
Fig. 6.37. T h e r m o d y n a m i c e x p r e s s i o n s for s o l i d / l i q u i d i n t e r a c t i o n s . See text. ( A d a p t e d f r o m Berg [70].)
From the equations above we see that an equilibrium thin liquid film on a solid substrate can only exist if e = 0 ~ In the partial wetting case a surface covered with drops is the equilibrium situation. However, from practical experience it is k n o w n that a liquid film can be m a d e on a solid substrate w h e n the static contact angle is greater than zero or even greater than 90 ~. This means that a formed film can be kinetically stable, i.e. dewetting does occur on a time-scale long enough to consolidate the liquid film by, for example, drying first. The fact that a homogeneous film can be formed on a partial wetting substrate is connected to the p h e n o m e n a of dynamic wetting and contact angle hysteresis which will be explained now. When a liquid drop is brought into contact with a tilted substrate there appears to be a tilting angle below which the liquid d r o p does not move d o w n w a r d (see Fig. 6.38). Apparently there exist several metastable states characterised by an advancing contact angle for an advancing contact line and a receding contact angle, usually m u c h lower, for a receding contact line. The advancing and receding static contact angles are defined, according to Fig. 6.38 for a drop just before it starts moving. When the drop moves we are in the d o m a i n of the dynamic advancing and receding contact angles. In Fig. 6.39 the d y n a m i c contact angle is plotted against the substrate speed in a coating situation. We see that with increasing substrate speed the advancing angle increases while the receding angle decreases. This is used in film-coating on hydrophobic substrates. The p h e n o m e n o n of contact angle hysteresis can be caused by surface roughness or chemical surface heterogeneity. The free energies involved are greater than kT (otherwise no metastable states could exist) or in case of vibrations greater than the vibration energy [67]. W h e n we define the surface roughness r as
6 -- PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING Liquid added through capillary tube
199
Liquid withdrawn through capillary tube
*
t
~ ~ ~ :.....:_2...........!..:-.......~::......~..:~-:_: (a) Advancing, 0a
~ ~ ..................... L_____~L__L__:L_____~_L_______~:L~;G~!:!:~ Receding, 0a
.....
0r
WIB~" (b) Drop at point of incipient motion Fig. 6.38. Techniques for measuring contact angle hysteresis. (After Bose, Chap. 3 ha Ref. [67].)
0a O
~ Hysteresis
negative (receding) 0 positive (advancing) Contact line speed Fig. 6.39. Illustration of contact angle hysteresis. (After Bose, Chap.3 in Ref. [67].)
Sr
r=~-c
(6.80)
w i t h Sr the c o n t o u r surface area a n d Sc the ' e n v e l o p e ' surface area of the substrate. The w e t t i n g p a r a m e t e r s are n o w given by W r d -"
~LV(r COS0 + 1)
(6.81)
W w = r ~'LVCOS0
(6.82)
W~ = YLV(r cos0 - 1)
(6.83)
W e see t h e n that the sign of Ww is unaffected, m e a n i n g that fluid p e n e t r a t i o n conditions in substrate pores are unaffected b y surface r o u g h n e s s . A d h e s i o n on a r o u g h surface takes place always, except for 0 > 90 ~ a n d r cos 0 > -1. D e w e t t i n g
200
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY D I P - C O A T I N G
is suppressed on a rough surface with 0 < 90 ~ if r cos 0 > 1 [68]. Our coating experience suggests that roughness may be important in the preparation of layer 2 but probably less important in the subsequent coatings.
6.3.2.4 Stability of Liquid Coatings Liquid horizontal films are always rupture resistant w h e n their thickness exceeds a certain value hi, given by: h 1- 2 q ~ _~L~
' sin
0
(6.84)
where YLis the surface tension of the liquid and 0 the stationary contact angle between a liquid drop and the surface, p and g are the liquid density and the gravitational force, respectively. Most liquid films which are formed during film-coating are metastable or instable. This especially involves coating with aqueous suspensions. With ceramic membrane support coatings with a wet coating as thin as 1000 nm, instability can already occur at relatively low contact angles. A film can only break up into droplets after a disturbance; the film locally thins to less than typically 1000 n m (see Fig. 6.40). In this region the interaction force (van der Waals, electrical double layer, for example) between the liquidsolid and liquid-air surface of the film becomes important. Attraction forces can rupture the thin film and a dry patch is nucleated. Such a film is called a non-wetting film. When the interaction between the two film interfaces is repulsive the so-called disjoining pressure (see also p. 162) of the film, i.e. the pressure difference between the film and bulk liquid, is negative. In the other case of negative disjoining pressures, it may also be called conjoining pressure,
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
Flat film
Disturbance thins film to a tenuous thickness
Disjoining pressuredewets solid
(a)
(b)
(c)
Fig. 6.40. A flat coated film of non-wetting liquid (a) usually greater than I ~rn thick can dewet if a disturbance thins the film (b) to the extent (ordinarilyless than I ~xn)that the effectof conjoining force is to dewet (c) the solid substrate. (After Kheshgi and Scriven [73].)
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
o
~~
m
~0
=
~
<
0
d ~
dh
Film thickness
.,..q
201
h
,,"
~ 0 ~
(a) Wetting Adsorbed layer ~'L ~
lnm Unstable film art> 0
~0 "N'
Fig. 6.41. Disjoining pressure profiles of a wetting liquid film (a) and a non-wetting liquid film (b). (After Kheshgi and Scriven [73].) Surface active aerosol particles
/
O
Surface active o particleemulsions :_. . . .
-..........
~=
.... _ - _ - ~ _ _ - _ :
_-:- :
:ssssss:::sssssssssssssssssssssssss-ss:sss:::::::::::::::::::s::::~ .............................................. ..................................................................
(a)
(b)
High Low High surface surface surface tension tension tension . . . . . . . . . . . .
Instabilities.~ :'::
Thermal gradients
i~
IA
Flow
.....
disturbance Suffactant ~.......~_~_....~_;.~..c,9~entra_t!0n
" ,
i~
~ - - - - ~ - : - _ - _ _ - :
,
::_-:_-:
~ ~
-
~
.....................................................
Cold
(c)
Hot
Cold
(d)
Fig. 6.42. Causes of surface tension gradients. (After Kheshgi and Scriven [73].)
for it is the conjoining force which tends to make liquid conjoin in beads, and hence dewets the substrate. Disturbances in the film tend to level if the disjoining pressure decreases monotonically with film thickness (Fig. 6.41). Disturbances are reinforced a s the conjoining force increases with film thickness.
202
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBY DIP-COATING High Low High surface surface surface tension tension tension
'
IIII
n
Surface-tension-gradient driven flow
I
ptt-Cold
Hot
(a)
Cold
Disjoining pressure active
~....~.. 9
. ~-,i',!ii.~
::: '""::: Rupture ..:.~i~:::i.:i:~ :i: .:
(c)
..-~:
(d)
Fig. 6.43. Surface tension gradient-driven flow draws liquid from regions of locally low tension (a,b). Film thinning might continue until conjoining forces become appreciable (c) and rupture the film (d). (After Kheshgi and Scriven [73].)
Figures 6.42 and 6.43 show how disturbances due to surface tension gradient driven flow occur. Causes for large disturbances can also be small air bubbles and vibrations. The nucleation and growth of dry patches depends on the viscosity of the liquid. Redon et al. [74] found recently that: 1 11
V - - - -
where v is the growth velocity and rl the viscosity. These authors found further, for the system they investigated, that the dimensionless growth velocity scales with 03. These dependencies are expected from theoretical predictions of wetting line velocities [71]. Hence, in ceramic coating operations the contact angle should be as low as possible and the low shear viscosity as high as possible in ceramic coating operations. See further Refs. [72-74]. The dewetting films above were treated as a continuum, i.e a fluid without (micro)structure. However, in most applications the coating fluids are suspensions which contain particles and often macromolecules. The fluids also behave as non-Newtonian. The presence of particles in the film can influence the stability of the wet coating. The non-Newtonian behaviour affects the dewetting kinetics. As far as we know, neither of these aspects has been investigated.
6 - - P R E P A R A T I O N OF A S Y M M E T R I C C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
203
6.3.3 Macromolecular Thickeners and Binders
In this section properties of macromolecular thickeners and binders are discussed. Usually, binder polymers also have thickening properties and in ceramics both functions are often denoted by the term binder [21]. Reasons for the presence of macromolecular additives in coating suspension formulations can be: imparting colloidal stability adjustment of the rheology of the coating suspension - improving the properties of the green coating with respect to drying behaviour and mechanical properties. It is beyond the scope of this chapter to give a full account of these reasons. Only the most important aspects will be highlighted. The discussion is further limited to the aqueous systems as usually used in the dip-coating of support coatings. The interested reader is referred to Refs. [28,75--80] for more information. In colloidal coating suspensions polyelectrolytes are usually used for obtaining colloidal stability. Adsorbing polyelectrolytes change the electrical double layer properties of the ceramic particle -water interface. This is illustrated in Fig. 6.44 where the electrophoretic mobility of (x-alumina particles is plotted as a function of the solution pH in the presence of the negatively charged ammonium polycarboxylic acid. Only at saturated adsorption of the polyelectrolyte is the zeta potential (i.e. mobility) sufficiently negative at the natural pH (about 9-10) of the suspension to obtain sufficient double layer repulsion for colloidal stability. Note the shift of the i.e.p, to the acid region. The particles in effect behave as weak acid negatively charged particles. But only at high adsorbed amount the properties of the alumina surface are masked. Ringenbach et al. [81] investigated the adsorption mechanism of polycarboxylic acid on alumina. The complex formation between polyanion and dissolved aluminium cations appears to play an important role in determining the properties of the adsorbed layer. This may explain the decrease in stability of polyelectrolyte suspensions with time. In the presence of a binder polymer such as PVA the electrostatic repulsion between the particles decreases to some extent [12]. However, no effect of PVA on the wet packing density was observed at maximum polyelectrolyte adsorption. On the contrary de Laat and Derks [82] observed that polyelectrolyte stabilised BaTiO 3 suspensions flocculated upon addition of PVA. These authors studied the steric stabilisation of aqueous BaTiO3 suspensions with block copolymers, which fulfil both the stabilising and binder function. Some block copolymers with PVA and polyacrylic acid blocks were found to be very suitable for this purpose. They found further that depletion flocculation occurs with random copolymers. In that case the homogeneity of dried layers prepared was lower. -
-
204
6 - - PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
~
:
3 -7c~ ~
~. 2 E
1
0
0 ,.Q 0
E -1
.u
-2 0
-3
-4
Electrophoretic mobility of submicrometer R-alumina as a function of the (diluted) suspension pH; no electrolyte added. Above: I m g / g ammonium polycarboxylate (unsaturated adsorption). Below: 5 m g / g ammonium polycarboxylate (saturated adsorption). Fig. 6.44.
Polymeric thickeners serve to increase the high shear and/or low shear viscosity of coating suspension without altering the colloidal stability. Inducing flocculation in a concentrated dispersion gives rise to an exponentional increase of the low shear viscosity but also alters the packing density in the wet and green coating which is undesired. In polyelectrolyte stabilised suspensions high molecular mass non-ionic water soluble polymers can be used to increase the viscosity 14 of the continuous phase liquid. When water is a good solvent for the polymer the effect on suspension stability is probably low when the polyelectrolyte is first adsorbed to full surface coverage. Non-equilibrium effects (which make the order of addition matter) may be important in suspension preparation with viscosifiers (thickeners). Most solutions of non-ionic water soluble polymers are pseudo plastic (shear thinning) but also viscoelastic behaviour can occur. In film-coating it is the shear thinning behaviour that can be used to control the drainage process which occurs at the low gravity determined shear stress in the wet film. What is needed is a moderate viscosity increase 14
Keep in mind that the viscosity of the suspension and of the continuous phase liquid depend on the shear rate (shear stress) in most cases.
6 -- PREPARATION OF ASYMMETRICCERAMICMEMBRANESUPPORTS BYDIP-COATING
205
104 103
Or
g 102 .
O
.
.
.
101 100 10- l
100
101
102
103
~(s-l) Fig. 5.45. Stationary flow curves of a-alumina suspensions 70 w / 0 phi = 0.37. (1) No methyl cellulose (MC) present; (2) 0.06% MC; (5) 0.1% MC; (4) 0.2% MC.
during coating so that the wet film thickness is mainly determined by a relatively low viscosity of the suspension in the dynamic meniscus region, but a higher viscosity in the film (low shear stress) far from the coating bead. In Fig. 6.45 flow curves of (~-alumina suspensions in methyl cellulose (MC) solutions of various concentrations are shown. Methyl cellulose is a binder and thickener frequently used in ceramic processing. The thickening properties are comparable to those of ethyl cellulose and hydroxyethyl cellulose which are often used for the same purpose (see for example Reed [21]). It is seen that the MC acts as a thickener over the whole region studied. The shear thinning effect of the polymer in the shear rate region 0.1/s to 100/s is however not sufficient for the film-coating purpose mentioned. The coating layer thickness of course increases with increasing MC content. In colloidal filtration it is the filtration rate which can be influenced by the addition of a thickener. Now the high shear rates at the pore openings play a role and it is the high shear viscosity which should be influenced. MC could be a suitable thickener for that purpose. The preparation of suitable polymer solutions is not a trivial matter. Preparing solutions without gel particles becomes increasingly more difficult with increasing molecular mass and polymer concentration. To prevent microbial growth the addition of biocides may be necessary a n d / o r storage at low temperature (2~ (see also for example Ref. [83]). Homogeneous solutions can be obtained by dispersing the polymer powder first at a temperature where the solvent quality is poor and then change the temperature. Another method is the dry mixing of polymer powder and ceramic powder in a ball mill followed by wet milling after addition of water. A disadvantage of the latter may be a decrease of milling efficiency due to the increase in viscosity. In all cases it is important that the polymer powder particles become separately wetted by liquid. Agglomerates of polymer powder
206
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING ~.
4O
30
% % %
9~
20-
% %
"0 % %
.,,q
%
10 -
% % %
r
Z~
% %
0 0.0
.
n , n ".., n 0.2 0.4 0.6 0.8 PVA content (ml PVA/ml boehrnite) !
,
, 1.0
Fig. 6.46. The effect of the PVA content of b o e h m i t e sols on the m a x i m u m stress d e v e l o p e d during drying for alumina membranes m a d e and dried under identical conditions. (Redrawn from Kumar [3].)
dissolve very slowly because a gelly polymer layer enveloping the agglomerate hinders strongly their dissolution. Binder polymer present in dip-coating suspensions can prevent the occurrence of cracking during drying. They also increase the green strength of the coating but this function is less important than in bulk ceramic parts. Particulate films with added binder such as PVA or HEC have a higher CCT not because the mechanical strength of the coating is larger (although this can play a role) but because drying stresses are much less due to stress relaxation. Lubrication properties of the binder are probably responsible for this effect. Stress relaxation was experimentally observed for drying boehmite films on alumina substrate obtained by colloidal filtration. In Fig. 6.46 [3] the maximum drying stress in the drying film is plotted as a function of the PVA content of the boehmite sol used. There appears to be a PVA concentration which gives rise to a zero drying stress. TABLE 6.5 A d v a n c i n g contact angle of boehmite sol with a PVA on a hydrophobised alumina substrate. The sol also contains non-ionic surfactant. A n g l e measurements were m a d e on small droplets on the substrate with a goniometer. Conc. PVA
Oa
(%)
(o)
0
90
0.5
70
1.0
68
2.0
63
6 - - P R E P A R A T I O N OF A S Y M M E T R I C C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
207
Another function of surface active polymers such as PVA is their wetting action. This can be seen from Table 6.5 where the advancing contact angle of small droplets of boehmite sol with non-ionic surfactant on a hydrophobised alumina substrate are given for increasing PVA concentration. Aging time of the droplets on the substrate was 4 min. Frothing is an unwanted effect of surface active water soluble polymers. Dynamic surface properties of the solution-air interface due to the presence of the polymer play an important role in foam formation and stability. The surface tension decrease due to adsorbed polymer plays a lesser role (see for example Ref. [84]).
6.3.4 Compact (Cake) Structure As shown in Section 6.2, the compact microstructure is the three-dimensional arrangement of the particles originating from the dispersion, firstly; in the wet coating, secondly, in the dry coating and thirdly, in the sintered coating. A defect structure can be superimposed on this microstructure. The main factor determining the pore properties of a coating is the particle size distribution in the dispersion. The microstructure of the packing (and of the complementary pore space), however, can be largely similar, whether or not defects are present. Capillary forces during drying minimise the large differences in packing density which exist in wet coatings from stable and coagulated or flocculated dispersions. Figure 6.47 shows the surface of a layer 2 film-coating after drying and sintering (only neck forming) prepared from a suspension of c~-alumina in water and charge stabilised with HNO3 and a suspension of the same powder flocculated near the iso-electric point at pH 8. It can be clearly seen that the stable suspension packing is more dense and thus more ordered than the coagulated packing. The packing surface from the stable suspension is also much more smooth than that of the coagulated surface and has a glossy appearance. The pore properties of cast bulk porous material and coating layers from the same suspension become different above sinter temperatures where intermediate stage sintering in the bulk starts (see Section 6.2.5). At lower temperatures pore properties of free casts determined with Hg porosimetry can be used to compare the pore properties of consolidated dispersion coatings. In Fig. 6.48 the porosity of free cast layers as a function of the suspension pH is given and in Fig. 6.49 the pore size distributions as determined with Hg porosimetry for free cast layers at pH 3.6 and 8.1 are given. It is shown that the porosity difference between consolidated coatings obtained from stable and coagulated suspensions is only 10% and the median pore diameter decreases from 210 nm to 160 nm going from instable to stable suspensions. The decrease
208
6 ~ P R E P A R A T I O N OF A S Y M M E T R I C C E R A M I C M E M B R A N E S U P P O R T S BY D I P - C O A T I N G
(a)
(b)
Fig. 6.47. Layer 2 film-coating after drying and sintering. (a) Electrostaticallystabilised suspension of 0~-A1203powder at pH 3. (b) Stronglyagglomerated suspension of the same powder at pH -~8 (i.e.p.).
in pore size is at the expense of porosity. Because wet coatings from unstable suspensions have a loose to very loose (random) packing, the linear shrinkage of flocculated wet coatings typical for the surface is 65% and that of the stable densely packed coatings is 30%. Hence, the flocculated coatings are much more amenable to cracking due to inhomogeneity shrinkage during drying than stable densely packed coatings. Furthermore, the rheology of flocculated suspensions is much more time dependent than that of the stable suspensions. This can be detrimental to coating operations.
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
209
50 after drying 9149
9 9149149149
O
~- 40 O
30 3
l
I
4
5
,
I
,I
6
7
....
I
I
8
9
pH
Fig. 6.48. Porosity of free cast coatings of 0~-A1203 as a function of the s u s p e n s i o n p H after drying and sintering (Hg porosimetry, cylinder model).
100 .=
'
'
' '''"I
'
'
''''"I
'
'
''''"I
80 I ! ! !
60 E
i
I i 1 i .......... p H 8 I I 1 1!
=.
O
~- 40
!
i!
! : !
20
!
' i
,
,,
,,,,I
10
i I ,,,ii
1
i! i
i I000
100 rp (nm)
Fig. 6.49. Pore size distributions of unsupported o~-A1203'coatings' obtained by Hg porosimetry of dried and sintered casts.
6.4 APPLICATIONS In this section we will put ourselves in the position of a ceramic engineer who wants to judge existing methods of preparing ceramic membrane supports and who is interested in developing alternative colloidal processing routes to make porous coatings on porous tubes. Before commencing experimental trials we
210
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
study the previous sections to understand the processes of film-coating and colloidal filtration and the relative merits of the existing patents on these topics. Suppose we wish to make a layer 2 coating on the inside of macroporous alumina tubes. The properties of the alumina substrate tube materials which are available are as follows: length Im outer diameter 14 mm inner diameter 8 mm porosity 0.4 hydraulic pore diameter 4 ~tm mean surface roughness Ra 5 ~tm max. surface roughness Rmax 30 ~tm Suppose further that at first we have suspension dip-coating in mind for the preparation of the layer to be obtained. The coating should be suitable as a substrate for a microfiltration membrane (layer 3) with a pore diameter of 200 nm. Which coating compound material is most appropriate? This depends on the application and on the substrate material. When there is no reason not to use alumina, this is the best choice because thermal shock cracks can then be avoided during heat treatment (sintering) of the coating. What pore size and porosity should be aimed at? For the particular application, the resistance of the layer 2 intermediate coating should be as low as possible which means large pore size, high porosity and low layer thickness. However, the pore size is largely restricted by the pore size of the tubular substrate (4 ~tm) and by the relevant coating mechanism. Colloidal filtration is selected as the dip-coating mechanism for the first trials in the development path. This means that cake filtration should occur when the suspension comes into contact with the substrate. So the particle size in the suspension should not be much smaller than I ~tm (approximately 4 times less than the mean pore size in the substrate) otherwise too much penetration and clogging of the substrate occurs prior to cake build-up. This would give rise to an extra high 'interfacial' flow resistance during application of the MF membrane. In assessing commercially available alumina powders we find Alox F as a promising candidate for the preparation of a coating suspension. Properties of this powder as provided by the manufacturer are: mean particle size 4 ~tm specific surface area <1 m2/g With this powder, cake filtration seems possible. It is always prudent to check the powder properties of each batch received. With static light scattering (for example, a Malvern Mastersizer) we find: d90 7 ~tm d50 2.8 ~tm dl0 0.4 ~tm
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
211
and as = 4.3 m2/g. So in a well dispersed sample there appears to be a tail in the distribution on the small side. These particles may penetrate the substrate during slipcasting. What porosity and pore size can be obtained with this powder? The porosity and pore size depend on the packing density obtained after coating and drying and the sintering temperature. Usually only neck forming without much shrinkage is sufficient to obtain a satisfactorily consolidated coating. So we are only concerned with the expected properties of the green coating. Drying shrinkage should be minimal. Further, the green microstructure should be as homogeneous as possible. Hence random packing of the particles should be aimed at. This means coating with a colloidally stable suspension containing no aggregates or agglomerates. Random packing properties of irregular, slightly anisotropic powders with a log normal size distribution such as for the powder chosen can not be predicted. But as a general rule the porosity of a dense random packing of such a powder is 0.3--0.4 and the median pore size is about 1/4 of the median particle size or about 0.7 for the Alox F powder. This size seems acceptable with respect to the application of the MF coating (layer 3). How do we prepare an Alox F suspension which is colloidally stable and fully deagglomerated? Colloidal stability can be obtained by the addition of a stabilising agent such as Darvan C (van der Bilt Corp.) to an aqueous powder mixture. Darvan C is a negatively charged polyelectrolyte which strongly adsorbs on the powder surface creating electrostatic repulsion between the particles and soft agglomerates, keeping them apart. The optimum dosage can be determined approximately by a viscosimetric titration of a concentrated dispersion with the dispersant to the viscosity minimum. However, the amount determined may be an underestimate because the powder was not fully deagglomerated initially and strong mechanical agitation may be needed in addition. Commercial submicrometer powders still contain a small fraction of hard agglomerates (aggregates) and a large fraction of soft agglomerates. The latter can be removed by some form of ball milling in the presence of the dispersant. The hard agglomerates can only be partly broken up in a milling step. Complete removal requires sedimentation fractionation. What suspension concentration is needed? Before this question can be answered we should first estimate the layer thickness needed. The minimum layer thickness which results in a fully covered substrate and which is free of pinholes, cracks etc. depends on the surface properties of the support: - surface roughness - heterogeneity in surface porosity. The latter factor was found experimentally (Section 6.2.6) and supported by the results of computer simulations [85]. The minimum layer thickness should be of the order of the maximum in surface roughness, i.e. 30 ~tm in this case. So we aim at 40 ~tm for the first trials
212
6 ~ PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
to allow a margin for uncertainty. This layer thickness is still appreciably lower than the critical cracking thickness (CCT). We assume further a drying shrinkage of 10%, so the wet layer thickness should be about 45 ~tm. The next step is to judge the slipcast capacity of the substrate. This capacity is determined by the porosity of the support and the wall thickness. The liquid penetration in the substrate h m is related to the coating thickness hc by hm = ~
- 1
(6.86)
s
where s is the substrate porosity and g~ and % are the solid volume fractions in the cake and suspension, respectively. (Pc is 0.5-0.6 for a dense packing of powders such as Alox F. We assume (Pc = 0.5. When the maximum substrate capacity is used it follows that % = 0.02 is the minimum volume fraction of solids which can be used. How long does it take to form that layer? We use Eq. (6.39) to estimate the filtration time. The permeability of the substrate is calculated from the clean water flux given and the permeability of the wet cake can be estimated from the Carman-Kozeny equation. The result is a relatively short contact time of t = 0.1 s. Filling the tube with suspension takes longer. So an approximately constant filtration time along the tube surface can not be achieved under these conditions. It is now evident that the filling and emptying time of the tube should be short compared to the filtration time in order to obtain a homogeneous coating along the substrate. Note that this problem does not occur with short test pieces or with the application of the layer on the outside of tubes according to Fig. 6.12. It could be argued that when the substrate is saturated the tayer thickness (cake) does not continue to grow and the times become unimportant. However, under these conditions there is no longer any pressure drop across the cake and because the cake is very likely to be still largely mobile (stable suspension) it is very susceptible to shear erosion or resuspension during draining the suspension from the tube. There are now two possible routes to explore: (1) make the suspension less stable, but still avoiding agglomeration in the suspension, that is, sufficient to induce the mobile-immobile transition across the whole layer; and (2) sustain the capillary pressure during drainage. The second method seems the most practical. Only a partly saturated support during drainage is needed. The correct conditions can be set-up as follows: (1) increase the volume fraction of solids in the suspension; (2) increase the viscosity of the continuous phase in the suspension. A volume fraction of 0.035 is needed to reach a half saturated support when the wet layer thickness becomes 45 ~tm. A filtration time of 5 min seems to be appropriate in order to be relatively independent from the filling and emptying time of the I m long tube. The desired viscosity can be calculated from Eq. (6.39).
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
213
Note that the driving force for the higher viscosity suspension may be lower than that for pure water because the surface tension may be lower depending on the thickener used. For aqueous polymer thickeners 7 may be 40-50 m N / m . A value of 40 m N / m is selected so that the driving pressure is 0.4 105 Pa. The viscosity calculated is then 4.7 Pa.s. This is a high viscosity and the filmcoat layer left after gravity drainage of the suspension from the tube is certainly not negligible. The thickness of this fluid film can be estimated if the drainage speed is known. Assuming Poisseuille flow and a fully developed velocity profile at all times the latter can be estimated to be about 4 mm/s. With Eq. (6.67) the wet filmcoat thickness is then calculated to be 1.2 mm. This may be an overestimation but if the adhering layer is assumed to be about 0.5 mm this gives rise to a green filmcoat layer of about 30 ~tm. This is the value aimed at with only the slip coat mode active. So it seems not to be possible to prepare a suitable coating on the inside of long tubes on the basis of colloidal filtration alone. Now we are going to analyse one of the patents available. In the French patent wo 85/01937 by Aureol and Gillot a method is disclosed to make good adherent microfiltration (layer 2 and 3) coatings on the inside of macroporous tubes, by filling and emptying (draining) a tube with a deagglomerated suspension and subsequent drying and sintering of the coating obtained. The terms slip casting, film-coating, filtration etc. do not occur in this patent. Filling and emptying times (speeds) as well as contact time of the suspension with the substrate are not mentioned. So the mechanism of layer formation does not become clear after reading this patent. We consider why their method should work and decide to analyse one of their examples along the lines already explored above. Aureol and Gillot give the following characteristics of their alumina support tubes" Outer diameter 19 mm Inner diameter 15 mm Wall thickness 2 mm Porosity 0.35 Mean pore size 12 ~tm Clean water permeance 120 m3/m2.h.bar Properties of layer 2 given in example 2 of their patent are: Material o~-A1203 Layer thickness 20 ~tm Porosity 0.3 Mean pore size 1.2 ~tm Clean water permeance 40 m3/m2.h.bar The preparation of this coating is as follows: (1) suspension preparation ; (2) filling the tube with suspension; (3) gravity drainage of the suspension from the tube;
214
6 ~ P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
(4) drying and sintering of the coating at 1550~ The suspension composition is given as (1) ~-alumina powder average grain size 2.5 ~tm; (2) alumina concentration 8% (m/m) ; (3) aqueous polyethylene glycol (Carbowax 4000C Union Carbide) solution 91.8% (viscosity 0.5 Pa s); (4) Darvan C dispersant 0.2%. According to the example the above mix is ball milled for 24 h in a 25 1 ball mill charged with 25 kg of alumina balls of 10 mm cross-section and 7 1 of suspension. The authors state that the milling operation is essential for breaking up the agglomerates and to disperse the particles well. This may be correct but the deagglomeration process may be more efficient by first milling in a mixture without thickener and then adding the polyethylene glycol (PEG) to the mixture in the ball mill. Further, the proportion of Darvan C seems to be rather large. The initial polyelectrolyte concentration is then about 2 g/1 in the aqueous phase. When a specific surface area of the alumina of about 1 m2/g is assumed and a plateau adsorption of the polyelectrolyte of 0.5 m2/g, We can easily calculate that the polyelectrolyte concentration in the aqueous phase after adsorption remains about the same. Hence it seems that there should be a large excess of polyelectrolyte in the solution. The deagglomeration optimum seems to be surpassed. Under these conditions the possibility of weak agglomeration due to depletion flocculation exists (see, for example, Ref. [86]). These agglomerates are weak enough to break up under the action of drying stresses and do not necessarily hinder dense packing of the particles in the coating process. An advantage may be the prevention of clogging of thesubstrate when the suspension contacts the substrate. But there is also a high concentration of polymeric thickener in the suspension. Adsorption competition between the Darvan C and the PEG may be the reason for the high Darvan C content necessary to obtain sufficient electrostatic repulsion between the particles for colloidal stability of the suspension. However, water is a good solvent for PEG and the adsorption on the alumina/water interface is likely to be small. Another possibility is complexation of PEG with the polymethacrylate (Darvan C) lowering the effectiveness of the dispersant. What is the function of the PEG addition? It could act as a binder preventing cracking during drying of the coating but the thickening effect may be beneficial or even necessary for the coating action itself as we have seen in the first exercises. So it could be asked whether slip casting is the main coating mechanism in this case. We assume the thickness of the wet cake to be 25 ~tm and the porosity 0.5. The volume fraction of alumina in the suspension is 0.02, which is rather low. From a volume balance (Eq. (6.36)) we calculate whether the slip cast capacity is large enough to form a layer of 25 ~tm. It is found that a substrate thickness of 1.9 mm is needed. This is about the wall thickness of the substrate
6 m PREPARATION OF ASYMMETRIC CERAMIC MEMBRANE SUPPORTS BY DIP-COATING
215
tube. Hence filtration coating (slip casting) seems possible. We calculate the permeability of the wet cake to be Kc = 1.7 1 0 -14 m 2 and the substrate permeability K 1 = 6.6 10-13m 2 and then we calculate the slip cast time necessary. Assuming good wetting of the substrate and a surface tension of 50 m N / m a filtration time of about 61 s is found. This time is relatively long because of the high viscosity of the dispersion liquid and the low volume fraction solids in the suspension. In the patent a waiting time between filling and emptying the tube is not mentioned. The patent states that the film of slip remaining after draining forms the coating, so perhaps film-coating is the filtration mode after all. In film-coating it is not the contact time but the drainage speed and suspension viscosity which are the crucial parameters. The increase in viscosity due to the presence of the alumina particles is low as can be seen from the Einstein equation (Eq. (6.21)). So the suspension viscosity is about the same as that of the dispersion liquid: 0.5 Pa s. Because the PEG is rather low in molecular mass, assuming the suspension rheology to be Newtonian seems reasonable as a first approximation. The free gravity drainage rate can now be estimated as 0.25 m / s and the wet filmcoat layer is then calculated to be about 2 mm. After drying this should give rise to a coating thickness of about 65 ~tm, much thicker than the thickness obtained in practice. But remember this may be an overestimation. Nevertheless, these estimates suggest that the film-coating mechanism is important too in this case. Assuming filling and emptying speeds to be the same and no waiting time in between, the contact time of suspension with substrate (assuming a tube length of I m) is about 13 s at the lower end and less than I s at the higher end. So the slip cast contribution is negligible at the higher end and about 10 ~tm at the lower end. This situation where both mechanisms seem to play a role is undesired from the point of view of process control. Coating long tubes by colloidal filtration alone does not seem possible if a homogeneous coating along the tube is the requirement. On the contrary, film coating can be the only mechanism when the capillary action of the substrate is completely suppressed as we have previously discussed. Filmcoat layers on hydrophobised substrates can be obtained both with stable and flocculated dispersions. To obtain the same layer thickness after drying (but a different microstructure and texture see Section 6.3.4) the agglomerated suspensions should have a higher viscosity at the shear stresses of importance. This is already the case at relatively low volume fraction solids. The strongly shear thinning behaviour of the flocculated suspension is advantageous from the coating point of view. It is however disadvantageous from the point of view of drying and sintering. The rheology of the suspensions together with the withdrawal speed determine the coating thickness. In Fig. 6.50 flow curves of Alox SF c~-alumina in water at pH 4 and pH 8 are shown. The low pH suspension is electrostatically stabilised using the positive
216
6 ~ P R E P A R A T I O N OF ASYMMETRIC C E R A M I C M E M B R A N E SUPPORTS BY D I P - C O A T I N G
I
I
I
I
I
I
pH 8 5
pH 4
4
I
I
I
I
I
I
20 40 60 80 100 ~(s-l) Fig. 6.50. Flow curves of Alox SF 0c-A1203in water at pH 4 and pH 8 (same shear history in both cases; see also text).. surface charge of the oxide present at low pH. At pH 8, the i.e.p, of the oxide, the suspension is strongly flocculated because of charge neutralisation. The solids volume fraction in the stable suspension is much higher than in the case of the flocculated suspension. However, the coating thickness is of the same order when the coating speed is similar. Both suspensions are non-Newtonian, although the stable suspension behaves more Newton-like because of the smaller yield value and linear flow curve. However, prediction of layer thicknesses on the basis of apparent viscosities and Newtonian film-coating theory lead to gross overestimation of the layer thickness. This was also reported by Kuhlmann [66] for the enamelling of steel with non-Newtonian filmcoat suspensions. The relationship between coating speed, rheology (solids volume fraction) and layer thickness needs to be investigated empirically. The low stress region of the flow curves is especially important in film-coating. In Fig. 6.51 the mean layer thickness after drying is plotted as a function of the withdrawal speed for a colloidally stable concentrated (q~ = 0.42) submicrometer 0~-alumina suspension. Three curves can be seen: (1) coating on a smooth hydrophillic non-porous glass tube; (2) coating on a water-saturated macroporous support tube; (3) coating on a hydrophobic silanised macroporous support tube. The coating set-up is as already shown in Fig. 6.12. The model case is the coating on glass (no capillary filtration effects are present and wetting problems are absent because the surface is hydrophilic). We see that the coating thickness increases with the withdrawal speed as expected (see Section 6.3.2). The slope of a log-log plot of coating thickness versus coating speed lies between 1/2 and 2/3 of the values for the gravity determined and surface tension determined cases for Newtonian coating flows (see Fig. 6.52). The deviation probably stems from the
6
-
-
PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING I
I"
I
'
I
I
stable a-Al203 suspension
50
"•40
'
217
A
3O
20
coating on:
9 smooth non-porous glass
10
9 porous support + capillary action II support without capillary action I
I
I
I
4
8
12
16
I
Vw(mm's-1)
Fig. 6.51. Mean layer thickness after drying as a function of the withdrawal speed for a colloidally stable concentrated (9 = 0.42) submicrometer oc-A1203suspension. log(h/mm) -1.8 -1.6 -1.4 -1.2 0.2
:
J
-
coating ~ suspension ~
I
0
:
J
0.2
~
!
0.4
stable coagulated
I
!
0.6
a
!
!
I
I
I
0.8 log (vw/mm.s-1)
!
o pH 4 p - 2.26 g/cm3 9 pH 8 19= 1.50 g/cm3
Fig. 6.52. Log-log plot of mean layer thickness versus withdrawal speed for film-coating of strongly agglomerated (pH = pH (i.e.p.) = 8) and stable alumina suspensions (pH = 4). fact t h a t the a l u m i n a s u s p e n s i o n s are p s e u d o p l a s t i c a n d s l i g h t l y thixotropic. T h e a p p l i c a t i o n of p o r o u s c e r a m i c c o a t i n g s c o m p l e t e l y d e t e r m i n e d b y the f i l m - c o a t i n g m o d e of w i t h d r a w a l c o a t i n g is o n l y p o s s i b l e u s i n g p l a i n concent r a t e d s u s p e n s i o n s w h e n the c a p i l l a r y s u c t i o n of the s u b s t r a t e is c o m p l e t e l y s u p p r e s s e d d u r i n g i m m e r s i o n of the s u b s t r a t e in a d i s p e r s i o n . E v i d e n c e for this b e c o m e s visible on c o m p a r i n g the c u r v e s for c o a t i n g on w a t e r s a t u r a t e d a n d
218
6 - - P R E P A R A T I O N OF ASYMMETRIC CERAMIC M E M B R A N E SUPPORTS BY D I P - C O A T I N G
hydrophobised support tubes. The coating behaviour on the hydrophobised support is similar to that on glass. A coating flow is possible because dynamic wetting forces the initial coverage of the surface despite a static contact angle of about 100-110 ~. The layer thickness is somewhat larger due to the larger roughness of the ceramic support. The roughness of the substrate also helps in forced wetting of the surface. The relatively high viscosity of the suspension under the coating conditions retards the tendency to break up (dewetting) of the coating. The water saturated tubes show a different coating behaviour especially at low coating speeds (long contact times suspension/support). There still appears to be a filtration effect determining the layer thickness at low speed. At higher speeds the film-coating mode is dominant. The result is a minimum in the coating thickness-withdrawal speed relationship. 6.5 FINAL R E M A R K S
In this chapter we have seen that application of porous ceramic coatings on porous substrates for preparing membrane supports is a complex process. Every step has to be carried out successfully to obtain substrates or membranes themselves which fulfil the requirements. We have seen that models of the coating processes are useful but still far from capable of describing the processes completely. We have seen that specific aspects such as prevention of defect formation still defy quantitative and sometimes also qualitative understanding. We have seen that scaling-up is not a trivial matter and that much work has to be performed to enable successful preparation of large surface areas. The manufacture of ceramic membrane supports is an emerging technology, based on skilful craftsmanship and innovative engineering science, with a wide range of possible applications.
Acknowledgements Firstly, I am greatly indebted to ECN for having offered me the opportunity to contribute this Chapter. I would also like to thank J. Heijn (BetaText) for his invaluable assistance in preparing and editing the manuscript. I am very grateful to P.T. Alderliesten, J. Heijn, R. Blackstone, C.W.R. Engelen, P.P.A.C. Pex, P.J.A.M. Blankevoorde and H.J. Veringa for their review of the manuscript. I acknowledge the late Mrs. M.J. Schoute for her early contributions to ECN membrane development. I would like to thank W.H. van 't Veen, J.A.J. Peters, A.J.G. Engel of the ECN Membrane Group for the experimental work on which this chapter is based; also G. Hamburg and Mrs. C.M. Roos for performing the necessary electron microscopy; R.G. Nyqvist of the ECN Characterisation Group for carrying out porosity measurements of the support materials; and, finally, to Mrs. N. 't Hart for skilfully drawing the diagrams and illustrations.
6 -- PREPARATION OF ASYMMETRICCERAMIC MEMBRANE SUPPORTS BY DIP-COATING
LIST OF SYMBOLS
A a
Ca D Do De e
AG
AGD AGH AGs AGel AGpol
g Go H h h
hc k k Kic Kc k0 Lc no //1 1/3 t/i
Nf Nr Nst Pe Pcap
P P1 Ps q R R
Hamaker constant [J] particle radius [m] capillary number mutual particle diffusion coefficient [m2.s-1] mutual particle diffusion coefficient, lim n ~ 0 [m2.s-1] Deborah number elementary charge (1.6021x10-19) [C] Gibbs free energy [J] G due to van der Waals attraction or dispersion [J] G due to hydrogen bond attraction [J] G due to solvation interaction [J] G due to electrostatic interaction [J] polymeric interaction [J] gravity acceleration [m-s-2] Goucher number separation distance Ira] Planck constant (6.626x10 -34) [J.s] layer thickness [m] CCT, critical cracking thickness Boltzmann constant (1.381x10 -23) [J.K-I] local permeability [m 2] fracture resistance [N.m -2] average permeability in/of compact [m 2] particle shape factor compact thickness [m] initial particle number concentration [m -3] ~ refractive index of particle material refractive index of dispersion material number density of ion i dimensionless number dimensionless number Stokes number Peclet number capillary pressure [N.m -2] dynamic pressure [N-m -2] local liquid pressure in the compact [N-m -2] local solid pressure in the compact [N-m -2] superficial fluid velocity [m.s- ] gas constant [J.K-1] centre to centre distance [m]
219
220 aBr Rsh
Rf Rs Ra
Rmax y /'max rmin t"s
So Sr Sc T tl/2 td tp t U Vw v
W Woo Wad
Wcoh Ww w~ Zi
Fsv I-'SL
YLV 7SV ~SL
Ys 7 8 E1 E E0 Er E1
6 ~ PREPARATIONOF ASYMMETRICCERAMICMEMBRANE SUPPORTSBY DIP-COATING
Brownian collision rate [m-3"s-1] shear collision rate [m-3-s-1] fast agglomeration rate Is-1] slow agglomeration rate [s-1] mean surface roughness [m] max surface roughness [m] surface roughness parameter maximum radius of curvature [m] minimum radius of curvature [m] substrate radius [m] specific surface of porous medium per volume solid [m -1] contour surface area [m 2] envelop surface area of substrate [m 2] absolute temperature [K] characteristic agglomeration time [s] diffusion time [s] characteristic process time [s] time [s] fluid velocity [m.s -1] withdrawal speed [m-s-1] growth velocity [m.s -1] stability ratio stability ratio in absence of repulsive interactions reversible work of adhesion [J.m-2] reversible work of cohesion [J.m-2] work of wetting [J.m -2] spreading parameter [J-m -2] valency of ion i adsorbed amount of vapour [mol-m -2] adsorbed amount of surfactant [mol.m -2] surface tension of liquid-vapour surface [N-m -1] surface tension of solid-vapour surface [N.m -1] surface tension of solid-liquid surface [N-m -1] surface tension of solid-vacuum surface [N.m -1] surface tension [N.m -1] shear rate [s-1] thickness of adsorbed layer [m] substrate porosity dielectric constant [C2.j-l.m -1] dielectric constant of vacuum [C2.j-l.m -1] relative dielectric constant of medium static dielectric constant of particle [C2-j-l-m -1]
6 -- PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING ~3
~o 0
-1
K
V /I~SV /I~SL
P (5
() ()2 ()m ()eft ()c ()0 Z
~Fo
221
static dielectric constant of dispersant [C2-j-l.m -1] viscosity of the dispersant m e d i u m [N.s.m -2] viscosity [N-s-m -2] contact angle Debye length [m] electromagnetic frequency [s -I] s p r e a d i n g pressure s o l i d - v a p o u r surface [N.m -1] s p r e a d i n g pressure solid-liquid interface [N.rn -1] density [kg.m -3] tensile stress [N.m -2] shear stress [N-m -2] solids v o l u m e fraction m e a n v o l u m e fraction p o l y m e r in the adsorption layer m a x i m u m effective particle packing v o l u m e effective v o l u m e fraction solids v o l u m e fraction in the cake
solids volume fraction in the suspension Flory-Huggins polymer interaction parameter surface potential IV]
REFERENCES
1.
2.
3. 4.
5. 6. 7. 8.
9.
B.C.Bonekamp, R.A. Terpstra and H.J. Veringa, On the forming of ceramic coatings by concentration polarization in dead-end and cross flow pressure filtration, Proc. 1st ECerS Conference Maastricht, in: O. de With, R.A. Terpstra and R. Metselaar (Eds.), Euro-Ceramics 1. Elsevier, London, New York.1989, pp. 1218-1222. A.J. Burggraaf and K. Keizer, Synthesis of Inorganic Membranes, in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991, p. 10-63. K.-N.P. Kumar, Nanostructured Ceramic Membranes, Layer and Texture Formation, Thesis Twente University, Enschede, The Netherlands 1993. P.P.A.P. Pex et al., Materials aspects of microporous inorganic gas separation membrane manufacturing, in: Eds. W.R. Bowen, R.W. Field and J.A. Howell (Eds.), Proc. Euromembrane "95. European Soc. Membrane Sci. and Technology, 1995. A.J. Pugh and L. Bergstr6m (eds.), Surface and Colloid Chemistry in Advanced Ceramics Processing, Surfactant Science Series 51, Marcel Dekker, New York, 1995. H. Sonntag and K. Strenge, Coagulation Kinetics and Structure Formation. Plenum Press, New York and London, 1987, Fig. 4.6, p. 141. F.M.Tiller and T.C. Green, The role of porosity in Filtration IX: Skin effect with highly compressible materials. AIChE, 7 (1973) 1266-1269. A.P. Philipse, B.C. Bonekamp and H.J. Veringa, Colloidal filtration and (simultaneous) sedimentation of alumina and silica suspensions: Influence of aggregates. J. Am. Ceram. Soc.. 73 (1990) 2720-2727. G.L. Witucki, A silane primer: Chemistry and applications of alkoxy silanes. J. Coat. Technol., 65 (1993) 57-60.
222
6 -- PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
10. B. Lindman and G. Karlstr6m, Polymer-surfactant systems, in: D.M. Bloor and E. Wyn-Jones (Eds.), The Structure, Dynamics and Equilibrium properties of Colloidal Systems. NATO ASI Series C 234, 1989, pp. 131-147. 11. B.C. Bonekamp, W.H. van 't Veen, M.J. Schoute and H.J. Veringa, Deflocculation of aqueous (x-A1203 suspensions in the presence of adsorting polymer and polyelectrolyte, Proc. 1st ECerS Conference Maastricht, in: O. de With, R.A. Terpstra and R. Metselaar (Eds.), Euro-Ceramics 1, 1989, Elsevier, London and New York. 12. A.W.M. de Laat and G.L.T. van den Heuvel, Colloids Surfaces A: Physicochem. Eng. Aspects, 70 (1993) 179. 13. D.C. Montgomery, Design and Analysis of Experiments, 3rd edn. Wiley, New York, 1991. 14. J.W. Goodwin, Rheological properties, interparticle forces and suspension structure, in: D.M. Bloor and E. Wyn-Jones (Eds.), The Structure, Dynamics and Equilibrium Properties of Colloidal Systems. NATO ASI Series C 324, Kluwer, The Netherlands, 1990, pp. 659-679. 15. H.A. Barnes, J.F. Hutton and K. Walters, An Introduction to Rheology. Rheology Ser., 3. Elsevier, Amsterdam, 1989. 16. W.B. Russel, The Dynamics of Colloidal Systems. The University of Wisconsin Press, Madison, WI, 1987. 17. F.A.L. Dullien, Porous Media, Fluid Transport and Pore Structure, 2nd edn., Academic Press, London, 1992, p. 237. 18. K.S. Sorbie, P.J. Clifford and E.R.W. Jones, The rheology of pseudoplastic fluids in porous media using network modelling. J. Colloid Interface Sci., 130 (1989) 508-533. 19. W. Kozicki, C. Tiu and A.R.K. Rao, Filtration of non-Newtonian fluids. Can. J. Chem. Eng., 46 (1990) 313-321. 20. K.K. Kunze and D. Segal, Modification of the pore structure of sol-gel derived ceramic oxide powders by water-soluble additives. Colloids Surfaces, (1991) 328-337. 21. J.S. Reed, Introduction to the Principles of Ceramic Processing, 2nd edn., Wiley, New York, 1995. 22. R.J. Hunter, Foundations of Colloid Science. Oxford University Press, Oxford, 1987. 23. P.C. Hiemenz, Principles of Colloid and Surface Chemistry. Dekker, New York, 1977. 24. T.G.M. van de Ven, Colloidal Hydrodynamics. Academic Press, London, 1989. 25. W.B. Russel, D.A. Saville and W.R. Schowalter, Colloidal Dispersions. Cambridge University Press, Cambridge, 1989. 26. J.N. Israelachvili, Intermolecular and Surface Forces, with Applications to Colloidal and Biological Systems. Academic Press, London, San Diego, New York, 1985. 27. R.G. Horn, Surface forces and their action in ceramic materials. J. Am. Ceram. Soc., 73 (1990) 1117-1135. 28. L. Bergstr6m, Rheology of concentrated suspensions, in: R.J. Pugh and L. Bergstr6m (Eds.), Surface and Colloid Chemistry in Advanced Ceramic Processing. Surfactant Science Ser., 51, Dekker, New York, 1994, p. 200. 29. J. Lyklema, Fundamentals of Interface and Colloid Science. I. Fundamentals. Academic Press, London, 1991. 30. H. Sonntag and K. Strenge, Coagulation Kinetics and Structure Formation. Plenum Press, New York and London, 1987. 31. D.H. Everett, Basic Principles of Colloid Science. Royal Society of Chemistry, London, 1988. 32. G.J. Fleer, M.A. Cohen Stuart, J.M.H.M. Scheutjens, T. Cosgrove and B. Vincent, Polymers at Interfaces. Chapman and Hall, London, 1993.
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
223
33. J.P. van Houten, On the Control of Powder Compact Microstructure through Methods of Wet Consolidation, Thesis, Delft University 1995. 34. A.P. Philipse, B.C. Bonekamp and H.J. Veringa, Colloidal filtration and (simultaneous) sedimentation of alumina and silica suspensions: Influence of aggregates. J. Am. Ceram. Soc., 73 (1990) 2720-2727. 35. M.T. Strauss, Polydispersity in Electrostatically Stabilized Ceramic Suspensions, Thesis, MIT, 1985. 36. M. Strauss, T.A. Ring and H.K. Bowen, Osmotic pressure for concentrated suspensions of polydisperse particles with thick double layers. J. Colloid Interface Sci., 118 (1987) 326-324. 37. B. Beresford-Smith and D. Chan, Electrical double layer interactions in concentrated systems. Faraday Discuss. Chem. Soc., 76 (1983) 65-75. 38. B. Beresford-Smith and D.Y.C. Chan, Discussion remark in General Discussion. Faraday Discuss. Chem. Soc., 76 (1983) 113. 39. (a) R.C. Chiu, T.J. Garino and M.J. Cima, Drying of Granular films: I. Effect of processing variables on cracking behavior. J. Am. Ceram. Soc., 76 (1993) 2257-2264. (b) R.C. Chiu and M.J. Cima, Drying of granular ceramic films: II. Drying stress and saturation uniformity. J. Am. Ceram. Soc., 76 (1993), 2769-2777. 40. G.W. Scherer, Theory of drying. ]. Am. Ceram. Soc., 73 (1990) 3-14. 41. C.J. Brinker and G.W. Scherer, Sol-Gel Science, The Physics and Chemistry of Sol-Gel Processing. Academic Press, Boston, 1990. 42. G.W. Scherer, Drying gels II. Film and plate. J. Non-Crystalline Solids, 89 (1987) 217-238. 43. A.J. Hurd and C.J. Brinker, Ellipsometric imaging of drying sol-gel films. Mater. Res. Soc. Syrup. Proc., 121 (1988) 731-742. 44. M.S. Hu, M.D. Thouless and A.G. Evans, The decohesion of thin films from brittle substrates. Acta MetalI., 36 (1988) 1301-1307. 45. B.C. Bonekamp and H.J. Veringa, Green microstructures and their characterization, in: R.J. Brook (Ed.), processing of Ceramics, Part 1. Materials Science and Technology, A Comprehensive Treatment, 17A (1995) 352. 46. R.K. Bordia and A. Jagota, Crack growth and damage in constrained sintering films. J. Am. Ceram. Soc., 76 (1993) 2475-2485. 47. R.K. Bordia and R. Raj, Sintering behavior of ceramic films constrained by a rigid substrate. J. Am. Ceram. Soc., 68 (1985) 287-292. 48. T.J. Garino and H.K. Bowen, Kinetics of constrained-film sintering. J. Am. Ceram. Soc., 73 (1990) 251-257. 49. G.W. Scherer and T. Garino, Viscous sintering on a rigid substrate. J. Am. Ceram. Soc., 68 (1985) 216-220. 50. A.F.M. Leenaars, Preparation, Structure and Separation Characteristics of Ceramic Alumina Membranes, Thesis Twente University, The Netherlands, 1984. 51. A.F.M. Leenaars and A.J. Burggraaf, The preparation and characterization of alumina membranes with ultrafine pores, Part 2: The formation of supported membranes. J. Colloid Interface Sci., 105 (1985) 27-40. 52. F.M. Tiller and C.-D. Tsai, Theory of filtration of ceramics: I. Slip casting. J. Am. Ceram. Soc., 69 (1986) 882-887. 53. F.M. Tiller and T.C. Green, Role of porosity in filtration, IX: Skin effect with highly compressive materials. AICHE J., 19 (1973) 1266-1269. 54. F.M. Tiller and N.B. Hsyung, Theory of filtration of ceramics: II. Slip casting on radial surfaces. J. Am. Ceram. Soc., 74 (1991) 210-218.
224
6 ~ PREPARATIONOFASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
55. F.A.L. Dullien, Porous Media, Fluid Transport and Pore Structure. 2nd edn., Academic Press, London, 1992. 56. J. Dodds and M. Leitzelement, The relation between the structure of packings of particles and their properties, in: N. Boccara and M. Daoud (Eds.), Physics of Finely Divided Matter. Springer Proc. Phys. Vol. 5, 1985, pp. 56-75. 57. J.D. Sherwood, Erosion and instability of a colloidal filter cake. Europhysics Lett., 4 (1987) 1273-1278. 58. K.J. Rushak, Coating flows. Ann. Rev. Fluid Mech., 17 (1985) 65-89. 59. L.E.Scriven, Physics and applications of dip coating and spin coating, in: Better Ceramics Through Chemistry III. MRS, 1988, pp. 717-729. 60. A. Prins, Liquid flow in foams as affected by rheological surface properties: a contribution to a better understanding of the foaming behaviour of liquids, in: J.P. Hulin, A.M. Cazabat, E. Guyon, F. Carmona (Eds.), Hydrodynamics of Dispersed Media. Elsevier, Amsterdam, 1990, pp. 5-15. 61. J.J. van Rossum, Viscous lifting and drainage of liquids. Appl. Sci. Res., A7 (1958) 121-144. 62. V.G. Levich, Physicochemical Hydrodynamics. Prentice Hall Englewood Cliffs, NJ, 1962. 63. J.A. Tallmadge and C. Gutfinger, Entrainment of liquid films, Drainage withdrawal and removal. Ind. Eng. Chem., 59 (!967) 19-34. 64. R.F. Probstein, Physicochemical Hydrodynamics, An Introduction. Butterworths, Boston, 1989. 65. D. Leighton and A. Acrivos, The shear-induced migration of particles in concentrated suspensions. J. Fluid Mech., 181 (1987) 415-439. 66. B. Kuhlmann, Emailschlickern hinsichtlich ihres Beschichtungsverhaltens von Stahlblech, Thesis Technische Universit~it Clausthal, 1985. 67. J.C. Berg (Ed.), Wettability. Surfactant Science Series 49, Marcel Dekker, New York, 1993. 68. T.C. Patton, Paint Flow and Pigment Dispersion, A Rheological Appr~ch to Coating and Ink Technology. Wiley, New York, 1979. 69. R.E. Johnson, Jr. and R.H. Dettre, Wetting of low energy surfaces, in: J.C. Berg (Ed.), Wettability. Surfactant Science Series 49, Marcel Dekker, New York, 1993, pp. 1-73. 70. J.C. Berg, Role of acid-base interactions in wetting and related phenomena, in: J.C. Berg (Ed.), Wettability. Surfactant Science Series 49, Marcel Dekker, New York, 1993, pp. 75-148. 71. P.G. de Gennes, Wetting: Statics and dynamics. Rev. Mod. Phys., 57 (1985) 827-863. 72. A. Sharma and E. Ruckenstein, Energetic criteria for the breakup of liquid films on nonwetting solid surfaces. J. Colloid Interface Sci., 137 (1990) 433--445. 73. H.S. Kheshgi and L.E. Scriven, Dewetting: Nucleation and growth of dry regions. Chem. Eng. Sci., 46 (1991) 519-526. 74. C. Redon, F. Brochard-Wyart and F. Rondelez, Dynamics of wetting. Phys. Rev. Lett., 66 (1991) 715-718. 75. T.A. Smith, Organic binders and other additives for glazes and engobes. Trans. Br. Ceram. Soc., 61 (1962) 523-549. 76. G.Y. Onoda, The Rheology of Organic Binder Solutions, in: G.Y. Onoda and L.L. Hench (Eds.), Ceramic Processing Before Firing. Wiley, New York, 1978. 77. J.E. Glass (Ed.), Water Soluble Polymers. Beauty with Performance. Adv. Chem. Series 213, ACS Washington DC, 1986. 78. E.A. Bekturov and R.E. Khamzamulina, Solution properties of water soluble nonionic polymers. JMS Rev. Macromol. Chem. Phys. C27 (2) (1987) 253-312.
6 -- PREPARATIONOF ASYMMETRICCERAMICMEMBRANESUPPORTSBYDIP-COATING
225
79. E. Carlstrom, in: R.J. Pugh and L. Bergstr6m (Eds.), Surface and Colloid Chemistry in Advanced Ceramic Processing. Surfactant Science Series 51, 1994, pp. 257-262. 80. D.J. Shanefield, Organic Additives and Ceramic Processing. With Applications in Powder Metallurgy, Ink, and Paint. Kluwer, Boston, Dordrecht, London, 1995. 81. E. Ringenbach, G. Chauveteau and E. Pefferkom, Adsorption of polyelectrolytes on soluble oxides induced by polyion complexation with dissolution species. J. Colloid Interface Sci., 161 (1993) 223-231. 82. A.W.M. de Laat and W.P.T. Derks, Colloidal stabilization of BaTiO3 with poly(vinyl alcohol) in water. Colloids Surf., A71 (1993) 147-153. 83. J. Ferguson and Z. Kemblowsky, Applied Fluid Rheology. Elsevier Applied Science, London and New York, 1991, 184 pp. 84. A. Prins, Liquid flow in foams as affected by rheological surface properties: a contribution to a better understanding of the foaming behaviour of liquids, in: J.P. Hulin, A.M. Cazabat, E. Guyon, F. Carmona (Eds.), Hydrodynamics of Dispersed Media. Elsevier, Amsterdam, 1990 pp. 5-15. 85. G.J.M. Janssen, W.J. Soppe and B.C. Bonekamp, The colloidal filtration step in the preparation of mesoporous ceramic membranes: a computer simulation. J. Colloid Interface Sci., 172 (1995) 161-170. 86. J. Cesarono III and I.A. Aksay, Processing of highly concentrated aqueous c~-alumina suspensions stabilized with polyelectrolytes. J. Am. Ceram. Soc., 71 (1988) 1062-1067.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 7
Sol-gel chemistry and its application to porous membrane processing Christian Guizard Laboratoire des Mat4riaux et Proc4d4s Membranaires (UMR 5635 CNRSENSC-UMII), Ecole Nationale Sup~rieure de Chimie, 8, rue de I'Ecole Normale, 34053 Montpellier cedex 1, France
7.1 INTRODUCTION The sol-gel process is one of the most appropriate methods for the preparation of functional oxide layers. Two sol-gel routes are generally described in the literature [1]. One is based on colloid chemistry in aqueous media, the other has to do with the chemistry of metal organic precursors in organic solvents, both being able to produce porous materials. These two routes can be used to prepare supported ceramic membranes with a common issue: how the porous structure is influenced by the different steps involved in the process, even the very first stage of precursor chemistry. The general method for membrane preparation through sol-gel processing is shown in Fig. 7.1. The first stage in the sol-gel process consists in the preparation of a sol using molecular precursors, either metal salts or metal organics. In both cases condensation reactions occur at the sol stage with formation of colloids or clusters which collide at the final stage to form the gel. In the case of membrane formation, it is important to note that coating of the active layer must be carried out at the sol stage with a rheological behaviour adapted to the porous substrate chosen as the membrane support. The presence of dust particles as well as a partial gelation in the sol must be avoided in order to prevent the formation of defects and pinholes in
228
7 -- SOL-GELCHEMISTRYANDITSAPPLICATIONTOPOROUSMEMBRANEPROCESSING I COLLOIDAL ( Me~!salt 1 ROUTE 1 . . ...... . . /Metal.Oorrganic . . . . . . . . ... \
(POLYMERIC ~, ROUTE
Precursors i
o
i
i
wa
rganic
,
i
media
:,o" o "-,.
~(
SOL'
),'a~':
i
._,
._.
"O,~,',,O.,', '
[~,col,oidalparticles) ',lw,' :.~[;:o'O,v'..'~d,~) ''" i
membrane coating
i
i
0
('COLLOIDAL) GEL )
fPoLYMERIC) ~, GEL ] i
,
.
ybrid organic-inorganic membran
t ~ (pure Sol~~i=eGmbrane)I Fig. 7.1. Diagramof the two sol-gel routes used in inorganicmembranepreparation.
the membrane. The drying and sintering steps will determine the nature of the membrane. A drying treatment performed in an intermediate temperature range (80-350~ results in a material containing residual organics. Pure inorganic membranes are generally obtained above 350~ after organic groups and residual carbon have been burned out. Finally, the consolidation of the membrane will be performed through viscous or conventional sintering depending on the amorphous or crystalline structure of the membrane material. During sol-gel processing of inorganic membranes, sols and gels evolve in a different way depending on the category of precursors used in the process. This evolution has a great influence on the porous structure of the final membrane materials.
7-
SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
7.2 P O R E F O R M A T I O N
229
IN SOL-GEL DERIVED CERAMIC MEMBRANES
Depending on the method used for membrane processing, colloidal or polymeric, two main structures of gel layers before drying and sintering can be described according to the literature: - physical gels in which steric or electrolytic effects in the sol dominate gel formation. The main characteristic of this type of gel is the way in which individual particles can be arranged during the process. These gels are concerned with aqueous media; - polymeric gels in which the relative rates and extents of chemical reactions (including polymerization) are critical for gel formation. In this case organic media are preferred.
7.2.1 Packing of Colloidal Particles When particulate gel layers are produced, the corresponding sols contain individual particles surrounded either by a steric barrier or an electrical double layer responsible for interparticle repulsion and sol stability [2]. In particular, the evolution of metal oxide colloidal suspensions in the presence of electrolytes can be described according to the DLVO theory [3,4]. The degree of aggregation of metal oxide particles in the sol is determined by the height of the potential barrier (zeta-potential) which results from an electrical double layer around the individual particles. This double layer is produced by acid-base reaction (peptization) at the particle/water interface due to the amphoteric behaviour of metal oxide surfaces. In this way, the intensity of the repul~sion-fokces depend essentially on the pH and the nature and concentration of the electrolyte. As shown in Fig. 7.2, the intensity of the interaction forces between particles has a direct effect on the residual porosity of the gel layer resulting from coating of the corresponding sol. A strong steric effect or a high repulsion barrier between particles at the sol stage provides a dense packed bed of particles during gel formation close to a perfect arrangement of spheres (green density d = 0.76 dth ). In this case membrane materials with a rather low porosity (< 30%) and capable of readily densifying at high temperature are generally obtained. For weaker interactions a partial aggregation of particles can occur in the sol. In this case a high porous volume is obtained in the gel layer and consequently in the resulting membrane material. These two effects have been exploited in the preparation of ceramic membranes exhibiting mesoporous structures [5-7]. However, sintering temperature is a second parameter which can affect the porous structure of ceramic membranes. In the case of membranes exhibiting a crystalline structure obtained from particulate sols, a grain growth phenomenon is generally observed responsible for an increase of pore size corresponding to the evolution of residual voids between the sintered particles.
230
7 -- SOL-GELCHEMISTRYAND ITS APPLICATIONTO POROUSMEMBRANEPROCESSING
/-.._stable sol
ilQ:, .... ] ",..-... ,,-o:i~..
coast
(a)
stronglycharged particle / ~ partially aggregated sol S
layer with a high porosity
COATING
charged
(b)
Fig. 7.2. Influence of colloidal sol stability on the porous structure of gel layers: (a) stable sol with non-aggregated particles; Co) partially aggregated sol with weakly charged particles.
7.2.2 Aggregation of Clusters Polymeric gel layers are produced through sequential polymer growth in the sol, providing a three-dimensional network at the gel stage. For polymer gels, the network gradually collapses under removal of solvent resulting in additional crosslinking as unreacted hydroxyl and alkoxy groups come into contact. If phase separation does not occur, it is expected that polymer gels will continue to collapse and crosslink until they can resist the compressive action of surface tension (at which point porosity is created). During this consolidation process the size and the structure of the pores created in the coated layer will depend on the structure of individual clusters resulting from polymerization. It has been shown in the literature [8,9] that either weakly or highly branched clusters can be obtained as a result of the conditions of hydrolysis and condensation
7 -- SOL-GELCHEMISTRYAND ITS APPLICATIONTO POROUS MEMBRANEPROCESSING
/--
231
low branched cluster ukramicroporous layer COATING iv
y///////, '//,i sP~176s
.•r..-
highly branched cluster
a)
microporous layer .--, .....................................
COATING
'////////,///, sP~176
(b)
Fig. 7.3. Influence of cluster structures in polymeric sols on the porosity of coated layers: (a) packing of interpenetrated low branched clusters; (b) packing of non-interpenetrated highly branched clusters.
(precursor concentration, acid or basic catalysis) applied to molecular precursors at the sol stage. As shown in Fig. 7.3a, low branched clusters can interpenetrate during gel collapse leading to a microporous material due to a highly compacted material. On the other hand, highly branched clusters, Fig. 7.3b, are not able to interpenetrate due to steric hindrance. In thiscase residual porosity will remain between collapsed clusters leading to micro or mesoporous materials. These phenomena have been seen mainly on silica membranes prepared from silicon alkoxides.
7.2.3 Utilization of Template Agents A more recent approach to prepare tailor-made porous structures in inorganic membranes has been to use the template effect of organic groups or organic molecules incorporated in gels during sol-to-gel transition. The role of the organic part is to generate a residual porosity after they are burned out under heat treatment. In the present case a pore size based on the size of the organic template molecule is anticipated. Porous volume and pore size (micro, meso or macropores) can be affected by the nature and the size of the templates. Preparations of membranes using this technique have recently been reported in the literature [10,11]. The creation of a residual porous structure by burning out organic templates incorporated in a gel layer is schematized in Fig. 7.4.
232
7 ~ SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
tailor-made b-a porous layer I /
template agents inserted in the \ :::" ":':':" i:i"
..~:~:i:::"
"
.::""
";:::::'-.
"
::.3::
AT
\T/ .. :~
,:::::::::
..:':::.$~'-.':':
.
:~.~.
:-:::::.:.-.
-.:.:.:.
'..'..//////////////.. Fig. 7.4. Formation of tailor-made p o r o u s m e m b r a n e materials by insertion and thermal decomposition of template agents in a gel layer.
7.3 C O L L O I D A L
SUSPENSIONS
TO PREPARE MESOPOROUS
MEMBRANES
7.3.1 Chemistry of Colloidal Sols The formation of colloidal particles can be obtained from hydrolysis and condensation of metal salts in aqueous media. The aqueous chemistry of metal salts is quite complicated owing to the occurrence of hydrolysis reactions which convert the ions to new ionic species or to precipitates. When dissolved in water a metal cation M z§ becomes solvated by the surrounding water molecules according to Z+
Three kind of ligands must then be considered in a non-complexing aqueous medium: aquo ligands (OH2), hydroxy ligands (-OH) and oxo ligands (=O). The formation of these three types of ligand can be summed up in a qualitative way using a "charge-pH" diagram as shown in Fig. 7.5. Such a diagram shows that low-valent cations (z < +4) give rise to aquohydroxo a n d / o r hydroxo complexes over the whole range of pH, while high-valent cations (z > +5) form oxo-hydroxo a n d / o r oxo complexes over the same range of pH. Tetravalent cations are on the borderline, and therefore lead to a large number of possible precursors. Starting from these precursors, condensation in water media operates following a very fast kinetic concerned with two reactions: - olation (nucleophilic substitution) M-O~-H + M~+--O6+I--I2~ M - O H - M + H20 - oxolation (nucleophilic addition with or without an OH leaving group) M - O H + H O - M --~ M - O - M + H20
233
7 - - SOL-GEL CHEMISTRY A N D ITS A P P L I C A T I O N TO POROUS MEMBRANE PROCESSING
8
---0-- aquo/hydroxo _-,,43-- hydroxo/aquo
7 6
0 2-
5 Z4
OH
3 2
H20
1
0
9
0
I
0
7
14
pH Fig. 7.5. Charge-pH diagram giving the existing domains in aqueous media for substituted ionic species. According to the type of precursor previously described, condensation with oxo-ions can only occur via addition w h e n the precursor is unsaturated while condensation cannot occur with aquo ions because no entering group is available. Following the c h a r g e - p H diagram means that it is necessary to move into the hydroxo d o m a i n in order to get condensed species. One can see that p H is a key parameter for precursors processed in aqueous media. This is a helpful model for selecting proper precursors and predicting condensation reaction in aqueous media [12]. Metal cations such as aluminum, titanium or zirconium currently involved in the preparation of ceramic m e m b r a n e belong or can be shifted through p H variatign i n the hydroxy complex area of the c h a r g e - p H diagram. -....... The normal w a y to obtain colloidal sols from oxide precursors is therefore a two-step process. In the first step, a precipitate of hydroxylated condensed species is formed from hydrolysed precursors. As described below, it can be seen that hydroxylated species capable of further condensation and precipitation in aqueous media can also be obtained from hydrolysis of metal alkoxides with excess water. In the second step this precipitate is transformed into a stable sol through a peptization reaction using basic or acid electrolytes. After adding appropriate organic binders, if requested, this sol can be directly used to form supported membranes.
7.3.2 Examples of Membrane Preparation G a m m a - a l u m i n a membranes were the first and most investigated mesoporous membranes to follow the colloidal preparation method. Based on a sol-gel process developed by Yoldas [13], a boehmite sol can be prepared by hydrolysis
234
7 ~ SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
2 403
~"
(1) Dry (~S0~ (2) T s = 5 0 0 ~ (3) Ts = 6 0 0 ~ (4) Ts = 7 8 0 ~
30
~ 2o
10
|
]
e
II ~
3
!
I
5
!
R (nm)
a Fig. 7.6. Porous characteristicsof ~,-aluminamembranesprepared by the Yoldas's method. (a) Pore size distributions versus sintering temperature from Ref. [15]; (b) opposite page.
of a l u m i n u m s-butoxide in water followed by peptization of the a l u m i n u m hydroxide with inorganic acids. The corresponding dried and fired gel body exhibits pore diameters of 4-10 nm with a narrow pore size distribution (+ 1 nm). This method was adapted in the 1980s for the preparation of crack-free supported membranes. Leenaards et al. [14,15] first published the characteristics of unsupported and supported ~/-alumina membranes using flat supports. Tubular supports were further used by Larbot et al. [16] to prepare 7-alumina membranes with almost the same characteristics. Figures 7.6a and b show pore size distribution and pore size evolution versus sintering temperature for these membranes. Since then these ~,-alumina membranes have been very popular with a number of scientists involved in the preparation and characterization of ceramic membranes [17-21]. However these membranes suffer from a poor chemical stability at high p H and a structural evolution under thermal conditions, which is w h y they have not been much applied at the industrial level although they are commercially available. More recently, methods based on the introduction of alkaline and rare earth oxide have been
7 - - SOL-GEL CHEMISTRY A N D ITS APPLICATION TO POROUS MEMBRANE PROCESSING
15
235
1200~ 55 nm ! ! ! ! !
I
10 I I
~
S
jr
f
500
T (~
1000
b
Fig. 7.6. (b) pore size evolution versus sintering temperature from Ref. [16]. proposed [22-24] to improve the thermal stability of y-alumina membranes. In the work by Lin et al. [22,23] the permeability of a pure alumina membrane was shown to steeply increase up to 1000~ due to pore growth versus heat treatment temperatures. Above 1000~ ~-~ transformation is observed. In comparison with pure alumina membranes, La-doped supported membranes were prepared showing that the top layer retains a monopore distribution after sintering at 1200~ Chai et al. [24] prepared membranes with a composition of hexaaluminate (BaA112019 and LaAlllOls). A small increase of H 2 permeability was noted for these membranes up to 1000~ but it drastically increased above 1000~ due to the crystallization of hexaaluminate. Since then, other colloidal oxide systems have been investigated in order to prepare ceramic mesoporous membranes designed for ultrafiltration. The preparation of an electronically conductive membrane from a RuO2-TiO2 mixed oxides sol and the application to an electro-ultrafiltration process [25,26], as well as the preparation of titania and zirconia ultrafiltration membranes [27], have been described following a colloidal process in which a partial destabilization of a metal oxide colloidal suspension is used to produce top layers with different pore size and pore volume in the mesoporous range. In agreement
236
7 ~ SOL-GEL CHEMISTRY A N D ITS A P P L I C A T I O N TO POROUS M E M B R A N E PROCESSING
with the DLVO theory, the pH, the ionic strength and the nature of the electrolytes in the colloidal suspension were pointed out as important parameters to control the degree of aggregation of particles. Porosity control was achieved from packing density of the particles obtained in the top layer during the coating process. Later on, Xu et al. [28] published almost the same results on particulate zirconia and titania membranes prepared in the same way. Due to their improved stability compared with y-alumina, these titania and zirconia membranes have been used to a greater extent in ultrafiltration processes. Recently, other examples of the colloidal method applied to the synthesis of mesoporous membranes have been given in the literature. A zirconia membrane with a pore diameter of 4 nm was obtained by Etienne et al. [29] from a particulate sol. This sol was synthesized by reaction of zirconium oxychloride with oxalic acid resulting in zirconium oxalate particles. Peptization of these particles was obtained in situ thanks to the HC1 released in the aqueous suspension during formation of zirconium oxalate. Kusakabe et al. [30] reported on the preparation of a BaTiO3 membrane from a colloidal sol. A macroporous alumina support was impregnated with this sol in order to produce a mesoporous membrane material. Nevertheless thermal stability of titania and zirconia membranes as well as other crystalline metal oxide membrane materials remains problematic owing to structure and porosity evolution under thermal and hydrothermal conditions [31]. Kumar et al. [32] showed that a titania membrane exhibits a higher anatase-rutile transformation temperature (slower rate of transformation) when supported on a porous substrate compared to an unsupported one. They also compared titania-alumina composite membranes to pure titania membranes [33,34].The presence of alumina in the membranes improved the thermal stability of the porous texture by retarding the anatase to rutile phase transformation and grain growth of the anatase phase. As an example, pure unsupported titania membranes lose their porosity completely after calcination at 600~ for 8 h, whereas a titania-50 wt% alumina composite membrane retained a porosity of ~40% even after calcination for 30 h at 800~ . Amorphous silica has also been mentioned as a starting metal oxide material for the preparation of particulate mesoporous membranes. These membranes were prepared from commercial sols, Ludox (DuPont) or Cecasol (Sobret), and coated on a macroporous ~-alumina support [35]. In contrast to crystalline membrane materials such as alumina, titania or zirconia, the evolution of pore size with temperature of amorphous silica membranes was revealed to be more sensitive to drying conditions than to firing temperature (Table 7.1). When heat-treated for several hours at 800~ the silica top layer transformed from an amorphous state to cristobalite. Except for silica, one common feature of the membranes described in this paragraph is their crystalline structure resulting from sintering of individual
7 - - SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
237
TABLE7.1 Mean pore diameters of SiO2membranes for different drying and firing temperatures [35] Drying temperature (~
Firing temperature (~
Mean pore diameter (nm) _+0.5nm
10 30 30 30 30 50 90
600 500 600 700 800 600 600
6.0 8.2 7.6 7.5 8.4 8.8 10.2
grains. So if aggregation is avoided at the sol stage, the pore size of the membrane is controlled by the particle size, larger particles yield larger mesopores. The final size of the pore can be adjusted by fixing the sintering temperature. The advantage of this approach is that the porosity of the membrane is independent of the particle size. For example, random dense packing of monosized particles always results in about 33% porosity. There are, however, negative effects of the colloidal approach regarding the preparation of microporous membranes. Consolidation of a ceramic material through conventional sintering of oxide particles (A1OOH, A120 3, TiO2, ZrO2) is generally obtained for temperatures and heating times resulting in grain growth. In most cases the size of these grains at the end of the sintering process is too large to produce a microporous structure.
7.4 I N O R G A N I C POLYMERS TO PREPARE M t C R O P O R O U S M E M B R A N E S
7.4.1 Formation and Aggregation of Clusters A quite different approach from that of colloidal sols in the preparation of sol-gel derived membranes utilizes polymeric sols. In this category of sols the dispersed phase results from the hydrolysis and condensation of metal organic precursors in organic media. In most cases this process deals with the polymerization of metal alkoxides in alcohol according to the following reactions: - hydrolysis M(OR)n + x H20 ~ M(OR)n_x(OH)x + x ROH
238
7 ~ SOL--GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
- condensation ( O R ) n _ I M - - O R + H O - M ( O R ) n _ I ( O H ) x _ 1 ---->
(OR)n_IM-O-M(OR)n_x(OH)x_ 1 + R O H
or/and (OH)x_I(OR)n_IM-OH
+ HO-M(OR)n_I(OH)x_I
--->
(OH)x_I(OR)n_IM-O-M(OR)n_x(OHx_ 1 + H 2 0
Silicon alkoxides exhibit very slow hydrolysis and condensation reactions compared with other alkoxides of aluminum, titanium or zirconium generally used for membrane preparation. Accordingly, acid or basic catalysts are used in the case of silicon alkoxides while methods for the control of hydrolysis are advisable with transition metal alkoxides [36,37]. In such reactive media branched clusters which do not contain fully condensed metal oxide cores are formed by kinetically limited growth processes. The structure of these clusters can be described using the fractal concept in which a mass fractal dimension D relates the cluster mass M to its radius rc according to Moor D
A concept of mutual transparency or opacity based on the relative evolution of fractal dimension and radius of the clusters has been developed by Mandelbrot [38]. The tendency of fractal systems to interpenetrate is inversely related to the mean number of intersections M1,2 of two mass fractal objects of size rc and mass fractal dimensions D1 and D2 placed in the same region of space of dimension d: _ DI+D2-d M 1 , 2 oc r c
In three-dimensional space (d = 3) and for two mass fractal objects of the same mass fractal dimensions D, there is a crossover point at a value of D - 1.5. If D < 1.5, the probability of intersection decreases with infinity as rc increases. Thus the structures of the clusters are mutually transparent and can interpenetrate. Due to clusters interpenetrating, an extremely fine texture is expected for the membranes prepared from these sols. Alternatively if D > 1.5 the probability of intersection increases with rc and clusters becomes mutually opaque. In this last case porosity will increase with rc as rc3-D 9From the preceding analysis a strategy for the preparation of microporous and ultramicroporous membranes has been described by several authors [39,40]. The prerequisite for preparing such membrane materials is to use sols containing clusters with controlled size and low mass fractal dimension. During deposition, clusters should interpenetrate and then, by capillary forces generated during drying, they will form an almost dense network with a residual microporosity. This type of material can be used
7 - - S O L - G E L C H E M I S T R Y A N D ITS A P P L I C A T I O N T O P O R O U S M E M B R A N E P R O C E S S I N G
239
to prepare microporous top layers and also for modification of mesoporous top-layers to reduce pore size down to the microporous range [41,42]. This is an alternative method to CVD processes proposed in the literature [43,44].
7.4.2 Examples of Membrane Preparation A first approach to preparing microporous layers has been described by Ulhorn et al. [45] based on the modification of y-alumina films. In this work silica and titania sols were prepared from alkoxide precursors under acidic catalysis conditions. A microporous texture was evidenced for a silica layer using a y-alumina mesoporous intermediate layer. Regarding titania systems, firing temperatures above 350~ caused pore growth due to formation of anatase. In fact, the microporous texture is maintained in the case of silica and not for titania because of the amorphous structure generally found for sol-gel derived silica materials. Later on, the formation of microporous materials from polymeric gels was better explained in terms of fractal dimension of the clusters involved in the sol-to-gel transition during membrane preparation. The preparation of microporous (pore radius < I nm) silica supported membranes has been described in the literature starting from silicon alkoxide derived sols [40,46,47]. The silicate sols were prepared using a two step acid-catalyzed hydrolysis of tetraethoxysilane under pH conditions where the condensation is low producing polymers of low fractal dimension. Based on the concept of mutual transparency previously discussed these polymers can interpenetrate during film deposition to provide amorphous layer with residual micropores. Two important parameters are pointed out by Brinker et al. in Ref. [46]. Aging of the sols caused both the polymer size and mass fractal dimension to increase (D = 1.0 after 2 h and 1.7 after 35 h). At low aging time (t/tg = 0.05), the acid catalyst concentration had a dramatic effect on the deposited membrane thickness and permeability. Using 1 N HC1 a discrete, very uniform membrane layer was obtained, while lower acid concentration (0.44 N) resulted in polymer penetration and filling of the support pores rather than the deposition of a discrete layer. Attention has also been focused by de Lange et al. [48] on the synthesis of amorphous silica and binary systems such as SiO2/TiO2, SiO2/ZrO2 and SiO2/ A1203. Acid catalysis of metal alkoxides with HNO 3 was used in this case. Three synthesis routes were used for the preparation of binary membrane materials" - a single-step prehydrolysis of silicon alkoxide followed by the addition of respectively the Ti-, Zr- or Al-alkoxide in alcohol, - a two-step hydrolysis for the synthesis of SiO2/TiO2, and SiO2/ZrO2 sols with a second addition of water and acid after the addition of titanium or zirconium alkoxides in the same conditions as for the previous route, - a separate prehydrolysis of silicon and titanium alkoxides followed by mixing of the two obtained sols to form a binary sol.
240
7 m S O L - G E L C H E M I S T R Y A N D ITS A P P L I C A T I O N T O P O R O U S M E M B R A N E P R O C E S S I N G
The conclusion arising from these experiments was that homogeneous polymeric silica-based binary sols can be made with the addition of a second component up to 30 mol%. Initially, the fractal dimension (~1.4) and the gyration radii (--2 nm) of the polymers were found to be low enough to obey the concept of mutual transparency. More details are provided in Chapter 8 on the preparation of such microporous membranes for gas separation.
7.5 THE CONCEPT OF N A N O P H A S E CERAMICS APPLIED TO THE P R E P A R A T I O N OF M I C R O P O R O U S M E M B R A N E S
The previous concept of cluster aggregation yielding microporous structures is only applicable to amorphous materials such as silica or mixed oxides systems. In ceramic membranes exhibiting a crystalline structure, pore sizes are related to the size of the individual grains forming the ceramic. For ceramic nanofilters, pore diameters resulting from grain sintering must be in the micropore range, smaller than 2 nm. Two main conditions must exist in order to prepare nanophase ceramics exhibiting a connected microporosity with a narrow pore size distribution. The first is to preserve individual grains of less than 10 nm to the sintered ceramic, the second is to prevent particle aggregation at the sol stage responsible for the formation of larger sintered grains leading to a residual mesoporosity. Porous structures down to the nanometer range can be attained by the sol-gel process. New strategies to prepare microporous ceramic membranes using either the colloidal route or the polymer route have been proposed by Guizard et al. [49,50]. Usually sol-gel processing of colloidal particles leads to mesoporous materials. Therefore new chemical aspects, recently developed in sol-gel science, can be advantageously applied with the aim of a microporous structure formation in ceramic membranes. This can be achieved by taking into account either the crucial role of counter ions in the growth of colloids in aqueous media [51] or the role of chelating agents as blocking functional groups in condensed species obtained in organic solvents [52]. Based on these considerations, ceramic membranes are described below which show the interest of sol-gel processing in the preparation of microporous top layers from nanophase ceramics.
7.5.1 Formation and Coating of Aqueous Nanoparticulate Sols In sol-gel processing, particulate sols of hydrous metal oxides can be formed using a peptization reaction to prevent aggregation of the primary particles. This is the case for ~,-alumina obtained from boehmite. Usually mesoporous membranes with a pore diameter down to 5 nm are easily elaborated from commercial boehmite using nitric acid as the peptization agent. Larbot et al. [53]
7 m SOL--GEL CHEMISTRY A N D ITS A P P L I C A T I O N TO POROUS MEMBRANE PROCESSING
241
described an improved process of synthesizing a microporous ~/-alumina membrane material according to the results on alumina gel formation previously reported by Yoldas [54]. The size of individual crystallites in the membrane was influenced by the water/alkoxide and the nitric acid/alkoxide molar ratios and by the pH and the concentration of alumina in the sol. These parameters were adjusted in order to form a very thin (~0.4 ~tm) supported microporous layer on mesoporous ~/-alumina used as intermediate layer. Boehmite precipitation was obtained from aluminum butoxide precipitated at 80~ with an excess of water, [water]/[alkoxide] = 100. A stable sol was formed at a pH of 3.9. Indeed acidic sols with pH < 3.4 led to infiltrated layers after coating. Regarding alumina concentration, the best results were obtained starting from a 2.4 wt% boehmite sol and then concentrating this sol to 35 vol% by evaporation. After coating and drying, the membrane was consolidated at 450~ The pore size was measured by N2 adsorption and membrane cut-off was determined using model solutes with small molecular weights [55]. A membrane cut-off of 450 Dalton was consistent with a pore diameter of about 1 nm calculated from the HorvathKawasoe model. Zirconia must be mentioned as a well-adapted membrane material for the preparation of ceramic nanofilters. Different zirconia polymorphs (tetragonal, monoclinic, cubic) can be encountered starting from the amorphous state obtained at room temperature. In order to obtain microporous membrane materials the two main parameters to deal with are phase stability of zirconia and crystallite size. According to basic principles established on the role of counter ions in the condensation of hydroxo complexes [51], nanoparticulate sols were obtained from ZrOC12 by Guizard et al. [49] by substituting chloride ions with nitrate ions. In a neutron diffraction study, Garcia [56] showed that crystallization kinetics and growth of particles obtained from ZrOC12 derived sols were related to the transformation sequences under different firing atmospheres (air, N 2 and H 2 / N 2 ) . Tetragonal crystallites of less than 6 nm were obtained at 360~ from the amorphous state up to the temperature of tetragonal/monoclinic phase transition, beyond which crystallite size changed abruptly from 6 nm to more than 12 nm. As shown in Fig. 7.7, depending on the firing atmosphere, this transition temperature was shifted from 550~ under air to 600~ under N 2 or H 2 / N 2 atmosphere. In each case individual tetragonal particle sizes were maintained under 6 nm when the firing temperature did not exceed the transition temperature. Homogeneous quite spherical consolidated particles of about 5-6 nm in size were observed by TEM on samples fired at 500~ [50]. The microporous structure of these samples was shown by N 2 adsorption measurements with pore sizes of less than 2 nm. Above the transition temperature micropores transformed into mesopores due to grain growth. The interest of these ceramic nanofilters for the separation of small molecules and ions in liquid media is stressed in Chapter 12.
242
7 -- SOL-GEL CHEMISTRYAND rrs APPLICATION TO POROUS MEMBRANE PROCESSING 70
"'
e
0< 6O 0
t~ [-
50
00
0
B
e
ee
~ 9
o*
9
E] E]E!E] E] EIo4[3e - , ~ _ % O ~ e
~
o~ ~
40
r~
E!
3O 30O
9
E!
qPqpd :~ o [] lab
[]
AIR
9
H2/N2
0
I
400
"
I
500
"
N2 I
600
"
700
TEMPERATURE (~ Fig. 7.7. Influence of firing atmosphere on the evolution of crystallite size versus temperature in the case of a tetragonal zirconia m e m b r a n e material [56].
Another example of the importance of counter ions on particle size was given by Chanaud et al. [57] on the preparation of homogeneous lanthanum chloride aqueous sols. When ammonia was added to a lanthanum chloride stirred aqueous solution, lanthanum hydroxide a n d / o r basic salt intermediate species were obtained in the form of an opalescent sol (pH ~ 8) yielding mesoporous materials. In order to prepare microporous materials, a precursor modification (by acetic acid) was carried out, leading to soluble acetate species at the working p H N 8. The addition of ammonia to the modified solution produced clear sols containing smaller particles than those obtained in the first preparation method. In a further work, these sols were used to prepare microporous membranes [58]. A binder and a plasticizer were added in the sol allowing a good behaviour of the casted films during the drying and firing treatments. In this case polyvinylic alcohol (PVA) fulfilled these functions and led to crack-free ceramic layers. The resulting casted gelled thin layer was then directly placed in an oven at 110~ in order to perform a rapid drying of the film and avoid crystallization of stable carbonate species. The crack-free dried film was heattreated at 800~ to convert it to a lanthanum oxychloride porous thin film. The mesoporous LaOC1 layer (prepared without acetic acid) was used as a support for the deposition of the microporous membrane. TEM observation of the microporous layer revealed small particles (about 6 n m in size) embedded in a denser phase yielding a microporous texture stable to 800~ The catalytic performance of such a material for oxidative coupling of methane has been described as a function of preparation conditions [59].
7 - - S O L - G E L CHEMISTRY A N D ITS A P P L I C A T I O N TO POROUS M E M B R A N E PROCESSING
243
7.5.2 Formation and Coating of Organic Nanoparticulate Sols Transition metal alkoxides can also be used as precursors to synthesize organic particulate sols with a view to microporous membrane preparation [60]. In order to avoid the precipitation of inhomogeneous hydroxide particles during the hydrolysis step, the alkoxide reactivity can be modified either by strong complexing ligands like acetylacetone (acacH) or by strong mineral acid (HNO3). Yamamoto [61] was one of the first to mention the exothermic chemical reaction that occurs between alkoxide and acetylacetone. AcacH reacts with the alkoxides to form mixed complexes which have different physico-chemical properties and, more accurately, which are more difficult to hydrolyze than alkoxy groups [62] M(OR)4 + acacH ~ M(OR)3(acac) + ROH with M = Ti or Zr This ligand acts as a functionality blocker when substoichiometric hydrolysis ratios are used. A ratio acacH/M greater than I prevents precipitation and leads to stable colloids or gels. Consequently, with a good formulation choice, sols can be prepared in air without any precipitation. Preparation of zirconia and attempts on titania microporous layers have been described by Julbe et al. [63] starting from acacH complexed alkoxide derived sols. Either titania or zirconia intermediate mesoporous layers have been used as supports for these membranes. The supported layers obtained after sintering at 500~ exhibited crystallized structures (anatase for titania and tetragonal metastable form for zirconia) and revealed a very fine texture when observed-by TEM. In the case of Ti with a Ti/AcacH ratio = 1, grains with defined faces can be observed whose size is about 20 nm. In the case of Zr with a Zr/acac ratio = 2, a finer texture was obtained, with a mean grain size of about 4 nm. In both cases, powder X-ray diffraction (Sherrer's formula) has been used for the determination of an average individual crystal size which is in good accordance with particle size in a supported membrane observed by field emission scanning electron microscopy (FESEM (Fig. 7.8). From the previous example it has been shown that acacH is less efficient for titania than for zirconia in promoting the formation of ceramic nanoparticles of less than 10 nm. Another method has been used dealing with the modification of titanium alkoxides in strong acidic conditions. According to the work of Yoldas [64], it is possible to prepare clear solutions which contain oxide constituents in a soluble polymerized form and from which uniform and continuous glass-like oxide films can be deposited on substrates at relatively low temperatures. In order to obtain nanoparticulate sols, solutions were prepared by mixing a titanium alkoxide, an alcohol, water and a small amount of strong mineral acid, HNO3 [63]. Several parameters were carefully adjusted, in
244
7 - - SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
~ -
. '
.
~
. . ...
...
:.~,~"~U,"~.~
Surface image
* ~ ~
"
,
,
,
~
1
7
6
1
200
7
6
' nm
;~:
Cross-section image
. , . , . . . . , . , . .
200nm
Fig. 7.8.FESEMsurface and cross-sectionimages of the activelayer of a zirconia nanofilter prepared from a sol of acacH complexed zirconium alkoxide precursor. particular, in the case of t i t a n i u m based sols, the hydrolysis ratio: h = H20/Ti(OR)4, the acid ratio: a = HNO3/Ti(OR)4 and the equivalent oxide content of titanium oxide given by the TiO2/Ti(OR)4 ratio (5% wt in the present case). For a specific a m o u n t of acid, stable and clear sols were synthesized. To prevent precipitate formation or self condensation, a n u m b e r of (OR) groups were kept unreacted in the alkoxide molecule by adjusting h to a sufficient low value. The sol stability at low acid rate appeared to be greater than those prepared for higher acid rate (a = 0.3). For instance, several compositions such as (h = 0.08, 0.03
7 m S O L - G E L CHEMISTRY A N D r r s A P P L I C A T I O N TO POROUS MEMBRANE PROCESSING
245
mesoporous layer with the aim of producing a microporous final layer. They used an organic sol of zirconia nanoparticles of less than 2 nm for impregnation of a y-alumina supported layer exhibiting a pore diameter of about 10 nm. These nanoparticles were synthesized using a reverse micelle method in which particle formation resulted from diffusion and hydrolysis of zirconium tetrabutoxide inside the micelles. The reversed micelle solution was prepared from water and a surfactant molecule (dioleyl phosphoric acid) in isooctane. By repeating a dipping-firing cycle, the mesopores of the y-alumina supported layer were packed with the zirconia particles. An activated transport of H 2 and N 2 w a s seen in these membranes showing that a microporous texture has been attained by this process.
7.6 TAILOR-MADE POROUS MEMBRANES VIA TEMPLATES
CONTAINING SYSTEMS Up to now the preparation of ceramic membranes by the sol-gel process has been described on the basis of two main routes using colloidal or polymeric sols. The colloidal route has mainly been exploited in the ceramic membrane industry to produce supported ultrafiltration membranes. As has been emphasized previously, recent developments in sol-gel science have resulted in the preparation of microporous inorganic membranes from polymeric or nanoparticulate sols. A later concept in designing tailor-madeporous structures is based on the incorporation of organic template agents in gel structures. These template agents can be organic groups chemically linked to the gel network, isolated molecules or clusters trapped in the sol-gel matrix or self-assembled systems participating in the formation of structured gels. Whatever the nature of the templates used they must be eliminated by solvation, solvolysis or burned out during heat treatment providing a residual porosity. The innovative aspect of these templates is that they allow the porous structure of sol-gel derived membrane materials to be tailored. For silica gels a number of parameters have been demonstrated to have a large effect on the evolution of porosity and subsequently on the resulting silica materials [1]. Almost dense, micro- or mesoporous silica materials can be obtained depending on the experimental conditions in which hydrolysis and condensation reactions of silicon alkoxides are carried out. This is not the case for transition metal alkoxides which are very sensitive to hydrolysis. They do not allow the adaptation of sol-to-gel transition in order to obtain controlled porous textures. Some years ago special attention was paid to the utilization of amphiphilic systems as reactive media to control hydrolysis and condensation kinetics with transition metal alkoxides [37]. In a more recent work Ayral et al.
246
7 - - SOL-GEL CHEMISTRY A N D ITS APPLICATION TO POROUS MEMBRANE PROCESSING
[10] described the role of these amphiphilic systems in the preparation of silica membranes exhibiting an ordered microporosity. In particular surfactant molecules and related self-assembled systems have been used as template agents during sol-gel processing of these membranes. In the following, more details are given concerning the template effect of non-ionic and ionic surfactants on silica gel formation. Regarding the role of the amphiphilic media, both passive and active effects on pore formation have been shown. Other methods have been applied based on the insertion in the gel matrix of passive template agents such as chemically fixed organic entities or physically trapped polymer particles. Then the elimination of the templates leads to the formation of a porous structure in which a relation between pore size and physical dimension of incorporated templates is anticipated.
7.6.1 Utilization of Amphiphilic Media Based on the utilization of amphiphilic media, Julbe et al. [66] proposed a method dealing with the effect on residual microporosity of non-ionic surfactants, alkylaryl polyether alcohols TRITON (X = 1, 30) of different molecular weights added to tetraethoxysilane sols. On the basis of sol and gel characterizations (gelafion time, 29Si NMR, QELS, SAXS) the effect of surfactant chain length (X) and TRITON/TEOS ratio on related membrane materials have been explored. Surfactants are susceptible to interacting with silica oligomers derived from TEOS by van der Waals forces or by interaction with OH groups. Due to these interactions, it can be assumed the formation around the clusters of an organic shell made of surfactant molecules. The resulting steric hindrance limits further condensation of cluster during sol aging (Fig. 7.9). A reduced number of Si-O-Si bonds between the clusters explains the formation of more stable sols with weak intercluster connections. A remarkable result in this work is the influence of the size and the concentration of surfactant molecules on membrane porous texture. During heat treatment of the gel layer, the elimination of the organic shell around the particles produced a homogeneous microporous membrane consisting of nanometric distinct silica particles. Nitrogen adsorption experiments performed on these materials showed that nitrogen molecules (kinetic diameter = 3.96 A) do not penetrate the structure of the material prepared without adding surfactant molecules. When surfactants were used, Type I isotherms were obtained which are characteristic of microporous materials. Results obtained with the TRITON series are given in Table 7.2. The mean hydraulic pore radius RH (estimated by the MP method) varied between 3.2 and 7.0 ~. This parameter was slightly affected by the surfactant/ precursor ratio (SAA/TEOS) but increased with the surfactant chain length X.
7 - - SOL-GEL CHEMISTRY A N D ITS APPLICATION TO POROUS MEMBRANE PROCESSING
Microporous , silica material ,
[
247
Dense silica material
,,
Fig. 7.9. Schematic representation of the influence of added non-ionic surfactants on cluster growth in a TEOS derived polymeric sol. (a) TEOS alcoholic solution; (b) TEOSstandard sol; (c) aged sol in the presence of surfactant; (d) aged sol without suffactant [65]. Furthermore the pore size distribution becomes wider w h e n X increases. From these results it has been assumed that the surfactant acts as a template agent by increasing pore size w h e n increasing surfactant molecule size but does not allow to control pore size distribution in the material. According to the authors too long polyether chains (X = 30) for the surfactant must be avoided because this leads to a large pore size distribution. Another effect of surfactant addition was the increase of porous volume and specific surface area w h e n increasing X
248
7 - - S O L - G E L C H E M I S T R Y A N D ITS A P P L I C A T I O N T O P O R O U S M E M B R A N E P R O C E S S I N G
TABLE 7.2 Influence of the size and concentration of TRITON X surfactants on the porous structure of unsupported silica membranes [66] x
1 3 3 3 10 30
SAA/TEOS SBET
0.55 0.16 0.35 0.55 0.55 0.55
Porousvolume (cc/g)
Porosity RH distribution (/~)
(m2/g)
Total
Microporous
(%)
Range
Mean
400 250 470 500 728 800
0.217 0.121 0.197 0.249 0.510 0.541
0.210 0.105 0.190 0.239 0.480 0.514
32.2 20.9 30.1 35.5 52.7 54.2
3--7 3--6 3--6 3-6 5--9 4-10
3.9 3.2 3.4 3.7 6.3 6.0
and SAA/TEOS ratio. An o p t i m u m effect was reported for X = 3-10 and SAA/TEOS = 0.55 with values for the porosity between 35 and 50%, 96% for the microporous volume and a specific surface area exceeding 500 m2/g. Crack-free m e m b r a n e s 200-500 n m thick were prepared from this type of sol (Fig. 7.10). This is consistent with an expected double effect of surfactants at the sol stage. The role of amphiphilic molecules is the first to increase the contact surface between the liquid reactive media and the growing particles (effect of surfactant concentration). The second effect is on pore size which can be adjusted depending on surfactant chain length (effect of surfactant size). Another efficient w a y of obtaining monodispersity of the pore size distribution and control of the pore size is to develop materials with an ordered microporous texture. For polymeric gels, the use of an ordered m e d i u m acting as a template for the growing inorganic network has been considered by Ayral et al. [67,68]. The conditions for a successful process is that both liquid crystal and inorganic network can form in the sol and that this gelation m e d i u m must be easily eliminated without the collapse of the solid part of the gel. In the first approach the template effect of lamellar systems m a d e of non-ionic surfactants on sol-gel derived silica materials has been investigated which led to mesoporous textures [69]. The second type of system investigated deals with the use of hexagonal phases as a gelation m e d i u m for silica microporous gels [68]. These phases consist of hexagonal packing of micellar cylinders surrounded by an aqueous phase. The chemical composition of the sols corresponds to an isotropic region in the water-surfactant binary diagrams. This is consistent with w h a t was observed experimentally: the sols were initially very fluid and as easy to deposit as crack-free coatings. The ordered mesostructures developed during the gelation process. The process of silica polymerization and lyotropic hexagonal
7 - - SOL--GEL CHEMISTRY A N D ITS APPLICATION TO POROUS MEMBRANE PROCESSING
249
Fig. 7.10.FESEMimages of a microporous silicamembrane (X = 3, SAA/TEOS = 0.55) deposited on a mesoporous silica intermediate layer. (a) surface, (b) cross-section. phase formation are simultaneous. The formation of the liquid crystal phase seems to be induced by the formation of silicate oligomers near the cationic surfactant head groups. Thus it has been assumed that a template effect is p r o d u c e d by the micellar cylinders on the silicate polymers which are growing in the aqueous phase. The template effect of the hexagonal phase on the structural and textural properties of the materials has been demonstrated after a first thermal treatment up to 450~ in a nitrogen atmosphere. A second thermal treatment at the same temperature in air, to remove residual organic carbons, did not produce any significant structural and textural change. This was supported by the presence of a diffraction peak on the diffractogram of thermally treated gels (Fig. 7.11). Thus an ordered porous texture seems to be preserved after the removal of the surfactant. Characteristics of the porous structure versus surfactant molecule size, obtained from nitrogen sorption measurements, are reported in Table 7.3.
250
7 - - S O L - G E L C H E M I S T R Y A N D ITS A P P L I C A T I O N T O P O R O U S M E M B R A N E P R O C E S S I N G
TABLE 7.3 Porous structure characteristics of unsupported silica materials exhibiting an ordered microporosity (N2 adsorption measurements) [68] Sample (Cx)
Surface BET (m2/g)
Porous volume (cc/g)
Porosity (%)
Microporosity (%)
C8 C10 C12 C14
1260 1040 1100 1090
0.67 0.60 0.74 0.78
60 57 62 63
81 88 86 71
t/tg=15
l(a:u.) ganized
:a gel
Cs
dh=24.8~.
dh I~r~ dhl2 Y
Y
20
b
silica n
oI
I
AT l(a.u.) d=22.o~
20
Fig. 7.11. Schematic representation of a silica membrane material exhibiting an ordered porous texture. Template effect obtained from self-assembled amphiphilic systems. (a) Wet gel containing a hexagonal liquid crystal phase and the corresponding X ray diagram. (b) Heat-treated gel with residual ordered porosity and the corresponding X-ray diagram [68].
7 - - SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
251
lO
R(A) ----<>--R (Tandford) R (HK) RH (MP)
8
I
I
10
12
'
I
14
Cx
Fig. 7.12.Comparisonof measured and calculatedaveragemicroporeradius versus surfactant chain length (Cx)in silica membrane materials [68]. Whatever the surfactant chain length used, the specific surface area is very large (>1000 m2/g). The pore volume is also large and essentially microporous leading to a porosity of about 60%. The micropore size distribution is generally narrow but not totally monodispersed. The average micropore sizes were measured using nitrogen adsorption methods according to two different models: MP and Horvath-Kawazoe. The average pore radius clearly appears to be proportional to the number x of carbon of the alkyl chain. The pore radius was estimated from calculation on the template unit (micellar cylinder), using the equation given by Tandford [70]. Assuming a full interpenetration of surfactant molecules in micellar cylinders, it should approximate to half the chain length. In Fig. 7.12, it can be seen that there is a very good agreement between measured and calculated values. Recently characteristics of supported membranes prepared from this material have been reported [71].
7.6.2 Insertion of Organic and Inorganic Entities or Polymer Particles in Gel
Layers Organic or inorganic entities as well as polymer particles can also be used as template agents in the preparation of porous ceramic membranes following either the polymeric or the colloidal sol-gel route. The strategy to control microstructure in porous material is illustrated in Fig. 7.13. The template agents are trapped during matrix formation and eliminated in a second step with the aim to define the pore size in the final material. An example related to the polymeric route has been given by Roger et al. [72] in which a tin alkoxide is modified in order to create a residual porosity in the matrix formed from this modified precursor. These compounds were prepared from a tin (IV) alkoxide modified with difunctional carboxylate ligands. Formation of the final porous material is obtained through a two step process. The first
252
7 - - SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
COLLOIDAL GEL
POLYMERIC GEL rganic
I
polymer V particle
AT
AT :SO- OF
cropore i
~ .'-i~,
-
..
|-~, ~ - . ~
.:~._
macropore ~J Fig. 7.13.Schematicrepresentation of tailor made porous structures obtained from thermal decomposition of passive organic templates. hydrolysis step consists of the removal of the alkoxide groups to create a three-dimensional network of oxo-bridged tin carboxylate species. In the second step, the bridging groups are removed by acid hydrolysis to leave pores without creating new Sn-O-Sn bonds. Similar experiments have been carried out with titanium alkoxide. Regarding the porous structure of these materials, mesopores are observed after the first hydrolysis step while micropores with significantly lower average pore radii were seen after the second hydrolysis under acidic condition. However no correlation was established in this work between the size of the pore and the size of the carboxylic ligand. In a further work these authors described general routes to porous metal oxides by removal of template molecules from inorganic polymers formed by sol-gel type hydrolysis and condensation of metal alkoxides [73]. The investigated template molecules for the generation of porous microstructures include organic polymers, copper (II) ions in hybrid copper oxide/silica sols and copper (II) bis(hexafluoroacetylacetonate) (hfac). Neutron scattering and BET measurements were used to characterize the materials. Polyacrylic acid (Mw = 2,000 Dalton) was
7 -- SOL-GEL CHEMISTRY AND ITS APPLICATION TO POROUS MEMBRANE PROCESSING
253
used as an organic template to generate after removal by hydrolysis skeletal voids in tin oxide matrix. In a second series of experiments, mixed copper silicon oxide xerogels were prepared by hydrolysis of mixtures of Si(OEt)4 and Cu(OCH2CH(CHB)N(CHB)H)2 in the ratios of Si:Cu = 2:1, 4:1, 9:1. Microporous silica materials were formed from these xerogels by selective removal (etching) of the inorganic templates. In the third strategy, Si(OEt)4 was hydrolyzed in the presence of Cu(hfac)2, a volatile, inert inorganic template, in a 4:1 molar ratio. Removal of the template from the xerogel at 100~ in vacuum afforded microporous silica. Recently Raman et al. [74] proposed a new approach to microporous silica membranes based on organic-inorganic polymers prepared by co-polymerization of tertraethoxysilane (TEOS) and methyltriethoxysilane (MTES). The hybrid polymers were deposited on commercial asymmetric alumina supports. Heat treatments were employed to densify the inorganic matrix and pyrolyse the methyl ligands, creating a continuous network of micropores. Ayral et al. [75] showed that the template method can also be applied to hydrosols of colloidal metal oxide particles in view of the preparation of tailor-made porous material. In order to produce macroporous silica materials
Fig. 7.14. Template effect on a silica layer of e m b e d d e d latex particles. (a) Untreated layer. (b) Layer treated at 250~ [75].
254
7 - - S O L - G E L CHEMISTRY A N D ITS A P P L I C A T I O N TO POROUS M E M B R A N E PROCESSING
latex particles were incorporated in silica colloidal sols. Compared with silica colloidal sols, polymeric latex is another kind of stable dispersion with particles between 10 and 30 nm in diameter and with a narrow particle size distribution. The general procedure consists in mixing a silica colloidal sol with a polystyrene latex suspension to form a stable binary sol. The resulting sol was used to prepare a thin supported film by dip coating. Electron microscopy images of the film after drying showed no segregation phenomena in the deposited layer. Thermal treatment at 250~ induced decomposition of the latex particles and the production of spherical pores with a pore size identical to the size of the latex spheres. Figure 7.14 shows the case of a silica layer made of particles of about 10 nm in which latex particles (particle size 0.2 ~tm) have been incorporated. The two images are related to the initial layer (a) before heat treatment and the final layer (b) with pore size defined by the latex particle size. 7.7 CONCLUSION In this chapter, different aspects of the sol-gel process have been described which were applied to the synthesis of inorganic membrane materials. Macro-, meso- or microporous as well as almost dense materials can be obtained depending on the preparation method. Some limitations in the control of pore size and pore size distribution are attached to conventional sol-gel methods, namely the colloidal and the polymeric routes, frequently used for the synthesis of inorganic membrane materials. Thanks to recent advances in sol-gel processing, a number of these limitations have been overcome introducing new concepts in the preparation of membranes with tailor-made porous textures. The synthesis of nanophase ceramics is one of these concepts, it allows microporous ceramic materials with ceramic grains in the nanometer range to be obtained. Research in the field of nanophase materials is very active. A number of results on the control of microstructure and temperature stability of metal oxide ceramics can be applied to membrane preparation. Works carried out on non-oxide ceramics such as silicon carbide, silicon oxinitride or alttminum nitride should be regarded in order to extend the domain of available membrane materials. The formation of controlled porous textures by removal of templates initially included in a solid matrix is a concept generally accepted by people working in the preparation of materials for adsorption, catalysis and, more recently, membranes. Recent works have shown the versatility of this process for membrane application. It can be adapted to different kinds of gel matrices in order to define uniform pore size in the macro-, meso- or micropore range. Nevertheless most of the results relate to silica materials and further work is needed to apply this concept to other sol-gel derived oxide materials. In the future there is no doubt that inorganic membrane materials will continue to benefit from new advances in sol-gel processing of metal oxides.
7 ~ SOL-GELCHEMISTRYANDITSAPPLICATIONTOPOROUSMEMBRANEPROCESSING
255
REFERENCES 1. C.J. Brinker and G. Scherer, Sol-Gel Science. Academic Press, New York, NY, 1990. 2. A.J. Rubin, Formation and stability of colloidal dispersions of fine particles in water. Mater. Sci. Res., 17 (1984)45. 3. B.V. Derjaguin and L.D. Landau, Theory of the stability of strongly charged lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes. Acta Physicochim. URSS, 14 (1941) 633. 4. E. Verwey and J.T.G. Overbeek, Theory of the Stability of Lyophobic Colloids. Elsevier, Amsterdam, 1948. 5. A. Larbot, J.P. Fabre, C. Guizard and L. Cot, Inorganic membranes obtained by sol-gel techniques. J. Membr. Sci., 39 (1988) 203. 6. M.A. Anderson, M.J. Gieselmann and Q. Xu, Titania and alumina membranes. J. Membr. Sci., 39 (1988) 243. 7. A.J. Burggraaf, K. Keizer and B.A. Van Hassel, Ceramic nanostructure materials, membranes and composite layers. Solid State Ionics, 32/33 (1989) 771. 8. C.J. Brinker, T.L. Ward, R. Sehgal, N.K. Raman, S.L. Hietala, D.M. Smith, D.-W. Hua and T.J. Headley, Utramicroporous silica-based supported inorganic membranes. J. Membr. Sci., 77 (1993) 165. 9. R.S.A. de Lange, K.-N.P. Kumar, J.H.A. Hekkink, G.M.H. Van de Vlde, K. Keizer, A.J. Burggraaf, W.H. Dokter, H.F. Van Garderen and T.P.M. Beelen, Microporous SiO2 and SiO2/MOx (M = Ti, Zr, A1) for ceramic membrane applications: a microstructural study of the sol stage and the consolidated state. J. Sol-Gel Sci. Tech., 2 (1994) 489. 10. A. Ayral, C. Balzer, T. Dabadie, C. Guizard and A. Julbe, Sol-gel derived silica membranes with tailored microporous structures. Catal. Today, 25 (1995) 219. 11. C. Roger and M.J. Hampden-Smith. J. Mater. Chem., 2 (1992) 1111. 12. J. Livage, M. Henry and C. Sanchez, Sol-gel chemistry of transition metal oxides. Prog. Solid State Chem., 18 (1988) 259. 13. D.P. Partlow and B.E. Yoldas, Colloidal versus Polymer gels and monolithic transformation in glass-forming systems. J. Non-Cryst. Solids, 46 (1981) 153. 14. A.F.M. Leenaars, K. Keizer and AJ. Burggraaf, The preparation and characterization of alumina membranes with ultra-fine p o r e s - 1. Microstructural investigations on nonsuppoerted membranes. J. Mater. Sci., 19 (1984) 1077. 15. A.F.M. Leenaars and AJ. Burggraaf, The preparation and characterization of alumina membranes with ultra-fine pores m 2. The formation of supported membranes. J. Colloid Interface Sci., 105 (1985) 27. 16. A. Larbot, J.A. Alary, C. Guizard, L. Cot and G. Gillot, New inorganic ultrafiltration membranes: preparation and characterization, High Tech. Ceramics, 3 (1987) 143. 17. H.P. Hsieh, R.R. Bhave and H.L. Fleming, Microporous alumina membranes. J. Membr. Sci., 39 (1988) 241. 18. M.J. Gislmann, M.A. Anderson, M.D. Moosemiller, C.G. Hill, Jr., Physicochemical properties of supported and unsupported y-alumina and titania membranes. Sep. Sci. Technol., 23, 12&13 (1988) 1695. 19. T. Okubo, M. Watanabe, K. Kusukabe, S. Morooka, Preparation of y-alumina thin membrane by sol-gel processing and its characterization by gas permeation. J. Mater. Sci., 25 (1990) 4822. 20. M. Zaharescu, C. Pirlog, A. Vasilescu, D. Crisan and A. Braileanu, Inorganic membranes obtained by the sol-gel process. Rev. Roumaine Chim., 35, 7-9 (1990) 909.
256
7 -- SOL--GELCHEMISTRYAND ITSAPPLICATIONTO POROUSMEMBRANEPROCESSING
21. P. Cini, S.R. Blaha, M.P. Harold and K. Venkataraman, Preparation and characterization of modified tubular ceramic membranes for use as catalyst support. J. Membr. Sci., 55 (1991) 199. 22. Y.S. Lin, K.J. de Vries and A.J. Burggraaf, Thermal stability and its improvement of alumina membrane toplayers prepared by sol-gel methods. J. Mater. Sci., 26 (1991) 715. 23. Y.S. Lin and A.J. Burggraaf, Preparation and characterization of high temperature thermally stable alumina composite membrane. J. Am. Ceram. Soc., 74 (1991) 219. 24. M. Chai, K. Sekizawa, M. Machida, K. Eguchi and H. Arai, Preparation of heat resistant microporous ceramic membranes for selective gas separation. Ceram. Lett., 99 (1991) 530. 25. C. Guizard, N. Idrissi, A. Larbot and L. Cot, An electronic conductive membrane from a sol-gel process. Br. Ceram. Proc., 38 (1986) 263. 26. C. Guizard, F. Legault, N. Idrissi, A. Larbot, L. Cot and C. Gavach, Electronically conductive mineral membranes designed for electro-ultrafitration. J. Membr. Sci., 41 (1998) 127. 27. A. Larbot, J.P. Fabre, C. Guizard, L. Cot and J. Gillot, New inorganic ultrafiltration membranes: titania and zirconia membranes. J. Am. Ceram. Soc., 72 (1989) 257. 28. Q. Xu and M.A. Anderson, Synthesis of porosity controlled ceramic membranes. J. Mater. Res., 6-5 (1991) 1073. 29. J. Etienne, A. Larbot, A. Julbe, C. Guizard and L. Cot, A microporous zirconia membrane prepared by the sol-gel process from zirconyl oxalate. J. Membr. Sci., 86 (1994) 95. 30. K. Kusakabe, K. Ichiki and S. Morooka, Separation of CO2 with BaTiO3 membrane prepared by the sol-gel method. ]. Membr. Sci., 95 (1994) 171. 31. C.-H. Chang, R. Gopalan and Y.S. Lin, A comparative study on thermal and hydrothermal stability of alumina, titania and zirconia membranes. J. Membr. Sci., 91 (1994) 27. 32. K.-N.P. Kumar, V.T. Zaspalis, F.F.M. De Mul, K. Keizer and A.J. Burggraaf, Thermal stability of supported titania membranes. Mater. Res. Soc. Syrup. Proc., 271 (1992) 499. 33. K.-N.P. Kumar, K. Keizer and A.J. Burggraaf, Textural stability of titania alumina composite membranes. ]. Mater. Chem., 3 (1993) 917. 34. K.-N.P. Kumar, K. Keizer and A.J. Burggraaf, Stabilisation of the porous structure of nanostructured titania avoiding the phase transformation. J. Mater. Sci. Lett., 13 (1994) 59. 35. A. Larbot, A. Julbe, C. Guizard and L. Cot, Silica membranes by the sol-gel process. J. Membr. Sci., 44 (1994) 289. 36. C. Guizard, N. Cygankiewicz, A. Larbot and L. Cot, Sol-gel transition in zirconia systems using physical and chemical processes. J. Non-Cryst. Solids, 82 (1986) 86. 37. C. Guizard, A. Larbot, L. Cot, S. Perez and J. Rouviere, Etude de la transition sol-gel en milieu micellaire inverse m Ih Principes fondamentaux. J. Chim. Phys., 87 (1990) 1901. 38. B.B. Mandelbrot, The Fractal Geometry of Nature. Freeman, San Francisco, CA, 1982. 39. C.J. Brinker, R. Sehgal, N. Raman, P.R. Schunk and T.J. Headley, Polymer approach to supported silica membranes. J. Sol-Gel Sci. Techol., 2 (1994) 469. 40. R.S.A. de Lange, K.-N.P. Kumar, J.H.A. Hekkink, G.M.H. Van de Velde, K. Keizer, A.J. Burggraaf, W.H. Dokter, H.F. van Garderen and T.P.M. Beelen, Microporous SiO2 and SiO2/MOx (M = Ti, Zr, A1) for ceramic membrane application: a microstructural study of the sol-stage and the consolidated state. J. Sol-Gel Sci. Technol., 2 (1994) 489. 41. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraff, Microporous sol-gel modified membranes for hydrogen separation. Key Engin. Mater., 61 & 62 (1991) 77. 42. R.J.R. Ulhorn, K. Keizer and A.J. Burggraff, Synthesis of ceramic membranes. Part II. Modification of thin alumina films: reservoir method. J. Mater. Sci., 27 (1992). 43. G.R. Gavalas, C.E. Megiris and S.-W Nam, Deposition of H2-permselective SiO2 films.
7 -- SOL-GELCHEMISTRYAND ITSAPPLICATIONTO POROUSMEMBRANEPROCESSING
44. 45.
46.
47.
48.
49.
50.
51. 52. 53. 54. 55.
56. 57. 58.
59.
60. 61. 62.
257
Chem. Eng. Sci., 44 (1989) 1829. C.L. Lin, D.L. Flowers and P.K.T. Liu, Characterization of ceramic membranes. II. Modified commercial membranes with pore size under 40 ~. J. Membr. Sci., 92 (1994) 45. R.J.R. Ulhom, K. Keizer and A.J. Burggraff, Gas transport and separation with ceramic membranes. Part II. Synthesis and separation properties of microporous membranes. J. Membr. Sci., 66 (1992) 271. C.J. Brinker, T.L. Ward, R. Sehgal, N.K. Raman, S.L. Hietala, D.M. Smith, D.-W. Hua and T.J. Headley, "Ultramicroporous" silica-based supported inorganic membranes. J. Membr. Sci., 77 (1993) 165. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Polymeric-silica based sols for membrane modification applications: sol-gel synthesis and characterization with SAXS. J. Non-Cryst. Solids, 191 (1995) 1. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterization of supported microporous ceramic membranes prepared by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57. C. Guizard, A. Julbe, A. Larbot and L. Cot, Nanostructures in sol-gel derived materials: application to the elaboration of nanofiltration membranes. J. Alloys Compounds, 188 (1992) 8. A. Julbe, C. Guizard, A. Larbot, L. Cot and A. Giroir-Fendler, The sol-gel approach to prepare candidate microporous membranes for membrane reactors. J. Membr. Sci., 77 (1993) 137. J. Livage, Complexation of inorganic precursors in aqueous solutions. Mater. Res. Soc. Symp. Proc., 271 (1992) 201. C. Sanchez and J. Livage, Sol-gel chemistry from metal alkoxide precursors, New. J. Chem., 14 (1990) 513. A. Larbot, D. Young, C. Guizard, R. Paterson and L. Cot, Alumina nanofiltration membrane from sol-gel process. Key Engin. Mater., 61 & 62 (1991) 395. B.E. Yoldas, Alumina gels that form porous transparent A1203. J. Mater. Sci., 10 (1975) 1856. S. Alami-Younssi, A. Larbot, M. Persin, J. Sarrazin and L. Cot, Gamma alumina nanofiltration membrane. Application to the rejection of metallic cations. J. Membr. Sci., 91 (1994) 87. F. Garcia, Elaboration de Membranes en Zircone sur Support M6tallique, PhD Thesis, University of Montpellier II, France, 1989. P. Chanaud, A. Julbe, P. Vaija, M. Persin and L. Cot, Study of lanthanum-based colloidal sols formation. J. Mater. Sci., 29 (1994) 4244. P. Chanaud, A. Julbe, A. Larbot, C. Guizard, L. Cot, H. Borges, A. Giroir Fendler, C. Mirodatos, Catalytic membrane reactor for oxidative coupling of methane. Part 1: preparation and characterisation of LAOC1 membranes. Catal. Today, 25 (1995) 225. H. Borges, A. Giroir Fendler, C. Mirodatos, P. Chanaud, A. Julbe, Catalytic membrane reactor for oxidative coupling of methane. Part 2: catalytic properties of LAOC1 membranes. Catal. Today, 25 (1995) 377. C. Guizard, C. Mouchet, R. Vacassy, A. Julbe and A. Larbot, Sol-gel processing of inorganic membranes. J. Sol-Gel Sci. Technol., 2 (1995) 483. A. Yamamoto and S. Kambara, Structures of the reaction products of tetraalkoxytitanium with acetylacetonate and acetoacetate. J. Am. Chem. Soc., 79 (1957) 4344. C. Sanchez, J. Livage, M. Henry and F. Babonneau, Chemical modifications of alkoxide precursors. J. Non-Cryst. Solids, 100 (1988) 65.
258
7 -- SOL-GELCHEMISTRYAND ITSAPPLICATIONTO POROUSMEMBRANEPROCESSING
63. A. Julbe, C. Guizard, A. Larbot, C. Mouchet, R. Vacassy, R. Metz and L. Cot, Sol-gel processing of titania and zirconia membranes, in: Y.H. Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, Worcester, MA, 1994 p. 17. 64. B. Yoldas, Hydrolysis of titanium alkoxide and effects on hydrolitic polycondensation parameters. ]. Mater. Sci., 21 (1986) 1087. 65. T. Yamaki, H. Maeda, K. Kusakabe and S. Morooka, Control of the pore characteristics of thin alumina membranes with ultrafine zirconia particles prepared by the reversed micelle method. J. Membr. Sci., 85 (1993) 167. 66. A. Julbe, C. Balzer, J.M. Barthez, C. Guizard and L. Cot, Effect of non-ionic surface active agents on TEOS-derived sols, gels and materials. J. Sol-Gel Sci. Technol., 4 (1995) 89. 67. T. Dabadie, A. Ayral, C. Guizard, L. Cot, C. Lurin, W. Nie and D. Rioult, Synthesis and characterization of silica gels obtained in lamellar media. Mater. Sci. Forum, 152-153 (1994) 267. 68. T. Dabadie, A. Ayral, C. Guizard and L. Cot, Liquid crystal templating effects on silica gels synthesized using quaternary ammonium surfactants. Mater. Res. Soc. Syrup. Proc., 346 (1994) 849. 69. T. Dabadie, A. Ayral, C. Guizard, L. Cot, C. Lurin, W. Nie and D. Rioult, Synthesis of inorganic gels in lyyotropic liquid crystal medium. I. Synthesis of silica gels in lamellar phases obtained from non-ionic surfactants. J. Sol--Gel Sci. Technol., 4 (1995) 107. 70. C. Tandford in: The Hydrophobic Effect, 2nd edn., Wiley-Interscience, New York, 1980. 71. T. Dabadie, A. Ayral, C. Guizard, L. Cot, J.C. Robert and O. Poncelet, New microporous silica materials for membrane processes, in: Y.H. Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, Worcester, MA, 1994, p. 411. 72. C. Roger, M.J. Hampden Smith and C.J. Brinker, Hydrolysis and condensation of modified tin (IV) alkoxide compounds to form controlled porosity materials. Mater. Res. Soc. Syrup. Proc., 271 (1992) 51. 73. C. Roger, D.W. Schaefer, G.B. Beaucage and M.J. Hampden-Smith, General routes to porous metal oxides via inorganic and organic templates. J. Sol--Gel Sci. Technol., 2 (1994) 67. 74. N.K. Raman and C.J. Brinker, Organic "template approach" to molecular sieving silica membranes. ]. Membr. Sci., 105 (1995) 273. 75. A. Ayral, C. Guizard and L. Cot, Synthesis and application of hybrid organic-inorganic colloidal gels. J. Mater. Sci. Lett., 13 (1994) 1538.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.I. Burggraay:and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 8
Fundamentals of membrane top-layer synthesis and
processing
A.J. Burggraaf Laboratory of Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
8.1 SYNTHESIS A N D P R O C E S S I N G OF S U P P O R T E D M E S O P O R O U S MEMBRANES
8.1.1 Introduction
The aim of ceramic membrane production is to obtain defect-free (no cracks, no pinholes) supported films with homogeneous thickness and a narrow pore size distribution. By far the most important are sol-gel based processing routes. They have in common the fact that a porous support is contacted with a colloidal precursor solution for a given time to form a film which is processed after this step. The general procedure can be broken down into the following steps: (1) initial layer formation in a precursor solution (2) growth of film thickness (lyogel state) (3) drying of the film to xerogel state (4) calcination/sintering to obtain the final membrane The final membrane properties and quality depend critically on the support quality, on the concentration and structure of the precursor solution and on details of the drying and calcination process. The drying process is particularly critical in order to avoid fracture and cracking of the membrane, as well as the formation of the final microstructure of the membrane.
260
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
Basic elements of drying theory relevant to stress formation and related defect formation will therefore be summarised. For derivations and detailed discussions, the book by Brinker and Scherer [1] should be used. 8.1.2 Film Formation
The process starts with contacting a porous support with the colloidal precursor solution in a dip-coating or a spin-coating process. Film formation starts either with a film-coating or a slip-casting mechanism as will be discussed below. 8.1.2.1 Initial Layer Formation
Directly after contacting a gel, plugs should form in the pore entrance to inhibit (excessive) penetration of the precursor particles into the pores of the support. This step is called "initial layer formation". Such penetration will result in partial blocking of the support pores in the later processing steps and will increase the resistance for permeation of gases and liquids which is usually not wanted. This process of pore plugging is essentially an interesting one because it, potentially, produces a membrane within the support. Up to now no sol-gel processes have been reported which can be controlled sufficiently to yield thin plugs with uniform thickness. Consequently, efforts have been directed towards obtaining membrane films on the porous support with no or only little penetration into the support pore system. The initial layer formation step has hardly been investigated. Some qualitative guidelines can be abstracted from theoretical considerations and experimental observations. The first intensive and systematic study of the kinetics and mechanism of membrane formation was performed by Leenaars et al. [2-4] on boehmite and 7-A1203membranes with pore diameters in the range of 3-5 nm. The critical parameter governing the initial layer formation was found to be the ratio of the particle diameter (in the colloidal suspension) to the pore diameter of the support. Given a certain colloidal suspension (characterised by its alumina concentration, kind and concentration of peptising acid used and ageing time) gel films could be formed on a support with pores below a critical diameter as shown in Table 8.1 [2,3]. Given a certain support diameter, the immediate formation of a gel film could be promoted by (i) increasing the boehmite concentration, (ii) ageing the colloidal solution and, (iii) in some cases the dipping time [3]. In later studies [12,14] additions such as PVA were added which enhance the viscosity of the colloidal solution and promote the formation of a lyogel film. These observations can be rationalised with the following (qualitative) models. To obtain ultra-fine pores in a structure by more or less regular packing of primary particles [1,4], these primary particles should be small. In the case of
8-- FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISANDPROCESSING
261
TABLE 8.1 The influence of the type of support (pore size) and the peptising acid on the formation of gel layers during dip-coating. In all cases the sol contains 1.2 tool A1 (boehmite) per litre. Support type (modal pore size) 1 (0.12~tm) 2 (0.34 gin) 3 (0.8 ~m)
Peptising acid used HC1
HNO 3
HC104
Gel layer formed Gel layer formed Gel layer formed
Gel layer formed Gel layer formed No gel layer formed
Gel layer formed No gel layer formed No gel layer formed
boehmite colloidal solutions discussed above, these primary particles are plateshaped with a diameter of at least 25 nm and a thickness of about 6-7 nm. Depending on the type and concentration of the dispersing acid, the concentration of particles and ageing time, the primary particles form loosely bound agglomerates (secondary particles) with increasing dimension the greater the concentrations a n d / o r ageing time. Reported secondary particle sizes are in the range of 80 nm for alumina and 30-250 nm for TiO2 [14]. Directly after dipping the substrate into the dip solution, water (liquid) is sucked into the support pores. When the secondary particles are considerably smaller than the support pores, they enter these pores to some extent and a pore clogging mechanism starts which reduces the effective pore diameter until, after some time, cake (gel film) formation starts. The pore clogging period is shortened when the concentration of solids is increased [5] and with larger (secondary) particle size. If the concentration a n d / o r the particle size is large enough, pore clogging hardly occurs and a gel film is immediately formed. It should be stressed that this is a very simplified model. For example, the magnitude and size of the electrical charges on the pore wall and particle surface will be important in cases where pore and particle size are not too different. No data are available on initial membrane formation. A similar situation exists to explain the positive effect of some additions (e.g. PVA to boehmite solutions). The trends in experimental observations and the qualitative model discussed above holds also for the formation of other types of mesoporous membranes (titania, zirconia, silica) [6] (Chapter 7).
8.1.2.2 Mesoporous Film Formation In this section the thickness of the formation of a lyogel film with reference to the development of its thickness will be discussed. In latter sections the development of its microstructure will be treated.
262
8 ~ F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
The production processes used are (i) dip-coating for plates and tubes, (and ii) spin-coating, mainly for plates. In the dip-coating process the porous plate or tube is contacted with the liquid for a given time. The way in which the contacting process is carried out is important because the porous support must always be put into and withdrawn from the liquid which causes start-up and drainage effects. The part of the support that leaves the liquid last is contacted for a longer time, and the liquid film is somewhat thicker at this place (due to drainage, see below). This causes a certain inhomogeneity in the final membrane thickness. In the case of tubes, the film can be applied on the inner or outer side depending on the position of the liquid. Moreover, the tube can be immersed in the liquid and withdrawn, or the liquid can be poured into the tube and removed. Strategies has been developed to minimise the formation of asymmetric/inhomogeneous films, e.g., by filling a tube with a limited amount of liquid (the height of the liquid reservoir being only a few centimetres) and pulling the tube along this liquid reservoir with the help of a liquid "slot" between the tube wall and reservoir (see Chapter 6). In the spin-coating process the liquid is applied to a rotating plate and is distributed across the plate by centrifugal forces [1,8]. Two main groups of lyogel film formation mechanisms can be distinguished: (i) film coating, and (ii) slip casting. In the film-coating process, capillary forces in pores do not play a role and this process can be used also on dense, non-porous substrates. The slip cast process is widespread in the production of bulk ceramics but has only been recently applied to the production of membranes [3,9]. In this process capillary forces play a dominant role.
The film-coating process The film-coating process is schematically shown in Fig. 8.1. After the immersion and start-up steps (Fig. 8.1A,B), a steady-state situation is established (Fig. 8.1C). The moving substrate exerts a viscous drag (~ rl4/h) upwards on the liquid and entrains liquid in a fluid mechanical boundary layer from which liquid is drained by gravitational force (pgh). When the substrate speed and liquid viscosity are not too large, as is often the case in sol-gel processing, this balance is modified by the ratio of the viscous drag to liquid-vapour surface tension (Tlv) and the film thickness h (Fig. 8.1B) is given by Eq. (8.1) [10]: 0.94 (rl U) 2/3 h = rlv'l/6'"~t,g)1/2
(8.1)
where U is the withdrawal speed, TI is the liquid viscosity, p the liquid specific mass. Equation (8.1) shows that thicker films are obtained with larger withdrawal speed and viscosity.
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
(A) IMMERSION
DEPOSITION & DRAINAGE
263
EVAPORATION
U
t
h
LIQUIO
BATH SURFACE 181
Fig. 8.1. The film-coatingprocess. (A) stages of the process. (B) Detail of the liquid flowpatterns. U is the withdrawal speed, h is the thicknessof the fluid film, S is the boundary layer. From Scriven [8]. In the discussion so far it has been assumed that the only interaction between liquid and substrate is by adhesion forces. This means that with porous substrates capillary forces should be absent. This situation can be obtained by filling the substrate pores with another liquid having a surface tension comparable to that of the colloidal solution or by making the substrate hydrophobic (in case of an aqueous solution). The film-coating process is applied using suspensions on macroporous substrates to produce intermediate films with macropores in order to obtain microfiltration membranes or to obtain composite, asymmetric supports suitable for production of ultra-fine, mesoporous membranes as discussed in Chapter 6.
264
8
-
-
FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
O
Un
/
~
mLM
.X:O
FiLM COLLJkPSE AND/OR PORE FORMATION
AGGREGATION
ALCOHOLrWATER EVAPORATION
GRAVITATIONAL ) DRAINING EVAPORATION
J ;L- 01Uolpg) ~r~
~
r ENTRAINED DILUTE SOL RESERVOIR
8U FACE
~xl
-
DILUTE SOL
Fig. 8.2. Schematicrepresentation of the sequential stages of the structural development of a film during a film-coatingprocess. From Brinker [1]. The applicability of Eq. (8.1) in film-coating processes with sols (sol-gel processing) is discussed by Scherer and Brinker [1] and a number of deviations between theory and experimental observations has been reported. This can be understood from Fig. 8.2 where it is shown that due to evaporation of the solvent, the colloidal solution (suspension) is strongly concentrated. This results first in aggregation, then in gelation and finally in drying and considerable reorganisation of the structure (see section on "drying and microstructure development"). This overlap between deposition and evaporation stages, with related changes in properties of the lyofilm, is the main cause of deviation from Eq. (8.1).
The slip-casting process In the slip-casting process a capillary pressure drop APc between the support pores (with radius r in case of cylindrical capillaries) and the liquid drives the liquid into the support and through the formed gel layer. This is schematically shown in Fig. 8.3. The total pressure drop APc (generated by the support pores)
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
265
/ pressure
aP0
l
-} &Ps
dry support wet s ~ r t
gel- I layer
(Lg)
ape
,,.....
sol
(Ls) Fig. 8.3. Schematicrepresentation of the pressure drop across the membrane system during the slip casting process (Leenaars [2]).
is proportional with 271vCOS~/r with ~lv the surface tension and [3 the contact angle between liquid and solid interface. When liquid flows through an initially formed gel layer, particles are deposited on this layer and the layer thickness increases as a function of time together with its flow resistance. This means that the flux decreases as a function of time. Membrane formation by slip casting with colloidal solids has been extensively and systematically studied by Leenaars et al. [2,3] and later by Tiller and Chum Dar Tsai [11], Uhlhorn et al. [12] and Kumar et al. [13]_ According to Leenaars et al [3], the layer thickness L of the gel layer can be expressed as a function of the time according to Eq. (8.2): Lg = K.
2 ~'lv cos ~ AP
g t 1/2 + La
(8.2)
n where K is a constant and La the adhering film thickness due to the film coating process (see above for the other parameters). The constant K incorporates factors which determine the flow resistance of the gel layer such as its porosity and tortuosity, pore size and shape, etc. Using the Kozeny-Carman relation, Leenaars [3] studied the relation between K, microstructure and process parameters. The value of the pressure drop across the gel layer APg is obtained from APc after correction for the pressure drop AP s in the support, which is usually very small. A plot of L versus t 1/2 gives a linear curve as shown in Fig. 8.4 for layer formation from a boehmite sol [12]. The process is perfectly described by a rapid slip casting mechanism. After a few seconds the layer thickness after calcination is of the order of 4-7 ~tm. The
266
8 ~
FUNDAMENTALS
OF
MEMBRANE
TOP-LAYER
SYNTHESIS
AND
PROCESSING
10
0
X
qWl IWI
"
C .1= ~t,t e--
5
4., 0 ..J
Z
0
I
0
_
,
_A
2
,,
l
4
,
I
6
,
8
Square root of dipping time (s 1/21 Fig. 8.4. Increase in layer thickness of a "f-A1203layer per dipping step as a function of the square root of time for a (boehmite) dip solution: (1) without PVA addition; (2) with PVA addition; (3) second dipping procedure without PVA. From Uhlhom et al. [12]. m i n i m u m obtainable layer thickness is determined by the film coating component (intersection of curves with the vertical axis) and is of the order of 0.5-1 ~tm. The film formation rate initially becomes larger w h e n the support pores become smaller (larger driving force &Pc). With decreasing support pore size however the flow resistance of the support increases, and so does the pressure drop AP s in the support. Consequently there should be an o p t i m u m support pore diameter to obtain a m a x i m u m driving force APg (Fig. 8.3) for layer formation [11]. Addition of organic molecules considerably increases the viscosity and so decreases the m e m b r a n e formation rate [12] (Fig. 8.4, compare curves 1 and 2). According to Eq. (8.1) an increase of the viscosity should lead to a larger value of La. This was not found by U h l h o m [12] as can be seen from Fig. 8.4. This is probably caused by a decrease of the surface tension of the liquid
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
267
after PVA addition which compensates for the increase in viscosity in the film-coating step and also reduces the driving force for slip casting and thus its formation rate.
Multiple step dip-coating process In the multiple step process, after calcination of the first layer, the complete cycle of dipping, drying and calcination is repeated. The final membrane consists then of a sandwich of X layers (where X is the number of dipping steps). Multiple dipping is important for several reasons: (1) The formation of second or more layers is helpful in repairing defects like pinholes or small cracks which are present in the first layer. (2) For making thicker membranes, which may be important in catalytic applications. (3) For making multilayer membranes with layers having different chemical compositions. As is shown in Fig. 8.4 the film growth rate is much smaller in the second step compared to that in the first one. This is mainly due to the additional resistance offered by the first applied layer. It should be noted that in both single and multiple dipping the driving force for film formation (capillary pull) is always exerted by the dry portion of the support. A multilayer membrane is shown in Fig. 8.5. Multiple dip-coating removes defects present in the first layer as shown in Fig. 8.5 for a three-layer y-alumina membrane system. From the non-zero slope of the permeability as a function of the trans-membrane pressure it follows that defects are present giving rise to non-Knudsen (viscous flow) contributions to the gas permeability (see Chapter 9). These
Fig. 8.5. SEM m i c r o g r a p h of a m u l t i l a y e r m e m b r a n e .
268
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
E c
200
d
150
o~
o
100
'\
b', \
~d
C
50
0
\
\ "x \
2
',\\
~L_
12
22
Dipping
ttme
32
42
~n Sec.
Fig. 8.6. Surface rouglmess of y-alumina membrane top layers formed on 0c-alumhla supports (pore diameter about 0.2 l.tm) with average roughness of 180 (a), 120 (b) and 40nm (c). From Kumar [13].
contributions are almost completely eliminated after the application of the second layer (Fig. 8.5). The drawback of this technique is the increase of the final layer thickness. To diminish this effect, the concentration of the dip solutions can be decreased in the subsequent dipping steps. The mechanism of the repairing process is most probably preferential layer (plug) -formation on defect sites (pinholes, cracks) in the first layer during subsequent dipping steps. This is due to the lower resistance against liquid transport (larger K value in Eq. (8.2)) on defect sites resulting in enhanced (gel) layer growth and self-repairing of these sites. The layer growth rate on regular (non-defect) areas of the layer is relatively small during the process. Multiple dipping finally increases the maximum allowable thickness of the (calcined) crack-free membrane with respect to that obtainable in a single step [13,14,18]. At each given set of experimental conditions there is a certain critical layer thickness which should not be exceeded if cracking during drying and / or calcination is to be avoided. This thickness differs greatly for different materials as shown in Table 8.2. Note that the thickness for unsupported membranes is much larger. As shown in Table 8.2 the maximum allowable thickness increases strongly using a multilayer system produced in a multi-step coating process. These effects will be discussed further in subsequent sections.
Some factors affecting mesoporous properties (1) Organic additives in precursor dip solutions. As has been discussed in the preceding section, organic additions strongly affect the kinetics of film formation. Even more important is their beneficial effect on the defect quality of the final membranes. Ulhorn et al. [12] showed that the reproducibility of the membrane formation process was greatly enhanced and the defect level of the final ?-alumina mem-
269
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
TABLE 8.2 Maximum allowable (critical) alumina and titania supported membrane thickness under given conditions on an cz-aluminasupport with a pore diameter of about 0.2/am Material
M a x i m u m Maximum thickness thickness first dip total
Sol particle size (nm)
Addition of organic additives
Support quality
7-A1203
--8/am
--24/am
TiO2
--2/am
>6/am
SiO 2
50-100 nm
50-100nm
40 (platelets) 25-30 (spheres) 0.2-2 (polymers)
increases surface roughness reproducibility < 300 nm necessary surface roughness <100 nm ? surface roughness -- 40 nm
branes was greatly reduced by addition of Poly vinyl alcohol (PVA) to the boehmite dip solution. A similar effect has been reported by Zaspalis et al. [14] for titania and composite membranes using a mixture of PVA and hydroxy propyl cellulose (HPC) as addition and by Z6ter et al. [15] for zirconia membranes. In addition to the above-mentioned effects these additions allow the drying and calcination (heating/cooling cycle) rates to be accelerated considerably [14]. Important parameters of the organic additions are their chemical nature, molecular weight, and concentration. The best results were obtained for boehmite sols with concentrations of typically 0.5-1.0 mol A1OOH/1 at p H = 4 with 35 g PVA/1 The PVA had an average molecular weight of 72000 Dalton. In the case of titania sols typical concentrations were 0.3 mol TiO2/1 and a mixture of 10 g PVA with 35 g HCP/1 with HPC having a molecular weight of 106 Dalton. It is important to use PVA and HPC (generally organic additions) of a quality which leaves no ash after calcination. Uhlhorn et al. [12] studied the effect of PVA with a molecular weight of 72000 Dalton on the microstructure of y-alumina membranes by gas absorption, water permeability and solute retention measurements and concluded that the structure of supported thin alumina films with and without PVA is similar. In a more recent study Z~ter et al. [15] found no significant effect of the molecular weight of the PVA (range 3000-155000 Da, fixed concentration) on the pore characteristics of the final (calcined) alumina or alumina-zirconia membranes. Pore diameters between 3.2 and 3.8 n m (with porosities of 52-56%) were found after calcination at 600~ for 3 h. Note that the gyration radii of the PVA molecules changes from 2.2 n m (3000 Da) to about 20 n m (155000 Da) and so are larger than the final pore diameter. In preliminary work with much larger molecular weights (of 106 Da) a m u c h broader pore size distribution was found. Zi.iter [15] reported an increase of the
270
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P- L A Y E R SYNTHESIS A N D P R O C E S S I N G
pore diameter in unsupported zirconia membranes after addition of PVA (using a molecular weight of 72000 Da). As a function of the added concentration, the pore diameter and the porosity showed a weak minimum or a plateau (depending on calcination temperature) around 2.5-5 ml PVA per ml solution after which these parameters increased with concentration (total investigated range is 20 ml PVA per ml solution). Finally, organic additives can affect the thermal properties (phase transformation behaviour) of mesoporous ceramic membranes. Ziiter [15] reported the transformation from monoclinic to tetragonal particles and back as a function of the amount of PVA. The precise mechanisms by which organic additives work is not known. Various theories concerning their role are reviewed by Zarzycki [16]. The strengthening of the gel network and the elimination of differential stresses originating from non-uniform (initial) pore structures appears to be important in minimising defects [17]. As will be shown below, additives decrease the drying stresses and so minimise crack formation during drying and subsequent calcination. (2) Support quality and roughness effects. As will be clear from the preceding discussions the quality of the support is important as is also discussed in detail in Chapter 6. A broad pore size distribution will give rise to pinhole formation due to locally insufficient initial layer formation (pore clogging). Local impurities may change the wetting behaviour, again resulting in pinhole formation. A minimum roughness of the support surface is also required to produce defect-free membrane layers. In the present context, surface roughness is defined as the average perpendicular (to the surface) distance between peaks and dips in the support surface. As discussed in Chapter 6, several other definitions of roughness can be given and used. The roughness of the support may limit the minimum achievable layer thickness. From a fracture mechanics point of view, surface roughness determines the maximum size and sharpness of flaws which can act as crack initiators (via the stress intensity factor). As shown in Table 8.2, the minimum required roughness of the support to obtain crack-free membranes is smaller for TiO2 membranes than for y-alumina (mesoporous) membranes. To obtain microporous silica membranes even more stringent requirements prevail. A smaller roughness can be obtained to some extent by polishing of the support. Beside economic arguments, this method cannot give roughness values below the limit given by the relation between particle size (and thus pore size) of the support particles and roughness. A better method is to coat the initial support with a layer consisting of smaller particles (and pores). As shown by Kumar [13] in Fig. 8.6, (x-alumina supports with different initial roughness can be smoothed by dip-coating with y-alumina layers.
8 -- FUNDAMENTALS
OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
271
The 7-alumina layers are obtained by dip-coating with boehmite sols, followed by drying and calcination at 600~ for 3 h. This results in mesoporous 7-alumina layers with pore diameters of 3-4 nm. It is interesting to note that the surface roughness of both two sets of membranes (a and b in Fig. 8.6) decrease with increasing thickness (dipping time) and became constant after about 10 s. The limiting roughness is about 40 nm and is obviously determined by the properties of the 7-alumina layer. Defect-free alumina and titania membranes on (z-alumina support could be made (under the given synthesis conditions) only on supports having average surface roughnesses of 300 and 100 nm respectively.
8.1.2.3 Theoretical Aspects of the Drying Processfrom Lyogel to Xerogel Film During the withdrawal process a lyogel film has been formed with sufficient mechanical stability to withstand gravitational and drainage forces (see Fig. 8.1). This lyogel film must be dried and form a xerogel before final consolidation by heating to high temperature. Drying is one of the most crucial steps in the formation of ceramic membranes for two reasons: firstly, because membranes tend to crack during this process, and secondly because basic features of the microstructural skeleton (pore size distribution) are determined in this stage. Both aspects are related to capillary forces acting on the gel and related stress (distribution) and deformation (shrinkage) of the gel which differ for bulk gel and gel films and which change during the process. The outline of this section is as follows: (1) A global picture is given of the several phenomenological stages of the drying process in bulk gels (constant rate period, first and second falling rate period, critical point). (2) Drying stresses result from the pressure gradient in the pores of the bulk gel. How they are related to microstructural features and to the drying rate is discussed. (3) Differences of stress development and fracture behaviour between bulk and films are discussed. (4) Strategies to reduce or eliminate cracking are discussed including chemical additives, supercritical drying and thickness effects. In the next section typical experimental results for ceramic membranes are discussed. For a detailed discussion of all theoretical aspects the reader is referred to the extensive treatment by Brinker and Scherer [1]. A useful summary of the drying literature is given by Simpkins et al. [14] and Fortes and Okos [20]. The drying of gels is treated by, e.g., Zarzycki [21] and by Dwivedi [22].
272
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
Phenomenological stages and critical point in idealised situations The first stage of drying a piece of gel is called the constant rate period (CRP) because the rate of evaporation per unit of area of the drying surface is independent of time. The evaporation rate is close to that of an open dish of liquid [22] and the volume shrinkage will be equal to the volume of water removed by evaporation. Liquid flows from the interior of the gel to replace that which evaporates and consequently the gel shrinks, typically by a factor 5-10. It should be noted that in membrane preparation this dramatic increase in concentration might result in precipitation of components (salts, acid) solved in the precursor solution (also called effiorence). The shrinkage in the CRP is caused by adsorption and capillary forces which oppose exposure of the solid phase. After some time the reduced volume of liquid is able to submerge the solid phase only after the formation of a meniscus as shown in Fig. 8.7a. This causes a capillary pressure P given by Eq. (8.3). This capillary pressure has the character of a tension in the liquid which is supported by the solid phase which therefore is in compression. P = - 271v cos0/r
(8.3)
where 7Iv is the liquid-vapour interfacial energy and 0 is the contact angle. Note that, by convention, the radius of curvature of the concave meniscus r has a negative sign. Initially the network is compliant in alkoxide derived gels and the small compressive forces cause it to contract into the liquid. The radius of the meniscus is m u c h larger than the pore radius (Fig. 8.7a). As drying proceeds the network becomes increasingly stiff, because particles come into contact, the porosity decreases, and the radius of the meniscus decreases continuously with a related increase of the capillary (compressive) force. Finally the radius of the f
w
(a)
(b)
(c)
Fig. 8.7. Schematic illustration of the formation of a liquid-vapour meniscus in the pores of a gel during drying: (a) and (b) are in the CRP with (b) the critical pohlt situation; (c) is in the FRP1 withf being adsorbed films at the unsaturated pore walls, w is the width of the drying front.
273
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
0.Z0
o o Evaporation . _ _ .g_ _ o D _ _ . - o - - - - rate of a ~__ o distilled water ___
o ,i
0.15
,~ 0.10 a
o
Q
C
R
J
P
U
~
FRP1
.....
"-1 -..
-
FRP2
J Q W O
=
(I.05
100
80
60 Water in gel
40
21)
0
(%)
Fig. 8.8. Rate of water loss from alumhla gel versus water content of gel durhlg drying for various initial thickalesses (0.75;<)3.0; gl 1.8;k 0.8 mm). From Dwivedi [22]. meniscus becomes equal to the pore radius (assuming for the moment a uniform pore size). This marks the maximum obtainable capillary pressure and is _ called the critical point. It is also the end of the CRP. Beyond the critical point the capillary pressure cannot overcome further stiffening of the network and shrinkage stops. The meniscus recedes into the pores as shown in Fig. 8.7c. From this point on the drying rate decreases as shown in Fig. 8.8 and this marks the beginning of the first falling rate period (FRP1). The drying rate decreases for several reasons. In the first place the vapour pressure Pv above a concave liquid meniscus with radius rp decreases during the process. According to the Gibbs-Thomson (or Kelvin) equation Pv is given by Eq. (8.4): / 2VmTlvCOS(O)) P~ - Po exp -
rp R T
(8.4)
where P0 is the vapour pressure above a flat (free) liquid surface. Furthermore when drying proceeds, the distance between the receding meniscus and the gel surface increases. According to Darcy's law this implies an
274
8 m FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
increasing resistance for liquid flow and a decreasing capillary pressure gradient both resulting in a decreasing drying rate. The rate of evaporation (drying rate) We is proportional to the difference between Pv and the ambient vapour pressure Pa: ~re --
k(Pv -
Pa)
(8.5)
where k depends on a variety of parameters, e.g., the draft rate and the design of the drying chamber, the thickness of the gel, the type of liquid, etc. As will be shown in subsequent sections the drying rate is very important for the final microstructure as well as the "defect state" of ceramic membranes. Liquid flow in the FRP1 is driven by the gradient in capillary pressure [20] along adsorbed films (Fig. 8.7c) or continuous patches of adsorbed film (the funicular state). When drying continues the liquid film near the surface first breaks up into isolated pockets (pendular state) and finally this state spreads over the complete thickness of the gel. This situation is called the second falling rate period (FRP2) where evaporation takes place inside the gel body and the principal transport process is expected to be Knudsen diffusion of vapour. It should be noted that the last traces of liquid will reside at the necks between gel particles and so will have a very strongly curved meniscus with consequently low vapour pressure. To remove the last traces of the liquid requires rather drastic drying conditions.
Additional effects in non-idealised situations In real systems there is a range of pore sizes and pore shapes in the body as well as a certain inhomogeneity in their spatial distribution. With a given vapour pressure and with a small drying rate, the system is close to equilibrium and all menisci in the FRP1 have the same curvature. For kinetic reasons the larger pores will empty first. With large drying rates, there is no equilibrium, menisci have different curvature and differences in drying rate between members of the pore size distribution become larger. In interconnected pore systems this leads to pockets of smaller saturated pores which are contacted or surrounded by regions of drained larger pores. This leads to local variation in the (capillary) stress and, as will be discussed later, in the drying stress distribution. The liquid-vapour interface in the pore system is called the drying front. As shown by Shaw [7] it is fractally rough on the scale of pores, but stable on larger scale. The smaller the pore (and particle) size, the smaller the width w of the drying front (see Fig. 8.7c) with w defined as the distance from the most advanced to the least advanced part of the front. Shaw found that the value of w decreased as the velocity of the front increased: W
oc V -m
(8.6a)
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
275
where m = 0.48. Because v is proportional to the pressure gradient VP in the liquid according to Darcy's law, Eq. (8.6a) can be rewritten as w ~ ( V P ) -m
(8.6b)
Note that Eq. (8.6b) implies that the drying front roughens (w increases) as it advances into the gel body because with a given external pressure, the pressure gradient in the drained region decreases and so does v. The existence of a certain width of the drying front will have important consequences for the drying characteristics of films as will be discussed later.
Microstructural development during drying and critical stress During drying the rising capillary pressure forces particles to arrange into a closer packing. This is initially easy because of the flimsy structure of the network. As shrinkage proceeds the particles makes more and more contact with each other and start to form a gel (gel point); ultimately, they become too crowded to rearrange further and shrinkage stops. Thus the critical point can occur after particles make contact. In any case the drying process around the critical point is also an "ordering" process which determines the final microstructure. The final lyogel and related xerogel structures are determined by the magnitude of the compaction pressure (determined by P, Eq. 8.3)) particle shape, the way particles interact and form bonds and the relative value of the rates of compaction (drying rate) and interaction. For spherical particles of uniform size with slightly repulsive forces acting among them, densely packed structures can be formed with body centred cubic (b.c.c.) or random close packings. These have coordination numbers z and porosities Vp of z - 8 and Vp - 0.32 respectively for b.c.c, structures and z is ca. 4.5 and Vp is ca. 0.36 for random close structures. The ease with which these packings are formed without residual stresses depends also on the particle shape and is probably different for plate-shaped boehmite particles and spherical titania or zirconia particles. The presence of agglomerates can give rise to hierarchical, random close packings as shown in Fig. 8.9. If the agglomerates in the lyogel are not eliminated during drying a bimodal pore size distribution results. The less the structure is compacted during drying, the larger the pore size and the wider the pore size distribution will be. For spherical particles there is a relation between the cavity and throat sizes of the particle packing which depends on the type of packing. For slit-shaped pores obtained from plateshaped particles, the pore size is related to the plate thickness [2,4]. For a b.c.c, structure with Vp = 0.32 the cavity and throat radii (defined as the radii of the inscribed circles) are 0.29 and 0.226 of the particle radius, respectively [1]. In the formation of microporous membranes from gels the building units are usually more or less branched, small silica molecules which can interpenetrate
276
8 m FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
(A)
(B)
Fig. 8.9. Schematic representation of two mesoporous xerogels. (A) Random close packing of spherical particles with tmiform size. (B) Hierarchical rmldom close packing of spherical agglomerates.
during packing and simultaneously can react (condensation reaction). Larger drying rates limit the time available for reaction. In the sections on microporous membranes these phenomena will be discussed. The critical stress (at the critical point) results from the largest obtainable pressure for compaction and is given by Eq. (8.3). This equation can be rewritten in easily measurable experimentally parameters as Spb
Pc -Ylv cos(0) [1 - p
/
(8.7)
where S is the specific surface area of the drained network, Pb is its bulk density and p its relative density. For alkoxide derived gels, using Ylv cos 0 values of 0 . 0 2 - 0 . 0 7 J / m 2 and S values of 200-800 m 2 this yields Pc values of 2-200 MPa. When the pore fluid is water (Ylv = 0.072 J / m 2, 0 - 0 ~ and the average pore radius is 2.0 nm this yields Pc-- 70 MPa.
Drying stress, deformation and cracking When the liquid evaporates from a piece of gel and the liquid-vapour meniscus recedes into the pores, the liquid goes into tension which is balanced by compressive stress in the solid network (resulting in capillary shrinkage). Liquid flows from the interior of the gel to the liquid vapour interface and the lower the permeability of the gel, the larger the pressure gradient is that develops. Related with this pressure gradient is a strain gradient, with the surface tending to contract more than the interior of the (free) gel phase. It is the difference between average and local free strain (the differential strain for an elastic material) or strain rate (for a viscous material) that provides the drying stress in the network. This is illustrated in Fig. 8.10. As shown in Fig. 8.10 the drying plate is thought to be cut into slabs parallel to the drying surface. The slabs closer to the surface would shrink more due to
277
8 - - F U N D A M E N T A L S O F M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
(a)
t [
'"
,
,
,
J
i
tl
t
.
. . . . . . . . . .
.
.
I 1
.
,2
,,,,
.
.
.
I
I
n t
. . . . . . .
.
,
I I
'i i i
......
-s
.
I I ........
i
.
i
", ," ", ,
,'
i..... I I
__1
(b) ..t-
..... 2,
, -
_]1~
4
m
m
1 I 1
I
I 2
I
I 3
I
! 41
Fig. 8.10.Schematic illustration of the origin of drying stresses. The length of the slabs in (a) represent the free strains, the uniform lengths in (b) represent the true strain. the higher tension in the liquid (pressure gradient). Because they are bound together they must shrink the same amount, so thesurface region is stretched and the interior is compressed. In addition to this overall stress variation, there are local stress variations due to a non-uniform distribution of pores a n d / o r porosity or of network properties (viscosity, visco-elastic relaxation). Equations for the pressure and drying stress distributions are given in a free gel body dried from both sides by Brinker and Scherer [1] for a number of different conditions. In most of the calculations a uniform porosity and permeability in the gel is assumed. This seems inconsistent with the above-mentioned differential strain (rate). According to Brinker and Scherer density gradients in dried gels are experimentally not observed and so must be small enough to be ignored in a first approximation. The results for the stress in a drying plate are given by Eqs. (8.8) and (8.9) [1]:
(~x_CNHg(_~_l[o~cOShsinh (00(~
(8.8)
278
8 -- FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
with
O~ =
~
TIL L 2
(8.8a)
DHg
or when (x is small ((x ___1): (Jx "~ CN ( Lrl2DVE )"~ : \[(7.2 -~ -
1"~/ 3]
(8.9)
with I)E is the constant evaporation rate (cm 3 per cm 2 of gel per second), 11is the viscosity of the liquid, D is the diffusivity (in Darcy's law), CN a constant relating the pressure in the liquid and the stress in the network, L the thickness of the plate, z the coordinate in the thickness direction, Hg is the viscosity (uniaxial modulus) of the gel. The equations show that there is a parabolic distribution of stress with the greatest tension at the drying surface (z = L). The stress increases in proportion to the thickness of the plate and the rate of evaporation; i.e., the stress becomes larger the steeper is the pressure gradient. These results account for the observation that cracking is most likely for thick gels (large L) and high drying rates. Note that the difference in pressure through the thickness of the body, giving rise to the stress, is usually much smaller than the value of the capillary pressure Pc at the critical point but there is a relation between them. The greater the capillary pressure, and the lower the permeability, the greater is the stress at the critical point. The maximum stress at the critical point approaches PRwhen the evaporation rate is very high and cracking is most likely at the end of the CRP and the beginning of the FRP1. Stresses in the CRP at the surface of drying plates, cylinders or spheres decrease in the ratio 1/3:1/4:1/5 respectively and so the highest stress level is found at the surface of plates.
Non uniform contraction and warping Local variations in pore size a n d / o r porosity as well as rigid inhomogeneities will give rise to local variations in the stress distributions and to residual stresses in the final xerogel. This happens because larger pores are emptied first by evaporation after the critical point with the consequence that the wall between adjoining pores is subjected to uneven stress which can cause cracking. Very few studies has been devoted to this aspect due to the absence of the necessary data. If a plate dries from only one side, the pressure distribution is asymmetric. Since the network is compressed more on the drying side, the plate becomes concave towards the drying side (so-called warping). As discussed by Brinker and Scherer [1] the curvature is greater when the evaporation rate is great and
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
279
the permeability and the viscosity of the gel network a n d / o r the gel thickness is small. An interesting observation is that in the course of the FRP the curvature of a free plate reverses and a permanent warping results. This is because, after the CRP, the unsaturated portion of the gel is compressed less than the saturated (non-drying) side. The fact that the warping is permanent means that viscous deformation of the unsaturated part of the gel is possible after the critical point even during the FRP2. Beyond the critical point (FRP1) the gel expands slightly as drying continues. This implies that around the liquid-vapour interface (the drying front) differential strains (and stress) gradients are relatively large. As the saturated region becomes thinner, its contraction is more effectively prevented by the larger unsaturated region. This raises the tension in the network of the saturated region and causes that cracks in drying gels often originate near the non-drying surface, as observed by Simpkins [19].
Stress in supported films The main difference between the stress in a plate and that in a supported film (which is a thin plate) is that the stress in the former is caused by the internal pressure distribution (differential strain), while the stress in the film results from the external constraint by the substrate. The situation can be visualised by Fig. 8.10 considering that the thick rigid substrate will be only very slightly deformed by the thin, much less rigid, drying film. This means that in a first approximation the deformation (stress) of the support can be ignored (steps 2-4 in Fig. 8.10b) and the strain adaption comes almost completely from the thin film (step I in Fig. 8.10b). Nevertheless the stress in the drying film is large enough to cause small, but measurable, warping of the total system. This will be used to measure the stress in the thin films as will be shown in a subsequent part of this chapter [13,18,26,27]. The strain in a supported film does not depend on the average pressure. The stress G(x) during drying in the CRP in the plane of the film can be described now by Eq. (8.8), omitting the figure -1 in the brackets of Eq. (8.8). For thin films and large drying rates and taking CN = 1 (small syneresis) this indicates that the stress in the film can be as large as the capillary stress in the liquid. As will be discussed in the section on typical experimental results, this enormous stress does not, remarkably enough, cause fracture of gel films below a critical thickness in the range of 0.5-10 ~tm for a single step process and depending on the material, the support, precursor conditions and process conditions during drying.
Elements of cracking theory Fracture of brittle materials depends on the presence of flaws (cracks) that amplify the stress applied to the body. The theory of linear elastic fracture
280
8 -- FUNDAMENTALS OF MEMBRANE TOP-LAYERSYNTHESISAND PROCESSING
!_
(~'x <
-
.
.
.
.
.
.
"IC
' ,....
.
.,., i,.
O'x W
.
.
.
.
.
II
Fig. 8.11. Schematic illustration of stress formation and stress relaxation by cracking at the tip of a crack with length C. r~x and r~c are the externally applied stress and the stress at the crack tip respectively. Zone I is the stress relief zone, w represents the irregular drying front zone w, hi zone II no stress relaxation occurs.
mechanics [22,23] predicts that catastrophic crack propagation occurs when r~i o~ C~xq-Cand that: (~x ~
--- KIC
(8.10)
where the materials property KIC is called the critical stress intensity factor (or fracture toughness), Ox is the applied stress, (~i is the critical stress at the crack tip and C is the length of the crack. As discussed by Brinker [1] and illustrated by Fig. 8.11, gels can relax stresses (delivered by the drying stress) by viscous and visco-elastic deformation. Consequently the upper part of the crack zone (Fig. 8.11) is free to relax (contract) in response to the compression applied by the liquid, but the gel network ahead of the crack (zone II, Fig. 8.11) is constrained. Large (tensile) stresses occur especially in the zone around the crack tip (zone W). Fracture occurs as soon as the value of C~csurpasses the strength of the gel and this occurs sooner the larger the (effective) crack length is and the larger are the drying stresses (and thus drying rates).
8.1.2.4 Consolidation to the Final Membrane Structure by Heating The dried xerogel must be stabilised by heating to a sufficiently high temperature to obtain a microstructural, mechanical and chemical stable ceramic membrane layer.
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
281
P
{81
{b!
Fig. 8.12. Schematic picture of contact points (a) and neck formation and sintering between particles (b), with radius R and neck width x. Path I represents matter transport from the grain boundary to the neck surface with curvature radius p.
In this overall process two sub-processes operate which partly overlap. At relatively low temperature, usually up to about 300-400~ the strongly hydrated amorphous gel particles are transformed to more crystalline, mainly dehydrated particles. This step is usually called calcination. In this step organic additives also have to be burned out. Usually considerable volume changes of the constituting crystallites occur. The particles formed, usually oxide, form a packing with contact points, the number of which per unit of volume is determined by their coordination number and particle radius (see e.g. Ref. [24]). On these contactpoints a "physical reaction" takes place which results in the formation of necks between the particles as shown in Fig. 8.12. This process is called the initial state of sintering. Sometimes real phase transitions occur during the calcination and/or sintering steps which are usually accompanied by volume changes of the constituting particles. It will be clear that all these volume changes will give rise to local stresses in the layer and between the layer and its support and so enhance the risk of cracking. An example of such a phase transition during calcination is the transformation of plate-shaped boehmite particles in the xerogel to ?-alumina particles in the calcined layer at T > 350~ During heating at higher temperatures (>300-400~ the initially formed necks become broader (Fig. 8.11b) and some shrinkage occurs and, finally, when the width of the neck increases the shrinkage becomes larger and the microstructure of the packing changes considerably (second stage of sintering). This is shown in Fig. 8.13. During the later part of the first sintering stage the pore size starts to change and the average pore size will grow because smaller particles are "consumed" by larger ones, resulting in larger mean pore size. The porosity will decrease in this process because particle centres approach each other (see Fig. 8.11).
282
8 -- FUNDAMENTALSOF MEMBRANE TOP-LAYERSYNTHESIS AND PROCESSING
(
Changes in pore shape
)
.
.
.
.
. ~ ,r
Change in sha~_ and shrinkag~
AT.
Fig. 8.13. Schematic picture of changes in pore shape during sintering.
Sintering mechanisms are (i) solid state diffusion from areas with relative small curvature to relative large curvature in the neck area (see Fig. 8.11) along the grain boundary in the neck, and (ii) viscous sintering of amorphous or glassy particles by viscous flow and deformation of the particles. In both cases the driving force for sintering is a decrease of the surface energy of the system with matter transport going from areas with relative large convex curvature (high Gibbs free energy) to areas with small concave curvature (low Gibbs free energy). In a rough approximation this driving force between sintering of two particles is given by the gradient of the "sintering pressure" Ps, with Ps given by: Ps ~ 7sv
/1 -
-
P
(8.11)
where R is the particle radius and p is the neck radius (Fig. 8.11). Because p is related to R the sintering activity and the sintering rate becomes larger the smaller are the particle and related pore sizes within the membrane layer and the larger the diffusivity in the solid or the lower its viscosity. A detailed discussion of sintering theory is beyond the scope of this chapter and the reader is referred to, e.g., Kingery [23] or to Ref. [1] for viscous sintering. The pore size can be increased at the cost of a decreasing porosity by controlled heating of the membrane in the temperature range of 400-1000~ for the most common membrane materials. Phase transformations are accompanied by an increase of the sintering activity and give rise to an enhanced pore growth [13] and so should usually be avoided. This goal can be achieved by suppressing the phase transformation or by shifting the phase transformation temperature to higher temperatures by
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
283
appropriate doping of the membrane material with certain elements, e.g., La in ~/-A1203 or TiO2 [36,37,40,41]. Finally the membrane system must be cooled down from the calcination or sintering temperature to room temperature (or the reverse in high-temperature applications). This will result in stresses due to mismatch of the thermal expansion coefficients between layer(s) and support with an enhanced risk of cracking or spalling of the layer as result. A discussion of this point is given in Section 8.1.3.4.
8.1.3 Illustrative Experimental Observations of Stress and Cracking in Membranes 8.1.3.1 Stress Measurements in Supported Porous Membranes As has been shown in the preceding sections the risk of cracking and defect formation increases with the magnitude of the stresses in the gel (film) during drying and calcination. Knowledge of the magnitude of the developing stresses as a function of synthesis conditions and the relation between stress levels and defect formation is important for adequate process control to produce crackfree membranes. Stresses in thin films on dense support can be measured by different techniques as summarised by, e.g., Hofman [24] and Campbell [25]. Stress measurements of porous films on a porous support involve additional complications. Voncken et al. [26] and Kumar et al. [13,27] developed a new method based on the "cantilever beam method" for measuring stresses during the formation of mesoporous membranes on porous (alumina) supports. The method is briefly summarised below. As shown in Fig. 8.14, a thin (mesoporous) layer is applied by dip-coating on a (macroporous) alumina support in the form of a strip (beam) with given properties such as precise dimensions, porosity, pore size, etc. The strip is clamped in a solid block which can be precisely located in space above a detector (see below). During drying, drying (tensile) stresses in the film causes an upwards "warping" (curvature) of the beam which cause a deflection ~5with respect to the initial position of the tip of the beam. This very small end deflection is measured in a triangular set up with a laser displacement meter assembled together with a detector which is sensitive to lateral displacements. The whole assembly of cantilever block and laser/sensor meter is placed in a drying chamber in which temperature, relative humidity (RH) and air velocity can be controlled and measured by a small humidity sensor and anemometer placed just above the drying membrane and by wind screens placed in the drying chamber. The observed deflection 3 is monitored as a function of time in a data acquisition system. To reduce the noise level the values of 50 scans are averaged, the average constituting one data point. For further details, see Voncken et al. [26].
284
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYERSYNTHESIS AND PROCESSING
Support with thin qel Ioyer on top ~f s
s
s
o,
II.
'6 ! ! .,J
Fig. 8.14. The cantilever b e a m principle for m e a s u r i n g stress in a s u p p o r t e d m e m b r m l e . The deflection at the end of the b e a m is ~i. F r o m Voncken et al. [26].
The end deflection 6 is related to the overall stress, ~, in the top layer by Eq. (8.12) [24,25]" c~=
Es d2s8 3L2(1 - Vs) df
(8.12)
where Es and Vs are the Youngs' m o d u l u s and Poisson ratio of the support, L and ds are the free length and thickness of support respectively and df is the thickness of the film (membrane) on top of the support. Equation (8.12) is valid w h e n ds >> df, which is always the case with supported membranes, w h e n L is greater than twice the width of the support strip and w h e n 6 < ds (which is also the case for thin membranes on rather thick and stiff supports). The Youngs moduli of the support has been measured with four point bending tests, Poisson's ratio is taken to be equal to that of dense alumina. Values of Es and Vs are 65 + 4 GPa and 0.23 respectively. To obtain measurable values of 6, ds is taken about 0.4 mm, while df is in the range of 1-10 gm. The values of 6 range from 0.05-0.15 mm. A typical example of a stress diagram is shown in Fig. 8.15 for two drying
8 -- FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISANDPROCESSING
285
ZSO ~ . . . . . . . . .
Tensile stress
s i
200
1 s"
Crackin(j
150
,- "
," "--'-.-,
]
-
Increase of temperature 11~
o
o-
~ 100 L,.
50-
01~lIl~" t ~ 5 0
0 m
Compressivestress !
1
T
I
~r
v
l
I
V
| I T
400
Y
V
~
Y
I
I
I
l
V
80 0
v
I'v
v
f
1
Y
T|
i
T
T
120 0 Time is)
y
y
v
Ir
l
T
|
T T T ' T
160 0
I
l
v
!
I
~'y'l
2000
j'
1j'l
l''l
v
l
2400
Fig. 8.15. Stress-(dryhlg) time diagram of boehmite gel layers dried at 60% RH at 25~ (dotted lhle) and 40~ (solid lhle) with a whld velocity of 3.25 ms-1 hi the dryhlg chamber. From Voncken et al. [26]. boehmite gel-layers (which subsequently are converted to ?-alumina membranes by calcination as described in preceding sections). Changes in deflection (stress) due to drying always took place within the first 20-30 min after the start of the drying process. Other features of Fig. 8.15 will be discussed below. It should be noted that despite noise-suppressing techniques, due to experimental uncertainties in the m a n y parameters of Eq. (8.12) and due to support effects (see below), a relative large scatter in the m e a s u r e d stress levels has been observed.
Stress measurement during calcination Stress measurements at high temperature were performed with a similar equipment as used in drying experiments. The clamped strip-top layer combination is placed in a small tubular furnace positioned horizontally by a suitable sliding arrangement [28]. A small hole is provided at the bottom of the furnace for the laser beam. The exit at the bottom of the furnace must be placed close to the detector and so must be provided with a radiation screen and be cooled adequately. Effects of the support on stress measurement in membranes Complications in the interpretation of the results arise from the presence of a porous support as s h o w n by K u m a r et al. [13,27]. Water sucked into the pores
286
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
of the support during the slip-casting process forms a small wet zone in the support just below the lyogel membrane. Capillary forces in these wet zones cause compressive stresses and so before drying an upward deflection of the strip results in addition to that caused by the layer (membrane). As has been shown by deflection measurements during drying with a partially water-filled porous support [13, 28], the removal of water by evaporation straightens the strip and this is measured as a negative deflection. During drying of the supported lyogel membrane most of the water evaporates from the top (membrane) side of the strip and therefore the water in the strip acts as a reservoir in the initial part of the drying process. When all the water in the wet zone of the support below the lyogel film has been removed in the drying process, the contribution of the support to the total deflection vanishes, the support straightens and because at this moment the lyogel is wet (no measurable compression stress) a negative deflection occurs, similar to that of the non-coated support. This is illustrated in the initial part of all drying curves (see, e.g., Figs. 8.15 and 8.18). This part of the stress diagram is called the pre-stress development region (PDR). It is characterised by a small negative deflection, corresponding with an apparent compressive stress. The stress diagrams can be corrected for this effect assuming that the resultant stress is a linear combination of both separate deflection (stress) contributions. A second effect of the wet zone in the support is that the drying rate of a supported lyogel film is considerably smaller than that of a unsupported film of comparable thickness. Because the depth of the wetted zone varies, depending on slip cast conditions, the drying rates exhibit some uncertainty even when all external drying conditions are controlled. Finally it was found [13] that non-coated supports, heat treated to temperatures in the range 400-600~ give a considerable deflection during heating and cooling, as shown in Fig. 8.20. Deflection (stress) measurements during calcination must be corrected for this support contribution, again assuming the final deflection is a linear contribution of support and film. Note that there is a small permanent deflection (warping) of the support after cooling to room temperature. The origin of this warping is not clear, but might be connected with the processing (polishing) of the support surface which, in the subsequent slip-casting process, will be coated. A last complication results from the measurement of the thickness of supported layers. This can easily be done with SEM for calcined layers. Wet lyogel, or even dry xerogel, layers cannot be measured in this way. Reproducible thickness measurements on wet lyogel films could not be obtained with other easy-to-perform methods. Consequently layer thicknesses were measured after calcination. Estimates of the shrinkage in the thickness direction were made for supported alumina (boehmite) membranes dried at 40~ and 60% RH made with a standard precursor solution of I mol A1OOH/1 stabilised at pH = 4 by
8 -- FUNDAMENTALS
OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
First stage of drytng
y
CSR |.
i
T
g
R
-
f\ ....
/ L
/S
E
a
-
....
S
n
R
287
i:: rate period
/._o !
S S
.
. CRP
1 .cRp2-
.
.
.
.
.
Time
_ ,
Fig. 8.16. Schematic representation of drying rate and stress development as a function of drying time. SDR and CSR are stress development and constant stress regions, respectively. H N O 3. The thickness of the film was found to decrease linearly with increasing drying time until after about 500 s the thickness became constant. Obviously this is the critical point discussed in preceding sections. The observed shrinkage from wet lyogel to critical point amounts to about 25 ~tm, the final thickness after calcination is about 5 ~tm which is somewhat smaller than the thickness at the critical point at which the m a x i m u m stress occurs. Sols with lower concentration yield similar final thickness.
8.1.3.2 Drying Characteristics of Membranes In order to relate stress development curves to drying rate characteristics, K u m a r et al. [13,29,31] measured the drying characteristics 6f thick (50-150 ~tm) alumina (boehmite) and titania films poured on glass plates and dried as such. Thin films could not be investigated due to insufficient sensitivity and accuracy of the measuring methods. In all cases, the films exhibited the behaviour predicted by drying theory: an initial constant drying rate period (CRP) followed by a falling drying rate period (FRP1) as schematically shown in Figure 16. The second falling rate period was not clearly observed. The change-over from CRP to FRP1 became smoother and more gradual with thinner layer. This behaviour is also observed in granular (clay) materials and has been explained by a smoother fluid concentration profile within the thinner drying samples [30]. Another reason might be the more pronounced effect of the (irregular) drying front with width w in thinner specimen. If w becomes comparable with the membrane thickness, CRP and FRP1 overlap and cannot be distinguished clearly.
8.1.3.3 Stress and Cracking in Membranes During Drying Extensive stress measurements has been reported in mesoporous alumina [13,26], titania and alumina-titania composite membranes [13, 27] supported by
288
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
(~-alumina supports with an average pore diameter of about 0.16 ~tm. A brief summary of the results is given in Ref. [31]. The membranes were synthesised from A1OOH (boehmite) sols with a typical (standard) concentration of I mol A1OOH/1, stabilised at pH = 4 with HNO3. TiO2 sols had a typical concentration of 0.3 mol TiO2/1, composite membranes of A1OOH and TiO 2 were obtained by mixing the sols. In a number of cases polyvinylalcohol (PVA) with an average molecular weight of 72000 Dalton was added to the boehmite sol (35 g PVA/1 sol). TiO2 sols were sometimes mixed with 3.5 g hydroxy propylcellulose (HPC) with an average molecular weight of one million Dalton. The final membrane thickness was dependent on dipping time and is usually given after calcination at 600~ resulting in y-alumina or titania (anatase) membranes. Drying rates are varied by controlled changes of conditions in the drying chamber. The wind velocity has been varied between 3.25 m / s to 0.5 m / s (wind screens present). The last value corresponds with a mild drying regime and is used unless stated otherwise.
Stress and cracking as a function of drying rate Figure 8.15 shows the stress development in drying boehmite membranes (layer thickness 2-3 ~tm after calcination) using a high wind velocity. This corresponds with a relative large mass transfer coefficient k in Eq. (8.5), and thus with a high evaporation rate. A higher temperature also results in a higher drying rate. The combination of these two factors results in a peak stress value of 160 MPa which then decreases sharply to 80 MPa. The peak value could be correlated by microscopic inspection with the sudden appearance of microcracks. These cause stress relaxation in accordance with the theory discussed before. Further peaks, followed by relaxation to lower stress values, could be induced in the constant stress region by a sudden increase of the temperature. A decrease of the drying temperature, and thus of the drying rate, resulted in continuous increase of the stress to about 180 MPa without the occurrence of stress peaks (cracking). The surprising conclusion so far is that these 2-3 ~tm thick gel layers are able to withstand stresses up to 180 MPa. These membranes are however very vulnerable and crack easily afterwards and lower stress levels are required. This can be obtained with a lower wind velocity (0.5 m / s ) as shown by Fig. 8.17. The membrane thickness (after calcination) was 4 ~tm. The stress increases with increasing drying rate (lower R.H.) sharply to about 50 MPa at 60% R.H. and more gradually to about 70 MPa at 20% R.H. A clear relation between stress and membrane thickness could not be found due to the too large scatter in the results and the uncertainty in the actual membrane thickness (before calcination).
8 w FUNDAMENTALS OF MEMBRANE TOP-LAYERSYNTHESISAND PROCESSING 80
40oA,.-,.
60
~ v
'-
40 u) ffl
V "V"
y'
289
20~lrl v
~
" l
20
4-'
U3
0 -20
" 0.00
~' 2.20
'
4.40
Time
6.60 in
8.80 11.C~ (Thousancls)
sec.
Fig. 8.17. Stress (drying) time diagram of boehmite gel layers dried at 40~ with a w h l d velocity of 0.5 ms -1 with stepwise changes hi relative h u m i d i t y (RH), starting with 70% RH. From K u m a r [13].
Stress patterns of titania without additives were always highly irregular with many peaks and rather low stress levels, indicating severe cracking of the top layer [13,31]. Additives are necessary to relax this problem (see below). Aluminatitania membranes with 25-40 mol% titania yield stress (deflection) curves without peaks (and so without observable cracking). These observations support the phenomenological experience that production of defect-free titania membranes by sol-gel techniques is more difficult compared with ?-alumina membranes.
Relaxation and bondformation in drying gel layers As shown by Fig. 8.18~he stress level of about 40 MPa obtained in a boehmite layer (thickness after calcination of 5 ~tm) after drying at 40~ and 60% RH could be relaxed to zero (after correction for the negative deflection due to support wetting) by an increase to 90% of the relative humidity and again to 40 MPa by turning back to 60% RH. Such a complete stress reversal was also noticed at 50~ and changes of 90% RH to 20% RH and back. Obviously the process of stress formation is, under these conditions, completely reversible. According to Eq. (8.4) it is estimated that at the average pore diameter of 3.0 nm obtained in boehmite membranes (from N 2 desorption measurements [4,13] menisci are formed and so pores are filled at about P/Po = 0.5 (50% RH) at 40 ~ The critical capillary pressure at this pore size is 70-90 MPa. So the observed stress at 40~ and 60% RH could be explained by capillary pressure. This is not the case, however, for the stress values obtained at lower RH values where, certainly at 20% RH and T > 40~ pores with a diameter of >3 nm are emptied. Nevertheless the gel layer can bear stresses up to 60 MPa. This points to the formation of bonds between gel particles in the dried layer which allow relaxation to varying stress levels when RH conditions are changed.
290
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
60
Tensile stress
50. 'X
r
%a J~ I/~
40.
~
I ~.
II
~.
I
" I ~ 'f
r
30.
I
,/
0
Q,.
P%,
E zo.
I
t_.
%i1
,,n 10.
9 0 % RH ~
04a,~ -104
60 % RX
\
J
60% RH
V
Compressive stress
4
-20 0 (c)
0 Time
Is)
Fig. 8.18. Reversiblestress diagram of boehmite membranes dried at 40~ with RH values changhlg cyclic between 60 and 90% RH. From Voncken et al. [26].
Effect of additives on stress and microstructure in membranes during drying It is well documented that the addition of organic polymeric additives to precursor sols promotes the formation of defect-free membranes [4,12,33]. To investigate this effect, stress measurements were performed on boehmite membranes obtained from standard sols (1 mole A1OOH/1) mixed with different amounts of PVA solutions (containing 35 ml PVA/1, molecular weight 72000). As s h o w n in Fig. 8.19 [13,31] the stress in the constant rate region decreases with increasing a m o u n t of PVA to zero at a weight ratio PVA/y-A1203 >_0.25 (0.7 ml PVA solution per ml A1OOH sol). More recently the effect of additives on the formation of zirconia and alumina-zirconia membranes was further investigated by Z6ter [32]. Z(iter showed that the molecular weight had no important influence but that the pore size showed a m i n i m u m at certain concentration of the a d d e d PVA. The addition of PVA strongly p r o m o t e d the formation of defect-free membranes. These results might be explained by either stress relaxation due to polymer molecules absorbed in the porous structure a n d / o r to a strong modification of the drying process by decreasing the drying rate and a partial elimination of the formation of menisci due to filling of the pores with PVA clews. This last fact is
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYERSYNTHESIS AND PROCESSING
291
40 fl_ E -q~ q~ d)
3O 20 10 9
O
-
0.00 PVA
a.
.
L,,
a,
0.20 cont.
0.40 ml/ml
0.60 of
AtOOH
0.80 sol
Fig. 8.19. Stress in the constant stress region of a drying b o e h m i t e m e m b r a n e as a function of the a m o u n t of PVA a d d e d to the precursor solution. Drying conditions: 40~ and 60% RH. From K u m a r [13].
not completely consistent with the dimensions of the PVA molecules. For a molecular weight of 72000 the gyration radius is 12 nm which is larger than the observed pore diameters. The linear polymer dimension is 16 nm but the thickness of the molecular chain is much less. Pore filling should be possible then by adsorption of the main chain in the plane of the particle (pore) surface. 8.1.3.4 Stress Formation in Membranes During Calcination The development of stress during calcination is shown in Fig. 8.20 for boehmite membranes calcined at 600~ (thickness after calcination is 5 ~tm). Curve c in Fig. 8.20 represents the curve which is corrected for support effects (see the preceding section on this subject). Three heating and cooling cycles are shown. During the first heating the Al-hydroxide particles of the gel are transformed to boehmite and subsequently to (hydrated) y-alumiNum oxide particles and the shape of the first peak of curve c differs from the subsequent peaks. The maximum tensile stress calculated from the deflection amounts about 30 MPa. In the first cycle the shape of the heating period (oxide formation) is different from the cooling period (thermal mismatch between layer and support). After the first cycle, the shape of the subsequent peaks are identical and all processes seem to be reversible. The maximum stress is obtained after each cycle and so no stress relaxation occurs. Note that the deflection measurements start with a dried membrane which already shows a certain deflection which is equivalent to a tensile stress level of 30-40 MPa. It is not clear at the moment whether it is allowed to sum up these two contributions or that the drying stress relaxes during heating and is replaced by stresses originating in the phase transformation/thermal mismatch processes. In any case when summing up is allowed the final stress in the y-alumina after cooling down is not greater than 30 MPa; in the other case it is zero.
292
8 -- FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
7O0
0.10.
v b-
x .. f ~
-0.~
t
",., V ",,I;'
,/ I
.300 I
..,
time (,see)
303(XX)
Fig. 8.20. Deflection (stress) versus time diagram during the cyclic heat treatment (calcination) of boehmite membranes converted to T-alumina at 600~ Heating and cooling rates were 60~ From Kumar [13]. Curve a (dotted): blank run, s u p p o r t only; curve b: actual run with supported membrane; curve c: deflection of b corrected for support effects.
8.1.3.5 A Model Discussion of Stress and Avoiding Cracking The overall scheme of stress formation and cracking From the preceding sections it has become clear that tensile stresses developing during the drying of the lyogel and subsequent calcination are important causes for defect formation by cracking. A tentative scheme to account for a number of data emerges but many details are unknown. Nevertheless some trends are qualitatively predictable. The relation between drying process and stress formation is shown in Fig. 8.21. In accordance with drying theory in porous membrane layers a constant drying rate period (CRP) and a falling rate period (FRP) can be distinguished. The transition between them is sharper with increasing thickness. A clear explanation has not been presented but is probably related to the width of the drying zone (see Fig. 8.7) which increases with width w of the pore size
distribution. In the first part of the CRP(1) particles in the lyogel layer are concentrated until shrinkage becomes hindered and a meniscus with a large radius starts to form. At a certain moment shrinkage stops because the network becomes rigid and the liquid meniscus starts to increase its curvature. This is the second stage of the CRP(2). At the end of the CRP2 the meniscus radius gets its minimum value, which is equal to the pore radius and with further evaporation the meniscus starts to recede into the pores of the now rigid network which marks the beginning of the FRP(1). The transition of CRP2 to FRP1 is called the critical
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
293
D First stage o f city,rig
r
y
i N g
R a
t e
CSF:~
/\
S T R E S S _ ...
',,,
=
C9 R I D 1 C R P 2 . 9
,
o
Time
---.
|;
Fig. 8.21. Schematic representation of dryhlg rate and stress development as a function of time.
point (occurring at the critical time). The formation of a liquid meniscus is accompanied by capillary tensile stresses in the liquid which cause shrinkage of the gel network. This shrinkage of the film is constrained by the support in a lateral direction and consequently tensile stresses develop in the gel network of the film which are related to the drying stress. During the CRP2 a considerable reorganisation of the gel network takes place and so drying is also a rearrangement process which determines the pore size distribution. This capillary stress (see Eq. (8.3)) is related to, and usually smaller than, the developing drying stress (see below). The larger the capillary stress the larger the drying stress which reaches its maximum at orjus t afte r the critical point (see Eq. (8.7). During the FRP1 the meniscus recedes into the pores of the network and the interface between the unsaturated (relatively dry) and the saturated part of the network will be called the drying front with a width w (see Fig. 8.7) related to pore size distribution and the drying front velocity v (which is related to the drying rate) as shown by Eq. (8.6). Liquid transport to the drying front is by liquid flow through the liquid films in the funicular state. Bonds are formed at the contact points of particles within the gel. Tensile stresses in the film network are caused now by two effects: (i) the tensile stress gradient over the film thickness and (ii) local stresses due to non-uniform contraction of the gel network which are related with non-uniform pore size distribution (large pores empty first) and non-uniform stress relaxation of the network. This overall stress gradient is larger with large drying rates, with lower permeability of the network and with larger film thickness and is lower with decreasing viscosity (or viscoelastic relaxation rate) as can be see from Eqs. (8.8) and (8.9). The gel network in the unsaturated dry part expands slightly when the drying front proceeds because capillary forces no longer act here and the stress
294
8 m F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
gradient changes relatively sharply in the drying front. When the drying front in the film approaches the support, tensile stresses become larger in the remaining wet part of the film and risk of cracking increases. When drying proceeds, all pores are emptied and only isolated areas containing liquid (pendular state) are present. This is the start of FRP2, characterised by a sharp decrease of the drying rate and by slow diffusion of gas (water) through the pores. Because the vapour pressure is decreased above concave menisci (see Eq. (8.4)) the last traces of liquid accumulate at the necks between particles in the gel (see Fig. 8.12). This might be important for stress relaxation due to varying drying conditions (see below).
Cracking phenomena Cracking most probably occurs at sites with large tensile stresses and at pre-existing defects which then grow as soon as the magnitude of the tensile stress surpasses a critical value (see Eq. (8.10)) which is dependent on the size of the defects and on the strength and toughness of the gel network. The macroscopic tensile stress is maximum at the critical point and at the end of the FRP1 and so the risk of cracking is largest here. After rewriting Eq. (8.9) it can be shown [1] that (~x ~ PR L2 I~L /DHg
(8.13)
with PR the maximum capillary stress and Hg the viscosity of the gel network respectively. The other parameters are defined for Eq. (8.9). This cracking will be reduced with all parameters reducing the magnitude of r~xas given by Eq. (8.13). In addition cracking may be caused by local stress concentration due to microscopic processes on the scale of microstructural inhomogeneities. One of these processes is the formation of subcritical microcracks. These form in a network with a wide pore distribution due to the existence of too large capillary forces across the pore walls between large and small pores causing fracture of this wall. If the inhomogeneity in the pore size distribution increases these microcracks may percolate into a macroscopic critical flaw and catastrophic failure occurs. These local stresses form especially at the irregular drying front and give rise to cracks with a typical length of the order of w. The combination of Eqs. (8.6) and (8.13) with m = 0.48 then yields [1] C~x~
~ (drying
rate)3/4
(8.14)
Equation (8.14) shows that the risk of cracking increases with increasing width of the drying front w (larger cracks) and with increasing drying rate.
Stress and cracking during calcination and sintering The total stress developed during calcination and sintering is the sum of: (1) stresses caused by volume changes due to the conversion of amorphous,
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
295
strongly hydrated phases to more or less crystalline, much less hydrated phases (calcination); (2) macroscopic sintering stresses due to constraints of the sintering film by the support; (3) local stresses due to different sintering rates between regions with different densities (inhomogeneities). Residual agglomerates and impurities are main causes of local stress; (4) stresses due to phase transformations occurring in a certain temperature region; (5) stresses due to thermal mismatch between the expansion coefficients of layer and support. These occur during cooling down after the sintering process and again in the heating up cycle during applications at high temperature. Contributions 1, 3, 4 and 5 can have positive as well as negative signs, causing tensile as well as compressive stresses; contribution 2 always causes tensile stresses in the layer. This means that hardly any prediction can be made of the total effect. It can be concluded however that the magnitude of the stress, and thus the risk of cracking and spalling off of the layer from its substrate, is minimised with better homogeneity, avoidance or suppression of phase transformations, small thermal mismatch and low heating and cooling rates (smaller temperature gradients). The sinter stress is proportional with the sintering pressure (see Eq. (8.11)) and so will be larger with smaller constituent grains. Thus the problems will be larger the smaller are the grains and pore size within the sintering films.
Discussion of the model The tentative model given above accounts for many of the experimental observations. It is qualitative in nature due to lack of data and it yields some results which are difficult to explain. The first uncertainty concerns the fact that the theory is derived for bulk materials, thin-walled bodies or thick films. It is not certain whether extrapolation to thin films is allowed. This is particularly the case when the width of the drying front approaches the film thickness. Nevertheless, the model predicts in the correct way the beneficial effect, avoiding cracking, of using liquids with a lower surface tension a n d / o r the addition of surfactants during drying as well as supercritical drying (elimination of capillary stress). The same holds for a decrease of the drying rate. In Section 8.1.3.3 it is shown that decreasing drying rates does indeed decrease the measured stress levels and thus the risk of cracking. It is not known at which place the critical point as defined by the point where the gel network becomes "rigid", is situated in Fig. 8.21. This implies that it is not known whether or not some relaxation of the structure (as expressed by Hg in Eq. (8.13)) takes place in the FRP1 region. That some relaxation takes place is
296
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
again shown in Section 8.1.3.3 by the occurrence of reversible changes in the observed film stress with changing drying conditions. The beneficial effect of additives might be explained by improved stress relaxation but m a y also be due to decreasing the drying rate a n d / o r the surface tension. That bonds are formed between particles is inferred by the fact that the gel layers are able to bear considerable stresses. These bonds are sensitive to the presence of stresses and allow stress relaxation to occur. The relation between stress relaxation and cracking on one hand and particle shape on the other hand is not known. The relative ease of preparing y-alumina membranes might be due to the relative ease of rearrangement of the particles and easy stress relaxation in plate-shaped boehmite particles and the isomorphous transitions to plate-shaped y-alumina at about 300~ the transition also being accompanied by a relatively small volume change [2-4]. With spherical particles (titania, zirconia) stress relaxation might be more difficult. The easier formation of defect poor composites of alumina and titania (with spherical particles) supports the beneficial effect of plate-shaped particles. With mild drying conditions the presence of capillary stresses might contribute to interparticle "bonds", at harsh drying conditions (20% RH, 90~ this is not the case. This raises questions about the character of the stress itself. In Section 8.1.3.3 it is shown that tensile stress levels of 40 MPa are observed without cracking in drying boehmite films with pore diameters of --3 nm. Under harsh drying conditions stress levels of 150-180 MPa are observed just before the m o m e n t of cracking. These stress levels far exceed the strength of bulk gel particles but are in accordance with the estimated capillary stresses. This suggests that even in the FRP2 at high drying temperature and harsh drying conditions residual capillary stresses at the contact points might be present or that thin films behave in a special way. The experimental observation that there exists a critical thickness above which cracking occurs cannot easily be explained. Brinker [1] discusses a theory which explains that very thin layers can bear much larger stresses because critical cracks cannot be formed unless a certain critical thickness is surpassed. This thickness is estimated to be equal to or less than I ~tm and Brinker comes to the conclusion that thicker films will always crack. This is certainly not the case for alumina, titania and zirconia films for which much larger (alumina) to larger (titania) thicknesses are observed. As shown in Table 8.2 critical thicknesses of a few ~tm in single-step dip-coated films occur and critical flaws are smaller than this thickness and so can be present. Surprisingly, the stress levels found after calcination are rather low. They are, however, the result of a number of contributions which might be favourable for 7-alumina on (x-alumina supports. Much more work is needed in this area.
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
297
8.1.4 Thermal Stability of Ceramic Membranes Mesoporous membranes calcined or sintered at relatively low temperature (300-400~ and for a short time (i.e., a few hours) are not thermally stable. Heat treatment for prolonged time a n d / o r at higher temperature causes an increase of the average pore diameter and a decrease of the porosity. The strategy to obtain thermally stable membranes with no further change of pore characteristics during application is twofold. Heat treatment of the membrane for prolonged time (100 h or more) at a temperature about 100~ above the intended temperature of use usually produces sufficiently stable membranes. The price to pay is an increase of the average pore diameter. To limit the pore growth, suppress phase transformations and further to enhance the pore stability, the membrane is doped with a few percent of metal ions such as La 3+. Van Veen et al. [34,35] showed that unsupported mesoporous y-alumina membranes, prepared by the sol-gel method, with a pore diameter of 3-4 nm after 5 h sintering at 600~ increased their pore diameter by about 20% after 600 h sintering at this temperature. Membranes sintered for 5 h at 600~ were not stable (pore growth) at temperatures above 425~ for prolonged time. After stabilisation for 600 h at 600~ no pore growth was observed after a further 600 h treatment at 500~ This type of phenomena has already been observed by Leenaars et al. [4] and systematically investigated by Lin et al. [36-38]. Unsupported y-alumina membranes heat treated for 30 h exhibited a continuous increase of the average pore diameter from about 3.2 nm at 450~ to 6-10 nm at 1000~ depending on the~synthesis conditions. Above the la~tt6r temperature a very sharp, explosive increase occurs due to the y-0t phase transformation [36]. In comparison with pure y-alumina, mixing the precursor boehmite sol with 3% LaNO3 or impregnation of calcined y-alumina with LaNO3 solution stabilised the unsupported membrane. After 120 h at 800~ the La doped system had a pore diameter of about 5 nm compared with about 9 nm for the pure y-alumina. Up to 1100~ the pore growth increased steadily to about 20 nm; above this temperature pore growth became explosive. The effects of doping with La and adding PVA on the pore growth of defect-free supported y-alumina membranes was reported in another paper by Lin and Burggraaf [37]. Permeability values show a trend in accordance with those of the pore growth for unsupported membranes. The addition of PVA only increases the average pore diameter but after addition of La to the PVA containing precursor this negative effect was suppressed. Chang et al. [38] continued the preceding study and divided the thermal stability region of y-alumina in two parts. Below a temperature of 900~ microstructural changes are dominated by sintering phenomena (sinter region), above 900~ phase transformation effects are believed to be dominant. The presence of steam accelerate pore growth and has a larger effect in the sinter region. Below 700~ the pore volume increases for the doped and decreases for the
298
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D P R O C E S S I N G
u n d o p e d samples; above this temperature, the pore volume decreases in all cases. For zirconia, Chang et al. [38] found a smaller effect of steam on the pore growth in the sinter region (below 700~ compared with 7-alumina, while the pore size of zirconia increases more strongly with temperature. PVA addition to zirconia precursor solutions resulted in an increase of the m e m b r a n e pore volume and the pore size compared to samples without PVA. Similar effects were found for titania membranes but at lower temperatures indicating a relatively small thermal stability of these titania membranes. This can be improved by doping with alumina [39,40]. Kumar [39] reports a considerably larger thermal stability for titania m e m branes in the rutile phase instead of the usual anatase form. The effect of the support on thermal stability has been reported by Kumar et al. [40,41]. Pure, non-supported titania (anatase) membranes lose their porosity completely w h e n calcined at 600~ for 8 h, where as the supported titania m e m b r a n e retained ca 30% porosity at 900~ (8 h). Unsupported titania-(50 wt%)alumina composite membranes retained a porosity of ca 40% at 700~ (8 h), supported ones retained porosity even at 900~ Finally, other examples of thermal behaviour of zirconia and titania m e m branes have been reported by Larbot et al. in a series of papers, e.g. Ref. [33]. Mesoporous membranes with high thermal stabilities to 1100~ have been reported by Chai et al. [65]. These membranes were obtained by dip coating an alumina-silica support (a 20-step process) into a mixed sol consisting of an alumina sol to which about 11 wt% Ba or La was added in the form of salts. The basic 7-alumina structure was heat treated at about 1150~ after which phase transformations start to occur. Calcination at T > 1300~ results in the formation of hexa-aluminate phases. These phases have a large resistance against sintering as has been proven by J. Kumari Kumar [66]. The pore diameter could be controlled in the range from about 4 to 8 n m (at 1150~ A full discussion of the available literature as well as theoretical considerations leading to a strategy to improve the thermal stability of porous materials is given by Z~iter [15], to which the reader is referred. 8.2 SYNTHESIS A N D PROCESSING OF SUPPORTED M I C R O P O R O U S MEMBRANES
8.2.1 Microporous Membranes Obtained by Sol-Gel Processes
8.2.1.1 Introduction and Overview of Film Formation The general scheme to obtain microporous membranes (pore diameter < 2 nm) is identical to that for mesoporous ones. Starting with a specific precursor, a wet film is formed by either film casting or slip casting, which is followed by
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
299
drying and sintering (see Fig. 8.25). Because the pore size scales with the particle size, microporous membranes require very small particles which are obtained by the polymeric route. In order to obtain reasonable fluxes and separation factors in applications (see e.g. Chapter 9 on gas permeation) the membranes should be (very) thin, preferably < 1 ~tm and "defect free". This requires high quality supports (smooth, defect-free, small pore size distribution) with small mean pore size. High quality (7-alumina) supported mesoporous films are generally used for this purpose. The drying rate for these very thin microporous layers is large and this causes probably rather dense, low porosity membranes (see below). The drying rate is, however, also controlled by the liquid content of the support system which acts as a reservoir similar to that discussed for mesoporous membranes. As discussed in Chapter 7 branching of the polymeric species, as characterised by its fractal dimension D, determines the number of contact points M a between two mass fractal objects of size rc (see Section 7.4.1, p. 238). During drying the gel network collapses and, the lower their fractal dimension, the more the particles interpenetrate and form an intertwined network. When particles come into contact there is the probability of a further reaction by condensation. This further reduces penetration. Therefore an important factor is the ratio of penetration rate (driven by evaporation/drying) and condensation rate which stiffens the structure and so increases the resistance to compaction. As discussed by Brinker and Scherer [1] the probability of forming irreversible bonds at the intersection points is given by their sticking probability which in turn depends on the condensation rate. In silicate systems the condensation r a t e - and thus the sticking probability ~ has a minimum around a pH near 2. Finally, their equations (p. 238) show that if D < 1.5 the probability of intersection ~ and thus the sticking factor at given condensation rate - - decreases as rc increases. The porosity can be controlled in two ways. The first method is based on the scaling of mass Mf and size rf of the mass fractal particles. Since density equals mass/volume, the density pf of a mass fractal object varies in three-dimensional space as: pf ~-
r~ /r 3
(8.15)
and the porosity as 1/pf
,-.rt 3-D)
(8.16)
Thus the porosity of a mass fractal decreases with its size and when complete interpenetration is avoided (which requires Df > 1.5) the porosity can be controlled by the size of the branched specimen during drying. Examples of this procedure are given by Brinker et al. [42]. This discussion reveals a dilemma: to obtain the smallest pores, interpenetration should be large and is obtained by D < 1.5 and low condensation rates. This leads however to low porosities.
300
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
To enhance the porosity, non-hydrolysable organic molecules are attached to the precursor molecules in the dip solution. During the film formation process these organic template molecules are incorporated in the film. Their removal by pyrolysis in the calcination step allows an independent control of pore size and volume through the size and volume fraction of the organic template (see Fig. 7.4 in Chapter 7). As discussed in Chapter 7, Guizard and co-workers as well as Brinker et al. [42,49] showed that microporosity can be introduced in this way. The synthesis of high quality, defect-free microporous membranes with this process has not been demonstrated so far. Recently Brinker [42] reported that complete pyrolysis of the template ligands at 500~ greatly diminished the (gas) permselective properties of the membranes probably by the formation of defects. To obtain defect-poor membranes, severe requirements should be imposed on the quality of the supports. Supported mesoporous membranes, usually y-alumina, with pore diameters in the range of 4-5 nm are used as supporting systems for microporous membranes. For a discussion of quality specifications of y-alumina membranes is referred to Section 8.1.3.2). Small pores in the support are requested to inhibit excessive penetration of the polymeric species into the support system during formation of the microporous layer. De Lange et al. [43,46] showed the activation energy of hydrogen permeation to correlate with the quality (separation factor) of silica microporous membranes (see also Chapter 9 on transport phenomena in membranes). They reported optimum results with a two layer y-alumina supporting membrane obtained in a two step dip-coating process on a polished (~-alumina macroporous support. The quality of silica microporous membranes could be improved slightly by a second dipping step. The maximum allowable temperature of silica microporous membranes is about 500~ Above this temperature cracking and sintering occurs. High quality microporous membranes are almost exclusively reported for silica or for binary silica-titania or silica-zirconia systems [42,46]. This is due to the very fast hydrolysis and condensation rates of the metal organic precursor of the metals relevant for membrane synthesis (Ti, Zr, Sn, A1). This usually results in too large particles in the precursor solution. Though many authors claim to have produced microporous materials by sol-gel methods (see e.g. Section 8.2.3), only a few have shown the synthesis of membranes of these materials and a still smaller number has characterised them with appropriate separation properties to be reasonably defect free. Therefore in the remainder of Section 8.2.1 a focus will be given to silica-based membranes.
8.2.1.2 Important Parameters in Precursor Synthesis Overview of basic elements of precursor chemistry As discussed in the preceding section, supported microporous membranes require precursor sols with weakly branched polymeric species with a fractal
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
301
dimension D < 2.0 (preferably around 1.5) and with a particle size as characterised by the gyration radius Rgof a size comparable to the size of the support pores [21,48]. As discussed in Section 8.2.3, these polymeric specimens are obtained from competitive hydrolysis and (poly)condensation of metal alkoxides M(OR)~ in solution with M = Si, A1, Ti or Zr. The degree of branching and the growth mechanism depends on the relative rate of these two processes. A detailed discussion of the chemistry of these hydrolysis-condensation reactions has been given by Brinker and Scherer [1]. A detailed discussion of the structure of polymeric species in precursor solutions to be used for membrane synthesis is given by de Lange et al. [44,52] and overviews are given by de Lange [43] and Brinker [42,49] and in Chapter 7. Important parameters are the concentration and concentration ratios of the components, the way the components are added to the final mixture, the temperature and ageing processes. Hydrolysis under strongly acidic conditions is necessary to obtain polymeric specimen. Characterisation is usually done by SAXS measurements.
Silica membranes The synthesis conditions which lead to weakly branched systems involve the use of an acid catalyst where pH < 2.2 (iso-electric point of silica) and the use of low to moderate water content (rw < 10). Hydrolysis (see reactions in Section 7.4) then takes place via a fast protonation of the alkoxide, followed by attack of water, resulting in the substitution of the alkoxy group with an hydroxyl group. Protonation becomes slower when more hydroxyls are present. The hydrolysis rate will therefore decrease with the extent of OH substitution. Acid catalysed condensation reactions proceed analogously where a protonated silanol species is attacked by water. The condensation reaction rate decreases with the number of condensed Si-O-Si groups. Condensation reactions under acid catalysed conditions are much slower than hydrolysis reactions and generally start when the hydrolysis process is almost complete. The largest differences in reaction rate constants for hydrolysis and condensation are reported for pH = 0.9 and these differences decrease if the pH is increased [56]. As a consequence a large amount of hydrolysed species is present at the moment condensation becomes significant. Further condensation reactions then take place between individual hydrolysed species (clusters) and lead to aggregated clusters. This is schematically represented for a simple case in Fig. 8.22 where dimers react with each other leading to a linear molecule. Further condensation reactions with other condensed polymers will take place preferentially at the end groups [54]. Cluster-cluster reactions are especially important during ageing. In general the hydrolysed species are not simple dimers but are weakly branched species
302
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
OR /
OR /
RO-Si-O-Si-OH \
OR
\
OR
OR
OR
/
/
\
\
+ HO-Si--O-SFOR OR
OR
,=
OR
OR
O~
OR
OR
OR
OR
OR
/ / / / RO- SP O- Si- O- SF O- Si- OR \ \ \ \
+ HOH
Fig. 8.22. Acid catalysed condensation reaction leading to linear polymers. (clusters as shown in Fig. 7.3, Section 7.2.2). The density of the clusters and aggregates and their D values critically depend on their growth mechanism and growth rate which is diffusion or reaction rate limited. A variety of aggregation growth models has been proposed and simulations based on specific conditions lead to different D values and structures (see for an overview Refs. [1,43]. Generally these models lead to relative large D values > 1.78. The tip to tip clustercluster model proposed by Jullien [55,56] can account for a smaller D value of 1.42 and so is important to account for the results of de Lange and of Brinker. In this model the clusters hardly penetrate and stick on tips (contact points). The hydrolysis and condensation rates are influenced by the size of the organic groups (R in Fig. 8.22) due to steric hindrance and to chemical effects (change of inductive effects of the R group and the metal ion, see Ref. [1]). Large groups usually cause a slower reaction rate. The solvent for the alkoxide is also important for several reasons. If the solvent is an alcohol, this can participate in the hydrolysis and condensation reactions as can be easily seen from the reaction equation given in Section 7.4. Again the reaction rate is decreased and in water-alcohol solutions an equilibrium can be obtained with a relative large amount of non-hydrolysed Si-OR groups. The hydrolysis reaction rate of transition metal alkoxide is much higher than for Si(OR)4 as discussed by, e.g., Livage et al. [57]. Their reactivity can be decreased by exchange of the R group by for a much larger one or by chelating bridge-forming agents such as acetylacetone which are difficult to hydrolyse. This very large reaction rate makes it difficult to synthesise polymer species with D < 2.00. Non-hydrolysable template ligands can be introduced into the polymeric silica sols by co-condensation of two different precursor molecules as discussed in Chapter 7.2. A recent example has been given by Brinker et al. [42,51] where mixtures of tetraethoxysilane (TEOS) and, e.g., methaacryloxypropylsilane (MPS) or TEOS and methyltriethoxysilane (MTES) are hydrolysed in ethanol, water and I M HC1 in a two-step process to obtain precursor sols for membrane synthesis. Other methods to influence the structure of the precursor sols are discussed in Section 7.6. The preparation of supported membranes has not however been reported with these sols.
8 -- FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
303
Effects of composition ratios and process parameters in silica precursors The synthesis route of silica membranes is schematically given in the upper part of Fig. 8.25. Tetraethylorthosilicate (TEOS) is not hydrolysed directly in water. To obtain a better control, the hydrolysis and condensation reaction rates were decreased by first diluting the TEOS in alcohol (ethanol) and then adding to this mixture a water-acid (HNO3) mixture dropwise under vigorous stirring. The mixture was kept for 3 h at 86~ under reflux conditions. Note that even with this procedure locally and for short times a relative large water excess exists in the reaction zone. De Lange [44] investigated systematically the relevant parameter values and found the best results with a composition given in Table 8.3. This is called the standard solution. TABLE 8.3 Composition of stmldard silica polymeric sol. The figures in parenthesis give the ratios X(reactrait)/TEOS as defined by rw, ralcohol mid ra
mol mol ratio (X/TEOS)
TEOS
H20 (rw)
C2H5OH (ralcoho~
HNO3 (1 M) (rH+)
0.094 (1)
0.6 (6.4)
0.36 (3.8)
0.008 (0.085)
Refluxing and ageing, effect of temperature The fractal dimension D and the gyration radius Rg of the polymeric specimen, as determined by SAXS, increased during refluxing from D = 1.1-1.3 after 1 h to D - 1.47-1.55 after 3 h, with Rg values of 10 nm (1 h) and 14-22 nm (3 h), respectively [43,44]. These figures did not change significantly after 3 days of ageing. After 10-14 days a transition occurs with a limiting value obtained after 14 days after which no further changes occurred. The observed transition corresponds with gelation of the sol. Limiting values are found of D = 1.8-1.9 and Rg ~ 4 nm respectively as shown by Figs. 8.23 and 8.24. Lowering the reaction temperature to 20~ caused a decrease of the D value to about 1.35 with Rg-- 2.0 nm. It can be concluded that weakly to moderated branched specimen (as indicated by D values) with particle sizes given by Rg can be obtained by careful control of processing parameters and of ageing time. Note that the particle size can be brought into the range of the pore diameter of the 7-alumina support (4 nm). Similar results are reported by Brinker et al. [42] using somewhat different synthesis parameter values.
304
8 m FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
ot 5O
0
+
0
t3~
0 o
0
0
.<~30 rr'
0 o~
o
++§
20 ~ D ~*D + +
+
(
10 0
ib
0
20'
30
Sol/gel age (Days)
4O
SO
Fig. 8.23. Evolution of the gyration radius Rgof silica sol/gel samples with standard composition as a function of aging time. From de Lange et al. [43,44].
2.2 2
%
1.8 9D
1.6
O ~
_~ 1.4 '
0
1I 0
10
- 20
30
Sol/gel age (Days)
40
~0
Fig. 8.24. Evolution of the fractal dimension D of silica sol/gel samples with standard composition as a ftu~ction of aging time. Data are obtained with two different camera lengths of the SAXS equipment in two different measurement runs. From de Lange et al. [43,44].
Effects of composition and acidity D e L a n g e [43,44] s y s t e m a t i c a l l y i n v e s t i g a t e d a n u m b e r of r e l e v a n t p a r a m e ters, s t a r t i n g w i t h T E O S s o l u t i o n s , b e f o r e a r r i v i n g at t h e o p t i m a l p r e c u r s o r solution. The results are summarised below.
8 -- FUNDAMENTALS
OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
305
With the standard solution as reference (see Table 8.3) it was found that: (1) small differences in the acid concentration in the range of rH = 0.074-0.096 (standard value is r H = 0.085) do not show significant differences in polymeric structure as characterised by the fractal dimension Df and gyration radius Rg. The r H value of 0.085 corresponds with a pH -- 0.82. It should be noted however that the measuring of a p H value in alcohol-water mixtures is not straightforward. Lowering the acid concentration to r = 0.043 (pH -- 1.12) results in lower values of Df (1.42) compared with D = 1.7 for the standard composition under the same conditions. (2) An increase of the water content to rw = 12 (rw = 6.4 for the standard) did not result in significant changes. Decrease of the r w values shows a trend towards lower fractal dimensions and cluster sizes. At rw = 1 (stoichiometric water concentration is at rw = 2) no SAXS scattering could be observed. (3) Lowering the reaction a n d / o r ageing temperature to 20~ or diluting the standard sol both resulted in a strong decrease of Df and Rg values. The effect of dilution (lowering the concentration) can be understood from the decrease of the condensation rates assuming the reaction is second order in both components. A further detailed study by Nair et al. in the same group [59] showed the importance of dilution to 0.05-0.2 M silica on the value of Df and of the parameters discussed above on porosity (see below).
Binary silica-metaloxide sols (M = Ti, Zr, AI) Sols of Ti, Zr or A1 with polymeric structure and defined fractal dimensions has not yet been reported in literature. Studies of hydrolysis-condensation behaviour of mixtures of Si and Ti alkoxides has been r e p o r t e d by de Lange [42,44,46] and Kamiyama et al. [50]. In order to control the hydrolysis rate three different routes were explored as s h o w n in Fig. 8.25. (1) Single step prehydrolysis of TEOS at low rw = 1, is followed by addition of Ti- or Zr n-butoxide in butanol. (2) Two-step hydrolysis: first TEOS is pre-hydrolysed at low rw -- 1 and mixed with Ti or Zr butoxide as in (1). In the second step an additional amount of a water-acid mixture is supplied. (3) Separate hydrolysis of both components (Si and Ti or Zr) and mixing of both sols. The best results for membrane synthesis are obtained with process (2) resulting in precursor solutions with Df = 1.3 (Si/Ti) or 1.45 (Si/Zr) and Rg ~ 1.7 (Si/Ti) and 0.8 (Si/Zr). With standardised solutions supported membranes could be made. Method (3) resulted for Si/Ti in Df >_2.04 and Rg ~ 3.6 nm. Kamiyama [50] reported that for hydrolysed solutions of Si/Ti alkoxides are spinnable w h e n Df < 1.79 and not spinnable for larger values.
306
8 -- FUNDAMENTALS OF MEMBRANE TOP-LAYERSYNTHESIS AND PROCESSING PREHYDROLYSlS OF TEOS
J~ 20 ~C or 80"C
(sH= 3 hydrJcond. ,) COMPOSITE SYNTHESIS SINGLE STEP
20"C <===ITi'lalcohol AI'Zr blnary sol single s l ~
TWO STEP HYDROLYSIS
(|
)
,.
20oc
"" .: .... :J ~.
wstm" HNO3
,
binwy 9ol two ~ep
Fig. 8.25. Synthesis scheme of polymeric silica and of binary polymeric sols by single step and two step hydrolysis. From de Lange et al. [43,44].
8.2.1.3 Illustrative Examples of Membrane Synthesis and Microstructure Development Silica and silica-titania/zirconia membranes with high quality combining high separation factors and high permeation values were first reported by Uhlhorn [12,58] and were further developed and analysed by de Lange et al. [43,46,47,60]. Further optimisation has been undertaken by Verwey and coworkers [59] in the same group. According to de Lange et al., standard precursor solutions were dip coated on top of high quality 7-alumina (pore diameter 4 nm) support layers which are supported by high quality (x-alumina plate supports (pore diameter about 0.2 Bm). The quality of the obtained, ultra-thin silica (titania) layers depends critically on the details of the synthesis route and support system [43,46,60]. It was shown that 7-alumina supports obtained in a two or three-step coating process (2-3 layers with total thickness 7-10 ~tm) in combination with a silica layer obtained in a single or two-step coating process give optimum results as determined from gas permeation and separation measurements [60] (see also Chapter 9 on gas transport). The layer thickness of the silica top layer is about 100 nm and was determined from sputter profiles with X-ray photo spectroscopy (XPS) as well as scanning auger microscopy (SAM). The sputter profiles showed a region with constant Si concentration followed by a zone with gradually decreasing Si and increasing A1 (from the support) concentration. The cross-over point of the two lines was considered to be the site of the original interface between the support and silica film. The width of this last zone is
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
307
Fig. 8.26.FE-SEMmicrograph of a cross section of a microporous silica membrane supported by a ~/-alumhla top layer with low roughness. From de Lange et al. [43,44]. larger than the sampling resolution of the method and indicates that the silica layer consists of "plugs" within the support pores (about 50 nm) and a part on top of it [46] with a thickness of >__50 nm. High resolution FE-SEM micrographs confirmed this picture as shown in Fig. 8.26. The average roughness after applying the silica-layer is not significantly changed. Atomic force microscopy results in a typical picture as shown in Fig. 8.27 [46]. This suggests that the silica film does not completely cover the mesopores of the support (diameter about 4 nm) but follows the roughness profile of the supporting layer with average roughness of 40 nm. Work by Elferink et al. [59b] has shown that further optimisation of the microstructure can be obtained by lowering the temperature of the hydrolysiscondensation reaction. This results in better control of porosity and pore size distribution. As yet, there are no methods available to determine directly the porosi~ and pore size of supported microporous membranes, supported by porous supports. To obtain information on the relation between precursor and layer characteristics, non-supported layers with a thickness of 50-100 gm were investigated with N 2 adsorption-desorption techniques by de Lange et al. [47] using similar standard solutions from which supported membrane systems are also made. A general discussion has been given by Brinker [1] and in Chapter 3. According to de Lange typical porosities of about 37% are obtained with a bi-modal pore size distribution having a maximum at an effective pore diameter
308
8 -- FUNDAMENTALS
.... ,,.,
0 ~
'-'~
nm
: ::::::::'..::7:~:!:::::.<.~.-..
i:: .... ::i:::::-:::::==========================
OF MEMBRANE
TOP-LAYER
..:.:: :::::::::!:::5:! ....
. ==================================== .........
"'~k"'--~~.u"
~
SYNTHESIS
AND
PROCESSING
::::::::::::::::::::::::::::::::::::::::
....I.. ~ 5 0 0 ~../
115 Fig. 8.27. Three dimensional picture of a ultra thin microporous silica membrane on a 7-alumh~a support (pore diameter 4 nm), obtahled with atomic force microscopy. From de Lange et al. [46].
of about 0.5 nm and a very weak second maximum at about 0.75 nm. Later research by Nair and Elferink in the same group [59] showed that the porosity could be purposely changed by variations of the precursor synthesis (rw, ra, rH) and reaction temperature without significant changes in pore diameter. Larger fractal dimension (Dr) values result in larger porosities. The drying rate proved to be important as indicated by differences in porosity between drying under ambient conditions and drying in a climate chamber at 40% and RH = 60% for 3 h. The last procedure resulted in a small porosity [43]. Because the drying rates for supported and unsupported membranes are different, porosities measured on supported membranes are also different from supported ones which, at the moment, are unknown. The comparison of the pore size measured on non-supported membranes by N2 absorption-desorption with that on supported silica membranes measured with gas permeation and separation with molecules of greatly different sizes indicates that the average pore diameter of supported silica layers is slightly smaller (0.40-0.45 nm). See Chapter 9 on gas transport. Finally, it has been shown by de Lange et al. [43,61] that ageing of the calcined membranes changes their microstructure. Long-term ageing under ambient conditions (i.e. 16 months at 20~ or 80 days at 40~ and 60% RH results in a few percent loss of porosity of non-supported layers. Ageing at 350~ for 10 days in an air atmosphere containing 1.7% H20 vapour caused a porosity decrease from 27 to 25%. Thermal ageing at high temperature of supported silica membranes causes densification of the silica layer, and the higher the temperature, the greater the
8 -- FUNDAMENTALS
OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
309
risk of cracking. Thermal treatment at 600~ instead of 400~ resulted in microcracking of the layer as indicated by H 2 permeation characteristics [43,46, 61]. This is discussed further in Chapter 9. These results indicate that for use at high temperature the thermal and chemical characteristics of the silica membranes should be improved. Brinker and coworkers [49] reported the synthesis of microporous silica membranes on commercial (membralox) 7-alumina supports with pore diameters of 4.0 nm. Ageing of the silica sols was shown to be effective to form discrete membrane layers with an estimated thickness of 35 nm on top of the support and to inhibit pore penetration of the silica. Sols with gyration radii Rg < rs (radius of support pores) penetrate the support to a depth of about 3 ~tm, which is the thickness of the 7-alumina support layer. Minimization of the condensation rate during film formation was considered to decrease the width of the pore size distribution without changing the average pore radius, which was estimated to be 0.35 < rp < 0.5 nm. The porosity of films deposited on dense supports was about 10% as calculated from refractive index measurements. Some gas permeation properties of these membranes are reported in Refs. [42,48]. Molecular sieving properties have not been observed [48] contrary to results obtained by the Lange. This might be due to a combination of the larger roughness of the commercial supports, the somewhat smaller hydrolysis ratio H20/Si (r ~- 5 instead of 6.4 used by de Lange), and the smaller layer thickness which all lead to an increased risk of defects in the layer. The ageing and drying conditions were also different. Kitao and Asaeda and coworkers reported the synthesis of microporous silica membrane systems in a series of investigations [e.g. Refs. 63,64]. Sol-gel coatings were prepared in a multiple dip-coating process (15 steps) using 4 types of silica sols on a smoothed (x-alumina support with a pore diameter of 1 ~tm. Most details are givenin [62]. The coatings were applied on a hot substrate (160-180~ resulting in very rapid drying and after each step the coating was fired at 570~ The first three groups of sols (A-C) consisted of particulate (colloidal) sols with decreasing particle size (sol A: 31 nm, sol C" 4.8 nm) and stronger dilution. With each sol the dip-coating process was repeated 3-4 times. As a matter of fact in this part of the process intermediate, mesoporous silica layers with decreasing pore size were applied on the macroporous alumina support. The final sol (D) was a diluted polymeric one, obtained from hydrolysis of TEOS under strongly acidic condition, with a concentration of 0.1-0.5 wt% TEOS. After dip-coating with this sol, the membrane was fired at 350~ for one day. Layer thicknesses were not determined accurately but were said to be a few ~tm, with the final layer < 1 ~tm. This very laborious method results in membranes that probably have microporous characteristics as is inferred from their gas [63,64] and vapour [62] permeation and separation properties which are discussed in Chapter 9 on gas transport.
310
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P- L A Y E R SYNTHESIS A N D P R O C E S S I N G
Template approach Membranes produced following a template approach are reported by Raman and Brinker [42,51] using precursor sols discussed in the preceding section and produced by copolymerisation of methyltriethoxysilane (MTES) and TEOS using a two-step acid catalysed process. Membranes were deposited by dip-coating on commercial Membralox supports (pore diameter 4 nm) and were dried and calcined in air at temperatures in the range 150-550~ for about 1 h. During this heat treatment the methyl ligands pyrolyse, creating a continuous network of micropores. The film thickness varies between 37.5 and 110 nm from one region to another, probably due to non-uniform thickness of the y-alumina support. EDS measurements indicate that silica penetrates the support to a depth of about 2 ~tm. The presence of non-hydrolysable ligands seem to promote compaction of the structure during drying resulting in smaller porosity a n d / o r pore size. Gas permeance measurements (CO2, CH4) show a large decrease with increasing temperature of calcination in the region of 400~ to 550~ This indicates densification of the membranes at high temperature. Permselectivity values of C O 2 / C H 4 w e r e calculated from single gas permeation measurements at 25~ and are well above Knudsen values and range up to 12. This value could be further enhanced by surface derivatization by treatment with monomeric TEOS to about 72 with CO2 fluxes of about 2.10 -4 c m 3 / c m 2 s-cm Hg (= 6.7x10 -8 mol m -2 s -1 Pa -1) compared with fluxes an order of magnitude larger before derivatization. These results can be explained rather well without microporous properties being present due to capillary condensation type of phenomena at 25~ (see Chapter 9 on transport properties). However the relative large permselectivities of He/N2 and He/SF6 couples indicate the presence of microporous properties. Complete removal of the template ligands by pyrolysis at 500~ greatly diminishes the separation factors, probably due to defect formation. 8.2.2 Microporous Membranes Obtained by CVD 8.2.2.1 CVD methods Chemical vapour deposition has been exploited to produce dense as well as microporous films on porous supports by several groups of investigators. Gavalas and coworkers reported the synthesis of SiO2 membranes in mesoporous Vycor tubes using different precursors like Sill4 or SIC14. In their more recent work the better results were obtained by deposition in the so-called opposing reactant geometry (ORG) [68] instead of by the one-sided geometry (OSG) [67]. In the OSG mode a mixture of SiC14 a n d H 2 0 in N 2 as carrier gas was applied to one side of the porous tube and reacted there at 600-800~ This resulted in a dense SiO 2 film. In the OSG mode one of the reactants in the carrier gas was applied
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP- LAYER SYNTHESIS A N D P R O C E S S I N G
311
to the outer side of the tube, the other reactant to the inner side. Reactants diffused into the pores of the tube and reacted there. The structure of the resulting film has been described in Ref. [69]. The SiO2 is deposited in a region with a depth of about 10 ~tm adjacent to the surface. The porosity of the deposit is highest in a region of about 0.5 ~tm adjacent to the surface and declines to zero at about 10 ~tm. The thin zone with highest density contains about 10% by volume trapped voids and is responsible for the permselectivity (for H 2 / N 2 this is above 500). Ha et al. [70] reported deposition of SiO2 within the pores of Vycor tubes by pyrolysis of TEOS with and without the presence of oxygen in the temperature region of 300-700~ Stable, H 2 selective membranes were prepared in the ORG mode only and with oxygen present in the gas mixture. Films produced from TEOS/O2 mixtures were said to be less dense than films made from hydrolysis of SiC14. C.L. Lin et al. [71] reported deposition of silica layers (plugs) with a thickness of about 1.5 ~tm within the pores of commercial, mesoporous y-alumina films (pore diameter 4 nm, thickness 1-3 ~tm) on (z-alumina supports (US filter). The deposits were obtained by reaction of TEOS-oxygen (10-20%) mixtures in He as carrier gas applied in the OSG mode to the mesoporous layer. No further details (e.g., temperature or pressure) were given. Depending on these unknown conditions, dense as well as microporous silica membranes with pores down to estimated values of 0.4-0.6 nm were obtained. These membranes have interesting combinations of permselectivity and flux values for several gas combinations (see Chapter 9 on gas transport properties).
Repairing defects and pore narrowing Films deposited by CVD can grow in several ways depending on the conditions. Usually it is not precisely known which is the dominant mechanism. One of the possibilities is film growth starting on the pore walls followed by a gradual increase of the thickness of the deposits (homogeneous deposition mechanism) resulting in pore narrowing. This process has been studied by Y.S. Lin and Burggraaf [72,73] and Cao et al. [76] in a systematic study of CVD deposition of Z r O 2 - Y 2 0 3 membranes on porous (7-cz)-alumina supports [74-76]. The results of Y.S. Lin and Burggraaf indicate that effective pore narrowing of a membrane by CVD processes requires a rather uniform pore size distribution of this membrane and conditions which promote strongly a homogeneous deposition mechanism. The results of Cao et al. [76] show that this goal can be attained within the mesopore range. A similar conclusion might be drawn from the results of C.L. Lin [71] and Kitao et al. [64]. Kitao treated a microporous silica membrane produced as described in Section 8.2.1 with a mixture of Sill 4 and oxygen in the OSG mode at 400~ in 1-10 short subsequent treatments. The permeability of N 2 decreased strongly while that of He was hardly affected. This was interpreted as closing or repair-
312
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
ing defects (i.e. pinholes) in the silica layer while the micropores through which He permeates were hardly affected. After a prolonged treatment the activation energy of He and H 2 increased which was interpreted as a decrease of the size of the micropores. It should be noted however that de Lange et al. [43,60] showed that defect reparation in silica films also increased the apparent activation energy of H 2 permeation. It therefore remains to be proven whether or not CVD can be used to control the pore size of microporous membranes to be used at high temperature, if the use of derivatization (reactive adsorption of large organic molecules at the pore wall) is excluded. An alternative might be the control of the pore opening as reported by Niwa et al. [77] by reactive adsorption of a suitable reactant at the surface of the membrane.
8.2.2.2 Other Methods and Microporous Membrane Systems Several other interesting methods to obtain microporous membranes are reported in literature. Except for zeolite membranes (see Section 8.2.3) they are probably of less practical or commercial importance and therefore will only be briefly summarised. Mesoporous glass ("Vycor type") can be produced by a combined heat-treatment and leaching procedure as summarised by Burggraaf and Keizer [78] in an extended overview of synthesis methods. Modification of this process leads to microporous hollow fibres with interesting properties (see Chapter 9). Details of this modification process are not known to this author. The oldest reported microporous membranes are based on carbon and are obtained by controlled pyrolysis of suitable polymeric precursors. Koresh and Soffer were the first to report properties of these membranes in a series of papers starting in 1980 (see refs. in Ref. [78]. Recently Linkov et al. [79] improved this method and arrived at mesoporous asymmetric hollow-fibre carbon membranes which could be transformed to microporous systems by coating the carbon membrane by e.g. vapour deposition polymerisation of polyimide forming precursors. Rao and Sircar [80] developed a new technique. A macroporous graphite sheet was coated with a suitable polymer (latex) which was pyrolysed subsequently. This process was repeated 4-5 times and resulted in a carbon layer thickness of 2.5 ~tm with an average pore diameter between 0.5 and 0.6 nm. The membrane has interesting properties (see Chapter 9). 8.2.3 Zeolite Membranes
8.2.3.1 Overview and Introduction to Zeolite Chemistry Zeolite membranes form the most recent branch of the inorganic membrane field for which characterised and properly described real microporous zeolite
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
313
membranes were reported [81-88]. The "older" literature concerns either dispersions of zeolite crystals on a support or very defective membrane systems which do not exhibit dominantly micropore transport (and so are not "real" microporous membranes) but sometimes have interesting properties especially for membrane reactors. Overviews of the existing literature on zeolite membranes are given by Geus [81] and Vroon [82] and briefly by Matsukata [85] and Burggraaf [89]. In this older literature the pioneering work of Suzuki [90] and of Lachmarm [91] yielding N a A / C a A and X or Y or mordenite zeolites should be mentioned. Zeolites are porous crystalline materials with a complex crystallographic structure. The characteristics of the void (pore) network is completely determined by the crystallographic space group and the composition. It is appropriate here to treat structural characteristics of zeolites. A brief summary has been given in the Chapter 9, Section 9.4.1 in which a typical structure is illustrated by Fig. 9.19 for MFI type (ZSM5, silicalite) zeolites, being almost exclusively used for defect-free membrane synthesis. For a more detailed discussion the reader is referred to the works of Breck [92] and of van Bekkum et al. [93].
Important synthesis parameters The synthesis mechanism of the very complex zeolite structure from precursor solutions is not known at present. It will certainly be a multiple-step process involving nucleation and growth steps. An overview has been given by Jansen [94]. It emerges that the result of the synthesis is very sensitive for the kind of reagents (reagent source) used, even if these have nominally similar composition and structure and on impurity levels. Obviously tiny differences in reactivity are important, and the type of organic template molecules used in the synthesis is also important (see Table 4 in Ref. [94]). In the case of MFI the synthesis mixture consists of a solution of silica, alumina, sodium hydroxide, water and the template molecule and the final zeolite composition a n d / o r structure depends critically on the precise compositions of the synthesis solution and the temperature of (hydrothermal) synthesis. Different reaction sequences take place in different temperature regions as indicated in Table 8.4. Reaction mixtures are usually prepared at lower temperature (< 60~ and are used subsequently at higher temperature. Drastic chemical and physical changes then take place. It is a point of debate in the zeolite literature whether nucleation of the zeolite starts at residual gel particles in the solution or can be formed directly on the surface of supporting systems (see Table 8.5). As shown by Jansen [95] on some support materials zeolite crystals are formed in a thin gel layer present on the support, on other supports this gel layer is not observed and a different crystal morphology, or even structure, can be formed.
314
8 - FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
TABLE 8.4 The subsequent events present in the course of the zeolite synthesis reaction (after Jansen [94]) Temperature (~
Subsequent events
Low (< 60)
Reactant solutions reactant mixture-gel formation
Low to high (<60 to <200)
Gel rearrangement Dissolution of gel Dissociation of silicate
High (<200)
Pre-nucleation phase Nucleation Crystallization
Low (<60)
Isolation
TABLE 8.5 Four cases of crystal growth environment and a schematic representation of nucleation and crystallisation (after Janssen [94]) Crystal growth environment
(a) Clear solution
Nucleation (,) Crystallization (--4)
,k
(b) Dispersed low density gel
(c) Separated high density gel
(d) Solid phase
I-X !1
8 - - F U N D A M E N T A L S OF M E M B R A N E TOP-LAYER SYNTHESIS A N D PROCESSING
315
The pH of the liquid phase is very important because OH-molecules act as a mineralising agent and its value is generally between 8 and 12. At high pH values monomeric Si ions are most abundant in the liquid at low Si concentration; at high concentrations these are cyclic tetramers. At lower pH neutral Si species are the dominant ones. The rate of the temperature increase of the reaction mixture in the autoclave from _<60~ to higher temperature affects the structure of the liquid and thus the subsequent course of the reaction. The main event occurring in the synthesis mixture at the reaction temperature is the formation of zeolite particles from amorphous material and involves [94]: (1) further reorganisation of the low temperature synthesis mixture; (2) homogeneous or heterogeneous primary and secondary (seed promoted) nucleation; (3) precipitation as a form of crystallisation. Four cases of nucleation and crystallisation are schematically presented in Table 8.5. As discussed above, no general conclusions can be reached about which situation prevails under certain conditions. Nucleation and crystallisation kinetics generally follow S-shaped crystallisation curves as shown for zeolite A in Figs. 8.28 and 8.29. This means that a rather long incubation time or nucleation period precedes the crystallisation (compare Figs. 8.28b for zeolite A and 8.29b for ZSM5). The general trend in the kinetics of zeolite (ZSM5) synthesis can be summarised as follows: (1) higher pH (alkalinity) leads to higher crystallisation rates and shorter incubation times; (2) higher temperature causes higher reaction rates, shorter incubation times and larger crystal sizes. The morphology of, e.g., ZSM5 crystals can be varied by a variety of parameters m e.g. Si concentration, template molecule, cations present, presence of growth inhibitors or promotors (the support of a membrane) ~ and is difficult to predict in actual cases. The most frequently found crystal form of ZSM5 is an elongated prismatic form which can be changed to a flat, cubic crystal form by increasing the Si concentration or by replacing tetrapropylammonium ions (TPA § as template molecules by divalent bi-quaternary ammonium ions [94].
What is required for zeolite membrane synthesis? For membrane synthesis the most preferred situation should be the formation of thin single crystalline, oriented layers on a porous support. This has not yet been realised. The next approach is the oriented deposition of monocrystals on this support. Some success has been obtained in literature by very laborious techniques, the main problem being the gaps between the crystals which must be filled [81a].
316
8 - - FUNDAMENTALS OF MEMBRANE TOP-LAYER SYNTHESIS AND PROCESSING
Z eolile A ['J{,I
100
-
H 2 0 / N a 2 0 - 20
H20/Na20-30 /'"
,
I I I I s ! !
Crystollilohon rote
H20/NazO=40 /'~
10.O
/
/ s
60 e/
l ! I ,,/ I
I
! 2
o
8o
, /
50"
a
4.0 2.0
-
i 3
'''l 4
. 5
O o
-/ -6
:~
4
6
e
[OH]
Fig. 8.28. The influence of alkalinity on (a) zeolite and (b) ZSM5crystallisation. From Jansen [94].
I
@
;8.~'C
9 2 0 0 " C j
leo!,te A [ % ]
80~ o f | ! ! !
60~
9
20
!
|
'// !
50
r
70~
r
|
9
~ ---
9 e
170"C
750"C
I
|
-
i .
.
.
'i
O o
.
b
a Timo {hc~rs)
_ _J__ ~o
.l 2o t(h)
.,J 3o
Fig. 8.29. The influence of temperature on the c r y s t ~ a t i o n rate of (a) zeolite A and (b) Z.SM5. From
Jansen [94]. As will be shown in the next section polycrystalline intergrown layers have been successfully deposited on a variety of porous supports. Some groups prefer large crystals to approach as much as possible the single-crystal situation [81], other groups [82-84] try to realise ultra-fine grained zeolite (ZSM5) layers to make use of properties of grain boundaries and to obtain very thin layers. Examples of both situations will be given in the next section. Defect-poor, supported membranes have been reported only for MFI-type crystals (ZSM, silicalite) because it has been proven that other membrane
8 - - F U N D A M E N T A L S OF M E M B R A N E T O P - L A Y E R SYNTHESIS A N D P R O C E S S I N G
317
systems are much more difficult to produce in layers free from defects. As will be clear from the above discussion, zeolite membrane synthesis is still in its infancy and predictions can hardly be made to select the most suitable synthesis conditions or zeolite systems.
8.2.3.2 Illustrative Examples of Zeolite Membrane Synthesis and Processing Geus et al. [96] reported the in situ synthesis of ZSM5 (MFI-type) deposits on several types of porous supports (clay, 0~-alumina, zirconia, metakaolin). A continuous film of randomly oriented zeolite crystals could only be obtained on porous clay supports. The synthesis mixture was chosen so as to obtain large cube-shaped crystals of silicalite (Si-rich MFI). A mixture consisting of silica (Aerosol 200), sodium hydroxide (Baker) and tetrapropylammoniumbromide (TPABr) in water in the molar ratio 100 SIO2:160 Na20:150 (TPA)20:16666 H20 was prepared and aged under ambient conditions for 1-6 h. The (formal) pH value was about 14. The films were prepared by positioning of the porous substrates on the bottom of the teflon-lined autoclave and hydrothermally treated under autogenous pressure at 180~ for 1-5 days in the above solution. After washing and drying the membranes were calcined at 400~ (heating rate 1~ min -1) for 20 h to remove the template molecules. The result was a MFI film with a thickness of 50-80 ~tm, strongly bound to the clay support. Crystal sizes were reported to be large, although no figures were given. In a later study, Geus reported that cracks were formed in the film during this calcination procedure despite the low heating rate. This is explained by the occurrence of tensile stresses which form in the layer by a combination-of shrinkage of the zeolite lattice due to template removal and thermal mismatch of the expansion coefficients of zeolite and support during cooling. These stresses set an upper limit on the application temperature of the supported zeolite membrane. The obtained gas permeation and separation properties of the clay supported MFI films were, however, rather low. This was ascribed to the low porosity of the clay support. In a subsequent study Geus et al. [81b] changed to high porosity supports of stainless steel consisting of small disks of sintered porous steel covered with a thin (50-150 ~tm) top layer of metal wool assembled in a non-isothermal procedure in a module system (see Ref. [81b] for details). The effective pore size of the metal wool was estimated to be about 10 ~tm. In a series of experiments the optimal composition of the zeolite precursor solution was determined to be given by the molar ratios 100 SIO2:230 TPA:75 OH-:14000 H20 (aged for 5 h) at a formal pH of 13.5. The main difference with the solutions used in the earlier experiments [96] is the much smaller sodium content (50 m g / k g in the present experiment compared with 1700 m g / k g in the earlier ones). This was done to ensure maximal incorporation of TPA (four per unit cell) in the zeolite frame
318
8 ~ F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
work. The support was put on the bottom of a teflon-lined autoclave and treated at 180~ for 45 h. After washing and drying this resulted in a continuous layer of randomly oriented, intergrown MFI-1 crystals with a thickness of about 50 ~tm. After calcination at 400~ the obtained zeolite membrane had good gas permeation/separation properties (see Chapter 9). Crystal size and shape were not given but were expected to be large. Only low levels of metal concentrations in the precursor solutions were observed after the hydrothermal treatments and it was concluded that the applied stainless steel can be considered inert under the required synthesis conditions. It should be noted however that this does not exclude relative large local metal concentrations in the zeolite at the zeolite-steel interface. The type of support plays an important role in the crystallisation process as observed by the comparison of layers formed on teflon or zirconia which exhibited different size/shape, formed with different kinetics and did not form completely continuous films (on teflon). Geus et al. concluded that to obtain total, continuous coverage of porous supports a relatively large layer thickness should be necessary. The minimal layer thickness is expected to be correlated to the maximal pore size of the porous support and on the smoothness of the support top face. Gas permeation values can be increased by decreasing the layer thickness. Therefore Vroon et al. [82-84,97,98] selected in situ synthesis conditions to obtain very small MFI crystals from which thin layers could be grown. Mesoporous ~-alumina and titania supporting layers were severely attacked by the precursor solution (with high alkalinity) and could not be used. Macroporous c~-alumina supports (pore diameter 0.16 ~tm, porosity -- 46%) could however be successfully applied. The standard zeolite synthesis solution chosen [82,97,98] was characterised by the molar ratios 100 SIO2:15 (TPA)20:5.3 Na20:1420 H20. The mixture was obtained by mixing sodium hydroxide (Merck, > 99%) and TPA hydroxide in water to which silica (Baker, 99.75%) was added. The silica was dissolved at 100~ in about 300 s after which the solution was cooled down to room temperature in 1.5 h and aged (at 25~ for an additional 1.5 h. The pH of the solution was > 12.5. To control crystal size and layer thickness, hydrothermal treatments were applied in a stainless steel autoclave under autogenous pressure at temperatures of 40-180~ for 10-200 h with the support on the bottom of the autoclave. The layers were very thoroughly washed (seven times at 60-80~ to remove all traces of sodium. In the case of multiple layer processing, the wet supported layer was again positioned at the bottom of the autoclave and the complete procedure was repeated. After the last hydrothermal treatment the multiplestep membrane was dried and the template molecules were removed by heating at 550~ for 16 h (heating rate 10~ cooling rate 20~
8 -- FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
319
TABLE 8.6 Influence of synthesis temperature and time on the growth characteristics of a MFI zeolite layer on top of a 0~-A1203 support [82,98] Synthesis temperature (K)
Synthesis time (h)
Top layer
Remarks
Thickness ( ~ t m )
Particle size (nm)
333 393 453
240 16 4
0.6-1.5 0.4-2 2-3
40-100 100-300 200-500
cracked crack free crack free
Crack-free continuous thin (< 3.0 ~tm) MFI layers were obtained on top of the support when the synthesis temperature was above 95~ This was correlated with the fact that at this temperature the average particle size, having a spherical shape, in the zeolite layer in the stationary stage (see below) is equal to or larger than the average pore size of the support (160 nm). Some typical particle sizes and layer thicknesses as a function of conditions are given in Table 6. Crack-free layers can be obtained with a layer thickness of 1-2 ~tm and particle sizes as low as 100 ~tm at a minimum temperature of 95~ The growth of the MFI layer can be divided into three periods: (1) the incubation period (no observable layer), (2) the layer growth period, and (3) the stationary period (no further increase of thickness with time). Incubation and layer growth period increase with decreasing synthesis temperature a n d a r e about 12.5 h (incubation period) and about 50 h (end of layer growth period) at 95~ compared with less than I h and 10-15 h respectively at 120~ MFI crystals were also formed to some extent inside the support pores as well as a discontinuous layer of MFI crystals was formed at the bottom of the support even if the support was placed with its bottom on that of the autoclave. The shape of the particles in the discontinuous bottom layer was different (coffinlike) from that of the top layer (spherical). From experiments with different positions (horizontal versus vertical at different heights in the autoclave) it was concluded that sedimentation on the support of nuclei formed in the solution followed by further growth on the support surface plays a role as well as preferred nucleation directly on the support surface. From gas permeation experiments (see Chapter 9) it was found that before template removal even very thin layers (1-2 ~tm) could be produced in a gastight state. After template removal a good membrane quality could be obtained with somewhat thicker layers. To obtain high quality, defect-free MFI membranes, two subsequent hydrothermal treatments resulting in a total layer thickness of about 3-4 ~tm give best results with excellent separation properties.
320
8 ~ F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
,.I]
La ~A 8-, Fig. 8.30.Schematicrepresentation of the microstructure of a fine grained zeolite MFI layer on top of a (~-alumina support. A = Grain boundary with fine slits between grain surfaces; B = "closed" intergrown interfaces;Lo= layer thicknessmeasured by SEM;Ls= effectivethicknessof separation layer. From Vroon et al. [82,98]. Layer thicknesses above 4 ~tm usually result in cracking during template removal. The structure and chemical composition of the layers was studied by XRD and TEM-EDX in combination with ion beam thinning of the layers. The XRD results indicate the presence of randomly distributed silicalite-1. Gas permeation experiments with xylenes indicate some catalytic activity which point to acid sites, not present in pure silicalite (A1/Si = 0) but which are present when A1 is built into the silicalite lattice. This indicates the occurrence of some reaction between the precursor solution and the 0c-alumina support during hydrothermal synthesis. The transport properties of the thin, fine crystalline layers prepared by Vroon et al. [82-84,98] are quite different from that of the thick, coarse crystalline layers prepared by Geus et al. [81]. A model of the structure of the fine crystalline layers is given in Fig. 8.30. Based on gas permeation properties, it is hypothesised that grain boundaries between the small crystals play a beneficial role on the transport properties of the layer. The intrinsic gas permeation values of the small crystals seem to be smaller than that of the large ones reported by Geus. Finally, Vroon et al. [82,97] reported the synthesis of continuous porous films of ZSM5 on top of ?-alumina supported membranes (pore diameter 4 nm) by slip-casting with a zeolite crystal suspension. The porous zeolite layers (thickness 1-2.5 ~tm) consist of densely packed zeolite crystals with a diameter of 70-80 nm and with micropores in the zeolite and mesopores (diameter 8-24 nm) between the zeolite particles. This zeolite layer can be used as a support for further processing, e.g., pore filling of the mesopores or deposition of catalysts. First experiments by Vroon et al. to fill the mesopores by in situ crystallisation of MFI in the pores did not result in gas-tight membranes Xiang and Ma [86] also recognised the problems of treating mesoporous ?-alumina membranes with highly alkaline zeolite precursor solutions. They
8 m F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
321
reported the synthesis of ZSM5 membranes with rather large A1 content on 0~-a1203 supports (pore diameter 0.2 ~tm) from aqueous-based solutions (sols) using in situ crystallisation in a similar way to that described above. The molar ratios in the precursor solution were 100 SIO2:10 Na20:10 (TPA)20:1600 H20 while hydrothermal treatment was conducted at 130~ for 48 h, followed by an additional treatment at 200~ for 16 h. After washing, drying and calcination at 600~ a zeolite membrane with good gas separation properties resulted. No further characteristics of the layer were given. y-Alumina s u p p o r t s were first treated with a mixture of alkoxides AI(OC4H9) 3 and Si(OC2H5) 4 with SiO2/A1203 = 30 at 90~ for 5 h to protect the y-alumina from being damaged when the synthesis was carried out. Then the sample was "coated" with a mixture of the above alkoxides and NaOC2H5 with ethanol (SIO2/A1203 = 30.14, Na20/A1203 = 3.0) and treated subsequently at 90~ for 6 h and 200~ for 12 h. No gas permeation properties or other characteristics were given. Jia/Noble and coworkers [87,88] reported the successhA synthesis of silicalite membranes on y-alumina composite supports using an interesting modification of the in situ crystallisation method. The support consisted of a short {x-alumina tube coated on the inside with a 5 ~tm thick y-alumina film with an average pore diameter of 5 nm, commercially available from US Filter. The precursor solution was put into the support tube after plugging both ends with teflon and the filled tube was then placed in a teflon-lined autoclave. Hydrothermal treatment was carried out at 180~ for 12 h. After removal from the autoclave and washing the formed zeolite layer with water, the procedure was repeated with the tube inverted from its previous orientation to obtain a uniform coating. As reported by Vroon et al. [82,84,98], Jia/Noble [88] also concluded that at least two synthesis steps are necessary to obtain defect-free membranes. Calcination was carried out at 455-480~ for 8 h after very slow heating (5~ The thickness of the silicalite layer was about 10 ~tm [87] and adhered well without peeling off. Crystals observable in the surface of the layer were elongated with an estimated length of 5-20 ~tm and a thickness of 3-4 ~tm). SEM pictures show no evidence of damage of the y-alumina layer. Good gas separation properties were reported (see Chapter 9). The precursor solution from which the silicalite was grown had a composition of 10 g silica, 2.1 g TPABr, 0.95 g NaOH and 125 g H20 and was used after one day ageing. This composition is equivalent to molar ratios of SiO2/Na20 = 100/7.4 and H 2 0 / N a 2 0 = 100/7.4. Comparison with precursor solutions used by Vroon et al. indicated a somewhat smaller pH value used by Jia et al., probably below 12. Nevertheless, the relative stability of the y-alumina layer under the conditions used remains a remarkable phenomenon. Finally Smith/Keizer et al. [99] reported the synthesis of a continuous silicalite layer (thickness 1 ~tm) on top of hollow-fibre carbon supports using a
322
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
precursor solution and conditions described by Vroon et al. (i.e. hydrothermal treatment at 120~ for 3 days). ZSM5 and mordenite layers could be formed only after activation of the carbon surface by absorption of tetraethoxysilane on the carbon surface followed by heat treatment in an oxidising atmosphere. This resulted in active surface sites (Si atoms). Hydrothermal treatment at 185~ for 3 days then resulted in the growth of ZSM5 with a thickness of 11 ~tm on the carbon support. This again shows the importance of the nature of the support surface for the in situ growth of zeolite layers.
8.3 CONCLUSIONS AND EVALUATION
Mesoporous, asymmetric multilayer membranes with graded pore diameters as small as 4 nm can be routinely produced on macroporous alumina or carbon substrates of different shapes (e.g., plates, tubes, multichannel monoliths). Film-coating or slip-casting processes, using colloidal suspensions with organic additions, are used to make membranes with a thickness down to about 4 ~tm. After some dipping steps the lyogel film is dried and heat treated at high temperature to stabilise the pore structure. Mesoporous membranes of 7-alumina, titania, zirconia or MFI-zeolite as well as their composites can be produced in this way. To obtain defect-free separation films, the use of support systems with a good quality (i.e., low roughness, reasonable pore size distribution without too large pores, reproducible wettability) is necessary and multiple dipping procedures are usually required. To obtain membranes on large surface areas and / or on complex support shapes, further optimisation of the process is necessary, e.g., the uniformity of the film thickness should be improved. Further study of the very first step of the layer formation process might lead to improved properties of the layer-support interface or to membrane "plugs" formed in the pore entrance instead of to films on top of the support. The drying process is important and becomes critical with small mesopore systems due to large tensile stresses which build up in the membrane layers. Techniques for the measurement of these stresses and theoretical and model descriptions for stress and crack development are given together with illustrative examples. Stress levels and membrane cracking can be decreased applying low controlled drying rates and using organic additions. Thermal stability differs widely between different membrane materials and can be improved by appropriate doping of the structure. Up to 1000~ good pore stabilities can be obtained for pore diameters larger than 6-10 nm and appropriately chosen materials. The chemical nature of the internal pore surface can be drastically changed by grafting organic functional groups onto this surface.
8 - - F U N D A M E N T A L S OF MEMBRANE TOP-LAYER SYNTHESIS A N D PROCESSING
323
Microporous membrane (pore diameter smaller than 2 nm) synthesis is still in its infancy. Microporous membrane layers of amorphous silica and silica-titania composites, zeolite, and carbon are reported on supports of (~ or 7) alumina (for silica and zeolite) or on stainless steel (for zeolite) or on carbon (for carbon or zeolite). Scaling up of the different processes used to obtain larger membrane surface areas have to be demonstrated. For amorphous silica layers the synthesis process is similar to that used for mesoporous membranes, except that now solutions of ultra small, polymeric silica particles, with fractal dimensions smaller than 1.5-2.0, are used as precursors. These are produced with a set of specific synthesis conditions (e.g. high acidity to control the relative rates of the hydrolysis and condensation reactions). With very smooth, high quality 7-alumina supporting layers on a o~-alumina support ultra-thin (100 nm) silica membranes with a pore diameter of 0.45-0.5 nm could be obtained. High quality membranes with low defect levels are obtained in a two-step coating process. The thermal stability of these systems is limited to about 500~ and relatively low water partial pressures. Research to improve this is necessary. The use of non-hydrolysable template molecules in the silica precursor solution allows modification of the porosity and pore size. These types of processes are interesting, especially in combination with derivatization, but need much more research to delineate their potential. Chemical Vapour Deposition (CVD) of microporous silicafilms with a thickness of about 1.5 ~tm onto mesoporous glass or y-alumina substrates are obtained by deposition from TEOS-oxygen mixtures at 300-700~ Pore sizes are estimated to be 0.4--0.6 nm or are virtually absent. CVD techniques may also be useful for repairing residual defects and for pore narrowing. Zeolite membranes of a good quality could be produced by two-step in situ growth of the zeolite from a precursor solution under hydrothermal conditions on y-~ alumina or stainless steel supports. Reliable results are reported only for MFI-type zeolites. The obtained layers consist of randomly oriented, intergrown MFI crystals with shapes and sizes depending on synthesis conditions and support. The thickness is in the range of 4-100 ~tm depending on conditions. The growth kinetics, crystal size and morphology, layer microstructure and properties are very sensitive to even small changes in the raw materials, precursor composition, process conditions and substrate material. Much more research is needed to establish the interesting prospects and to broaden the field to zeolite types other than MFI. Carbon microporous membrane layers with a thickness of 2.5 ~tm could be obtained by pyrolysis of selected polymeric precursor films (obtained in a multistep coating process) on carbon supports.
324
8 m FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
REFERENCES
1. C.J. Brinker and G.W. Scherer, Sol-Gel Science. Academic Press, New York, 1990. 2. A.F.M. Leenaars, Preparation, structure and separation characteristics of ceramic membranes, PhD Thesis, 1984, University of Twente, Enschede, The Netherlands. 3. A.F.M. Leenaars and A.J. Burggraaf, The preparation and characterisation of alumina membranes with ultrafine pores. Part 2. The formation of supported membranes. J. Colloid Interface Sci., 105 (1985) 27-40 4. A.F.M. Leenaars, K. Keizer and A.J. Burggraaf, The preparation and characterisation of alumina membranes with ultrafine pores. Part 1. Microstructural investigations on non supported membranes. J. Mater. Sci., 19 (1984) 1077-1088. 5. D. Freilich and G.B. Fanney, J. Colloid Interface Sci., 64 (1987) 362. 6. L. Cot, C. Guizard and A. Larbot, Novel ceramic material for liquid separation process. Industr. Ceramics, 8 (1988) 143-148. 7. T.M. Shaw, Movement of a drying front in a porous material, in: C.J. Brinker, D.E. Clark and D.R. Uhlrich (Eds.), Better Ceramics Through Chemistry II. Materials Research Society Symposia, Vol. 73. Materials Research Society, Pittsburgh, PA, 1986, pp. 215-223. 8. L.E. Scrivin, Physics and applications of dip-coating and spin coating, in: C.J. Brinker, D.E. Clark and D.R. Ulrich (Eds.), Better Ceramics Through Chemistry III. Materials Research Society, Pittsburgh, PA, 1988, pp. 717-729. 9. R.R. Bhave, Inorganic Membranes. Van Nostrand Reinhold, 1991. 10. L.D. London and B.G. Levich, Acta Physiochim. URSS, 17 (1942) 42-54 11. F.M. Tiller and Chum Dar Tsai, Theory of filtration of ceramics. I: Slipcasting. J. Am. Ceram. Soc., 69 (1986) 882-887. 12. (a) R.J.R. Uhlhorn, Ceramic membranes for gas separation. PhD Thesis 1990, University of Twente, Enschede, The Netherlands. (b) R.J.R. Uhlhorn, M.B.J. Huis in 't Veld, K. Keizer and A.J. Burggraaf, Synthesis of ceramic membranes. Part I. Synthesis of non-supported and supported ~,-alumina membranes without defects. J. Mater. Sci., 27 (1992) 527-537. 13. K.N.P. Kumar, Nanostructured ceramic membranes. PhD Thesis, 1993, University of Twente, Enschede, The Netherlands. 14. V.T. Zaspalis, W. v. Praag, K. Keizer, J.R.H. Ross and A.J. Burggraaf, Synthesis and characterisation of alumina, titania and binary membranes. J. Mater. Sci., 27 (1992) 1023-1035. 15. J.M. Z~iter-Hofman, Chemical and thermal stability of (modified) mesoporous ceramic membranes. PhD Thesis, University of Twente, Enschede, the Netherlands. 16. J. Zarzycki, Monolithic xero- and aerogels for gel-glass processes, in: L.L. Hench and D.R. Uhlrich (Eds.), Ultra Structure Processing of Ceramics, Glasses and Composites. Wiley, New York, 1984, pp. 27-43. 17. L.L. Hench, Use of drying control chemical additives (DCCAs) in controlling sol-gel processing, in: L.L. Hench and P. Uhlrich (Eds.), Science of Ceramic Chemical Processing. Wiley, New York, 1986, pp. 52-65, 18. A.J. Burggraaf, Key points in understanding and development of ceramic membranes, Yi Hue Ma (Ed.), Proceedings of the third International Conference on Inorganic Membranes July 10-14 1994. Worcester Polytechnic Institute, Worcester, MA, pp. 1-16/ 19. P.G. Simpkins, D.W. Johnson Jr. and D.A. Fleming, Drying behavior of colloidal silica gels. J. Am. Ceram. Soc., 72 (1989) 1816. 20. M. Fortes and M. Okos, in: A.S. Mujumdar (Ed.), Advances in Drying, Vol. 1. Hemisphere,
8 -- FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISANDPROCESSING
325
New York, 1980, pp 119-154. 21. J. Zarzycki, Processing of gel glasses, in D.R. Uhlmann and N.J. Keidl (Eds.), Glass: Science and Technology. Academic Press, Orlando, FL, 1984, pp 209-249 22. R.K. Dwivedi, J. Mater. Sci. Lett., 5 (1986) 373-376. 23. W.D. Kingery, H.K. Bowen and D.R. Uhlmann, Introduction to Ceramics. Wiley, New York, 1976. 24. R.W. Hofmann, in: B. Schwartz and N. Schwartz (Eds.), Measurement Techniques for Thin Films. The Electrochemical Society, New York, 1967, p. 312. 25. D.S. Campbell, in: L. Maisel and R. Glang (Eds.), Handbook of Thin Films. McGraw Hall, New York, 1970, p 123. 26. J.H.L. Voncken, C. Lijzenga, K.P. Kumar, K. Keizer, A.J. Burggraaf and B.C. Bonekamp, New method for the measurement of stress in thin drying gel layers, produced during the formation of ceramic membranes. J. Mater. Sci., 27 (1992) 472-478. 27. K.P. Kumar, K. Keizer and A.J. Burggraaf, Stress development during the formation of ceramic membranes, in: Proceedings of the 2nd European Conference on Advanced Materials and Processes, Cambridge, July 1991. TransTech Publications, Ziirich, Switzerland, 1992, pp 204-210. 28. J.H.L. Voncken and A.J. Burggraaf, Internal report no. CT90/443/128, 1990, Faculty of Chemical Technology, University of Twente, 7500 AE Enschde, The Netherlands. 29. K.N.P. Kumar, V.T. Zaspalis, K. Keizer and A.J. Burggraaf, Drying process in the formation of sol-gel derived ceramic membrane. ]. Non-Cryst. Solids, 147/148 (1992) 375-381. 30. P.J.A.M. Kerkhof, Technical University of Eindhoven, Eindhoven, The Netherlands. Private communication, 1993. 31. H. Verweij, K.N.P. Kumar, K. Keizer and A.J. Burggraaf, Formation and formation mechanisms of alumina and titania membranes, in: Yi Hua Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes (ICIM3) July 10-14,1993. Worcester Polytechnic Institue, Worcester, MA 016109, USA, pp. 27-35. 32. J.M. Hofman-Ziiter, Chemical and thermal stability of (modified) mesoporous ceramic membranes. PhD Thesis 1995, University of Twente, Enschede, The Netherlands. 33. A. Larbot, J.P. Fabre, C. Guizrd and L. Cot, New inorganic ultrafiltration membranes: titania and zirconia membranes. J. Am. Ceram. Soc., 72 (1989) 257. 34. H.M. van Veen, R.A. Terpstra, J.P.B.M. Tol and H.J. Veringa, Three layer ceramic alumina membranes for high temperature gas separation applications, in: L. Cot (Ed.),
35.
36.
37. 38. 39.
Proceedings of the First International Conference on Inorganic Membranes, 3-6 July 1989, Montpellier, France, pp. 329-335. H.M. van Veen, J.P.M. Top, R.A. Terpstra, A.J.G. Engel, H.J. Veringa, The calcination and thermal stability of ceramic A1203 membranes for gas separation, in: G. de With, R.A. Terpstra and R. Metselaar (Eds.),Euroceramics L Vol. 3. Elsevier, Amsterdam 1989, pp. 585-589. Y.S. Lin, K.J. de Vries and A.J. Burggraaf, Thermal stability and its improvement of the alumina membrane toplayer prepared by sol-gel methods. J. Mater. Sci., 26 (1991) 715-720. Y.S. Lin and A.J. Burggraaf, Preparation and characterisation of high temperature thermally stable alumina composite membrane. J. Am. Cer. Soc., 74 (1991) 219. C.H. Chang, R. Gopalan and Y.S. Lin, A comparative study on thermal and hydrothermal stability of alumina, titania and zirconia membranes. J. Membr. Sci., 91 (1994) 27-45. K.N.P. Kumar, K. Keizer, A.J. Burggraaf, T. Okubo and H. Nagamoto, Synthesis and
326
8 ~ FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISAND PROCESSING textural properties of unsupported and supported rutile (TiO2) membranes. J. Mater.
Chem., 3 (1993) 923-929. 40.
K.N.P. Kumar, K. Keizer and A.J. Burggraaf, Textural stability of titania-alumina composite membranes. J. Mater. Chem., 3 (1993) 917-922. 41. K.N.P. Kumar, K. Keizer, A.J. Burggraaf, T. Okubo and H. Nagamoto, Textural evolution and phase transformations in titania membranes. Part 2. Supported membranes. J. Mater. Chem., 3 (1993) 1151-1159. 42. C.J. Brinker, N.K. Raman, L. Delatre and S.S. Prakash, Sol--gel strategies for amorphous inorganic membranes exhibiting molecular sieving characteristics, in: Yi Hua Ma (Ed.),
Proceedings of the Third International Conference on Inorganic Membranes (ICIM3) July 10-14, 1993. Worcester Polytechnic Institute Worcester, MA, pp. 63-73. 43. R.S.A. de Lange, Microporous sol-gel derived ceramic membranes for gas separation, PhD Thesis 1993, University of Twente, Enschede, The Netherlands. 44. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Polymeric silica based sols for membrane modification applications: sol-gel synthesis and characterisation with SAXS. J. Non-Cryst. Solids, 191 (1995) 1-16. 46. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterisation of supported microporous ceramic membranes by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57-75. 47. R.S.A. de Lange, K. Keizer and A.J. Burggraaf, Characterisation of microporous non supported membrane toplayers using physisorption techniques. J. Porous Mater., 1 (1995) 139-153. 48. C.J. Brinker, T.L. Ward, R. Sehgal, N.K. Raman, S.L. Hietale, D.M. Smith, D.W. Hua and T.J. Headly, Ultra microporous silica based supported inorganic membranes. J. Membr. Sci., 77 (1993) 165-179. 49. C.J. Brinker, R. Sehgal, N. Raman, P.R. Schunk and T.J. Headly, Polymer approach to supported silica membranes. J. Sol-Gel Sci. Technol., 2 (1994) 469-476. 50. T. Kamiyama, M. Makumi and K. Suzuki, A SAXS study of the gelation process of silicon and titanium. J. Non-Cryst. Solids, 150 (1992) 157-162. 51. N.K. Raman and C.J. Brinker, Organic template approach to molecular sieving membranes. J. Membr. Sci., 105 (1995) 273-281. 52. R.S.A. de Lange, K.N.P. Kumar, J.H.A. Hekkink, G.M.H. van der Velde, K. Keizer, A.J. Burggraaf, W.H. Dokter, M.F. van Garderen and T.P.M. Beelen, Microporous SiO2 and SiO2/MOx (M = Ti, Zr, A1) for ceramic membrane applications: a microstructural study of the sol and the consolidated state. J. Sol-Gel Sci. Technol., 2 (1994) 489-495. 53. J.C. Ro and J.J. Chung, Sol-gel kinetics of tetraethylorthosilicate (TEOS) in acid catalyst. J. Non-Cryst. Solids, 130 (1989) 26-32. 54. C.W. Turner and K.J. Franklin, Studies of the hydrolysis and condensation of tetraorthosilicate by multinuclear NMR spectroscopy. J. Non-Cryst. Solids, 91 (1987) 402-415. 55. R. Jullien, A new model of cluster aggregation. J. Phys. A: Math. Gen., 19 (1986) 2129-2136. 56. M. Kallala, R. Jullien and B. Carbane, J. Phys II (Paris), 2 (1992) 7. 57. J. Livage, M. Henry and C. Sanchez, Sol-gel chemistry of transition metal alkoxides. Prog. Solid State Chem., 18 (1988) 259-341. 58. R.J.R.Uhlhom, K. Keizer and A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part II. Synthesis and separation properties of microporous membranes. J. Membr. Sci., 66 (1992) 271-282. 59. (a) B.N. Nair, W.J. Elferink, K. Keizer and H. Verweij, Sol-gel synthesis and characterisation of microporous silica. 1. SAXS study on the growth of polymeric structures.
8 -- FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISANDPROCESSING
327
J. Coll. Interface Sci., 178 (1996) 565-570. (b) W.J. Elferink, B.N. Nair, R.M. de Vos, K. Keizer and H. Verweij, Sol-gel synthesis and characterisation of microporous silica membranes. Part II. Tailor making porosity. J. Coll. Interface Sci., 180 (1996) 127-134. 60. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, J. Microporous Mater., 4 (1995) 169-168. 61. R.S.A. de Lange, K. Keizer and A.J. Burggraaf, Ageing and stability of microporous sol-gel modified ceramic membranes. Ind. Eng. Chem. Res., 34 (1995) 3838-3847. 62. M. Asaeda, A. Yamamichi, M. Satoh and M. Kamamura, Preparation of porous silica membranes for separation of propylene/propane gaseous mixtures in: Yi Hua Ma (Ed.),
Proceedings of the Third International Conference on Inorganic Membranes, July 10-14, 1994. Worcester Polytechnic Institute, Worcester, MA, pp. 315-325. 63. S. Kitao, H. Kameda and M. Asaeda, Gas separation by thin porous silica membranes of ultra fine pores at high temperature. Membrane, 15 (1990) 222 64. S. Kitao and M. Asaeda, Gas separation performance of thin porous silica membranes prepared by sol--gel and CVD methods, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.),
Proceedings of the Second International Conference on Inorganic Membranes, 1-4 July 1991, Montpellier. Transtech Publ., Z/~rich, 1991, pp. 267-273. 65. M. Chai, M. Machida, K. Eguchi and H. Arai, Preparation and characterisation of sol-gel derived microporous membranes with high thermal stability. ]. Membr. Sci., 96 (1994) 205-212. 66. Jalaya Kumari Kumar, Textural stability of high temperature catalyst supports, PhD Thesis 1995, University of Twente, Enschede, The Netherlands. 67. M. Tsapatsis and G.R. Gavalas, A kinetic model of membrane formation by CVD of SiO2 and A1203. AICHE ]., 38 (1992) 847 68. M. Tsapatsis, S. Kim, S.W. Nair and G.R. Gavalas, Synthesis of H2 permselective SiO2, TiO2, A1203 and B203 membranes from the chloride precursors. Ind. Eng. Chem. Res., 30 (1991) 2152. 69. M. Tsapatsis and G. Gavalas, Structure and ageing characteristics of H2 permselective SiO2/Vycor membranes. J. Membr. Sci., 87 (1994) 281-296. 70. H.Y. Ha, S.W. Nam, S.A. Hong and W.K. Lee, Chemical vapor deposition of hydrogenpermselective silica films on porous glass supports from tetraorthoethyl silicate. J. Membr. Sci., 85 (1993) 279-290. 71. C.L. Lin, D.L. Flowers and P.K.T. Liu, Characterisation of ceramic membranes. II: Modified commercial membranes with pore size under 40 ]~. ]. Membr. Sci., 92 (1994) 45-58. 72. Y.S.Lin and A.J. Burggraaf, Experimental studies of pore size change of porous ceramic membranes after modification. J. Membr. Sci., 79 (1993) 65-82. 73. Y.S. Lin, A theoretical analysis on pore size change of porous ceramic membranes after modification. ]. Membr. Sci., 79 (1993) 55. 74. Y.S. Lin and A.J. Burggraaf, Modelling analysis of CVD processes in porous media for ceramic composite preparation. Chem. Eng. Sci., 46 (1991) 3061-3080. 75. Y.S.Lin, K.J. de Vries, H.W. Brinkman and A.J. Burggraaf, Oxygen semipermeable solid oxid membrane composites prepared by electrochemical vapor deposition. J. Membr. Sci., 66 (1992) 211-226. 76. G.Z. Cao, H.W. Brinkman, J. Meijerink, K.J. de Vries and A.J. Burggraaf, Pore narrowing and formation of ultra thin yttria-stabilised zirconia layers in ceramic membranes by chemical vapor deposition. J. Am. Cer. Soc., 76 (1993) 2201-2208.
328
8 ~ FUNDAMENTALSOFMEMBRANETOP-LAYERSYNTHESISAND PROCESSING
77. M. Niwa and Y. Murakami, CVD Zeolites with controlled pore opening size. J. Phys. Chem. Solids, 50 (1989) 487. 78. A.J. Burggraaf and K. Keizer, Synthesis of Inorganic Membranes" in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Application. Van Nostrand Reinhold, New York, 1991, pp. 10-63. 79. V.M. Linkov, R.D. Sanderson and E.P. Jacobs, Preparation of hollow fibre and composite hollow fibre carbon membranes, in: Yi Hua Ma (Ed.), Proceedings of the 3rd International Conference on Inorganic Membranes 10-14 July 1994, Worcester. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 471-480. 80. M.B. Rao and S. Sircar, Nanoporous carbon membranes for separation of gas mixtures by selective surface flow. J. Membr. Sci., 85 (1993) 253-265. 81. (a) E.R. Geus, Preparation and characterisation of composite inorganic zeolite membranes with molecular sieve properties. PhD Thesis 1993, Technical University of Delft, Delft, The Netherlands. (b) E.R. Geus, H. van Bekkum, W.J.W. Bakker and J.A. Moulijn, High temperature stainless steel supported zeolite (MFI) membranes: preparation, module construction and permeation experiments. Microporous Mater., 1 (1993) 131-147. 82. Z.A.E.P. Vroon, Synthesis and transport studies of thin ceramic supported zeolite MFI membranes. PhD Thesis 1995, University of Twente, Enschede, The Netherlands. 83. Z.A.E.P Vroon, K. Keizer, H. Verweij and A.J. Burggraaf, Transport properties of a ceramic thin zeolite membrane, in: Yi Hua Ma (Ed.), Proceedings of the 3rd International Conference on Inorganic Membranes 10-14 July 1994, Worcester. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp 503-508. 84. Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Transport of alkanes through ceramic thin zeolite membranes. J. Membr. Sci., 113 (1996) 293-300. 85. M. Matsukata, N. Nishiyma and K. Keyama, Preparation of a thin zeolite membrane, in: Y. Wehkamp, H.G. Karge, H. Pfeiffer and W. H61derlioh (Eds.), Studies in Surface Science and Catalysis, Vol. 84. Elsevier, Amsterdam, 1994, pp. 1183-1290. 86. S. Xiang and Y.H. Ma, Formation and characterisation of zeolite membranes from sols, in: Yi Hua Ma (Ed.), Proceedings of the 3rd International Conference on Inorganic Membranes 10-14 July 1994, Worcester. Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA, pp. 95-105. 87. Meng Dong Jia, B. Chen, R.D. Noble and J.D. Falconer, Ceramic zeolite composite membranes and their application for separation of vapor/gas mixtures. J. Membr. Sci., 90 (1994) 1-10. 88. C. Bai, M.D. Jia, J.L. Falconer and R.D. Noble, Preparation and separation of silicalite composite membranes. J. Membr. Sci., 105 (1995) 79-87. 89. A.J. Burggraaf, K. Keizer, R.S.A. de Lange and Z.A.E.P. Vroon, Ceramic membranes for separation and reactions, in: R. van Ballmoos (Ed.), Proceedings of the 9th International Zeolite Conference, Montreal, 1992. Butterworth-Heinnenam-Reed, 1993, pp. 47-70. 90. H. Suzuki, Composite membrane having a single layer of an ultra thin film of cage shaped zeolite and process for production thereof. US patent 5.069.794. 91. J.M. Lachman, Method of crystallisation of a zeolite on the surface of a monolithic ceramic substrate. U.S. patent 4.800.187, 1989. 92. D.W. Breck, Zeolite Molecular Sieves: Structure, Chemistry and Use. Wiley, New York, 1974. 93. H. van Bekkum, E.M. Flarigan and J.C. Jansen (Eds.), Introduction to Zeolite Science and Practice. Studies in Surface Science and Catalysis, Vol. 58. Elsevier, Amsterdam, 1991.
8 -- FUNDAMENTALSOF MEMBRANETOP-LAYERSYNTHESISAND PROCESSING
329
94. J.C. Jansen, Preparation of molecular sieves, in: H. van Bekkum, E.M. Flarigan and J.C. Jansen (Eds.), Introduction to Zeolite Science and Practice. Studies in Surface Science and Catalysis, Vol. 58. Elsevier, Amsterdam, 1991, pp. 77-136. 95. J.C.Jansen, Zeolite crystal growth and the structure on an atomic scale. PhD Thesis 1992, Technical University, Delft, The Netherlands. 96. E.R. Geus, M.J. den Exter and H. van Bekkum, Synthesis and characteristics of zeolite (MFI) membranes on porous ceramic supports. J. Chem. Soc. Faraday Trans., 88 (1992) 3101-3109. 97. H. Deckman, A.J. Jacobson, J.A. McHenry, K. Keizer, A.J. Burggraaf, Z.A.E.P. Vroon, L.R. Czernetski, F.W. Lai, A.J. Bons, W.J. Mortier, J.P. Veringa and E.W. Corcoran, Molecular sieve layers and process for their manufacture. US Patent Application WO94/25152, priority date 23/04/93. 98. Z.A.E.P. Vroon, K. Keizer, H. Verweij and A.J. Burggraaf, Preparation and characterisation of thin zeolite MFI mebranes; influence of the microstructure on the transport properties. Microporous Materials, in prep. 99. S.P.J. Smith, V.M. Linkov, R.D. Sanderson, L.F. Petrik, C.T. O'Connor and K. Keizer, Preparation of hollow fibre composite carbon zeolite membranes. Microporous Materials, 4 (1995) 385-390.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 9
Transport and separation properties of membranes with g a s e s a n d vapours A.J. Burggraaf Department of Chemical Technology, Laboratory of Inorganic Materials Science, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
9.1 INTRODUCTION
9.1.1 Chapter Outline Transport phenomena in porous solids have been the subject of many studies [1-6,10]. Quantitative solutions are obtained however only in a number of limiting cases of generally formulated problems or in relatively simple cases. Such a case is, e.g., the permeation of a single gas in a membrane system with a relatively simple pore architecture and under conditions when a single mechanism is predominantly operating. Transport of mixtures is more complicated, especially in membrane systems with a more complex architecture and operated with large pressure gradients. In such cases quantitative solutions for permeation and separation efficiency (selectivity) are not available in a generally applicable form. Specific solutions have to be obtained by approximations and by combining solutions for limiting cases. The description in this chapter takes account of this situation. First a number of important points will be summarised including a brief discussion of definitions and terminology. In subsequent sections a brief overview will be given of the most important theoretical aspects (equations) of single gas permeation and of accepted ways to combine several, simultaneously
332
9 ~ TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
operating mechanisms in simple membrane architectures. This is followed by a brief description of those transport properties of mixtures which can be applied in membrane systems. Next, permeation and separation in real, but simple, porous membrane systems will be discussed in more detail with a focus on operational applicability. Some more complex systems (multilayered, hollow fibre) will be briefly treated. Finally, a discussion will be given of the validity of important approximations made in preceding sections, of important problems (multi-component mixtures) and opportunities, and of some interesting models (e.g., molecular sieving). In all sections macro-, meso-, and microporous (molecular sieving) systems will be treated separately. The focus will be on the most promising systems to obtain high selectivity (separation factors) in combination with reasonable permeation values.
9.1.2 Overview of Important Points For single gases a number of transport mechanisms exists (Sections 9.2.3.19.2.3.3, 9.4). Depending on the pore diameter distribution and/or the temperature-pressure combination one of these mechanisms might be dominant. In many cases some of them act simultaneously and addition rules must be formulated and each contribution has to be "weighted" according to its own driving force. This is generally not the pressure gradient, but the gradient of the thermodynamic potential. As a consequence a thermodynamic correction factor has to be applied in diffusion or permeation equations expressed in terms of pressure or concentration. Even then appropriate descriptions cannot always be obtained (see, e.g., Section 9.4.) In gas mixtures the permeation of components (and thus the selectivity) is only identical with that of single gases under special conditions (high temperature and low pressure). This difference is of importance in the transition region between molecular diffusion (Poiseuille flow) and Knudsen diffusion and in that of Knudsen to configurational diffusion. In multicomponent gas mixtures general descriptions make use of Stefan-Maxwell equations and e.g. the extended Dusty-Gas Model. For binary gases these more complicated models converge to Fickian type of equations and relatively easy-to-obtain solutions for permeation (and thus for ideal) separation factors. In systems consisting of a macro-and/or mesoporous support and a mesoor microporous separation (top) layer, the permeation is a system property and the driving force for transport is distributed over the system components. In studying the permeation and separation properties of the top layer, corrections must be made on the permeation of the total system to find that of the top layer, unless it is shown that the flow resistance of the support is negligible compared to that of the top layer. Even when the permeation of the support is much larger
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
333
and its flow resistance therefore much smaller ~ than that of the top layer, there can be a considerable effect on the effective separation factor of the total system. This last one is usually different from the ideal separation factor of the top layer. In all cases the value of the pressure on the downstream (permeate) side of the membrane is important and should be as low as possible (back-diffusion, concentration polarisation). Relations between the effective and the ideal separation factors can be obtained in a number of cases. Almost all physical models use simple pore geometries. Practical pore systems are, however, very complicated and contain parameters which are difficult to measure or which have a wide distribution of their characteristic parameters. The applicability of a rigorous treatment and of very refined models and physical expressions is therefore doubtful. The treatment in this chapter will make use mainly of phenomenological equations which allow description of data, data reduction and some extrapolation and which rely on experimentally determined parameter values. Gas kinetic theory and expressions based on the microscopic (atomic) level will be used only to estimate some parameter values and to predict trends. For practical applications a combination of high selectivity and high permeation is required. As will be shown below, these two requirements are more or less contradictory and so an optimal compromise has to be sought. In this chapter a certain focus will be given to mechanisms with a large potential for high separation factors and at least reasonable permeation values. This leads to microporous systems or capillary condensation type of phenomena. Complete membrane systems can be operated in a variety of modes with e.g. co- or counter flow of feed (high pressure side) and permeate (low pressure side) streams and with membrane modules coupled in different ways. Permeation and separation in these complex engineering systems will not be treated in this chapter. Heat and mass transfer limitations on the gas-membrane surfaces or interfaces can be important with high fluxes a n d / o r strongly adsorbing gases as well as in membrane reactors. These effects will not be treated explicitly but are introduced in experimental results, e.g., by variation of sweep rates of permeated gases.
9.2 GAS T R A N S P O R T IN SIMPLE M E M B R A N E STRUCTURES
9.2.1 Important Concepts Transport data of membranes can be expressed in terms of flux (mol/m 2 s) or as flux normalised per unit of pressure (mol/m 2 s Pa). Following the IUPAC convention this last parameter is called permeation (note: in the literature the better term 'permeance' is frequently used). Using 'permeation' is meaningful
334
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
however only if there is a linear relation between flux and pressure. Despite the fact that this relation in many cases does not hold, transport data in the literature are expressed as permeation. To facilitate comparison of data the permeation can be normalised per unit of thickness and is then called permeability (tool m / m 2 s Pa). This should be done only if the thickness of the separation layer is known. In many cases only an unknown part of this layer is really active and use of the parameter permeability gives rise to large values compared with the real intrinsic ones. Therefore, in case of doubt, flux values should always be given together with the (partial) pressure of the relevant components at the high pressure (feed) and low pressure (permeate) sides of the membrane, as well as the apparent membrane thickness. It is convenient to distinguish between permeation measurements in which the flux is measured under a known (and constant) pressure gradient and those in which the flux of a component i is driven by a concentration difference between the membrane faces under a constant and equal total pressure at both sides (Wicke--Callenbach [3]). Either of these two main methods may be performed under steady state or under transient conditions. Whether or not component fluxes and diffusivities measured with both methods give similar or different values depends on the conditions and on the type of the dominant diffusion mechanism. An overview of the transport mechanisms in porous membranes is given in Table 9.1. TABLE 9.1 T r a n s p o r t r e g i m e s in p o r o u s m e m b r a n e s Transport type
Pore diameter
Selectivity
Viscous flow
> 20 n m
-
Molecular diffusion
> 10 n m
-
Knudsen diffusion
2-100 n m
1/
Surface d i f f u s i o n
+
Capillary condensation
++
M i c r o p o r e (config.) d i f f u s i o n
< 1.5 n m
++
Viscous (Poiseuille) flow and molecular diffusion are non-selective. Nevertheless they play an important role in the macroporous substrate(s) supporting the separation layer and can seriously affect the total flow resistance of the membrane system. Mesoporous separation layers or supports are frequently in the transient-regime between Knudsen diffusion (flow) and molecular diffusion, with large effects on the separation factor (selectivity).
9
-
-
TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
335
Configurational diffusion in microporous (molecular sieve) membranes will be treated separately. Here the driving force must be described in terms of a chemical potential gradient, which is coupled to partial pressure via adsorption isotherms. In cases where several mechanisms operate simultaneously, the problem of additivity arises and in real membrane systems simplifying assumptions have to made. 9.2.2 Pore Characteristics and Membrane Architecture
Porous materials have a very complex structure and morphology and many studies have been devoted to describing and characterising them [1-3,8]. Roucquerol et al. [8] in their IUPAC report give useful advice for terminology, definitions and characterisation strategies. Parameters which influence transport properties are porosity, pore size distribution, pore shape, interconnectivity and orientation. Indirectly particle size distribution and shape are important in the way they affect the uniformity of the pore size distribution, the pore shape and the roughness of the internal surface area. A schematic picture of different types of pores is given in Fig. 9.1 and of main types of pore shapes in Fig. 9.2. In single crystal zeolites the pore characteristics are an intrinsic property of the crystalline lattice [3] but in zeolite membranes other pore types also occur. As can be seen from Fig. 9.1, isolated pores and dead ends do not contribute to the permeation under steady conditions. With adsorbing gases, dead end pores can contribute however in transient measurements [1,2,3]. Dead ends do also contribute to the porosity as measured by adsorption techniques but do not contribute to the effective porosity in permeation. Pore shapes are channel-like or slit-shaped. Pore constrictions are important for flow resistance, especially when capillary condensation and surface diffusion phenomena occur in systems with a relatively large internal surface area.
'
ex2,
Fig. 9.1. Schematic picture of pore types in a porous solid, a: Isolated pore; b,f: dead end pores; c,d: tortuous a n d / o r rough pores (d), with constrictions (c); e: conical pore.
336
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
I
B
I
C
I
-
I
Fig. 9.2. Schematic representation of main types of pore structures and membrane architectures: (A) straight cylindrical pores; (B) straight asymmetric pores; (C) tortuous pore system.
A very important concept is the interconnectivity and the related tortuosity (~), as illustrated in pore d of Fig. 9.1. This parameter is used in almost all equations and will be discussed below in some detail. Burggraaf and Keizer [9] distinguish between different main types of pore and membrane structures as shown in Fig. 9.2. These different structures are related to the way they are fabricated. There are straight and parallel pores running from one side of the membrane to the other side with a constant pore diameter or conical shaped pores (Fig. 9.2A and B respectively). The tortuosity has in this case a value of about unity. In the case of conical pores as shown in the figure the membrane is asymmetric and combines a 10w flow resistance (large pores across a considerable fraction of the membrane thickness) with a relatively large selectivity (small pores on the top side of the membrane). This structure is relatively simple and systems designed in this way are useful for model experiments. Systems used in practice have a spongy structure (porous glass or carbon) or have the structure common in ceramic membranes. The latter have an interconnected, tortuous and randomly oriented pore network with constrictions and dead ends (Fig. 9.1) and are formed by packing of particles. The pore structure of zeolite membranes is formed by arrays of intergrown zeolite particles or zeolite particle packings with interparticle pores filled with another material. The intracrystalline pores are a part of the crystallographic structure and are the ones which should be responsible for the selectivity. The architecture of the ceramic membrane system is that of a multi-layer asymmetric system (Figs. 9.2C and 9.3). The separation activity is concentrated mainly in the top layer, the other parts form the supporting systems with
9 m TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
337
3
2
I
Fig. 9.3. Architecture of an asymmetric composite membrane. (1) Porous support (1-15 ~tm pores). (2) hltermediate layer(s); pore diameter dp = 100-1500 nm. (3) Mesoporous separation layer; dp = 3-100 nm. (4) Modification of 3 to microporous separation layer; dp = 0.5-2 nm.
relatively large pores to minimise the transport resistance. Hollow-fibre membranes do not need 'supports' and are thin single wall membranes. Several types o f defect, such as pinholes or cracks, can exist in the morphologic structure which reduce the selectivity even when the effect on permeation is limited.
9.2.3 Single Gas Permeation in Macroporous and Mesoporous Systems The properties of gas flow in porous media depend on the ratio of the number of molecule-molecule collisions to that of the molecule-wall collisions. The Knudsen number Kn is a characteristic parameter defining different regions of this ratio. Its value is defined by Kn = ~,/dp with ;~ being the average free path length of the gas molecules and dp the characteristic pore diameter (sometimes the hydraulic pore radius is taken). The magnitude of Kn separates three main flow regimes of gaseous diffusion (see also Table 9.1): (a) Viscous flow: Kn <<1, ~ << dp (b) Knudsen diffusion (flow): Kn >>1, ~ >> dp (c) Transition flow: Kn = 1, ~ = dp When the pore walls strongly absorb gas molecules, surface diffusion and / or capillary condensation accompanied by (surface) flow occurs. Usually this is the case with gases which condense rather easily at moderate temperaturepressure conditions (in any case being below their critical point) and we are dealing with 'vapour' flow. Configurational diffusion is a separate class and occurs when the pore diameter is a factor of 1-5 larger than the molecular diameter.
9.2.3.1 Viscous Flow When the number of intermolecular collisions is strongly dominant (Kn << 1), forced flow under a pressure or concentration gradient in a capillary can be
338
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
described by Darcy's law [14]. In capillaries with small diameter the flow is laminar and, if the gas velocity near the pore walls equals zero (no slip), the molar flux can be described by a Hagen-Poiseuille type law: r 2 P dP
Jv = -
81"1 RT dz
(9.1)
In real porous media, Eq. (9.1) m u s t be modified to account for the n u m b e r of capillaries per unit volume (porosity) and the complexities of the structure (tortuosity 1;). This leads to: 2 P dP
s
Jv = -
"r 81"1RT dz
(9.2)
In the steady state the fluxes into and out of any cross section of a pore are equal. Therefore p(dP/dz) is constant and the integration of Eq. (9.2) over the thickness L of the porous m e d i u m gives the permeation:
Jv E 1-2 =~ ~ Pm AP 8r1"r RTL
Fv = - ~
(9.3)
with the m e a n pressure P m = 0.5(P1 + P2) and P1 and P2 the pressure at the inlet and outlet respectively. Equation (9.3) shows that the permeation is proportional to" - the (hydraulic) radius squared (r2); - the mean pressure.
9.2.3.2 Knudsen Diffusion and the Transition Region Kn udsen diffusion W h e n the n u m b e r of molecule to wall collisions is strongly d o m i n a n t (Kn >> 1) the flow of a single gas in a long capillary under the action of a concentration pressure gradient can be described by the Knudsen equation [15]: 2Jk = - ~
v
dc r d---z
2_
(9.4a)
1 dP
Jk = - ~ v r R--T-d---7
(9.4b)
with the thermal m e a n velocity of the gas molecules v given by: v = (8RT) ~
(9.5)
In real porous media geometrical effects play an important role, similar to that discussed for viscous flow (Section 9.2.3.1.), and Eq. (9.4) m u s t be modified by a term I~/'l~Kn.
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
339
Furthermore in the derivation of (9.4) it is assumed that after collisions with the wall the molecules are specularly reflected. This is usually not the case. The walls have a certain roughness and this causes diffuse reflections. This effect is accounted for by a factor 1/OKn; for smooth walls 0 equals unity, otherwise it is larger. Taking these effects into account and inserting (9.5) into (9.4b) the expression for the Knudsen flow in a porous membrane is obtained: 2 ~r ( 8 ) ~ Jk" = - 3 "~0k ~RTM dz
(9.6)
Several authors introduce a shape factor [~ in (9.6) which accounts for the pore size distribution. After integration of (9.6) over the membrane thickness L the permeation is found to be: Jk,
Fk . . . .
2~r
AP
(
8
31:.0 k L rcRTM
)0.5 (9.7)
Equation (9.7) shows that the permeation of a single gas in the Knudsen regime is: proportional to the average pore radius r; independent of the pressure (note this is an important difference from viscous flow); proportional to M -~ Note that the Knudsen diffusion coefficient is obtained by introducing the above mentioned geometrical coefficients in Eq. (9.4). -
-
-
Transition flow
Transition flow occurs when viscous flow and Knudsen diffusion both play a role, that is in the region with Knudsen number values around unity. Estimates of the value of Kn can be made with the help of the gas kinetic expression for )~: ~, = ~
1
~J2- ~ 2
RT
~
P
(9.8a)
with the average collision diameter, o, of different molecules c~1and c~2defined as: (~1 } (~2
c~=~
(9.8b)
For example for Argon (M = 40) at a pressure of I Bar and a temperature of 293 K, the value of ~, equals 6.9x10 -8 m and with d = 10 nm the value of Kn = 7. This means that in pores with d < 10 nm and under ambient conditions, Knudsen diffusion occurs; pores with a size of 100 nm however fall in the transition region.
340
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
Most attempts at describing the Knudsen-viscous Poiseuille transition involve a combination of Eqs. (9.2) and (9.6). For single gases this is a good approximation and after some reorganisation and integration of Eqs. (9.2) and (9.6) and assuming a linear pressure drop across the membrane this yields the expression for the total flux Jt: It =
-
~ (Ar2 -P+ Br) AP ,r L
(9.9a)
with A = 1/(8TIRT) and B = (2/30k)~I(8/TIRTM). For further discussion, see Section 9.2.4.2 on binary gases. The expression for F t follows from:
Ft= Jt /AP
(9.9b)
This is a very useful equation which is used for several purposes (see below). It is noted that the occurrence of slip flow has been ignored so far. This point will be discussed below. An alternative, semi-empirical, expression has been proposed by Schofield et al. [16]: Jt = a. pb. AP
(9.10)
In this equation the exponent b of the dimensionless pressure Pd (= P/Pref) is a measure for the extent of viscous flow (b = 1 for viscous flow and b = 0 for Knudsen diffusion). The reference pressure Pref is chosen as a typical or average pressure for the range of applications concerned. Equation (9.10) can be substantiated by manipulation of (9.2) and(9.6), using the kinetic molecular expression for the gas viscosity 11:
1 1PMq(BRT I n =-~ NM v X- 2 RT ~M
(9.11)
This yields an expression for the permeation which can be approximated by Eq. (9.10) provided P and Pref do not differ by more than a factor of 3. The advantage of Eq. (9.10) as an engineering equation compared with (9.9) is that it does not require knowledge of the membrane properties, r and z but expresses the gas flux in terms of a membrane property a and of operation conditions b.
Geometric aspects Many authors have derived expressions for the transport in porous media based on specific models, taking into account pore size distribution under different assumptions for the interconnectivity (tortuosity) of the pore network. An overview has been given by Cunningham [1], some discussions are presented by Karger and Ruthven [3] and Dullien [2]. Sometimes reasonable
9 w TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES w r r H GASES AND VAPOURS
341
predictions can be made, which require however a detailed knowledge of the pore shape as well as pore shape distribution. The most simple definition of the tortuosity is that of the increased flow (diffusion) path length with respect to the shortest distance in the flow direction. This means 1: = Le/L w i t h L e the effective length. However other effects influencing the transport resistance play a role (e.g., roughness, interconnectivity, etc.) and are specific for the transport regime under consideration. For example the reflection coefficient 0 plays a role in the Knudsen diffusion but hardly at all in the viscous flow regime. Therefore some authors have split the tortuosity factor into a purely geometrical one (Le/L and in a part accounting for all other aspects. Other authors put all the structural complexities in a single parameter and some of them call it-torfuosity (which is incorrect). Karger and Ruthven define the tortuosity 1:with the help of: D=
aDp
(9.12)
where Dp is the diffusivity for a straight cylindrical pore and D is the experimentally determined diffusion coefficient. So all complexities of the structure are hidden here in ~. In practice it is simpler to treat the tortuosity 9as an empirical factor and to determine it experimentally (see below). The same holds for other geometrical parameters like 0 and [~, discussed in connection with Eqs. (9.6 and (9.2). In porous pellets of packed particles a correlation of the type ~/~ = constant is frequently found [3]. The validity of this expression is not shown however for low values of the porosity (r < 0.30) and very small pore sizes. Experimental tortuosity values generally fall inthe region 2 < "r < 5, but in special cases much larger values havebeen reported. Leenaars et al. [17] reported values of "r = 6-7 for membranes consisting of a packing of plate-shaped (boehmite, gamma alumina) particles.
Experimental determination of geometric parameters of membranes The expression for the total flux Jt of a single gas through a homogeneous, single wall membrane as given by Eq. (9.9) can be used for several purposes. For mesoporous membranes which are definitely in the Knudsen regime, the permeation plotted versus the average pressure P should give a horizontal line because F k is pressure independent. If the curve has a certain slope this points to a contribution of viscous flow and this in turn means that there are defects in the membrane in such a number and size that they cause a measurable viscous flow contribution. A plot of F t v s P for defect-free membranes which are definitely in the transition region yields a curve with a certain slope, which intersects the permeation axis. This is shown in Fig. 9.5.
342
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANESWITH GASES AND VAPOURS
2.20 A
o O x f0 n
H2
4,
1.6S
m "m
E d)
O
E
H 9
1.10
"...,
>,
N2
l u
.O ca 4)
E
0.SS
I--
4) a.
CO
U. -;
.... '
'
~' ..... i
0.5
' '
....... '
'
I :-~'
1.0
~
--'
' .... i : '
1.5
~
'
~ ....
2.0
p, Average pressure across membrane (Pa x 10 "s)
(a)
Fig. 9.4. Gas permeabilifies versus average pressure at 20~ (a) and 538~ (b). Layer thickness of top layer is 3-5 ~tm. After J.C.S. W u et al. [19].
From the slope the value of r can be calculated,while the intersection with the permeation axis (P = 0) yields the value of ~/(~.e) and so e can be calculated as well as the Knudsen contribution to the total flow. The pore size should be well defined in these cases and so the pore size distribution should be reasonably sharp. When the total porosity really is representative for all the active pores (thus, e.g., not many "dead ends" should be present), the value of the tortuosity ~ can then be calculated. Otherwise the parameter ~ is used as a fitting parameter. Examples of this type of analysis are given by, e.g., Eichmann and Werner [18] for Nuclepore membranes with a pore diameter of 30 nm. The Knudsen permeability is given for several non-condensable gases and is reported to be 3.6x10-8/L and 1.1x10-8/L mol m / m 2 s Pa for H2 and N2 at room temperature respectively. Because the thickness L is not given, the actual flux obtained cannot be recalculated from their results. Wu et al. [19] and Keizer et al. [20] reported permeability data on thin gamma alumina layers with a porosity of about 0.5 and a layer thickness of 4.0 ~tm (Keizer et al. [20]) and 4.0-7.0 ~tm (Wu et al. [19]), supported by an (~-alumina supporting system. Wu did not correct for the support resistance. Some of Wu's results are shown in Fig. 9.4.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
343
2.0 A
o
H2
,i.=,
o I-=
x t~
.=
1.5-
(h
t'~i"
.=
.=
He
.=
.=
d)
1.0-
O
E >,
==
N2
mm =mm
.Q
0.5-
O
E
i
illili
A=L
I
. . . . . . . .
A ..,.
Ill
III
--
---,.
Im
(I} CL
CO
LI=
i
,r
'
'
I
'
'
'
0.5 p,
Average
' i .... '
"~
'
1.0 pressure across me~ne
(b)
'
I +'
'
'
1.5
'
2.0 ( P a x 1 0 4)
Fig. 9.4 ( c o n t i n u e d ) . C a p t i o n o p p o s i t e .
The Knudsen permeabilifies (obtained for P = 0) at 20~ are 1.9x10 -1~ and 5.5x10 -11 mol m / m 2 s Pa for H 2 and N 2 respectively. Taking an average value of 4.0 ~tm for the thickness of the g a m m a alumina layer and assuming the resistance of the supporting system to be negligible, this yields values for the permeation F at room temperature of 5.5x10 -s and 1.1x10 -5 m o l / m 2 s Pa for H 2 and N 2 respectively. Keizer et al. [20] reported permeation data on a similarly m a d e g a m m a alumina membrane, supported by a different alpha support and corrected for the support resistance (see Section 9.5.2). Their results are shown in Fig 9.5. The N 2 permeation at 20~ is reported to be 4x10 -6 m o l / m 2 s Pa in reasonable agreement with the value reported by W u et al. It m u s t be noted that in the above-mentioned treatment the absence of surface flow is assumed. \
Relative contributions of viscous and Knudsen flow; some data As will be discussed in Section 9.3, small contributions of viscous flow to the total flow in the transition region can have a considerable effect on the selectivity in separations. Therefore some typical data are given in Table 9.2. for N 2 a s a reference gas. Note that for light gases (H2) the contribution of the viscous flow differs considerably from that given in Table 9.2.
344
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
l
|
.
13)
0 0
.
,
,
.. 0
_
0
m,,i
0
9
0
e.d
•
ll)
0 ,i,a
0~ o
(2)
0
100
"
tOO Pressure (kPal
Fig. 9.5. Nitrogen permeation as a function of pressure for a supported 7-A1203 m e m b r a n e at 20~ (1) support; (2) support + top layer; (3) top layer. TABLE 9.2 Permeation data of macroporous supports and mesoporous layers for N2 at 20~ and an average pressure p = I bar in the transition region of viscous to Knudsen flow. The fraction of the viscous flow (b) to the total flow is given by Fr, the remainder is the Knudsen contribution (a) Thiclaless (10-6 m)
Permeation ( m o l / m 2 s Pa) Pore radius (10-6 m)
2000
10
1
0.1
(a) 1.25x10 -'4
1.25x10 -5
1.25x10-6 2.8x10 -7
(b) 2.8x10 -3
2.8x10 -5
sum 2.92x10 -3
4.05x10 -5
1.55x10 -6
Fr 0.92
0.69
0.18
(a) 2.5x10 -4 (b) 2.8x10 -5
5.1x10 -5
lx10 -5
sum 2.8x10 -4 F r 0.10
lx10 -5 0.025
9 ~ T R A N S P O R T A N D S E P A R A T I O N PROPERTIES OF M E M B R A N E S W I T H GASES A N D V A P O U R S
345
S lip flow The summation of viscous and Knudsen flow as given in Eq. 9.9 is not strictly valid. In long cylindrical capillaries a minimum in the permeation has been observed at low pressure when plotting the permeation in the transition region versus the pressure [1,2]. This minimum has been described already by Knudsen, but it has not been observed in porous media with a tortuous network, although it remains a controversial point in the literature. This minimum is caused by the occurrence of 'slip'. When the velocity of gas molecules at the wall is not zero, a slip (wall) velocity must be taken into account. This effect becomes significant when the mean free path ~ of the gas molecules is of comparable magnitude to the pore size (so in the transition region) and is negligibly small when ;~ is much smaller than the pore size. After the last collision of a gas molecule with the wall it travels a certain distance. "Wall velocity" means now the average flow velocity in the immediate vicinity of the wall, but still in the gas phase. At a distance from the wall equal to the mean free path, the gas molecules have, on average, a non-zero velocity and as the mean free path becomes an increasingly greater fraction of the capillary diameter, the wall velocity increases in significance relative to the average velocity. Starting with P = 0 and increasing the pressure, first the decrease of ~,/dp dominates (the flight length decreases) and so the flux decreases. At higher pressures intermolecular collisions increase and so does the flux. The effect of slip flow can be treated either as an extension of a pure viscous flow or as an extension of a Knudsen flow. The simplest method is by adding an additional term (R/2~) 9(P/RT) 9dP/dz to Eq. (9.2), with ~ being the slip coefficient which is proportional to P. As shown [1] the Dusty Gas Model expresses the slip flux in terms of a Knudsen diffusion. This implies that the slip flow is inversional proportional to the square root of the molecular mass and this has the interesting consequence that slip flow can contribute to segregative properties in gas mixtures.
9.2.3.3 Surface Diffusion and Capillary Condensation Surface diffusion When the temperature of the gas is such that adsorption on pore walls is important, experimental results show that the preceding laws for gaseous flow are no longer valid. Overviews of the subject have been given by Uhlhorn and Burggraaf [21a,b] and have been treated by many authors in detail [22-26]. The mechanism of surface flow is rather complicated and three main groups of mechanism can be distinguished [22b]: The hydrodynamic model: In this model the adsorbed gas is considered as a liquid film, which can 'glide' along the surface under the influence of a pressure gradient.
346
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
The hopping model: This model assumes that the molecules can move over the surface by hopping over a certain distance with a certain velocity. The random walk model: This model is based on the two-dimensional form of Fick's law and is most frequently used in the literature. For relatively low surface concentrations, the surface flux Js for a single gas is generally described by the two-dimensional Fick law:
dCs Is = - 1: Ds dz
(9.13a)
where Cs is the surface concentration (mol/m2). Expressed in terms of directly measurable parameters this gives: J s - - P (1 - 8) (~] Dsdq dz
(9.13b)
with Cs- qp(1 - 8) and p(1 - 8) the density of the porous material. To demonstrate the influence of the pore size on the magnitude of the surface flow it is considered that: q - 0s" Sw "Csa t
(9.14a)
and s
Sw = 2--= pr
(9.14b)
where 0s is the occupancy, defined as the mole fraction occupied by adsorption relative to a monolayer with sorption capacity Csat in m o l / m 2 and Sw is the surface area of the porous medium. Substituting (9.14) into (9.13) one obtains: Is--
2~ 2
Ds dO - Csat r dL
(9.15)
This expression shows that Js increases strongly with decreasing average pore size. Assuming local adsorption equilibrium (adsorption processes are fast), Eq. (9.15) can be converted in terms of pressure instead of concentration using dq / dL = d q / d P . d P / d L and the expressions for the adsorption isotherm which relate q or 0s to P, e.g., for Henry's law q - b. P
(9.16a)
for Langmuir adsorption KiP 0i = ~ 1+KIP
(9.16b)
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
347
with Ci,sat the saturation concentration of i (mol/kg or m o l / m 2) and q - 0.Ci,sa t (kg/mol) = 0 . C s a t (mol/m2).Sw . Generally an Arrhenius (exponential) type of relation represents the diffusion coefficient as a function of the temperature, with AQa the activation energy of diffusion. Similarly the parameters b and K (9.16) can be expressed with Arrhenius functions with Qa the (isosteric) heat of adsorption. Consequently Js is also activated with a total apparent activation energy of (Qa-AQa). For chemisorption AQa has about the same value as Qa [1]. For physical adsorption the value of AQa is < (0.5--0.66)Q a. Since the surface flux is small at very low temperature as well as very high temperature there must be a maximum. The possibility of observing this maximum depends on the relative magnitudes of Qa and AQa. Note that it is assumed so far that D~ as well as Qa are independent of 0. This is not true for larger values of 0. The value of Csat usually decreases with increasing temperature. Some data Recently, Bai et al [27] reported permeation and separation data of zeolite membranes on a supporting system consisting of a thin gamma alumina layer (thickness 5 ~tm, pore diameter 5 nm) on an m-alumina substrate. The log-log plot of the (measured) permeation of the y-~ supporting system as a function of temperature for H2, Ar, SF 6 and isobutene was activated and gave linear curves with a slope of-0.66 to -0.76 depending onthe gas and conditions. For ideal Knudsen transport a slope of -0.5 is expected. Furthermore, the single gas permeation ratio of isobutene/Argon equals 2.4 (Knudsen ratio is 0.83) at room temperature and equals 2.2 at 770 K. This means that even at high temperature the transport of C4 hydrocarbons (in the Knudsen regime) is significantly increased by surface diffusion in the y-alumina layer. Uhlhorn et a1.[28] also reported surface diffusion on modified y-alumina layers with a pore diameter of 4 nm and unsupported layer thickness about 20-30 ~tm) and found that at 20~ about 30% of the total flux of CO2 through the membrane was carried by surface diffusion. Modification of the y-alumina with 2 wt% MgO strongly increased the adsorption (0 and C s a t in Eq. (9.15) increase), but this did not increase the value of Is due to the strong increase of Qa. Modification of y-alumina with 17 wt% of finely dispersed Ag increased the flux of H 2 considerably above the Knudsen level as shown in Fig. 9.6. At 25~ and P = 60 kPa the flux by surface diffusion is 2.5 times the Knudsen flux. Increasing the H 2 pressure decreased the contribution of the surface diffusion. This is due to saturation of the adsorption (e approaches unity in Eq. (9.15)) with increasing pressure, causing the surface flow to become constant while the Knudsen flow continues to increase. Finally, Sloot et al. [29] reported a surface flux contribution of about 40% of the total flux of SO2 in ~-alumina membranes with a pore diameter of 350 nm
348
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
12 T
Zg
-
298
K
o\ O~Q,~
_,~ ._~ .m
0
m
,,,,b m
,
0
~
,,,,,, m ,
m
u w m - , , m
expiwm,mnta~ rario
I
I
m
u
m
m
m
tl'e~~
|
m
w
u
I ,,
100
m
i
rst~)
2O0
m
~
I
w
g
~
m
..............
3OO
p ~a) Fig. 9.6. Surface diffusion as s h o w n b y c o m p a r i s o n of the e x p e r i m e n t a l ( e ) a n d theoretical flux ratio of H2 a n d N2 at 25~ on a n o n s u p p o r t e d 7-A1203 layer m o d i f i e d w i t h 17 w t % Ag. After U h l h o m et al. [28].
and modified with impregnated y-alumina ~ in the temperature region 170-290~ and with P = 2-6 bar. This means that the membrane was in the molecular flow regime (note: Wicke--Callenbach measurements, no absolute pressure gradient) and the surface diffusion flux was combined with the flux from molecular diffusion in the gas phase. An overview of data for different gas-membrane combinations is given by Uhlhorn [21]. It is concluded that in all treatments in the literature the surface flux is taken as an additional contribution to the gas flow and usually the total permeation is obtained as a linear combination of gas and surface permeation,
9 -- TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANES wrrH GASES AND VAPOURS
349
0
,M
0 0
surfaceflow
Temperature Fig. 9.7. Schematic view of total flow (permeation) as a function of temperature for the combination of gas and surface flow.
as derived from Eq. (9.15).This is an ad hoc assumption for which no justification is given. The generally observed trend of the total permeation of a single gas versus temperature (including surface diffusion) is given in Fig. 9.7.
Multilayer diffusion and capillary condensation An extensive analysis of data and theories describing permeation by surface flow and capillary condensation is given by Uhlhorn [21a]. A fully satisfactory explanation of surface flow mechanisms has not been provided. Some very useful models and equations are however available and will be discussed below. With increasing pressure and at temperatures below the critical temperature the surface coverage (occupancy) can become larger than unity. In this case the adsorbed molecules behave like a sliding film on the internal surface of the porous membrane under the action of a bi-dimensional spreading pressure related to the gas pressure. This situation is best described by a hydrodynamic model first proposed by Flood and Huber [30] and further developed by Gilliland et al. [31,32] and by Tamon and Toei [33,34]. These models cover the complete range of coverages including capillary condensation. According to Gilliland it follows that the permeation Fsm due to multilayer flow (and not too far from a monolayer coverage) is
350
9 -- TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
I ,M
" i
i
0
I I
I
2
relative ~euure
Fig. 9.8. Schematic picture of the permeation as a function of the relative pressure in the presence of capillary condensate. After U h l h o m et al. [21]. (1) Onset of multilayer adsorption; (2) pores are completely filled. 2
RT Cs
Fsm
= ~ ~Cr~L p
(9.17)
where Cr is the flow resistance, Cs the surface concentration and the constant ~t incorporating geometrical characteristics of the pore system. With the onset of multilayer flow the measured flow strongly increases (see Figs. 9.8 and 9.10). It should be noted that in small pores the increasing thickness of the adsorbed layer decreases the effective radius of the pore for diffusion through the gas phase. This is important for the selectivity in binary mixtures. At temperatures below the critical point of the diffusing gas, the increase of pressure first leads to multilayer adsorption until finally all pores are filled with liquid. This phenomenon is called capillary condensation and this process starts when the gas pressure P surpasses the pressure Pt given by the Kelvin equation which is for a cylindrical capillary: RT
Pt
r~s c o s ~l/
V----~In ~00-- 2
r
(9.18)
where P0 is the saturated vapour pressure above a flat surface. This equation predicts that the smaller the pore radius, the lower the pressure at which capillary condensation starts, provided a good wettability of the pore surface by the condensate is present. The general picture of flow due to capillary condensation is given in Fig. 9.8 for a narrow pore size distribution. An important conclusion from Tamon and Toei's studies is that permeation
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS I
I
,ll
II I
II
I
II I
t
I
ITI
351
P2
pt
t2
l."r<,t,2
for vapour flow is higher than for liquid flow. This implies that as soon as all pores are filled with liquid,the permeation drops (see Figs. 9.8 and 9.10). This qualitative picture was quantified by Lee and Hwang [35], based on Gillilands hydrodynamic model. In order to describe the transport, Lee and Hwang proposed six flow modes, which differ according to the site where the meniscus is formed, as illustrated by Fig. 9.9. In this model also the decrease of the pore radius due to the formation of an adsorbed layer is incorporated. Flow I in Fig. 9.9 is the case of combined Knudsen molecular diffusion in the gas phase and multilayer (surface) flow in the adsorbed phase. In case 2, capillary condensation takes place at the upstream end of the pore (high pressure P1) but not at the downstream end (P2), and in case 3 the entire capillary is filled with condensate. The crucial point in cases 3 and 4 is that the liquid meniscus with a curved surface not only reduces the vapour pressure (Kelvin equation) but also causes a hydrostatic pressure difference across the meniscus and so causes a capillary suction pressure Pc equal to Pc = -
2Gs cos
(9.19)
1"
The overall capillary pressure drop across the cylindrical pore is given by APc='~m
ln~0-1n
-" Wm Pm
(9.20)
352
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
where Pm is the mean pressure and APg = P1-P2. Note that Eq. (9.19) predicts that the actual pressure drop is two orders of magnitude larger than the gas pressure drop. Actually the capillary pressure Pc works on the effective pore radius, which equals ( r - t) with t the adsorbed condensate film thickness and which is assumed to be immobile; this point is discussed below. A force balance for one end of the capillary with length L and radius r then yields the effective capillary pressure Pc,eff: 2c~(rPc,eft = -
t ) 2 COS ~ r2
(9.21)
Combining (9.21) with the Kelvin equation (9.18) and using Darcy's law for liquid flow through porous media (as similarly done before for gas flow, see Eq. (9.1)) yields an expression for the gas flux Jcf3of capillary condensate in case 3 of Fig. 9.9:
~tRT [(r-t1) 2 P1 (r-t2) 2 P21 r----d---In p--~- r--------d---In -~o
Jcf3 - n VmL
(9.22)
The other cases, which describe situations where the capillary is not completely filled and/or where the meniscus on one side is not present, follow from (9.22) by adjusting the relevant capillary pressure term and using the filled length z instead of the real length L. Note that case 6 of Fig. 9.9 obeys Eq. (9.2). The film thickness t is estimated with the help of the BET adsorption equation (t plot) giving: t = qBETWt/St
(9.23)
where Vt is the specific volume of the adsorbed layer, St is the specific surface area of the porous membrane and qBETis the amount of adsorbed condensate. The geometric constant ~t contains characteristic pore parameters (porosity, tortuosity) and is determined by liquid permeation experiments.
Some illustrative data A few publications have reported the permeation of capillary condensate in inorganic, mesoporous membranes. Lee and Hwang [35], using their equations (9.22) and (9.23) found a good agreement between measured and calculated permeabilities in Vycor glass membranes for Freon at 19-41~ and a reasonable agreement for water vapour at 70~ Maxima in the permeation (or the permeability) are indeed found at relative mean vapour pressures Pm ranging from 0.6-0.8 and with permeabilifies of 20-50 times the Knudsen gas permeability. Similar maxima have been found by Rhim and Hwang [36] for C2H 6, n-butane and CO2 in Vycor glass membranes and by Uhlhorn et al. [37] for propylene at 263 K in l-alumina membranes(pore diameter of about 3 nm) as shown in Fig. 9.10.
9
-
-
TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS | 0
35
X
3O
r |
~opylene
353
~..
desorption
25
~dsorption
pml 0
oM ~
r
5
0 ......
0,00
nitrogen ;
~ A v v
.,nlk v
i
ii
_
,6.,=. vv---
0,5O
9 "T
_ A A ,vv
_
1.00
Rel. pressure Fig. 9.10. Permeation of a s u p p o r t e d 7-alumina thin film for nitrogen and p r o p y l e n e at 263 K as a function of the relative pressure of pr0pylene. After U h l h o m et al. [21,37].
Uhlhorn analysed his results in terms of the model of Lee and Hwang taking into account the slit-shaped pore geometry of ~-alumina membranes. This means that capillary condensation in the adsorption mode did not take place. The meniscus in the adsorption mode is formed due to the fact that the adsorbed layer thickness completely fills the pore width.In the desorption mode the usual description in terms of capillary condensation holds. Asaeda and Du [38] separated alcohols from water condensate by flow through a silica-alumina membrane with 3 nm pores. Sperry et al. [39] demonstrated that capillary condensation can be effective in permeation and separation with modified mesoporous 7-alumina membranes (pore diameter 4 nm) at elevated temperatures, provided the applied pressure of the condensing gas is increased. Capillary condensation of methanol was observed up to a temperature of 473 K at a partial pressure of 23 bar. This is about 0.65 Psat, so a large range of operation pressures is possible. Also here the permeability drops sharply at the onset of capillary condensation and is in this region about three times the Knudsen permeation. A limiting factor for the maximum allowable total pressure drop across the membrane is the requirement of a stable condensate which should not be blown out of the pores. For methanol in pores of 4 nm in gamma alumina the total allowable pressure drop is 0.28 bar at 373 K and 0.05 bar at 473 K [39].
354
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
Modified membranes and the transition to micropore diffusion Modification of mesoporous membranes can result in (i) a decreased pore size which increases the contribution of surface diffusion, and (ii) a change in the nature of the pore surface and consequently a change in all types of interaction energies with the gas phase. Both phenomena have an effect on permeation and separation. Abeles et al. [39a] investigated some fundamental aspects of capillary condensation and surface flow. The pore system of Vycor glass (pore radius 3.1 nm) was treated with ClsH38 molecules and in this way the pore radius was decreased to 2.1 nm. The permeation of toluene was then studied on both systems. A similar type of model was derived as used by Lee and Hwang and was investigated for case 2 in Fig 9.9. The main conclusion was that the effective viscosity of liquid toluene increases with decreasing pore size due to a rather strongly adsorbed monolayer on the pore walls. The flow of this layer is determined by a thermally activated friction parameter whose activation energy is considerably larger than that of the bulk liquid. Note that this layer should be similar to the layer assumed by Lee and Hwang. Surface diffusion is primarily due to molecules adsorbed on top of the first layer. The friction coefficient of the adsorbed layer on membranes treated with C18H38 is somewhat larger than that of the non-treated one due to a changed interaction energy. Local equilibrium is indeed obtained because the exchange rate of molecules between vapour and liquid (10-1 mol/cm 2 s) is orders of magnitude larger than the molecular flux (10-6 mol/cm 2 s). Okubo et al. [40,41] treated Vycor glass membranes with tetra-ethoxysilane which was initially adsorbed and finally decomposed on the pore wall by heat treatment. The pore size was expected to be decreased by this treatment. As a result of this modification the permeation decreased and the permeation as a function of temperature increased (compared with that of the non-modified glass) for the gases He, 02, N2, Ar, H 2 and CO2 and became activated. The authors argue that surface diffusion cannot explain this result and suggest that the modified system is in the transition region of Knudsen to molecular sieving (micropore diffusion). Rao and Sircar [42] made nanoporous carbon membranes (thickness 5 ~tm) by repeated deposition of a polymeric latex film on a mesoporous graphite substrate, followed by decomposition of the polymer to Carbon. Permeation and selectivity in separation of He, H 2 and some hydrocarbons were studied. The average pore diameter of the carbon membrane was estimated to be 0.5-0.6 nm. All hydrocarbons exhibited larger pure gas permeabilities than H2 because of their preferential adsorption on the carbon surface. The permeation was activated and followed the order H 2 < CH 4 < C2H6 > C3I-I8 > C4H10. For the higher hydrocarbons (>C2) the specific amount of adsorbed material increases as the molecular weight increases but the adsorption strength also increases
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
355
causing a decreased mobility at the surface and resulting in a net decrease of the permeability. This is a similar phenomenon to that observed by Uhlhorn et al. [37] for CO2 transport on y-alumina modified by MgO. The transport mechanism suggested by Rao is surface diffusion, but from the size of the pores it is clear that the transport is at least in the transition region of surface diffusion to micropore diffusion.
9.2.4 Permeation in Binary Gas Mixtures in Macroporous and Mesoporous Membranes
9.2.4.1 General Considerations Transport of components of a mixture through a porous medium is usually caused by mole fraction gradients (in isobaric systems) as well as total pressure gradients. The transport in the Knudsen regime, with mainly molecule-wall interactions, and in the continuum regime (bulk molecular diffusion) with mainly molecule-molecule interactions, are well understood. In the transition region between Knudsen and continuum diffusion interpolating models (e.g. the Bosanquet equation) are used to describe the effective diffusivity. A superimposed pressure gradient across the system forces convective motion of the components. In order to take into account the effect of both mechanisms, more complicated models have been proposed [1,2,11,43-45]. Overviews have been given by Uhlhorn et al. [21] and more recently by Veldsink [46]. The models differ in the way the different mechanisms are combined and which coupling terms are taken into account. The most important coupling effects are the occurrence of 'drag effects' in mixtures and of momentum transfer between different species. Drag effects on molecular species a and b occur in isobaric binary mixtures a-b due to differences in molecular velocities between species a and b, which induce internal pressure differences causing a net flow of the mixture which has to be superimposed on the diffusive fluxes of a and b. The Dusty Gas Model (DGM) is one of the most suitable models to describe transport through membranes [11]. It is derived for porous materials from the generalised Maxwell-Stefan equations for mass transport in multi-component mixtures [1,2,47]. The advantage of this model is that convective motion, momentum transfer as well as drag effects are directly incorporated in the equations (see also Section 9.2.4.2 and Fig. 9.12). Although this model is fundamentally more correct than a description in terms of the classical Fick model, DGM/Maxwell-Stefan models yield implicit transport equations which are more difficult to solve and in many cases the explicit Fick type models give an adequate approximation. For binary mixtures the DGM model can be solved explicitly and the Fickian type of equations are obtained. Surface diffusion is
356
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
/
Feed Qf Xo, PF
Retentate
Qr
Y, P~
membrane Sweep
,
"/
\ /
T
,,
Permeate
Qp
Y, Pp
Fig. 9.11. A perfect mixing model for gas separation, x and y are mole fractions, Q is the molar flux at pressure p. only incorporated in the later developments (extended DGM and Fick models [46]) but its implementation is not as straightforward as that of the other mechanisms. In membrane systems, which require segregative properties e.g. in gas separation, usually large permeation in combination with a good separation factor (selectivity) is required. This is obtained by applying an external pressure gradient and a low partial pressure at the permeate (low pressure) side of the membrane (see also Section 9.3). A frequently used membrane system is schematically given in Fig. 9.11 as an example. The gas is applied as a mixture to the retentate (high pressure) side of the membrane, the components of the mixture diffuse with different rates through the membrane under the action of a total pressure gradient and are removed at the permeate side by a sweep gas or by vacuum suction. Because the only segregative mechanisms in mesopores are Knudsen diffusion and surface diffusion/capillary condensation (see Table 9.1), viscous flow and continuum (bulk gas) diffusion should be absent in the separation layer. Only the transition state between Knudsen diffusion and continuum diffusion is allowed to some extent, but is not preferred because the selectivity is decreased. Nevertheless, continuum diffusion and viscous flow usually occur in the macroscopic pores of the support of the separation layer in asymmetric systems (see Fig. 9.2) and this can affect the separation factor. Furthermore the experimental set-up as shown in Fig. 9.11 can be used under isobaric conditions (only partial pressure differences are present) for the measurement of diffusivities in gas mixtures in so-called Wicke--Callenbach types of measurement. Isobaric applications in the continuum regime, making use of molecular bulk diffusion a n d / o r some viscous flow are found in catalytic membrane reactors. The membrane is used here as an intermediating wall or as a system of microreactors [29,46]. For this reason some attention will be paid to the general description of mass transport, which will also be used in Sections 9.4 and 9.5.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
357
The treatment in Section 9.2.4 will first start with some simple limiting cases (Knudsen diffusion and viscous flow in mixtures), followed by a comparison of an extended Fick model with the DGM model derived equations for binary gas mixtures. Subsequently a treatment will be given of a direct application to membrane separation of a set of equations derived from the model of Present and Bethune by Wu et al. [18] and by Eichmann and Werner [19]. 9.2.4.2 Knudsen Diffusion For Knudsen diffusion collisions between particles are negligible and molecules of different species move entirely independent of each other under the action of their own concentration (or partial pressure) gradient. There is no fundamental difference between flow and diffusion. The resulting expression for the total flux Jk,t of a mixture with component fluxes Jk,1 and Jk,2is Jk, t = Jk,1 + lk,2
(9.24a)
where Jk,1 and Jk,2 are related by /k,__}_l= Jk,2
Xl
(9.24b)
1 - Xl
where xl is the mole fraction of component 1. The expression for Jk,2 and Jk,2 is given by Eq. (9.6). If the mean molecular weight <M> is defined by <M> -~ = Xl' Mi ~ + (1 - Xl)" M2 ~
(9.25)
then the Knudsen permeation of the mixture is obtained by inserting <M> -~ from (9.25) for (M) -~ from (9.7). Equation (9.24b) in combination with (9.6) predicts for a non-isobaric and equimolecular mixture (xl = x2 = 0.5) that the ratio Jk,1/Jk,2 is proportional to sqrtM2/M1. This is the ideal permselectivity of the mixture. 9.2.4.3 Viscous Flow and the Transition Region
Viscousflow The viscous flow of a binary mixture which is fully in the continuum regime does not affect the concentration of both gases and relation (9.2) applies for the mixture as for a single gas with the mixture viscosity rl(x) of the mixture with a constant mole fraction x. The viscous flow Jv~ of each species i of the mixture equals the total flow Jv,t multiplied by the mole fraction x (proportional to the partial pressure p): Jv,i = xi. Jv,t and Jv~ given by Eq. (9.2).
(9.26)
358
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
The transition region of Knudsen and continuum diffusion or viscous]low Two important cases must be considered: (i) non-isobaric, and (ii) isobaric situations. The non-isobaric situation will first be discussed. - Estimate of magnitude of different contributions: According to Eq. (9.9a) the viscous flow increases with r 2 and with P, while the Knudsen diffusion increases with r and is independent of pressure. This means that the contribution of the viscous flow to the total permeation increases with r and p. Using relations (9.3) and (9.4b) or (9.7) it can easily be shown that in a first approximation the total permeation F can be written as: F=Fk.
(
3F'P/ l+l--6.rl.v
(1-~) =Fk" 1+ .A
(9.27a)
Using the gas kinetic relations between r, rl, v and ;~ we find A = n K~ and so:
(
f =f k 91 + ~ 16 9K n
/
(9.27b)
Equation (9.27b) is useful to estimate the contribution of viscous flow to the total permeation. For argon at 1 bar and 293 K it is found that with r = 10 n m (K n = 7), 98% is Knudsen diffusion, with r = 1 ~tm (Kn = 0.07), 67% is viscous flow and 33% is Knudsen diffusion. So with larger pores and higher pressure in non-isobaric systems viscous flow is the dominant contribution and molecular diffusion can be assumed to be negligible. Note that in this treatment m o m e n t u m transfer is ignored. - The extended Fick model: An extended Fick type of equation is used by Veldsink [46] to incorporate this m o m e n t u m transfer. The total flux Ji of component i can be written as a superposition of the total pressure driven viscous flow on the diffusional flow component. 5(xiP) Bo 1 De ~ + ~ x ~ P J i - - RT 5z 11
5P)
(9.28)
where Bo is the permeability coefficient, xiP the partial pressure and D e is the effective diffusion coefficient of i in the mixture. The term 'effective' indicates that geometric effects of the pore structure are incorporated in D e and Bo (with D e = TI/'cD ~ with D ~ the expression for a cylindrical pore). In the transition region the transport resistances are assumed to be in series as expressed by the Bosanquet equation: 1
1
~ = ~ +
Die,j
~
1
Diem Diek
(9.29)
9 m TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
359
where D ei,m and D ei,k a r e the effective diffusion coefficients for continuum and Knudsen diffusion respectively. Di,k is given by the gas kinetic expression. So it follows:
4 ,N/8RT e D i,k -----T, -3 Kn ~M i
(9.30)
If the mixture continuum diffusivity is unknown it can be estimated using Blanc's law: tl
De z,m
= ~1 1
--
y__, Died 9xj X i
(9.31a)
j=l, j~l
For binary diffusion (9.31a) reduces to d In p _ Dll (Fick) Diem = D12 d In c i
(9.31b)
Dij (here D12) is the diffusion coefficient of the pair i-j. It can be experimentally measured by Wicke-Callenbach type (isobaric) measurements (see Sections 9.2.4.3 and 9.4.2,3) or calculated with the help of the first order approximation Chapman-Enskog relation [1,4] which is written as
.N/,r3 D12-
0.00262
L
+__M___2.1
M 2 P (~12 ~'~12
j
(9.32)
where oh2 is the collision diameter (taken as the arithmetical mean of the individual component diameters), ~'~12 is the first order collision integral, which is tabulated by e.g. Hirschfelder [4] and which is a function of the temperature. P is the pressure in atm and D12 is obtained in c m 2 s -1.
The Dusty Gas Model (DGM) In the DGM model as presented by Mason and Malinauskas [11a] all the different contributions to the transport are taken into account. The wall of the porous medium is considered as a very heavy component and so contributes to the momentum transfer. The model is schematically represented in Fig. 9.12 for a binary mixture (in analogy with an electrical network). As can be seen from this figure, the flux contributions by Knudsen diffusion Jk,i and of molecular (continuum) diffusion of the mixture Jm,12are in series and so are coupled. The total flux of component i (i = 1,2) due to these contributions is Ji, km" Note that Jk,i = Jm,12. The contribution of the viscous flow Jv,i and of the surface diffusion Js,i are parallel with Ji, km and so are considered independent of each other (no coupling terms, e.g. no transport interaction between gas phase and surface diffusion).
360
9 -- TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANESWITH GASES AND VAPOURS
L
P Jr
I
L_
J v,i
J,.,
J m,1I2 =''"
,i. Fig. 9.12. Schematic representation of the Dusty Gas Model. Ji is the molar flux of component i; k = knudsen, m = molecular diffusion, v = viscous flow, s = surface diffusion.
The flux expression for a single species i in a multi-component mixture with n components according to the DGM model results in 1"/
~_~ xi ' Jj- x.j. ' Ji Ji 1 ~)xi xi l BoP ) SP j=l, j,i Pi Dfj - p . Di,---~k- RT 8z + PRT ('11 Di,--~k+ 1_ ~
(9.33)
with Diek- g/T,.Di~ given by (9.30) and D,~ = 8/t.D~ given by (9.32)or directly measured. B0 is the permeability coefficient for a porous medium (m 2) and it can be obtained from the slope of the curve obtained by plotting the permeation F (in the transition region) versus the average pressure, as discussed in Section 9.2.3.2. For multi-component mixtures the flux ]i as described by (9.33) can only be obtained in implicit form. For binary mixtures (9.33) can be solved directly in explicit form.
- Comparison of DGM and extended Fick models, some data: A comparison of DGM and the extended Fick model for the transition region has been made by Veldsink et al. [46] and is illustrated by many transport data and applied to describe transport in a macro-porous membrane reactor. Their main conclusion is that for ternary mixtures the use of the DGM model is necessary and predicts the transport of a gas mixture within a few percent (5%). For binary gases usually the extended Fick model is sufficient, but with an overall pressure over the membrane the accuracy is less than that obtained by use of the DGM. A further discussion will be given in Section 9.7.
9 ~ TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
361
TABLE 9.3 Relative importance of molecular (continuum), Knudsen diffusion and Poiseuille flow for air at 20~ in a straight cylindrical pore (after Karger and Rutven [3]) Dpoiseuille
p (atm)
Dm (cm2/s)
r (cm)
Dk (cm2/s)
D (cm2/s)
DPoisuille (cm2/s)
Dtotal (cm2/s)
1.0
0.2
10-6 10-5 10-4
0.03 0.3 3.0
0.027 0.121 0.19
0.0007 0.07 7.0
0.027 0.19 7.2
10 -6
0.03
0.012
0.007
0.019
0.37
10
0.02
10 -5
0.3
0.019
0.7
0.719
0.97
10-4
3.0
0.020
70
70
1.0
Dtotal
0.026 0.37 0.97
The relative importance of different transport contributions in a porous structure is given in Table 9.3 which shows that the contribution of Poiseulle (viscous) flow becomes important in larger pores (range 0.1-0.3 ~tm). At high pressure (10 bar) the Poiseuille flow is already important in pores with a radius of 10 nm. - The extended P - D model:
Present and De Bethune [48] were the first to develop a model (P-D model) including diffusion, intermolecular m o m e n t u m transfer and viscous flow. Based on the P-D model, Eickmann and Werner [18] incorporated two parameters (n k and [5) i n t h e P-D equations to account for geometric and reflection characteristics of a real membrane. This extended P-D model is very successful to describe the effect of a variety of parameters on permeation and separation [18] and will also be used in Section 9.3. Note that surface diffusion is not incorporated in the model. The flux of component i in a binary mixture is given by: Ji
g'[ = L
~176 d ( x . P ) ~f~ dP , dP] 1 + B'--------P d-----~-+ 1 + B ' P d z + x A P--~z
(9.34)
with the mol fractions for components 1 and 2 (i = 1 or 2) given by x and l-x, respectively. The terms in (9.34) describe the Knudsen diffusion (1st term), m o m e n t u m transfer (2nd term) and viscous flow, respectively. The different coefficients in (9.34) are described below:
8r[ g = -~
~ 2kTM
(9.34a)
362
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
g ' = nkr I r 2 ~g
(9.34b)
Mathematically nk[-] accounts for the porosity (r and the tortuosity (z) for gas permeation dominated by Knudsen diffusion (see Eq. (9.16)). ~[-] is used to correct for behaviour deviating from the ideal Knudsen behaviour, e.g., due to reflection conditions deviating from elastic specular collisions with the pore wall. 3r
A=~ 16rlv2
(9.34c)
and
A
A' = --
(9.34d)
B=8r qrckT q Mx 1 3---ff"- - -2--M " M1 + M2 PD12
(9.34e)
B' = B. [3
(9.34f)
= ~ / M 1 + (1 - x ) ~4M 2
MIM2
M+=
(9.34g) (9.34h)
M 1+ M 2
(9.34i) with M2 > M1
o0: - x/f 8kT /
(9.34j)
34k
D12 in (9.34e) can be calculated from (9.32) or directly measured. Equation 9.34 is used by Eichmann and by Wu et al [19] to study separation in porous media and this will be discussed in Section 9.3. Wu et al. [19] used (9.34) for single gas permeation (see Fig. 9.4a,b) to obtain values of n k and ~ in an asymmetric membralox membrane consisting of a top layer of T-A1203 (thickness 3-4 gm, pore radius r = 4(-7.8) nm) supported on an o~-A1203 support.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
363
Gases studied were He, N 2, H 2 and CO in the temperature range 20-815~ and pressure range 3-38 bar. The pore diameter calculated from the measurements is not the same for all gases. The same holds for the values of n k which vary from 2.43x10 -12 m -2 (H2) to 4.3x10 -16 m -2 (N2). The values of [3 (representing reflection conditions of molecules after colliding with the pore wall) decrease with increasing temperature for all four gases and strongly different values are found for the different gases. Especially the ~ values formed for CO are much lower: 0.27 (20~ and 0.06 (T = 538~ compared with that for N 2 w i t h ~ = 0.40 (20~ and 0.024 (815~ respectively. This unexpected behaviour of CO may be attributed to the interaction of CO with the aluminium oxide surface. The small value of [~ explains the much lower permeation of CO compared with the theoretical Knudsen diffusion in the membrane (for the other gases there is a good agreement) (see Fig. 9.4a,b). It should be noted that surface diffusion of CO is possible, but is probably negligible because the permeation is decreased (with respect to expectations based on Knudsen diffusion) instead of increased (if surface diffusion is important).
- Determination of effective diffusion coefficients: The effective diffusion coefficient, and so the permeation of a component in a mixture, can be determined with the so-called Wicke-Callenbach cell [7]. The cell has a similar design to that given in Fig. 9.11 but in this Wicke-Callenbach type of measurement there is no total pressure difference across the membrane (isobaric). The feed is in this case gas a, the permeate in Fig. 9.11 is replaced by an incoming flow of gas b (countercurrent configuration). Gases a and b diffuse through the membrane (counter diffusion) with fluxes Ja and Jb, and so the retentate (Fig. 9.11) is now a flow of gases a+db, the outgoing stream ('sweep' in Fig. 9.11) is b+da. In the measurement of D a, the volume flow ~v,d of the gas mixture b+da in the bottom compartment (d) and the concentration Ca,e in Qba,d are measured; this gives the mol fraction Ja,e. In the equilibrium state using a mass balance over the cell and using the DGM expression for a binary gas (under isobaric conditions) it can be described that [49] 2r
P d ' Ya,d " Tcell
" a - "- (I)v'd Pcell 9Ya,cell " T d = D e ' a
1 - 2ya, d Ya, cell
AP + K e ' a Pcell
(9.35)
This equation takes into account that usually P, y and T are measured not in the cell but at a different site in the measuring equipment. A plot of the left-hand side of Eq. (9.35) v e r s u s AP/Pcell yields the effective flow factor Ke~ from the slope of the curve. The value of De,a c a n be calculated from the intersection of the curve at AP/Pcell = 0 because the mol fraction Ya,d is known.
364
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
9.3 S E P A R A T I O N OF BINARY MIXTURES IN SIMPLE M E S O P O R O U S MEMBRANES
9.3.1 Important Concepts The separation of gas mixtures in practice can be performed in a variety of modes e.g. counter- or concurrent flows and cross flow (dead end mode) with different conditions concerning the variation of pressure and concentration on the feed and permeate chambers and along the membrane surface. Examples are discussed by e.g. Eichmann and Wemer [18]. The most simple experimental set-up, suitable to define some important parameters, is given in Fig. 9.11. Assumptions made here are well mixed flow on both permeate and retentate streams (this means constant concentration), no pressure drop throughout the permeate and retentate sides respectively and ideal gas behaviour. These assumptions hold usually in small modules with not too large membrane permeation. In large modules (long tubes, capillaries or large plates) with high membrane fluxes other conditions prevail. This will be discussed in Section 9.5. The parameter to describe the separation efficiency for a binary mixture is the separation factor (xwhich is a measure of the enrichment of a gas component after it has passed the membrane. (x=
y .1-x 1-y x
(9.36)
with x and y the mol fractions of feed and permeate respectively. For a given mixture, 0~is influenced by the membrane and the process specific parameters. In mesoporous membranes the most effective separation mechanism outside the capillary condensation region is Knudsen diffusion. In this case the ideal separation factor 0~*equals the square root of the ratio of masses: 0~* = ~ M 2 / M 1 with
M 2 >
M1
(9.37)
In general c~* is not equal to ~ due to back diffusion, caused by non-zero pressure at the permeate side, or to contributions of non-separative mechanisms to the total flow and concentration polarisation on feed or the permeate side. Also the presence of surface diffusion influences the ideal separation factor. Back diffusion due to a non-zero value of the pressure at the permeate side is a very general phenomenon to decrease the value of (x. The permeant gases at the permeate side of the membrane are removed by pumping or by a sweep gas. In the last case the total pressure is usually relatively large, but the partial pressure of the permeant is low. Using a sweep gas makes the mixture effectively a ternary system and ignoring the effect of the sweep gas (as is frequently done) is not always allowable as will be discussed in Sections 9.4 and 9.5.
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
365
If the pressure at the downstream (permeate) side is in the transition or continuum regime and is not negligible, there is a back-diffusional flux into the membrane decreasing the value of c~. Equation 9.38 gives the effect of back diffusion on the actual separation factor [23,24]. a = 1+
(1 - Pr) (a* - 1) 1 + Pr(1 - y )
(9.38)
(0~*- 1)
where Pr is the ratio of permeate pressure divided by the feed pressure. It is obvious from (9.38) that the permeate pressure directly after the separation layer should be kept low. This is in principle possible with single wall, symmetric membranes. With asymmetric (supported) membranes the support represents always a certain flow resistance and this means that the actual, or partial, pressure of the interface between separation layer and support is larger than the pressure at the permeate side of the support. This implies that the flow resistance of the support should be as small as possible to minimise back-diffusional effects. The separation factor a as determined from gas mixtures is generally not the same as the permselectivity which is defined as the ratio of the permeation of the single gases at a given membrane thickness. They are similar only when all interactions between the different phases and between gases and the pore wall can be neglected, e.g., in the Knudsen region and at high temperature (surface diffusion negligible).
9.3.2. Separation in the Knudsen and Transition Regions As discussed above, the ideal separation factor (x" in the case of pure Knudsen diffusion is given by Eq. (9.37) and is equal to the permselectivity provided that surface diffusion is not present (high temperature). As can be seen from (9.37) the highest ideal separation factors are obtained for mixtures of light and heavy gases. Back-diffusion effects are taken into account by Eq. (9.38) to give the real separation factor. The support can have a considerable influence on the separation factor of the membrane consisting of separation layer and support when its flow resistance is not negligible and the gases in the support pores are in the transition or viscous flow regime [20]. This point will be discussed in Section 9.5. In the transition region intermolecular momentum transfer decreases the separation factor considerably. The effects of the pressure ratio Pr, with feed pressure as a parameter of temperature of pore size and of concentration, are analysed by Wu et al. [18] and by Eichmann and Werner [19]. Wu et al. used Eq. (9.34) to simulate the permeate composition and separation factor for H2/N2, H2/Co and H e / O 2 gas mixtures and compared them with experimental results obtained on a Membralox asymmetric membrane system,
366
9 w TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
4.00
II
3.S0.
1
3.00
2
~
_ I_
]
II
..
".'~_-~
I I
II
I
II ILl
Ilill
lab
I
I
In
_
;
I
Ill
.
i
-__] . . . .
Ideal
|
I
I
I
I
sepmidon
2.S0 3.O0
8. I . s o m
1.00
0.S0
I
. . . . .
m
m
i
m
m
m
-
m
m
m
m
nomuaaon -
0.( lO
9" " '
I ""'" "
0.20
9 I 0.40
9 i
,
" I
'
0.60
I I I
I
0.80
I'l
,
m
J m
9 9 I 1.00
Pressure Ratio, Pr Fig. 9.13. Feed pressure effect on separation of H2-N2 mixtures at T = 538~ feed H2X0 = 0.5, stage cut = 0.01, p o r e diameter 5.6 run. Feed pressures (1) I atm; (2) 7 atm; and (3) 34 atm. After Wu et al. [19].
whose characteristics are described in Section 9.2.4.3. Correction for the support resistance was not applied. The simulation is generally in good agreement with the experimental results generated for a wide range of operation conditions (20-815~ P = 1-34 atm, P~ = 0.1-0.8, stage cut 0.01-0.36). Deviations between predicted and experimental mole fractions are within 10%, with a consistent overestimate of the light component in the permeate. The effect of the pressure ratio Pr of permeate and feed and feed pressure on the c~ value of a H2/N2 membrane is given in Fig. 9.13 for T = 538~ and feed pressures ranging from 1-34 arm using a separation layer of 5.6 nm. As is shown in Fig. 9.13 for a given pressure ratio, the higher the feed pressure, the lower the separation factor. At all pressures (1-34 atm) the separation factor decreases continuously with P~ (0.10--0.70). At P~ = 0.70 allseparation factors converge to a value of 1.5. Note that even at the lowest pressure (1 atm) and lowest value of P~ = 0.10 the value of (~ = 3.20 which is considerably smaller than the ideal value (~* = 3.70) as given by Eq. (9.37). So even a small amount of non-Knudsen contribution to the total flow in a pore considerably influences the separation. At higher temperature the separation factor increases because the mean free path increases and consequently less momentum loss is expected for H2. The effect is stronger at lower Pr value, and at Pr = 0.10 and P = 7 arm the values of
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
367
for H 2 / N 2 are 2.50 and 3.00 at 20~ and 815~ respectively for the same membrane as used in Fig. 9.13. As the H 2 concentration in the discussed H 2 / N 2 mixture decreases the separation factor also decreases under selected operating conditions. When the partial pressure (concentration) of H2 decreases the number of H 2 to N2 collisions increase relative to that of the H 2 to H 2 collisions and consequently more H 2 momentum is lost at low H 2 concentration and the separation efficiency decreases. The effect is weak however compared to that of pressure and temperature, because collisions with the pore wall are much more frequent compared with intermolecular collisions. Finally the 'stage cut' Sc = Qp/Qb(Qp,f = feed and permeate flow respectively, see Fig. 9.11) is important. At high stage cut the driving force for gas separation in terms of a partial pressure difference is reduced to maintain the material balance. At low Pr the effect of Sc is largest and the lower the value of Sc the larger the separation. For H 2 / N 2 and the conditions given for Fig. 9.13 with Pr = 0.01 and P = 7 atm the values of ~ are 1.90 and 2.90 at Sc = 0.4 and 0.01 respectively (ideal separation is 3.70). The effect of the pressure, temperature and pore radius on the separation factor is investigated also by Eichmann and Werner [19] using Eq .(9.34) with a constant and experimentally determined value of ~ for all gas membrane combinations, in contrast to Wu et al. Who fitted the value of ~ for each gas membrane combination. Figure 9.14 shows the effect of the pressure ratio Pr for different mean pressure levels P (assuming a linear pressure drop in the membrane) on the separation factor of a N 2 / C O 2 mixture (ideal separation factor equals 1.25) in a membrane with pore radius Rp = 0.03 ~tm. In contrast to the situations given in Fig. 9.13, maxima can be seen which shift to larger Pr values with higher pressures. Similar curves are obtained for different pore radii as shown in Fig. 9.15, where the maxima become smaller and shift to larger Pr values with increasing pore radius. The maximum is caused by a viscous flow contribution in the relatively large pores (0.015-0.12 ~tm) considered here. In the rising part of the curves the (non-separative) viscous flow contribution decreases with increasing Pr (smaller pressure difference). The contribution of the viscous flow decreases with decreasing pore radius and with small enough pores the maximum vanishes and continuously decreasing curves are obtained which exhibit greatly reduced pressure dependency. This is shown in Fig. 9.16 with similar shapes to those of Fig. 9.13. The results of Fig. 9.16 are obtained on membranes of y-A1203 with a pore radius of 2.5 nm as prepared by Leenaars and Burggraaf [17b]. In conclusion it can be said that the key operating parameters to approach the ideal Knudsen separation factor (determined by mass ratio) in mesoporous membranes are: small pore diameters; low pressure ratio, adjusted to produce maximum separation; relatively low pressure level; and high process temperature.
368
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
1.15 =2x10s
1.10 C 0 0
1.05
Ix10s
1.00 0
0.2 0.4 0.6 0.8
1.0
PR : I~ I p . Fig. 9.14. Influence of pressure ratio Pr on the separation factor of N2/CO2 mixtures. Pore radius is 0.03 l~n. After Eickmannand Wemer [18]. To increase the separation factor above the ideal Knudsen separation factor requires contribution of surface diffusion a n d / o r capillary condensation or the presence of micropore systems.
9.3.3 Separation with Surface Diffusion and Capillary Condensation The permeation of gases in membranes due to surface diffusion and capillary condensation has been discussed in Section 9.2.3.3. together with some illustrative data. The total flux of a single gas is usually calculated as the sum of the flux by surface diffusion and the flux through the gas phase. As shown the surface flux can contribute considerably to the total flux (increased by factor 2-3 of gas diffusional flux), especially with smaller and uniform pore sizes (compare Eqs. (9.9a) and (9.15). With decreasing pore size the flux through the bulk gas decreases while the surface diffusional flux increases. With very small pore diameter (< 2-3 nm) the effective diameter for bulk gas transport is less than the geometric pore diameter due to the thickness of the absorbed layer which
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
I
,
9
I
I
I
.
.
.
.
.
.
.
369
.
1.15 i.
0
o
90 015 pm
0
C o
....
1.10
,.mlw
o
0 Q. o ul
0.03
1.0S
1.00
0
(106
0.20.t.
06 (18 1.0
P,,
Fig. 9.15.Influence of pore radius r onthe separation factor of N2/CO2mixture at a pressure of 2 bar. After Eickmann and Wemer [18]. decreases the space available for the gas phase. With gas mixtures this means that the bulk gas phase diffusion of a non absorbing molecule is decreased by absorption of an adsorbing molecular species in the mixture resulting in an increase of the separation factor. This is especially the case with lower temperatures of a few hundred degrees and intermediate pressures which give rise to partial blocking by capillary condensation. Some illustrative examples and special phenomena will be discussed below.
Separation by surface diffusion With gas mixtures, enhancement of the separation factor can be obtained by preferential sorption of mobile species of one of the components of the gas mixture. Adsorption does not always lead to enhanced separation. In a mixture of light non-adsorbing molecules and heavy molecules, the heavy molecules move slower than the lighter ones but in many cases are preferentially adsorbed. Consequently the flux of the heavier molecules is better enhanced by surface diffusion and the separation factor decreases. This occurs, e.g., in CH4/CO 2
370
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
I 1.20
6,2,~os./.z a,@ O
u 1.15
0
2
-O,
D 1.10-
0
1
a,10s
1.05 0
0.2 0.~, 06
0.8
1.0
Pn = p, I P, Fig. 9.16. Influence of pressure level on the separation factor of a N2/CO2 mixture. Pore radius is 2.5 nm. After Eickmann and Wemer [18].
mixtures in Vycor glass membranes. With two adsorbing molecular species, competition for the adsorption sites m a y exist and sorption isotherms for single gas species are no longer valid. Uhlhorn et al [28] reported for a H 2 / N 2 mixture a separation factor of about 9 compared to the Knudsen value of 3.74. As shown in Fig. 9.17 the ratio of the H 2 flux over that of the N 2 flux decreases from 9 at a pressure of 50 kPa to 5 at 200 kPa. This result is obtained on ~'-A1203 membranes (thickness 100 ~tm, pore diameter 2.5-4.0 nm) impregnated with 17 wt% (finely dispersed) Ag. The increase of the H 2 flux is obtained by the Ag impregnation. Probably the decrease of the separation factor is caused by a decreasing contribution of the surface diffusion to the total flux with increasing pressure due to saturation of the adsorption. Keizer et al [20] found a similar p h e n o m e n o n for C O 2 / N 2 separation (with C - O ~as 2 the fastest diffusing species) on non-modified ~'-A1203 m e m b r a n e s (0~ = 1.5-2.0 at 240 K, c( = 0.8 (Knudsen value) at 360 K, pressure I bar). In order to
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
371
12
Z 9
e~e--.~....e 9l i o r e ~ r A i
~..
0
.
..._
.
.
A
100
ratm
___1.,.~,,,,~...~,~.-,~.,,,~,,,,~
2O0
30O
Fig. 9.17.Experimental (o) and theoretical flux ratio of H2 and N2at 25~ on a nonsupported 7-Al203 layer modified with 17 wt% silver, measured in counter-diffusion configuration. After Uhlhom et al. [28]. enhance the surface contribution the 7-A103 membrane was modified with 2.2 wt% MgO [28,20]. The result was a decrease of the separation factor to 1.0 due to the formation of strongly bonded, immobile CO2 species, the total concentration of adsorbed CO2 remaining constant. As shown by Eq. (9.15) this results indeed in a lower CO2 surface flux.
Separation by multilayer diffusion and capillary condensation (see also Section 9.2.3.3) Brief overviews are given by Keizer et al. [50] and Sperry et al. [39] and these show that very high separation factors in combination with large permeation can be obtained in cases of mixtures of an easily condensable gas (vapour) and a difficult (non)-condensable gas which has a low solubility in the condensed phase. Pore blocking by capillary condensation takes place at 0.5--0.8 of the saturated vapour pressure (depending on pore size) and is preceded by multilayer diffu-
372
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
A
X a.
0
~ii 5
3o
A
-
,
.
.
.
.
.
.
.
prol:)ytene
.
.
.
.
.~
.
.
.
30
-'
.....
-
B
cleS~:~Dhon
2o .15
"
~ads~ption
Oesor
E C 0
m
0
E
10
s
r.trogen
0 0.00
(xs0
O
t .00
0.00
.
.
.
Rel. prestmm
.
.
.
.
0.50
1.00
R~a.~ m
ql"
r
20
9O
"E 0
E
Q.
.
.
.
'
.
.
D
.
im
o
~ 60
~.
cle~pt~n
J l
rutrooen (~
.
aclsorptton + c~mcxpt|~
to
m
E
'.
p, o o y J ~ / "
c 0
Q
-
C
~
o~oo
iior 30
, O
~L p ~
-
ooo
''
oa0
,~x~
Fig, 9.18. Permeation and separation factors of supported y-alumina thin film for nitrogen and propylene at 263 K (A,B). Propylene is the preferentially permeating component; dashed line gives the relative pressure at which the maximum in the permeation of Fig. 9.18a. occurs. (C) and (D) as (A) and (B) but for a supported film modified with MgO. After Uhlhom et al. [37].
sion and an increased flux of the condensable gas and an increased separation factor. Uhlhom et al. [37] reported separation factors of c~up to 27 for propene/N2 (60:40) mixtures at 263 K (with propene the fastest permeating species). Note that the Knudsen factor is 0.8 and the permselectivity (ratio of single gas phase fluxes) amounts 7.4. As shown in Fig. 9.18, the region with the highest separation factors coincides with the maxima in the permeation curves which in turn are determined by the blocking of pores by adsorbate (capillary condensation). The permeation of propylene at the maximum amounts 30x10 -6 m o l / m 2 s Pa. A further improvement of the separation factor is obtained by modification of the ~/-A1203 membrane with the reservoir method [51]. The membrane pores are filled up to 85% of the pore volume with MgO. This process enhances the value of 0c to 85 with a corresponding decrease of both propylene and N2 permeation values to 15x10 -6 m o l / m 2 s Pa for propene (equivalent to 300 N m 3 / m 2 day bar.
9 ~ T R A N S P O R T A N D S E P A R A T I O N PROPERTIES OF M E M B R A N E S W I T H GASES A N D V A P O U R S
373
The shape of the curves (steeper, shift of desorption branch to lower relative pressures) indicates a narrower pore size distribution with smaller average pore size (below 3 nm) and less defects for the modified membrane. Sperry et al. [39] reported capillary condensation up to 473 K in methanol/H2 mixtures for certain pressure ranges. They used a similar type of membrane as used by Uhlhorn, but treated with NaOH to poison the surface for chemical dehydration reactions. Using the Wicke-Callenbach method (no absolute pressure drop) the highest value of o~ equals 680 (methanol being the faster permeating species) and is obtained at 373 K and 2.2 bar methanol pressure, with a methanol permeability of 51x10 -6 cm3(STP) c m / c m 2 s cmHg. At higher temperatures the maximum obtainable values of both {x and permeability decrease and {x = 110 (with methanol permeability is 4.2x10 -6 m o l / m 2 s Pa) at 473 K. (Note: 1 cmB(STP) c m / c m 2 s cmHg is equivalent to 3.12x10 -6 mol m / m 2 s Pa). Capillary condensation takes place at Pr = 0.60. This is considerably lower than predicted by the Kelvin equation (9.18) for pores with a diameter of 4 nm. Separations with a pressure drop must be carried out with pressure drops smaller than 0.25-0.28 at T < 448 K or 0.05 bar at 473 K due to blow-out of the condensate under these conditions. The observed flow rates in the capillarycondensation regime are larger than those obtained for Knudsen diffusion at lower pressures. Together with the results reported by Sperry e t al., the conclusion is that separation by capillary condensation yields a combination of large separation factor and high permeation even at increased temperature provided the appropriate temperaturepressure, pore size combination is chosen. A disadvantage is the sensitivity of the process for pressure changes (blow-out phenomena). Finally, Asaeda and co-workers [52,53,64] reported separation results using membranes which are modified in such a way that pore sizes below the mesopore range (<2 nm) are obtained; no definitive pore characteristics are given however. A type of pore blocking by one of the components occurs but whether this is capillary condensation is not certain. Asaeda and Du [38] reported values up to o~ > 100 for water-light-alcohol mixtures at 70-90~ in alumina-silica membranes. The water permeability is dependent on its concentration in the mixture. At atmospheric pressure and 20% water a typical water permeation value is 1.3X10 -2 m -2 s -1 (= 20 1 H 2 0 (liquid) m -2 day-l). Azeotropic points can be bypassed in this way with an alcohol concentration much higher than the azeotropic concentration. Similar results are given for mixtures of water and organic acids (acetic, propionic, acrylic) by Kitao and Asaeda [52] for rather thick (10 ~tm) silica membranes supported by 7-A1203 and made in a multi step process (up to 15 layers on top of each other). A permeation mechanism and a model for the pore
374
9 ~ TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D V A P O U R S
structure is proposed by Kitao et al [53]. The pore shape is assumed to be conical, changing from rather wide on the support side to very small at the surface. Here the 'neck' diameter is suggested to be 0.4 nm. Equations for the (preferentially) permeating water flux are given. Near the surface an additional resistance to the flow builds up due to osmotic effects caused by rejection of organic molecules at the pore entrance. 9.4 PERMEATION A N D SEPARATION IN MICROPOROUS MEMBRANES
9.4.1 Introduction and Important Concepts Existing ceramic, mesoporous membranes (with a 4 nm pore diameter) perform most gas separations according to Knudsen diffusion. The obtainable separation factors (Section 9.3.2.) are usually not economical for most gas separations and provide incremental but limited conversion enhancement in catalytic membrane reactor applications. Capillary condensation and preceding surface flow yield economically interesting separation factors but this mechanism is limited to easily condensable gases and is limited to rather low pressure drops due to stability problems (Sections 9.2.3. and 9.3.3.). To enhance the separation factor the average pore diameter should be decreased considerably. According to Eqs. (9.9a) and (9.15) the contribution to the total gas flux of the gas (Knudsen) diffusion decreases and at the same time that of surface flow (diffusion) increases with decreasing pore radius. In recent years modification of existing membranes for improving their separation efficiency has been actively pursued especially by attempts to decrease the pore size of membranes. This resulted in different types of microporous membranes. According to IUPAC convention these are porous systems with a pore diameter below 2 nm. In the literature the name 'microporous' is frequently misused and this should be avoided. An overview of microporous membrane types is given in Table 9.4. The oldest microporous membranes are based on carbon and are reported by Koresh and Softer in a series of papers from 1980 to 1987 (see overviews in Refs. [6,42]). They are made by pyrolysis of a suitable polymer (hollow fibre) as reviewed by Burggraaf and Keizer [9]. More recently Rao and Sircar [42] developed a new technique. A macroporous graphite sheet was coated with a suitable polymer (latex) which was pyrolysed subsequently. This process was repeated 4-5 times and resulted in a total carbon layer thickness of 2.5 ~tm with an average pore diameter between 0.5 and 0.6 nm. The membrane has interesting properties (see Section 9.4.3). Finally, very recently Linkov and Sanderson et al. [55] modified and improved the method reported by Koresh and Softer and produced flat sheets as well as hollow-fibre systems.
9m TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
375
TABLE9.4 Microporous membrane types Type .
2. 3. 3.1 3.2 4.
Ref. Carbon hollow fibre, film on (C) support Porous silica glass (Vycor) Amorphous silica based systems Sol-gel techniques C.V.D. Zeolite films on supports (alumina, steel)
42,54,55 56 21,57-63,64 65-68 69-79
Mesoporous glass (Vycor type) can be produced by a combined heat-treatment and leaching procedure [9]. Modification of this process can lead to microporous hollow-fibre systems with interesting properties as discussed by Shelekhin, Ma et al [56]. For further discussion see Sections 9.4.2 to 9.4.4. The most promising results from the viewpoint of a combination of large separation factors and reasonable-to-large flux values are reported for supported silica based systems. Burggraaf and co-workers reported in a series of publications [21,57,63] the sol-gel, two-step synthesis of silica and silica-titania films supported by a composite membrane of mesoporous 7-A1203 and macro-porous (x-al203. The film has a thickness of 50-100 nm and is situated for about 50% within the mesopores of the y-A1203 and for the rest on top of it. The pore diameters are around 0.5 nm. A combination of large separation factors and large fluxes was reported for several gas combinations [60,61] (See sections 9.4.2-9.4.4). As described by de Lange et al., the precursor sol consist of a polymeric silica solution with low fractal dimension [59,62] and the support quality (roughness) is important to obtain defect-free membranes [59,60]. Asaeda et al. [64] produced a microporous film directly in a macroporous c~-A1203 support with a 15-step coating process starting with colloidal silica solution and ending with a polymeric silica solution. This means that a mesoporous intermediate silica layer was first produced. The final top layer was said to have a pore diameter < I nm. The system shows very interesting (isomer) separation properties (see Sections 9.4.3-9.4.4). In a series of papers (1989-1994) Gavalas and co-workers reported the synthesis of silica films in porous Vycor glass substrates with chemical vapour deposition (CVD) techniques [65]. A similar technique was used by Heung et al. [66]. The separation factors reported by Gavalas and by Heung are very high but the fluxes are low. In fact the silica layers are non-porous (no interconnected pore network). Wu et al. [67,68] improved the method used by Gavalas using a
376
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
composite support of 0~-A1203with a 3-5 ~tm thick 7-A1203 in the top region of the 0~-A1203.A 1.5-3 ~tm thick silica film was deposited in the 4 nm pores of the 7-A1203. The smallest obtained effective pore diameter in the silica plugs was estimated to be -- 0.5 nm. A combination of large separation factors and reasonable fluxes was reported for H2/N 2 and H2-isobutane mixtures (see Section 9.4.3). Zeolite membranes form the most recent branch of the inorganic membrane field. It is only very recently that well characterised and properly described real microporous zeolite membranes have been reported [69,72-78,88,89]. Geus et al. [69,70] and Bakker et al. [70] described the synthesis of 50 ~tm thick silicalite (MFI) membranes on porous stainless steel supports; Vroon et al. synthesised 3 ~tm thick silicalite membranes on o~-A1203supports [72-74]. These membranes consist of very small crystals (100-200 nm). Jia and Noble and co-workers et al. reported a 10 ~tm thick silicalite membrane on a composite support of c~-A1203 [27,77]. Finally, Xiang and Ma [76] partially filled the pores of a microporous (~-alumina support with ZSM5 crystals. All the authors used an in situ hydrothermal crystallisation method to grow directly polycrystalline zeolite layers. The layers reported by Jia et al. and by Xiang and Ma contain a relatively large number of defects, in contrast to that of Geus/Bakker and Vroon, but nevertheless show interesting separation and flux properties provided that good condensable gases are present (e.g. methanol, xylenes). The microstructure of the layers plays an important role as shown by Vroon et al. [72,74] as well as does the support (compare clay with stainless steel) as shown by Geus et al. [69,75]. Examples of properties will be discussed in Section 9.4.3. Zeolite membranes on porous support with good to reasonable quality has been reported so far only for silicalite and (related) ZSM5 systems. In the literature since 1985 a number of other systems are reported including a series of patents. They are reviewed by Geus [69] and Vroon [72] and briefly by Matsukata et al. [78] and Burggraaf [79]. This older literature concerns either membrane systems which are not real (but very defective) membrane systems but sometimes have interesting properties for membrane reactors or concern single crystal work or very fragile non-supported membranes on which important fundamental studies have been performed. In the first category belong the pioneering work of Suzuki (patents 1985, 1987) and of I.M. Lachmann (patent 1989) yielding N a A / C a A and X or Y or mordenite zeolites. Unsupported ZSM5 layers were prepared by Haag and Tsikoyannis (1992) and Sano (1991/1992). Work on single crystals of NaX and silicalite were reported by Wernick and Osterhuber (1985) and Geus [69] respectively. For literature references see cited overview papers. It is not the place here to treat structural characteristics of zeolites. Nevertheless a very brief summary with a focus on silicalite/ZSM5 systems is necessary
9 m TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
377
as a background for permeation/separation studies in Section 9.4.3. For details the reader is referred to books of e.g. Breck [80], Meier [81] and van Bekkum et al. [82]. Zeolites can be represented by the empirical formula [80]
M 2/nO"al 203"xSi O2"yH 20 in which n is the cation (M) valence, x/2 is the Si/A1 ratio (equal to or larger than two). The cations M are present to balance the negative charge introduced in the crystalline framework by the substitution of Si 4+ by A13+. These cations can be exchanged (in exchange reactions). The aluminium-rich zeolites are hydrophilic (high affinity for water), the silica-rich zeolites are hydrophobic (small affinity for water) a n d / o r organophilic. Also the thermal stability increases with increasing Si/A1 ratio. The crystalline framework consists of a three-dimensional network of SiO 4 and A104 tetrahedra, linked to eachother by sharing the oxygen atoms. The framework structures contain channels of voids interconnected by ring openings. These channels can be isolated from each other (one-dimensional) or are interconnected by ring openings and form two or three-dimensional network structures. More then 85 different framework structures are known [81]. Silica-rich zeolites are ZSM5 with a Si/A1 ratio of 11/1000 and silicalite (Si/A1 > 1000). Both have a similar structure (i.e. MFI type) but ZSM5 contains some cations and is more hydrophilic. The structure of MFI-type zeolites is given in Fig. 9.19. The structure has two sets of intersecting channels (10-membered oxygen rings, see Fig. 9.19b), one set consisting of straight channels with ring openings of 0.52x0.57 nm, the other set consists of sinusoidal channels of 0.53x0.56 nm (Fig. 9.19a). At the intersection points cavities are formed with a size of about 0.9 nm. The lattice of ZSM5 is stable up to 1175 K; that of silicalite to a somewhat higher temperature. Both zeolites have a good stability in strongly acidic environments, are relatively easy to prepare and have a low affinity for water, which is important for (gas) separation properties. In recent years zeolites with very large pores (supercages) and ring openings up to 0.6xl.32 nm (cloverite) have been synthesised.
9.4.2 Phenomenological Description of Single Gas Permeation The theory of transport in microporous solids is complex and involves many aspects and steps. Although many aspects has been treated separately (e.g., sorption, diffusion, simulation studies, mechanisms, etc.) there are no coherent descriptions of permeation and separation in microporous membranes covering all the important aspects. In this chapter an attempt is made to introduce such a description. It is useful to give a qualitative picture first (Section 9.4.2.1).
378
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
(.) (b) Fig. 9.19. Schematic picture of zeolite MFI structure. This will show that a quantitative description involving all the complexities in simple microporous membranes is not available (if possible). However a number of boundary cases can be described quantitatively, as in Section 9.4.2.2, and trends in more complex situations can be predicted in combination with the qualitative pictures based on mechanistic considerations.
9.4.2.1 Qualitative Description of Gas Permeation As discussed in Section 9.4.1, the contribution of Knudsen diffusion to the total flux decreases with decreasing pore radius of the membrane material. Initially the selectivity of binary mixtures of gases is constant and equal to the Knudsen value. Lin et al. [67] reported in the region between pore diameters of 3.0-2.0 nm small negative deviations for H e / N 2 mixtures, but with pore diameters < 2.0 nm a strong increase occurs to values above the Knudsen value. This is a typical phenomenon for microporous systems together with the onset of activated gas permeation. As will be shown, it is useful to distinguish microporous membranes in systems with relatively large, intermediate and small pores. This is discussed by de Lange and Burggraaf et al. [59,63] and is schematically shown in Fig. 9.20. Note that here the location of the minima and the shape of the potential as a function of z is given schematically and is not exact. Simulation results yielding pictures as given for region c2 are reported by Petropoulos and Petrou [83]. For mesopores the minimum in the potential curves is equal to the (isosteric) adsorption heat at 'free' surfaces with respect
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
deft
, B
,........,
1!1 a
r
379
nt bl
b2
- si c1
'i c2
2-
0
Z
A
II <--2 N 4-
r/OA.-
0.9
1.086 1.239
2
3
Fig. 9.20. The relative potential U = U(Z)/UA* of molecule A as a function of the distance Z from the pore centre for several values of the relative (dimensionless) pore radius r~r~Afor three different regions (a-c). UA*is the absolute value of the (Lennard Jones) potential on a free surface, OAthe molecular diameter. After de Lange et al. [59,63].
to the ambient atmosphere (horizontal line in Fig. 9.20, the effects of curvature are neglected here). In the central region of the pore the gas molecules move freely (Knudsen). At the b o u n d a r y of meso- and micropore region the potential fields of both walls start to overlap and the potential curve shows two minima separated by a m a x i m u m (region c2). This m a x i m u m decreases (larger negative values) with decreasing pore diameter. This means that molecules are adsorbed more strongly. For relatively wide pores the enhancement of the sorpfion energy with respect to a free surface is significant and is calculated by Everett et al. to be 30-50% for R/r~ = 3 and cylindrical pores [84]. The m a x i m u m e n h a n c e m e n t is by a factor of 3.39 at R/r~ = 1.086 (region bl). With decreasing pore size the desorpfion energy from the wall to the gas phase within the pores (the m a x i m u m in the curves in Fig. 9.20) becomes smaller but remains positive. This implies that the molecules in the central part of the pore behave in a Knudsen-like m a n n e r (i.e. no intermolecular collision) and can pass each other (region q, upper part of Fig. 9.20) but nevertheless are not 'free' and follow curved trajectories (see Ref. [83] and Sections 9.4.3.1-2). In this region c we can speak of a surface flow enhanced micropore difhasion (SEMP). Because in surface diffusion the activation energy is a fraction of the adsorption heat (see
380
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
Section 9.2.3.3) this implies that the transport in the SEMP regime is activated. With further decreasing pore size, a single minimum is finally obtained at R = 1.24 G (region b2). When the effective pore diameter deft < (~A+ (~B(region bl in Fig. 9.20) the molecules in cylindrical pores cannot pass each other (except e.g. at wider intersections or channels) and the molecules interact strongly during diffusion. Finally in region a the pore diameter is about equal to the molecular diameter and molecules have increasing difficulties to enter the pore (less negative sorption heat) and relatively large molecules cannot enter the pore at all. This is the size exclusion region with mixtures. In regions b and c we cannot speak any longer of 'bulk' gas phase in the pore and the diffusion shows strong similarities with solid state diffusion (solutiondiffusion models). From the discussion so far it follows that sorption as well as diffusion play a role. Their relative importance depends on such sorbent material characteristics as pore size, sorption strength, and gas properties such as molecular size and shape, concentration, etc. A theoretical model describing gas transport in microporous (single crystal zeolite) membranes in regions a and bl of Fig. 9.20 was proposed by Barrer [85] and is schematically given in Fig. 9.21. Essentially the flux J consist of two parallel and additive components F1 and F 2. The flux F1.J c o m e s directly from the gas phase to the first sites 01 in the micropore via the pore entrance. The flux F2"Jconsist of two parallel fractions f2"J andfl-J, each consisting of several sequential steps, the first being an adsorption step. Infl.J adsorption takes place at sector 00 at the external surface near the pore entrance. Jumps from 00 -~ 01 may require the passage of an energy barrier. The flux f2"J involves first adsorption at the external surface at sites 00~urf followed by surface diffusion from sites 00~urftOsites 00 and subsequently from 00 ~ 01. The sites 01 are occupied to a degree (concentration) depending on the sorpfion isotherm. In all cases the steps at the surface are followed by micropore diffusion in the pore channels from sites 01 t o the other side of the membrane. Here desorpfion takes place directly to the gas phase or via desorpfion from the pore to the external surface and than to the gas phase. Equations for the several steps and the total flux have been derived and will be discussed in Section 9.4.2.2. Some important conclusions can be drawn however from this qualitative picture: (i) Direct entrance from the gas phase (FI-J) is important for relatively small, spherically shaped molecules and will then dominate the flux for weakly adsorbing molecules or at high temperature. For larger, branched molecules (e.g. hydrocarbons) direct entrance is unlikely and an adsorption step at the external surface is necessary. (ii) The adsorption step is important at lower temperature (increasing concentration). With increasing temperature the diffusion rate becomes more im-
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
381
GAS PHASE f2.J
fl .J
F1 .J
A
A~
1 +F2"19
6n.,,rt
PORE Fig. 9.21.Schematicmodelof gas permeationin microporousmembranes.FluxF consistof fractions F1(direct from gas phase to site 01)and F2 (=fl +f2) (via adsorption at the external surface). portant, while the concentration decreases (adsorption isotherm). So maxima in the flux can be expected at a certain temperature given the pressure. (iii) With mixtures (gases A and B) competitive adsorption can take place and a strongly adsorbing gas can exclude a weakly or non-adsorbing one from being sorbed. This will severely affect the flux component F 2 and may enhance the selectivity (sorption selectivity) depending on the differences in mobility of A and B (diffusion selectivity). (iv) In the size exclusion range adsorption of a large, non-penetrating molecule at the external surface (at sites O0) can block or strongly hinder the flux of the smaller, penetrating component. This decreases flux and selectivity. (v) At larger concentrations (occupancy degree) within the pores of components A and B, strong interactions will take place and in regions a and bl of Fig. 9.20 selectivity by mobility differences vanish and only selectivity by sorption remains. As will be shown later all these phenomena have been observed. The consequence is that the permeation behaviour in mixtures depends strongly on the character Of the mixture and it is necessary to distinguish several categories in terms of combinations of weakly (W) and strongly (S) adsorbing gases as shown in Table 9.5 [72,74]. The quantitative description of permeation and separation in terms of operational equations is today only reported in a few limiting cases mainly characterised by relatively low concentrations (Henry regime, initial part of Langmuir type regime) or by size exclusion at higher temperature where adsorption at the external surface is not important and we have essentially permeation of a single gas species. Finally should be stressed that firm conclusions on the magnitude of permeation and separation factors are only possible after appropriate control of the defect level of the membranes (non- micropore/Knudsen contributions to the
382
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
TABLE 9.5 Overview of categories of (binary) gas mixtures in terms of separation regimes as a function of relative concentrations and mobilities Regime
Code
Concentration on the external surface
Concentration in the zeolite pore
Mobility in the zeolite pore
I
W-W W-S W-S S-S S-S S-S (SE)
A:low;B:low A:low; B:high A:low; B:high A:high; B:high A:high; B:high A:high;B:high A:low; B:low A:high; B:high
A:low; B:low A:low; B:high A:low; B:high A:high; B:high A:high; B:high A:high; B:low* A:low; B:low* A:low; B:low*
A
IIa IIb IIIa IIIb IV V VI
AB A=B A > B or B > A A > B, A >> B A>B A>B
w and S are weakly or strongly adsorbing components respectively of mixtures: W-W, etc. SE is size exclusion. *B is large molecule.
flow). The best w a y to p e r f o r m this is d e t e r m i n a t i o n of p e r m e a t i o n values and ~ separation factors in binary mixtures of gases consisting of small, w e a k l y a d s o r b i n g and very large molecules at high temperature. This is the size exclusion regime u n d e r exclusion of strong a d s o r p t i o n on the external surface of the larger molecule.
9.4.2.2 Quantitative Description of Gas Permeation and Separation Single gas permeation The equations given below are derived for single wall or u n s u p p o r t e d m e m b r a n e s u n d e r similar conditions to those given in Section 9.2.4 a n d Fig. 9.11. These are h o m o g e n e o u s and u n i f o r m concentrations (well mixed) a n d pressures on the feed and p e r m e a t e sides of the m e m b r a n e and near equilibr i u m b e t w e e n concentrations in the bulk gas phase and in the m e m b r a n e surface. As discussed in Section 9.4.2.1, small and large micropores s h o u l d be distinguished. This treatment will start w i t h a general description w h i c h is applied to small micropores. Subsequently the consequences for larger micropores will be treated. U n d e r isothermal conditions it follows from irreversible t h e r m o d y n a m i c s [1-3] for the flux Ji in a mixture of k components:
Ji = - ~_~ Lik V ~tk k
(9.39)
9 n TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
383
_
So the real driving force is the sum of the gradients of the chemical potentials as is also implicit in the general Maxwell-Stefan formulation [87-89]. For a single gas this reduces to: 3 In P dqi
J i - - ~tDo,i 0 ha qi dz
(9.40a)
Here the term 31nP/31nq is the so-called thermodynamic factor (hereafter called F), Do(q) is the corrected or intrinsic diffusion constant and Ix is a correction term (see notes below). With q = qsat'0, Eq. (9.40a) becomes:
Ji--
~tqsat, i D 0 ( 0 ) F
d0i d---~-
(9.40b)
Notes on Eqs. (9.40a,b): (1) When qsat~ is expressed in m o l / k g the density ( k g / m 3) enters the nominator of Eqs. (9.40a,b) and ~t = p. (2) If the zeolite is supported with a support having porosity ~, the effective surface area of the zeolite available for transport is ~.m2/m 2 and the term ~ enters the nominator of Eqs. (9.40a,b). If the flux is measured on a supported system and one wants to calculate the intrinsic zeolite properties, ~ enters the denominator of Eqs. (9.40a,b). (3) The term Do(q).blnP/blnq is identical to the Fick diffusion coefficient DF, while the intrinsic diffusion coefficient Do(q) is identical to the Maxwell-Stefan diffusion coefficient Dms. The thermodynamic factor F corrects for differences in activities (chemical potentials) of different gases which can exist with similar concentration gradients. It is similar to the factor that has been described in solid state diffusion by Darken and is sometimes named after him. Equations (9.40a,b) can be integrated over the thickness L of the membrane to yield expressions for the flux of specimen i: qp
Ji dz = - ~ , D 0 (q) F dq
(9.41)
qf
with q - qf (feed) at z = 0; q = qp (permeate) at z = 1. Note that qp and qf are steady-state concentrations which are not necessarily equal to the equilibrium concentrations. Equation (9.41) can be integrated under a number of different boundary conditions which will be treated below.
The Langmuir and Henry adsorption regions In many cases single gas adsorption in zeolites can be adequately describe d by a Langmuir-type adsorption isotherm as given in Section 9.2.2.3.:
KiP I+KiP
O~= ~
(9.16b)
384
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
with Ki the equilibrium Langmuir adsorption constant (Pa-1). Inserting (9.16b) in the expression for F yields the equation for the thermodynamic factor in the case of Langmuir adsorption: F-
1 1-0
(9.42a)
At higher values of 0 small deviations from the Langmuir isotherm are corrected in (9.42a) by introduction of an empirical constant k [86b]" F-
k 1-0
(6.42b)
Substitution of Eqs. (9.16b) and (9.42a) in (9.41) and integration, assuming D is not dependent on q, yields an explicit equation for the single gas flux in the Langmuir regime in terms of sorption and diffusion parameters: Do, i" qsat,i In
L
Ji-
qsat, i - qf, i qsat, i - qp, i
_ ~ D o , i qsat, i In 1 + KPf, i
L
(9.43a)
(9.43b)
1 + KPp, i
Equation (9.43a) can easily be converted in terms of occupancies by dividing numerator and denominator of the In term by qsatd. Note again that Do,i is the intrinsic diffusion coefficient and that DFick = D0/(1 - 0 ) and so DF increases strongly when e assumes larger values. At low occupancy we are in the Henry regime and Eqs. (9.16b) and (9.43b) can be simplified because KiP < 1: Ji = ~ D ~
" qsat'i " K
L
(Pf, i - Pp, i)
(9.44)
The temperature dependency of Ji can be introduced using a van 't Hoff-type relation for K and an Arrhenius relation for D:
i: 0iexp Do, i = D~,i exp -
(9.46)
where Ed is the activation energy for diffusion in the micropores. Insertion of (9.45) and (9.46) in (9.44) yields the temperature dependency of the flux in the Henry regime of a supported zeolite (~t ~ 1):
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
Ji = -~ D~,i Ko, i qsat,i exp
(Ed, i - Qa, i) (Pf- Pp) RT
385 (9.47)
Here D~,i is the pre-exponential coefficient of the intrinsic diffusion coefficient. Note that K'qsat is the Henry constant b as given in Eq. (9.16a) (mol kg -1Pa -1 or mol m -3 Pa -1) which can be directly determined from experiments without separate knowledge of the value of qsat. Equation (9.47) shows that the flux Ji is activated with an apparent activation energy (Ed~- Qa~) which is determined directly from permeation experiments. Since both parameters are positive quantifies, positive as well as negative values can be expected and the flux can be increase as well as decrease with temperature depending on the relative values of Ed~ and Qa~. Equation (9.47) has been used by several authors to describe, analyse a n d / o r simulate permeation and diffusion in silica [59,63,92] and in zeolite membranes [69,72,75]. 1
- Some limiting cases and discussion: At high concentration (high 0, low temperature, relatively large pressure), but within the Langmuir regime, KP >> I and with (9.16b) and (9.43b) or (9.41) one finds d ln pi li = - ~t . Do, i(q) 9qsat,i dz (9.48a) and
bt ]i = -~ Do, i qsat.i
P Li
(9.48b)
i n /~p,i
and with (9.45) and (9.46) assuming qsat~is independent of F:
~t, ln Pf, i ~ Pp, i exp Ji = ~ D~ qsat'iln
E(_~I
(9.48c)
Equation (9.48c) shows that at high values of 0 (low temperature) the apparent activation energy of the permeation equals that of the diffusivity provided that intra-crystalline diffusion is still the controlling mechanism. Outside the Henry region calculation of the permeation from adsorption and diffusion data requires knowledge of the value of qsat~" Especially for weakly adsorbing gases the value is not always known nor can be easily determined from experiments. As discussed by Kapteyn et al. [88] the value of qsat can be estimated from the molar volume which is obtained from extrapolation of the liquid state [90] or from volume filling theory [91]. Some results will be discussed below (binary gas permeation). In the Henry regime separate values of qsat are not necessary as discussed above and the product K'qsat~= b (Henry coef.) 1
Ed~can be larger than Qa,i because molecules can penetrate pores directly without preceding adsorption.
386
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
can be directly measured and used in the permeation equation. To obtain Eqs. (9.43), (9.47) and (9.48) it was necessary to assume that the intrinsic diffusion coefficient D o should be independent of the concentration (occupancy). This is only correct when there are no intermolecular interactions, so for lower values of the occupancy 0. An extensive discussion has been given by Xiao and Wei [86]. Based on model calculations and analysis of experimental results, they showed that Do is approximately constant until 0 -- 0.5-0.6 and then starts to change in a way depending on the value and the character of the interaction energy. This is expressed by a parameter W = A E / R T being a non-dimensional energy change from the non interacting to the intermolecular interacting situation. Analysing their model results it is obvious that for W = 2 and until 0 = 0.85 the relation between DFickand Do is almost similar to that obtained from the Langmuir adsorption type of isotherm. This is equivalent to an occupancy independent D o until large values of 0. For 0 _<0.5 this holds for all values of W _<3.
9.4.2.3 Permeation and Separation in Binary (Ternary) Gas Mixtures For multicomponent gas mixtures the generalised Maxwell-Stefan (GMS) equations should be used. Krishna [87b] derived an expression for the flux of specimen Ji: n
Oi li V~ti = ~,, O/Ji Oi J/ + RT qsat, i Di,j qsat, i " D j=l
(9.49) i
j~i
The first term at the right-hand side represents the friction due to the counterexchange of adsorbed molecules, while the second term represents the friction with the zeolite. Note that this equation shows a strong similarity with the GMS equation (9.33) for gaseous diffusion. For a two-component gas mixture, Eq. (9.49) reduces to a set of three equations [89]" J1 ~t " qsat,1 " D1
001
A1
3z
D1
302
A2
~)z
and
D2
J2
+
~ " qsat,2 " D2
01
+
-
A1
+
D1
q- 02 ~ - D12
A2
-
0
ln(1 -
D2
01 - 02)
(9.50)
0z
01 ~ D12
A(~_~ 1 - 02
A2) A1 A2 + ~22 - 02 ~D12+ 01 ~D12
(9.51a)
A1and A2defined by (9.51).
(9.51b)
9 -- TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANESWITH GASES AND VAPOURS
387
For small por~s with a pore diameter of about the same size as the molecular diameter (situation b 2 in Fig. 9.20) counter-exchange will not take place and terms with D12 a r e absent. This is the so-called single file diffusion for which can be written [89]" D1 { ~01 C)02l J1 - - ~tqsat,/(1 - 01 - 02) (1 - 02) -~z + 01 --~-zj
(9.52a)
{
(9.52b)
12-"--~tqsat,2
1-
D2
01 -- 0 2
(1 - 0~)
~Z
+ 02
l
-~zJ
The second component affects the diffusion of the first component in two ways: (i) by occupying a number of sites which now are not any longer available for component 1 (term (1 - 02) in denominator of D), and (ii) by an entrainment contribution due to the gradient of component 2, which can be either positive or negative depending on conditions. If single file diffusion prevails the single component permeation data (which determine Di) and single component adsorption data are sufficient to predict the binary fluxes via Eq. (9.52). The values for the concentration qi or occupancy 0i must be determined from the competitive Langmuir adsorption isotherm: 0i = ~
qi
qsat, i
Ki Pi
=
n 1+
E
(9.53)
KiPi
i=1
In most cases these adsorption data from gas mixtures are not available and estimates must be made by insertion of single gas data (K and Qa values) in (9.53) and setting /l
0i = 1
(9.53a)
i=1
- Permeation in large micropores:
The situation in large micropores is schematically represented in situation C1 and C2 of Fig. 9.20 and as discussed in Section 9.4.2.1. A part of the molecules move rather "free" in the gas phase (central part of pores) with a free length of diffusion in the Knudsen range [Xiao, 86] but with curved path trajectories [Petropoulos, 83], an increased potential with respect to the gas phase outside the sample and a small activation energy. Shelekhin et al. [92] have modeled this situation while in the transition region Xiao [86] describes the total micropore diffusion coefficient D t as:
388
9 ~ TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
-1 Dtl= DK1 + Dconf
(9.54)
where Dconf is the configurational diffusion coefficient in small micropores. Shelekhin developed his model for a single wall (Vycor type) membrane with pore diameter of approximately 1.5 nm. The large pores in the tortuous network are interconnected by pore openings(windows) with a smaller diameter in the range 0.5-0.6 nm. As discussed in Section 9.4.2.1. and Fig. 9.20 there are two possible states for the diffusant molecules inside relatively large micropores (1.0 < dp < 2.0 nm). In the SEMP model a certain fraction of the gas molecules move through the "gas phase" in the pore, while the remainder reside on the pore walls. According to Shelekhin [92] the total gas concentration CT inside the membrane is determined as: CT = Cg + Ca
(9.55)
where Cg and Ca are the gas phase and adsorbed phase concentrations respectively in m o l e / m 3 (membrane). The permeation is described by Eq. (9.40a) for both phases and the total permeation is taken as the sum of both gas (in the pore) and surface flow. An expression for the adsorbed concentration is obtained with the help of the Dubinin-Radushkevitch adsorption isotherm for microporous materials:
CR=
(9.56a)
Wpp Vm
where pp is the density of the porous medium (kg/m3), Vm is the adsorbate molar volume (m3/mol) and W the adsorbed volume given by: W = W0 exp
I
A
(9.56b)
where W0 is the limiting adsorption volume, [31 is an empirical factor (affinity coefficient), e is the adsorption potential e = RT ln(po/p) which is equal to the free energy to remove one mole of adsorbate molecules from the surface to the gas phase and W is the adsorbate volume at temperature T and relative pressure P/P0. Note that the meaning of W and W0 are equal to that of q and qsat in (9.40) in the case that CT = C~ (see below). The diffusion process is described now by a single diffusion coefficient D: D=~vv
(9.57a)
v = v g. v E exp -
(9.57b)
where v is the molecular mean velocity, ~ is the molecular mean free path, Vg is
9 -- TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
389
the probability that the gas molecule jumps in the desired direction (a geometrical parameter) and VEis the probability that the molecule has sufficient kinetic energy to surmount the energy barrier AE. Assuming all the obstructions in the way of diffusing molecules are rigid, expressions for Vg can be derived. For the case discussed this yields" 1 a2
V g = 3 dp2
(9.58)
where dn is the diameter of the window betweenthe large pores with diameter dp. Expression (9.58) accounts to some extent for shape selectivity which occurs with non-spherical molecules. The thermodynamic factor F in Eq. (9.40) can be calculated now with Eq. (9.56) and yields for the adsorbed phase: ~2 P0 F =-~- (RT)2 In ~ P
(9.58a)
with A an empiricalfactor. Substituting (9.57b) in (9.57a), using the gas kinetic expression for v, taking s =dp yields expression for D in both gas and surface phases:
(8az~ 0"5
DO,ads =Vgdp ~/t M )
/AEad,/
exp - ~
aT )
(8RT/~ (kEgas I RT-
Do,gas =Vgdp (=M)
exp - ~
(9.59a)
(9.59b)
where Do is the corrected (intrinsic) diffusion coefficient which is related to the Fickian diffusion coefficient by DF = D0.F (see Eqs. (9.40b) and (9.58a)). The total transmembrane flux is obtained now by summing up the expressions obtained from (9.40) for bulk gas and surface flow:
l=-g
dCads DO,ads F - - ~ p +
D0, ga~) RT L ap
(9.60)
with g = 0/~ and assuming D is independent of the occupancy (concentration). Note that the form of (9.60) resembles that of (9.47) for small pores. Shelekhin defines a parameter Tisowhich defines the temperature where the amount of gas adsorbed on the pore walls equals that in the gas phase (iso concentration point).
390
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
At T > Tiso + AT and C T ~ C g the first term in Eq. (9.60) can be ignored. For T < Tiso - AT and C T ~ C a the last term in Eq. (9.60) is negligible. From adsorption isotherms the value of T~o was determined for a number of gases. For highly adsorbable gases like CO2, the magnitude of T~somay be as high as 160~ At p = 5 bar, and in the investigated membrane T~o equals -20, 30, 70 and 160~ for N2, 02, CH4 and CO2 respectively. Adsorption is negligible for He and H 2 at temperatures above ambient. With decreasing pore size the contribution of the gas phase decreases and that of the surface flow increases and interaction between gas molecules and pore wall increases (see Section 9.4.2.1.). This is the case in the small windows between the larger pores. This transition situation from Knudsen-like to configurational diffusion has been modelled by Xiao and Wei [86] for zeolite systems. The activation energy of the gas phase is calculated using LennardJones potentials for the interaction energy, the activation energy for the adsorbed molecules is determined as a difference between the potentials in the pores and in the necks. The ratio of molecular diameter om and pore diameter dm at which the transition takes place depends on molecular shape and zeolite pore characteristics but is situated in the region 0.6 < ~m/dm < 1. A maximum permeability coefficient was estimated by Shelekhin using Eq. (9.60) assuming porosity r and tortuosity 9 values equal to 0.3 and 2 respectively, a pore diameter of 1.5 nm and a micropore volume of 0.11-0.13 m3/g. For gases with T >> T~s~ and so in the regime where bulk gas diffusivity with AEg is dominant, the permeability is strongly dependent on the magnitude of AE. Permeability values for He at T = 90 K are estimated to be 5000 and 9000 Barrer for z~E = 10 and 6 kJ/mol respectively (note: for AE = 0 (Knudsen) this value is 3500 Barrer). With a membrane thickness of 30 ~tm, estimated permeation values for He are 5x10 -~ and 10 -7 mol m-2s -1Pa respectively. Hassan et al. [95] using porous Vycor glass with a pore diameter of about 0.8 nm reported a separation factor ~ equal to 11.5 for O 2 / N 2 at 298 K and of 0~ = 4.6 at 423 K which values are about 20% larger than the perm selectivities. This is due to competitive adsorption in which the relatively strongly adsorbing component (02) saturates the surface and blocks the transport of the weakly adsorbing component (N2). Similar results are reported for C O 2 / C H 4 mixtures (~ 186-122 in the same temperature range). This explanation seems qualitatively in accordance with sorption data of Shelekhin [92] giving a sorption of 2 c m 3 / c m 3 membrane for 02 which is a factor 100 larger than that of N 2 at 30~ -
- Diffusion coefficients and kinetic information: The simplest way to obtain kinetic information is to perform permeation measurements under transient conditions with a non-adsorbing gas in a Wicke-Callenbach experiment [3]. In this case the total amount of permeant qt that has passed through the membrane as a function of time is given by
9 m TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
qt 1 De' t L" c-----~= 6 g 2
2 ~2
(-1) n ~
/,/2
De" t exp
_//2/i;2
g2
391
(9.61a)
n=l
which for t --4 ~ approaches the asymptote:
OeC0/t - - ~'2e /
qt= L
(9.61b)
which yields a straight line with intercept (time lag) equal to L2/6Deon the time axis. A similar result is obtained by plotting any quantity which is directly proportional to qt. Here Co = c(z = 0,t) and D e is the effective diffusion coefficient (porosity and tortuosity effects are incorporated in De). If the upstream (high) pressure is constant and much larger than the downstream (low) pressure, the slope of the asymptote will correspond to the steady state and so it is possible to determine the diffusivity under both steady state and transient conditions from a single permeation experiment. With a narrow and unimodal pore size distribution both methods yield reasonable consistent values. Large discrepancies point to strong microstructural effects (bimodal broad distribution, many dead ends, many defects).
9.4.3.4 Illustrative Examples of Permeation and Separation with Microporous Membranes Usually membranes investigated in literature do not have the simple architecture assumed in the preceding theoretical treatment. This requires a number of corrections or modified equations before data of the separating layers can be compared and analysed. This problem is treated in Section 9.5 but results will be used in this section.
(a) Large micropores Shelekhin et al. [92, 56] reported some interesting results for Vycor type of hollow-fibre membranes (for membrane characteristics see Section 9.4.2.2). The theory of permeation of hollow fibre systems will be treated in Section 9.5.
- Pressure dependence of permeation: For He, H2, 02 and N2 a linear dependence of the transmembrane flux on the pressure gradient across the membrane was observed. So the permeation is constant and independent of pressure as expected for Knudsen diffusion and sorbed gases in the Henry regime (and accordingly to the sum of both mechanisms).
392
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
T'I~
~
It, in I
m
120
O~
T-3,0~
I e l
.a
g
U
t-
eq
o 3"heory imi
0
4
8 12 Pressure, s t m
16
Fig. 9.22. Pressure d e p e n d e n c y of the permeability of CO2. After Shelekhin et al. [92].
For C O 2 , which is a highly adsorbable gas, the permeability (Barrer) as a function of pressure at T = 30 and 100~ is given in Fig. 9.22. At 100~ there is a weak maximum above which the permeability slightly decreases with increasing pressure, at 30~ there is a continuous decrease. The two curves could be described with Eqs. (9.60) using Eqs. (9.59) and (9.56) with values for AEads and AEgasof 21 kJ/mol and 10 kJ/mole, respectively. These values were obtained from a best fit of the curves to the experimental results (note: AE=E in Fig. 9.22). The maximum was explained with Eq. (9.60) considering a pressure independent bulk gas term (second term in (9.60)), while the first term for highly adsorbable gases may initially increase or decrease and then decrease with increasing pressure.
- Temperature dependence of permeation: For He the theoretically predicted permeability (Barrer) using Eq. (9.60) exhibits a maximum as a function of temperature for AEgas-4 kJ/mole. Note that the adsorbed gas phase is hardly present here. For larger values of AEgas
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
393
TABLE 9.6 Activation energies of diffusion and molecular kinetic diameters for different gases in microporous silica and zeolite membranes Gas
(~m (nm)
Eperm (kJ/mol)
Ed (kJ/mol)
He H2 CO2 02 N2 CH 4 C~-I8 n-C4H8 iso-C4H8 benzene
0.26 0.289 0.33 0.346 0.364 0.38 0.43
22.5 (b)
0.52 (a)
3.68 (b) 13.0 (b) 13.0 (b) 23.4 (b)
9.9 (a) 10.6 (a) 18.2 (a) 28.7 (a)
Eperm (kJ/mol)
Ed (kJ/mol)
15-21 [63] --10
21 [63] 32 [63]
30 [88,89] 0.5 0.585
(a) Ref. [92]; (b) Ref. [56].
there is a c o n t i n u o u s increase of the permeability. N o discussion w a s g i v e n of the occurrence of this m a x i m u m . Such a c o n t i n u o u s (non-linear) increase of the p e r m e a b i l i t y as fiT) w a s o b s e r v e d i n d e e d for CO2, O~- N2 a n d CH4 in the t e m p e r a t u r e r e g i o n of 300520 K. The activation energies of the p e r m e a b i l i t y w e r e o b t a i n e d w i t h a non-linear least-squares fit to the exp. curves a n d are given in Table 9.6 t o g e t h e r w i t h the theoretical calculated ones. These activation energies reflect the activation e n e r g y of diffusion E d in the m e m b r a n e . This table illustrates that, g i v e n the d i a m e t e r of the p o r e there is an increase of the v a l u e of E d w i t h increasing kinetic (molecular) diameter. The reverse t r e n d is f o u n d in the p e r m e a b i l i t y values. Theoretically it is p r e d i c t e d that the v a l u e of B = D.(M/T)l/2should c o n v e r g e to a single v a l u e for all gases for T ~ co. A plot of literature d a t a of B v a l u e s as af(T) for a large n u m b e r of gases yields v a l u e s of 1.1x10 -4 for the Vycor m e m b r a n e a n d 2.2x10 -4 for ZSM5 (Shelekhin [92]) in g o o d a g r e e m e n t w i t h theory.
- Selectivity: Based on p e r m e a b i l i t y data, permselectivities (selectivity factor a = FA/FB) w e r e calculated for a n u m b e r of pairs of gases A-B. At 30~ s o m e typical v a l u e s are a -- 4190 for H2/CH4, (z = 2.5 for H 2 / C O 2 a n d a = 1675 for C O 2 / C H 4. All selectivity factors decrease w i t h t e m p e r a t u r e e.g. at 250~ ~ = 62 for C O 2 / C H 4.
394
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
(b) Small micropores - Silica membranes (permeation and separation): Silica microporous membranes combining high separation factors and high permeation values were first reported by Uhlhorn et al. [28,58] and were further developed and analysed by de Lange et al. [59-63]. More recently silica membranes made by a CVD process with similar qualities were reported by Lin et al. [67] and by Wu et al. [68]. The membranes, synthesised by Uhlhorn and by de Lange et al., were formed from polymeric silica solutions in an ultra-thin layer of about 100 nm thick partly on top (50 nm), partly within (50 nm) the pores of the y-A1203 support (pore diameter ~- 4 nm, thickness 3-8 nm) which was in turn supported by an ~-alumina support (disc) with a pore diameter of about 0.2 ~tm. The characteristics of the silica layer depend strongly on details of the synthesis procedure and a high quality supporting system is required (with low roughness and no or few defects) to obtain good quality membranes. The pore diameters were in the range 0.4-0.5 nm. Discussion of permeation and separation requires some characteristic parameter for the membrane quality. As shown below the apparent activation energy for H 2 permeation gives a good correlation with the separation factor and is used as a measure of quality. Furthermore, the total measured permeation has to be corrected for influence of the support to obtain permeation and activation energy values characteristic for the silica layer (see also Section 9.5). The experimental permeation results could be consistently described using Eqs. (9.43b) and (9.47) for Langmuir and Henry sorption respectively as shown by de Lange in a full analysis of sorption, permeation and separation results of five different gases [63]. This description requires knowledge of adsorption isotherms which could be measured only on unsupported membranes. To use these data for calculation of the permeation of supported membranes requires the assumption of equal pore characteristics in both cases. As discussed by de Lange et al. this is probably not correct in the case of silica layers. Based on sorption data a microporosity of about 30% and a pore size distribution with a peak at 0.5 nm is found. Analysis of permeation data point to a pore diameter of -- 0.4 nm and a considerably smaller porosity. Table 9.7 summarises the sorption data. H 2 and C H 4 have relatively low (isosteric) adsorption h e a t s (qSt) while CO2 and isobutane strongly adsorb. Henry behaviour in the pressure range up to 125 kPa exist at temperatures larger than the limiting temperature Z l i m i t , H e n r y given in Table 9.7. At ambient temperature (323 K) C H 4 showed Henry behaviour up to 8 bar while H 2 exhibited Henry behaviour to at least 15 bar. CO2 exhibited Langmuir behaviour at I bar (323 K).
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
395
TABLE 9.7 Henry constants (b), isosteric heats of adsorption qst and lower limiting temperature for sorption behaviour Zlimifl-Ienry for CH4, H2, CO2 and isobutane in microporous silica at P < 125 kPa. After de Lange et al. [59,63] Gas (--~)
CO2
CH4
H2
iso-C4Hlo
qSt (kJ/ mol)
22.3
10.3
6.1
22.9
(-~) b (mol kg-1 Pa-1)
348
273
194
T (K)
($)
~) 0.43 4.2x10-6 2.3x10--6
1.5x10-4 3.7x10-7 1.8x10-7
1.7x10-4
(~)
Zlimit~
Henry (K)
77 194 273 303 323 348 373 473
3.2x10-5 7.8x10-6 5.9x10-6 3.2x10--6 2.1x10-6 4.8x10-7
w
-
2.9•
-5
2.4x10 -5 9.4•
-6
TABLE 9.8 Typical values of permeation and activation energies of microporous silica membranes. Phigh~ 3 bar. After de Lange et al. [59,63] Permeation (10-7 tool m -2 8-1 Pa-1) Gas H2 CO2
50~ (H2) 28~ (CO2) 4.1 (4.5) 2.3 (3.0)
Apparent Eact* (kJ mo1-1) 200~ 21.7 (52.7) 6.8 (32.3)
14.9 (21.7) 6.1 (14.9)
*Values between brackets corrected for support influence.
T h e s u r f a c e c o v e r a g e (0) for C O 2 w a s m a x i m u m 20% at 273 K a n d 125 k P a a n d the isosteric h e a t w a s p r a c t i c a l l y i n d e p e n d e n t of c o v e r a g e . T h i s r e s u l t i n d i c a t e t h a t for all o t h e r g a s e s in the p r e s s u r e r a n g e u p to -- I b a r c o v e r a g e w a s also low. C o n s e q u e n t l y , Eq. (9.47) c a n be u s e d to d e s c r i b e t h e p e r m e a t i o n results. T y p i c a l p e r m e a t i o n r e s u l t s are g i v e n in T a b l e 9.8. T h e p e r m e a t i o n v a l u e s for H2, CH4 a n d CO2 at T > Tiso,Henry w e r e a p p r o x i m a t e l y i n d e p e n d e n t of p r e s s u r e (as e x p e c t e d for H e n r y b e h a v i o u r ) a n d in-
396
9 m TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
creased with temperature from about 4.5-20x10 -7 m o l / m 2 s Pa (425-473 K) for H 2 and 2.3-7x10 -7 mol/m2 s Pa (273-473 K) for CO2 [63]. For membranes with lower quality the increase is less pronounced due to the smaller apparent activation energy. Typical values for iso-butane are 0,6-0.35x10 -7 m o l / m 2 s Pa at 50~ and 200~ respectively. Note that the permeation in this case decrease with increasing temperature. Similar conclusions were drawn by Wu et al. [68] who reported apparent activation energies in the range 11-20 kJ/mol for considerably lower H 2 permeation values. Table 9.8 shows that in the case where the flow resistance is not negligible, corrections should be applied on the total permeation value of the system to obtain the true permeation values of the silica separation layer. Consequently the true values of the activation energy may also differ considerably compared to the apparent ones. (See further Section 9.5). The conclusion of de Lange et al. [61] is that the activation energy of permeation of H2 exhibit a good correlation with the quality of the membranes (permeation, separation factor) and high quality membranes should have an apparent activation energy of at least 10 kJ/mol. Sometimes a weak maximum in the permeation of CO2 as a function of the feed pressure of a similar type as reported by Shelekhin et al. [92] has been observed by de Lange [61] also for small micropores. Separation factors (defined by Eq. (9.36)) obtained from mixtures are usually smaller than permselectivities obtained from the ratio of single gas permeation (see qualitative discussion in Section 9.4.2.1.). At hightemperature and lower concentrations the mixture separation approaches the permselectivities which in turn tend to approach the Knudsen value at high enough temperatures. Typical values for some gas mixtures in combination with permeation data (in the mixture) for different silica membrane systems are given in Tables 9.9a and 9.9b respectively. Several interesting conclusions can be drawn from Tables 9.8 and 9.9. The synthesis method and related membrane quality strongly determines the obtainable combination of permeation and separation values (as characterised by E~ct,H2).High quality membranes have activation energies for permeation (after correction for support influences) in the range 15-22 kJ/mol for H 2 and 10-15 kJ/mol for CO2 with typical permeation values at 200~ of 20x10-7 for H2 and 5x10-7 m o l / m 2 s Pa for CO2 respectively. The permeation value of isobutane at 200~ is very small which indicates a pore size close to that of the kinetic diameter of i-butane and the absence of (larger) defects. Separation factors are in the order of 20-30 for H2/CH4 and 150-200 for H2/isobutane. Lower quality membranes (lower values of Eapp,H2 tend to give larger permeation and smaller separation values for non-adsorbing gases. For strongly adsorbing gases (i-butane, CO2) even with moderate quality membranes ( E a p p , H 2 = 5-10 kJ/mol) good separation factors can be obtained up to about 200~
9 m TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
397
TABLE 9.9a Separation factors (defined by Eq. (9.36)) for some gas mixture-silica membrane combinations Membrane
Gas mixture Separation factor a at T (~ ~-25
A13Sil a A13Sil a a. A13Si2 b. A13Sil A13Si2 A13Si2Ti AllSi2
H2/C3I~ H2/CO 2 H2/N2 H2/Ct-I4 H2/CH4 H2/CH4 H2/CH4
AllSil AllSi2
H2-iC4H8 H2-iC4H8
CO2/CH 4
50
48 13 1.7 15-36 2 2 12
100 65 62 2.5
-
Eapp
150
4.5
200
>_250
kJ/mol
28 156 5.5
270 6.6
7-(10) 7-(10) 5-(7)
3 3 50 =9
5 5 150 10
8 10 200 11
11 18 165 11
80 80
105 130
110 170
110 180
12 30-40 11 110 170
5-(7) 7-(9) 12-(16) -
Values are taken from de Lange et al. [59-63], unless otherwise referred. aTaken from Uhlhom et al. [58]. bTaken from Shelekhin et al. [56]. Membrane code: AlxSiy with x and y are number of A1203 and SiO2 layers respectively. Eapp is the apparent activation energy of permeation for H2. Figure in parenthesis is corrected for support influence.
TABLE 9.9b Permeation values of some gas mixtures in different silica membranes Membrane
Gas mixture
A13Sil a A13Si2
CO2/CH4 H2/CH 4
A13Si2 A13SilTi
H2/CH4 H2/CH4
Permeation F at T (~ 50
100
4 (CO2) 3.7 (H2) 1.8 (CH4) 1.8 (H2)
10
2 (H2)
150
200
3 (H2)
50 (H2) 4 (H2)
Membrane code: see Table 9.9a. Permeation given in 10-7mol/m 2 s Pa.
For non-adsorbing or weakly adsorbing gases (H2, CH4, N2, 02) the permeation increases with temperature (for high quality membranes). This is in accordance with data of Wu et al. [68] who reported increasing permeation values (H2) for membranes with lower quality (characterised by positive values of Eap p for
398
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES w r r H GASES AND VAPOURS
N2). Permeation values of Wu at 600~ are in the range 0.03-1.0x10 -7 m o l / m 2 s Pa for H 2 (for N 2 a factor of 20-70 lower) with the highest value for the lower quality membrane (Eact,H2-~ 11 kJ/mol). As will be discussed in Section 9.5 the different values cannot be compared directly because of the strong influence of pressure conditions and support effects.
- Silica membranes (diffusion data): Equation (9.47) can be used to calculate the activation energy for "intracrystalline" micropore diffusion Ed,/ of specimen i provided sufficient sorption data are available. The value of Ed,/ follows from: Edd = Eapp- Qa,i
(9.62)
where Qa,/is the isosteric heat of adsorption and Eapp is the measured apparent activation energy of permeation (Eq. (9.47)) after correction for support influences. With typical values of Eap p equal to 15 kJ/mol (H2) and 10 kJ/mol (CO2) and typical values of Qa equal to 6 (H2) and 23 kJ/mol (CO2) [63] the resulting calculated activation energies of the intra channel (micropore) diffusion are about 21 kJ/mol for H2 and 32 kJ/mol CO2 [59,93]. This is in accordance with the expectation that larger molecules will have a larger activation energy for diffusion than smaller ones [92,82]. Equation (9.47) is also used by de Lange [63] to calculate the value of the diffusion coefficient of several gases in silica membranes. The term ~t in (9.47) takes the form ~t = p(1 - ~)/~ with the skeletal density of silica p = 2.2 k g / m 3, the silica porosity ~ = 0.4 and the membranes thickness L = 100 nm. Taking all the sorption terms together in the Henry constant b (which can be directly measured) and substitution in (9.47) yields: J = 3.3 x 101SD x b
(9.63)
Typical values of b and D for a range of membranes are given in Table 9.10 together with some other parameter values. The range in D values reflects differences in membrane quality, the smallest D values being formed in high quality membranes. The diffusion coefficients become smaller in the same order as the kinetic molecular diameter (see Table 9.6) increase. The large differences in the D values indicate that the pore diameter is of the order of the molecular diameters (0.4~.5 nm). The differences in D values are much larger than the differences in permeation values and indicate the effect of the sorption term even for weakly adsorbing gases (compare H 2 and CH4). The absolute magnitude of the diffusion coefficient is rather uncertain, because all uncertainties concerning the value of ~t are reflected by the D values.
9 ~ TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
399
TABLE 9.10 Typical values for the diffusion coefficient of different gases in silica membranes at room temperature after de Lange et al. [63] Gas
b (289 K) ( m o l / k g Pa)
D (,298 K) (m'/s)
Ed,i (kJ/mol)
qst (kJ/mol)
H2
1.4x10 -7
5.25x10 -11
13-21
--6
CO2
1.4x10 -5
6-9x10 -13
--30-33
=23
CH4
1.6x10 -6
5-15x10 -13
--10
iC4H10
7.3x10 -5
3x10 -14
---23
The value of the porosity is taken from adsorption measurements on unsupported silica membranes and probably the porosity of supported silica membranes is considerably smaller and the calculated D values give a lower limit. A comparison with zeolite data and effects of surface reactions will be discussed below. - Zeolite membranes:
Permeation and separation data on well defined, high quality zeolite membranes are only reported for MFI (ZSM-5, silicalite) zeolites grown in situ directly from the precursor solution on top of a substrate. The experimental single gas permeation results could be in a number of cases consistently described using Eqs. (9.43)-(9.48) for the Langmuir and Henry regimes. Geus et al. [70] give a detailed description of the synthesis of a MFI layer with a thickness of about 50 ~tm on top of a porous steel support. Vroon et al. [72,74] synthesised thin (2-6 ~tm) MFI layers on a (x-A1203 support and varied the crystallite size (0.1-0.4 ~tm) in the layers by varying the synthesis temperature and using a very high pH (-~ 12.5). Both groups of authors investigated the quality of their membranes. Both groups of authors measured a very small flux of gas molecules which are much too large to pass the pores of the MFI structure indicating that some larger pores were present in the layer. The measured fluxes for iso-octane (or of 2-2-di-methylpentane) were more than 5 orders of magnitude smaller than that of C H 4 indicating a good membrane quality. This conclusion is supported by the large observed separation factors for e.g. H2/butane, CH4/butane and n-butane/i-butane (see below). Vroon had to apply two silicalite layers on top of each other in order to obtain this good quality. Typical single gas permeation data for relatively thick MFI membranes are given in Fig. 9.23 [71]. At 673 K all the gases show a linear dependence on the (feed) pressure (Henry behaviour) as is the case at 300 K for the noble gases and for C H 4 , whereas butane and ethane exhibit saturation at low and higher
400
9 - - TRANSPORTAND SEPARATION PROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
10
22
(a)
20 8 18
N
16
s
~4
6
E E
~2
5
~'~
10
0
ft.
-~
@
T
~W"
.,..m
O.
W
0
E 0
.E 0
4
e
~" , . b u t a n e * 100 A
W W
o
3
O. t
0
20
40
60
80
I O0
Partial pressure in feed (a)
(kPa)
Fig. 9.23. Steady-state flux and permeate pressure as a function of partial feed pressure for different gases at 300 K (a) and at 673 K (b). After Bakker et al. [71]. (0) neon, (+) argon, (V) krypton, ( 9) methane, (A) ethane, ( I ) n-butane, (&) isobutane, (O) CFC-12.
pressure (30 and 80 kPa) respectively. The permeation increases with temperature for all gases except krypton and CH4 which were almost independent of temperature. The maximum observed permeation values (673 K) of noble gases and CH 4 are about the same and correspond with a permeation of 1.6-2.3x10 -7 m o l / m 2 s Pa. Permeation values of 1.2 and 0.9x10 -7 m o l / m 2 s Pa are found for n- and i-butane respectively. Similar results concerning the trends in the permeation values as a function of pressure are reported by Vroon et al. [72,73,74] for CH4, ethane, propane and butane. The absolute values of the permeation reported by Vroon et al. for these
9 -- TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
401
10
22
(b)
20
8
18
O.
..~ Q)
A(/)
16
7
E
~4
6
12
S
.E
4
~
3
ID ~" t2.
2
"~.
0
E E x
~..
10
8 II
0
20
40
60
80
m
E
~00
Partial pressure in feed ( k P a ) (b) Fig. 9.23 (continued). Caption opposite. MFI m e m b r a n e s on (x-A120 3 supports showed for CH 4 a decrease of the permeation from lx10 -7 (298 K) to 0.6x10 -7 m o l / m 2 s Pa (473 K) with increasing temperature. Plots of the flux of butane (Fig. 9.24), propane and ethane versus temperature exhibit a (weak) m a x i m u m which values shifts from 440 K for n-butane to 350 K for ethane at 100 kPa. This m a x i m u m depends on the (partial) pressure of the gas (e.g. for n-butane at 8 kPa pressure the m a x i m u m is situated at about 390 K). Similar maxima are found [72,74] in the curves of H 2 and CO2 vs temperature as s h o w n in Fig. 9.25 and are also reported by Kapteyn for n-butane [88,89]. The absolute values found for CH4, CO 2 and n-butane can be compared with that obtained by Bakker [71], Geus [75] and Kapteyn et al. [88]. At 473 K the values obtained by Vroon are lower by a factor of about 2.5 compared to that of
402
9
-
TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
-
15 0
0 0
A
qll I
E
,/.
0 o
00 O
10
0 o
0
0
0@0000
m m
0
E E v
X
0
O O0
o
vvVVVVVVVv
vvvvVVVVYV
m
u
vvVt~ vvv vV V
0 273
i
,
I
,
I
373
473
Temperature (K) Fig. 9.24. Comparison of the methane and n-butane flux measured by the dead-end method (O,V) and the Wicke-Kallenback method (o,v). After Z. Vroon [72-74].
20
,..
15
oO~176176176176176176176176176176
10r
oO
~1 v v v V V F
0
-vvvvvvvvvvvv
+ "l" "1""1"+ + + + + + + + ,1" + +
50
100
150
V
l /
+++t 200
Temperature (~ Fig. 9.25. Flux of hydrogen (O), helium (V), carbon dioxide (+) and sulphur hexafluoride A as a function of temperature at a feed pressure of 100 kPa. After Vroon et al. [72-74].
9 -- TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
403
Bakker. This means that the permeation values corrected for thickness differences (permeability) of the thin membranes with small crystallites on a support with relatively narrow pores (0.2 ~tm) are much lower than of the thicker layers on a steel support. This point will be discussed later. Kapteyn et al. [88,89] and Vroon et al. [72,74] could model and describe their single gas permeation measurements for C H 4 and n-butane rather well with Eqs. (9.41)-(9.48) taking the thermodynamic factor in the Langmuir regime from adsorption measurements. Using Eqs. (9.40) and (9.43), Kapteyn showed that the corrected (intrinsic) value of the diffusion constant Do at 300 K of n-butane is independent of pressure (up to 1 bar) and Do (300 K) equals 0.4x10-6 c m 2 / s (contrary to the Fickian diffusion constant which is strongly increasing with increasing 0). The maximum in the flux versus temperature for n-butane could be correlated with a strong decrease of the occupancy 0 at higher temperatures. Occupancies at p = 0.5 bar vary almost linearly from 0 = 0.8 at 350 K to 0 = 0.2 at 450 K and the maximum in the c u ~ e is situated at 0 -- 0.4. Initially the change in 0 is less. So at lower temperature the diffusion coefficient increases more rapidly than the concentration (occupancy) decreases, at higher temperature the reverse is true and this give rise to the observed maximum asflT). A similar result is reported by Vroon et al. [72,74] who calculated the flux of C H 4 and n-butane using also Eqs. (9.40) and (9.43) and using diffusion constants taken from literature [94] and measured on twinned single crystals by the membrane method. The sorption data for methane taken from literature agree within 20% from data obtained by Vroon, for n-butane not sufficient literature data are available and measured data (gravimetric method) are used. The set of data used in the calculations is given in Table 9.11 and the calculation results in Figs. 9.26 and 9.27. TABLE 9.11 Henry constants and saturation concentrations obtained from the gravimetric sorption measurements on silicalite particles and diffusion constants obtained by the membrane method of methane and n-butane. After Vroon et al. [72-74] Gas
Temperature (K)
Henry constant (mol Pa -1 m -3)
Saturation concentration (mol m -3)
Diffusion coefficient (m 2 s -1)
Methane
298 323
8.6x10 -3 5.4x10-3
-
0.7x10-1~ 1.0x10-1~
n-Butane
298 323 348
17.5 4.0 1.0
2.2x10 3 1.8x103 1.4><103
1.0xl0 -12 2.2x10 -12 4.3x10 -12
404
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
III A
E
vI
t~
I
o v
E E
x
@
_~ u.
1
iF
9
A 9 @
0
20
40
60
80
100
Feed pressure (kPa) Fig. 9.26. Calculated n-butane flux as a function of the feed pressure at 298 K (I), 323 (II) and 348 (1II) K. Permeate pressure is 0 kPa. Curve A is experimentally obtained. After Vroon et al. [72-74].
10
8
A
E
T
w 0
E E
4
X n
u.
2
0
1 Permeate
I
I
I
2
3
4
pressure
5
(kPa)
Fig. 9.27. Calculated methane (I) and n-butane (II) fluxes at a feed pressure of 100 kPa and 298 K as function of the permeate pressure. After Vroon et al. [72-74]. Figures in Curve II have been multiplied by a factor of 3.
9 - - T R A N S P O R T A N D S E P A R A T I O N PROPERTIES OF M E M B R A N E S W I T H GASES A N D V A P O U R S
405
The activation energy for intra-crystalline diffusion for n-butane is 30 kJ mo1-1 [72,74,89]. For the isosteric heat of adsorption values of =38 kJ mo1-1 are reported by Kapteyn and by Vroon, which value is considerably lower than other values (=50 kJ mo1-1) reported in literature. For CH4 a good agreement between calculated and measured fluxes is obtained, for n-butane the agreement is reasonable to bad at low pressure and good at higher pressure. A difficult problem is the value of the saturation concentration. In many cases no reliable experimental data are known and theoretical estimates have to be made usually under the questionable assumption that qs is independent of temperature. For n-butane the theoretical sorpfion capacity of MFI/silicalite equals about 2.1x10 -3 mol g-1 (equivalent to 12 molecules per unit cell). The Delft group (Kapteyn, Bakker) reported for their large MFI crystals experimental values of qs in the range of 1.6-210x10-3 mol g-i, whereas Vroon et al. for small crystals (0.1-0.4 ~tm) reported qs = 1.2x10-3 mol g-1. Obviously the sorpfion capacity depends on synthesis conditions and/or crystal size. For C H 4 a theoretical estimate of qs = 4 x 1 0 - 3 mol g-1 is reported by Kapteyn et al. [89], whereas the qs values decrease with increasing molecular weight in the C1--C 4 series. Figure 9.27 shows a decrease by a factor of 2 of the calculated flux values, when at a constant feed pressure of 100 kPa the permeate pressure changes from I to 10 kPa. This illustrates the large influence of the permeate pressure on the flux. In the supported membranes under consideration the driving force may be decreased due to a pressure gradient in the support (decreased occupancy at interface) a n d / o r equilibrium conditions may not exist at the interface between support and zeolite due to a slow desorption process and/or due to insufficient removal of the permeate. Bai et al. [27] observed that a pressure drop across the membrane leads to higher selectivities than obtained from W-C type of measurements using a sweep gas to remove the permeant. Comparison of the flux values obtained by Bakker/Kapteyn and by Vroon et al. show comparable values after correction for differences in support porosity. This is unexpected because the layer thickness of the membranes made by Vroon et al. is about 3 ~tm, a factor 17 smaller than that prepared by Bakker/Kapteyn. The reason for this discrepancy is not known and may be originated in a difference of the effective and nominal thickness, e.g., by a strongly porous part of the thick layers. Another reason might be effects of crystal size and number of grain boundaries and of synthesis conditions (lower crystallinity at lower synthesis temperatures). These points have been investigated by Vroon et al. [72,74]. The results are tabulated in Table 9.12 for the methane flux in membranes with different crystal sizes and thickness. As shown in the table the membrane flux corrected for the layer thickness (J.L) decrease with increasing synthesis temperature (and so with decreasing particle size). A direct correlation with bulk properties due to particle size is however not present because the flux is
406
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
TABLE 9.12 Methane flux of typical Mb1 membranes with different micro structures. After Vroon et al. [72-74] Membrane L code* (~tm)
r (nm)
A (L/r)
J**
J.L***
J.r***
J.A***
P100 P20 P150 P180
275 350 550 700
13 10 9 8
13 10 7 5
46 35 35 28
3.6 3.5 3.8 3.5
2.7 1.2 0.6 0.4
3.5 3.5 5 5.5
*PlOO,P20, P150 and P180 are membranes grown in situ in two steps at 373, 393, 423 and 453 K, respectively. **Methane flux (mmol m-2 s-1) at a feed pressure of 100 kPa and 298 K. ***J-Lis the flux corrected for the layer thickness, J.r is the flux corrected for the particle size, J.A is the flux taken into account the munber of interfaces perpendicular to the support surface. independent of the crystal size itself (product J.r is constant). So obviously the interfaces (grain boundaries) between zeolite particles play a role and an increase of the n u m b e r of grain boundaries results in an increase of the flux (product J.A is largest for largest A value). The membranes show large separation factors including size exclusion for large molecules, and this exclude continuous and rapid transport p a t h w a y s w i t h large dimensions. A model is d e v e l o p e d therefore of grain b o u n d a r i e s containing small microvoids and slits of a size similar to the pore size w h i c h p r o m o t e the flux. This does not explain h o w e v e r the observed difference b e t w e e n flux values of Vroon and of K a p t e y n / B a k k e r . - Separation and permeation with binary gas mixtures in zeolite membranes: Separation results with s u p p o r t e d MFI m e m b r a n e s of good quality are reported by the Delft group (Kapteyn, Geus, Bakker [69-71,89]) with m e m branes on porous steel supports and by the Twente group (Burggraaf, Vroon et al. [72,73]) with m e m b r a n e s on porous 0~-alumina s u p p o r t a n d of m e d i u m (defect quality) by, e.g., the g r o u p s of Noble and coworkers [27,77,97] on y-alumina supports, of Ma and coworkers [76,96]. Some other g r o u p s report results w i t h o u t sufficient indication of m e m b r a n e quality [64,75,98]. It is interesting to note that m e m b r a n e s with m e d i u m quality ~ as indicated by relatively low separation factors for non- or w e a k l y adsorbing gases ~ can have good separation factors for mixtures with a strongly adsorbing, easily condensing c o m p o n e n t (e.g., H2-methanol). Obviously the defects are blocked by filling the defect with the easily condensing c o m p o n e n t (capillary condensation type of blocking [98]). This makes it clear that w e m u s t distinguish the several categories of gas mixtures as mentioned in Table 9.5 w i t h the m a i n categories of mixtures of (i) w e a k l y ( W ) - w e a k l y (W) adsorbing gases ii) weakly ( W ) - strongly (S) adsorbing and iii) strongly (S) - strongly (S) adsorbing gases.
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
407
For a combination of two weakly adsorbing gases the separation factors approach the permselectivity values but are usually somewhat lower whereas the permeance values in the mixture are somewhat decreased with respect to single gas permeances. This is precisely what can be expected for supported membranes with little interactions of molecules within the pores. Interesting results are described by Kapteyn [89] and by Krishna and van der Broek [100] for isomer separation in the range of intermediately adsorbing (between S and W) gases (ethane/ethene, propane/propene) using the Maxwell-Stefan approach (Eqs. (9.50) and (9.51)) and the competitive Langmuir adsorption model (Eq. (9.53)). Krishna et al. also simulated the behaviour of H2/n-butane mixtures. The exchange contributions were assumed to be negligible (single file diffusion) and so diffusivity data of the single gas permeation experiments were used and Eqs. (9.50) and (9.51) were numerically integrated. The best description was obtained with diffusivity values of ethane which are decreased and those of ethene slightly increased with respect to single gas measurements, while for propane/propene the diffusivifies became equal. In this latter case high occupancies (98%) were present and the faster component is decelerated by the slower one. The alkanes in the mixture exhibit higher (measured) fluxes compared with their unsaturated analogues and some selectivity over the corresponding alkanes exist. At 292 K and for 50 kPa: 50 kPa mixtures the selectivity factors are 1.9 (ethane/ethene with a flux value of 17x10 -3 mol m -2 s -1 for ethane) and 1.3 (propane/propene with a flux of 5x10 -3 mol m -2 s -1 for propane). Typical curves for a combination of weakly and strongly adsorbing gases (S-W) are given in Fig. 9.29 (after Kapteyn [99]) and Fig. 9.28 (after Vroon [72,74]). A similar result to that shown in Fig. 9.29 is also observed for H 2 / C O 2 mixtures [71]. Typical in all the cases is the occurrence of a maximum in the permeation curve of the S component in the mixture as is also observed in their single gas permeation curves. As shown in Fig. 9.28 the permeation of the S component (n-butane) is only slightly lower than the single gas permeation. The W component (H2) is drastically decreased and, in contrast to the single gas permeation, increases with increasing temperature. When the temperature increases sufficiently, the permeation values in the mixture first become equal and finally cross each other (see Fig. 9.29), with the W component becoming faster permeating. So we have a conversion of the selectivity factor ~ > I for the S / W mixture at lower temperature to values ~ < 1 at high temperature (note a > I for the W / S combination is equivalent to ~ < 1 for the S/W). This can be explained by preferential sorpfion of the strongly adsorbing component which excludes (or decreases) the concentration of the S component. With increasing temperature the concentration of the S component decreases much more strongly than that of the W component, the "blocking" effect decreases and finally vanishes and at high temperature the mixture starts to behave in a similar way to a mixture of two W components.
408
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
6
A
t
4
I
o
E E I
V v
2
N l
V
la,
0
,
9 ,
,
,i,,,
9
i
Temlmrature (K)
Fig. 9.28. Permeation and separation behaviour of a mixture of 50 EPa C1-~ ( 9and 50 EPa n-butane
(V) as a function of temperature of a MFI membrane. Single gas permeation values are added: CH4 (O), n-butane V). After Vroon et al. [72-74]. 25
-'" ?
E
,,,,,=
O
E E
f
2o
~s n-Butane
/
10
"0 e ~
X
ft.
s
Hydrogen/ 9
300
,
~
&
400
500
,,i
600
Temperature (K) Fig. 9.29. Separation b e h a v i o u r of a H 2 / n - b u t a n e m i x t u r e (1:1) as a function of t e m p e r a t u r e of a MFI (silicalite) m e m b r a n e at 100 kPa. After K a p t e y n et al. [99].
A q u a n t i t a t i v e t r e a t m e n t of this complex b e h a v i o u r is not yet p u b l i s h e d . The case of a m i x t u r e of t w o S c o m p o n e n t s is e v e n m o r e c o m p l e x a n d general qualitative descriptions h a v e not yet b e e n published. E x a m p l e s of p e r m e a t i o n
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
10-' V V
A (N I
E
T I m o
E E
x =l m u,,,
V'
V
V
V
V
409
V V
T
10"
V V
A
A
10-*
A
A
A
of equipment
10" 273
373
473
Temperature (K) Fig. 9.30. Separation and permeation behaviour of a mixture of 0.31 kPa paraxylene (V) and 0.26 kPa o-xylene (A) as a function of temperature. Single gas permeation data are also given: 0.62 kPa px (A) and 0.52 kPa ox (V). The total pressure was 100 kPa, the balance being He. After Vroon et al.
[72-74]. and separation results of these combinations are reported by Vroon et al. [72,74] for n-butane/i-butane, benzene/cyclohexane, methane or hexane/2,2-dimethylbenzene and p/o-xylene mixtures. The separation behaviour of a p / o xylene mixture is given in Fig. 9.30. The permeation of the paraxylene is much larger than that of the o-xylene at higher temperature, the last one has a permeation which is at the detection limit of the equipment used. The molecule has a diameter which is larger than that of the pore diameter of the MFI and so we have here an example of separation by size exclusion. The flux of p-xylene shows a weak maximum as a fiT) and consequently the separation factor does the same with a peak value of c~ = 100 at ~400 K under the given conditions. The separation factors and the permselectivities are equal as expected for the size exclusion mechanism. Xiang and Ma [76] reported a value of ~ = !5 for p/meta-xylene separation with a flux of 35 ml m -2 h -1 (=--4.3x10-7molm -2 s-1) for the m-xylene at room temperature. An even more straightforward example of size exclusion is exhibited by the mixture of n-hexane and 2-2 dimethylbutane where the flux of the hexane is three to four orders of magnitude larger than that of the 2-2 dimethylbutane up to the highest temperature measured (473 K) and the flux of the 2-2 dimethylbutane
410
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
is of a similar magnitude of that of o-xylene Vroon [72,74]. Obviously a small number of small defects accounts for the remaining flux of the large molecule. Finally it should be noted that isomers like n-i-butane [73,89] and cis-butane/ 2-trans-butanes [76] exhibit different permeances indicating the shape selective properties of the zeolite membranes. - Diffusion in zeolite membranes:
Diffusion data can be obtained by a wide variety of different techniques which yield diffusion data which, for the same material, can differ by more than 4 orders of magnitude. So it is outside the scope of this paper to give a full discussion of diffusion data. Some relevant aspects for membrane permeation will be mentioned. A comparison of diffusivities of n-butane in silicalite obtained by different techniques is given by Kapteyn et al. [88]. Compared with his experimental results obtained from steady-state permeation measurement using Eqs. (9.50) and (9.51), values obtained by single crystal (membrane) measurements [94] are too low by more than two orders of magnitude. These single crystal data however reproduce reasonably the permeation results of Vroon et al. [72,74] as discussed in the preceding sections. The diffusion data of Kapteyn et al. agree well with diffusion data obtained by frequency response (FR) and square wave (SW) methods. Kapteyn argues that the diffusivity of n-hexane in silicalite is not influenced by the fact that the crystals in the membrane are intergrown and assumes that the same holds for n-butane. The intrinsic (corrected, Maxwell-Stefan) diffusion coefficient Do of n-butane in silicalite can be described by an Arrhenius equation with a pre-exponenfial coefficient D~ -- 0.053 c m 2 s -1 and an activation energy for diffusion E d = 29.8 kJ mo1-1. At 300 K this gives a value of Do = 4x10 -7 cm 2 s-1. It should be noted that the absence of effects due to intergrown particles does not mean that grain boundaries do not play a role, as has been shown by Vroon [72,74] and discussed in the preceding section. Vroon et al. report values obtained by transient measurements on their silicalite membranes using Eq. (9.61b) and find a good agreement with values obtained from steady-state membrane measurements. Values obtained from transient measurements in sorption experiments on powdered material are two orders of magnitude smaller. Geus et al. [75] reports diffusion data at 21 and 145~ for H2, N2, C H 4 , C O 2 and CF2C12 in silicalite membranes on a clay support which are obtained with the similar transient permeation technique as used above by Vroon. The diffusion coefficients for methane are about two orders of magnitude smaller than those obtained by PF-NMR methods. Usually this last technique gives relatively large diffusion coefficient values, which in the case of n-butane are of the same order of magnitude as reported for FR techniques and membrane techniques as reported by Kapteyn.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
411
Geus ascribes his low values to the influence of the support which has a low porosity. Indeed, uncertainties in geometric aspects of the separation layer and the membrane system affect the value of the measured diffusion coefficients. The conclusion so far must be that synthesis and sample preparation techniques play an important role. Diffusion data to be used in permeation experiments should be measured on membranes with techniques which reflect as closely as possible the transport phenomena during permeation. This also minimises heat effects due to adsorpfion/desorption which play an important role in diffusion experiments based on large crystals, but is of minor importance in membrane experiments [101].
9.4.4 Surface Effects on Permeation in Microporous Membranes In the preceding discussion it was assumed that the transfer from molecules from the gas phase to the solid (porous) membrane was not the rate-determining step in the permeation. This assumption will be evaluated in this section because in oxygen permeation of dense oxidic membranes surface reactions become clearly rate determining for several groups of materials (see also Chapter 10). For the best permeating dense materials (perovskites) with relatively large exchange coefficients, surface reactions become rate determining with membrane thicknesses in the range 0.3-1.0 mm corresponding with flux values in the range 0.4-4.0x10-6 mol m -2 s -1 (corresponding to a permeation value of 0.44.0x10-7 mol m -2 s -1 Pa -1 with pressures of I bar and about zero at the feed and permeate side respectively). This high oxygen permeation is comparable with or somewhat lower than many of the permeation values for microporous membranes. De Lange [63] used a gas kinetic expression to estimate the total number of molecules Zwan colliding per second with the walls of a volume: 1N_
(9.64)
Zwall -" 4 V v
where V is the molar volume, v the mean molecular velocity and N / V is the number of molecules per unit volume. At 1 atm and 300 K the calculated value of Zwall is approximately 1.8 mol c m -2 s -1. A typical hydrogen flux through the microporous membranes is 10x10-6 mol c m -2 s -1 (calculated from a permeation of 1 0 X 1 0 -7 mol m -2 s - 1 P a -1 at a pressure difference of I bar). Not every collision leads to penetration of the molecule into the membrane. This is expressed by the sticking factor t as defined by Eq. (9.65)"
P 1
R a = t [ (2~MRT)I/2 e x p -
E('R-T-/
(9.65)
412
9 a TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
The sticking factor gives the ratio of the number of activated collisions divided by the total number of collisions, whereas Ra in Eq. (9.65) gives the rate of adsorption (in mol cm -2 s-l), with an activation energy Earsfor adsorption at the external surface, the other parameters having their usual meaning. According to Turkdogan [5] the maximum value of R a in Eq.(9.65) is obtained by setting t equal to unity and zero activation (adsorption) energy (Ea = 0). Equation (9.65) then transforms to the classical Hertz-Knudsen equation for the number of moles striking a unit surface area per unit time Rmax: Rmax =
P
2 ~ ( M R T ) 1/2
(9.66a)
and with p given in atmosphere: Rmax - 44.3p
- (MT)I/2
(9.66b)
with Rmax in mol cm -2 s -1. Note that Eqs. (9.66b) and (9.64) will produce figures of similar orders of magnitude and that Eq. (9.66a) also gives the maximum rate of vaporisation from a non-contaminated surface at low pressures. For microporous membranes only the porous part of the surface (~) is available for penetration; the solid is assumed n o t to accept molecules. For small molecules hitting the surface under not too low angles it is reasonable to assume a low value of the activation energy for pore penetration (this is process F1 i n Fig. 9.21). A pessimistic estimate for microporous silica membranes using values of ~ = 0.01 and t = 0.01 yields at 300 K and I atm a collisional flux (of H2) which is at least one order of magnitude larger than the permeation (flux) values found by de Lange et al. [63]. The conclusion is that for relatively small molecules (H2, CO2, etc.), permeation in microporous (silica) membranes is not limited by surface reactions and direct penetration in the pores is the dominant mechanism in a wide range of temperature and pressure conditions [63]. This conclusion does not hold for large non-spherical molecules. Here sorption is necessary, the sticking coefficient becomes very important and surface reactions probably will limit the permeation as soon as bulk permeation becomes appreciable. To the knowledge of the present author, no investigations of this phenomenon in microporous membranes have yet been reported. In dense, non-porous membranes, surface limitations to oxygen permeation are a common phenomenon as can be understood from the very low adsorption levels and large activation energies on the dense membrane materials (see Chapter 10). For hydrogen permeation in dense metal membranes estimates have been made by Govind [105].
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
413
The implication of the theoretical considerations given above is that the permeation can be increased in cases of low adsorption and sticking coefficients by application of a mesoporous top layer with better sorption properties on top of the microporous membranes. Selective sorption should then also lead to an enhanced separation factor (see Eq. (9.71)). Indications for this effect are reported for dense membranes by Deng et al. [106] and for microporous silica membranes by Nair [107]. 9.5 PERMEATION A N D SEPARATION IN MORE COMPLICATED SYSTEMS
Real membrane systems to be used in practice usually do not have the simple architecture assumed in the preceding quantitative treatments (single-wall, non-supported) nor do they fulfil basic boundary conditions, i.e. well mixed gas mixtures, homogeneous gas compositions and pressure (no gradient) across the membrane length (flow direction of feed/permeate). In those cases the aerodynamic conditions of the feed and permeate flow, the precise design and the type of permeate removal (sweep gas, vacuum suction) are important. In the case of supported membranes the effect of the support has always to be evaluated, and if not negligible, corrections for support effects should be applied even with simple membrane architectures. A full description of permeation and separation in practical systems is out of the scope of this paper. Two important cases will be treated for illustration because of their importance for laboratory experiments.
9.5.1 Hollow Fibres In the case of hollow fibres, or long cylindrical tubes, the pressure drop across the membrane length is not negligible. In the case of hollow fibres with a characteristic ratio of length-to-inside-diameter of 104 this pressure drop is very large and the gas densities at inlet and. outlet differ considerably. Then the gas flow is a compressible flow. Shelekhin et al. [56] derived a set of three expressions to describe the permeation of single gases through a micro porous hollow fibre (Vycor type) which, in the general case, should be solved numerically. In the special case of a relatively low permeable gas, the pressure drop along the fibre becomes again negligible and the permeation Fp (mol m -2 s -1Pa -1) can be calculated directly: 9 T . r i In Fp =
S(po-
(ro/ri)
(9.67)
P3)
with (I) T the transmembrane flow rate, ro and ri the outer and inner radii of the fibre, S the membrane surface area (m 2) and P0 and P3 the inlet pressure and the pressure on the permeate side of the membrane, respectively.
414
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
The pressure distribution along the fibre was expressed as P/Po and was calculated with the complete set of equations. Values of P/Po across the complete fibre length (so Po/Pl with Pl is outlet pressure) for He at 30 and 250~ are equal to 0.8 and 0.5 respectively and use of Eq. (9.67) gives wrong results. This is illustrated by a comparison of the permeation versus temperature curves of He which give a maximum if Eq. (9.67) is used but give a continuously decreasing function when the pressure drop is taken into account. In the case of gas mixtures the gas composition of both feed and permeate flows changes along the membrane length. There is a difference in behaviour between co-current and counter-current flow of feed and permeate streams. A brief description for separation in a single stage module with ideal mixing and of a coupling of modules to form cascades or membrane rectification units is given by Eichmann and Werner [18]. It is illustrated that the concentration at the permeate side changes across the membrane length. An implicit expression to calculate the concentration of a binary mixture at the outlet of the membrane system as a function of the inlet concentration is given and so the separation factor can be calculated. This equation gives a good description of the actual behaviour of gases as illustrated e.g. for H2/CO 2 mixtures. The effect of several important parameters (e.g. average pressure, feed or permeate pressure, feed or permeate pressure at outlet, temperature) is illustrated and the necessity to select an optimum set of parameters, given economical boundary conditions, is shown. An extensive treatment of this type of problem is given by Sengupta and Sirkar [114]
9.5.2 Multilayered, Asymmetric Supported Systems The use of supports in asymmetric, supported membranes introduces a number of complications in the interpretation of permeation and separation data as well as in the optimalisation of membrane systems. If the flow resistance of the support is not negligible, there is a pressure drop across the support. This implies that the pressure and so the occupancy at the interface of separation layer and support is different from the (directly accessible) pressure at the support surface, usually the permeate side. Consequently, the driving force for permeation through the separation layer is different from the total driving force across the membrane system. In cases where one wants to calculate or compare transport properties of the separation layer material, it is necessary to correct for this effect (for illustration see below). Expressions to calculate the pressure Pint at the interface of top layer and substrate and thus to calculate the pressure drop across the top layer only are originally derived by Uhlhorn et al. [21] and further developed and used by Lin et al. [103,104] and de Lange et al. [59,60]. More recently Uchytil [102] used and refined this method for different cases. De Lange [60] gives an illustration of the
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
415
calculation on a typical, supported mesoporous y-A1203 membrane. It is assumed that the flow resistance in single gas permeation experiments is a series combination of the flow resistances of support and top layer, respectively, where the permeation is a reciprocal resistance. The support permeation can be expressed as (see also Section 9.2.3.): F0,s = gs + ~ls" Pav
(9.68a)
where gs expresses the Knudsen component, [3s- Pay expresses the Poiseuille (Laminar) flow component and Pav is the average pressure (Phigh- P!o:w). For gs and as gas kinetic expressions can be given [103]. A similar expression to Eq. (9.68) usually fits very well the behaviour of the membrane (F0,m: support + top layer) so: F0,m = g m + ~ m " P a v
(9.68b)
The values of gs and [3s are calculated from measured permeation data for non-adsorbable gases (He, Ar, H2) using Eq. (9.68a). The permeation or permeability properties of the top layer are calculated now by subtracting the permeation data of the support only from the measured permeation data of the membrane using the series model. Note that Ph (high pressure) and P1 are measured at the interfaces of gas/top layer and gas (permeate side) / support respectively. The pressure Pi at the interface of top layer and support can be calculated by
Pi= 1~s -gs +
2 + [3s. P~ + 2gs" P1 + 2
(9.69)
where ~ is the flow rate (tool or m 3 s -1) and A the membrane surface area. The support permeation for the actual experiment is given by: F~
-
A(Pi-
P1)
(9.70)
The theoretical validity of Eq. (9.68b) is discussed by Lin et al. [104] and it is shown that this equation is a special simplified case of a more general, but very complicated expression which strictly holds for the case that ~m/gm = ~s/gs. Uchytill [102] also devotes an extended discussion to this problem. Typical examples of the value of Pi and of the magnitude of the corrections are given in the cited literature. For ~/-A1203 top layers (thickness -- 4 ktm, pore diameter 4 nm) on an ct-A1203 support (thickness -- 2 ram, pore diameter -- 0.2 ktm) Uhlhorn [21] reports a value of Pi = 55 kPa with Ph = 80 kPa and P1 = 7 kPa and H 2 as the permeating gas. This means that in this case only 30% of the total pressure drop is across the 7-A1203 top layer; the remainder is across the support. De Lange et al. [60] applied a
416
9 --TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
microporous silica layer on top of supporting system similar to that used by Uhlhorn. The silica layer has a thickness of about 100 nm and a pore diameter of 0.4-0.5 nm. At 50~ the relative pressure drop across the support is about 4%, while at 250~ this is 15% (with the same gas flow 9 of 7.1x10-6 mol s-1) and the main pressure drop being across the silica top layers. At 250~ the value of Ph = 0.71 bar and of Pi = 0.12 bar. So in this case the correction increases with increasing temperature and with increasing total permeation values due to a decreasing contribution to the permeation of the silica top layer.
Effects of the support on separation The low pressure PI is measured and can be manipulated only at the interface gas(permeate)/support. This implies that when the support resistance is not negligible the value of Pi at the interface of support and top layer can be considerably larger than P1. Especially important is the case with strongly adsorbing gases where even a small increase of Pi c a n lead to a large increase of the occupancy at this interface and consequently to strong effects on the relative permeation contributions (separation) in gas mixtures. According to Eq. (9.38) this also means that the real separation factor of the top layer is decreased with respect to the ideal separation factor by back diffusion from the support (see Sections 9.3.1 and 9.3.2). This becomes especially serious when the conditions are such that the support is in the viscous flow or in transition regime from viscous flow to Knudsen flow. (This means that the support has no or hardly any separation properties itself.) Even relatively small amounts of non-Knudsen contributions in the diffusive transport (which hardly affects the permeation) can decrease the separation factor considerably (see Eqs. (9.38) and (9.34)). This implies that to obtain maximum separation factors the support resistance should be as small as possible and vacuum suction is preferred above use of a sweep gas to remove the permeate (from the permeate side). If the conditions are such that the mesoporous support is in the Knudsen regime, and so has some separation properties, the separation factor can be enhanced when the feed is applied from the support side. In this case the gas composition at the interface between support and separation layer is enriched already somewhat. This effect is reported by Keizer et al. [20] and could be described by the empirical relation: Or
= O~supp- O~oplayer
(9.71)
9.6 OVERVIEW OF I M P O R T A N T RESULTS
In this section a brief overview will be given of the most important results of permeation and separation. It is not the intention to give a complete review of
9 ~ TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
417
all available literature but merely to illustrate the state of the art, to show possibilities and to compare results with porous systems with competing dense membranes.
9.6.1 Introductory Remarks Permeation and separation data reported in the literature are difficult to compare directly. This is due to the variety of parameters which influence the absolute value of permeation and separation data and which are usually badly described and sometimes cannot even be adequately described. As is shown in the preceding sections the pressure conditions and the flow dynamics (aerodynamic conditions) play a very important role. These pressure conditions are not always adequately described and data describing the external flow conditions do not directly reflect flow conditions in the membrane (model design a n d / o r membrane architecture playing a role). Flux data in mol (or m 3) per unit of time and surface area are the preferred data. To obtain data reduction and to make comparison easier permeation (permeance) data are usually given. One should realise however that this is only meaningful if the flux is a linear function of pressure (difference), so in the Henry region. Permeation data given as permeation (permeance) must be accompanied by information concerning the validity region (pressure boundaries) and the form of the pressure dependency. In the latter case this leads generally to a dimension of mol m -2 s -1 Pa -x with 0 < x < 1. A membrane material with a high permeation which is valid only in a small pressure range and which "saturates" at low pressure is inferior compared with a membrane material with lower permeation which is valid in a wide pressure range. Data given in the form of permeability (mol m / m 2 s Pa) are usually meaningful only in symmetric membranes (single, homogeneous wall, non-supported). In asymmetric supported membranes the use of permeability data can give rise to much confusion and erroneous conclusions for several reasons. In most cases the layer thickness is not precisely known and usually it is not known whether this layer is homogeneous or has property gradients (e.g. a "skin" and a more porous part). In many cases the material of the layer penetrates the support to some extent and so it is not possible to separate properties of separation layer and support without giving account of the interface effect. Finally, even if all these complications can be avoided, a comparison based on separation layer properties expressed in terms of permeabilities can give a completely wrong impression of the practical possibilities (as done in e.g. Ref. [109]). This is illustrated by comparison of hydrogen permeabilities of ultra-thin silica layers (see Tables 9.14-9.16) with other materials such as zeolites and metals. The "intrinsic" material properties of these silica layers are not impressive;
418
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
nevertheless m e m b r a n e s give the highest permeation values reported in literature in combination with good separation properties. This is due to a "technology factor", i.e., the possibility to make them extremely thin. This cannot be obtained so far with other materials. Comparison per unit of thickness ~ which is the essence of p e r m e a b i l i t y - gives the impression that equal thicknesses for all materials can actually be obtained, which is not the case. Furthermore, limitations imposed by surface reactions becoming rate limiting at a given thickness are not taken into account.
9.6.2 Typical Permeation and Separation Data for Porous Membranes Most of the data are taken from an overview of Burggraaf [108] which has been u p d a t e d with results reported later. Some typical results obtained with capillary condensation a n d / o r surface diffusion as transport mechanism are given in Table 9.13. A discussion of these data is given in Section 9.2.3.3. As is shown, interesting combinations of high to very high separation factors with reasonable to good flux values can be obtained. Typical results for supported microporous silica m e m b r a n e s are given in Table 9.14 and are partly discussed in Section 9.4.3. The data given by de Lange (see Table 9.14) are all in the Henry regime and the permeation of H 2 and CO2 is in the range of 10-25 and 3-6x10 -2 mol m - 2 s - 1 bar -1 TABLE 9.13 Some typical results with capillary condensation and surface diffusion in meso- and macroporous membranes Membrane
Thickness Pore dia- Separation (~Jm) meter factor (nm)
Permeation (mol m-2 S-1 Pa-1)
T-A1203modified with Ag
10
3--4
H2/N2:8
H2:35x10-6
T-A1203
4
3
C3H6/N2:
T-A1203 m o d i -
C3H6/N2:
fied with MgO T-A1203 5--8
4
T-A1203 + silicate
4-5 0.5
10
26 84
Temp.(~ Ref. Press.(bar)
230 0.1 (H2) C3H6:30x10-6 -10 C3H6:1.6x10-6 -10
Methanol/H2: Methanol: 680 =20x10-6 Methanol/H2: Methanol: 110 =1.6x10-6 Methanol/H2: Methanol: >1000 5.6x10-6
100 2.2 200 23 100 7.7
Ulhorn et al. [28] Ulhom [37] Ulhorn et al. [37] Sperry et al. [39] Sperry et al. [39] Bai/Jai Noble [27]
Permeation is measured at relative pressure p/po where pore is filled (capillary condensation). p0 is condensation (saturation pressure) of free liquid.
9 -- TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
419
TABLE 9.14 Overview of typical flux and separation data of supported microporous silica membranes m a d e by different processes Material / gas
o~
Flux (10-2 tool m -2 s -1)
Temp. (~
SiO2 (microporous) on 3'-(x-A1203 support
Pressure / thickness de Lange [5963] d=100 n m
H2/CO 2
4-10
H2: CO2:
10-25 3-6
200
I bar (-~ small
H2/CH 4 H2/n/i-butane
30--40 160
CH4: i-but:
0.9-2.5 0.3--0.6
200 250
" "
O2/N 2
2-4
02:
2-6
200
"
O2/N2
11.5 4.5
? ?
H2/N2
15.7 4.2
H2: H2:
H2/i-butane
243 40
H2: i-but: i-but:
75 21
1.6 3.2
C3H6/C3H8
(C3H6) (C3I-I6)
25 150
Hassan [95] ?
1 0.3
620 340
Wu [68]
2.7 --0.010 --0.1
300
Pfeed: 0.3-0.8 Pperm:small
300 35 150
C3H 8 (single)
1.5
50-100
Asaeda [64] d < 1 ~tm pfeed: 6 bar Pf = 2 bar
C3H 6 (single)
2.5
50-100
Pperm = 1 bar
respectively in combination with reasonable separation factors. High pressure data up to 20 bar for H 2 (de Lange et al.) indicate the possibility of very high permeation values. H 2 / N 2 mixtures are investigated by Kim and Gavalas [65] using Vycor glass supports with silica deposited partly in and partly on top of the support. At 500~ they report a separation value of c~ equal to 1000 and a permeation of 3.6x10 -3 mol m -2 s-1 bar -1 for H2. More recently Wu et al. [68] improved this method and at 600~ reported (x = 12-16 with a permeation of 1 x 1 0 -2 mol m -2 s -1 bar -1 which is about one order of magnitude smaller than that reported by de Lange. This is partly explained by the rather thick plugs (2.0-2.5 ~tm) of silica, completely deposited within the support pores. Interesting results are reported by Hassan et al. [95] using hollow fibre silica with an estimated pore size of 1.3 nm. For O 2 / N 2 the separation factor c~ = 11.5 at 298 K and c~ = 4.6 at 423 K. Permeation data are not given.
420
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
De Lange (Table 9.14) reports a value of (~ = 2 at 473 K with a permeation of 2-6x10-2 mol m -2 s -1 bar -1, which is reasonable in agreement with Hassan's data considering the strong decrease of c~with increasing temperature. These results indicate that interesting combinations of flux and separation factor in air separation can be obtained with silica membranes. A similar conclusion can be drawn for separation of saturated-unsaturated hydrocarbons as shown by Asaeda et al. [64] for propane-propene (see Table 9.14). In this case permeation values cannot be calculated from the flux values due to non-linear behaviour of the flux as function of pressure. Finally Rao and Sircar [42] report data for microporous carbon layers (thickness 2.0-2.5 ~tm) deposited on carbon supports. For C4H10/H2 mixtures, 0~ equals 94 (at 295 K) which is much larger than the permselectivity (for explanation see Sections 9.4.2 and 9.4.3). The permeability for C4H10 is reported to be 112 Barrer. The author of this chapter recalculated this value to a permeation of 1.4x10 -3 mol m -2 s -1 bar -1. It is, however, very questionable to assume a linear TABLE 9.15a Overview and separation data of typical supported microporous zeolite (MFI) membranes Gas
o~
Flux (10-2 tool m -2 s -1)
Temp. (~
Pressure/ thickness 50 kPa" 50 kPa
H2/CO 2
1-2
H2: CO2:
2-3 1.7
200-350
d=50 ~tm (Bakker) stainless support
1.25
CO2: H2:
0.1 0.12
200
d=3.0 ~tm (Vroon) A1203 support
CO2/H 2
10
H2: CO2:
0.18 1.8
25
d=50 ~tm (Bakker) [71]
14
CO2:
0.15--0.5
25
d=3.0 ~tm
H2/CH 4
low 1.9
CH4: CH4:
? 0.5-0.7
200 200
Bakker [71] Vroon [72-74]
H2/n-butane
2.5 1.0 >100
n-butane: n-butane: n-butane:
1.0 1.5 0.5
350 200 25
Bakker Bakker Bakker
n/i-butane
50 27
n-butane: ?
0.2
25 25
Vroon Bakker
O2/N2 p / o xylene
1 1 =60 25
02: p-xylene 3.5x10 --6 3.5x10 -6
0.5-0.35
25-200 25 100-150 200
Vroon Vroon [72-74] 0.36:0.26 kPa 0.36:0.26 kPa
25 200
Vroon 4.6:4.6 kPa 4.6:4.6 kPa
n-butane/H2
Benzene / cyclohex. 5 4.5
2.6x10 -7 18x10 -7 (benzene)
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
421
TABLE 19.15b Overview of typical flux and separation data of supported zeolite (MFI) membranes Material
0~
Flux (10-2 tool m -2 s-1)
Temp. (~
p / m xylene (triisopropyl benzene)
15
35 ml m -2 h* (m-xylene)
25
n-butane i-butane
-
5.5** 4.5
25 25
CH3OH/CH4
190
1.7 (CH3OH) 0.001 (CI-~
100
CH3OH/CH 4
29
1.8 (CH3OH) 0.007 (CH4)
100
n/i-butane
3-6
?
?
Pressure / thickness Xiang [76]: Pfeed" 17 bar Pperm: 10 Torr ratio 1:1:1 Pfeed' I bar Pperrn' zero Jia [27,97]: Ptotal:1100kPa Pfeed(CH3OH): 440-220 Ptotal:1500kPa Pfeed (CH3OH): 400-165 kPa ? thickness: 10.0 ~tm
*Not clear whether this is ml liquid or gas. **Calculated from given permeance assuming linear relation with partial pressure (=40x10-2. relationship b e t w e e n the flux a n d the p r e s s u r e for b u t a n e at 295 K. In other m i c r o p o r o u s s y s t e m s such as silica a n d zeolites this is not the case. In a n o t h e r p a p e r Rao a n d Sircar [Gas Separation and Purification (1993) 279-284] r e p o r t e d for C O 2 / H 2 m i x t u r e s at 296 K, c~ = 5 (Phigh = 2.36 bar) to 0~ = 20 (Phigh = 3.7 bar) w i t h a p e r m e a b i l i t y of 1200 Barrer in b o t h cases. This indicates an increase in the s e p a r a t i o n factor w i t h p r e s s u r e , w h e r e a s the p e r m e a t i o n r e m a i n s constant. Typical results for zeolite (MFS) m e m b r a n e s are collected in Table 9.15 a n d p a r t l y d i s c u s s e d in Section 9.4.3. As is s h o w n in the table the s e p a r a t i o n factor of m i x t u r e s of w e a k l y a n d s t r o n g l y a d s o r b i n g gases (see Section 9.4.3) s h o w s a c o n v e r s i o n as a function of t e m p e r a t u r e . Interesting s e p a r a t i o n values can be o b t a i n e d for C O 2 / H 2 a n d n - b u t a n e / H 2 m i x t u r e s at low t e m p e r a t u r e . These are in the s a m e r a n g e as those o b t a i n e d in carbon m e m b r a n e s ; for the flux v a l u e s a similar conclusion holds. I s o m e r s e p a r a t i o n is d e m o n s t r a t e d b y several a u t h o r s (see Table 9.15). G o o d s e p a r a t i o n factors (27-50) are r e p o r t e d for m i x t u r e s of n- a n d i s o b u t a n e b y V r o o n et al. [72,74] a n d Bakker et al. [71] with, h o w e v e r , m o d e s t flux values. S e p a r a t i o n of para- f r o m ortho-xylene is r e p o r t e d b y V r o o n et al. [72,74] w i t h c~ equals 60 in the t e m p e r a t u r e r a n g e of 100-150~ a n d (x = 25 at 200~ a n d a flux of 3.5x10 -6 m o l / m -2 s q for the fastest p e r m e a t i n g p-xylene (100-150~ with h o w e v e r a v e r y small d r i v i n g force.
422
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
The partial pressure at feed side (high pressure side) is only 0.36 kPa. Using higher partial pressures and increasing the temperature might bring the flux in the range 10-3-10-4 m o l / m -1 s-1. Xiang and Ma [76] reported results for mixtures of para- and meta xylene. At room temperature the value of (x equals 15 with a flux for m-xylene of 35 c m 3 m -2 h -1. Assuming that the permeating gas volume is expressed a s c m 3 gas under standard conditions (this is not defined) this permeation value corresponds to -- 8x10-s mol m -2 s -1 for m-xylene and with 3x10 -7 mol m -2 s-1 for p-xylene which is one order of magnitude smaller compared to values reported by Vroon et al.
9.6.3 Comparison of Permeation and Separation Data of Porous and Dense Membranes Typical data for dense membranes are collected in Table 9.16. A full discussion of these data is outside the scope of this chapter. Using permeation values the reader should be aware of the fact that the pressure dependence of the flux is usually strongly non-linear, but takes the form of a power law with values for the exponent around 0.5. This makes direct comparison on the basis of permeance or permeability not meaningful. Furthermore, the permeation value is limited by surface reactions with a critical thickness varying between 0.1 and 2 mm depending on material and condition. Finally, dense (i.e. non-porous) membranes permeate 02 o r H 2 o n l y and so are important only in applications where these gases play a role such as in air TABLE 9.16a C o m p a r i s o n of typical flux data of microporous and d e n s e m e m b r a n e s Hydrogen
Permeation (mol m -2 s-1)
Temp. (~
SiO2 a m o r p h o u s silica (measured)
6-20x10 -2
25-250
Calculated
>300x10 -2
Zeolite (silicalite) on steel
1-3x10 -2
100--400
Bakker et al. [71] Thickness 50 ~tm
Zeolite (silicalite) on alumina
0.5-0.85x10 -2
25-250
Vroon et al. [72-74] thickness 3-4 ~tm AP = 1 atm. ( 1 ~ 0 )
Pd resp. P d / A g films on alumina
3.0-4.5x10 -2
400-900
A r m o r [115]: AP = 2 bar; H2 thickness: 4.5; resp. 22 ~tm
Pd film
0.1x10 -2
100
N a g a m o t o [116] AP = I bar H 2
Pd film within pores of 0r
10-40x10 -2 0c > 1000 H2/N2
300
Y a n / M o r o o k a [113]; AP = 1 bar H 2 thickness 2 p m
AP = 1 bar (1---~0) AP> 1 5 b a r
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
423
TABLE 9.16b Comparison of typical flux data of microporous and dense membranes Oxygen
Flux (mol m -2 s-1)
Temp. (~
La0.6Sr0.4Co3_~
4.0x10-2
900
Teraok~ [110]
La0.3Sr0.?Co3_8
0.3-0.4x10 -2
900
v. Doom/Bouwmeester [119] Thickness I mm air vs. 10-2 bar
Y0.05BaCo0.9bO3.6
0.4x10- 2
900
Brinkman et al. [118] air vs. 10-2 bar
La0.2Sr0.8Fe0.6Co0.403-8
0.2x10-2
850
Balachandral [117] Thickness 0.25-1.2 mm air vs. CH4/H2 (4:1)
ZY-Pd (40 vol%)
0.1-0.2x10 -2 0.2-0.5x10 -3
1100 900
0.6x10-2
1100
Chen et al. [112] air vs. 10-2 bar Thickness 0.5 mm air vs. CO/CO2
BiEr-Au (40 vol%) BiEr-Ag (40 vol%) BiEr-Ag (40 vol%)
0.68x10-3 0.17x10-2 0.85x10-3
850 850 750
Chen et al. [112] Thickness 1-1.5 mm air vs. --2x10-2bar
BiY-Ag (35 vol%)
1.0x10-2
750
Shen et al. [111] Thickness 90 ~tm air vs 6x10-5 bar
si02 microporous film on alumina
2.0-5.0x10 -2
35-200
de Lange [59--63] thickness 100 nm AP = I bar
s e p a r a t i o n a n d d e h y d r o g e n a t i o n or p a r t i a l o x i d a t i o n r e a c t i o n s in m e m b r a n e reactors. A s is s h o w n in T a b l e 9.16 a n d b y c o m p a r i s o n of T a b l e 9.16 w i t h T a b l e 9.14 t h e v a l u e s of o b t a i n a b l e s e p a r a t i o n f a c t o r s of m i c r o p o r o u s m e m b r a n e s is m u c h l o w e r t h a n t h o s e o b t a i n e d w i t h d e n s e m e m b r a n e s ( w h i c h s h o u l d b e i n f i n i t e in t h e case of c o m p l e t e l y d e f e c t - f r e e d e n s e m e m b r a n e s ) . V e r y r e c e n t l y o x y g e n p e r m e a t i o n v a l u e s r e p o r t e d b y S h e n et al. [111], C h e n et al. [112] a n d T e r a o k a et al. [110] s h o w t h a t t h e o b t a i n a b l e flux v a l u e s at high
424
9 - - TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D V A P O U R S
temperature (>600~ are at least a factor of 5-10 lower than those obtainable with microporous membranes at ambient or somewhat increased temperature (200~ For hydrogen a similar situation exists, except for the results reported by Yan and Morooka [113]. In this case the flux data are comparable with those obtained by de Lange et al. but with c~> 1000 for H2/N2 mixtures.
9.7 CONCLUSIONS AND EVALUATION A general description of gas transport properties of inorganic membranes with complex architecture and for multicomponent gas mixtures is not yet available. Quantitative descriptions based on phenomenological (thermodynamic) equations a n d / o r microscopic models can be given in a number of limiting cases like single gases or binary gas mixtures and single wall, unsupported membranes or small plate shaped, asymmetric supported ceramic membranes. In the latter case the support properties are important and must be taken into account in the description of the membrane system and of the separating top layer. In mesoporous membranes the maximum obtainable separation factor for non-condensable gases is limited to the Knudsen separation factor. For adsorbing gases below their critical point, surface flow can play an important role and high values of the permeation and of the separation factor can be obtained in some cases up to temperatures of 300~ In the case of macro- and mesoporous supports their flow resistance should be as small as possible. If the transport resistance is not negligible corrections must be applied in the study of the separation properties of the separating layers. It is shown that even small pressure gradients across the support can cause a considerable decrease of the permeation and of the separation factor of the top layer, especially in the case of adsorbing gases. The absolute value of the permeate pressure is important in addition to the pressure ratio of feed and permeate streams. Increasing support resistance causes an increase of the permeate pressure on the interface between support and separation (top) layer in the case of supported membranes. High separation factors can be obtained with microporous membranes with a pore diameter smaller than 2 nm and are realised with carbon, silica and zeolite membrane systems. The description of these systems is still in its infancy. In some cases reasonable agreement is obtained between calculated and measured permeation and separation properties. Permeation values of a single gas and of that gas in a mixture are generally different and so the separation factor of binary mixtures and the permselectivity (ratio of single gas permeation values) is also different.
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
425
The permeation of a gas is strongly affected by the sorption properties of the combination of gas and membrane and by the ratio of the molecular diameter of the gas molecule and the pore diameter. Mixtures must be classified on the basis of these two properties and the transport properties of these classes differ considerably. The highest separation factors are obtained in the case of: (i) mixtures of strongly (S) and weakly (W) adsorbing gases at intermediate temperature and pressure values, and (ii) the size exclusion regime; here one of the gases in the mixture has a molecular diameter which is larger than the pore diameter. Typical values for permeation and separation factors of microporous membranes are given in Tables 9.14 and 9.15. A comparison is also made with dense membranes in Table 9.16.
LIST OF SYMBOLS
A A* b B0 B* sat
Ci C dp
D D* Ed
;Co: F Fp
gs
J k K
Kn L M Mx,M* FIk
Surface area (m 2) Coefficient in Eq. (9.34). Subscript o: per mol. absorbed Henry constant (mol/kg Pa) Permeation coefficient (m) Coefficient in Eq. (9.34). Subscript o: per mol. absorbed Saturation concentration in material (mol/kg or mol / m 3) Concentration. Subscripts: s, surface; sat, saturated Pore diameter Diffusion coefficient (m2/s) Pre-exponential coeff, in Arrhenius equation Activation energy for diffusion Coefficient in Eq. (9.34) Permeation (mol/m2 s Pa) Permeability (mol m / m 2 s Pa) Fitting parameter in Eq. (9.65) Molar flux (mol/m 2 s). Subscripts" v, viscous; k, Knudsen; c, capillary condensation Correction term in Eq. (9.41b) Langmuir constant (Pa -1) Knudsen number Thickness (m) Molecular mass or molecular weight (kg/mol) Eq. (9.34g-h)) Fitting parameter in Eq. (9.34a) (m -2)
426
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
P P Pd Pr q Q Qa AQa Y Ra
R Sc
t U V
Vm x Z Zwall
Partial pressure (Pa) Total pressure (Pa) Dimensionless pressure: P/Pref Ratio Plow/Phigh Amount absorbed gas (mol/kg) or (mol/m 3) Molar flow (tool/s) Heat of absorbtion (kJ/mol) Activation energy for surface diffusion (kJ/mol) Pore radius (m) or particle radius (m) Rate of adsorption Eq. (9.65) Gas constant (8.314 J/mol K) Stage cut (Qp/Qf) Sticking factor, defined by Eq. (9.65) Potential energy (kJ/mol). Subscript r: relative Molecular velocit~ (m/s) Molar volume (m/mol) Mol. fraction Distance coordinate (m) Number of molecular collisions with the walls of a volume ( c m - 3 S -1 )
Greek letters
F
0
Ok
V
Gs
Separation factor. Subscript 0: ideal separation factor (Eqs. (9.36) or (9.38)) Fitting parameter in Eq. (9.34b) (-) Fitting parameter in Eq. (9.68) Affinity coefficient in Eq. (9.56) (J/mol) Thermodynamic factor (-), defined in Eq. (9.40) Porosity (-) Dynamic viscosity (Pa s) Occupancy (c/cs) (-) Reflection factor in Eqs. (9.6) and (9.9) Molecular mean free path length (m) Geometric constant of pore structure (-) Jump probability, Eq. (9.57) Collision diameter (m 2) Surface tension (J/ m 2) Tortuosity (-) Volume flow (m3/s) or mol. flow (mol/s) Contact angle (-)
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
427
REFERENCES
1. R.E.Cunningham and R.J.J. Williams, Diffusion in Gases and Porous Media. Plenum Press, New York 1980. 2. F.A.L. Dullien, Porous Media: Fluid Transport and Pore Structure. Academic Press, New York, 1979. 3. J. Karger and D.M. Ruthven, Diffusion in Zeolites and Other Microporous Solids. Wiley, New York 1992. 4. J.O. Hirschfelder, C.F. Kurtiss and R.B. Bird, Molecular Theory of Ga~es and Liquids. Wiley, New York 1965. 5. E.T. Turkdogan, Physical Chemistry of High Temperature Technology. Academic Press, New York, 1980. 6. R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1990. 7. E. Wicke and R. Kallenbach, Koltoid Z., 97 (1941) 34-41. 8. J. Rouquerol, D. Avnir, C.W. Fairbridge, D.H. Everett, J.H. Haynes, N. Pernicone, J.D.F. Ramsay, K.S.W. Sing and K.K. Unger, Recommendations for the characterisation of porous solids, Pure Appl. Chem.,66 (1994) 1739-1758. 9. A.J. Burggraaf and K. Keizer, Synthesis of inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes: Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991, pp. 10-63. 10. R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport Phenomena. Wiley, New York 1960. 11. (a) E.A. Mason and A.P. Malinauskas, Gas Transport in Porous Media: The Dusty Gas Model. Chem. Eng. Mon., 17. Elsevier, Amsterdam, 1983. (b) R.B. Evans, G.M. Watson and E.A. Mason, Gaseous diffusion in porous media at uniform pressure, ]. Chem. Phys., 35 (1961) 2076-2083; Effect of pressure gradients, I. Chem. Phys., 36 (1962) 1894-1902. 12. E.A. Mason, A.P. Malinauskas and R.B. Evans, Flow and diffusion of gases in porous media. J. Chem. Phys., 46 (1967) 3199-3216. 13. H.P.G. Darcy, Les Fontaines Publiques de la Ville Dyon. Victor Dalmont, Paris, 1856. 14. J.L. Poiseuille, Comptes Rendus, 11 (1840) 961. 15. M. Knudsen, Ann. Physik (Leipzig), 28 (1909) 75. 16. R.W. Schofield, A.G. Fane and C.J.D. Fell, Gas and vapour transport through microporous membranes. 1, Knudsen-Poiseuille transition. J. Membr. Sci., 53 (1990) 159-171. 17. (a) A. Leenaars and A.J. Burggraaf, The preparation and characterisation of alumina membranes with ultra fine pores. Part 3: The permeability for pure liquids. J. Membr. Sci., 24 (1985) 245-260. (b) A. Leenaars and A.J. Burggraaf, Part 2: The formation of supported membranes. I. Colloid Interface Sci., 105 (1985) 27-40. 18. W. Eickmann and U. Werner, Gas separation using porous membranes. Ger. Chem. Eng., 8 (1985) 186-194. 19. J.C.S. Wu, D.F. Flowers and P.K.T. Liu, High temperature separation of binary gas mixtures using microporous membranes. ]. Membr. Sci., 77 (1993) 85-98. 20. K. Keizer, R.J.R. Uhlhorn, R.J. van Vuren and A.J. Burggraaf, Gas separation mechanisms in microporous modified ~'A1203membranes. ]. Membr. Sci., 39 (1988) 285-300. 21. (a) R.J.R. Uhlhorn, Ceramic membranes for gas separation. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1990.
428
22. 23. 24.
25. 26. 27. 28. 29.
30. 31. 32. 33. 34. 35. 36. 37.
38.
39. 39a. 40.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESAND VAPOURS (b) R.J.R. Uhlhom and A.J. Burggraaf, Gas separation with inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes. Van Nostrand Reinhold, New York 1990, pp. 155-176. R.L. Rowell, S.A. Carrano, A.J. de Bethune and A.P. Malinauskas, Gas and vapor permeability. ]. Colloid Interface Sci., 37 (1971) 242-246. R.W. Tock and K. Kammermeijer, Temperature-separation factor relationship in gaseous diffusion. AIChE J., 15 (1969) 715-718. K. Haraya, Y. Shindo, T. Hakuta and H. Yoshitome, Separation of H2-CO2 mixtures with porous glass membranes in the intermediate flow region. J. Chem. Eng. Jpn., 19 (1986) 186-190. Y. Shindo, T. Hakuta, H. Yoshitome and H. Inoue, Separation of gases by means of a porous glass membrane at high temperature. ]. Chem. Eng. Jpn., 17 (1984) 650-652. M. Okazaki, H. Tamon and R. Toei, Interpretation of surface flow phenomena of adsorbed gases by a hopping model, AIChE ]., 27 (1981) 262-270. C. Bai, M. Dong Jia, J.L. Falconer and R.D. Noble, Preparation and separation properties of silicalite composite membranes. ]. Membr. Sc/., 105 (1995) 79-89. R.J.R. Uhlhom, K. Keizer and A.J. Burggraaf, Gas and surface diffusion in modified Talumina systems. ]. Membr. Sci., 46 (1989) 225-241. (a) H.J. Sloot, A non-permselective membrane reactor for catalytic gas phase reaction. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1991. (b) H.J. Sloot, C.A. Smolders, W.P.M. van Swaaij and G.P. Versteeg, Surface diffusion of hydrogen sulfide and sulfur dioxide in alumina membranes in the continum regime. J. Membr. Sci., 74 (1992) 263-278. E.A. Flood and M. Huber, Can. J. Chem., 33 (1955) 203. E.R. Gilliland, R.F. Baddour and J.L. Russel, Rates of flow through microporous membranes. AIChE J., 4 (1958) 90-96. E.R. Gilliland, R.F. Baddour, G.P. Perkinson and K.J. Sladek, Diffusion on surfaces. Ind. Eng. Fundam., 13 (1974) 100-105. H. Tamon, S. Kyotani, H. Wadu, M. Okazaki and R. Toei, Surface flow phenomenon of adsorbed gases on activated ~-alumina. ]. Chem. Eng. Jpn., 14 (1981) 136-141. H. Tamon, M. Okazaki and R. Toei, Flow mechanism of adsorbate through porous media in capillary condensation. AIChE J., 27 (1981) 271-277. K.H. Lee and S.T. Hwang, The transport of condensable vapors through a microporous Vycor glass membrane. ]. Colloid Interface Sci., 110 (1986) 544-555. H. Rhim and S.T. Hwang, Transport of capillary condensate. J. Colloid Interface Sci., 52 (1975) 174-181. R.J.R.Uhlhom, K. Keizer and A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part 1: Multilayer diffusion and capillary condensation. J. Membr. Sci., 66 (1992) 259-269. M. Asaeda and L.D. Du, Separation of alcohol/water gaseous mixtures by thin ceramic membranes. I. Chem. Eng. ]pn., 19 (1986) 72-77. (a) M. Asaeda, L.D. Du and M. Fuji, Separation of alcohol/water gaseous mixtures by an improved ceramic membrane. J. Chem. Eng. Jpn., 19 (1986) 84-86. D.P. Sperry, J.L. Falconer and R.D. Noble, Methanol-hydrogen separation by capillary condensation in inorganic membranes. J. Membr. Sci., 60 (1991) 185-193. B. Abeles, L.F. Chen, J.W. Johnson and J.M. Drake, Capillary condensation and surface flow in microporous Vycor glass. Israel J. Chem., 31 (1991) 106. T. Okubo and H. Inoue, Introduction of specific gas selectivity to porous glass mem-
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
429
branes by treatment with tetra ethoxy silane. J. Membr. Sci., 42 (1989) 109-117. 41. T. Okubo and H. Inoue, Single gas permeation through porous glass modified with tetra ethoxysilane. AIChE ]., 35 (1989) 845-848. 42. M.B. Rao and S. Sircar, Nanoporous carbon membranes for separation of gas mixtures by selective surface flow. J. Membr. Sci., 85 (1993) 253-265. 43. L.D. Rothfeld, Gaseous conterdiffusion in catalyst pellets. AIChE J., 9 (1963) 19-24. 44. N. Wakao, J. Otani and J.M. Smith, Significance of pressure gradients in porous materials, 1. Diffusion and flow in fine capillaries; 2. Diffusion and flow in porous catalysts. AIChE l., 11 (1965) 435-439; 439-445. 45. D.S. Scott and F.A.L. Dullien, Diffusion of ideal gases in capillaries and porous solids. AIChE J., 8 (1962) 113-117, 293. 46. (a) J.W. Veldsink, A catalytically active, non permselective membrane reactor for kinetically fast, strongly exothermic heterogeneous reactions. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1993. (b) J.W. Veldsink, G.F. Versteeg and W.P.M. van Swaaij, An experimental study of diffusion and convection of multicomponent gases through catalytically and non catalytically membranes. ]. Membr. Sci., 92 (1994) 275-291. 47. R. Taylor and R. Krishna, Multicomponent Mass Transfer. Wiley, New York, 1993. 48. R.D. Present and A.J. De Bethune, Separation of a gas mixture through a long tube at low pressure. Phys. Rev., 75 (1949) 1050-1057. 49. J.J.M. Oude Egberink, Gas transport in a porous medium. M.Sc. Thesis, Faculty of Chemical Technology, University of Twente, Enschede, The Netherlands, 1993. 50. K. Keizer, R.J.R. Uhlhorn and A.J. Burggraaf, Gas separation using inorganic membranes, in: R.D. Noble and S.A. Stern (Eds.), Membrane Separations Technology: Principles and Applications. Elsevier, Amsterdam, 1994, pp. 553-588. 51. R.J.R, Uhlhorn, V.T. Zaspalis, K. Keizer and A.J. Burggraaf, Synthesis of ceramic membranes. Part II: Modification of alumina thin films: reservoir method. J. Mater. Sci., 27 (1992) 538-552. 52. S. Kitao and M. Asaeda, Separation of organic acid-water mixtures by thin porous silica-membranes. ]. Chem. Eng. Jpn., 23 (1990) 367. 53. S. Kitao, M. Ishizaki and M. Asaeda, Permeation mechanisms of water through fine porous ceramic membranes for separation of organic solvent-water mixtures. Key Eng. Mater., 61 & 62 (1991) 175-180. 54. J.E. Koresh and A. Softer, The carbon molecular sieve membranes. General properties and the permeability of CH4/H2 mixtures. Sep. Sci. Technol., 22 (1987) 972-982. 55. V.M. Linkov, R.D. Sanderson and E.P, Jacobs, Preparation of inorganic hollow-fibre membrane and composite hollow-fibre carbon membranes, in: Y.H. Ma (Ed.), Proceed-
ings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 471-481. 56. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. II. Permeation of gases in microporous glass membranes. ]. Membr. Sci., 75 (1992) 233-244. 57. R.J.R.Uhlhorn, M.H.B.J. Huis in 't Veld, K. Keizer and A.J. Burggraaf, High permselectivities of microporous silica modified ~'alumina membranes. J. Mater. Sci. Lett., 8 (1989) 1135-1138. 58. R.J.R.Uhlhorn, K. Keizer and A.J. Burggraaf, Gas transport and separation with ceramic membranes. II. Synthesis and separation properties of microporous membranes.
430
9 -- TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESAND VAPOURS
J. Membr. Sci., 66 (1992) 271-287. 59. R.S.A. de Lange, Microporous sol-gel derived ceramic membranes for gas separation synthesis, gas transport and separation properties. Ph.D. Thesis, 1993, University of Twente, Enschede, The Netherlands. 60. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterisation of supported microporous ceramic membranes prepared by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57-75. 61. R.S.A.de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Permeation and separation studies on microporous sol-gel modified ceramic membranes. Microporous Mater., 4 (1995) 169-186. 62. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Polymeric silica based sols for membrane modification applications sol-gel synthesis and characterisation with SAXS. ]. Non-Cryst. Solids, 191 (1995) 1-16. 63. R.S.A. de Lange, K. Keizer and A.J. Burggraaf, Analysis and theory of gas transport in microporous sol-gel derived ceramic membranes. J. Membr. Sci., 104 (1995) 81-100. 64. M. Asaeda, A. Yamamichi, M. Satoh and M. Kamakura, Preparation of porous silica membranes for separation of propylene/propane gaseous mixtures, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), 10-14 July 1994, Worcester MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd., Worcester, MA 01609, USA. pp. 315-325. 65. M. Tsapatsis and G. Gavalas, Structure and aging characteristics of H2 permselective SiO2-Vycor membranes. ]. Membr. Sci., 87 (1994) 281-296. 66. Heung Yong Ha, Suk Woo Nam, Seong Ahn Hong and Won Kook Lee, Chemical vapor deposition of hydrogen permselective silica films on porous glass supports from tetra ethyl orthosilicate. J. Membr. Sci., 85 (1993) 279-291. 67. C.L. Lin, D.L. Flowers and P.K.T. Liu, Characterisation of ceramic membranes II. Modified commercial membranes with pore size under 40 ~. J. Membr. Sci., 92 (1994) 45-48. 68. J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers and P.K.T. Liu, Characterisation of hydrogen permselective microporous ceramic membranes. J. Membr. Sci., 96 (1994) 275-287. 69. E.R. Geus, Preparation and characterisation of composite inorganic zeolite membranes with molecular sieve properties. Ph.D Thesis, 1993, Technical University of Delft, Delft, The Netherlands. 70. E.R.Geus, H. van Bekkum, W.J.W. Bakker and J.A. Moulijn, High temperature stainless steel supported zeolite (MFI) membranes: preparation, module construction and permeation experiments. Microporous Mater., 1 (1993) 131-147. 71. W.J.W. Bakker, G. Zheng, F. Kapteyn, M. Makkee and J.A. Moulijn, Single and multicomponent transport through metal supported MFI zeolite membranes, in: M.P.C. Weynen and A.A.H. Drinkenburg (Eds.), Precision Process Technology. Kluwer, 1993, pp. 425-434. 72. Z.A.E.P. Vroon, Synthesis and transport properties of thin ceramic zeolite MFI membranes. Ph.D thesis, 1995, University of Twente, Enschede, The Netherlands. 73. Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Transport properties of alkanes through ceramic thin zeolite MFI membranes. J. Membr. Sci., 113 (1996) 293-300. 74. (a) Z.A.E.P. Vroon, K. Keizer and A.J. Burggraaf, Preparation and characterization of thin zeolite MFI membranes. J. Membr. Sci., in prep.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
75.
76.
77. 78.
79.
80. 81. 82. 83. 84. 85. 86.
87.
88.
89.
90.
431
(b) Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Multi-component permeation through thin zeolite MFI membranes. Microporous Mater., in prep. (c) Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Single gas permeation through thin zeolite MFI membranes. Microporous Mater., in prep. E.R. Geus, M.J. den Exter and H. van Bekkum, Synthesis and characterisation of zeolite (MFI) membranes on porous ceramic supports. ]. Chem. Soc. Faraday Trans., 88 (1992) 3101-3109. Shouhe Xiang and Yi Hua Ma, Formation and characterisation of zeolite membranes from sols, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 95-105. Meng Dong Jia, Baoshu Chen, R.D. Noble and J.L. Falconer, Separation of vapor/gas mixtures using ceramic zeolite composite membranes. J. Membr. Sci.,90 (1994) 1-10. M. Matsukata, N. Nishiyama and K. Keyama, Preparation of a thin zeolite membrane, in: J. Weitkamp, H.G. Karge, H. Pfeiffer and W. H61derich (Eds.), Zeolites and Related Microporous Materials: State of the Art 1994. Studies in Surface Science and Catalysis, Vol. 84, Elsevier, Amsterdam, 1994, pp. 1183-1190. A.J. Burggraaf, K. Keizer, R.S.A. de Lange, Z.A.E.P. Vroon and V.T. Zaspalis, Ceramic membranes for separation and reactions, in: R. von Ballmoos (Ed.), Proceedings of the 9th International Zeolite Conference, Montreal 1992. Butterworth-Heinemann/Reed, USA, 1993, pp. 47-70. D.W. Breck, Zeolite Molecular Sieves: Structure, Chemistry and Use. Wiley, New York, 1974. W.M. Meier and D.H. Olson, Atlas of Zeolite Structures, 3rd Edn., Butterworth Heinemann, 1992. H. van Bekkum, E.M. Flanigan and J.C. Jansen (Eds.), Introduction to Zeolite Science and Practice. Studies in Surface Science and Catalysis, Vol. 58, Elsevier, Amsterdam, 1991. J.H. Petropoulos and J.K. Petrou, Simulation of molecular transport in pores and pore networks. J. Chem. Soc. Faraday Trans., 87 (1991) 2017-2022. D.E. Everett and J.C. Powl, Adsorption in slit-like and cylindrical pores in the Henry's law region. J. Chem. Soc. Faraday Trans., 72 (1976) 619. R.M. Barrer, Porous crystal membranes. J. Chem. Soc. Faraday Trans., 86 (1990) 11231130. (a) Jirong Xiao and J. Wei, Diffusion mechanism of hydrocarbons in zeolites. I. Theory. Chem. Eng. Sci., 47 (1992) 1123-1141. (b) Jirong Xiao and J. Wei, Diffusion mechanism of hydrocarbons in zeolites. II. Analysis of experimental observations. Chem. Eng. Sci., 47 (1992) 1143-1159. (a) R. Krishna, A unified approach to modelling of intraparticle diffusion in adsorption processes. Gas Sep. Purif., 7 (1993) 91-104. (b) R. Krishna and J.A. Wesseling, Mass Transfer. Ellis Horwood, New York, 1990. F. Kapteyn, W.J.W. Bakker, Guhong Zheng and J.A. Moulijn, Temperature and occupancy dependent diffusion of n-butane through a silicalite 1 membrane. Microporous Mater., 3 (1994) 227-234. F. Kapteyn, W.J.W. Bakker, Guhong Zheng, J. Poppe and J.A. Moulijn, Permeation and separation of light hydrocarbons through a silicalite-1 membrane. Application of the generalised Maxwell Stefan equations. Chem. Eng. J., 57 (1995) 145-153. H. Abdul Rahman, M.A. Hasanain and K.F. Loughlin, Quaternary, ternary, binary and pure component sorption on zeolites. Ind. Eng. Chem. Res., 29 (1990) 1525.
432
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
91. E.J. Magim, A.T. Bell and D.N. Theodoru, Transport diffusivity of methane in silicalite from equilibrium and non-equilibrium situation. J. Phys. Chem., 97 (1993) 4173. 92. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular sieve membranes. AIChE J., 41 (1995) 58-67. 93. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer, A.J. Burggraaf and Y.H. Ma, Gas transport in microporous sol-gel derived ceramic membranes. II. Sorpfion studies. J. Porous Mater., 2 (1995) 141-149. 94. (a) D.T. Hayhurst and A.R. Paravar, Diffusion of C1-C5 normal paraffins in silicalite. Zeolites, 8 (1988) 27. (b) D.B. Shah, S. Chokchai and D.T. Hayhurst, Measurements of transport rates of C4 hydrocarbons across a single crystal silicalite membrane. J. Chem. Soc. Faraday Trans., 89 (1993) 3161-3167. 95. M. Hassan, J.D. Way, P.M. Thoen and A.C. Dillon, Separation of gas mixtures using hollow fiber silica membranes, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 325-335. 96. Yi Hua Ma and Shouhe Xia, Formation of zeolite membranes from sols. Patent application WP191-04 dated 3 Dec. 1992. 97. Meng Dong Jia, B. Chen, R.D. Noble and J.L. Falconer, Ceramic zeolite composite membranes and their application for separation of vapor/gas mixtures. J. Membr. Sci., 90 (1994) 1-10. 98. Meng Dong Jia, K.V. Peinemann and R.D. Behling, Ceramic zeolite composite membrane: preparation, characterisation and gas permeation. J. Membr. Sci., 82 (1993) 15-26. 99. F. Kapteyn, W.J.W. Bakker, G. Zheng, J.A. Moulijn and H. van Bekkum, Permeation and separation behaviour of a silicalite (MFI) membrane, in: 1st International Workshop on
Catalytic Membranes, 26-28 September, 1994, Lyon, France. 100. R. Krishna and L.P.J. van der Broeke, The Maxwell-Stefan description of Mass transport across zeolite membranes. Chem. Eng. J., 57 (1995) 155-162. 101. D.M. Ruthven, Lecture: Diffusion in zeolites. 4th International Symposium European Science and Engineering Program on Advanced Separation Science and Engineering, 4-6 October 1994, Leuven, Belgium. 102. P. Uchytil, Gas permeation in ceramic membranes. Part 1, Theory and testing of ceramic membranes. J. Membr. Sci., 97 (1994) 139-144. 103. Y.S. Lin and A.J. Burggraaf, CVD of solid oxides in porous substrates for ceramic membrane modification. AIChE J., 9 (1992) 251-263. 104. Y.S. Lin and A.J. Burggraaf, Experimental studies in pore size change of porous ceramic membranes after modification. J. Membr. Sci., 79 (1993) 65-82. 105. R. Govind, Future prospects and application for metal based membranes. Key Eng. Mater., 91/92 (1991) 319. 106. Huiming Deng, Minyan Zhou and B. Abeles, Diffusion reaction in mixed-electronic solid oxide membranes with porous electrodes. Solid State Ionics, 74 (1994) 75-84. 107. B.N. Nair, Internal Report, Lab. of Inorganic Materials, Faculty of Chem. Technology, University of Twente, Enschede, The Netherlands, 1994. 108. A.J. Burggraaf, Key Points in understanding and development of ceramic membranes" in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd., Worcester, MA 01609, USA. pp. 1-17.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
433
109. T.L. Ward, Assessment of inorganic membrane technology for petrochemical applications, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. 110. Y. Teraoka, T. Nobunaga and N. Yamazoe, Effect of cation substitute on the oxygen semipermeability of perovskite type oxides. Chem. Lett., (1988) 503-506. 111. Y. Shen, M. Liu, D. Taylor, S. Balagopal and A. Joshi, Mixed ionic electronic conductors based on Bi-Y-O-Ag metal-ceramic system, in: T.A. Ramanarayan et al. (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conductors. Proceedings Vol. 94.12, The Electrochemical Society Inc., pp. 574-583. 112. (a) Chu-sheng Chen, Fine grained zirconia-metal dual phase composites, oxygen permeation and electrical properties. Ph.D Thesis, University of Twente, Enschede, The Netherlands, 1994. (b) C.S. Chen, H. Kruidhof, H.J. Bouwmeester, H. Verweij and A.J. Burggraaf, Oxygen permeation through erbia stabilized bismuth oxide noble metal composites. Solid State Ionics, submitted. 113. S. Yan, H. Maeda, K. Kusakabe and S. Morooka, Thin palladium membrane formed in support pores by metal-organic chemical vapor deposition and application to hydrogen separation. L&E.C. Res., 33 (1994) 616-622. 114. A. Sengupta and K.K. Sirkar, Analysis and design of membrane permeators for gas separation", in: R.D. Noble and S.A. Stern (Eds.), Membrane Separations Technology, Principals and Applications. Elsevier, Amsterdam, 1995, pp. 449-550. 115. J.N. Armor, Challenges in membrane catalysis. Chemtech, 22 (1992) 557-563. 116. H. Nagamoto and H. Inoue, A hydrogen reactor with catalytic membrane. Chem. Eng. Commun., 34 (1985) 315-323. 117. U. Balachandral, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L. Mieville, P.S. Maiya, M.S. Kleefish, S. Pei, T.P. Kobylinski and A.C. Bose, Methane to syngas via ceramic membranes. Am. Ceram. Soc. Bull., 74 (1995) 71-75. 118. H.W. Brinkman, H. Kruidhof and A.J. Burggraaf, Mixed conducting yttrium-bariumcobaltoxide for high oxygen permeation. Solid State Ionics, 68 (1994) 173-176. 119. R. van Doorn, PhD Thesis, University of Twente, Enschede, The Netherlands, 1996.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 10
Dense ceramic membranes for oxygen separation H.J.M. Bouwmeester and A.J. Burggraaf Laboratory for Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
10.1 INTRODUCTION Dense ceramic membranes exhibiting high oxygen ionic and electronic conductivity have become of great interest as a potentially economical, clean and efficient means of producing oxygen by separation from air or other oxygencontaining gas mixtures. In addition to infinite permselectivity, notably high oxygen flux values are measured through selected mixed-conducting oxides with the perovskite structure. These may be in the range exhibited by microporous membranes, albeit that sufficiently high temperatures are required, typically above about 700~ It is generally accepted that, provided they can be developed with sufficient durability and reliability, mixed-conducting oxide membranes have great potential to meet the needs of many segments of the oxygen market. It is further expected that the Oxygen fluxes can be improved by thin film deposition on a porous substrate preferably of the same material to avoid compatibility problems. The applications envisioned range from small-scale oxygen pumps for medical applications to large-scale usage in combustion processes, e.g. coal gasification [1-4]. As oxygen, but also nitrogen, ranks among the top five in the production of commodity chemicals in the United States [5] successful development of the mixed-conducting oxide membranes could thus have clear economic benefits, at the expense of market share from more traditional supply options. Whilst the targeted membranes will be most competitive at small and intermediate scale level in which flexibility of operation is desired, they may
436
10 n DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
eventually challenge the present commercial status of cryogenics, pressureswing adsorption (PSA) and polymeric membranes [1-4]. Another application of mixed-conducting oxide membranes is to be found in the field of chemical processing, including the partial oxidation of light hydrocarbons, e.g. natural gas to value-added products such as ethane/ethene [6--11] and syngas [12-14], waste reduction and recovery [15]. The catalyst may be either the membrane surface itself or another material deposited in particulate form on top of the membrane. Besides the controlled supply or removal of oxygen to or from the side where the catalyst and the reactants are located, a promising feature is that the oxygen flux may alter the relative presence of different oxygen species (02,0-) on the catalyst surface, thereby providing species that may be more selective for partial oxidation reactions. This review addresses recent developments in the area of mixed ionic-electronic conducting membranes for oxygen separation, in which the membrane material is made dense, i.e. free of cracks and connected-through porosity, being susceptible only for oxygen ionic and electronic transport. Current work on different mixedconducting oxides is reviewed using concepts from electrochemistry and solidstate chemistry. Emphasis is on the defect chemistry, mass transport and the associated surface exchange kinetics, providing some basic background knowledge which aids further development of these materials into membranes for the aforementioned applications. There is no attempt to discuss inroads against competing technologies, or to speculate on new opportunities that may result from successful development. New developments in dense ceramic membrane research could offer very economical ways of separating hydrogen such as the proton-conducting ceramics or thin Pd-foils. These are not considered in this review. For a general discussion on the topical area of membrane technology and its impact in various applications the reader is referred to specific reviews, for example, see Refs. [16-20] and other chapters in this textbook. 10.2 GENERAL SURVEY
In this section, a brief overview is given of major membrane concepts and materials. Besides membranes made from a mixed ionic-electronic conductor (MIEC), other membranes incorporating an oxygen ion conductor are briefly discussed. Data from oxygen permeability measurements on selected membrane materials are presented.
10.2.1 Major Membrane Concepts In this chapter, a membrane is regarded as a barrier between two enclosures which preferentially allows one gas (i.e. oxygen) to permeate owing to the presence of a driving force such as a pressure or electric potential gradient.
10 m DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
437
.__ 02"_>
__ O2-.~ P 02,,
PO 2 '
*e
e
(b)
(a)
P02 "
P02 '
(c)
e
,#
02
(d)
Fig. 10.1. Different membrane concepts incorporating an oxygen ion conductor: (a) mixed conducting oxide, (b) solid electrolyte cell (oxygen pump), and (c) dual-phase membrane. Also shown is the schematics of an asymmetric porous membrane (d), consisting of a support, an intermediate
and a barrier layer havhlg a graded porosity across the membrane. The separation of oxygen using an MIEC membrane is schematically shown in Fig. 10.1a. The driving force for overall oxygen transport is the differential oxygen partial pressure applied across the membrane. As the MIEC membrane is dense and gas-tight, the direct passage of oxygen moleculesis blocked, yet oxygen ions migrate selectively through the membrane. Dissociation and ionization of oxygen occurs at the oxide surface at the high pressure side (feed side), where electrons are picked up from accessible (near-) surface electronic states. The flux of oxygen ions is charge compensated by a simultaneous flux of electronic charge carriers. Upon arrival at the low pressure side (permeate side), the individual oxygen ions part with their electrons and recombine again to form oxygen molecules, which are released in the permeate stream.
438
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
Mixed conduction also plays an important role in many other processes, e.g. in improving electrode kinetics and catalytic behaviour [21]. In fact, all oxides exhibit to some degree mixed ionic and electronic conduction, and selective oxygen permeation has been reported even for dense sintered alumina above 1500~ [22,23]. Although it is common to speak of mixed conduction when the total conductivity is provided by near equal fractions (transference numbers) of the partial ionic and electronic conductivity, respectively [24], from the point of view of oxygen permeation it is more useful to relate mixed conduction to their absolute values. Volume diffusion theories treating ambipolar transport in oxides clearly indicate that higher currents (fluxes) are obtained when either the electronic or the ionic conductivity increases, or both increase simultaneously. The flux at a given total conductivity is maximum when the ionic and electronic transference numbers are equal, i.e. 0.5. In this view, alumina is not a good mixed conductor. Materials showing predominant electronic conduction may thus prove to be excellent mixed conductors when their ionic conductivity is also substantial. The general objective for optimum membrane performance therefore is to maximize the product of mobility and concentration of both ionic and electronic charge carriers in appropriate ranges of temperature and oxygen partial pressure. Owing to the ability to conduct both oxygen ions and electrons, the MIEC membrane can operate without the need of attachment of electrodes to the oxide surface and external circuitry. The latter represents an inherent advantage over traditional oxygen pumps in which a solid oxide electrolyte is sandwiched between two gas-permeable electrically conductive electrodes (Fig. 10.1b). An advantage of electrically-driven oxygen separation may be its ability to deliver oxygen at elevated pressures, eliminating the need for compressors [25]. Figure 10.1c shows a dual-phase membrane, which can be visualized as being a dispersion of a metallic phase into an oxygen ion conducting host or matrix, e.g, Pd metal into stabilized zirconia. This challenging approach was first described by Mazanec et al. [26] and offers an alternative use of oxide electrolytes in the field of dense ceramic membranes. Industrially important solid oxide electrolytes to date are mainly based on oxygen-deficient fluorite-related structures such a s Z r O 2 and C e O 2 doped with CaO o r Y 2 0 3 . Unless operated with an internal or external circuitry, the oxygen flux through these materials in usual ranges of temperature and oxygen pressure is negligibly low, preventing their practical use as oxygen separation membrane. The existence of a non-vanishing electronic conduction in the ionic domain, and concomitant oxygen semipermeability, however, can be detrimental considering their use as solid electrolytes in fuel cells and oxygen sensors [27,28]. While past efforts were focused on expanding the electrolytic domain of oxygen ion conducting fluorite-type ceramics, more recently one has begun to introduce enhanced electronic conduction in fluorite matrices. Extrinsic elec-
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
439
tronic conduction in ionically conducting matrices can be obtained by dissolution of multivalent cations in the fluorite oxide lattice. Notable examples include yttria-stabilized zirconia doped with either titania [29,30] and ceria [31, 32]. Electronic conductivity in these solid solutions is reportedly found to increase with increasing dopant concentration, but may be limited by the solid solubility range of the multivalent oxide. As conduction occurs via a small polaron mechanism (electron hopping) between dopant ions of different valence charge, its magnitude will strongly vary with temperature and oxygen partial pressure. In general, the extent of mixed conductivity that can be induced in fluorite ceramics is limited, which restricts its possible use as ceramic membrane, unless very high temperatures of operation (> 1400~ and stability down to very low values of oxygen partial pressure are required as, e.g., in the production of gaseous fuels CO and H2 by direct thermal splitting of CO2 and H20, respectively, and extraction of the oxygen arising from dissociation [33]. Since the first report on high oxide ion conductivity in some of the rare earth aluminates in the mid sixties [34,35], materials with oxygen-deficient perovskite and perovskite-related structures receive much attention for the development of new solid electrolytes and mixed conductors for numerous applications [36]. Currently, extensive research is conducted on acceptor-doped perovskite oxides with the generic formula Lal_xAxCOl_yByO3_~(A = Sr, Ba, Ca and B = Fe, Cu, Ni). Teraoka et al. [37-39] were the first to report very high oxygen fluxes through the cobalt-rich compositions, which perovskites are known to become highly oxygen anion defective at elevated temperatures and reduced oxygen partial pressure. The oxygen-ion conductivity in the given series can be 1-2 orders of magnitude higher than those of the stabilizedzirconias, though in usual ranges of temperature and oxygen partial pressure electronic conduction in the perovskite remains predominant [39,40]. Besides potential use of these perovskite compositions as catalytically active electrodes in, e.g. fuel cells, oxygen pumps and sensors, the compounds have a bright future for use as oxygen separation membrane. The precise composition may be tailored for a specific application, but this has not yet been fully developed. Structural and chemical integrity of the cobaltites, however, is a serious problem and needs to be addressed before commercial exploitation becomes feasible. For the sake of completeness, a schematic representation of a porous ceramic membrane is given in Fig. 10.1d. The majority of porous ceramic membranes are composite or asymmetric in structure. They include materials like 0~-A1203, ~-A1203, TiO2 and SiO2, and generally consist of a thin layer of either a mesoporOUS (2
10 -- DENSECERAMICMEMBRANESFOR OXYGENSEPARATION
440
10.2.2 Data: Oxygen Permeability of Solid Oxide Membranes Table 10.1 lists data of steady-state oxygen permeability measurements on various solid electrolytes, mixed conducting oxides and dual phase membranes taken from various literature reports. Measurements most commonly are performed by imposing a gradient in oxygen partial pressure across the membrane, usually by passing an oxygen rich and lean gas, e.g. air and inert gas, respectively, along opposite sides of a sealed ceramic disk or tube wall, without the use of external circuitry such as electrodes and power supplies. The number of moles, volume or mass of oxygen passing per unit time through a unit of membrane surface area is measured down stream using, e.g., on-line gas chromatography or an oxygen sensor, from which data the oxygen flux is calculated. Table 10.1 also includes data for solid electrolyte cells used in the oxygen pump mode. A graphical presentation of selected data is given in Fig. 10.2.
-5.00 -6.00
'm
(1)
' + ~
-7.00
o,I
E o o E
"~
-8.O0
(9)
~
9
N
0
....~
o'~
-9.00
o
-10.00 -11.00
r
0.65
.
.
.
.
.
.
i
.
.
1.00
.
.
.
.
1.35
1000/T [l/K] Fig. 10.2. Arrhenius plots of oxygen permeation for: (1) La0.3Sr0.K~oO3-8[7]; (2) Ba0.9Y0.1CoO3-6[19]; (3) YSZ-Pd (40 vol%), continuous Pd-phase [23,24]; (4) YSZ-Pd (30 vol%), discontinuous Pd-phase [23,24]; (5) La0.sSr0.sCoO34 [7]; (6) (ZrO2)0.7-(Tb203.5)0.228-(Y203)0.072[4,5]; (7) SrCo0.sFe0.203-~ [9]; (8) BE25-Ag 40 vol% [24,29]; (9) Ba0.66Y0.33CoO34 [19]; (10) (Bi203)0.75-(Er203)0.25 (BE25) [22]; (11) BY25-Ag (35 vol%), thickness 90 wn [27]; (12) BY25-Ag (35 vol%), thickness 1.5 mm [27]. Air and inert gas are passed along opposite sides of the membrane. Unless specified otherwise, the membrane thickness varies between I and 2 mm. For references, see footnote to Table 10.1.
~~o~
~ ~ ' ~ ~, ~.~" ~. ~
o
~cl:l -~'~
E[
~: ~.~o
c',l
ra~
~
~.~'~ ~
,
Ref.
0
0
8
~0
Range of measurements t~O
0
~
Temp. Thickness Oxygen fluxa Experimental conditions T L J 0, "C mm kmol cm-' s-'
~.~ ~,~
hi:)
u
o
~.~
Membrane material
~
~o
~
~
v~
~
..
~
.~.~
mixed conductors: (tion<< 1) on fluorite basis (see Section 10.5) m
O~
o
o
m
T=827-1523"C;x = 0.09-0.36 '
~d
ceramic tube (013 x 5 mm and 013 x 15 mrn); He
cq
X
1
~
%
1 2
~
0
d
o b
II
0o"7
"~1" ~
0
~0
0 ~
ceramic tube (08 x 50 mm); Po,' T=13O5-148l0C;x = 0.075, 0.1; = 0.122 atm; sweep gas HJCO, = -lo-" atm mixed gas (Po," = lo-'' atm)
x~
d
.-
28 x 10-3 0.20 o
650
~
o
900 900 900
oo
d
0
~
d d
~o
~
o i
c~
2.0 1.5
1523 1481
v~ X
~
(q ~
13 x
y=0.36
~
X
(q.-
2.0
S
--
o
870
o
c~ o I"
~
~
o~
i
(Z~2),.,,-(Ce0,),-(CaO~y x=0.09;
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
TABLE 10.1
Oxygen fluxes through ceramic membranes, for a given temperature and membrane thickness, together with the experimental conditions during measurements. When not specified otherwise, the Pe-gradient corresponds to Po; = 0.21 atm (air) versus inert gas (sweep method). Indicated are the sweep rate of the inert gas (in sccm) and disc diameter 0 (in mm), which parameters in conjuction with the oxygen flux determine the oxygen partial pressure Po/ in the permeate stream. In a number of cases the value of Po/ is specified. The full range of measurements covered by the experiments is also given. For references see end of table
S
~0"
"7~
~
"7,
0
T=700-11Oo0C;Po>,= 0.07-1 atm
o
0=12 mm; He (12.5 sccm)
0
o
T=65O-81O0C;Po,' = 0.01-1 atm
b
X
d
',0 (.q
II
~'~ 0
d
i
O"
o
0
0
103
o o
X
o
o
oo
0.18
o o00
o
0=12 mm; He (20 sccm)
b
Po,' = 0.21-1 atm T=900-11OO0C; Po,, = 0.01-1 atm
~,
% . -X
@=I2mm; Ar 0=12 mm;He (20 sccm)
(t"~
~.
~1
o
(q
~
o
~1 ~
o
o~
~
o
i
c~
1.7
5.2 x 0.037 x 10.' 0.026 x X
oo
o~
c~
2.0 2.0 2.0
3 43 4.5
6
.~
~9~
on perovkite basis (see Section 10.6) 7
~
(continued)
441
!
II
c~
0
~5
0.022
"7,
2.0
o
900
e.q
o
x=0.2
=
c~
Lal.,SrxCo03.,
442
TABLE 10.1 (continuntion)
b P
f3
Range of measurements ffl
0
r~
~0 o,..,~
r0j
q~
~
~
Temp. Thickness Oxygen flux" Experimental conditions T L 10, "C mm pmol cmw2s-l cL
,-O
Membrane material
Ref.
ou
t'--
9
E
o
t~ r~
9
OO
txl
~
0
9 o
9
O0 O0 (:~ O~
9
O~
~o
0o
T=3W-865"C t~
o
0 er
9
(:~
0o
0=20 mm; He (30 sccm) II
0o
9
0~,
oo oo
9
0~,
oo
II
9
~
oo
II
9
O~
II
0.70 1.oo 1.32
9
(:~
oo
II
0.42
9
0~,
0o
0.80 1.07 0.33 0.39
9 9 9
(:~
o
0o
T=75O-82O0C t~
oo
0 er
II
0=20 mm; He (30 sccm)
1
DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
O0
0.256
10-
,,o
.. o
~
1.5 1.5 1.5 1.5 1.5 1.5
X
865 865 865 865 865 865
1.5
II
~ , ~
0.61 0.69
It
0.47
1.5 1.5 1.5
o ~
.
1.5
oo
B=Mn B=Cr B=Fe B=Co B=Ni B=Cu
820 820 820 820 820
II
o
gB0 2O3-6
0.188
II
o~ e~
o~
o
Lrb6sr04c00
1 .o 1 .o
+
0
8 8 8
~o0~
o
u
o
Ln=La Ln=Pr Ln=Nd Ln=Sm Ln=Gd
0.104
oo
x=0.4
1 .o
oo
x=0.3
1000 1000 1000
8
Po,' = 0.01-1 atm; Pol" = (0.440) x 10.' atm (He: 5-70 sccm)
0.004 atm) x=0.2
cSo,-
ii
II
~P~ oo
0
c~
9
0 = 12 mm; He (efluent Pol"= T= 850-1050°C; L = 0.5-2.0 mm; ,~,
0.019
7 7 7
,~,
1 .o
.....He (10 sccm) !
1000
x=o. 1
o~
0
(:~
2.0
x=0.6 x=O.8
0
c~.
~
900
0.20 0.49
c~
2.0 2.0
0.074
9
900 900
x=0.4
O ~ OO
o
~ ~
O ~ OO
1.o
O ~"~ OO
I .o
O ~'~ OO
0.55
d
O
d
~
~
~n
~ ~ OO
II
~
9,lO 9,lO O
~
o o0
~)
~. o
850 850 850 850
~, 9
~.
x=o x=0.4 x=O.8 x= 1
0.15
T= 650-950°C
11
-
II c~
~ o
850
0=25 mm;Pol' = 1 atm. N,: (efluent Po," = 0.latm) ~
~.
x=0.6
O
~n
c~
O
850
O
~--
~
9.10 9,lO 9.10
~
II
o
m
0=lO mm; He (30 sccm)
~"
T= rt.-875"C
~ d d
12 13 14
v
~
O
A
9
i
8
v)
m
1 51
B 0
~'~
~:)
15
~'~ ~ ~
~
-
9. 0 Pod |
II
OO O~
o~ 8
T= 900-1 100°C
16,17 18
P
T=700-95OoC
o ~
~;
o
~
O
O~
19 20 O
" o !
T=700-1oOo0C; Po2" = 0.01-0.13 atm iI
~c~
x
O
O
OO
0.39 7.8 x 10-3
~
~
~
0.56 0.17
15 12 16.17
T = rt.-875"C T = 620-920°C; k1.0-5.5 Pol' = 0.04-0.90 atm II
t'~
o o
m
~
~
d
,...,
o
d
~
O
OO
2.0 1.o
~ d! . ~ P
L
~
d
,..4
@=I2 mm; He (10 sccm) ceramic tube; evacuated to Po," = 0.079 atm
~
o o
OO
o~
900 800
2.0 1.o
~ 00 8
oo
~
~ .
N d
O
0=12 mm; He (10 sccm)
820 850
12 12 12 12
T = rt.-875"C T = 730-1ooO"C T=800-1ooO0C; L = 0.5-2.0 Po,' = 0.01-0.21 atm; Po2"= (350) x 10.' atm (He: 5-70 sccm). T=600-90O0C II
~
o U ~
t'~
~
d
o
d
~
.....He (40 sccm)
,-4
OO
@=lomm; He (30 sccm) 0=13 mm; He (20 sccm)
O
"~-O
OO
~ ~ . O -- d d
c~
OO
0.33 1.3 240 pm 0.85 1.o 2.11 0.43 I .O
r
o o o o
OO
,-.,
~
II
o
o.
"~"
~
II
0=13 mm; He
1.06
0.026 0.054
o
o
v
< 0.01 0.12 0.40 1.49
@=lomm; He (30 sccm) @=n.s.; He (30 sccm) 0=15 mm; He (effluentPo," = 0.017 atm)
o
r
x= 1
11
O
II
850 850 850 820
o
II
oO
x=0.4
850 850 850
o.
II
"~"
o~
O~
x=0.2
o
II
o~
O
La,.,Sr,Coo4Feod4.6
0.33
c. 0
10 - - D E N S E CERAMIC M E M B R A N E S FOR O X Y G E N SEPARATION
O ~ OO
1.o
O
O ~ OO
1.o
o
O ~ OO
1 .O
u
O ~ OO
1.o
m
O ~ OO
~0
2.5
o
II
~"
2.5
0.01 0.16 0.42 1.26 I .43
m
Vh ~ OO
1.5
o
II
~
O
~ ~O OO
1.5
865 865 865 865 865
0=20 m m ;He (30 sccm)
d
~ ~O OO
II
~
~.
~O OO
II
I .5 I .5 I .5
A=Co A=Na A=Sr A=Ca A=Ba II
23'6.
0
o
O
6A04C008Fe,l
t
443
(continued)
E
TABLE 10.1 (continuation)
Temp. Thickness Oxygen fluxa Experimental conditions
Membrane material
T
"C
L mm
Range of measurements
Ref.
10,
pmol m-*s-'
oxide electrolytes (t,,, = I ) (see Section 10.4.2) (ZlO,),
9 2 - ( y 2 0 , ) 0 08
(YSZ)
(Bi203),,7j-(Er203)o 2j W 2 5 )
800
1.0
1U8 -
average value .from compilation by [I1
1200 650
1.0 0.7
1 0 . ~- 10-3 0.76 x l o 3
810
0.7
6.7 x
21 @=I2nun; Po>'= 1 atm; He (40 T = 650-810°C; Pol, = 0.01-1 atm; 22 sccm) Po," = 1.1 x 1 0 . ~atm 22
21
I
internally short-circuited: dual-phase composites (see Section 10.4.3.1) YSZ-Pd
metal phase fr.
,
YSZ-In, pPro 0,5.6
30 vol%
1100
2.0
0.26 x l o 3
40 vol%
1100
2.0
76 x
1100
0.5
0.25
1100
0.5
0.87
50 vol%
1100
0.8
1.44
50 vol%
1100 1100
0.8 0.25
1.56 4.17
volume fraction below percolationT=1050-1150"C;Polr = 0.01-1 rhreshold; @=I2 mm; He atm; Pol'' = (0.6-0.18) x l o 3 atm (effluenrPol'' = 0.17 x 10.' atm) (He: < 20 sccm). ......volume fraction above T=900-1100"C; L = 0.5-2.0 mm; percolafion threshold He Pol' = 0.01-1 atm; Pol'' = (4.3-14) (effluent P," = 0.14 atm) x 10'' atm (He: < 20 sccm). .....He (effluent Pol" = 12 x l o 3 " atm) .....sweep gas He/CO/CO, T=900-1100"C (effluent Pol" = 0.26 x atm) 0=25 mm; sweep gas 90% H, in Ar (68 sccm) .....0=31 mm
.....0=n.s.
23.24
23,24
24 24 25,26 25,26 25.26
8
10
3
0
--
0=25 mm; N, (120 sccm)
,m
d
II
27
t'--
~
,--,
<
200
;>
O o
10 um
o
$
3-7
o
0o-o" c5 d
O O
100 nm
T=5OO-75O0C;1-50 ~ 0 1 %Ag; L = 0.09-5.0 IIUII
0
d
O O kO
4 1.5
0
1.0
s
6
750
vh
1.6
II
750
:
o c'q
90 pm 1.0
o
750 720
o ~r~
o p.
40 vol%
d
BE25-Ag
o
750
,m
35 vol%
25,26
"
c',,I
x
~
~.-,
;>
('-,I
o~
"
c~h c,,I
24.29
~C~
0=15 mm; He (efluent Po," = atm) 16 x .____He (efluent Pol" = 2.1 x atm)
27 28
T=600-72OoC;0-40 vol% Ag; L = 0.25-2.0;UTIII Po>'= 0.01-1 atm; PO2"= (0.6 - 30) x 10.' atm (He: 5-70 sccm). T=60O-85O0C ,, ~
II -"
~r~
00
O
0
0.085
-~.
o~
II -~
cxl
c5 o
"~
@=15 mm; He (effluent Po," = 6.3 x 10.' atm)
t--, o o
k~D
o
O ~
o p.
0.96 0.12
:
0.016
ox
o "" X
kO
O
0
40 vol%
BE25-Au
24,29
"N
~!
N
externally short-circuited: oxygen pump (see Section 10.4.3.2) ~
er~
b--
O oh
kid
30 31
co-pressed Au-grids electrodes sputtered Au electrodes
~o
+
0.78 0.5-0.7
e~
0
600 630
Bi0571Pb042801 285 (+0.187 mol% ZrO,) BE25
DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
0.31
BY25-Ag
~c~
1.0
1100
c~I
.....0=n.s.
o'~
5.39
o
0.3
O p.
O o
YSZ-Ino,,Pr,,o,iZro,,iO, i.6 50 vol%
r.~
porous membranes m
35-200
o
~
SiO, film (d,= 0.5 nm) on layered y - a A1,0, support ~ O
400
o
Meso-porous (d, = 10 nm) (calculated value)b
activated diffusion; sel. ct = 2-4 ( 0 2 - N j ; abs. press. direrence I atm theor. expected value based upon Knudsen diffusion
b~
O
445
II
day-' = 386 mA cm-, oo
II
c-4
o
II
c~
o
o
o
(D
.,..q
--d
.o
..~ o o
o
>.<
(a) Values may be converted to other dimensions using equalities 1 pmol c m 2 s-' = 1.47 sccm cm-, = 21.1 m3 (b) Additional parameters used in calculation are the porosity E = 0.5 and the tortuosity z = 2.
32
446
,.:.;
~
REFERENCES USED IN TABLE 10.1: v
Mc,i
--
1. Y. Nigara, J. Mizusaki and M. Ishigame, Solid State lonics, 79 (1995) 208-211. 2. H. Arashi, H. Naito and M. Nakata, Solid State Ionics, 76 (1995) 315-319. 3. H. Iwahara, T. Esaka and K. Takeda, Mixed conduction and oxygen permeation in sintered oxides of a system Zr02-Tb~07,in: S. Somiya, N. Yamamoto, H. Yanegida (Eds.),Advatices it1 Ceramics, Vol. 24A. The American Ceramic Society, Inc., Westenrille, OH, 1988, pp. 907-916. 4. G.Z. Cao, X.Q. Liu, H.W. Brinkman, K.J. De Vries and A.J. Burggraaf, Mixed conduction and oxygen permeation of Zr02-Tb203.5Y2O3 solid solutions, in: S.P.S. Badwal et al. (Eds.), Science and Technology of Zirconia V. Technomic Publications, Lancaster, 1993, pp. 576-583. 5. G.Z. Cao, I. Appl. Electrochem., 24 (1994) 1222-1224. 6. H. Kruidhof and H.J.M. Bouwmeester, unpublished results. 7. R.H.E. Van Doorn, H. Kruidhof, H.J.M. Bouwmeester and A.J. Burggraaf, Oxygen permeabilty of strontium-doped LaCoO3.6 perovskites, in: G.-A. Nazri, J.-M. Taracson and M.S. Schreiber (Eds.), Materinls Resenrcli Society Syntpsitini Proceedings, v d 369, Solid State Iotiics IV. Materials Research Society, Pittsburgh, PA, 1995, pp. 377-382. 8. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, Solid State Ionics, 81 (1995) 97-110. 9. Y. Teraoka, T. Nobunaga and N. Yamazoe, Cliem. Lett. (1988) 503-506. 10. y. Teraoka, T. Nobunaga, K. Okamoto, N. Miura and N. Yamazoe, Solid State Ionics, 48 (1991) 207-212. 11. W.J. Weber, J.W. Stevenson, T.R. Armstrong and L.R. Pederson, Processing and electrochemical properties of mixed conducting Lal-xAxCo1-yFey03-s (A = Sr, Ca), in: G.-A. Nazri, J.-M. Taracson and M.S. Schreiber (Eds.), Materials Resenrcli Society Synipiisiirtn Proceedings, Vol. 369, Solid State Ionics IV. Materials Research Society, Pittsburgh, PA, 1995, pp. 395-400. 12. Y. Teraoka, H.M. Zhang, S. Furukawa and N. Yamazoe, Clwm. Lett. (1985) 1743-1746. 13. C.-Y. Tsai, Y.H. Ma, W.R. Moser and A.G. Dixon, Simulation of nonisothennal catalytic membrane reactor for methane partial oxidation to syngas, in: Y.H. Ma (Ed.), Proceedings of the 3rd Internutionul Conference on Inorganic Membranes, Worcester, 1994, PP. 271-280. 14. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, Appl. Catal. A: General, 130 (1995) 195-212 15. N. Miura, Y. Okamoto, J. Tamaki, K. Morinag and N. Yamazoe, Solid State Ionics, 79 (1995) 195-200. 16. L. Qiu, T.H. Lee, L.-M. Lie, Y.L. Yang and A.J. Jacobson, Solid State lonics, 76 (1995) 321-329. 17. Y.L. Yang, T.H. Lee, L. Qiu, L. Liu and A.J. Jacobson, Oxygenpermeationstudiesof S~oo.sFeo.zOss, in: G.-A. Nazri, J.-M. Taracson and M.S. Schreiber (Eds.),Materials Research Society Symposium Proceedings, V01369, Solid State loiiics IV. Materials Research Society: Pittsburgh, PA, 1995, pp. 383-388. 18. H. Kruidhof, H.J.M. Bouwmeester, R.H.E. van Doom and A.J. Burggraaf, Solid State Ionics, 63/65 (1993) 816-22.
.~ o
p.~'~
~
~
,~
.~
zz
10 -- DENSE
o ,~
o =
e2
~
=~
~.,
o-~
c~
, . -
~-:
~
o
~=
~
x ~
o ~~
"-
~
.~
~ ~_-~
o ~
L~ ~
'
~
~9 ~
.__~
o
~o
._
I
L4 Z
x ~
= 9~
n
,a~
E ~
a,,m 9
~
~ ~
v)
~~ . ~
~
~-=~
$l
r162
FOR OXYGEN
9
t
MEMBRANES
~~.
= ~,
~- ~
~ ~ ~~
..~ ~ ~
~
0
2
88
,~
~
SEPARATION
!
""~
~
o~
~
"~
~~
"~' ~~~'~>Z" ~x >~~~ ~~~ ~ ~
~
=o
o--~s
9
,
P
3
CERAMIC
z
o
~
,~~--~' ~ ~=~.
~ -:~:~:~ ~ ~
~
~~"
~
~~ >~.~ ~ . - ~ g~~
~~,~~ ~
~~zo~ >~ b xO.~ ~,
+
0
F
~9~
~
~'~
9
~
~
~
~
,~ ~
T
~
~ ~
!
6-~
m
"Q "~ t',.
!
.
"~ ~
.
.
I
.
~0
~-~ ~ -
72~ z ~
~
u.iu~ B-~O:~ 0~~
.-,,~,m SN
~
"~
~
"i
u~ ~
28. 29. 30. 31. 32.
_:,_:, ~ : ':-:u:.~ ~ ~~0 ~~ : u5 ~'~
24. 25. 26. 27.
H.W. Brinkman, H. Kruidhof and A.J. Burggraaf, Solid State lonics, 68 (1994) 173-176. H. Iwahara, T. Esaka and T. Mangahara, I. Appl. Electrochem., 18 (1988) 173-177. J. Fouletier, P. Fabry and M. Kleitz, I. Electrochem. Soc., 123(2) (1976) 204-213. H.J.M. Bouwmeester, H. Kruidhof, A.J. Burggraaf and P.J. Gellings, Solid State lonics, 53/56 (1992) 460-68. C.S. Chen, B.A. Boukamp, H.J.M. Bouwmeester, G.Z. Cao, H. Kruidhof, A.J.A. Winnubst and A.J. Burggraaf, Solid State loizics, 76 (1995) 23-28. C.S. Chen, PhD Thesis, University of Twente, The Netherlands, 1994. T.J. Mazanec, T.L. Cable and J.G. Frye, Solid State lonics, 53/56 (1992) 111-118. T.J. Mazanec, and J.G. Frye Jnr., Eirr. Patent Awl. 0399 833 A1 (1990). Y.S. Shen, M. Liu, D. Taylor, S. Bolagopal, A. Joshi and K. Krist, Mixed ionic-electronic conductors based on Bi-Y-0-Ag metal-ceramic system, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings ofthe 2nd International Symposillm uti Ionic nizd Mixed C o n d i ~ t i q Ceramics, Vol. 94-12. The Electrochemical Society, Pennington, NJ, 1994, pp. 574-595. J.E. Ten Elshof, D.N.Q. Nguyen, H.J.M. Bouwmeester and H. Verweij, Solid State lonics, submitted. C.S. Chen, H.J.M. Bouwmeester, H. Venveij and A.J. Burggraaf, Solid State lonics, submitted. M. Dumelie, G. Nowogrocki and J.C. Boivin, J.C. Solid State lonics, 28/30 (1988) 524-528. I.C. Vinke, K. Seshan, B.A. Boukamp, K.J. d e Vries and A.J. Burggraaf, Solid State lonics, 34 (1989) 235-242. R.S.A. De Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Microporous Materials, 4 (1995) 169-186.
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
3
19. 20. 21. 22. 23.
0
I
rn 0
5m
! n
B
E
s<
@#
F8 z
447
448
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN S E P A R A T I O N
High oxygen fluxes are found through selected perovskite-structured ceramics. Table 10.1 shows for several perovskite systems the trend in permeation flux as a function of the type and concentration of applied dopants. In the range 800-900~ the highest flux was measured by Teraoka et al. [37] for SrCo0.sFe0.2034 but, as for a number of other compositions, different values have been reported by other groups. Such conflicting results reflect the experimental difficulties in measuring oxygen permeation of sealed ceramic discs at high temperatures, but may also be due to factors that influence the effective P%-gradient across the membrane, sample preparation, etc. This is further discussed in Section 10.6.7.2. For the sake of comparison, Table 10.1 contains limited data for the oxygen flux through micro- and mesoporous membranes. As noted before, the lastmentioned category of membranes falls outside the general scope of this chapter. It is seen that the oxygen fluxes observed through membranes formed from the mixed-conducting perovskite-type oxides, such as La>xSrxCo1_yFeyO34, approach those exhibited by the porous membranes. It should be noted, however, that these types of membrane have different requirements. The high temperature needed for operation using membranes based on oxygen ion conductors may be restrictive in certain applications, but beneficial to others, e.g. coal gasification and partial oxidation of light paraffins [25].
10.2.3 Factors Controlling Oxygen Permeation The rate at which oxygen permeates through a non-porous ceramic membrane is essentially controlled by two factors, the rate of solid state diffusion within the membrane and that of interfacial oxygen exchange on either side of the membrane. The oxygen flux can be increased by reducing the thickness of the membrane, until its thickness becomes less than a characteristic value, Lc, at which point the flux of oxygen is under conditions of mixed control of the surface exchange kinetics and bulk diffusion [41]. Below Lc, the oxygen flux can only marginally be improved by making the membrane thinner. For predominant electronic conductors like, for example, the perovskites Lal_xSrxCOl_yFeyO3~, Lc is determined by the ratio of the oxygen self diffusivity and surface exchange coefficient. Both parameters can be measured simultaneously by using 180-160 isotopic exchange techniques. Calculations show that Lc may vary from the ~tm-range to the cm-range, depending on material and environmental parameters. Modelling studies, however, show that significant increase in the rate of interfacial oxygen transfer and, hence, in the oxygen flux can be achieved by deposition of a porous MIEC layer on top of the (thin) non-porous membrane [42--44]. Since a number of simplifying assumptions is made, such as neglect of changes in material parameters with variation in the chemical potential of oxygen, the models developed are valid only in the limit of small Po2-gradients across the MIEC membrane. For a more rigorous approach,
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
449
referring to actual operating conditions of oxygen separation membranes, much more work is needed to arrive at a better understanding of the transport processes under oxygen potential gradients. In particular, our present understanding of the factors that govern the surface exchange kinetics is rather poor. Effects related to microstructure, including grain boundary diffusion and (local) order-disorder phenomena, may also influence overall oxygen transport. Besides the processing into defect-free thin films and associated problems of compatibility between deposited membrane layer and the porous substrate material, chemical stability at high temperatures, effects induced by the presence of an oxygen potential gradient like segregation of impurities to the surface and to grain boundaries, kinetic demixing and kinetic decomposition could affect membrane performance or limit operational life. In many cases, these difficulties remain to be overcome before commercial exploitation becomes viable. All these factors are important and govern the selection of materials. In the following sections, the emphasis is on the basic elements of mixed ionic and electronic transport through dense ceramic membranes. Due to size considerations, we shall mainly focus this chapter to mixed-conducting acceptordoped perovskite and perovskite-related oxides. Other membrane concepts are also discussed, but only briefly. The examples chosen illustrate the fundamental factors determining the oxygen fluxes through dense ceramic membranes, which is the primary aim of this chapter. 10.3 F U N D A M E N T A L S
10.3.1 Bulk Transport The basic assumption of the theory presented in this section is that the lattice diffusion of oxygen or the transport of electronic charge carriers through the bulk oxide determines the rate of overall oxygen permeation. Moreover, oxygen is transported selectively through the membrane in the form of oxygen ions, rather than molecules, under the driving force of a gradient in oxygen chemical potential. The flux of oxygen ions is charge compensated by a simultaneous flux of electrons or electron holes, which is enabled without the use of external circuitry. We only briefly review the fundamentals of solid state diffusion through mixed conducting oxides and the reader is referred to Refs. [45-47] for a more complete discussion.
10.3.1.1 WagnerEquation Considered here is the case where the interaction of gaseous oxygen with the oxide lattice can be represented by a chemical reaction of the form 1 1 The notation adopted for defectsis from Kr6gerand Vink [48].
450
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
1 -~ 0 2 4- W6 4-
2 e'= O~)
(10.1)
a s s u m i n g that oxygen vacancies are the mobile ionic defects. These m a y be obtained, e.g., by d o p i n g of the oxide lattice with aliovalent cations. The intrinsic ionization across the b a n d g a p can be expressed by nil = e' + h.
(10.2)
The single particle flux of charge carriers, with neglect of cross terms b e t w e e n fluxes, is given by (3"k
jk =
z~k F2
Vnk
(10.3)
w h e r e Z k is the charge n u m b e r and C~kthe conductivity of charge carrier k, F the F a r a d a y constant and Vrlk the gradient of the electrochemical potential. The latter comprises a gradient in chemical potential V~tk a n d a gradient in electrical potential VO, for each individual charge carder k given by VT~k = V ~ k + Zk F
V~)
(10.4)
The charge carrier diffusing m o r e rapidly causes a gradient in the electrical potential V~, in which the transport of carriers with opposite charge is accelerated. At steady state, no charge accumulation occurs. The fluxes of ionic a n d electronic defects are therefore related to each other by the charge balance 2 j v 6 = je" - jh.
(10.5)
Equation 10.5 can be used together with Eqs. (10.3) a n d (10.4) to eliminate the electrostatic potential gradient. The flux of oxygen vacancies is then obtained in terms of the chemical potential gradients only. If it is further a s s u m e d that internal defect chemical reactions are locally not disturbed by the transport of matter, the chemical potential gradients of individual charge species can be converted into the virtual chemical potential of gaseous oxygen, ~to2. The following differential relations hold at equilibrium 2
2
It is tacitly assumed here that the chemical potential of lattice oxygen ~to~)is constant. The present formulation of the defect equilibrium for the formation and annihilation of oxygen vacancies and electrons by the reaction of the solid with environmental oxygen, however, is written in terms of the 'virtual' chemical potentials of the constituent structure elements. In so doing, one does not properly take into account the so-called site-exclusion effect, because the chemical potential of the oxygen vacancy 1/6 and that of lattice oxygen O~) cannot be defined independently from one another. In the present context, it suffices to say that the derived equations are in agreement with those obtained from a more rigorous thermodynamic treatment based upon the 'true' chemical potential for the building unit vacancy, i.e. (V6-O~). For further reading concerning the definition of chemical potentials, the reader may consult Refs. [45] and [46].
10 u DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION 1
451
-~ V~to2 + V~tv6 + 2V~te, = 0
(10.6)
V~I,e, 4- V~h.---- 0
(10.7)
where ~tv6 denotes the chemical potential of the oxygen vacancy, ~te' and ~h. denoting the chemical potential of electrons and electron holes, respectively. The flux of oxygen through the membrane can be derived by combining Eqs. (10.3)-(10.7), using the relationship Jo2 = -1/2 jr6. One finds
J02
=-
1
~ ((~e'4- (~h.)(~V 6
J .........
42F 2 k ((~e' 4- (~h.) 4- O'V6 j
JV~t02
(10.8)
or in a more generalized form 1 (~elt~ion Jo2 = -- 42 F 2 V~to2 Gel 4- (~ion
(10.9)
where (~ion -- (~V 6 and (3"el ---- (~h. + (~e' are the partial ionic and electronic conductivity, respectively. The conductivity term in Eq. (10.9) is equivalent to teltiont~total = tiont~el = telt~ion, where tel and tion a r e the fractions (transference numbers) of the total conductivity (~totalprovided by electronic and ionic defects, respectively. Integration of Eq. (10.9) across the oxide membrane thickness, L, using the relationship V~to2 = ORT In Po2/Ox (x = distance coordinate) and assuming no divergence in the fluxes, yields the Wagner equation in the usual form In P"o2
RT Jo2 = - 2------42 f ~
(~elt~i~ d In Po2 ~
(10.10)
(~el + (~ion
lnP'02
The limits of integration are the oxygen partial pressures maintained at the gas phase boundaries. Equation (10.10) has general validity for mixed conductors. To carry the derivation further, one needs to consider the defect chemistry of a specific material system. When electronic conductivity prevails, Eqs. (10.9) and (10.10) can be recast through the use of the Nernst-Einstein equation in a form that includes the oxygen self-diffusion coefficient Ds, which is accessible from ionic conductivity measurements. This is further exemplified for perovskitetype oxides in Section 10.6.4, assuming a vacancy diffusion mechanism to hold in these materials.
10.3.1.2 Chemical diffusion coefficient The preceding theory was used by Wagner to describe oxide film growth on metals [49,50]. The driving force for diffusion is not a concentration gradient,
452
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
but rather a chemical potential gradient. An important and necessary assumption is that the internal defect reactions are fast enough to attain local chemical equilibrium so that the concentrations of involved ionic and electronic (electrons or holes) charge carriers at any distance coordinate in the oxide are fixed by the local value of the virtual chemical potential, ~to2. The effective transport is still that of neutral oxygen atoms by which the theory fits that of a chemical diffusion process in terms of Fick's first law 3co jo = - D 3x -
(10.11)
where the driving force for diffusion is the gradient in neutral oxygen, 3Co/3X. The coefficient of proportionality, denoted by D, is called the chemical diffusion coefficient. By virtue of Eqs. (10.9) and (10.11 ), one obtains ~_
1
(3"el(3"i~
3~t~
(10.12)
8 F 2 (~el + ($ion OCO
Here we note that Jo2 = 1/2jo. Because 3co/3X =-3Cv/3x, a similar expression is obtained when diffusions were dominated by neutral vacancies. The thermodynamic factor 3~o/3Co in Eq. (10.12) can be determined directly from experiment by measuring the oxygen stoichiometry as a function of oxygen partial pressure, either by gravimetric or coulometric measurements. In view of Eqs. (10.6) and (10.7), it comprises contributions from both ionic and electronic defects, which reflect their non-ideal behaviour. For materials with prevailing electronic conductivity Eq. (10.12) may be siml~lified to yield an exact relation between the chemical diffusion coefficient D and the oxygen tracer diffusion coefficient D*: D* D=~ HR
1,/23~to2
R T 3 In Co
(10.13)
Here Ha is the Haven ratio, defined as the ratio of the tracer diffusion coefficient D* to the quantity D ~ derived from dc ionic conductivity measurements, (~ion
Da = ~
RT
c o Z2 F 2
(10.14)
The Haven ratio may deviate from unity when correlation effects and possibly different jump distances and jump frequencies can not be neglected [51]. For a vacancy diffusion mechanism Ha equals the well-known tracer correlation factorf.
10 -- DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
453
10.3.1.3 Trapping of Electronic and Ionic Defects Equation (10.10) and those derived from it are valid as long as fully ionized oxygen defects contribute to transport. Different equations are obtained if valency changes of oxygen defects occur. Wagner [50] proposed to put the influence of reactions between ionic and electronic defect species in the cross terms of the Onsager equations. Maier [52-54] explicitly attributed individual diffusivities and conductivities to the new defect species, using the concept of a conservative ensemble accounting for free and trapped species. Following his approach, the reversible reaction between electrons and oxygen vacancies, Vo + e ' : Vo (10.15)
Vo + e': V ~ leads to the following expression for the oxygen flux,
Jo2 =
((~e' + (~h') ((~V6 + 4rYv6) + (~v6 r~v6 ] i [ I V,o2 (10.16) 42 F214s[vo~ + ((~e"+ (~h')+ ((~V6 + (~V6)
J
where we have adapted Eq. (33) in Ref. [53] (Part I) i n t o a form to be similar to Eq. (10.8), in which ionic transport is by doubly ionized oxygen vacancies only. The Onsager coefficient SWoaccounts for the contribution of neutral defects, enabling oxygen transport even when the electronic conductivity of the oxide is zero. We further note that the counter-diffusion of two Vo and a single Vo would result in a net neutral oxygen flux, as reflected by the last term in the numerator of Eq. (10.16). Maier [53] also examined the case in which electronic or ionic defects are associated (trapped) with immobile centres such as dopant ions. Trapping inevitably leads to a decrease in concentrations of the charge carriers available for transport. The impact of these phenomena is that the transport equations for evaluation of data obtained from electrochemical measurements like, for example, ionic conductivity, concentration cell, permeability and Hebb-Wagner polarization experiments should accordingly be modified. It is shown by Maier how these are influenced by trapping effects observed in perovskite SrTiO3, and by the transport properties of the high-temperature superconductor YBa2Cu306+x. Because of the large oxygen excess possible in the latter material it is assumed that transport occurs by differently ionized ionic defects, partly even by neutral oxygen species. For references, see the papers by Maier [52-54].
454
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
10.3.1.4 Empirical Equations Evaluation of j% from Eq. (10.10) requires that data exist for the partial conductivities r~io~and (3"elas a function of oxygen partial pressure between the limits of the integral. In what follows, some special relations for either prevailing electronic or ionic conduction are discussed. For the sake of approximation, in defect chemical studies often an empirical power law is used for the partial conductivity of the rate determining species, O'ir s u c h
asr
odP%) = r~~ P~2
(10.17)
where ~o is the conductivity at standard state. The value of n can be derived from experimental data of steady-state oxygen permeation. For proper evaluation it is necessary that the Po~-gradient across a specimen is varied within the assumed range of validity of the empirical power law. Inserting Eq. (10.17) in Eq. (10.10), one finds after integration, assuming ~i << (~totab J%
._ ~_
__
ppYt
(10.18)
[P'~)2 P 021
where ~ = r~~ For large positive values of n the rate of oxygen permeation is predominantly governed by the oxygen partial pressure maintained at the feed side, P02'. Likewise for large negative values it is predominantly governed by the oxygen partial pressure maintained at the permeate side, P02". For either a small value of n or a small Po2-gradient, the flux becomes proportional to In (Po~'/P02"). Provided that the electronic transference number is known, the ionic (and electronic) conductivity may be obtained by differentiation of experimental data. Assume that we have produced a data set for different P%-gradients, keeping the oxygen partial pressure P%" at the permeate side fixed. Differentiating Eq. (10.10) with respect to the lower integration limit yields,
La
/
RT = tel(~io n X 42
F 2L
(10.19)
The ionic conductivity at a given pressure Po2 is thus obtained from the slope of the jo~- In Po2' plot at that Po2. Similarly, the ionic conductivity can be evaluated from oxygen flux values measured by varying the oxygen partial pressure at the permeate side, keeping the one at the feed side fixed. The two data sets in general will yield the ionic conductivity over a complementary range in oxygen partial pressure.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
455
So from oxygen permeability measurements, it is possible to get information on the transport properties of oxides without making use of electrodes and external circuitry. In practice, however, experiments are plagued by the usual problems of sealing oxide ceramic discs at high temperature. A further complication is that the activity of oxygen at the oxide surfaces may not be precisely known, due to the influx and effiux of oxygen. In addition to these polarization effects, the results may be significantly affected by the flow patterns that possibly exist on feed and permeate sides. It was further assumed by Wagner that the surface reactions proceeding at the gas phase boundaries have reached a quasi-equilibrium condition relative to diffusion through the oxide. However, awareness is growing that in many cases the surface reaction may exert a partial control over the transport kinetics [41]. The extent of surface control varies with membrane thickness, temperature and oxygen pressure difference imposed across the membrane. Other limitations of the theory, some of which are briefly discussed in subsequent sections, include solid state diffusion of matter along preferred paths such as grain boundaries, porosity and, especially at large departures from the ideal stoichiometric composition, the formation of point defect clusters and ordering.
10.3.2 Surface Oxygen Exchange The exchange of oxygen between oxide surfaces and the gas phase has been recognized to involve a series of reaction steps, each of which may be rate determining [21]. Possible steps for the reduction of oxygen include adsorption, dissociation, charge transfer, surface diffusion of-intermediate species, e.g., O2ads, Oads and Oads, and finally incorporation in the (near-) surface layer. Generally, it is assumed that the reoxidation of oxygen anions follows the same series of steps in the reverse direction. The surface reaction is thus associated with transport of charge. The charge carriers in the interior of the oxide maintain thermodynamic equilibrium, according to Wagner's theory, but not at the surface if rate limitations by the surface exchange kinetics come into play. Liu [55] has presented a detailed analysis of the oxygen separation rates of mixed conductors, using the well-known Butler-Volmer formalism to model interfacial mass and charge transfer. At first glance such a description is not applicable because of the absence of electrodes. But, the adsorbate may be regarded as to replace the electrode material. Accordingly, the electrical double layer (Helmholtz layer) formed between the ionosorbed adsorbate and the oxide dominates the charge transfer kinetics, thus leading to the introduction of transfer coefficients in the rate constants for reduction and oxidation. In general, however, even at equilibrium, there may also be a double layer, called space charge, next to the surface extending into the oxide interior (MottSchottky layer). The width of the space charge layer will be of the order of the
456
10 - - D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
length, LD, the value of which depends on the volume concentration of all mobile charge carriers. Alteration of the space charge leads to a change in bending of the energy bands, and this influences the occupation of electronic levels near the band edges at the oxide surface. The total potential drop across the interface thus becomes distributed between the Helmholtz layer and the space charge layer [56], which complicates the analysis of the charge transfer kinetics. Furthermore, if strong electrical fields are developed in a situation where LD is greater than the size of the ionic charge carrier, which is favoured by a low concentration of mobile charges, the migration of ionic charge carriers through the space charge zone can be described by an expression similar to the Butler-Volmer equation [57,58]. The possibility of space charge in ionic crystalline solids, and the attendant redistribution of lattice components, has a great influence on the properties of boundary regions. Their defect chemistry may depart considerably from the predominant one in the bulk [59]. In oxides, there are quite a number of experimental observations to support the existence of space charge induced segregation [60,61]. Other factors have been recognized to contribute to segregation, like the misfit strain energy of the solute ion and the surface tension considering adsorption equilibria. Owing to the segregation phenomenon, the composition and crystal ordering of oxide surfaces and grain boundaries differ from that in the bulk. In some cases this leads to the formation of a second phase. It is clear that these phenomena will have a significant, often controlling influence on the properties of oxide materials, which includes the heterogeneous kinetics of the gas/solid interface.
10.3.2.1 Characteristic Membrane Thickness A simple yet valuable criterion for candidate membrane material selection is the characteristic membrane thickness Lc [41]. In the theory, the transport equations for diffusion in the solid, and for the surface exchange are linearized. It should therefore strictly be used when small Podgradients are imposed across the membrane. In the next section, methods for measuring Lc are briefly discussed. The starting point is to divide the membrane into a central bulk (Wagner) zone and adjacent interfacial zones, emphasizing the importance of both solid state diffusion and surface oxygen exchange to the magnitude of the oxygen flux. This is schematically shown in Fig. 10.3. The available driving force 9 total A~o 2 is distributed across the membrane such that the rate determining process receives the greater proportion. Diffusion is rate determining, if the membrane is made sufficiently thick. Upon reducing the thickness, the flux is controlled by the limited transfer of oxygen across the interfaces. Within the approximation of linear kinetics for diffusion and interfacial exchange, the flux equation can be written as [41]
10
-
-
457
DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
interracial
bulk
interfacial
zone
zone
I
I
I I I I
"~ I ~1 t -- ....
PO l '
T
I I I
~.
I
'~.
~.
PO=
I I
",,
I
I
I I
I ~,~ I ,,
I I9
I I
Fig. 10.3.Drop in chemicalpotential 1102across bulk m~dinterfacial zones of a membrane imposed to an oxygen partial pressure difference, P02' > P02". The largest drop occurs across the least permeable zone.
1 teltion(Stotal J02 = - 1 + (2Lc/L) 42 F 2
A,
total
/-l~tO 2
L
(10.18)
noting that in the limit of small Po2-gradients, the exchange rates at opposite interfaces will be identical. The quantity Lc, has been called the characteristic
thickness, RT Lc = 42 F 2 x
teltion(~total Je~
(10.19)
and determines the transition from p r e d o m i n a n t control by diffusion to that by surface exchange. The parameter j~ (mol 02 cm -2 s -1) expresses the balanced exchange rate in the absence of oxygen potential gradients and relates to the surface exchange coefficient ks (cm s -1) accessible from data of 180-~60 isotopic exchange, je~ = 1/4 ks Co
(10.20)
in which co is the volume concentration of oxygen anions at equilibrium. Exchange rates are customarily defined in terms of moles of anions or molecules per unit time and area. Often the geometric area is used in calculation, in spite of the fact that the true surface area on a microscopic scale may be substantially larger. C o m p a r i n g Eq. (10.18) with the Wagner equation, we see that the diffusional flux of oxygen across the m e m b r a n e is reduced by a factor (1 + 2Lc/L) -~, relative to that in the absence of transfer limitations across the interfaces. In general, the surface exchange rate will be different in, e.g., O2-N2, CO-CO2, H2-H20 atmospheres even at approximately equal oxygen partial pressures. If opposite sides of the m e m b r a n e s are exposed to different gas ambients, then the interface with the lesser performance dictates the overall exchange behaviour.
458
10 -- DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION 1.00
~
_. ),
1.00
0.80 0.60
0
-,=,=,,t
0.10 0.40 oo
.............
0.20
0.01 0.01
. . . .
9, L - " " " "
7|"
0.10
1.00
10.00
0.00 100.00
L/L,, Fig. 10.4. Thickness dependence of the dimensionless oxygen flux j'o2 calculated using Eq. (10.18). The quantity j'o2 is defined by the ratio of the oxygen flux over the maximum achievable oxygen flux in the surface exchange limited regime. Only if L >> Lc, the oxygen flux becomes proportional to 1 / L~with 7 = I in agreement with the Wagner theory. For smaller thicknesses, 7ranges between I and 0 (right-hand scale).
The influence of the membrane thickness on oxygen flux is shown in Fig. 10.4. When L > Lc, the oxygen flux varies inversely with Lv where g = 1, in agreement with Wagner's theory (cf. Eq. (10.10)). Departures from this inverse relationship are observed when the oxygen flux becomes partly governed by the surface exchange kinetics. The value of 7, at a given L, corresponds with the negative slope in the double logarithmic plot of the oxygen flux versus membrane thickness at that L. Taking the logarithm of Eq. (10.18), partial differentiation with respect to log L shows that y is equal to the reduction factor (1 + 2Lc/L) -1, the value of which gradually decreases with decreasing thickness to become zero for L << Lc, as is shown in Fig. 10.4. The latter situation corresponds with the maximum achievable flux, which for a symmetrical membrane is given by: 1/2 ]ex' "o -- total A~l.O2 9 Equation (10.19) may be simplified for predominant electronic conduction, assuming that the classical Nernst-Einstein relationship can be represented as co D s z2 F 2 O'i~ --
RT
(10.21)
where Ds is the self-diffusion coefficient of oxygen anions with valence charge zo (= -2). Making the appropriate substitutions, the characteristic thickness Lc becomes Ds D* Lc- ks - k--7 (tel = 1) (10.22)
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
459
In the second part of Eq. (10.22) the fact has been used that, if correlation effects can be neglected, the tracer diffusion coefficient, D*, is equal to the self-diffusion coefficient, Ds. It is important to note once again that Eq. (10.18) is valid in the limit of small Podgradients only. Since both D* and ks for a given material are a function of its specific defect chemistry, in general, Lc will be a function of process parameters Po2 and temperature. The picture that emerges is that, at given experimental conditions, for thicknesses below Lc no appreciable gain in the oxygen flux can be obtained by fabricating thinner membranes, unless the value of ks can be significantly increased. Similar criteria can be formulated for fuel cell electrodes as has been advanced by Steele [62,63] and Kleitz [64]. In the limit of small overpotentials (low field approximation), i.e. assuming ohmic behaviour for the relevant surface kinetics, whatever the rate controlling mechanism, the electrode resistance can be correlated with the electrolyte resistivity.
10.3.2.2. MeasuringLc The quantity D*/ks appears to be a fundamental parameter governing tracer diffusion bounded by a limiting surface exchange between lattice oxygen and oxygen from the 1802-enriched gas phase. Experimental methods for isotopic exchange include monitoring the change of 1802 concentration in the gas phase upon exchange in a fixed volume of 1802-enriched oxygen using a mass spectrometer [65-69], weight measurements [70], depth probing of 180/(180 + 160) diffusion profiles using secondary ion mass spectroscopy (SIMS) after exchange at high temperature for a selected time [71-73] and combined approaches [74-76]. Fitting the acquired data to the appropriate diffusion equation allows both D* and ks to be obtained from a single experiment. Selected data of D* and ks from 180-160 isotope exchange measurements of perovskite-type oxides are compiled in Table 10.2. High D* and ks values are reported for the ferrites and cobaltites, in which solids the anions are assumed to move via a vacancy diffusion mechanism. The mixed perovskites are notable for being excellent electronic conductors. Usually the electronic conduction is found to be predominant, in spite of the fact that the ionic conductivity may also be substantial (Section 10.6). The examples given in Table 10.2 clearly emphasize the importance of the surface exchange kinetics, relative to diffusion, in limiting overall oxygen transport through the perovskites. Analyzing published data for D* and ks, Kilner [73] noted that the two parameters seem to be correlated. A square root dependence of ks with D* is found for perovskite oxides ABO3, albeit that the data show substantial scatter, yielding an average value for D*/ks of about 100 gm. For fluorite oxides M O 2 the two parameters are correlated almost linearly, while the corresponding value of D*/ks ranges between mm and cm. The results suggest that related
460
10 -- DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
TABLE 10.2 Tracer-diffusion coefficient D*, surface exchange coefficient ks and characteristic thickness Lc for selected perovskite-type oxides Perovskite
T (~
PO2 (kPa)
D* (cm 2 s -1)
ks (b) 1 (cm s- )
Lc (a) (cm)
Ref.
Lao.sSro.sMnO3-6
700 800 900
70
2x10 -15 8x10-14 3x10-12
1x]0 -8 l x l 0 -7 9x10~
2x10-7 8x10 -7 3x10 -5
1
Lao.sSr0.2CoO3,5
700 800 900
70
l x l 0 -8 2x10 -8 4x10~
3x10~ 5x10-6 2x10 -5
3x10 -3 4x10 -3 2xlO -3
1
Lao.6Cao.4Coo.8Feo.203-5
700 800 900
70
2x10~ 1x10-7 3x10 -7
4x10~ 2K10-s 4x10-5
5x10-3 5x10-3 7x10 -3
1
Lao. 6Sr0.4C00.8Ni0.20345
700 800 900
70
3x10 -8 lx10 -7 4XlO--7
2KIO-6 2x10 -6 2XlO4
2x10 -2 5x10 -2 2xlO q
2
Lao.oSro.4Coo.6Nio.403.6 (c)
700 800 900
70
2X10-9 6x10 ~ 3x10 -7
7x10-7 3x10 -6 3x10 -6
3XlO-3 2x10-2 lx10 -1
2
Lao.oSro.4Coo.4Nio.603-6
700 800 900
70
lx10 ~ 7x10~ 6x10 -7
3x10-7 2x10 -6 2x10 -4
3x10-2 3x10-2 3x10 q
2
LaCoO3_8 (single crystal)
700 800 900
4.5
9x10-13 2XlO-11 6x10 q ~
1x10-9 3KI0-7 lx10 -6
4x10 -4 7x10 -5 4x10-4
3
LaFeO3_5 (single crystal)
900 1000
7
lx10 -i2 5x10q2
4x10-8 2x10 -7
3x10 -5 3x10 -5
4
Lao.9Sro.lCoO3.8
900 1000
4.5
3x10-9 2x10~
lx10 -6 2x10~
2XI0-3 l x l 0 -2
5
Lao.gSro.lFeO3.6
900 1000
6.5
3x10 -9 lx10 -8
5x10-7 2x10-6
6x10-3 6x10-3
5
Lao.oSro.4FeO3.6
1000
6.5
6x10-7
lx10 -5
5x10 -2
5
(a) Calculated using Eq. (10.22). (b) Value can be equated to ]~• in accordance with Eq. (10.20), b y multiplication with a factor 0.022. Strictly speaking, this holds for LaCoO3. (c) A u t h o r s report a two-phase mixture [2]. REFERENCES USED IN TABLE 10.2: 1. S. Carter, A. Selcuk, J. Chater, R.J. Kajda, J.A. Kilner a n d B.C.H. Steele, Solid State Ionics, 53-56 (1992) 597--605. 2. Ch. Ftikos, S. C a r t e r a n d B.C.H. Steele, J. European Cer. Soc., 12 (1993) 79-86. 3. T. Ishigaki, S. Y a m a u c h i , J. M i z u s a k i , K. Fueki a n d H. T a m u r a , J. Solid State Chem., 54 (1984) 100-107. 4. T. Ishigaki, S. Y a m a u c h i , J. M i z u s a k i , K. Fueki a n d H. T a m u r a , J. Solid State Chem., 55 (1984) 50-53. 5. T. Ishigaki, S. Y a m a u c h i , K. Kishio, J. M i z u s a k i a n d K. Fueki, J. Solid State Chem., 73 (1988) 1
rwt-~
.t ~t,-n
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
461
point defect processes must be common to both diffusion and surface oxygen exchange [73]. However, mechanisms responsible for the apparent correlations remain obscure, reflecting our poor knowledge at present of the factors that control the oxygen exchange kinetics. As discussed in previous work by the present authors [41], calculations of Lc using combined data from ionic conductivity measurements and 180-160 isotopic exchange for a number of fluorite oxides were found to be in good agreement with values estimated from oxygen permeability measurements. Finally, it may be noted that also relaxation methods offer a useful tool for measuring Lc, without the requirement of high-temperature seals as in oxygen permeation experiments. The (re-)equilibration may follow after an instantaneous change of the oxygen activity in the gas phase and involves the propagation of a composition gradient through a thin slab or single crystal of the oxide. The change in stoichiometry brings about a change in weight, or in electrical conductivity, which can be monitored experimentally as a function of time. It is not possible to give a full account of these techniques in this chapter and the reader is referred to, for example, Refs. [77-79].
10.3.2.3 The Effect of Surface Roughness and Porosity The parameter Lc does not represent an intrinsic material property, but may depend through the value of ks on the roughness or porosity of the membrane surface. This has recently been exploited by Thorogood et al. [42] and Deng et al. [43], showing that the oxygen flux can be significantly improved if the thin dense membrane is coated on either one or both surfaces with a porous layer. Based on a simple effective medium model, and linearized transport equations, Deng et al. [43] calculated the oxygen flux through the modified membrane whose dense layer thickness is assumed to be small enough so that the drop in chemical potential across it can be neglected. The rate limiting step is thus ionic diffusion and surface exchange in the porous solid. The latter is modeled by a simple cubic array of consolidated spherical grains. In the limit of small Poagradients, the maximum enhancement in the oxygen flux through the symmetric membrane, over the non-coated membrane, is given by = qLcS(1 - e)/~
+ e
(10.23)
where S is the pore wall surface area per unit volume, 0 the porosity and ~s the tortuosity of the solid phase in the porous structure. Maximum enhancement is achieved for a membrane whose porous layer thickness L >> Lp where ap - qac(X - 0)/ST
s
(10.24)
462
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
and refers to the active part of the porous layer. To achieve near full enhancement L - 3Lp suffices. With Lc = 100 ~tm and surface area S = 10-6 cm -], the authors calculated an enhancement in oxygen flux, over the non-coated m e m brane, of almost two orders of magnitude. In a separate paper, Deng et al. [44] showed that the factor ~ is substantially reduced when the rate limiting step is the transport of gas molecules in the pores. Finally, we note the potential enhancement in oxygen fluxes, at thicknesses below Lc, by coating the dense m e m b r a n e with an exchange active layer.
10.4 SOLID OXIDE ELECTROLYTES
10.4.1 Introduction
A key factor in the possible application of oxygen ion conducting ceramics is that, for use as solid electrolyte in fuel cells, batteries, oxygen p u m p s or sensors, their electronic transport n u m b e r should be as low as possible. Given that the mobilities of electronic defects typically are a factor of 1000 larger than those of ionic defects, a band gap of at least 3 eV is required to minimize electronic contributions arising from the intrinsic generation of electrons and holes. Useful solid oxide electrolytes to date are those with a fluorite or fluorite-related structure, especially ones based on ZrO2, ThO2, C e O 2 and B i 2 0 3 [80]. Mixed conduction occurs only at sufficiently low or high values of Po2, where electronic defects are generated for charge compensation of the excess of ionic defects relative to the stoichiometric composition. This latter mechanism proceeds only w h e n the oxygen defects m vacancies or interstitials - - introduced by equilibration of the oxide with the gas phase have relatively low ionization energies and thus ionize under the selected conditions. Taking into account the mobilities of ionic and electronic defects, the range in nonstoichiometry can be correlated with the width of the electrolytic domain. For the stabilized zirconias the domain width, at 1000~ typically extends to values below Po2 = 10-30 atm [81]. On the other hand, the domain width of ceria electrolytes is limited, reflecting the ease of reduction of C e 4+ to C e 3+ relative to that of the transition metal ions in stabilized zirconia. For example, at the same t e m p e r a t u r e (CeO2)0.95--(Y203)0.05has its domain boundary at about Po~ = 10-10 atm, the value of which has been taken for tion = 0.5 [82]. As is clear solid oxide electrolytes are not useful for applications as oxygen separation membrane, unless operated with external circuitry (oxygen p u m p ) or as a constituent phase of a dual-phase membrane. Both modes of operation, classified in this paper as electrochemical oxygen separation, are briefly discussed in Section 10.4.3. But we first start with a discussion of the models that have been developed to describe the oxygen semi-permeability of solid oxide
10 -- DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
463
electrolytes, originating from the residual electronic conductivity in the electrolytic domain. These models can be translated easily to mixed conductors. Examples are drawn from experimental studies on calcia-stabilized zirconia and erbia-stabilized bismuth oxide, clearly emphasizing the importance of both bulk diffusion and surface exchange in determining the rate of oxygen permeation through these solids.
10.4.2 Oxygen Semi-permeability of Oxide Electrolytes 10.4.2.1 Diffusion of Electronic Charge Carriers In the absence of interactions between defects, the following reactions determine the defect concentrations in oxides with the fluorite structure, O2,g 4- 2 V6 ~
2 0 ~ + 4h.
(10.25)
O~ ~
O / ' + Vo
(10.26)
nil ~
e' + h.
(10.27)
with the respective equilibrium constants, p4 [0~)]2 ]2
(10.28)
K F = [Oi"] [W~]
(10.29)
Ke =np
(10.30)
Kg =
Po~ IV6
Here n and p denote the concentrations, expressed as mole fractions, of electrons and electron holes, respectively. The anti-Frenkel defects Oi" and V6 are assumed to be fully ionized, which is usually observed at elevated temperatures. Whilst phases derived from 8-Bi203 show substantial disorder in the oxygen anion sublattice, reaching a maximum value at the disordering temperature, doping with aliovalent impurities is essential to achieve high ionic carrier concentrations in oxides such as ThO2, HfO2 and ZrO2 [83,84]. Doping sometimes serves to stabilize the cubic fluorite structure down to working temperatures, e.g. for HfO2 and ZrO2. In the following, we use notations D and A for dopant and acceptor cations, respectively. The electronic conductivity, comprising p-and n-type contributions, is obtained by multiplying each of the concentrations with their respective charge and mobility. Upon substitution of the defect concentrations established by the above equilibria in the electroneutrality relation, 2[Vo] + p + [D'] = 2[Oi"] + n + [A']
(10.31)
the following expression for the partial electronic conductivity can be derived,
464
10 w DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
(~el --pFUh + nFue (10.32) o D1/4 -b o = (3"p ~ O 2 (~n
P
O12/4
where Uh, Ue are the mobilities of electron-holes and electrons, respectively, and r~, c~~ the corresponding partial conductivities extrapolated to unit oxygen partial pressure. In deriving Eq. (10.32), the concentration of oxygen vacancies [Vo in the electrolytic domain is taken to be fixed, either by the Frenkel equilibrium given in Eq. (10.26) or by the net acceptor dopant concentration: [A'] - [D]. Substitution of Eq. (10.32) into Wagner's equation (Eq. (10.10)) yields for the oxygen flux, after integration RT Jo2 - 4F 2L [r~ t~o2 ,o ,1/4 - "02 D --1/4,j - r~, O t~o2 ,O ,-1/4 - Po2 ,,-1/4 )]
(10.33)
noting that the ionic transference number, tion, in the electrolytic domain has been set to unity. If the experimental conditions are chosen such that the oxygen pressure at the permeate side is not too low, i.e. neglecting the n-type contribution to the electronic conductivity, Eq. (10.33) reduces to (cf. Eq. (10.18)),
jo -T-orO
,1/4
O
,,1/4,
J
(10.34)
o RT/4F 2 where [3 = r~p Because of its great technological importance, much literature on the 'detrimental' oxygen semipermeability flux occurring in solid electrolytes is available [85-98], of which a great number deals with the stabilized zirconias, but also includes electrolytes based o n T h O 2 [97,98], H_fO2 [87] and Bi20 3 [96] Results on the stabilized zirconias up to 1976 have been reviewed by Fouletier et al. [99]. For relatively thick electrolyte membranes the results are consistent with Eq. (10.34). But, often a value between 1/4 and 1/2 is found for the exponent [91-96]. Sometimes this has been taken as evidence for electronic trapping, i.e. a different mechanism for the incorporation of oxygen in the oxide lattice [91,92]. Dou et al. [95] were the first to invoke a surface reaction to reconcile the apparent conflict with a diffusion controlled mechanism, i.e. the overall kinetics is determined both by surface reactions and by bulk diffusion. 10.4.2.2 Modelling Equations
In this section, a model is presented for solid oxide electrolytes based upon two consecutive steps for oxygen permeation; one for the surface exchange process at the oxide surface on both sides of the membrane, and another for the joint diffusion of oxygen ions and electron-holes through the solid.
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
465
Considering oxygen exchange between the gas phase and the oxide surface via the reaction given by reaction (10.25), one may distinguish many steps, like adsorption, dissociation, surface diffusion, charge transfer and incorporation in the (near) surface layer, and reversed steps, each of these steps may impede interfacial transfer of oxygen. Following Dou et al. [95], the surface reaction may be represented by a simple two-step scheme: O2,g ~
2Oad s
(10.35) O ad s 4- V 6 ~
O~) 4-
2 h"
If it were supposed, for example, that the first of these reactions is at equilibrium and the second is rate-determining, the net flux of molecular oxygen through the interface at the high pressure side can be described by, 1/2 _ 1/2 k_2 [O~)] p2 jo~ = 1/2 k2 K~/2 [V61 Po2
(10.36)
where k2 and k-2 are the forward and backward rate constant for the ionization and incorporation reaction (2nd step in (10.35)), K1 is the equilibrium constant for the adsorption reaction (1st step in (10.35)). The reverse equation holds for the rate of the surface reaction at the opposite side of the membrane (permeate side). As mentioned above, in the electrolytic domain the concentration of oxygen vacancies, [Vo] (and that of lattice oxygen, [O~)]) is constant. Accordingly, by applying the law of mass action for the overall exchange reaction given by Eq. (10.28) the actual concentration of electron holes p at either side of the membrane can be equated to a virtual oxygen partial pressure, i.e, the oxygen partial pressure if equilibrium were established with the gas phase. Considering both surface reactions and bulk diffusion, wearrive at the following set of equations for the oxygen flux rD ,1/2 j o 2 = (X [zO2
-
1/2] FOu(I )
jo 2=~- [Po2q)1/4 - Po2(II)1/4]
(10.37)
jo 2= 0~ [Po2(H)1/2- Po2''1/2] where ~ = 1/2 k2 K~ [V6] and Po~(~),PO~(H)indicate the virtual oxygen pressures at the feed and permeate side, respectively. The usual assumption of fast equilibration of the oxide surface with the imposed gas atmosphere would imply that PoS = Po~(i) and Po~" = PO~(H).Finally it may be noted that the set of equations given in Eq. (10.37) can only be solved numerically.
10.4.2.3 Examples (a) Calcia-stabilized zirconia Dou et al. [95] studied the isothermal oxygen permeation through calcia-stabi-
466
10 -- DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
I0 0
K, 8 x
% 4
_ 0.005
9
Oo
I
I
I 0.1
I
r
-
-
-
I, 0.2.
,
cm Fig. 10.5. The effect of sample thickness on oxygen permeation through calcia-stabilized zirconia at 1230~ Solid lines are theoretical results calculated using Eq. (38). Reprinted from Dou et al. [95].
lized zirconia (CSZ) tubes at 960-1450~ and oxygen pressures Po2' of 1 0 -3 to 1 atm. The oxygen which permeated into the interior of tubes with known wall thicknesses was immediately pumped away by a diffusion pump and measured with a gas burette (10-6 < Po2" < 10-4 atm). The data obtained qualitatively agreed with an oxygen pressure dependence in accordance with Eq. (10.18), in which the value of the exponent n is allowed to vary between 1/4 and 1/2. Figure 10.5 shows the measured effect of sample thickness on oxygen flux. Assuming Po2" = 0 atm, and with neglect of rate limitations at the low pressure interface, Dou et al. [95] arrived at the following expression for the oxygen flux, Jo2 - 20~L 2
I/4o2,o 1
+
~2
~02
-- I
]
(10.38)
which equation matched the experimental results well. The parameters c~ and showed similar activation energies" 191 + 5 kJ mole -1 and 206 + 11 kJ mole -1, respectively. The quantity ~/offPo~') 'a has the unit of length. The meaning of it is more or less similar to that of the parameter Lc defined earlier in Section 10.3.2.1. At 1230~ and oxygen pressure Po2' of 1 atm, the transition from predominant control by bulk diffusion to that by surface exchange would occur at a sample thickness of about 2.7 x 10-2 cm. Dou et al. [95] performed their experiments on CSZ tubes with a homogeneous composition, having 10% pores by volume. The authors estimated that the
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
467
true surface area would be about 10% greater than the geometrical one used in calculation, and the surface exchange parameter a needed accordingly to be reduced for an ideal surface without pores. For the bulk transport of oxygen, the effect of non-connected micropores would be either to shortcut the solid state diffusion, in the case of fast surface exchange kinetics, or to enlarge the diffusion path to an extent which is a function of the pore size. For fully dense ceramics, the bulk diffusion parameter ~ was expected to be about 14% greater. Besides the possibility of fitting the experimental data to Eq. (10.38), the authors demonstrated that a more complicated mechanism for the surface exchange reaction may be invoked. Interestingly, steady-state oxygen permeation measurements by Dou et al. [95] provided no evidence of a surface rate limitation for CSZ tubes containing a segregated impurity phase. The order with respect to oxygen remained close to 1/4. This second phase consisted of a mixture of metal silicates with a composition similar to that of a SiO2-CaO-A1203 eutectic. It was suggested that surface oxygen exchange on this second phase would be very rapid. In subsequent studies, the same authors used the 'time-lag' method (in which the transient process towards steady-state oxygen permeation is monitored) and a desorption technique to study chemical diffusion in CSZ with different impurities [100-102]. The observed kinetics scaled with the presence of Fe203, in such a way that faster equilibration rates were observed for samples containing a smaller impurity content. The results were taken to be consistent with a mechanism in which electron-holes are trapped at iron impurity sites.
(b) Erbia-stabilized bismuth oxide Indications of a limitation by a surface process have also been reported for the oxygen flux through sintered dense ceramics of bismuth oxide stabilized with 25 mol% erbia, (Bi2OB)0.75--(Er203)0.25(BE25) [96]. As is known, this material exhibits high ionic conductivity at a temperature distinctly lower than for the stabilized zirconias [103]. Using glass-sealed discs, the amount of oxygen which permeated into a closed reservoir, being flushed with helium gas prior to measurement, was monitored as a function of time. The oxygen flux was calculated from the corresponding slope, which was found invariant as long as PO2' > > P o 2 " .
In the range of temperatures (610-810~ and oxygen pressures (10-4 - 1 atm) covered by experiment, the concentration of minority charge carriers, i.e., electron-holes, in BE25 is proportional to P~2 with n = 1/4. However, the apparent value derived from experiment increases gradually from 1/4 to higher values upon decreasing specimen thickness from 0.285 cm to 200 gm, indicating permeation to be limited by two or more processes differing in order. The activation energy of the oxygen flux was found to increase too in the same direction. The observed behaviour can be attributed to the change-over from diffusion to
468
10 - - DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
surface control u p o n decreasing sample thickness. The experimental data can be fitted well by means of Eq. (10.37), though it is necessary to adapt the kinetic order of the surface reaction with respect to oxygen to a value of 5/8. The parameters c~ and 13obtained from numerical fitting appear to exhibit different activation energies; 136 + 4 kJ mole -] and 99 + 4 kJ mole -1, respectively, which indicates that the surface process is less limiting at higher temperatures. Isotopic exchange measurements on sintered dense discs of BE25 showed a P~2 dependence with m = 0.60 at 550~ and m = 0.54 at 700~ for the overall surface oxygen exchange rate [67,104]. Figure 10.6 shows that the value for the surface oxygen exchange rate j~x (= (~ Pc~2),normalized to air, obtained from the fit of the data agrees with that measured by isotopic exchange. The thickness, at which point the oxygen flux is half of that expected under conditions of pure diffusion-controlled kinetics, imposing opposite sides of discs to pure oxygen and helium gas, was calculated at 0.16 cm at 650~ and 0.09 cm at 800~ These values were found to be in good agreement with estimates of the parameter Lc as noted before in Section 10.3.2.2.
-6.50
=r~
opic exchange
-7.50
0
oxygen .
u.._.l
X,,,.~
o
-8.50 O
-9.50 0.90
1.00
1.10
1.20
1.30
IO00/T[K] Fig. 10.6. Data for the surface oxygen exchange rate, normalized to air, of 25 tool% erbia-stabilized bismuth oxide (BE25) from (a) isotopic exchange and (b) oxygen permeation measurements. Reprinted from Bouwmeester et al. [96].
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
469
10.4.3 Electrochemical Oxygen Separation
10.4.3.1 Oxygen Pump The open-cell emf generated across an oxygen concentration cell such as
O2(Po2' ), Pt I CSZ IPt, O2(Po2" )
(10.39)
with each side maintained at a different oxygen partial pressure Po2' and Po2" is given by, Eeq =
(1 -
_
RT
Po2'
tel ) - ~ - I n ~ Po2"
(10.40)
where tel is defined as a mean electronic transference number. In the absence of any electronic conduction, i.e. when tel = 0, Eq. (10.40) simplifies to the Nernst equation. When the cell arrangement delivers a current I under load conditions, the cell voltage drops below the value Eeq, due to ohmic losses IRi (Ri = electrolyte resistance) and polarization losses at both Pt electrodes. As an approximation,
E=
Eeq - I R i - n
(10.41)
where 1] represents the total cathodic and anodic polarization loss. Upon shortcircuiting both Pt-electrodes, the emf of the cell drops to zero while oxygen is transported from the high pressure side PO2' t o the!ow pressure side P O 2 " . By applying an external power source, the applied dc voltage can be used to enhance the magnitude of the current but also to reverse its sign. That is, oxygen may be pumped in both directions; the rate of transport equals I/4F according tO Faraday's law. This is the basic principle of electrochemical oxygen separation. An important phase during device development is optimization of the pumping rate, i.e. ohmic and polarization losses must be kept as low as possible. Much efforts have been concentrated on development, fabrication and testing of zirconia-based separators. For example, Clark et al. [105] has described the performance of a multi-stack yttria-stabilized zirconia (YSZ) based separator. Each cell contained a 125 ~tm thick YSZ layer of diameter 6.35 cm, whereas porous strontium-doped lanthanum manganite electrodes were used to eliminate the need for costly Pt. The largest of these separators, built with 20 cells, was found to be capable of an oxygen flux up to 1 1 min -1 at an operating temperature of 1000~ Factors influencing the efficiency of the oxygen separation process and systems analysis of conceptual oxygen production plants are also addressed. A major drawback of ZrO2-based materials is the high temperature required for operation, typically 900-1000~ expressing the need for development of oxide electrolytes which exhibit significant levels of ionic conduction at modest temperatures. Several alternative materials may be considered. To provide a
470
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
reference point for discussion, the ionic conductivity of YSZ is about 0.1 S c m -1 at 950~ This value is found in bismuth oxide stabilized with dopants such as E r 2 0 3 and Y203 and in cerium oxide doped with Gd203, Sm20 3 or Y20 3 already in the range 650-700~ [62,63] which electrolytes are less useful in, for example, fuel cells or sensor applications due to the presence of rather reducible ions Bi3+ and Ce 4+ and, hence, a non-negligible contribution of electronic conduction. The suitability of Bi0.571Pb0.42801.285 as electrolyte membrane has been proposed for temperatures as low as 600~ [106]. This material suffices however from structural instabilities. Having its mechanical properties enhanced by incorporating Z r O 2 into the starting material, the optimized membrane is able to operate continuously up to 300 mA c m 2 at 600~ Fast ionic conduction at modest temperatures has also been reported in Bi4V2_yCUyOll (BICUVOX) 3 [107-109], which phases possess an intergrowth structure consisting of Bi2 O2+ blocks alternating with perovskite blocks. The material Bi2V0.gCu0.105.35was found to exhibit an ionic conductivity of I x 10 -3 S c m -1 already at 240~ which is about two orders of magnitude higher than that of stabilized bismuth oxide [108]. In most cases the ability of these electrolytes for electrochemical oxygen separation has not yet been fully explored. Thus, it can not be excluded that relevant properties like, for example, oxygen ion conductivity, phase stability, gas tightness, mechanical strength and compatibility with electrode materials will not be affected during prolonged operation. Of course, the current-voltage characteristics and operational life are influenced not only by the quality of the solid electrolyte but also by the properties of the electrodes. For a recent review on oxygen electrode kinetics, see Ref. [64].
10.4.3.2 Dual-phase Composites As seen from Table 10.1 impressive oxygen fluxes have been reported through 25 mol% yttria-stabilized bismuth oxide (BY25) [110] and 25 mol% erbia-stabilized bismuth oxide (BE25) [111,112], which oxide electrolytes were rendered electronically conductive by dispersion with silver metal. A prerequisite is that both constituent phases in the composite membranes do form a continuous path for both ionic and electronic conduction, having their concentrations above the critical (percolation threshold) volume fraction ~)c.The latter quantity determines the minimum volume fraction in which conduction is possible and is a function of, for example, the relative dimensions and shape of the particles of both constituent phases [113]. In actual composite materials, 3
It may be noted that BICUVOXrepresents only one member of a family of Bi203-based solid electrolyte phases, whichmay be derived from Bi4V2Ollbysubstitution of copper for vanadium. Many cations may be substituted for vanadium and the general acronym BIMEVOXwas given to these materials, which have been claimed for electrochemical oxygen separation at temperatures as low as 500 K [109].Besides copper, high oxide ion conductivity is reported for substituents titanium and niobium [212].
10 m D E N S E C E R A M I C M E M B R A N E S FOR OXYGEN S E P A R A T I O N
471
however, the interconnectivity between particles will not be ideal. These may be linked up to form so-called dead-ends or isolated clusters, which do not contribute at all to the conductance of the percolative system. Accordingly, conduction is expected to proceed through a significantly smaller fraction of consolidated particles or grains, which implies that the actual volume fraction of each phase should always be somewhat in excess of ~c. The optimum volume ratio is just above ~c of the high conducting phase, i.e. the metal phase, in order to have the highest effective ionic conductivity of the composite. Dual-phase membranes made of BY25-Ag [110] and YSZ-Pd [114] behave quite similar in having their conductivity threshold at about 33-35 vol% of the metal phase. These membranes were made by conventional ceramic processing techniques. The value of ~c obtained for these composite materials agrees well with the high concentration limit predicted by simple effective medium theory in which the composite is described as a three-dimensional resistor network [115]. The effective ionic conductivity is reduced relative to that what is expected purely on the basis of the volume fraction of the ionic conducting phase, which originates, at least partly, from the enhanced tortuosity of the migrating path for the oxygen anion due to partial blocking by the metal phase. It is therefore expected that a further gain in the oxygen flux can be realized through proper design of the microstructure [112,114]. The optimum situation would correspond with the one in which the particles of each phase line up in strings (or slabs) parallel to the applied gradient in oxygen partial pressure. Even though, theoretically, the critical volume fraction of the metal phase could be reduced in this way to a value practically equal to zero, such an approach is bounded by the additional requirement for practical membranes of fast surface exchange kinetics, especially for very thin membranes. The exchange reaction at the composite surface is confined to the three-phase boundary (tpb) between the gas, metal and electrolyte formed by particle grains being connected to the percolative network. Fast oxygen transfer can be sustained only if the corresponding length or area available to oxygen exchange is large enough, where it should be noted that the exchange reaction can only take place at a point remote from the tpb line which is shorter than the spill-over distance of electro-active species across the surface. The electrical field necessary to guide the current becomes distorted in the vicinity of the surface of a coarse-grained composite, where the separation between adjacent tpb lines is too large and, hence, only part of the surface is effective towards oxygen exchange. This contribution is stressed in the SOFC literature and is known as the constriction effect [116]. Often, it is the synergism between electrode and electrolyte material that leads to fast exchange characteristics. The oxygen flux through disc membranes made of BE25-Au (40 vol%) was found to increase almost one order of magnitude by substituting gold for silver in the composite [112]. This observation can be related to the higher activity of silver in the
472
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
oxygen exchange reaction on BE25, compared with gold, imposing less limitations on overall oxygen transport. Materials like, for example, Bi2CuO44 [117], TiN [112], MgLaCrOg,s [26] have been proposed to replace the inert metals. Even though, in the examples chosen, ionic and electronic transport are confined to separate phases, mixed ionic-electronic conductors could be useful. A systematic evaluation of dual phase membranes, however, is too new so far to come to definite conclusions. Besides simple modelling in terms of a short-circuited oxygen concentration cell, to our knowledge no one has yet described oxygen permeation through dual-phase membranes, taking into account the distinct three-dimensional aspects of the microstructure that may arise in practical composite materials. Besides high values for the oxygen flux (and permselectivity), commercial use of membrane systems will demand chemical, mechanical and structural integrity of applied materials in appropriate ranges of temperature and oxygen partial pressure. Dual-phase membranes have the obvious potential to distribute specific requirements among the system components. INTRODUCING ELECTRONIC CONDUCTION IN FLUORITE-TYPE OXYGEN ION CONDUCTORS
10.5
10.5.1 Introduction Stimulated by the search for candidate materials for electrodes in solid oxide fuel cells (SOFC) and oxygen separation membranes, researchers have explored the possibility of introducing electronic conductivity in oxygen-ion conducting fluorite-type matrices by doping with multi-valent dopants. The major factors which establish electronic conduction in the mixed-conducting oxide solid solutions obtained are, at a given temperature, (i) the multi-valent dopant fraction, (ii) its redox characteristics and (iii) oxygen partial pressure. The suggested mechanism for electronic conduction is the hopping of electrons between adjacent dopant ions of different valence charge. However, experimental data of oxygen permeation is still scarce. In the following sections, we briefly focus on the defect chemistry, which includes some fundamentals of mass transport, and shall summarize relevant work on selected oxides. 10.5.2 Defect Chemistry
The topic of 'mixed conduction in nonstoichiometric oxides" was reviewed by Tuller [24], and his comprehensive paper is recommended to the reader interested in more detail concerning the role of multivalent dopants on the defect chemistry of fluorite and fluorite-related oxides, and corresponding transport properties. Equations which express the oxygen flux in solid solutions of, e.g.,
10 -- DENSECERAMICMEMBRANESFOR OXYGENSEPARATION
473
ceria in stabilized zirconia, as a function of temperature, oxygen partial pressure and dopant concentration have been developed recently by Ling et al. [118] and Marques et al. [119]. In addition to the defect reactions given in Section 10.4.2.1, one extra reaction needs to be considered, i.e. the ionization of the multivalent cation. On using the general notation N for the multivalent cation one may write, N M' ~--- N~ + e
(10.42)
with equilibrium constant, [N~] n
KN=
[NM']
(10.43)
Mass conservation requires that [NM' ] + [N~4 ] -[NM]tota 1
(10.44)
Electroneutrality relation Eq. (10.31) must be rewritten to include the charged species NM': 2[Vo] + p + [D'] = 2[00" ] + n + [A'] + [NM']
(10.45)
With the aid of experimentally derived equilibrium constants, Eqs. (10.42)(10.45) may be used to construct the Kr6ger-Vink defect diagram, from which expressions for the partial conductivities of the mobile ionic and electronic defects can be derived [24]. The defect diagram obtained by Marques et al. [119] for lightly ceria-doped ZrO2-Y203, ignoring the possibility of defect association, is shown schematically in Fig. 10.7a. The corresponding conductivity diagram (Fig. 10.7b) can be obtained by multiplying each of the mobile species by their respective charge and mobility, which leads to the following expression for the total electronic conductivity, Gel = (~p 4- (~n 4- (~h
(10.46)
where (~n and o v represent the intrinsic n-type and p-type conductivities, respectively, and Oh is the extrinsic electronic conductivity owing to the multivalent cations. As distinct from the n- and p-type contributions, for which a band-like mechanism is assumed, the extrinsic contribution to electronic conduction is assumed to proceed via a small-polaron mechanism, involving the activated hopping of electrons between adjacent dopant cations of different valence charge. As the small polaron mobility includes the fraction of sites not already occupied by electrons [24], the extrinsic electronic conductivity Oh depends on both [Ce~r] and [Cezr'] and is given by (~h = F[Cezr'] [Ce~r] u~ e x p ( - E H / k T )
(10.47)
474
10 ~ DENSECERAMICMEMBRANESFOR OXYGENSEPARATION
(a)
23
.....
,//
---
Vo
. FM
-N-.-'-"-" =--" --a::~-"~--.----
.....
21 A |
E
0 ,,0
"10
19
0
17
-
i
/,
",,
9 I
..... I
I
I
-30
-20
-10
0
o ,
I 10
Iog(Po=/Pa)
(b) - ~ ~ "
.-;-,
E
0 6O
b 0
" "
~ ~
" - "
- , r ~ b
--
,
,,
,,,
,,
,
-4
-8
,.~~
~
-12 /./'/
P1/2 !.,
I,
',
-30
-20
-10
.
I
0
Iog(Po=/Pa) Fig. 10.7. (a) Defect and (b) conductivity diagram for ceria-doped YSZ at 1000~ The relevant parameters to construct the diagrams are given in Ref. [119]. The theoretical dependence of ionic transference number tionand oxygen permeability jo~ are given in (c). Dashed lines in (c) refer to YSZ. FM' (Yzr') represents the aliovalent dopant used. Reproduced (slightly adapted) from Marques et al. [119].
10
-
-
475
DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
(c)
,,
,,=
,
i
,,,,
,,,i
,,
,
,,
,,
,
,,
,,,i,,
II
0.25
";',
I
o
[I
I III II
0.8
0.20 I I I I
|
E
I
0.15
0.6
i,.0
0
E
0.10
0 -., 9
0.05
0.4
-
0.2
I
i
J i ....
-20
-10
,,
|
,
.
0
Iog(Po2/Pa) Fig. 10.7c. C a p t i o n o p p o s i t e .
where u ~ is the pre-exponential of the mobility and EH the hopping energy. Accordingly, r~h displays a maximum at a critical oxygen pressure, P~/~, characterized by equal concentrations of [Ce~.r] and [Cezr']. On assuming that the oxygen vacancy concentration remains fixed by the aliovalent dopant concentration, one may derive from Eqs. (10.42)-(10.45) that the Po2-dependence of the small polaron conductivity at given temperature takes the form, K ~ D1/4 J-O 2
D1/4 + 1)2 u~ exp(-EH/kT) (10.48) (K, ~o~ where K~ -" K~[ C e z r ] t o2t a l and K~ = 21/2Ke/(KceKg1/4[Yzr ,]1/2). The partial ionic and electronic conductivities may be substituted into the Wagner equation (Eq. (10.10)) to derive an expression for the oxygen flux. Typical results of such calculations are given in Fig. 10.7c, in which the oxygen flux at given temperature is plotted against Po~", assuming air to be present at the feed side of the membrane. It can be seen that the oxygen flux saturates upon lowering Po2". An inflection point occurs at Po2" = P1/2. At the lowest values of Po2", the curve bends upwards again due to the onset of the intrinsic electronic conduction. For extended discussion we refer the reader to the original papers [118,119]. c~h - F
10.5.3 Examples (a) Ceria-doped ZrO2-Y203 Electrical properties of solid solutions ZrO2-CeO2-Y203 have been investi-
476
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
gated thoroughly by Cal6s and Baumard [31,32]. The main features have been confirmed by others [120,121]. The amount of 10 mol% yttria used by Cal6s and Baumard ensured a minimum concentration of oxygen vacancies in a wide range of experimental conditions. For all ceria dopant levels and temperatures (1000-1400~ ionic conduction is found to predominate at high Po2 values. Doping with ceria decreases the ionic conductivity up to (ZrO2)o.45--(CeO2)o.45(Y203)0.1, beyond which it increases again up to the composition as high as (CeO2)0.9--(Y203)0.1.
For not too low ceria contents the total electrical conductivity displays a maximum at reduced P02 values, shifting to lower P02 values as the temperature is decreased. It thereby follows the predictions of the preceding section, which is generally taken as evidence that the electrons in the ceria-based solid solutions move by a hopping mechanism. The maximum can be correlated with the presence of nearly equal concentrations of Ce 4+ and Ce 3+ if one takes into account the concomitant change in ionic conductivity with decreasing P02" Cal6s and Baumard deduced that, for (ZrO2)o.8r(CeO2)o.09-(Y203)0.1 at an oxygen pressure of about 10-13-10-14atm and temperature 1200~ (~elis of the same order of magnitude as Oionand approximates 0.06 S cm d. The contribution of the electronic to the total conductivity, at a given P02 and temperature, increases with increasing ceria content, albeit at the expense of the ionic conductivity (up to the composition (ZrO2)0.45--(CeO2)0.45--(Y203)0.1). For the most reducing conditions the conductivity becomes predominantly ionic again, albeit that the corresponding value is significantly less than that observed at high Po~. This may cause surprise knowing that the major fraction of the cerium ions under the reduced conditions adopts the trivalent state and, hence, the concentration of oxygen vacancies will be enhanced. The reduced ionic conductivity at low Po2 is attributed to enhanced defect interactions and/or lattice distortions. This type of behaviour is reminiscent to that of zirconia and ceria based electrolytes, for which it is observed that the ionic conductivity increases with the extent of aliovalent doping up to a certain limit beyond which defect ordering or formation of defect associates lowers the ionic conductivity [73]. In solid solutions with a high ceria content, for example, (ZrO2)0.45--(CeO2)0.45-(Y203)0.1, the ionic conductivity at 1100~ decreases rapidly at an oxygen partial pressure below about 10-11 atm due to the formation of ordered pyrochlore-type domains, as confirmed by XRD measurements. The more recent work on 10 mol% ceria-doped YSZ (5.8 mol% yttria) by Ramanarayanan et al. [122] showed that the ionic transference number tion decreases with reduction in grain size. This observation suggests that the preferred path for electronic conduction is via the grain boundary. TEM imaging confirmed a strong tendency for cerium cations to segregate to the grain boundary, showing enrichments up to about 20 mol% compared with the value of 11 mol% observed in the lattice.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
477
To the best of our knowledge literature reports from oxygen permeation measurements on solid solutions ZrO2--CeO2-Y203 are not available. Recent data from measurements on the related system ZrO2--CeO2-CaO are included in Table 10.1.
(b) Titania-doped ZrO2--Y203 Pure titania has the rutile structure and therefore has limited solubility in YSZ. The observed linear decrease in lattice parameter with increasing titania concentration in these solid solutions suggests that titanium cations enter the lattice substitutionally for zirconium. Concordant with the data from XRD measurements [29,30,123] the cubic fluorite structure is retained upon addition of 12-20 mol% titania, above which a second phase appears, claimed to be ZrTiO4 [123]. The spread in data of the solubility limit produced by different authors may be due to slight differences in, e.g., yttria concentration, sample processing, sintering temperature and impurity content in the cited studies. Microstructural investigations based on SEM and TEM indicated that precipitates of the second phase actually may appear already at lower titania contents [123,125]. Contrary to the earlier observations [29,30], recent studies on electrical conductivity by Marques et al. [124] and Lindegaard et al. [125] indicate that the lattice ionic conductivity decreases with the extent of incorporation of titania into YSZ. Results confirm that the ionic conductivity of 10 mol% titania-doped YSZ in air, at a typical temperature of 1000~ is about ten times less than that of undoped YSZ. The hopping electronic conductivity at this temperature is estimated to be ---10-7 S cm -1 [121]. For similar dopant levels addition of titania appears to be more effective in enhancing the electronic conductivity of YSZ than ceria, which is not expected considering the redox behaviour of pure ceria and titania. Using thermogravimetric measurements on YSZ with ceria and titania additions up to 10 mol%, Marques et al. [124] confirmed that Ce 4+ cations in these solid solutions are more easily reduced than Ti4+. The observed increase in grain boundary conductivity with increasing titania concentration [29,30, 126] and decreasing Po~ [125] would indicate that electronic conductivity occurs at the grain boundaries. Liou and Worell [29,30] presumed segregation of Tizr to the grain boundary region, but within experimental uncertainty of EDS (Energy Dispersive Spectroscopy of X-rays) no evidence was found for titania-rich grain boundaries in the already cited study by Marques et al. [124]. The higher electronic conductivity of titania-doped YSZ was interpreted to reflect the formation of highly mobile electronic defects (large polarons) in the bulk, by comparison with the low mobility of small polarons formed in ceria-doped specimens. At 1000~ significant levels of electronic conduction in titania-doped YSZ, as in ceria-doped specimens, are found only under strongly reducing atmospheres. Data of oxygen permeability have been presented for the ZrO2-Y203TiO2 system by Arashi and Naito [127] (see also Table 10.1). By virtue of its
478
10 m DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
stability up to 2000 ~ it is proposed to be used as a m e m b r a n e for direct hydrolysis of water to produce hydrogen [128].
(c) Miscellaneous materials Terbia has been dissolved in pure ZrO2 to form mixed-conducting solid solutions with Tb203. 5 concentrations as high as 50 mol% by Iwahara et al. [129]. Briefly, the electrons in these mixed oxides move by hopping between Tb 3+ and Tb 4+ ions, the coexistence of which ions has been confirmed using the XANES (X-ray absorption near edge structure) technique [130]. The relative contribution of the electronic to the total conductivity measured by Iwahara et al. increases with increasing terbia concentration, in spite of the fact that the relationship between the latter quantity and (Itotalr at a given temperature and oxygen pressure, turns out to be very complex. At 900~ and atmospheric pressure, (~totalfor (ZrO2)0.7-(Tb203.5)0.3 is 1.8 x 10 -2 S cm -2, and tio n is 0.30. Oxygen from permeation, at 900~ was found at the argon side of a 2-3 m m thick disc m e m b r a n e of this composition, at a rate of about 5 x 10 -9 mol c m -2 s -1, the value of which was measured with ambient air maintained at the feed side. W h e n terbia is dissolved in YSZ, this ensures a m i n i m u m value for the oxygen vacancy concentration, which is then fixed by that of yttria. Cao et al. [131,132] examined the electrical conductivity and oxygen permeation of selected compositions, including (ZrO2)07-(Tb2035)03-y--(Y2OB)y with y = 0, 0.025, 0.05 and y = 0.072. At 900~ (~totaldecreases f r o m l . 2 x 1 0 - 2 S c m -2 for y = 0 to 0.86 • 10-2 S cm -2 for y = 0.072, where tioniS 0.37 and 1, respectively. The oxygen flux, at 900~ passing from the air to the helium side of 2 m m thick disc-shaped membranes varied in the range 2.6-3.7 x 10-11 mol cm -2 s-1. Hardly any effect of the yttria-content on oxygen fluxes was measured. Based upon these results, amongst some additional experimental facts, e.g., the Po2-dependence of the oxygen flux, it is concluded that the surface exchange reaction is the rate limiting step for oxygen permeation. Regrettably, no account is given as to w h y the data of oxygen permeation are almost two orders of magnitude lower than the one claimed for y = 0 by Iwahara et al. [129]. Dense thin films of several microns could be grown successfully on different porous ceramic substrates by electrochemical vapour deposition (ECVD) [133,134]. An oxygen permeation flux of 7 x 10-1~mol c m -2 s -1 at 953~ was measured for a film (ZrO2)0.86--(Tb2OB.5)0.10--(Y203)0.04 of thickness 8 ~tm deposited on a coarse a-alumina substrate, which value increased to 3 x 10-8 mol c m -2 s -1 if the helium line was switched to CO/CO2 having Po2-5 x 10-16 atm. In these experiments, air was supplied to feed side of the membrane. Finally, we briefly describe the observations recently m a d e in the present authors' laboratory in an attempt to increase the oxygen permeation flux through stabilized bismuth oxide by substitution of the 8-Bi203 host with 40 mol% terbium on the bismuth sites (BT40). Measurements using the concentration cell method and ac impedance confirmed that BT40 exhibits good p-type
10 n DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
479
conductivity and is an excellent mixed conductor with ionic transference numbers, tio n = 0.74 at 650~ and tion = 0.85 at 800~ in air [135]. Using ambient air as the source of oxygen and helium as the sweep gas on the other side of dense BT40 disc membranes, in the range of thickness 0.07-0.17 cm and temperature 600-800~ did not yield the expected increase in the oxygen flux, over BE25 [136]. Isotopic exchange measurements in the relevant range of oxygen partial pressure and temperature showed that both oxides exhibit an almost equal activity in oxygen exchange [104], which is in support of the conclusion made from oxygen permeation measurements that the oxygen fluxes through BT40, at the conditions covered by the experiments, are limited by the surface exchange kinetics. Additional attempts have been presented to render hosts with the fluorite and the related pyrochlore structure electronically conductive by doping with mixed-valence a n d / o r shallow dopants. The list of dopant materials examined includes oxides of elements of, for example, Ti, Cr, Mn, Fe, Zn, Fe, Sn, Ce, Pr, Gd, Tb and U. In general, however, the extent of mixed conductivity that can be obtained in fluorite-type ceramics is rather limited, by comparison with the corresponding values found in some of the perovskite and perovskite-related oxides considered in the next section. 10.6 A C C C E P T O R - D O P E D PEROVSKITE A N D PEROVSKITE-RELATED OXIDES
10.6.1 Introduction
The general trend observed from the pioneering studies on oxygen permeation through perovskites of the type L n l _ x a x C O l _ y B y O 3 _ a (Ln = La, Pr, Nd, Sm, Gd; A = Sr, Ca, Ba; B= Mn, Cr, Fe, Co, Ni, Cu) by Teraoka et al. [37-39] is that higher oxygen fluxes are facilitated by increased A-site substitution, and a lower thermodynamic stability of the particular perovskite. Clearly, not all these perovskite compositions are useful for oxygen delivery applications. For example, ceramics based on Lal_xAxCrOB-a (x = Sr, Ba, Ca), Cal_xSrxCrl_yMnyO3_a and Cal_xCaxCrl_yCoyO3_a have been proposed for use as interconnection material (separator) in solid oxide fuel cells (SOFC), and therefore should be dense and impermeable in order to prevent burning off of the fuel without generating electricity [137,138]. Selected perovskite compositions are also targeted in basic SOFC research for use as potential electrode material for the cathodic reduction of oxygen. The most promising cathode materials to date are the manganites Lal_xSrxMnO34 [137,138]. The composition with x = 0.15 scarcely permeates oxygen up to 900~ as was measured by feeding air and helium to opposite sides of a dense sintered membrane of I mm thickness [136]. The observed behaviour is consis-
480
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
tent with the low value of the oxygen self-diffusivity in La0.sSr0.5MnO3~, determined by 180-160 isotopic exchange, and can be attributed to the small negative departure from oxygen stoichiometry exhibited in the range of temperature and oxygen pressure covered by experiment [139]. On the other hand, oxygen transport is actually predicted to be quite fast under conditions of high oxygendeficiency, i.e. low oxygen partial pressures, as the oxygen vacancy diffusion coefficient of La1_xSrxMnO34 was found to be comparable in magnitude with that of Fe- and Co-based perovskites [140]. Emerging from the first of these studies by Teraoka et al. [37] is that in the series La1_xSrxCo1_yFeyO34the oxygen fluxes increase with Co and Sr content, the highest flux being found for SrCo0.sFe0.2034. Data were obtained with air on one side of a 1 mm thick disc specimen, using helium as sweeping gas on the other side, up to a maximum temperature of 1150 K. The observed oxygen fluxes were found to be roughly proportional to the ionic conductivity of the perovskites, which is in agreement with the fact that the electronic conductivity of compositions in this series can be extremely high, typically in the range 102103 S cm -1 [40]. Four-probe dc measurements using electron blocking electrodes showed that the ionic conductivity at 800~ in air can be 1-2 orders of magnitude higher than that of stabilized zirconia [40]. These findings have been confirmed by others, apart from scatter in the published data, which partly reflects the experimental difficulties in measuring the ionic conductivity in these predominantly electronic conductors [141-144]. In a subsequent study, Teraoka et al. [38] investigated the influence of A and B site substitution on oxygen permeation through La0.6A0.4Co0.sFe0.2034(A = La, Na, Ca, Sr, Ba) and La0.6Sr0.4Co0.8B0.203,s(B = Cr, Mn, Fe, Co, Ni, Cu). As seen from Figs. 10.8 and 10.9, the oxygen permeability in the two series increases in the respective orders La < Na < Sr < Ca < Ba and Mn < Cr < Fe < Co < Ni < Cu, which differ from trends in the periodical system, as far as comparison is meaningful. Results from ionic and electronic conductivity measurements of La0.6A0.4Co0.8Fe0.203_8 (A = La, Ca, Sr) and La0.6Sr0.4Co0.8B0.203.8(B = Fe, Co, Ni, Cu) suggest that oxygen permeation is governed by the ionic conductivity [39]. In the homologous series Ln0.6Sr0.4CoO3_a,the oxygen flux was found to increase in the order La 3+ < Pr 3+ < Nd 3+ < Sm 3+ < Gd 3+ which corresponds with a decrease in radius of the lanthanide-ion [38]. Since the initial observations by Teraoka et al., a considerable number of studies have appeared. Selected perovskite compositions have been re-examined, while a few others have been adapted in an attempt to optimize the oxygen fluxes. The list of materials for which oxygen permeation data are presently available has been extended to include, LaCoO34 [145], La1_xSrxCoO34 [146-149], La1_xSrxFeO34 [150,151], Lal_xAxCol_yFeyO34 (A = Sr, Ca) [12,139,144, 152], SrCo0.8Fe0.2034 [13,41,153,154], SrCoo.sBo.203_8 (B = Cr, Co, Cu) [154], SrCol_• (B = Cr, Mn, Fe, Ni, Cu, x = 0...0.5) [155], and Y1_xBaxCoO34 [156]. In general, fair agreement
10 -- DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
481
1.50
9
Ba
9
Ca
9
Na
9
La
Sr A ,-;.,
1.00
E o
o
E ,,, O
0.50
0.00 300
500
700
900
temperature (~ Fig. 10.8. Temperature variation of the oxygen permeation rate from the air to the helium (30 cm 3 min -1) side of disc membranes La0.6A0.4Co0.8Fe0.2034(A = Na, Ba, Ca, Sr), 20 m m in diameter and 1.5 m m thick, after Teraoka et al. [38]. Reproduced (data re-scaled) from Teraoka et al. [38].
is obtained with data produced by Teraoka et al., albeit that in a number of studies the observed oxygen fluxes are reportedly found to be significantly lower [12,153,154]. The pioneering studies by Teraoka et al. [37-39] have opened a very challenging research area as the perovskites, e.g. Lal_~SrxCo1_yFeyO34, have a bright future for use as oxygen separation membrane. The precise composition may be tailored for a specific application, but this has not yet been fully developed. One of the important issues is considered to be the low structural and chemical stability of the perovskites, especially in reducing environments, which remains to be solved before industrial applications become feasible. In order to meet this challenge, it is necessary first to understand the factors that limit and control the quality criteria for any given application. The perovskite and related oxides exhibit a great diversity of properties, like electrical, optical, magnetic, catalytic properties, which have been studied extensively. In the following sections, we mainly focus on those properties affecting the magnitude of the oxygen fluxes through these materials.
482
10 n DENSECERAMICMEMBRANESFOR OXYGENSEPARATION 1.50
~"
1.00
+
9
Cu
9
Ni
9
9
Co
9 9
Fe Cr Mn
.'~~ 0.50
0.00
500
~
~
,
600
I
700
temperature
,
I
800
,
900
( ~ C)
Fig. 10.9. Temperature variation of the oxygen permeation rate of La0.6Sr0.4Co0.8B0.203-8(B = Cr, Mn, Fe, Co, Ni, Cu) after Teraoka et al. Experimental conditions are specified in the legend of Fig. 10.8. Reproduced (data re-scaled) from Teraoka et al. [38].
10.6.2 Structure and Defect Chemistry 10.6.2.1 Perovskite Structure The ideal perovskite structure ABO3 consists of a cubic array of corner-sharing BO 6 octahedra, where B is a transition metal cation (Fig. 10.10). The A-site ion, interstitial between the BO 6 octahedra, may be occupied by either an alkali, an alkaline earth or a rare earth ion. In m a n y cases the BO 6 octahedra are distorted, or tilted, due to the presence of the A cation, which is generally larger in size than the B cation. Alternatively, the perovskite structure m a y be regarded as a cubic close-packing of layers AO 3 with B cations placed in the interlayer octahedral interstices [157]. The latter turns out to be more useful in distinguishing different structural arrangements (stacking sequences) of perovskite blocks. The tolerance limits of the cationic radii in the A and B sites are defined by the Goldschmidt factor, which is based on geometric considerations: t = (ra + ro) / (~-(rB + ro)), where rA, rB and ro are the radii of the respective ions [158]. When the distortion becomes too large, other crystal symmetries such as
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
483
~ A @0 QB ::::iiii:i::..
Fig. 10.10.Ideal perovskite structure. orthorhombic and rhombohedral appear. Nominally, the perovskite structure should be stable between 1.0 < t < 0.75. The ideal perovskite lattice exists only for tolerance factors t very close to one. Clearly, it is the stability of the perovskite structure that allows for large departures from ideal stoichiometry, resulting either from the substitution with aliovalent cations on the A or B-site or from redox processes associated with the presence of transition metal atoms which can adopt different formal oxidation states. Oxygen vacancies are free to move among energetically equivalent crystallographic sites as long as the perovskite structure exhibits ideal cubic symmetry. The degeneracy between sites disappears upondistortion of the lattice towards lower symmetries. The onset of electronic conductivity mainly depends on the nature of the B-site cation. The total electrical conductivity can be either predominantly ionic as in the acceptor-doped rare earth aluminates or predominantly electronic as in the late transition metal containing perovskites considered below.
10.6.2.2 Nonstoichiometry Important contributions to the area of defect chemistry of the acceptor-doped Lnl_xAxBO3, perovskites, where B is selected from Cr, Mn, Fe or Co, have been made by a number of investigators. Particular reference is made to reviews provided by Anderson [159,160] and Mizusaki [161]. The substitution of divalent alkaline-earth ions on the A-site increases the concentration of oxygen vacancies. Temperature and oxygen partial pressure determine whether charge compensation occurs by an increased valency of the transition metal ion at the B-site or by the formation of ionized oxygen vacancies. Thermogravimetric studies have indicated that in, for example, LaCrO3, YCrO3 and LaMnO3 the
484
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
native nonstoichiometric ionic defects are cation vacancies, leading to oxygenexcess stoichiometries [160]. For simplicity, it is assumed here that extrinsic ionic defects generated by A-site substitution prevail, i.e. only oxygen-deficient stoichiometries are considered. Furthermore, crystallographic sites available for oxygen are taken to be energetically equivalent. For the purpose of our discussion, LaFeO3 is considered to be the host for substitution. The dissolution of SrFeO 3 into this material can be represented by, SrFeO3
LaFeO 3 )
SrLa' + FeFe + 30~)
(10.49)
The incorporation of Sr 2§ thus leads to charge compensation by the formation of Fe 4+ ions, which is in accord with the Verwey principle of controlled ionic valency [162]. The extent of oxygen non-stoichiometry is established by the following defect chemical reactions, 2FeFe + O~) ~ 2Fete ~
2Fete + V'o + 1/2 0 2
FeFe' + FeFe
(10.50) (10.51)
with the corresponding equilibrium constants, [Fete] 2 [V'o] Po1/2 2 Kg = [FeFe]2 [O~)]
Kd =
[FeFe'] [FeFe] [Fete] 2
(10.52)
(10.53)
The oxygen vacancies formed at elevated temperatures and low oxygen partial pressure are assumed to be doubly ionized. The thermally activated charge disproportionation reaction given by Eq. (10.51) reflects the localized nature of electronic species and may be treated as equivalent to the genera.tion of electrons and electron holes by ionization across a pseudo band gap (cf. Eq. (10.27)). The associated free enthalpy of reaction may be taken equal to the effective band gap energy. At fixed A / B site ratio the following condition must be fulfilled [FeFe'] + [Fete] + [FeFe] = 1
(10.54)
and the condition of charge neutrality is, [SrLa'] + [FeFe'] = 211/'O] + [FeFe]
(10.55)
In the absence of extended defects, i.e. no interaction between point defects, Eqs. (10.52)-(10.55) may be used with the aid of experimentally determined equilibrium constants to construct the Kr6ger-Vink defect diagram, from which ex-
10 - - DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
485
pressions for the partial conductivities of the mobile ionic and electronic defects can be derived. Oxygen nonstoichiometry of the perovskites Lal_xSrxBO3.s(B = Cr, Mn, Co, Fe) and its relationship with electrical properties and oxygen diffusion has been studied extensively [159-161]. Typical nonstoichiometry data for La1_xSrxFeO34 and for some other perovskites as obtained from gravimetric analysis and coulometric titration are given in Fig. 10.11. At small oxygen deficiency, acceptor dopants are the majority defects. The charge neutrality condition then becomes, [SrLa'] = [FeFe]
(10.56)
In this region, one finds for the oxygen non-stoichiometry 5, oc p o~/2 2
(10.57)
noting that 8 = [V6], by definition. A plateau is observed around the point of electronic stoichiometry, 8 = x/2, where the charge neutrality condition reads, [Srca'] = 2[V6]
(10.58)
3.05
& &
A
.
.
.
.
.
-
/
2.95 o3
2.9
Lao.9Sro.lCo03.6 9 Lao.oSro.lFeO3.s
0
o Lao.3Sro.7Cr03-8 zx Lao.2Sro.sMn03-8
2.85
.8
lu
-20
~
I
I
I
-16
I
I
,
I,
I
I
-12
I,
.
I
I
-8
I
I
I
I
-4
I
I
'
0
Iog(P 0 z/atm) Fig. 10.11. Data of oxygen nonstoichiometry of Lao.75Sro.25CrO3-~, Lao.gSro.lFeO3-~, Lao.gSro.lCoO3-~, and Lao.sSro.2MnO3-~ at 1000~ as a function of oxygen partial pressure. Solid lines are results from a fit of the random point defect model to the experimental data. Reproduced (slightly adapted) from Van Hassel et al. [185].
486
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
corresponding with a minimum in the electronic conductivity of La1_xSrxFeO34 [163,164]. In this region, the oxygen non-stoichiometry is virtually constant. As the oxygen activity decreases further, oxygen vacancies are again generated, down to the oxygen activity at which decomposition of the perovskite structure occurs. The onset of the different regions depends on the nature of the transition metal B-cation. The incentive of B-site substitution can therefore be to optimize oxygen transport in appropriate ranges of oxygen partial pressure and temperature. As discussed below, doping may also increase stability or suppress cooperative ordering of oxygen vacancies. 10.6.2.3 Localized versus Delocalized Electrons Given the relative success of the above point defect scheme to model the experimental data of oxygen nonstoichiometry and electrical conductivity for Lal_vSrxFeO34 [165,166] and Lal_xSrxCrO3.s [167], its use is less satisfactory for Lal_xSrxCoO34 and Lal_xSrxMnO3~, which compounds show notably high values for the electronic conductivity. Nonstoichiometry of the compounds La1_xSrxCoO34 (x = 0, 0.1, 0.2, 0.3, 0.5 and 0.7) in the range 10-5 < Po2 < 1 atm and 300< T < 1000~ was investigated by Mizusaki et al. [168] using thermogravimetric methods. At 800~ 5 in Lal_xSrxCoO3_s varies almost proportional to Po2n with n = - 1 / 2 for x = 0 to n = - 1 / 1 6 for x = 0.7 (see Fig. 10.12). No plateau is observed around 8 = x/2. Fitting the 5-Po2 relationship in accord with the random point defect model leads to very large concentrations of disproportionation reaction products Coco" and Coco. A corollary is that the pseudo bandgap must be very small. The model fit, however, is less satisfactory for high Sr substitutions [169]. A similar e x p l a n a t i o n holds for La1_xSrxMnO34, d i s r e g a r d i n g the oxygen-excess stoichiometries seen in this system at high oxygen partial pressures. At high oxygen deficiency of the perovskite, the validity of the ideal mass action equations (based upon dilute solution thermodynamics) cannot be assumed a priori. In addition, interaction and association between defects are expected at high defect concentrations. A further limitation concerns the nature of electronic defects. The general assumption, that in the first row transition metal perovskites changes in the oxygen content leads to changes in the 3d electronic configuration, may be too naive. It is based implicitly on the idea that oxygen is strongly electronegative and, by comparison, the 3d electrons can be easily ionized. There is substantial evidence from soft-X-ray absorption spectroscopy (XAS) based studies that the electron holes introduced by doping with divalent earth-alkaline ions go to states with significant O 2p character [170]. This has also been reported for the perovskite-related oxide YBa2Cu306§ [171]. In a localized description, i.e. assuming a narrow bandwidth of the hole band derived from the O 2p band, this would imply that 0 2- is effectively converted into O-.
10 -- DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
I
i''
I
-"
487
~
' I
"'
-i
....
-1
.3
-2
a
0
0 ,-4
m
-3
m
....
i
I
_
1
-4
1
log
1,,
-2
(P
02
I
/atm)
l
0
Fig. 10.12. O x y g e n pressure dependence of 6 in Lal-xSrxCoO3-6 for different strontium contents at 800~ Reprinted from Mizusaki et al. [168].
A proper description of electronic defects in terms of simple point defect chemistry is even more complicated as the d electrons of the transition metals and their compounds are intermediate between localized and delocalized behaviour. Recent analysis of the redox thermodynamics of La0.sSr0.2CoO34based upon data from coulometric titration measurements supports itinerant behaviour of the electronic charge carriers in this compound [172]. The analysis was based on the partial molar enthalpy and entropy of the oxygen incorporation reaction, which can be evaluated from changes in emf with temperature at different oxygen (non-)stoichiometries. The experimental value of the partial molar entropy (free formation entropy) of oxygen incorporation, Aso2, could be
488
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
fitted by assuming a statistical distribution among sites on the oxygen sublatfice, Aso2 = s~ - 2k In
(3 - 8 ) 8
(10.59)
where s o is a constant. That is, no entropy change associated with electron annihilation can be identified. The partial molar enthalpy (free enthalpy of formation of vacancies) associated with oxygen incorporation was found to decrease almost linearly with 8. A first inclination might be to assume that the mutual repulsion between oxygen vacancies increases with increasing oxygen deficiency. But this interpretation immediately raises the question why such a behaviour is not found in the case of La1_xSrxFeO34 [166,167]. Instead, the experimental data are interpreted to reflect the energetic costs of band filling. With increasing oxygen nonstoichiometry in La0.sSr0.2CoO34 the two electrons, which are needed for charge compensation of a single oxygen vacancy, are donated to an electron band broad enough to induce Fermi condensation characteristic of a metallic compound. The average density of electron states at the Fermi-level is determined to be 1.9_+0.1 eV-1 per unit cell. The physical significance of the work is that the defect chemistry of La0.8Sr0.2CoO3-4 cannot be modeled using simple mass action type of equations. An empirical model for the oxygen non-stoichiometry of La0.sSr0.2CoO3-8 is proposed, which demonstrates that the density of states is related to the slope of the log-log plots of 8 versus Po~. In support of these interpretations, it is noted that XAS has not been successful in detecting charge disproportionation in LaCoO34, due to localization of electrons, in the temperature range 80-630 K [173]. The nonstoichiometry data obtained for La0.8Sr0.2CoO3-4are found to be in good agreement with earlier results from gravimetric analysis in the series La1.~rxCoO34 obtained by Mizusaki et al. [168], which authors arrived at more or less similar conclusions regarding the role of electronic states in the energetics of oxygen incorporation into these compounds.
10.6.3 Oxygen Desorption and Perovskite Stability As seen from Fig. 10.11, the value of (3-8) in Lal_xSrxCoO34 falls off with decreasing oxygen activity much more rapidly than for the other compounds shown. The general trend at which the perovskites become nonstoichiometric follows that of the relative redox stability of the late transition metal ions occupying the B-site, i.e. C r 3+ > Fe 3+ > Mn 3+ > C o 3+. The reductive nonstoichiometry of the cobaltites increases further by partial B-site substitution with copper and nickel. The reductive (and oxidative) nonstoichiometry and the stability in reducing oxygen atmospheres of perovskite-type oxides was reviewed by Tejuca et al. [174]. Data from temperature programmed reduction (TPR) measurements indicate that
10 E D E N S E C E R A M I C M E M B R A N E S FOR OXYGEN S E P A R A T I O N
489
the stability (or reducibility) of the perovskite oxides increases (decreases) with increasing size of the A ion, which would be consistent with the preferred occupancy of the larger Ln 3+ ion in a 12-fold coordination. The trend is just the reverse of that of the stability of the corresponding binary oxides. The ease of reduction increases by partial substitution of the A ion, e.g., La 3+ by Sr 2+. Trends in the thermodynamic stabilities of perovskite oxides have been systematized in terms of the stabilization energy from their constituent binary oxides and the valence stability of the transition metal ions by Yokokawa et al. [175]. The stability of the undoped perovskites LaBO3_a, at 1000~ expressed in terms of Po2 decreases in the order LaCrO34 (10 -20 atm) > LaFeO34 (10 -17 atm) > LaMnO3_~ (10-15 atm) > LaCoO34 (10 -7 atm), noting that the cited value for LaCrO34 corresponds with the lowest limit in a thermogravimetric study by Nakamura et al. [176]. The same trend was found by means of TPR [174]. Tabata et al. [177] and Seyama [178] both described significant differences in the chemical composition of the surface, due to Sr segregation, compared with the bulk composition in a series of powders Lal_~SrxCoO3_~. This indicates a behaviour of the surface different from that of the bulk in these compounds. Not only can this account for a number of observations made in the total oxidation of CO and C H 4, as discussed by the authors, but it is also considered to be an important factor when one tries to correlate the composition of a perovskite with its activity in surface oxygen exchange. The sorpfion kinetics of oxides is certainly influenced by their corresponding defect structure. A number of interesting observations were made by Yamazoe and co-workers [179,180], showing that for perovskites LaMO3_~ (M = Cr, Mn, Fe, Co, Ni), Lal_vqrxCoO3_~(x = 0, 0.2, 0.4 and 1) andLa0.8Ao.2CoO3_8 (A = Na, Ca, Sr and Ba), two distinct types of oxygen are desorbed upon heating in a helium stream after a pre-treatment step in which the oxide was saturated in an oxygen-rich atmosphere at high temperature, followed by slow cooling to room temperature. The oxygen desorbed in a wide range at moderate temperatures, referred to as c~-oxygen, was found to be correlated with the amount of partial substitution of the A ion. The onset temperature of the so-called ~-desorpfion peak observed at high temperature was correlated with the thermal decomposition temperature of the corresponding transition metal oxides. Accordingly, the ~-peak corresponds with the reduction of the transition metal ion from B 3+ to B2+. The partial substitution of Co by Fe in the series Lal_xSrxCOl_yFeyO3_~ stabilizes the Co B+oxidation state (no -peak observed), while shifting the o~-type of desorpfion to lower temperatures [181,182].
10.6.4 Equationsfor Oxygen Transport Equations for oxygen transport can be derived from the point defect equilibria discussed in Section 10.6.2.2. This provides us with some general insight
490
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
into the transport behaviour of oxygen-deficient perovskites. Strictly speaking, the equations presented below are valid at low defect concentrations only, i.e. assuming oxygen defects to be randomly distributed. Oxygen transport in the perovskites is generally considered to occur via a vacancy transport mechanism. On the assumption that the oxygen vacancies are fully ionized and all contribute to transport, i.e., oxygen defects are not associated, the Nernst-Einstein equation reads, Gi~
4 F 2 [ V o ] Dv ----- RTVm
(10.60)
where Dv is the vacancy diffusion coefficient and Vm is the perovskite molar volume. Since electronic conduction in the perovskites predominates, i.e. Gel > Glow,the integral in the Wagner equation (Eq. (10.10)) involves only Gionover the applied oxygen partial pressure gradient. Using Eq. (10.60), we may rewrite the Wagner equation, to give In P"o2 jo 2 =
Dv 4VmL ~ ~)d In Po2
(10.61)
In P'o2
by virtue of 6 = [Vo]. Evaluation can be performed numerically provided that Dv and the 5-1n(Po2) relationship are known. The ability of Eq. (10.61) to quantitively fit experimental data of oxygen permeation is illustrated for La0.9Sr0.1FeO3~ in Fig. 10.13. Similar results have been presented for, e.g., La0.75Sr0.25CrO3~ [183] and La0.70Ca0.30CrO3~ [184]. The analytical solution of the integral given by Eq. (10.61) incorporating random point defect chemistry has been given by Van Hassel et al. [185]. When data of oxygen nonstoichiometry follows a simple power law 8 ~ P"02 , integration of Eq. (10.61) yields an expression similar to that of Eq. (10.18) having ~ = DvS~ n. Examination of the data from oxygen permeability measurements on disc specimens of thickness 2 mm in a series Lal_xSrxCoO3_8 (0_ x _< 0.8) in a study by Van Doom et al. [148] indicate that the results, at 1000~ can be fitted well by this equation, the validity of which is usually restricted to a small range in oxygen partial pressure. For compositions x _<0.6, the values of n obtained from fitting were found to be in excellent agreement with the corresponding slopes of the In 5--ln Po2 plots derived from data of thermogravimetry in the range 10-4_
10 m DENSE CERAMIC MEMBRANESFOR OXYGENSEPARATION
491
0.3 Ishigaid et al.
E o
E
E
O4
O "~
Dv~ =6.0.1 -e
-2.5
-2
-1.5 log
-1
-0.5
0
(po2/bar)
Fig. 10.13. T h e o r e t i c a l fit of feed-side Po2-dependence of o x y g e n p e r m e a t i o n t h r o u g h La0.9Sr0.1FeOa~, at 1000~ The best fit is obtahled when Dv equals 6 x 10.4 cm 2 s-1, which slightly deviates from the corresponding value obtahled from isotopic exchange. Reprinted from Ten Elshof et al. [1511.
ing oxygen partial pressure), to an extent depending on the particular solid. In accord with classical diffusion theory, the probability of vacancy hopping, hence, Dv is proportional to a factor (1-5/3), which represents the site occupancy of lattice oxygen anions [186]. Moreover, complications due to local stresses resulting from a change in cell volume with decreasing oxygen partial pressure may need further consideration, especially if the nonstoichiometry becomes relatively large. With the help of Eq. (10.60) the vacancy diffusion coefficient Dv can be determined from ionic conductivity measurements provided that data of oxygen nonstoichiometry are available. The direct measurement of (~ionin these materials requires the use of auxiliary electrolytes such as doped zirconia or ceria to block the electronic charge carriers. The ionic current that is passed through the sample is measured by the electric current in the external circuit. The problems faced are that of interfacial charge transfer between the blocking electrodes and the mixed conductor and that it is very difficult to effectively block all of the electronic current. Short-circuiting paths for oxygen transport can occur such as diffusion along the oxide surface or via the gas phase through rapid exchange, leading to overestimates of the ionic conductivity [143]. Also, there is the possibility of an interfacial reaction between the perovskite and the blocking electrode material or the glass used for sealing to suppress the parasitic contributions to oxygen transport [187]. The relation between Dv and the tracer diffusion coefficient D* can be expressed as 8 D* = f (3 L 8i Dv
(10.62)
492
10 m DENSE C E R A M I C M E M B R A N E S FOR O X Y G E N S E P A R A T I O N
where f is the correlation factor for diffusion of oxygen vacancies across the ideal perovskite anion sublattice: f = 0.69 (for small values of 8) [75]. Data of Dv thus obtained have been published by Ishigaki, Mizusaki and coworkers for single crystals of LaCoO34 [74,188] and LaFeO34 [189]. The value of D* at 950~ in both oxides is found to be proportional to P~2, with n =-0.34 + 0.04 and -V2 n = -0.58 + 0.15, respectively, where a Po2-dependence is expected in the high Po2-regions covered by experiment. These results provide firm evidence that diffusion in the perovskites occurs by a vacancy mechanism. Though the values of D* observed in these oxides differ about 3 orders of magnitude, the corresponding values of Dv are nearly the same. Additional data have been reported for single crystals of La0.gSr0.1CoO34 (x = 0.1) [75], La1_xSrxFeO34 (x = 0.1, 0.25 and 0.4) [75], and polycrystaIIine phases La1_xSrxFeO34 (x = 0, 0.4, 0.6 and 1.0) [190] and La0.70Ca0.30CrO3.8 [184], revealing that the diffusivities at elevated temperatures may be similar to those observed for fluorite and fluorite-related oxides, albeit that the associated activation energy generally tends to be slightly higher in the perovskite structure (see Ref. [161]). When electronic conduction predominates, one may derive the following relationship between Dv and the chernical diffusion coefficient D (by combining Eqs. (10.12) and (10.60)), N D~ O In Po~ D =--~-- 31n8
(10.63)
Measurements of the weight change following a sudden change of the oxygen activity in the gas phase were carried out for determining D in La1_~SrxCoO34 (x = 0 and x = 0.1) at 800-1000~ in the range 10-5 < Po2 < 1 atm [191]. The calculated values of D v agree with those obtained from the above studies, which suggests that the random point defect model holds well for the cobaltites at low Sr contents. Fair agreement is also obtained with results from relaxation experiments on La1_xSrxCoO34 (x = 0 and 0.2) in which the time change of the conductivity was traced [192]. A typical value of D taken from these studies is 10-s c m 2 s -1 at 900~ Relaxation experiments were also carried out to study chemical diffusion in, for example, SrCo1_yFeyO34 (y = 0.2, 0.5 and 0.8) [187,193], Lal_xSrxCOl_yFeyO3_ ~ (0.2 < x < 1.0 and 0 < y < 1.0) [194], Lal_xCax_ CrO3_ ~ (x = 0.1, 0.2 and 0.3) [195], Lal_~SrxMnO3,s (x = 0.05--0.20) [140] and Lal_ xSrxMnO34 (x = 0.20 and 0.5) [196,197].
10.6.5 Electronic Conductivity The late transition metal-containing perovskites exhibit high electronic conductivities. In the materials which receive prime interest for oxygen delivery applications, the electronic contribution at high temperature of operation is usually predominant. The values for e.g., La1_~SrxCol_yFeyO34 at 800~ in air
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
493
range between 102-103 S cm -1, whilst 10-2-1 S cm -1 is found for the ionic conductivity [40]. The ionic transference numbers in this series vary between 10-4-10-2. Departures of the above behaviour can occur at reduced oxygen partial pressures, due to the loss of p-type charge carriers [194,198]. This is most serious at the point where the hopping-type of conductivity goes through a minimum, to become n-type, e.g., in La1_xAxFeO34 (A = Sr, Ca) [163,164,199], provided that the correspondingly low oxygen pressure is maintained during experiment. At 1000~ the minimum in La0.75Sr0.25FeO3-soccurs below Po2 values of 10-12 atm [164]. In the following, we discuss a number of characteristics which are considered to control electronic conductivity in this class of oxides and give some examples of their behaviour. Electronic conduction in usual ranges of temperature and oxygen pressure is reported to be p-type, and is commonly explained by assuming a small polaron mechanism with a thermally activated mobility [159,160]. This behaviour may be masked by substantial oxygen loss, and concomitant decrease in the concentration of p-type charge carriers, seen at the highest temperatures and at reduced oxygen partial pressures. The direct overlap between transition-metal d orbitals is known to be small, being across a cube face. The hopping transport of mobile charge carriers between two neighbouring B-cations in the perovskite lattice is mediated by the O 2p orbital, B n+ - 0 2 - - B ( n - l ) + --4 B ( n - l ) + - O -
-
B ( n - l ) + --4 B ( n - l ) + - 0 2 - - B n+
which is known as double exchange, first discussed qualitatively by Zener [200]. This process is favoured by a strong overlap of empty or partly filled cation orbitals of the d-manifold (involving e~ and t2~ orbitals) with the filled 0 2p orbital of neighbouring anions, and reaches a maximum for a B-O-B angle of 180~ corresponding with ideal cubic symmetry. Lal_xAxCrO3_~ (Sr, Ca) and Lal_xAxFeO3_s (Sr, Ca) are typical examples of which the data of small-polaron transport can be explained in terms of simple defect chemistry, including the thermally activated charge disproportionation among the B-cations. The predominant mechanism of hopping in Lal_xAxFeO3_s (Sr, Ca) in the p-type region is between Fe4+ and Fe3+valence states, changing to that between Fe2+ and Fe3+ in the n-type region, upon lowering the oxygen partial pressure. The electrical conductivity thereby passes through a minimum at the point of electronic stoichiometry, where the concentrations of Fe4+ (FeFe') and Fe2+(FeEd)are equal [163,164,199]. Charge disproportionation has also been used to account for results from electrical conductivity and thermopower measurements of selected substitutionally mixed oxides LaMnl_xCOl_xO3-s[201], LaMnl_xCrl_xO3_~ [202], Lal_xCaxCOl_yCryO3_~[203] and Lal_xSrxCOl_yFeyO3_s[204]. Preferential electronic charge compensation may occur in these compounds, i.e. the charge carrier may (at low temperature) be temporarily trapped at the small polaron site which is lower in energy, thereby decreasing the electrical conduc-
494
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
tivity. This effect disappears at high temperature, when the thermal energy is sufficient to surpass the barrier between the traps and the more conductive hopping sites, or when the population of low energy sites exceeds the percolation limit. In the latter case the electrical conductivity is controlled by the short-range hopping among the lower energy sites. Based upon results from electrical conductivity and thermopower measurements, in conjunction with some other techniques, Tai et al. [204] concluded that in La0.sSr0.2COl_yFeyO3. 8 compensation of Fe 3+ --~ Fe4+ is more likely than Co 3+ --~Co4+. The electrical conductivity and the spin-transition state of LaCoO34 and La1_xSrxCoO34 has been studied extensively. The covalent mixing of orbitals induces itinerant behaviour of the charge carriers in La1_~SrxCoO34 [205]. Racah and Goodenough [206] claimed a first order localized-electron to collectiveelectron phase transition in LaCoO3 at 937~ in air, though the electrical conductivity was found to be continuous at the transition. Electrical conductivity and differential thermal analysis (DTA) behaviour of ACoO3. 6 (A = Nd, Gd, Ho, Y, La) were investigated by Thornton et al. [207]. Evidence was adduced that the transition previously noted in LaCoO3 (and in the other cobaltates) may be caused by the presence of binary cobalt oxides. The endothermic heat effect observed in DTA can be correlated with a similar feature obtained from a sample Co30 4. Instead, a gradual semiconductor-to-metal transition with increasing temperature is suggested in compounds ACoO3,s. For LaCoO3& which adopts a rhombohedrally distorted perovskite structure at room temperature, this range was found to extend from 110 to 300~ Analysis was based in part upon the temperature independent magnetic susceptibility observed at high temperature, which was attributed to Pauli-paramagnetism (with possible Van Vleck type of contributions) of the conduction electrons. Data produced by Mizusaki et al. [208] suggest a close relationship between the transition temperature of the conductivity and the rhombohedral angle, which gradually decreases with increasing x in Lal_xSrxCoO3_~.The conduction becomes metallic when the rhombohedral angle becomes smaller than 60.3 ~ noting that the value of 60 ~ corresponds with ideal cubic symmetry. Another factor is the extent of oxygen non-stoichiometry. In the metallic region, the electrical conductivity of Lal_xSrxCoO3.5 decreases almost linearly with increasing 8. Apart from changes in d-band occupancy, it is assumed by the authors that band narrowing takes place with increasing oxygen nonstoichiometry. Finally, the data of electrical conductivity and thermopower of Lal_~Sr• at elevated temperature has been treated by a hopping mechanism for x < 0.2, and by a band model for the semi-metallic behaviour observed at x > 0.3 by Mizusaki [161]. On the other hand, to explain their results of Lal_xSrxMnO3_~and Lal_xSrxMnO34 for compositions with 0.30 < x _<0.80, Stevenson et al. [209] included the thermally activated charge disproportionation of Mn 3+ into Mn 2+ and Mn 4+pairs.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
495
10.6.6 Extended Defects and Vacancy Ordering As is known for fluorite and fluorite-related oxides, increased defect interactions are likely if the oxygen vacancy concentration exceeds 1 mol% [81]. The interaction between defects and defect association effectively lower the concentration of 'free' oxygen vacancies available for oxygen transport. In the perovskites, it is not uncommon to have an oxygen deficiency of 10 tool% or more. As noted before, the assumption of randomly distributed point defects at such large vacancy concentrations probably is an oversimplified picture. Van Roosmalen and Cordfunke [210] showed that by assuming divalent transition metal ions in undoped perovskites LaBO3_~(M = Mn, Fe, Co) to be bound to oxygen vacancies, forming neutral defect clusters of the type , the model fit to the experimental data was greatly improved. Similar, but probably more complicated, extended defects were suggested to be the structural building elements of highly defective perovskites. In general, the tendency to form ordered structures progressively grows with increasing defect concentrations. As a matter of fact, ordering of oxygen vacancies in the perovskite and perovskite-related structures seems more common than their random distribution between the perovskite slabs. Sometimes the two limiting cases are linked for the same composition by an order-disorder transformation, driven by the gain in configurational entropy of oxygen vacancies in the disordered state at elevated temperature. It has also been suggested that any ordering of the oxygen vacancies in the defective perovskite and perovskite-related oxides, thereby confining vacancy transport to two-dimensional layers, may give rise to fast ionic conductivity at significantly reduced temperatures [211,21], a point to which we return below.
10.6.6.1 Static Lattice Simulation
Attempts were made by Kilner and Brook [213] to model the ionic conductivity in perovskites LnA103 using static lattice simulation techniques. Here, the minimum energy positions for the mobile ions, and the activation energy barrier that they must surmount to migrate through the rigid crystal lattice, are calculated by minimization of the total lattice energy. The results show that aliovalent dopants might act as trapping centres for oxygen vacancies through the formation of defect associates, e.g., V6-SrLa'. It is further found that the size proportion of A and B cations is of significant importance in determining the minimum migration enthalpy for oxygen transport in the ABO3 structure. During diffusion, the migrating 0 2- ion must pass the saddle point formed by two A ions and one B ion, as shown in Fig. 10.14. The associated energy barrier to migration decreases with increasing size of the B cation and decreasing size
496
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
B cation
A cations
Fig. 10.14. Saddle-pohlt configuration for oxygen anion migration.
of the A cation. This work has been extended recently by Cherry et al. [214] to include perovskites LaBO3 (Cr, Mn, Fe, Co), showing the importance of relaxation effects at the migration saddle point, which were not invoked in the previous study by Kilner and Brook. Profile mapping shows that the migrating 0 2- ion does not prefer a linear path between adjacent sites of a B O 6 octahedron, but rather follows a curved route with the saddle point away from the neighbouring B site cation. The pioneering work by Kilner and Brook was further expanded by Cook and Sammells [215] and Sammells et al. [216] to include additional empirical relationships for predicting oxygen ionic conductivity in perovskite solid solutions, such as the average metal-oxygen bonding energy, the lattice free volume and the overall lattice polarizibility towards anion migration. A marked correlation found is that between the activation energy of oxygen anion migration in the perovskite lattice and the free volume. A smaller activation energy is apparent in ABO3 perovskites which possess an inherently larger free volume. For fluorite-structured oxygen ion conductors, again linear but opposite correlations are found. This contradiction might be solved if there is optimum value of the lattice free volume at which the Coulombic, polarization and repulsive contributions to the migration enthalpy in both type of structures are best balanced. The results obtained led to the identification of perovskite oxide electrolytes BaTb0.9In0.1034, CaCe0.9Gd0.1034 and CaCe0.9Er0.1034, which indeed exhibit low activation energies varying between 35-53 kJ mo1-1 for the ionic conductivity [216]. Even, though, the suggested correlations guide the selection towards materials exhibiting a low value of the ionic migration enthalpy for oxygen anions and protons (if assumed to occur via OH-), none of the above authors did address the problem how to optimize the magnitude of the pre-exponential term in the expression of the overall ionic conductivity. The pre-exponential term is related in part to the density of mobile oxygen anions and the
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
497
availability of sites (e.g., vacancies) to which they might jump. Its value will thus be determined by the state of order in the oxygen sublattice 4.
10.6.6.2 Vacancy Ordering X-ray diffraction, electron diffraction and HRTEM have provided ample evidence that, particularly at low temperature, nonstoichiometry in oxygen-deficient perovskites is accommodated by vacancy ordering to a degree which depends both on oxygen partial pressure and on the thermal history [219,220]. A family of intergrowth compounds is observed for, e.g., the ferrites, which can be regarded as being composed of perovskite (ABO3) and brownmillerite (A2B205) structural units stacked along the superlattice axis. The intergrowth structures fit into a homologous series expressed by AnBnO3n_l, with end members n = (perovskite) and n = 2 (brownmillerite). For non-integral values of n, disordered intergrowths are observed between nearby members of the ideal series. The concept of 'perovskite space' has been introduced by Smyth [221,222] in order to systematize the intergrowth structures exhibited by various oxygendeficient and oxygen-excess perovskite systems. In the proposed diagram, the close structural relationship between the parent and intergrowth structures is expressed by plotting the value of n along with the compositional excursion from ideal perovskite stoichiometry. The perovskite systems are found to be quite specific in their tendency to form ordered structures in the sense that only selected values of n are found for a particular system. The observed vacancy patterns and three-dimensional structures vary along with lateral shifts in the stacking sequence of successive layers AO3_6. Anderson-et al. [223] showed the driving force towards ordering to be strongly dependent u p o n the size and electronic configuration of the B-cation, in addition to the size and coordination preference of the A-cation. In general, c o m p o u n d s that contain ordered vacancies are found for n = 5, 4, 3, 2,1.5,1.33, and 1, that is for overall oxygen contents 2.8, 2.75, 2.67, 2.5, 2.33, 2.25 and 2 [223]. Unfortunately, the structural studies are usually carried out at room temperature and little is k n o w n about the extended defects at elevated temperatures, and their behaviour during quenching and cooling. 4
Ordering reduces the number of free ionic charge carriers and, in general, has a negative impact on the magnitude of the ionic conductivity. An exception to this rule is apparent in selected pyrochlores with composition Ln2Zr207 (Ln = La-Gd), and is best illustrated for Gd2Zr207 [217,218]. The pyrochlore structure can be derived from that of fluorite by ordering of both cations and vacancies on their respective sublattices. Electron microscopy has shown that the actual microstructure of these solids consists of ordered pyrochlore domains embedded in a disordered fluorite matrix. The degree of ordering can be varied by thermal annealing. Ordering causes both a lower pre-exponential term and activation enthalpy, and leads to an overall ionic conductivity for well ordered Gd2Zr207 competitive with values as measured for stabilized zirconia. The observed phenomena have been interpreted to reflect the presence of high diffusivity paths in the pyrochlore structure.
498
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
10.6.6.3 Microdomain Formation
The structural rearrangements accompanying the redox phenomena sometimes lead to a texture of the sample consisting wholly of microdomains, which may be of varying size and composition. The superstructures observed depend upon local composition, while their orientation within each microdomain may be randomly distributed. As the domains are usually smaller than the sampling size (<500 A) of X-ray and neutron diffraction, such investigations require the use of high resolution transmission electron microscopy and electron diffraction. As an example, we briefly describe here the observations made for Lal_xCax_ FeO3_~. In this solid solution system, a single perovskite-brownmillerite intergrowth structure (n = 3) is found at composition x = 2/3 [224]. The intergrowth structure exhibits ideal stoichiometry, i.e. LaCa2Fe308, at reduced oxygen partial pressures near the minimum observed in the electrical conductivity [199]. For any other composition, a disordered intergrowth is observed. Oxidizing LaCa2Fe3Os in air at 1400~ and quenching to room temperature produced a material which appeared cubic to X-rays. The most simple explanation would have been that oxygen vacancies become disordered in a highly defective perovskite structure, but transmission electron microscopy and diffraction revealed the existence of three-dimensional intergrowths of domains of the parent phase [224]. Oxygen excess is therefore considered to be accommodated at the domain walls, suggesting that the size of the domains is closely related to the oxygen excess, i.e. the greater the oxygen excess the smaller the domains. In the regions 0 < x < 2/3 and 2/3 < x < 1, disordered intergrowths are observed with members n = 2 and n = ~, respectively. In the reduced samples, it is assumed that the local composition is very heterogeneous and a variable amount of oxygen can be accommodated as the Ca/La ratio changes. Oxidizing the disordered intergrowth structures again led to quenched samples with a microdomain texture having an estimated size of domains of about (100 A) 3. In each of these cases up to six types of microdomains were observed, three for each of the end members, e.g. LaCa2Fe308 and Ca2Fe205 for the region 2/3 < x < 1. Since the fringe separation within each of the microdomains was found to be very regular, the oxygen excess is probably again located at the domain walls [225]. In fact, the domain texture appeared to be not very stable either when kept in air at room temperature or under reducing conditions (as present in the electron microscope). After long operation in the electron microscope up to nine sets of microdomains were observed in the oxidized composition LaCa2Fe3Os.2~, three for each member in the ideal series, n = 2 (Ca2Fe2Os), n = 3 (LaCa2Fe3Os) and n = ~ (LaFeO3). Other materials exhibiting a three-dimensional microdomain texture include CaMnl_xFexO3_~ [226,227], BaxLal_xFeO3_~ [228] and Sr2CoFeO5 [229], while a few
10 - - DENSE C E R A M I C MEMBRANES FOR OXYGEN SEPARATION
499
more examples are given below. Although it may be anticipated that mass transport is strongly influenced by the presence of a microdomain texture, the question is open as to whether these microdomains are also present at high temperatures. In a study by M6ssbauer spectroscopy, X-ray diffraction and magnetic susceptibility, Battle et al. [229] found a microdomain texture in partially oxidized samples Sr2CoFeOs~ quenched in air from 1200~ but not in substoichiometric samples obtained either after quenching in liquid nitrogen from 1200~ or after annealing under argon at 800~ Furthermore, the brownmillerite microdomains thus produced in air-quenched Sr2CoFeO54 were found to be significantly larger in size in the centre of a pellet, while oxidation at the surface had produced a new phase, presumed to be perovskite. The authors arrived at the conclusion that the microdomains were produced by the accommodation of excess oxygen in the domain walls, arising from the diffusion of oxygen into the oxide lattice during quenching, and therefore would not be in thermodynamic equilibrium.
10.6.6.4 Brownmillerite Structure The well-known brownmillerite structure with ideal stoichiometry A2B205 can be derived from the ideal perovskite lattice by regularly removing one sixth of the oxygen atoms, resulting in an orthorhombic ~ q2-a x 4a x ~-a supercell. The ordering of oxygen vacancies thus creates layers of corner-shared B O 6 octahedra (O) alternating with layers of corner-shared B O 4 tetrahedra (T) stacked in an ...OTOTOT... sequence, as shown in Fig.t0.15. No less than six unique vacancy patterns are known for this type of structure [223]. The structures tend to disorder at elevated temperature, so as to produce a material with a high intrinsic concentration of mobile oxygen vacancies [36], i.e. not achieved by aliovalent doping. Indeed, Ba2In20 5 which has the brownmillerite structure of Ca2Fe205 becomes a fast oxygen conductor, with ionic conducfivities exceeding those of stabilized zirconia, after a first-order transition at about Tt = 930~ at which the oxygen sublattice disorders [211]. The situation is analogous to that of 5 - B i 2 0 3 , which undergoes a first-order order--disorder transition at T t - 730~ [230]. On the other hand, no increased ionic conductivity is found after a similar first-order transition in the isostructural Sr2Fe205 at Tt = 700~ which suggests an ordered arrangement of oxygen vacancies above Tt, while Ca2Fe205 remains stable up to at least 1100~ [231,232]. Examination of Sr2Fe205 by M6ssbauer spectroscopy showed that tetrahedral coordination of Fe persists even in the apparently disordered material. It was therefore suggested that the ordering breaks up into microdomains of a size not detectable by XRD measurements [233]. Seeking for solutions to lower Tt in Ba2In205, Goodenough et al. [211,212] extended the foresaid investigations to include Ba3In2MOs (M = Zr, Hf, Ce). Ideal ordering of the oxygen vacancies in these compounds would render them
500
10 m DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
O T
[110]c i
OT (a)
~
(b)
Fig. 10.15. Idealized structure of (a) cubic perovskite (CaTiO3) and b) orthorhombic brownmiUerite (Ca2Fe205) lattice with ordered oxygen vacancies if-I) along the cubic [110] direction.
isostructural with Ca3Fe2TiO8, in which two octahedral (O) layers alternate with one tetrahedral (T) layer in an ...OOTOOT... sequence along the c-axis [234]. Accordingly the structure may be regarded as an intergrowth (n = 2) of perovskite layers alternating with brownmillerite layers. Enhanced ionic conduction of, in particular, Ba3In2Zr208 at modest temperatures was interpreted to reflect the disordering of the In 3+ and Zr4+-cations over the cation sublattice, which would lead to interstitial oxygen in the T-layers and correspondingly to oxygen vacancies in the O-layers, due the preference of the Zr4+-cations for octahedral coordination versus tetrahedral coordination. In a subsequent study [235], however, these high values could not be reproduced. In fact, long-range ordering does not occur, and the residual low-temperature conduction is due to facile insertion of oxygen a n d / o r water even below 400~ In moist air, the ceramic discs crumbled on repeated thermal cycling. Only short-range ordering of the oxygen vacancies is anticipated. It is further suggested that any ordering of the oxygen vacancies in the oxygen deficient perovskites so as to create layers of corner-shared B O 4 tetrahedra (T), as in Ca3Fe2TiO8, makes more facile the insertion of oxygen a n d / o r water at modest temperatures.
10.6.6.5 High TemperatureNMR A point defect model has been proposed for the brownmillerite structure, based on a study of electrical conductivity and emf measurements of Ba2In20 5 [236]. The experimental results matched well with modelling, while evidence was found for protonic conduction at low temperatures. In modelling, the unoccupied oxygen sites relative to the perovskite structure, below Tt, a r e regarded as structural units and therefore are potential interstitial sites for
10 - - D E N S E CERAMIC MEMBRANES FOR OXYGEN SEPARATION
501
oxygen. Charge neutrality is then dominated by intrinsic anion Frenkel disorder, i.e. [Oi"] = [Vo']. Above Tt, charge neutrality would be determined by the condition 2[Vo'] - [ I n l n ] . Partially at variance with these findings, Adler et al. [237] arrived at the conclusion that even above the order-disorder transition observed at T t = 925~ local order in the oxygen sublattice persists. Vacancy transport in Ba2In205 would be confined mainly to two-dimensional layers, whilst oxygen anions are bounded and thus immobile in adjacent layers. Using high-temperature 170NMR and X-ray diffraction, it was demonstrated that the material retains its orthorhombic brownmillerite structure until ~1075~ at which point it becomes cubic. The density of mobile oxygen anions was found to increase continuously between 925~ and ~1075~ and only above -~1075~ the full population of oxygen anions becomes mobile. The investigations were extended by Adler et al. [238] to include BaIno.67Zro.3303. & BaIno.67Ceo.3303_ 6, and the cobalt-containing perovskites La0.5Ba0.sCo0.?Cu0.3034i and La0.6Sr0.4Co0.sCu0.2034i. Whilst room temperature X-ray and neutron powder diffraction revealed simple long-range cubic symmetry, HRTEM images and electron diffraction patterns indicated that these materials possess microdomains with a layered-like structure on a length scale 50-500 A. Measurements of high-temperature 170-NMR made apparent that in all these phases only a few oxygen vacancies are mobile below 800~ suggesting that the remainder is trapped in locally ordered layers. Owing to intrinsic paramagnetism of the cobalt-containing compounds, the application of 170NMR is not possible at room temperature. Only by rapid motion of oxygen, the spectra become visible. For both cobalt-containing perovskites, the signal intensity above 800~ was found to increase steadily with temperature, up to the maximum temperature of 950~ in the experiments by Adler et al. [238], suggesting a concomitant increase in the number of mobile oxygen anions. In the case of La0.6Sr0.4Co0.8Cu0.2034, the value of the ionic migration enthalpy for oxygen transport estimated from NMR, viz., 1.13 + 0.10 eV, showed good agreement with the 1.04 eV obtained from 4-point ionic conductivity measurements. It is clear that high-temperature NMR is an important tool to study the local structure and oxygen dynamics in the highly defective perovskites, which certainly merits its further use. In the case of La0.6Sr0.4Co0.sCu0.2034, Adler et al. [238] assumed that oxygen nonstoichiometry is accommodated by local ordering of MOs square pyramids in a manner similar to that in YBaCuFeO5 [239] and YBaCo2_xCuxO544 [240]. A similar type of structure was hypothesized for Lal_xSrxCoO34i (x = 0.50 and 0.70) by Van Doorn et al. [241]. As seen in Fig. 10.16, two different sites for oxygen are present in the proposed structure for Lal_~SrxCoO3_s. Four oxygens are in the basal plane of the square pyramid and one at the apex. Electron diffraction and HRTEM of powders, annealed about 15 h in air and furnace-cooled, indicated that the presence of the tetragonal super-
502
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
i
i
i
9 (~ @ I-I
Co La/Sr 0 O-vacancy
Fig. 10.16.Ideal structureproposed for Lal-,SrxCoO34(x = 0.5 and x = 0.7).Aftervan Doom et al. [241].
structure, corresponding to a doubling of the pseudocubic perovskite unit cell, gives rise to a texture consisting of microdomains with a mean size estimated at ca. (500 ~ ) 3 .
10.6.7 Observations from Permeability Measurements This section discusses selected observations from oxygen permeation experiments. General trends have been summarized already in Section 10.6.1.
10.6.7.1 SrCoo.sFeo.203.6 The highest oxygen flux at 850~ in the series Lal_~SrxCol_yFeyO3_~studied by Teraoka et al. [37] was found in SrCoo.sFe0.2034 (see Table 10.1) and this composition has been reexamined by a number of investigators [13,41,153,154]. In a study of oxygen permeation through SrCo0.8B0.2034 (B = Cr, Fe, Co, Cu), Kruidhof et al. [154] showed for the first time that in SrCo0.8Fe0.2034 a change in permeation mechanism, corresponding with an order-disorder transition, occurs at about Tt = 790~ This change has been observed by others too [13,153] but was not noted in the previous study by Teraoka et al. [37]. Results from thermal analysis and X-ray powder diffraction under controlled oxygen partial pressures indicate that the observed phenomena can be attributed to the transition of SrCo0.sFe0.203_6from vacancy-ordered brownmillerite to defective perovskite, occurring at reduced oxygen partial pressure. Both the structures of the cubic p e r o v s k i t e a n d o r t h o r h o m b i c b r o w n m i l l e r i t e - t y p e p h a s e s of SrCo0.8Fe0.2034 have been refined using powder neutron diffraction data by
10 - - D E N S E C E R A M I C M E M B R A N E S FOR OXYGEN S E P A R A T I O N
503
Harrison et al. [242]. A phase diagram was presented by Qiu et al. [153], showing that only at relatively high oxygen partial pressure (>0.1 atm) and high temperature the perovskite phase is thermodynamically stable. At relatively low oxygen partial pressure and low temperature a perovskite-brownmillerite two-phase region is found. The brownmillerite phase has only a small homogeneity region around 3 - 5 = 2.5. Below Tt, the situation during flux measurements therefore becomes very complicated, considering the fact that the Po2gradient across the membrane also may cross the two-phase region provided, of course, that such a gradient is imposed during experiment. The studies report slow kinetics of transformation between the brownmillerite and perovskite phases in view of the long times for the oxygen flux to reach steady-state conditions at these modest temperatures. Kruidhof et al. [154] attributed these to a progressive growth of microdomains of the ordered structure in a disordered perovskite matrix. Based on experiments, in which the membrane thickness was varied in the range 5.5-1.0 mm, Qiu et al. [153] arrived at the conclusion that the surface oxygen exchange process is the rate limiting step in the overall oxygen permeation mechanism. Further experimental evidence that the oxygen fluxes through SrCo0.8Fe0.2034 are limited by the surface exchange kinetics was given by the present authors [41]. Fitting the oxygen permeation fluxes obtained from measurements at 750~ under various oxygen partial pressure gradients to Eq. (10.18) yielded a positive slope of n = +0.5, where a value between 0 and -0.5 is expected from the experimentally observed In 5--ln PO2 relationship" However, these results merit further investigation as the flux data were taken at a temperature just below the order--disorder transition in this material. It is already known for some time that SrCoO3_ 6 transforms reversibly from a brownmillerite-like structure to defective perovskite at about Tt = 900~ in air. Kruidhof et al. [154] observed that the transition temperature is not, or only slightly, affected if SrCoO3_ a is substituted with either 20 mol% Cr or Cu at the Co-sites. Interesting to note is that the oxygen flux for the undoped and doped specimens is very small below T t, as expected for an ordered arrangement of oxygen vacancies, but is found to increase sharply (between 5-6 orders of magnitude) at the onset of the phase transition to defective perovskite, up to values between 0.3-3 x 10 -7 mol c m -2 s -1. In view of these results, the perovskite phase in SrCoO3_ a s e e m s to be stabilized by the partial substitution of Co with Fe, but not with Cu or Cr, thereby suppressing the brownmillerite-perovskite two phase region to lower oxygen partial pressures.
10.6.7.2 Experimental Difficulties In a number of studies, the oxygen fluxes through, e.g., SrCo0.sFe0.203_a have been reported to be significantly lower than claimed by Teraoka et al. [37]. Such
504
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
conflicting results reflect the difficulties in measuring the oxygen fluxes at high temperatures and may, at least partly, be due to specific conditions, including (1) edge-effects associated with the required sealing of sample discs to avoid gas bypassing, giving rise to non-axial contributions to the oxygen flux, (2) possible interfacial reactions when a glass is used for sealing, (3) undesired spreading of the glass seal (when its softening temperature is too low) over the oxide disc surface, and (4) the precise value of the Po2-gradient across the membrane. With regard to the first point, it is frequently the cross-sectional area of the disc that is used in the calculation of the oxygen flux. In the usual experimental arrangements however an appreciable portion of the membrane is 'clamped' between impermeable annular plates or glass rings. This edge effect means that the usual assumption of one-dimensional diffusion is not strictly correct. Another contribution to non-axial transport is that of flow of oxygen through the side walls of disc specimens, if left uncovered. Appreciable errors creep in if these edge effects are neglected as shown, for example, on the basis of a solution of Fick's second diffusion equation (with a constant diffusion coefficient) by Barrer et al. [243]. This is further demonstrated in Fig. 10.17, showing the effect of sealing edges on the departure from one-dimensional diffusion. These results were obtained from a numerical procedure to solve the steady-state diffusion equation in cylindrical coordinates [112]. Neglecting edge effects corrupts analysis of experiments in which the membrane thickness is varied, and may lead to erroneous conclusions when one tries to infer from the acquired data the influence of the surface exchange kinetics on overall oxygen permeation. Finally, it cannot be excluded that the observed oxygen fluxes are specific for the particular sample under investigation and may be affected, for instance, by microstructural effects, a point to which we return in Section 10.6.7.5. The gas flow rate of, in particular, the inert gas used to sweep the oxygen-lean side of the membrane affects the Po~ -gradient across the membrane. Under ideal gas mixing conditions, the Po2 at the oxygen-lean side of the membrane is determined by the amount of oxygen permeating through the membrane. If the flow rate is not adjusted to obtain a constant Po2 at this side of the membrane, but a constant gas flow rate is used, the Po2-gradient gets smaller with increasing oxygen flux. This may give rise to an apparent activation energy for overall permeation, which may depart significantly from the one derived if a constant Po2 were maintained at this side of the membrane [148,149,153]. The adjustable range of the sweeping gas flow rate (to a constant Po2 at the outlet of the reactor) may be limited during experiment, being determined by the requirement that the reactor behaviour remains close to that of a CSTR (continuous stirred tank reactor). Using a constant value of Po2 at the oxygen-lean side of 2 m m thick disc membranes of Lal_~SrxCoO3,s (x = 0.2, 0.3, 0.4, 0.5 and 0.6), Van Doorn et al. [148,149] showed that the activation energy Eact for oxygen permeation in the
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
505
2b
(a)
membrane r
seal
2a
(b)
1.40
1.30~
3.75
C9 t_
0
dk.J
o r o
. I t_.
1.20 7.5
(1)
E 0
ID 1.10
1.00
, 0.70
, 0.80
,
, 0.90
, 1.00
a/b Fig. 10.17. (a) Schematic cross-section of a disk membrm~e. Dashed parts indicate insulating boundaries. (b) Influence of sealing edge-effects oil the departure from one-dimensional diffusion. A geometric factor G is used for correction of the flux (normalized to surface area with diameter 2a). Relevant parameters are defined in Fig. 10.17a.
r a n g e 900-1100~ decreases from 164 kJ mole -1 for x = 0.2 to 81 kJ mole -1 for x = 0.6. O p p o s e d to these results, Eact decreased from 121 kJ mole -1 to 58 kJ mole -1 w h e n a constant gas flow rate o f the h e l i u m w a s used. Besides an i m p r o v e d fit to the A r r h e n i u s equation in the former case, Eact can be correlated w i t h the s u m of the enthalpies for m i g r a t i o n a n d that for the formation of oxide ion vacancies for each of the investigated compositions. Such a correlation is expected if o x y g e n t r a n s p o r t is d r i v e n b y the g r a d i e n t in o x y g e n nonstoichiom e t r y across the m e m b r a n e d u e to the i m p o s e d P o d g r a d i e n t . It suggests that o x y g e n vacancies are free and non-interactive in Lat_xSrxCoO34 u n d e r the con-
506
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
ditions covered by experiment. Oxygen permeation fluxes for strontium-doping levels above x = 0.6 were found to be partially controlled by the surface exchange kinetics, as already mentioned in Section 10.6.4. In contradiction to the observed behaviour at high temperatures, results from thermal analysis and oxygen permeation measurements indicated that a phase transition, with a small first order component, probably related with order-disorder of oxygen vacancies, occurs in selected compositions Lal_xSrxCoO3,s in the range 750-775~ [148,149]. Long times extending to over 30 h were needed for equilibration towards steady-state oxygen pemeation at these modest temperatures. Such a behaviour is reminiscent of that observed for SrCo0.8Fe0.203@ where this can be attributed to the slow kinetics of the transformation between the brownmillerite and perovskite phases at modest temperatures. In the case of La1_vSrxCoO34 (x = 0.50 and 0.70), microdomains were observed in electron diffraction and HRTEM, corresponding to ordered arrangements of oxygen vacancies in these compounds at room temperature, as mentioned in the previous section. Another factor that is considered to be responsible for a reduced oxygen flux is the surface modification of the perovskite oxide membrane by reaction with impurities in the gas phase, as emphasized by Qiu et al. [153]. Referring to the surface degradation by reaction with minor amounts of CO2 and corresponding deterioration of the properties observed for YBa2CuO6+ x superconducting thin films [244], a similar modification effect could occur when, e.g. ambient air is used as the source of oxygen at the membrane feed side. With the help of N2 and 02 admixed to feed side pressure Po2' = 0.21 atm, Qiu et al. found the oxygen fluxes through SrCo0.8Fe0.2034 in the range 620-920~ to be larger by a factor of about 6 than when ambient air was used as feed gas, but still a factor of about 5 smaller than measured by Teraoka et al. Similar experiments were conducted in our study on SrCo0.8Fe0.203,s [41,154], where this effect was not noted in the temperature range 700-950~ so that we are inclined to believe that other factors must account for the disagreements in oxygen fluxes. This interpretation is supported by experimental evidence disclosed in a number of patents: that the oxygen fluxes through perovskite membranes remain stable as long as these are operated above certain critical temperatures, the precise value depending on the type of alkaline-earth dopant applied. Below these temperatures, a loss in oxygen flux may be observed over a period of about 100 h by as much as 30-40% when a membrane is exposed to CO2 and H20 impurities in the feed gas. This is further exemplified in Section 10.7.
10.6.7.3 Surface Exchange Kinetics Attention has already been drawn to the importance of the surface exchange kinetics in determining the rate of oxygen permeation through mixed-conduct-
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
507
ing oxides in Section 10.3.2.2. Though for the perovskites a value of 100 ~tm is often quoted for the characteristic membrane thickness Lc, at which the change over from bulk to surface control occurs, in a number of cases much higher values are found, up to about 3000 ~tm (Table 10.2). As was emphasized earlier, the parameter Lc is not an intrinsic material property and, hence, may be specific to the sample under investigation and experimental conditions. The basic assumptions made in the derivation, notably that of small Po -gradients across the membrane, may restrict its use in practical situations, where these gradients can be substantial. Experimental evidence that the oxygen fluxes are limited by the surface exchange kinetics has been found in a number of cases, as discussed elsewhere in this text.
10.6.7.4 Behaviour in Large Po2-Gradients The mixed-conducting perovskite oxides have attracted particular interest for use as dense ceramic membrane to control partial oxidation of methane to C2-products or syngas. Such a process bypasses the use of costly oxygen since air can be used as oxidant on the oxygen-rich of the membrane. Using SrCo0.8Fe0.2034 tubular membranes fabricated by an extrusion method, Pei et al. [13] observed two types of fracture of the tubes during the process for generating syngas. The first fracture, occurring short (within 1 h) after initiation of the reaction at 800~ resulted from the Po -gradient across the membrane and the accompanying strain due to lattice mismatch and the brownmillerite-perovskite phase transition. The second type offracture, occurring after prolonged exposure to the reducing environment, resulted from chemical decomposition towards SrCO3, and elemental Co and Fe. Similar observations have been reported for tubes made of La0.2Sr0.8Co0.4Fe0.603-~ [14], and in that study an optimized composition was also claimed, but not given, showing stable performance for up to 500 h. Using a rhodium-based reforming catalyst inside the tubes, methane conversions over 99% were achievable. Ten Elshof et al. [10] studied the oxidative coupling of methane using a disc reactor with La0.6Sr0.4Co0.sFe0.2034 as the catalyst membrane for the supply of oxygen to the methane feed stream. Examination of the oxygen fluxes measured under various Po2-gradients in the range of thickness 0.55-0.98 mm suggested that the surface exchange reaction limits the rate of oxygen permeation. The oxygen flux was found to increase only slightly when methane was admixed with the helium used as the carrier gas. The methane was converted to ethane and ethene with selectivities up to 70%, albeit with a low conversion, typically in the range 1-3% at operating temperatures 1073-1173 K. The selectivity observed at a given oxygen flux and temperature was about twice as low if the same amount of molecular oxygen was co-fed with the methane feed stream in a single chamber reactor design, suggesting that the membrane-mode of operation
508
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
is conceptually more attractive for generating C2-products. Decomposition of the oxide surface did not occur as long as molecular oxygen could be traced at the reactor outlet, which emphasizes the importance of surface-controlled oxygen flux for membrane-driven methane coupling. That is, for a bulk diffusion-controlled oxygen flux the surface would become reduced by the methane, until the depth of reduction has progressed up to a point where the oxygen flux counterbalances the consumption of oxygen by methane. On the one hand the slow surface exchange kinetics observed on La0.6Sr0.4Co0.aFe0.2034limits the magnitude of the oxygen fluxes, on the other hand its existence prevents the oxide surface from reduction, i.e. as long as the rate of oxygen supply across the membrane exceeds the rate of (partial) oxidation of methane. Noteworthy is that segregation of strontium occurred on both sides of the membrane, as confirmed by depth-profiling Auger analysis. The extent of segregation appeared to be influenced by the imposed Podgradient across the membrane, and was also found if a pure helium stream was passed along the oxygen-lean side of the membrane. Van Hassel et al. [150] studied oxygen permeation through Lal.~SrxFeO3,s (x = 0.1, 0.2) membranes in a disc reactor using CO-CO2 based gas mixtures to control the Po2 at the oxygen-lean side. Ambient air was used as the oxygen source at the opposite side of the membrane. At 800-1100~ the oxygen flux was found to increase linearly with the partial pressure of CO. Deposition of a 50 nm thin porous Pt layer on this side of the membrane increased the oxidation rate and likewise the oxygen flux, by a factor of about 1.8. In a separate study [245], the oxygen flux was found to be invariant with the thickness of the membrane in the range 0.5-2.0 ram, while no effect was observed upon varying the Po~ at the oxygen-rich side. It was concluded that the oxygen flux is fully limited by the carbon monoxide oxidation rate. The experimentally determined rate constants scale with Sr-content in the extended range of composition 0.1 < x < 0.4. The latter can be accounted for, in view of the fact that the oxygen deficiency of the ferrites is fixed by the dopant concentration in a wide range of oxygen partial pressure, by assuming that oxygen vacancies act as active sites in the oxidation reaction of CO on the perovskite surface following either an Eley-Rideal or a Langmuir-Hinselwood type of mechanism.
10.6.7.5 Grain Boundary Diffusivity Besides the possibility of surface exchange limitations, oxygen transport through dense ceramics is necessarily influenced by the presence of high diffusivity paths along internal surfaces such as grain boundaries. A systematic study investigating to which extent these preferred diffusion paths contribute to the diffusivity in the perovskite oxides is however still lacking. Both impurity and solute segregation take place at grain boundaries (and the external surface) or in their close proximities (less than 3 or 4 atomic distances) during sintering
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
509
and subsequent heat treatments. An obvious consideration is that, in general, these significantly alter the magnitude of ionic transport along and across the grain boundaries. In many cases the ceramics invariably contain impurities present in the starting powder or added as a sintering aid to lower the sintering temperature and/or to achieve high density. It therefore can not be excluded that disagreements in the literature regarding the magnitude of the oxygen fluxes can be explained on the basis of different ceramic processing techniques used by various authors. In general, the presence of high diffusivity paths is important in ceramics where lattice diffusion is slow. Analyzing 180 depth profiles using secondary ion mass spectroscopy, Yasuda et al. [246] noted a significant contribution of the grain boundary diffusion to the diffusivity in the interconnect material La0.?Ca0.35CrO34, where the tracer diffusivity is of the order of ~10 -13 cm 2 s -1 at 900~ Erroneous results were obtained when isotopic exchange was performed by gas phase analysis, which resulted in apparent tracer diffusion coefficients that were almost 2 orders in magnitude higher. More recently, Kawada et al. [184] confirmed the existence of high diffusivity paths along grain boundaries in La0.~Ca0.3CrO34 using depth profiling and imaging SIMS of lso-160 exchanged specimens. But, oxygen permeation measurements suggested negligible contribution of grain boundary diffusion to the steady-state oxygen flux. These data were obtained at 1000~ for a sample of thickness 0.75 mm. An oxygen pump and sensor were used to control the permeate side Po2. The results are well-described by the Wagner equation assuming a random point defect scheme for La0.~a0.3CrO34, as discussed in Section 10.6.2.2. For fast ionic conductors grain boundary diffusion will have little influence, or indeed may become blocking to the diffusion from one grain to the next as is recognized in the interpretation of impedance spectra from ionic conductivity of zirconia and ceria-based solid electrolytes. In these ceramics silicon is the most common impurity detected along with enhanced yttrium segregation. Various models to account for the effects of segregation at grain boundaries and how these affect the electrical properties have been discussed by Badwal et al. [247]. Although there is no unique model describing the ceramic microstructure, the most widely adopted model for doped zirconia and doped ceria is the brick-layer model. In this model bricks present the grains and mortar the grain boundary region, i.e. assuming the grain boundary phase to completely wet the grains [248,249]. The grain boundaries in series with the grains, along the direction of charge flow, mainly contribute to the grain boundary resistivity. For doped zirconia and ceria the grain boundary resistivity can be of similar order of magnitude or higher than the bulk resistivity. Pores at the grain boundaries can have a positive effect on the oxygen transport. It is evident that more detailed studies are needed to aid in the interpretation of oxygen transport through the mixed-conducting perovskite oxides, where similar blocking effects can be expected.
510
10.7
10 - - D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
FINAL
R E M A R K S
The considerations in this chapter were mainly prompted by the potential application of mixed-conducting perovskite-type oxides to be used as dense ceramic membranes for oxygen delivery applications, and lead to the following general criteria for the selection of materials - high electronic and ionic conductivity, - high catalytic activity towards oxygen reduction and reoxidafion, - ability to be formed into dense thin films, free of micro-cracks and connected-through porosity, chemical and structural integrity (i.e. no destructive phase transition) within appropriate ranges of temperature and oxygen partial pressure, - low volatility at operating temperatures, thermal and chemical compatibility with other cell components, low cost of material and fabrication. The precise perovskite composition may be tailored for a specific application. To obtain a high performance membrane, however, many technical and material problems remain to be solved. This final section will focus on several issues, which are not yet well understood, but are thought to be of importance for further development of the membrane devices. In the first place our understanding of factors that control and limit the interfacial kinetics is still rudimentary, and therefore should be a fruitful area for further investigation. The apparent correlation between the surface oxygen exchange coefficient ks and the tracer diffusion coefficient D* for different classes of oxides, the fluorite-related and the perovskite-related oxides, as noted by Kilner et al. [73], clearly indicate the potential of isotopic 180--160 exchange. However, a problem remains how to relate the observations (at equilibrium) from isotopic exchange to the conditions met during membrane operation. In chemical relaxation experiments, the oxide is studied after perturbation of the equilibrium state. These methods are thus complementary and probably their combined application, whenever possible together with spectroscopic techniques, such as FT-IR, UV and EPR, has a great capacity to elucidate the kinetics of surface oxygen exchange. Though, at first glance, the limited exchange capability of the perovskites, relative to diffusion, puts limits on attempts to improve the oxygen fluxes or to lower the operating temperatures by making thinner membranes, it is expected that the surface exchange kinetics can be significantly improved by surface modification. One approach is coating with a porous surface layer which will effectively enlarge the surface area available to exchange, as discussed in Section 10.3.2.3. Improvements can also be expected by finely dispersing precious metals or other exchange active second phases on the oxide surface. It is clear that further investigations are required to evaluate these innovative approaches.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
511
As yet, more work is also required to gain insight in the role of the ceramic microstructure in the performance values of membranes, and to evaluate different processing routes for the fabrication of perovskite thin films. Besides the technological challenge of fabrication of dense and crack-free thin perovskite films, which need to be supported if its thickness is less than about 150 ~tm, a number of other problems relate to the long-term stability of perovskite membranes, including segregation, a low volatility of lattice components, etc. Some of these problems are linked to the imposed oxygen pressure gradient across the membrane. Aside from the lattice expansion mismatch of opposite sides of the membrane, attention is drawn to the potential problem of demixing, which arises in almost all situations where a multicomponent oxide is brought into a gradient of oxygen chemical potential. The available theories predict that, if the mobilities of the cations are different and non-negligible at high temperatures, concentration gradients appear in the oxide in such a way that the high oxygen pressure side of the membrane tends to be enriched with the faster moving cation species. Depending on the phase diagram, the spatially inhomogeneous oxide may eventually decompose. The latter may cause surprise, if the (homogeneous) oxide is stable in the range of oxygen partial pressures covered by experiment. This is why these processes have been termed kinetic demixing and kinetic decomposition by Schmalzried et al. [250,251] who were the first to study them. Degradation phenomena have been shown to occur in, for example, Col_xMgxO, Fe2SiO4 and NiTiO3. Internal oxidation or reduction processes sometimes lead to precipitation of a second phase in the matrix of the parent phase. Another possible consequence of the demixing process is the morphological instability of the (moving) low pressure interface due to formation of pores, which may eventually penetrate throughout the ceramic. The above phenomena have been the subject of a number of theoretical and experimental studies in the last decade [252-256], to give only a brief number. A review up to 1986 has been written by Schmalzried [257]. To our knowledge, no report has been made up to now of demixing phenomena in mixed oxide ion-electronic conductors. Since they cannot be excluded to occur on the basis of theoretical arguments, this is also why the phenomena deserve (more) attention in order to be able to control deterioration of membrane materials. Intergrowth structures in which perovskite-type blocks or layers are held apart by non-perovskite ones could offer a new strategy for identifying new materials, as was suggested earlier by Goodenough et al. [212]. In such structures, vacancy transport is confined to two-dimensional layers or to sites which link up to form channels extending throughout the crystal. An interesting variation to the BIMEVOX compounds, already discussed in Section 10.4.3.1, is found in derivatives of Sr4Fe6013. Its orthorhombic structure can be described as built of perovskite layers alternating with sesquioxide Fe203 layers perpendicular to the b-axis. The discovery of high levels of oxygen permeation through mixed
512
10 m DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
metal oxide compositions obtained by partial substitution of iron for cobalt, for instance, SrCo0.sFeOx recently translated into a patent for this class of materials [258]. Tubes made from the given composition showed oxygen fluxes similar to those through known state-of-the-art materials having a perovskite structure, but did not fracture in the process for preparing syngas as was found for some of the perovskite materials. As noted before, the membrane performance could be affected by the presence of H20, CO2 or other volatile hydrocarbons in the gas phase of both compartments. As laid down in patent literature [1-3], the oxygen fluxes through Mg-, Ca-, Sr-, and Ba-doped perovskites deteriorated over time, roughly 30-50% over a time period of about 100 h, if the air used as feed gas contained several percent of H20 and amounts of CO2 on a hundreds of ppm level. It was claimed, that either no deterioration is found or the fluxes can be restored to their initial values if the temperature is raised above certain critical values, 500~ for magnesium, 600~ for calcium, 700~ for strontium and 810~ for barium. Though no explanation was given, it is possible that carbonate formation took place. One may further note that the tendency for carbonate formation increases at lower temperatures. A surprising observation was recently made in the author's laboratory in a study of oxygen permeation through Lal_xSrxFeO34 (0.1 < x < 0.4) [151]. Long times to reach steady-state oxygen permeation at 1000~ extending over hundreds of hours were observed, yet could be avoided by exposing the permeate side surface of the membrane for a 1-2 h to 1:1 CO/CO2 gas mixture. A clear explanation cannot yet be given for this observation, which is still under investigation, though a reconstruction of the surface by the reducing ambient cannot be excluded. The oxygen permeability measured if helium was used again as the sweeping gas on this side of the membrane, was found to be limited by diffusional transport of oxygen across the membrane [151]. A similar type of observation was made by Miura et al. [152], who noticed the oxygen flux through slib-casted membranes of La0.6Sr0.4Co0.sFe0.203,stobe greatly improved if these were freed from surface impurities, like SrO, following an acid treatment. One final point to note is the ability of acceptor-doped perovskite oxides to incorporate water, and some contribution of proton conduction therefore cannot be excluded. If water insertion occurs at low temperature, this might lead to residual stresses in the ceramics. Besides water may play an active role in the surface oxygen exchange. For example, on Bi2MoO6, which has an intergrowth structure consisting of Bi2O2+ blocks alternating with MO42- layers of cornershared MO 6 octahedra, exchange with 1802-enriched oxygen could not be observed experimentally [259]. On the other hand, Novokova and Jiru [260] demonstrated that exchange of water with lattice oxygen on an industrial bismuth molybdate catalyst proceeds rapidly at 200~ and is even measurable at room temperature.
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
513
Acknowledgements The authors are indebted to colleagues H. Kruidhof, R.H.E. van Doorn, J.E. ten Elshof, M.H.R. Lankhorst and B.A. van Hassel for many useful discussions and for providing experimental data. Paul Gellings and Henk Verweij are gratefully acknowledged for valuable comments and careful reading of the manuscript. The Commission of the European Communities and the Netherlands Foundation for Chemical Research (SON) are thanked for financial support.
List of Abbreviations and Symbols
Abbreviations: BE25 25 mole% erbia-stabilized bismuth oxide BICUVOX Bi4V2-yCUyOll BIMEVOX general acronym for materials derived from Bi4V2O11, like BICUVOX BT40 40 mole% terbia-stabilized bismuth oxide BY25 25 mole% yttria-stabilized bismuth oxide CSZ calcia-stabilized zirconia ECVD electrochemical vapour deposition EDS energy dispersive spectroscopy (of X-rays) EPR electron proton resonance electromotive force emf fourier transform infrared spectroscopy FT-IR high resolution transmission electron microscopy HRTEM mixed ionic-electronic conductor MIEC nuclear magnetic resonance NMR scanning electron microscopy SEM secondary ion mass spectroscopy SIMS solid oxide fuel cell SOFC transmission electron microscopy TEM three phase boundary tpb temperature programmed reduction TPR ultra-violet spectroscopy UV X-ray absorption near edge structure XANES X-ray absorption spectroscopy XAS X-ray diffraction XRD yttria-stabilized zirconia YSZ
Symbols: Ci N
D D*
mole fraction or concentration of species i chemical diffusion coefficient tracer diffusion coefficient
514
Ds Dv dp e
E Eeq
F G
I
ji .0
Jex
ks k K L Lc Ld Lp n
P Po2 PO2' Po2"
R Si o si
S t tel
ti tion T Tt ui o bli
Vm zi
10- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
self-diffusion coefficient vacancy diffusion coefficient pore diameter elementary charge
emf emf at equilibrium Faraday constant geometric factor used to account for non-axial contributions to the oxygen flux Haven ratio electrical current flux of species i balanced surface exchange rate atequilibrium, mol 02 c m -2 s -1 surface exchange coefficient, cm s reaction rate constant equilibrium constant for a reaction membrane thickness characteristic thickness of membrane Debye-Hiickel screening length characteristic thickness (active width) of porous coating layer frequently used to designate the mole fraction of electrons, yet its use is multipurpose mole fraction of electron holes oxygen partial pressure oxygen partial pressure at feed side of the membrane oxygen partial pressure at permeate side of the membrane radius of species i gas constant entropy of species i entropy of species i at standard state surface area Goldschmidt factor electronic transference number transference number of species i ionic transference number temperature transition temperature electrical mobility of species i electrical mobility of species i in standard state molar volume charge number of species i (positive for cations and negative for anions)
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
Greek: (x
8
1~i 0 o
(~el (3"h o
(~ion (~n (3"p (~total Xs
515
surface exchange coefficient bulk diffusion coefficient reduction factor deviation from ideal oxygen stoichiometry e n h a n c e m e n t factor overpotential electrochemical potential of species i porosity chemical potential of species i s t a n d a r d chemical potential of species i electronic conductivity polaron h o p p i n g conductivity electrical conductivity of species i conductivity of species i at s t a n d a r d state ionic conductivity n-type electronic conductivity p-type electronic conductivity total conductivity tortuosity electric potential of phase (Galvani potential) critical (percolation threshold) v o l u m e fraction
REFERENCES
1.
2.
3.
4. 5. 6. 7. 8.
M.F.Carolan, P.N. Dyer, J.M. LaBar Sr. and R.M. Thorogood, Process for recovering oxygen from gaseous mixtures containing water or carbon dioxide which process employs ion transport membranes. US Patent 5,261,932, 1993. M.F. Carolan, P.N. Dyer, S.M. Fine, J.M. LaBar Sr. and R.M. Thorogood, Process for recovering oxygen from gaseous mixtures containing water or carbon dioxide which process employs barium-containing ion transport membranes. US Patent 5,269,822, 1993. M.F. Carolan, P.N. Dyer, J.M. LaBar Sr. and R.M. Thorogood, Process for restoring permeance of an oxygen-permeable ion transport membrane utilized to recover oxygen from oxygen-containing gaseous mixtures. US Patent 5,240,473, 1993. M. Liu, A.V. Joshi, Y. Shen and K. Krist, Mixed ionic-electronic conductors for oxygen separation and electrocatalysis. US Patent 5,273,628,1993. R.M.Thorogood, Developments in air separation. Gas Sep. Purif., 5 (1991) 83-94. E.A.Hazbun, Ceramic membrane for hydrocarbon conversion. U.S Patent 4,791,079,1988. E.A. Hazbun, Ceramic membrane and use thereof for hydrocarbon conversion. U.S. Patent 4,827,071, 1989. W. Wang and Y.S. Lin, Analysis of oxidative coupling in dense oxide membrane reactors, J. Membr. Sci., 103 (1995) 219-233.
516
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
9. J.E. ten Elshof, B.A. Van Hassel and H.J.M. Bouwmeester, Activation of methane using solid oxide membranes. Catal. Today, 25 (1995) 397--402. 10. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, Oxidative coupling of methane in a mixed-conducting perovskite membrane reactor. Appl. Catal. A: General, 130 (1995) 195-212. 11. T. Nozaki and K. Fujimoto, Oxide ion transport for selective oxidative coupling of methane with new membrane reactor. AIChE J., 40 (1994) 870-877. 12. C.-Y. Tsai, Y.H. Ma, W.R. Moser and A.G. Dixon, Simulation of nonisothermal catalytic membrane reactor for methane partial oxidation to syngas, in: Y.H. Ma (Ed.), Proceedings of the 3rd International Conference on Inorganic Membranes, Worcester, 1994, pp. 271-280. 13. S. Pei, M.S. Kleefisch, T.P. Kobylinski, K. Faber, C.A. Udovich, V. Zhang-McCoy, B. Dabrowski, U. Balachandran, R.L. Mieville and R.B. Poeppel, Failure mechanisms of ceramic membrane reactors in partial oxidation of methane to synthesis gas. Catal. Lett., 30 (1995) 210-212. 14. U. Balachandran, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L. Mieville, P.S. Maiya, M.S. Kleefisch, S. Pei, T.P. Kobylinski, C.A. Udovich and A.C. Bose, Methane to synthesis gas via ceramic membranes. Am. Ceram. Soc. Bull., 74(1) (1995) 71-75. 15. A.G. Dixon, W.R. Moser and Y.H. Ma, Waste reduction and recovery using O2-permeable membrane reactors. Ind. Eng. Chem. Res., 33 (1994) 3015-3024. 16. H.P. Hsieh, Inorganic membrane reactors. Cat. Rev.-Sci. Eng., 33(1/2) (1992) 1-70. 17. W.R. Gurr, An operators view on gas membranes, in: M.K. Turner (Ed.), Effective Industrial Membrane Processes m BeneJits and Opportunities. Elsevier, London, 1991, p. 329. 18. V.T. Zaspalis and A.J. Burggraaf, Inorganic membrane reactors to enhance the productivity of chemical processes, in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991, pp. 177-207. 19. T.J. Mazanec, Prospects for ceramic electrochemical reactors in industry. Solid State Ionics, 70/71 (1994) 11-19. 20. G. Saracco and V. Specchia, Catalytic inorganic membrane reactors: Present experience and future opportunities. Catal. Rev.-Sci. Eng., 36(2) (1994) 303-384. 21. P.J. Gellings and H.J.M. Bouwmeester, Ion and mixed-conducting oxides as catalysts. Catal. Today, 1 (1992) 1-101. 22. D. Hayes, D.W. Budworth and J.P. Roberts, Selective permeation of gases through dense sintered alumina. Trans. Br. Ceram. Soc., 60 (1961) 494-504. 23. D. Hayes, D.W. Budworth and J.P Roberts, Permeability of sintered alumina materials to gases at high temperatures. Trans. Br. Ceram. Soc., 62 (1963) 507-523. 24. H.L. Tuller, Mixed conduction in nonstoichiometric oxides, in: O. Sorenson (Ed.), Nonstoichiometric Oxides. Academic Press, New York 1981, pp. 271-337. 25. J.W. Suitor, D.J. Clark and R.W. Losey, Development of alternative oxygen production source using a zirconia solid electrolyte membrane, in: Technical progress report for fiscal years 1987, 1988 and 1990. Jet Propulsion Laboratory Internal Document D7790, 1990. 26. T.J. Mazanec, T.L. Cable and J.G. Frye, Jr., Electrocatalytic cells for chemical reaction. Solid State Ionics, 53-56 (1992) 111-118. 27. M. Kleitz and M. Siebert, Electrode reactions in potentiometric gas sensors, in: T. Seiyama (Ed.), Chemical Sensor Technology, Vol. 2. Elsevier, Amsterdam, 1989, pp. 151-71. 28. W.C. Maskell and B.C.H. Steele, Solid state potentiometric sensors. J. Appl. Electrochem., 16 (1986) 475-489.
10 -- DENSECERAMICMEMBRANESFOROXYGENSEPARATION 29. 30.
31. 32. 33. 34. 35. 36. 37. 38. 39.
40. 41.
42. 43. 44. 45. 46. 47.
48.
49. 50. 51. 52.
517
S.S. Liou and W.L. Worrell, Electrical properties of novel mixed-conducting oxides. Appl. Phys. A., 49 (1989) 25-31. S.S. Liou and W.L. Worrell, Mixed conducting electrodes for solid oxide fuel cells, in: S.C. Singhal (Ed.), Proceedings of the 1st International Symposium on Solid Oxide Fuel Cells. The Electrochem. Society, Pennington, NJ, 1989, pp. 81-89. B. Cal6s and J.F. Baumard, Electrical properties of the ternary solid solutions (ZrO21-xCeO2 x)0.9-Y203 0.1 (0 < x _<1). Rev. Int. hautes Tempdr. Rdfract., Fr., 17 (1980) 137-147. B. Cal6s and J.F. Baumard, Mixed conduction and defect structure of ZrO2-CeO2-Y203 solid solutions. J. Electrochem. Soc., 131 (1984) 2407-2413. Y. Nigara and B. Cales, Production of carbon monoxide by direct themal splitting of carbon dioxide at high temperature. Bull. Chem. Soc. Jpn., 59 (1986) 1997-2002. T. Takahashi and H. Iwahara, Denki Kagaku, 35 (1967) 433. T. Takahashi and H. Iwahara, Energy Conversion, 11 (1971) 105. B.C.H. Steele, Oxygen ion conductors and their technological applications. Mat. Sci. Eng., B13 (1992) 79-87. Y. Teraoka, H.M. Zhang, S. Furukawa and N. Yamazoe, Oxygen permation through perovskite-type oxides. Chem. Lett., (1985) 1743-1746. Y. Teraoka, T. Nobunaga and N. Yamazoe, Effect of cation substitution on the oxygen semipermeability of perovskite oxides. Chem. Lett. (1988) 503-506. Y. Teraoka, T. Nobunaga, K. Okamoto, N. Miura and N. Yamazoe, Influence of constituent metal cations in substituted LaCoO3 on mixed conductivity and oxygen permeability. Solid State Ionics, 48 (1991) 207-212. Y. Teraoka, H.M. Zhang, K. Okamoto and N. Yamazoe, Mixed ionic--electronic conductivity of Lal-xSrxCol-yFeyO3-8. Mat. Res. Bull., 23 (1988) 51-58. H.J.M. Bouwmeester, H. Kruidhof and A.J. Burggraaf, Importance of the surface exchange kinetics as rate limiting step in oxygen permeation through mixed-conducting oxides. Solid State Ionics, 72 (1994) 185-194. R.M. Thorogood, R. Srinivasan, T.F. Yee and M.P. Drake, Composite mixed conductor membranes for producing oxygen; US Patent 5,240,480,1993. H. Deng, M. Zhou and B. Abeles, Diffusion-reaction in porous mixed ionic-electronic solid oxide membranes. Solid State Ionics, 74 (1994) 75-84. H. Deng, M. Zhou and B. Abeles, Transport in solid oxide porous electrodes: Effect of gas diffusion. Solid State Ionics, 80 (1995) 213-222. H. Schmalzried, Solid State Reactions. Verlag Chemie, Weinheim, 1981. H. Rickert, Electrochemistry of Solids, An Introduction. Springer-Verlag, Berlin, Heidelberg, New York, 1992. L. Heyne, Electrochemistry of mixed-ionic electronic conductors, in: S. Geller (Ed.), Solid Electrolytes, Topics in Applied Physics. Springer, Berlin, Heidelberg, New York 1977, pp. 167-221. F.A. Kr6ger and H.J. Vink, Relations between the concentrations of imperfections in crystalline solids, in: F. Seitz and D. Turnbull (Eds.), Solid State Physics, Vol. 3. Academic Press, New York, 1956, p. 307. C. Wagner and W. Schottky, Beitrag zur Theorie des Anlaufvorganges. Z. Phys. Chem., Bll (1930) 25-41. C. Wagner, Equations for transport in solid oxides and sulfides of transition metals. Prog. Solid State Chem., 10(1) (1975) 3--16. G.E. Murch, The Haven ratio in fast ionic conductors. Solid State Ionics, 7 (1982) 177-198. J. Maier and G. Schwitzgebel, Theoretical treatment of the diffusion coupled with
518
53.
54. 55.
56. 57. 58.
59. 60. 61.
62.
63. 64.
65. 66. 67.
68.
69.
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION reaction, applied to the example of a binary solid compounds MX. Phys. Stat. Sol. (B), 113 (1982) 535-547. J. Maier, Mass transport in the presence of internal defect reactions-concept of conservative ensembles: I, Chemical diffusion in pure compounds. J. Am. Ceram. Soc., 76(5) (1993) 1212-1217; II, Evaluation of electrochemical transport measurements, ibid., 1218-1222; III, Trapping effect of dopants on chemical diffusion, ibid., 1223-1227; IV, Tracer diffusion and intercorrelation with chemical diffusion and ion conductivity, ibid., 12281232. J. Maier, Diffusion in materials with ionic and electronic disorder. Mat. Res. Soc. Syrup. Proc. Vol. 210, 1991, pp. 499-509. M. Liu, Theoretical assessment of oxygen separation rates of mixed conductors, in: T.A. Ramanarayanan and H.L. Tuller (Eds.), Ionic and Mixed Conducting Oxide Ceramics 91-12. Electrochemical Society Inc., NJ, 1991, pp. 95-109. Yu.V. Pleskov and Yu.Ya. Gurevich, Semiconductor Photoelectrochemistry. Consultants Bureau, New York, 1986, Chapter 4. M.J. Dignam, D.J. Young and D.G.W. Goad, Metal oxidation, I. Ionic transport equation. J. Phys. Chem. Solids, 34 (1973) 1227-1234. D.J. Young and M.J. Dignam, Metal oxidation, II. Kinetics in the thin and very thin flim regions under conditions of electron equilibrium. ]. Phys. Chem. Solids, 34 (1973) 12351250. J. Maier, Kr6ger-Vink diagrams for boundary regions. Solid State Ionics, 32/33 (1989) 727-733. J. Nowotny, Surface segregation of defects in oxide ceramics materials. Solid State Ionics, 28-30 (1988) 1235-1243. Z. Adamczyck and J. Nowotny, Effect of the surface on gas/solid equilibration kinetics in non-stoichiometric compounds, in: J. Nowotny (Ed.), Diffusion and Defect Data, Solid State Data, Part B, Vols. 15 & 16, 1991, pp. 285-336. B.C.H. Steele, Isotopic oxygen exchange and optimisation of SOFC cathode materials, in: F.W. Poulsen, J.J. Bentzen, T. Jacobsen, E. Skou and M.J.L. f~sterg/ird (Eds.), High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Riso National Laboratory, Denmark, 1993, pp. 423--430. B.C.H. Steele, Interfacial reactions associated with ceramic ion transport membranes. Solid State ionics, 75 (1995) 157-165. M. Kleitz and T. Kloidt, Dessemondt, Conventional oxygen electrode reaction: facts and models, in: F.W. Poulsen, J.J. Bentzen, T. Jacobsen, E. Skou, M.J.L. f~sterg/ird (Eds.), High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Riso National Laboratory, Denmark, 1993, pp. 89-116. K. Ando, Y. Oishi, H. Koizama and Y. Sakka, Lattice defect and oxygen diffusivity in MgO stabilized ZrO2. J. Mat. Sc. Lett., 4 (1985) 176-180. Y. Oishi and W.D. Kingery, Self-diffusion of oxygen in single crystal and polycrystalline aluminum oxide. ]. Chem. Phys., 33 (1960) 480-486. B.A. Boukamp, K.J. de Vries, A.J. Burggraaf, Surface oxygen exchange in bismuth oxide based materials, in: J. Nowotny, W. Weppner (Eds.), Non-Stoichiometric Compounds, Surfaces, Grain Boundaries and Structural Defects, Kluwer, Dordrecht, 1989, pp. 299-309. B.A. Boukamp, I.C. Vinke, K.J. de Vries and A.J. Burggraaf, Surface oxygen exchange properties of bismuth oxide-based solid electrolytes and electrode materials. Solid State Ionics, 32/33 (1989) 918--923. E.Kh. Kurumchin and M.V. Perfiliev, An isotope exchange study of the behaviour of
10- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
519
electrochemical systems. Solid State Ionics, 42 (1990) 129-133. 70. K. Park and D.R. Olander, Oxygen diffusion in single-crystal tetragonal zirconia. J. Electrochem. Soc., 138(4) (1991) 1154-1159. 71. J. Kilner, B.C.H. Steele and L. Ilkov, Oxygen self-diffusion studies using negative-ion secondary ion mass spectroscopy (SIMS). Solid State Ionics, 12 (1984) 89-97. 72. R.J. Tarento and C. Monty, Influence of nonstoichiometry on the oxygen self-diffusion in Col-xO single crystals. Solid State Ionics, 28-30 (1988) 1221-1229. 73. J. Kilner, Isotopic exchange in mixed and ionically conducting oxides, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics, Vol. 94-12. The Electrochemical Society, Pennington, NJ, 1994, pp. 174-190. 74. S. Yamauchi, T. Ishigaki, J. Mizusaki and K. Fueki, Tracer diffsuion coeffcient of oxide ions in LaCoO3. Solid State Ionics, 9&10 (1983) 997-1003. 75. T. Ishigaki, S. Yamauchi, K. Kishio, J. Mizusaki and K. Fueki, Diffsuion of oxide ion vacancies in perovskite-type oxides. J. Solid State Chem., 73 (1988) 179-187. 76. L.A. Simpson and R.E. Carter, Oxygen exchange and diffusion in calcia-stabilized zirconia. J. Am. Ceram. Soc., 49(3) (1966) 139-144. 77. P.E. Childs and J.B. Wagner Jr., Chemical diffusion in wfistite and chromium-doped manganous oxide, in: G.R. Beltane and W.L. Worrell (Eds.), Heterogeneous Kinetics at Elevated Temperatures. Plenum Press, New York, 1970, pp. 269-342. 78. P.E. Childs, L.W. Laub and J.B. Wagner Jr., Chemical diffusion in nonstoichiometric oxides. Proc. Br. Ceram. Soc., 19 (1973) 29-53. 79. F. Morin, Mesure du coefficient de diffusion chimique dans les oxydes par la methode de relaxation. Can. Metall. Q., 14(2) (1975) 97-104. 80. E.C. Subbarao and H.S. Maiti, Solid eletrolytes with oxygen ion conduction. Solid State Ionics, 11 (1984) 317-338. 81. J. Kilner and B.C.H. Steele, Mass transport in anion-deficient fluorite oxides, in: O. Sorenson (Ed.), Nonstoichiometric Oxides. Academic Press, New York, 1981, pp. 233-269. 82. H.L. Tuller and A.S. Nowick, Doped ceria as a solid oxide electrolyte. J. Electrochem. Soc., 122 (1975) 255-259. 83. T.H. Etsell and S.N. Flengas, The electrical properties of solid oxide electrolytes. Chem. Rev., 70(3) (1970) 339-376. 84. W.L. Worrell, Oxide solid electrolytes, in: S. Geller (Ed.), Solid Electrolytes. Topics in Appl. Phys., 21. Springer Verlag, Berlin, 1971, pp. 143-168. 85. J.-H. Park and R.N. Blumenthal, Electronic transport in 8 mole percent Y203-ZrO2. J. Electrochem. Soc., 136(10) (1989) 2867-2876. 86. M. Iwase and T. Mori, Oxygen permeability of calcia-stabilized zirconia. Met. Trans. B, 9 (1978) 365-370. 87. A.W. Smith, F.W. Meszaros and C.D. Amata, Permeability of zirconia, hafnia, and thoria to oxygen. J. Am. Ceram. Soc., 49(5) (1966) 240-244. 88. W.A. Fisher, Permeation of oxygen and partial electronic conductivities in solid oxide electrolytes, in: W. van Gool (Ed.), Fast Ion Transport in Solids. Elsevier, New York, 1973, pp. 503-512. 89. S.F. Palguev, V.K. Gilderman and A.D. Neujmin, Oxygen permeability of stabilized zirconia solid electrolytes. J. Electrochem. Soc., 122(6) (1975) 745-748. 90. R. Hartung, H.-H. MObius, Bestimmung der Oxydationshalbleitung van ZrO2-Festelektrolyten aus Permabilit/itsdaten. Z. Phys. Chem., 243 (1970) 133-138. 91. L. Heyne and N.M. Beekmans, Electronic transport in calcia-stabilized zirconia. Proc. Br.
520
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
Ceram. Soc., 19 (1971) 229-262. 92. C.B. Alcock and J.C. Chan, The oxygen permeability of stabilized zirconia electrolytes at high temperatures. Can. Metall. Quart., 11(4) (1972) 559-567. 93. K. Kitazawa and R.L. Coble, Use of stabilized ZrO2 to measure 02 permeation. J. Am. Ceram. Soc., 57 (1974) 360-363. 94. B. Cales and J.F. Baumard, Oxygen semipermeability and electronic conductivity in calcia-stabilized zirconia. J. Mat. Sci., 12 (1982) 3243--3248. 95. S. Dou, C.R. Masson and P.D. Pacey, Mechanism of oxygen permeation through lime-stabilized zirconia. J. Electrochem. Soc., 132 (1985) 1843-1849. 96. H.J.M. Bouwmeester, H. Kruidhof, A.J. Burggraaf and P.J. Gellings, Oxygen semipermeability of erbia-stabilized bismuth oxide. Solid State Ionics, 53-56 (1992) 460-468. 97. H. Ullmann, Die Sauerstoffpermeabilit/it von oxidischen Festelektrolyten. Z. Phys. Chemie, 237 (1968) 71-80. 98. T. Reetz, Sauerstoffpermeabilit/it oxidionenleitender Festelektrolyte. Z. Phys. Chemie, 249 (1972) 369-375. 99. J. Fouletier, P. Fabry and M. Kleitz, Electrochemical semipermeability and the elctrode microsystem in solid oxide electrolyte cells. J. Electrochern. Soc., 123(2) (1976) 204-213. 100. S. Dou, C.R. Masson and P.D. Pacey, Chemical diffusion in calcia-stabilized zirconia ceramic. Solid State Ionics, 18/19 (1986) 736-740. 101. S. Dou, C.R. Masson and P.D. Pacey, Induction period for oxygen permeation through calcia-stabilized zirconia ceramic. J. Am. Ceram. Soc., 72(7) (1989) 1114-1118. 102. S. Dou, C.R. Masson and P.D. Pacey, Diffusion of electronic defects in calcia-stabilized zirconia from a deorpfion rate study. J. Electrochem. Soc., 137(11) (1990) 3455-3458. 103. M.J. Verkerk, K. Keizer and A.J. BurggraaL High oxygen ion conduction in sintered oxides of the Bi203-Er203. J. Appl. Electrochem., 10 (1980) 81-90. 104. B.A. Boukamp, I.C. Vinke, K.J. de Vries and A.J. Burggraaf, Surface oxygen exchange properties of bismuth oxided based solid electrolytes and electrode materials. Solid State Ionics, 32/33 (1989) 918--923. 105. D.J. Clark R.W. Losey and J.W. Suitor, Separation of oxygen by using zirconia solid electrolyte membranes. Gas Sep. Purif., 6(4) (1992) 201-205. 106. M. Dum61i6, G. Nowogrocki and J.C. Boivin, Ionic conductor membrane for oxygen separation. Solid State Ionics, 28-30 (1988) 524-528. 107. F. Abraham, M.F. Debreuille-Gresse, G. Mairesse and G. Nowogrocki, Phase transitions and ionic conductivity in Bi4V2Oll an oxide with a layered structure. Solid State ionics, 28-30 (1988) 529-532. 108. F. Abraham, J.C. Boivin, G. Mairesse and G. Nowogrocki, The BIMEVOX series: A new family of high perfomance oxide ion conductors. Solid State Ionics, 40/41 (1990) 934-937. 109. J.C. Boivin, R.N. Vannier, G. Mairesse, F. Abraham and G. Nowogrocki, BIMEVOX: A new family of Bi203 solid electrolytes for the separation of oxygen. ISSI Lett., 3(4) (1992) 14-16. 110. Y.S. Shen, M. Liu, D. Taylor, S. Bolagopal, A. Joshi and K. Krist, Mixed ionic-electronic conductors based on Bi-Y-O-Ag metal-ceramic system, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings of 2nd International Symposium on Ionic and Mixed Conducting Ceramics, Vol. 94-12. The Electrochemical Society, Pennington, NJ, 1994, pp. 574-595. 111. C.S. Chen, H.J.M. Bouwmeester, H. Verweij and A.J. Burggraaf, Oxygen permeation through stabilized bismuth oxide-noble metal composites. Solid State Ionics, accepted. 112. C.S. Chen, Fine grained zirconia-metal dual phase composites, PhD Thesis, University
10- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
521
of Twente, The Netherlands, 1994. 113. D.S. McLachlan, M. Blaszkiewics and R.E. Newnham, Electrical resistivity of composites. J. Am. Ceram. Soc., 73(8) (1990) 2187-2203. 114. C.S. Chen, B.A. Boukamp, H.J.M. Bouwmeester, G.Z. Cao, H. Kruidhof, A.J.A. Winnubst and A.J. Burggraaf, Microstructural development, electrical properties and oxygen permeation of zirconia-pallladium composites. Solid State Ionics, 76 (1995) 23-28. 115. S. Kirkpatrick, Percolation and conduction. Rev. Mod. Phys., 45(4) (1973) 574-588. 116. H. Tannenberger and H. Siegert, The behaviour of silver cathodes in solid electrolyte fuel cells. Adv. Chem., 90 (1969) 281-300. 117. Y.S. Shen, A. Joshi, M. Liu and K. Krist, Structure, microstructure and transport properties of mixed ionic-electronic conductors based on bismuth oxide. Part I. Bi-Y-Cu-O system. Solid State Ionics, 72 (1994) 209-217. 118. S. Ling, M.P. Anderson and T.A. Ramanarayanan, Optimization of ionic transport through mixed conducting oxide ceramics. Solid State Ionics, 59 (1993) 33-45. 119. R.M.C. Marques, F.M.B. Marques and J.R. Frade, Characterization of mixed condcutors by dc techniques. Part I: Theoretical solutions. Solid State Ionics, 73 (1994) 15-25. 120. H. Arashi, H. Naito and M. Nakata, Electrical properties in the ZrO2-CeO2-Y203 system. Solid State Ionics, 76 (1995) 315-319. 121. R.M.C. Marques, F.M.B. Marques and J.R. Frade, Characterization of mixed conductors by dc techniques. Part II: Experimental results. Solid State Ionics, 73 (1994) 27-34. 122. T. Ramanarayanan, S. Ling, M.P. Anderson and A. Ozeckni, Effect of microstructure on ionic and electronic conduction in the ZrO2-Y203-CeO2 system, in: F.W. Poulson, J.J. Bentzen, T. Jacobsen, E. Skou and M.L. f~stergArd (Eds.), Proceedings of the 14th Riso International Symposium on Materials Science. High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Riso National Laborary, Roskilde, Denmark, 1993, pp. 381-388. 123. M.T. Colomer, J.R. Jurado, R.M.C. Marques and F.M.B.Marques, Evaluation of titania doped YSZ for SOFC anodes, in: S.C. Singhal and H. Iwahara (Eds.), Proceedings of the 3rd International Symposium on Solid Oxide Fuel Cells, The Electrochem. Society, Pennington, NJ, 1989, pp. 523-532. 124. R.M.C. Marques, J.R. Frade and F.M.B. Marques, Ceramic materials for SOFC anode cermets, in: S.C. Singhal and H. Iwahara (Eds.), Proceedings of the 3rd International Symposium on Solid Oxide Fuel Cells. The Electrochemical Society, Pennington, NJ, 1989, pp. 513-522. 125. T. Lindegaard, C. Clausen and M. Mogenson, Electrical and electrochemical properties of Zr0.77Y0.13Ti0.101.93, in: F.W. Poulson, J.J. Bentzen, T. Jacobsen, E. Skou and M.L. f~stergArd (Eds.), Proceedings of the 14th Risa International Symposium on Materials Science. High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Riso National Laboratory, Roskilde, Denmark, 1993, pp. 311-318. 126. K.E. Swider, S. Loughin and W.L. Worrell, in: T.A. Ramanarayanan and H.L. Tuller (Eds.), Proceedings of the First International Symposium on Ionic and Mixed Cond. Ceramics. The Electrochemical Society, Pennington, NJ, 1991, pp. 73-83. 127. H. Arashi and H. Naito, Oxygen permeability in ZrO2-TiO2-Y203 system. Solid State Ionics, 53-56 (1992) 431-435. 128. H. Naito and H. Arashi, Hydrogen production from direct water splitting at high temperatures using a ZrO2-TiO2-Y203 membrane. Solid State Ionics, 79 (1995) 366-370. 129. H. Iwahara, T. Esaka and K. Takeda, Mixed conduction and oxygen permeation in sintered oxides of a system ZrO2-Tb407, in: S. Somiya, N. Yamamoto and H. Yanegida
522
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
(Eds.), Advances in Ceramics, Vol. 24A. The American Ceramic Society, Inc., Westerville, OH, 1988, pp. 907-916. 130. H. Arashi, S. Shin, H. Miura, A. Nakshima, M. Ishigame and O. Shimomura, Investigations of valence change of Tb-ions in ZrO2-Tb407 mixed conductors using XANES measurements, Solid State Ionics, 35 (1989) 323-327. 131. G.Z. Cao, X.Q. Liu, H.W. Brinkman, K.J. de Vries and A.J. Burggraaf, Mixed conduction and oxygen permeation of ZrO2-Tb203.5-Y203 solid solutions, in: S.P.S. Badwal et al. (Eds.), Science and Technology ofZirconia V. Technomic Publications, Lancaster, 1993, pp. 576-583. 132. G.Z. Cao, Electrical conductivity and oxygen semipermeability of terbia and yttria stabilized zirconia. J. Appl. Electrochem., 24 (1994) 1222-1224. 133. G.Z. Cao, J. Meijerink, H.W. Brinkman, K.J. de Vries and A.J. Burggraaf, Growth of thin dense gas tight (Tb,Y)-ZrO2 films by electrochemical vapour deposition. J. Mater. Chem., 3(7) (1993) 773-774. 134. H.W. Brinkman and A.J. Burggraaf, Ceramic membranes by electrochemical vapour deposition of zirconia-yttria-terbia layers on porous substrates. J. Electrochem. Soc., 142(11) (1995) 3851-3858. 135. I.C. Vinke, B.A. Boukamp, K.J. de Vries and A.J. Burggraaf, Mixed conductivity in terbia-stabilized bismuth oxide. Solid State Ionics, 57 (1992) 91-98. 136. H. Kruidhof and H.J.M. Bouwmeester, unpublished results. 137. A. Hammou, Solid oxide fuel cells, in: H. Gerischer and C.W. Tobias (Eds.), Advances in Electrochemical Science and Engineering, Vol. 2. VCH Verlaggesellschaft mbH, Weinheim, Germany, 1991, pp. 87-139. 138. N.Q. Minh, Ceramic fuel cells. J. Am. Ceram. Soc., 76(3) (1993) 564-588. 139. S. Carter, A. Selcuk, J. Chater, R.J. Kajda, J.A. Kilner and B.C.H. Steele, Oxygen transport in selected non-stoichiometric perovskite-structure oxides. Solid State Ionics, 53-56 (1992) 597-605. 140. I. Yasuda and M. Hishinuma, Electrical conductivity and chemical diffusion coefficient of Sr-doped lanthanum manganites, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics. The Electrochemical Society, Pennington, NJ, 1994, pp. 209-221. 141. W.L. Worrell, Electrical properties of mixed-conducting oxides having high oxygen-ion conductivity. Solid State Ionics, 52 (1992) 147-51. 142. W.L. Worrell, P. Han and J. Huang, Mixed conductivity in oxygen exhibiting both p-type electronic and high oxygen-ion conductivity, in: F.W. Poulsen, J.J. Bentzen, T. Jacobson, E. Skou and M.J.L. Osterg/ird (Eds.), High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Riso National Laboratory, Denmark, 1993, pp. 461-466. 143. H.U. Anderson, C.C. Chen, L.W. Tai and M.M. Nasrallah, Electrical conductivity and defect structure of (La,Sr)(Co,Fe)O3, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics. The Electrochemical Society, Pennington, NJ, 1994, pp. 376-387. 144. W.J. Weber, J.W. Stevenson, T.R. Armstrong and L.R. Pederson, Processing and electrochemical properties of mixed conducting La]-xAxCol-yFeyO3-8(A = Sr, Ca), in: G.-A. Nazri, J.-M. Taracson and M.S. Schreiber (Eds.), Materials Research Society Symposium Proceedings, Vol. 369: Solid State Ionics IV. Materials Research Society, Pittsburgh, PA, 1995, pp. 395-400. 145. C.H. Chen, H. Kruidhof, H.J.M. Bouwmeester and A.J. Burggraaf, Ionic conductivity of
10J DENSECERAMICMEMBRANESFOROXYGENSEPARATION
523
perovskite LaCoO3 measured by oxygen permeation technique. J. Appl. Electrochem., accepted for publication. 146. P. Shuk, V. Charton and V. Samochval, Mixed conductors on the lanthanide cobaltites basis. Mat. Sci. Forum, 76 (1991) 161-164. 147. N. Itoh, T. Kato, K. Uchida and K. Haraya, Preparation of pore-free disk of LalxSrxCoO3-~ and its oxygen permeability. J. Membr. Sci., 92 (1994) 239-246. 148. R.H.E. Van Doom, H. Kruidhof, H.J.M. Bouwmeester and A.J. Burggraaf, Oxygen permeability of strontium-doped LaCoO3-a perovskites, in: G.-A. Nazri, J.-M. Taracson and M.S. Schreiber (Eds.), Materials Research Society Symposium Proceedings, Vo1369: Solid State Ionics IV. Materials Research Society, Pittsburgh, PA, 1995, pp. 377-382. 149. R.H.E. Van Doom, H. Kruidhof, H.J.M. Bouwmeester and A.J. Burggraaf, Oxygen permeation through dense ceramic membranes of Sr-doped LaCoO3-5 membranes. To be submitted. 150. B.A. Van Hassel, J.E. ten Elshof and H.J.M. Bouwmeester, Oxygen permeation flux through Lal-ySryFeO3-~ limited by carbon monoxide oxidation rate. Appl. Catal. A: General, 119 (1994) 279-291. 151. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, Oxygen transport through LalxSrxFeO3-5 membranes. I. Permeation in air/He gradients. Solid State Ionics, 81 (1995) 97-109. 152. N. Miura, Y. Okamoto, J. Tamaki, K. Morinag and N. Yamazoe, Oxygen semipermeability of mixed-conductive oxide thick-film prepared by slip casting. Solid State Ionics, 79 (1995) 195-200. 153. L. Qiu, T.H. Lee, L.-M. Lie, Y.L. Yang and A.J. Jacobson, Oxygen permeation studies of SrCo0.8Fe0.203-8. Solid State Ionics 76 (1995) 321-329. 154. H. Kruidhof, H.J.M. Bouwmeester, R.H.E. van Doorn and A.J. Burggraaf, Influence of order-disorder transitions on oxygen permeability through selected nonstoichiometric perovskite-type oxides. Solid State Ionics, 63--65 (1993) 816-822. 155. V.V. Kharton, E.N. Naumovich and A.V. Nikolaev, Oxide ion and electron conjugate diffusion in perovskite-like SrCol-xMxO3-8 (M = Cr..Cu; x = 0..0.5). Solid State Phenomena, 39-40 (1994) 147-152. 156. H.W. Brinkman, H. Kruidhof and A.J. Burggraaf, Mixed conducting yttrium-bariumcobalt-oxide for high oxygen permeation. Solid State Ionics, 68 (1994) 173-176. 157. L. Katz and R. Ward, Structure relations in mixed metal oxides. Inorg. Chem., 3(2) (1964) 205-211. 158. V.M. Goldschmidt, Akad. Oslo., A42 (1946) 224. 159. H.U. Anderson, Review of p-type doped perovskite materials for SOFC and other applications. Solid State Ionics, 52 (1992) 33-41. 160. H.U. Anderson, Defect chemistry of p-type perovskites, in: F.W. Poulsen, J.J. Bentzen, T. Jacobson, E. Skou and M.J.L. Ostergard (Eds.), High Temperature Electrochemical Behaviour of Fast Ion and Mixed Conductors. Rise National Laboratory, Denmark, 1993, pp. 1-19. 161. J. Mizusaki, Nonstoichiometry, diffusion, and electrical properties of perovskite-type oxide electrode materials. Solid State Ionics, 52 (1992) 79-91. 162. E.J.W. Verwey, P.W. Haaijman, F.C. Romeijn and G.W. van Oosterhout, Philips Res. Rep., 5 (1950) 173. 163. J. Mizusaki, T. Sosamoto, W.R. Cannon and H.K. Bowen, Electronic conductivity, seebeck coefficient, and defect structure of LaFeO3. ]. Am. Ceram. Soc., 65(8) (1982) 363-368.
524
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
164. J. Mizusaki, T. Sosamoto, W.R. Cannon and H.K. Bowen, Electronic conductivity, seebeck coefficient, and defect structure of Lal-xFexO3-8 (x = 0.1, 0.25). J. Am. Ceram. Soc., 66(4) (1983) 247-252. 165. J. Mizusaki, M. Yoshihiro, S. Yamauchi and K. Fueki, Nonstoichiometry and defect structure of the perovskite-type oxides Lal-xSrxFeO3-8. J. Solid State Chem., 58 (1985) 257-266. 166. J. Mizusaki, M. Yoshihiro, S. Yamauchi and K. Fueki, Thermodynamic quantitites and defect equilbrium in the perovskite-type oxides solid solution Lal-xSrxFeO3-8. J. Solid State Chem., 67 (1987) 1-8. 167. J. Mizusaki, S. Yamauchi, K. Fueki and A. Ishikawa, Nonstoichiometry of the perovskite-type oxide Lal-xSrxCrO3-5. Solid State Ionics, 12 (1984) 119-124. 168. J. Mizusaki, Y. Mima, S. Yamauchi, K. Fueki and H. Tagawa, Nonstoichiometry of the perovskite-type oxides Lal-xSrxCoO3-5. J. Solid State Chem., 80 (1989) 102-111. 169. H.J.M. Bouwmeester, unpublished results. 170. M. Abbate, F.M.F. de Groot, J.C. Fuggle, A. Fujimori, O. Strebel, F. Lopez, M. Domke, G. Kaindl, G.A. Sawatzky, M. Takana, H. Eisaki and S. Uchida, Controlled valence proerties of Lal-xSrxFeO3 and Lal-xSrxMnO3 studied by soft-x-ray absorption spectroscopy. Phys. Rev. B., 46(8) (1992) 4511-4519. 171. P. Kuiper, G. Kruizinga, J. Ghijsen, M. Grioni, P.J.W. Weijs, F.M.F. de Groot, G.A. Sawatzky, H. Verweij, L.F. Feiner and H. Petersen, X-ray absorption study of the O 2p hole concentration dependence on O stoichiometry in YBa2Cu3Ox. Phys. Rev. B., 38(8) (1988) 6483-6489. 172. M.H.R. Lankhorst and H.J.M. Bouwmeester, Determination of oxygen nonstoichiometry and diffusivity in mixed conducting oxides by oxygen coulometric titration. Part II: Oxygen nonstoichiometry and defect model for La0.8Sr0.2CoO3-5.J. Electrochem. Soc. submitted 173. M. Abbate, J.C. Fuggle, A. Fujimori, L.H. Tjeng, C.T. Chen, R. Potze, G.A. Sawatzky, H. Eisaki and S. Uchida, Electronic structure and spin-state transition in LaCoO3, Phys. Rev. B., 47(24) (1993) 16124-16130. 174. L.G. Tejuca, J.L.G. Fierro and J.M.D. Tasc6n, Structure and reactivity of perovskite-type oxides. Adv. Catal., 36 (1989) 237-228. 175. H. Yokokawa, N. Sakai, T. Kawada and M. Dokya, Thermodynamic stabilities of perovskite-type oxides for electrodes and other electrochemical materials. Solid State Ionics, 52 (1992) 43-56. 176. T. Nakamura, G. Petzow and L.J. Gauckler, Stability of the perovkite phase LaBO3 (B = V, Cr, Mn, Fe, Co, Ni) in reducing atmosphere. I. Experimental results. Mat. Res. Bull., 14 (1979) 649-659. 177. K. Tabata, I. Matsumoto and S. Kohiki, Surface characterization and catalytic proerties of Lal-xSrxCoO3. ]. Mat. Sci., 22 (1987) 1882-86. 178. T. Seiyama, Surface reactivity of oxide materials in oxidation-reduction environment, in: J. Nowotny and L.C. Dufour (Eds.), Surface and Near Surface Chemistry of Oxide Materials. Materials Science Monographs, Vol. 47. Elsevier, Amsterdam, 1988, pp. 189-215 179. N. Yamazoe, S. Furukawa, Y. Teraoka and T. Seiyama, The effect of sorption on the crystal structure of Lal-xSrxCoO3-8. Chem. Lett. (1982) 2019-2022. 180. Y. Teraoka, M. Yoshimatsu, N. Yamazoe and T. Seiyama, Oxygen-sorptive properties and defect structure of perovskite-type oxides, Chem. Left. (1984) 893-896. 181. Y. Teraoka, H.-M. Zhang and N. Yamazoe, Oxygen-sorptive properties of defect perovskite-type Lal-xSrxCol-yFeyO3_8.Chem. Lett. (1985) 1367-1370.
10m DENSEC E ~ C MEMBRANESFOROXYGENSEPARATION
525
182. H.-M. Zhang, Y. Shimizu, Y. Teraoka, M. Miura and N. Yamazoe, Oxygen sorption and catalytic properties of Lal-xSrxCol-yFeyO3-~perovskite-type oxides. J. Catal., 121 (1990) 1367-1370. 183. B.A. Van Hassel, T. Kawada, N. Sakai, H. Yokokawa and M. Dokiya, Oxygen permeation modelling of Lal-yCayCrO3-8. Solid State Ionics, 66 (1993) 41-47. 184. T. Kawada, T. Horita, N. Sakai, H. Yokokawa and M. Dokiya, Experimental determination of oxygen permeation flux through bulk and grain boundary of La0.TCa0.3CrO3. Solid State Ionics, 79 (1995) 201-207. 185. B.A. Van Hassel, T. Kawada, N. Sakai, H. Yokokawa, M. Dokiya and H.J.M. Bouwmeester, Oxygen permeation modelling of perovskites. Solid State Ionics, 66 (1993) 295-305. 186. J.B. Goodenough, Fast ionic conduction in solids. Proc. Roy. Soc. (London), A393 (1984) 214-234. 187. K. Nisancioglu and T.M. Gfir, Oxygen difusion in iron doped strontium cobaltites, in: S.C. Singhal and H. Iwahara (Eds.), Proceedings of the 3rd International Symposium on Solid Oxide Fuel Cells. Electrochemical Society, Pennington, NJ, 1994, pp. 267-275. 188. T. Ishigaki, S. Yamauchi, J. Mizusaki, K. Fueki and H. Tamura, Tracer diffusion coefficient of oxide ions in LaCoO3 single crystal. J. Solid State Chem., 54 (1984) 100-107. 189. T. Ishigaki, S. Yamauchi, J. Mizusaki, K. Fueki and H. Tamura, Diffusion of oxide ions in LaFeO3 single crystal. Solid State Chem., 55 (1984) 50-53. 190. M.C. Kim, S.J. Park H. Haneda, J. Tanaka, T. Mitsuhashi and S. Shirasaki, Self-diffusion of oxygen in Lal-xSrxFeO3-~.J. Mat. Sci. Lett., 9 (1990) 102-104. 191. K. Fueki, J. Mizusaki, S. Yamauchi, T. Ishigaki and Y. Mima, Nonstoichiometry and oxygen vacancy diffusion in the perovskite type oxides Lal-xSrxCoO3-5, in: P. Barret and L.C. Dufour (Eds.), Proceedings of the l Oth International Symposium on Reactivity of Solids 1984. Elsevier, Amsterdam, 1985, pp. 339-343. 192. Y. Denos, F. Morin and G. Trudel, Oxygen diffusivity in strontium substituted lanthanum cobaltite, in: T.A. Ramanarayanan, W.L. Worrel and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics. The Electrochemical Society, Pennington, NJ, 1994, pp. 150-158. 193. K. Nisancioglu and T.M. Gfir, Potentiostatic step technique to study ionic transport in mixed conductors. Solid State Ionics, 72 (1994) 199-203. 194. S. Sekido, H. Tachibana, Y. Yamamura and T. Kambara, Electric-ionic conductivity in perovskite-type oxides, SrxLal-xCOl-yFeyO3-~.Solid State Ionics, 37 (1990) 253-259. 195. I. Yasuda and T.J. Hikita, Precise determination of the chemical diffusion coefficient of calcium-doped lanthanum chromites by means of electrical conductivity relaxation, Electrochem. Soc., 141(5) (1995) 1268-1273. 196. A. Belzner, T.M. Gfir and R.A. Huggins, Oxygen chemical diffusion in strontium doped lanthanum manganites. Solid State Ionics, 57 (1992) 327-337. 197. T.M. Gfir, A. Belzner and R.A. Huggins, A new class of oxygen selective chemically driven nonporous ceramic membranes. Part I. A-site doped perovskites. J. Membr. Sci., 95 (1992) 151-162. 198. L.-W. Tai, M.M. Nasrallah and H.U. Anderson, Lal-xSrxCol-yFeyO3-~,A potential cathode for intermediate temperature SOFC applications, in: S.C. Singhal and H. Iwahara (Eds.), Proceedings of the 3rd International Symposium on Solid Oxide Fuel Cells. The Electrochemical Society, Pennington, NJ, 1994, pp. 241-251. 199. J.P. Bonnet, J.C. Grenier, M. Onillon, M. Pouchard and P. Hagenmuller, Influence de la pression partiele d'oxygene sur la nature des defauts ponctuels dans Ca2LaFe3Os+x. Mat. Res. Bull., 14 (1979) 67-75.
526
10B DENSECERAMICMEMBRANESFOROXYGENSEPARATION
200. C. Zener, Phys. Rev., 82 (1952) 403. 201. R. Koc, H.U. Anderson, S.A. Howard and D.M. Sparlin, Structural, sintering and electrical properties of the perovskite-type (La,Sr)(Cr, Mn)O3, in: S.C. Singhal (Ed.), Proceedings of the 1st International Symposium on Solid Oxide Fuel Cells. The Electrochemical Society, Pennington, NJ, 1989, pp. 220-244. 202. R. Rafaelle, H.U. Anderson, D.M. Sparlin and P.E. Perris, Transport anomalies in the high-temperature hopping conductivity and thermopower of Sr-doped La(Cr, Mn)O3. Phys. Rev. B, 43 (1993) 7991-7999. 203. S.R. Sehlin, H.U. Anderson and D.M. Sparlin, Electrical characterization of the (La,Ca)(Cr, Co)O3 system. Solid State Ionics, 78 (1995) 235-243. 204. L.-W. Tai, M.M. Nasrallah, H.U. Anderson, D.M. Sparlin and S.R. Sehlin, Structure and electrical properties of Lal-xSrxCol-yFe~3. Part I. The system La0.8Sr0.2COl-yFeyO3.Solid State Ionics, 76 (1995) 259-71; Part 2. Lal-xSrxCo0.2Fe0.803, ibid., 273-283. 205. J.B. Goodenough, Metallic Oxides, in: H. Reiss (Ed.), Progress in Solid State Chemistry. Pergamon, Oxford, 1971, pp. 145-399. 206. P.M. Racah and J.B. Goodenough, Phys. Rev., 155(3) (1967) 932-943. 207. G. Thornton, F.C. Morrison, S. Partington, B.C. Tofield and J. Williams, The rare earth cobaltites: localised or collective electron behaviour. Phys. C: Solid State Phys., 21 (1988) 2871-2880. 208. J. Mizusaki, J. Tabuchi, T. Matsuura, S. Yamauchi and K. Fueki, Electrical conductivity and seebeck coefficient of nonstoichiometric Lal-xSrxCoO3-8. J. Electrochem. Soc., 136(7) (1989) 2082-2088. 209. J.W. Stevenson, M.M. Nasrallah, H.U. Anderson and D.M. Sparlin, Defect structure of Yl-yCayMnO3 and Lal-yCayMnO3. I Electrical properties. J. Solid State Chem., 102 (1993) 175-184; II. Oxidation-reduction behaviour, ibid., 185-197. 210. J.A.M. Van Roosmalen and E.P.H. Cordfunke, A new defect model to describe the oxygen deficiency in perovskite-type oxides. J. Solid State Chem., 93 (1991) 212-219. 211. J.B. Goodenough, J.E. Ruiz-Diaz and Y.S. Zhen, Oxide-ion conduction in Ba2In205 and Ba3In2MO8 (M = Ce, Hf, or Zr). Solid State Ionics, 44 (1990) 21-31. 212. J.B. Goodenough, A. Manthiram, P. Paranthanam and Y.S. Zhen, Fast oxide-ion conduction in intergrowth structures. Solid State Ionics, 52 (1992) 105-109. 213. J.A. Kilner and R.J. Brook, A study of oxygen ion conductivity in doped nonstoichiometric oxides. Solid State Ionics, 6 (1982) 237-252. 214. M. Cherry, M.S. Islam and C.R.A. Catlow, Oxygen ion migration in perovkite-type oxides, J. Solid State Chem., 118 (1995) 125-132. 215. R.L. Cook and A.F. Sammells, On the systematic selection of perovskite solid electrolytes for intermediate temperature fuel cells. Solid State Ionics, 45 (1991) 311-321. 216. A.F. Sammells, R.L. Cook, J.H. White, J.J. Osborne and R.C. MacDuff, Rational selection of advanced solid electrolytes for intermediate temperature fuel cells. Solid Sate Ionics. 52 (1992) 111-123. 217. M.P. Dijk, K.J. de Vries and A.J. Burggraaf, Oxygen ion and mixed conductivity in compounds with the fluorite and pyrochlore compounds. Solid State Ionics, 9/10 (1983) 913-919. 218. M.P. Dijk, K.J. de Vries, A.J. Burggraaf, A.N. Cormack and C.R.A. Catlow, Defect structures and migration mechanisms in oxide pyrochlores. Solid State Ionics, 17 (1985) 159-167. 219. C.N.R. Rao, J., Gopalakrishnan and K. Viyasagar, Superstructures, ordered defects and nonstoichiometry in metal oxides of perovskite and related structures. Ind. J. Chem., 23A (1984) 265-284.
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
527
220. P. Hagenmuller, M. Pouchard and J.C. Grenier, Nonstroichiometry in the perovskitetype oxides: An evolution from the classic Scottky-Wagner model to the recent high Tc superconductors. Solid State Ionics, 43 (1990) 7-18. 221. D.M. Smyth, Defects and order in perovskite-related oxides. Ann. Rev. Mater. Sci., 15 (1985) 329-357. 222. D.M. Smyth, Defects and structural changes in perovskite systems: from insulators to superconductors. Cryst. Latt. Def. Amorph. Mat., 18 (1989) 355-375. 223. M.T. Anderson, J.T. Vaughy and K.R. Poeppelmeier, Structural similarities among oxygen-deficient perovskites. Chem. Mater., 5(2) (1993) 151-65. 224. M.A. Alario-Franco, M.J.R. Henche, M. Vallet, J.M.G. Calbet, J.C. Grenier, A. Wattiaux and P. Hagenmuller, Microdomain texture and oxygen excess in the calcium-lanthanum ferrite: Ca2LaFe308. J. Solid State Chem., 46 (1983) 23-40. 225. M.A. Alario-Franco, J.M. Gonz~ilez-Calbet and J.C. Grenier, Brownmillerite-type microdomans in the calcium lanthanum ferrites: CaxLal-xFeO3-y. I. 2/3 < x < 1. J. Solid State Chem., 49 (1983) 219-231. 226. M.A. Alario-Franco, J.M. Gonz~ilez-Calbet and M. Vallet-Regi, Microdomains in the CaFexMnl-xO3-y ferrites. ]. Solid State Chem., 65 (1986) 383-391. 227. J.M. Gonz~les-Calbet, M.A. Alario-Franco and M. Vallet-Regi, Microdomain formation: A sophisticated way of accomodating compostional variations in non-stoichiometric perovkites. Cryst. Latt. Def. Amorph. Mat., 16 (1987) 379-385. 228. J.M. Gonz~iles-Calbet, M. Parras, M. Vallet-Regi and J.C. Grenier, Anionic vacancy distribution in reduced barium-lanthanum ferrites: BaxLal-xFeO3-x/2 (1/2 < x < 2/3). J. Solid State Chem., 92 (1991) 100-115. 229. P.D. Battle, T.C. Gibb and S. Nixon, The formation of nonequilibrium microdomains during the oxidation of Sr2CoFeO5. J. Solid State Chem., 73 (1988) 330-337. 230. T. Takahashi, H. Iwahara and T. Esaka, High oxide conduction in sintered oxide of the system Bi203-M203. J. Electrochem. Soc., 124(10) (1977) 1563-1569. 231. S. Shin, M. Yonemura and H. Ikawa, Order-disorder transition of Sr2Fe205 from brownmillerite to perovskite at elevated temperature. Mat. Res. Bull., 13 (1978) 1017-1021. 232. S. Shin, M. Yonemura and H. Ikawa, Crystallographic properties of Ca2Fe2Os. Difference in crystallographic properties of brownmillerite-like compounds, Ca2Fe205 and Sr2Fe2Os, at elevated tmperature. Bull. Chem. Soc. Jpn., 52 (1979) 947-948. 233. J.C. Grenier, E. Norbert, M. Pouchard and P. Hagenmuller, Structural transitions at high temperature in Sr2Fe205. J. Solid State Chem., 58 (1985) 243-252. 234. J. Rodriguez-Carvajal, M. Vallet-Regi and J.M. Gonz~les-Calbet, Perovskite threefold superlattices: A structure determination of the A3M308 phase. Mater. Res. Bull., 24 (1989) 423-429. 235. A. Manthiram, J.F. Kuo and J.B. Goodenough, Characterization of oxygen-deficient perovskites as oxide-ion electrolytes. Solid State Ionics, 62 (1992) 225-234. 236. G.B. Zhang and D.M. Smyth, Defects and transport in brownmillerite oxides, in: T.A. Ramanarayanan, W.L. Worrel and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics. The Electrochemical Society, Pennington, NJ, 1994, pp. 608-622. 237. S.B. Adler, J.A. Reimer, J. Baltisberger and U. Werner, Chemical structure and oxygen dynamics in Ba2In205. J. Am. Ceram. Soc., 116(2) (1994) 675-681. 238. S. Adler, S. Russek, J. Reimer, M. Fendorf, A. Stacy, Q. Huang, A. Santoro, J. Lynn, J. Baltisberger and U. Werner, Local structure and oxide-ion motion in defective perovskites. Solid State Ionics, 68 (1994) 193-211.
528
10 n D E N S E C E R A M I C M E M B R A N E S FO R O X Y G E N S E P A R A T I O N
239. L. Er-Rakho, C. Michel, P. LaCorre and B. Raveau, YBa2CuFeO5+~: A novel oxygen-deficient perovskite with a layer structure. J. Solid State. Chem., 73 (1988) 531. 240. L. Barbey, N. Nguyen, V. Caignart, M. Hervieu and B. Ravenau, Mixed oxides of cobalt and copper with a double pyramidal layer structure. Mat. Res. Bull., 27 (1992) 295-301. 241. R.H.E. Van Doorn, E.G. Keim, T. Kachliki, H.J.M. Bouwmeester and A.J. Burggraaf, in preparation. 242. W.T.A. Harrison, T.H. Lee, Y.L. Yang, D.P. Scarfe, L.M. Liu and A.J. Jacobson, A neutron diffraction study of two strontium cobalt iron oxides. Mat. Res. Bull., 30(5) (1995) 621-629. 243. R.M. Barrer, J.A. Barrie and M.G. Rogers, Trans. Faraday Soc., 58 (1962) 2473. 244. J.P. Zhou and J.T. McDevitt, Corrosion reactions of YBa2Cu3OT-x and T12Ba2Ca2Cu3010+x superconductor phases in aqueous environments. Chem. Mater., 4 (1992) 952-959. 245. E.J. Ten Elshof, H.J.M. Bouwmeester and H. Verweij, Oxygen transport through Lal-xSrxFeO3-~ membranes. II. Permeation in air/CO,CO2 gradients. Solid State Ionics, submitted. 246. J. Yasuda, K. Ogasawara and M. Kishinuma, Oxygen tracer diffusion in (La,Ca)CrO3-5, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conducting Oxide Ceramics, Vol. 94-12. The Electrochemical Society, Pennington, NJ, 1994, pp. 164-173. 247. S.P.S. Badwal, J. Drennan and A.E. Hughes, Segregation in oxygen-ion conducting solid electrolytes and its influence on electrical properties, in: J. Nowotny (Ed.), Science of Ceramic Interfaces. Elsevier, Amsterdam, 1991, pp. 227-285. 248. T. Van Dijk and A.J. Burggraaf, Grain boundary effects on ionic conductivity in ceramic GdxZrl-xO2-x/2 solid solutions. Phys. Stat. Sol. (A), 63 (1981) 229-240. 249. M.J. Verkerk, B.J. Middelhuis and A.J. Burggraaf, Effect of grain boundaries on the conductivity of high purity ZrO2-Y203 ceramics. Solid State ionics, 6 (1982) 159-170. 250. H. Schmalzried, W. Laqua and P.L. Lin, Crystalline oxide solid solutions in oxygen potential gradients. Z. Naturforsch., 34A (1979) 192-199. 251. H. Schmalzried and W. Laqua, Multicomponent oxides in oxygen potential gradients. Oxid. Met., 15 (1981) 339-353. 252. D. Monceau, C. Petot and E. Petot, Kinetic demixing profile calculation under a temperature gradient in multi-component oxides. J. Europ. Ceram. Soc., 9 (1992) 193-204. 253. M. Martin and R. Schmackpfeffer, Demixing of oxides: influence of defect interactions. Solid State Ionics, 72 (1994) 67-71. 254. P.K. Gallagher, G.S. Grader and H.M. O'Bryan, Effect of an oxyge gradient on the Ba2YCu3Ox superconductor. Solid State Ionics, 32/33 (1989) 1133-1136. 255. K. Vedula, Modelling of transient and steady-state demixing of oxide solid solutions in an oxygen chamical potential gradients. Oxid. Met., 28 (1987) 99-108. 256. T. Ishikawa, S.A. Akbhar, W. Zhu and H. Sato, Time evolution of demixing in oxides under an oxygen potential gradient. J. Am. Ceram. Soc., 71(7) (1988) 513-521. 257. H. Schmalzried, Behavior of (semiconducting) oxide crystals in oxygen potential gradients. React. Solids, 1 (1986) 117-137. 258. U. Balachandran, M. Kleefish, T.P. Kobylinski, S.L. Morisetti and S. Pei, Oxygen ionconducting dense ceramic, Patent Appl. PCT/US94/03704. 259. J.W. Keulks, The mechanism of oxygen atom incorporation into the products of propylene oxidation over bismuth molybdate. J. Catal., 19 (1970) 232-156. 260. J. Novakova and P. Jiru, A comment on oxygen mobility during catalytic oxidation. J. Catal., 27 (1972) 155-156.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.I. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 11
Current developments and future research in catalytic m e m b r a n e reactors Jose S a n c h e z ~ a n d Theodore T. Tsotsis2 1Laboratoire des Materiaux et Proc6d@s Membranaires, UMR 9987 CNRS ENSCM UMII, 2, Place E. Bataillon, cc 024, 34095 Montpellier Cedex 5, France 2Department of Chemical Engineering, University of Southern California, Los Angeles, CA 90089-1211, USA
11.1 INTRODUCTION Membrane reactors combine two distinctly different functions, i.e., reaction and separation into a single operation. For the high temperature catalytic membrane reactors, which are the main topic of this chapter, the reaction function is most often carried out by a conventional bed (packed, fluidized or moving) of catalyst particles; the membrane (metal or ceramic especially suited for the high temperature operation) is placed inside the reactor and carries out the separation function. The membrane reactor concept has evolved from the simpler design concept shown in Fig. 11.1, where the reaction and separation functions are carried out by two different processing units, the secondd_ unit being a membrane separator. For bioengineering applications the process of Fig. 11.1 is finding widespread applications [1-3]. For the high temperature catalytic applications discussed here the process economics are only recently beginning to look attractive with the development of permselective, high temperature resistant membranes. It will become obvious to the reader that membrane reactors, where reaction and separation are carried out simultaneously in a single physical unit, provide significantly better design options than the concept of Fig. 11.1. Combining reaction and separation in the same unit often creates a synergy.
530
11 - - CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
Feed Reactants
t
v'--l
Reactor
I
! Membrane ] Separator ]
Products
Reactants Recycle
Fig. 11.1. A conventional reactor and a membrane separator with reactant recycle. The presence of the membrane enhances the "per-pass" conversion and, in turn, enhanced "per-pass" conversions diminish the downstream separation requirements. Throughout this chapter, the reader should look for the need for this synergy. It is where a difficult separation problem exists, coupled to a "per-pass" conversion or selectivity or equilibrium limitation problem that the application of membrane reactors makes the best sense. The earlier applications of membrane reactors involved for the most part various equilibrium limited dehydrogenation reactions, where the role that the membrane played was well defined and simple to describe; remove hydrogen efficiently and by so doing increase the yield of the reaction, by shifting the conversion to the right of the equilibrium. The earlier reactor designers envisioned two main reasons why the membrane reactor concept would improve upon the economics of the more conventional processes. These are the capital and operational savings realized by eliminating the processing steps required for separating the hydrocarbons from the hydrogen rich stream, and the energy savings realized from the lower operating temperatures required due to the reactor yield enhancement. Three decades or so later the promise still remains mostly unrealized for the high temperature catalytic dehydrogenation applications the early designers envisioned. Though small-scale commercial applications exist we are not aware of any large-scale commercial units. The reasons are many and varied and we will touch upon them throughout this chapter. In the meantime membrane reactors are finding increasing utilization in bioengineering applications both for the production of fine chemicals via the use of both enzyme and whole-cell bioreactors [2] and for large-scale environmental clean-up type applications [4]. Membrane bioreactors remain an exciting area with many important new processes coming on line or in the pilot-plant stage. The topic, however, goes beyond the scope of this chapter; furthermore membrane bioreactors use for the most part polymeric or organic membranes, which are outside the theme topic of this book. Those interested for further reading on the topic (and everyone working in the area of catalytic/high temperature membrane reactors should be since many of the basic concepts are similar and some of the smart ideas developed could have widespread applications) can get started with a number of fairly recent review papers [1-3]. Some of the exciting things that are recently happening in the area of high temperature catalytic membrane reactors (which we will discuss in more detail
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
531
later in this chapter) no longer involve equilibrium limited dehydrogenation reactions. The membrane instead aims to separate intermediates and products from the reacting zone so that they do not deactivate the catalyst or undergo further undesirable reaction. In some applications the membrane is not required to be permselective since its only role is to provide a controlled reactive interface between reactants flowing on opposite sides of the membrane. [5,6]. Whether such devices (which is this chapter are referred to by the acronym CNMR, namely catalytic nonpermselective membrane reactors) should be called membrane reactors in the first place will stay a matter of debate among the purists; nevertheless these 'membrane reactors' are attracting a growing share of attention with a number of hydrogenation, oxidation and partial oxidation reactions studied. Membrane reactors as a concept date back to the 1950s. In the limited space of a book chapter, one cannot, of course, even begin to do justice to the topic and the many interesting studies that have been published. Fortunately the field has been blessed with a number of well written review papers focusing on both general and special aspects of the problem [7-16]. For those who are beginning in this area in additionto these papers we would also recommend the proceedings volumes of the International Conferences on Inorganic Membranes and a Volume 25 of Catalysis Todayon the topic, which gives a good snapshot of what is currently happening in the area. It is a challenging feat to compose a chapter on a topic as old but also as vibrant as catalytic membrane reactors are. The risk is real that one is obsolete by the time the ink dries. The approach in the writing of this chapter is motivated by the belief that a technical book is first and foremost a reference guiae; we have, therefore, tried to be inclusive by attempting to cover as wide an area of topics as the space allows. For the benefit of breadth, we have sometimes sacrificed depth. For those desiring further in-depth study the long (but by necessity very selective) list of technical citations at the end of the chapter should provide welcome additional insight. The choice of illustrations in the discussion of various topics, furthermore, is more a reflection of personal taste rather than of a judgment of technical quality. More often than not the choice was among many equally illustrative and important examples. Before proceeding further it would be appropriate for our readers to familiarize themselves with the few additional acronyms that will be used in this chapter and which are listed in Table 11.1. They are used to describe some of the most common membrane reactor configurations that have been studied in the technical literature. By far the most commonly referred to reactor is the PBMR, in which the reaction function is provided by a packed bed of catalysts in contact with the membrane. The membrane is not itself catalytic at least not intentionally so. Some of the commonly utilized inorganic and metal membranes, on the other hand, are intrinsically catalytically active. The PBMR classification, therefore, should be assigned with caution. When the packed bed
532
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
T A B L E 11.1
Types of membrane reactors Acronym
Description
CMR CNMR PBMR PBCMR FBMR FBCMR
Catalytic membrane reactor Catalytic nonpermselective membrane reactor Packed-bed membrane reactor Packed-bed catalytic membrane reactor Fluidized-bed membrane reactor Fluidized-bed catalytic membrane reactor
is replaced by a fluidized bed the FBMR configuration results. The concept is really interesting but FBMRs have yet to gain widespread acceptance. In the CMR configuration the membrane provides both the separation and the reaction function. The concept, however, has found wider acceptance in the bioreactor area than with catalytic reactors. Finally, the PBCMR (and FBCMR) uses both a catalytic bed and a permselective membrane. This configuration would appear to be ideal for situations where a bifunctional catalytic function is desirable; we are not aware, however, of many examples of such PBCMR use. Other reactor configurations and concepts have also been discussed in the technical literature. Most commonly cited are hybrid concepts, where the membrane reactor is used as an add-on stage to an already existing conventional reactor. This particular configuration has a number of attractive features, especially for applications involving conventional type porous membranes, which are characterized by moderate (Knudsen-type) permselective properties. Staged membrane reactors have received mention and so have reactors with multiple feed-ports and recycle. To facilitate the transport across the membrane in laboratory studies one often applies a sweep gas or a vacuum in the permeate side or a pressure gradient across the membrane. It is unlikely that the first two approaches, effective as they may be in laboratory applications, will find widespread commercial application. 11.2 D E N S E METAL M E M B R A N E REACTORS
The earlier membrane reactors involved the use of Pd and Pd alloy membranes. Pd together with a handful of other metals is permeable to hydrogen but virtually impermeable to other gases and, of course, liquids. The diffusion process through Pd, furthermore, is an activated process and at high temperatures such membranes show very reasonable permeances. The pioneering work on Pd membrane reactors was done by Gryaznov and coworkers in the former Soviet Union and some industrial groups in the U.S. and Europe. Gryaznov and
11
-
-
533
CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
coworkers have studied a great many systems involving both hydrogenation and dehydrogenation reactions. Their work has been very ably reviewed by Gryaznov et al. [17], Gryaznov [18] and more recently Shu et al. [19]. Their focus on recent years has been on the production of high added value, specialty chemicals. Some examples of these applications involve: the synthesis of vitamin K from quinone and acetic anhydride [17] and the cis/trans-2-butene-l,4 diol hydrogenation to cis/trans-butanediol [20]. (a)
It
(b)
SECTION " A A "
Double spiral coiled plate membrane reactor.
Flat double spiral tube membrane reactor. Corrugated Spacer Plate
Pd- Based
Vlembrone
Dehydrogenation Side
Hydrogenation Side
Schematic of a cross-flow catalytic membrane reactor Fig. 11.2.Various types of Pd membrane reactors. Reproduced from Shu et al. [19]with permission.
534
11 m CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
Gryaznov and coworkers have paid particular attention to the mechanical design aspects of the metal membranes and have come up with very elaborate designs as can be seen in Fig. 11.2. Small scale Pd-alloy membrane reactors became available in the eighties for CH4/CH3OH reforming for H2 production for stationary fuel-cell type applications in remote locations [14,21]. Large-scale industrial applications have yet to materialize, however, because of a number of concerns. These include:
11.2.1 Cost and Availability Self-supporting, mechanically resistant Pd or Pd-alloy dense membranes must be typically a few hundred ~t thick. Pd being an expensive metal the cost of such membranes is, of course, non-trivial. What is often not mentioned, furthermore, in the discussion about dense metal membranes is the effect that the increased demand of the metal will have on worldwide markets and prices, if metal membrane reactors were to be adopted for large scale chemical processes like for example methane reforming. In recent years various groups have instead focused their attention on the development of composite metal membranes whereby a thin (a few ~t) film of the metal is deposited on (or in the pores of) an underlying porous support. This kind of membrane aims to have the mechanical resistance of the inorganic porous matrix and the high selectivity of a dense membrane with better permeabilities. Porous glass, ceramic and metal supports have been utilized [19], the metal film deposited by a variety of conventional techniques like vacuum sputtering [14], pyrolysis [23] and electroless plating [24-25]. Metal composite membranes are not free of problems themselves. In addition to the phase transition problem to be discussed below other problems include loss of mechanical stability, due to the significant differences in thermal expansion coefficients between the metal film and the support and metal atom/ion counter diffusion (this latter phenomenon is also of concern for some of the advanced microporous membranes, see discussion in the appropriate section). Such problems are being extensively investigated by Edlund and coworkers [26] at Bend research and by other groups [22].
11.2.2 Mechanical and Thermal Stability Pure Pd membranes become brittle upon thermal cycling in a H 2 atmosphere due to a phase transition between the different Pd hydride phases (c~ and ~) with distinctly different crystal lattice parameters (see Fig. 11.3). Alloying with various other metals (for example Ag, Ru, Rh) tends to lower the phase transition temperature and undersome circumstances improves the hydrogen permeability. The phase transition problem is not technically insurmountable, if good care is exercised during start-up (when the membrane is first exposed to hydrogen)
11 -- CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
lOJ-
535
/Fzeo'c r 288, "C
o
I:L o4 2: 0.
0.1
,
o.o11 0.001 0
\
I
I
i.......... .....
/ i ,m
\
20"C 0.1
0.2 ~
0.:5
0.4
0.5
0.6
0.7 n
Fig. 11.3. Equilibrium relationship between the hydrogen pressure and the H / P d ratio (n) in the metal. Shown are the domains of existence of the R and ~ palladium hydride phases. Reproduced from Shu et al. [19] with permission.
and shutdown operations (when the membrane is purged completely by hydrogen). The concern is, however, still there. Mechanical and thermal stability problems real or perceived have plagued the future of dense metal membranes in terms of large scale chemical process applications. Membrane cost and reactor safety issues notwithstanding, the shutdown costs of a large chemical plant brought upon by the need for membrane replacement can be astronomical.
11.2.3 Poisoning and Carbon Deposition Problems Poisoning by S and C1 containing gas-phase impurities and deactivation due to carbon deposition are matters of concern with Pd membranes. Ali et al. [27], for example, recently studied the effect of gas phase impurities on Pd-Ag composite membranes during their use in a membrane reactor for methylcyclohexane (MCH) dehydrogenation to toluene (TOL). The presence of small amounts of dimethyl disulfide (DMDS) in the MCH feedstream were shown to significantly reduce the membrane permeability. For example, after exposure to the MCH-DMDS feed for about I h the I--I2 permeation rate decreased to 12% of its value with the S free feedstream. Similar behaviour was observed with a C1 containing gas phase impurity (CC14). The results of Ali et al. are consistent with reports from the Bend Research labs [26] which show that poisoning by
536
11 -- CURRENT DEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
H2S can have significant effects on dense Pd membranes resulting in some instances in complete m e m b r a n e failure in short periods of time. Short of replacing Pd with other metals, it is unclear at this point how one overcomes the problem of Pd sensitivity to sulphur and chloride impurities, which are unfortunately present in m a n y industrial hydrocarbon streams. 100
f
J
80
60 @
It
~ 40
",,,,
@ r..)
"' ",, ,,,, .,, .,, " -' .,, .. ,,...
"'-.....
20
equilibrium I
I
,
I
I
10
2 4 6 8 Reaction pressure / arm 100 80r..) @
60,.
Q r,.9
,, .,. " S'''''
O S
,,,,,.""
40am_
200
v
I..
600
I
,,
I
650 700 750 Temperature / K
I
800
Fig. 11.4. Top figure shows the effect of pressure on the reaction side of the membrane on methane conversion in the Pd membrane reactor; bottom figure shows the effect of temperature. The solid line and the symbols (o) are for the Pd membrane reactor. The dotted line is the calculated equilibrium conversion and the symbols (r'l) are for a membrane reactor using a porous Vycor glass membrane. Reproduced from Uemiya et al. [29] with permission.
11 - - CURRENT D E V E L O P M E N T S A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
537
Less understood is the effect of carbon deposition. Ali et al. [27], for example, showed that after an initial pretreatment, membrane exposure to MCH + H 2 or TOL + H 2 mixtures did not significantly influence its permeability. Surface carbon on the other hand, which often results from exposure to hydrocarbons, has a tendency to dissolve in the bulk of the metal. In a study on the topic [28] this resulted eventually in membrane mechanical failure. Their problems notwithstanding Pd membrane reactors have been shown very effective through the years in enhancing the yield of both hydrogenation and dehydrogenation and dehydrocyclodimerization reactions. Figure 11.4 is from the study of CH4 steam reforming by Uemiya et al. [29] using a Pd membrane coated on a Vycor glass tube by an electroless plating technique. The significant enhancement in methane conversion over the calculated equilibrium is evident. Interestingly the effect of reactor pressure in the membrane reactor is quite the opposite of what would be expected for the same reaction being run in a conventional reactor. According to Kikuchi and coworkers this is indicative that what limits the membrane reactor is the hydrogen permeation through the membrane, which is a strongly activated process, thus the significant effect of temperature. This is not always, however, true and in some of the studies reported where the membrane reactor failed to produce significant enhancements in yield [30] slow kinetics have been blamed for the failure. An interesting application of palladium membranes to C H 4 reforming was proposed by Adris et al. [31]. They combined a fluidized catalytic bed with palladium membranes for the C H 4 steam reforming. C H 4 o r CH3OH steam reforming remains the reaction of choice with no less than six different research group reporting studies using Pd membrane reactors over the last year alone [32-37]. The only other metal that has received some serious attention for membrane reactor applications is Ag [14,38] which is permeable to oxygen. Ag has similar thermal/mechanical stability problem as Pd and in addition its oxygen permeability is orders of magnitude lower. 11.3 P O R O U S I N O R G A N I C MEMBRANE REACTORS
The earlier applications of membrane reactors using porous inorganic membranes involved the use of glass membranes, i.e., Vycor glass. Reactions studied included HI and H2S decomposition [39-41]. The latter reaction is still receiving attention for porous glass membrane reactor applications [41]; H2S catalytic decomposition is a very important but also truly challenging experimental reaction system. Porous Vycor glass is a mesoporous system (average pore diameter ~40 ~. Self-supporting, mechanically resistant Vycor glass membranes must have a thickness of ~1 ram. The permeances of such membranes are, therefore, 2-3 orders of magnitude less than the conventional mesoporous asymmetric sol-gel A1203 membranes.
538
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
A few years ago PPG produced an exciting new microporous glass membrane, which exhibits molecular sieving properties [42]. The membranes became available in a hollow fibre configuration and have since proven an interesting topic of fundamental research by a number of U.S. groups [43]. These glass membranes are plagued by mechanical stability problems, however, and to the best of our knowledge PPG is not pursuing their further commercialization. Hollow fibre inorganic membranes have from time to time been announced and soon after forgotten; it appears, at least to those not intimately involved with the technology, that it will be a challenging technical task to produce a hollow ceramic fibre that will be able to withstand the kind of environments one expects to encounter in high temperature catalytic reaction applications. Mesoporous Vycor glass membranes have been used as a porous substrate for the deposition of various metal or other inorganic films in order to create asymmetric type highly permselective membranes. We have already discussed the membranes made by Kikuchi and coworkers [24,25] who deposit Pd films on Vycor glass membranes by electroless plating. As shown in Fig. 11.4 for the C H 4 steam reforming reaction carried out in a PBMR using the unmodified Vycor glass membrane alone conversions are at or slightly above the calculated equilibrium; on the other hand the composite Pd/Vycor glass membrane attains conversion significantly higher than equilibrium. Using CVD techniques various groups are also depositing both dense and m i c r o p o r o u s SiO 2 films within the porous structure of Vycor glass membranes [44,45]. Dense SiO 2 like Pd is permeable to hydrogen but not to other gases, though in SiO 2 it appears that transport involves molecular rather than atomic hydrogen. Ioannides and Gavalas [46] used such a membrane to study isobutane dehydrogenation in a PBMR but problems with the catalyst itself seemed to overshadow the beneficial membrane effects. Dense SiO 2 has very low permeability and membranes with reasonable permeances must be very thin (~ a few hundred ~). Such membranes have been recently produced by Gavalas and Jiang [47] using a technique which involves the deposition of an organic template within the membrane pores which is then burned away after the deposition of the SiO 2 film by CVD. A number of concerns still remain with the dense SiO 2 membranes. Some of these are not unique to SiO 2 but concern all dense or molecular sieve type membranes. Due to their small permeabilities such membranes are sensitive to microcrack development. A single crack of an area of few ~t2 allowing a convective flow can potentially neutralize the permselective properties of a membrane area many orders of magnitude its size. Poisoning and coking is always a concern when membrane transport is preceded by dissolution/adsorption type processes. One unique concern for dense SiO 2 [48] membranes is the fact that these membranes undergo a densification process, especially in the presence of steam. It is unclear what is the true long term effect of this densification process;
11 ~ C U R R E N T D E V E L O P M E N T S A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
539
the data that have been reported so far are for a limited set of conditions and relatively short periods of time. Until very recently all membrane reactor applications using porous inorganic membranes (other than glass) involved alumina membranes. With the exception of the use of alumina membranes for H2S decomposition in the late seventies and early eighties [40,49-50] all other applications are more recent. In the middle to late eighties Alcan Laboratories in England developed an efficient technique for producing both symmetric and asymmetric type anodic alumina membranes. These membranes are ideally suited for fundamental studies of transport and reaction because they have straight nonintersecting pores. Alcan workers reported on the use of such membranes in a catalytic membrane reactor for C2H6 dehydrogenation and C2H4 hydrogenation and hydrogenolysis [51]. Tsotsis and coworkers [52] prepared their own anodic membranes by anodization of aluminum in various electrolytes. They used the resulting membranes for studying cyclohexane dehydrogenation in a PBMR [53]. Though anodic aluminas are ideal for fundamental investigations of membrane reactor systems they do not appear promising for large-scale industrial applications. The films that are available today are appropriate for small-scale laboratory filtration type applications; they do not appear to have the mechanical strength required for large-scale industrial reactor applications. More appropriate for reactor applications are asymmetric sol-gel type alumina membranes. These typically consist of a mesoporous thin 7-A1203 film placed on the top of one or more macroporous o~-A1203 support layers. The 7-A1203 film is deposited by a sol-gel process [54-56]. The commercially available alumina membranes consist of several macroporous support layers each layer with a successively smaller pore diameter. This technique seems to help in avoiding crack formation and to eliminate significant infiltration of the deposited layer into the underlying support layer. Unless special precautions are taken, however, some infiltration is unavoidable [57], and in modelling calculations this must be taken into account [13]. The Process and Media Technology Group has recently reported [58] a series of studies in which they measured the thickness of the permselective layer along the membrane length of commercial A1203 membranes. They report significant variations for the same membrane and among different membranes. This is an alarming finding pointing out the difficulty in standardizing the preparation of such membranes and the need for additional fundamental studies and insight. In focusing on the challenges involved in the preparation of the thin permselective sol-gel films what one often ignores is the fact that a major component in determining the final cost of such membranes is the quality and cost of the underlying macroporous support layers. Several groups are recently doing work in this area [59-60] and a critical need exists for major developments here. Mesoporous alumina membranes have been studied extensively mostly for
540
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
60 "@
@
55
5O 45 40 "N 35 o
30
o
25
~0
I~, 20
15 [0
600
650
700
750
800
Temperature (K) Fig. 11.5. Formaldehyde yields in various reactors under similar experimental conditions. V, conventional plug-flowreactor; e, a PBMRusing a Pd/A1203dense compositemembrane; A,a PBMR using a mesoporous A1203membrane; O, a CMRushlg a mesoporous A1203membrane catalytically impregnated by a sol-gel technique. Reproduced from Deng and Wu [61] with permission. equilibrium limited dehydrogenation of various lower molecular hydrocarbons and alcohols [7-16]. Figure 11.5, for example, is from a very recent study by Deng and Wu [61]. These authors studied the CH3OH dehydrogenation and compared the performance of a variety of membrane reactors including a PBMR using a Pd/A120 3 composite dense membrane, a PBMR using a mesoporous alumina membrane and a CMR in which the mesoporous alumina membrane was itself rendered catalytic during sol-gel processing by the incorporation of catalytic components. Though the PBMR using the Pd/alumina composite membrane showed the best performance it proved to be sensitive to temperature cycling. The CMR improved the formaldehyde yield by on the average 10%, which in the lower temperature region corresponds to almost doubling the attained yield in the conventional reactor. Chai and coworkers [62] have recently used mesoporous alumina membranes to study the CH 4 steam reforming reaction. The mesoporous alumina membranes were impregnated by a variety of catalytic metals. As can be seen in Fig. 11.6 under the same experimental conditions including the same space velocity and catalyst bed weight and the same sweep gas flow rate the PBCMRs, in which the membrane was impregnated by Ru and Rh, showed significant better performances than the PBCMRs using Pd or Pt impregnated membranes and also the PBMR using the original unimpregnated mesoporous alumina
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTICMEMBRANE REACTORS
100 . . . . . .
,
--
,
.......
,
541
.......
fpj.,.
80
C
.o_ 60 r C
8 4o
o
20
0 200
300
400
500
600
Temperature I "(3 Fig. 11.6. Effect of temperature on reactor conversion with all other experimental conditions the same. The catalyst hi the bed is Ru/A1203. e, Rh-A1203 membrane; A, Ru-A1203 membrane; O, Pd-A1203 membrane; V, Pt-A1203 membrane; A, A1203 membrane; r-l, conventional PFR reactor. R e p r o d u c e d from Chai et al. [62] with permission.
membrane. The study of C H 4 steam reforming in a membrane reactor using a porous alumina membrane was reported some years back by Vasileiadis et al. [63]. This group is studying the application of membrane reactors to the CH4 steam reforming reaction in the context of the Chemically Recuperated Gas Turbine (CRGT) concept, which we will briefly describe later in this chapter. Porous membranes, in contrast to their dense counterparts, allow other species to penetrate through in addition to the preferentially permeable component. Typically this, all other things being equal, will result in some loss of attainable yield (see for example Fig. 11.5). Reactant permeation through the membrane, furthermore, negatively impacts on the economics of the membrane reactor processes because of the need for additional downstream separation requirements, which membrane reactors promise to eliminate in the first place. The latter problem is a serious one and can only be improved upon by the use of more permselective membranes. Several such membranes are currently under development by various groups. The earlier efforts involved the use of conventional sol-gel techniques [14] for the deposition of microporous A 1 2 0 3 (as well as zirconia and titania [64,65]) films within the pores of various mesoporous and macroporous alumina supports. Some of the resulting films proved to be prone to densification and sintering. Techniques to avoid this involve adding various additives, which are thought to inhibit the sintering process [66]. A number of groups are experimenting with techniques which
542
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
involve the use of various organic templates [67]. In the final preparation stage the organic additives are burned away to create the porous structure. Recently, microporous silica (and silica-titania) membranes have also been developed by de Lange et al. [131,132] following Uhlhom [140]. These are reported to combine high hydrogen fluxes with large separation factors. We have mentioned in passing zirconia and titania porous membranes. These membranes are finding applications in ultra and nanofiltration but little use, if any, in membrane reactor applications. A type of microporous membranes that have attracted considerable attention are zeolite-type membranes. These membranes are made by hydrothermal techniques traditionally practised in the area of zeolite synthesis. The earlier efforts involved the preparation of unsupported membranes [68]. Most recent efforts, and they are many, involve the deposition of zeolite films on various macroporous and mesoporous supports including porous metals [69-70] and various aluminas [71-77]. The preparation of silicalite, ZSM-5 and various A type zeolites has been reported. Zeolite membrane preparation represents a very active new thrust for the field of inorganic membranes one that has attracted significant attention. A number of reports exist on the measurement of the transport properties of such membranes and recently even some membrane reactor applications [78] have been reported. Our fundamental understanding of these materials still remains incomplete and much more remains to be learned. Zeolites being catalytically active systems care must be exercised in their catalytic reaction applications. Zeolites like ZSM-5 are prone to coking due to their tendency to catalytically crack various hydrocarbons. Being polycrystalline materials one must strike a fine balance between intra- and intercrystalline transport. Due to their nature these materials are sensitive to poisoning, plugging of the narrow pores due to coking (particularly true for the zeolites with the finer pores) and metal atom/ion counterdiffusion from the underlying mesoporous/macroporous support layers. Without doubt zeolite membranes represent a very new exciting research area. We will return to the discussion of micro and mesoporous membranes but before we do so we will discuss a number of recent studies which utilize macroporous alumina membranes, where the membrane itself acts as a means of bringing together reactants flowing on its opposite sides and to create a well controlled reactive interface. Earlier efforts involved the use of macroporous membranes as efficient contactors for various gas/liquid phase reactions. These studies were motivated by the desire to improve upon the mass transfer characteristics of multiphase reactors. Mass transfer in these reactors is one of the most important parameters controlling efficiency. The volatile reactant limits the reaction extent due to its poor mass transfer through the liquid film covering the catalyst particles. Strategies that have been used to overcome this limitation include varying the catalyst size and shape, the operating parameters
11
-
-
CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
543
and the reactor configuration. Typically one utilizes, slurry, trickle bed, fluidized bed and cross flow reactors. The application of a membrane reactor to multi-phase catalytic reactions aims to improve upon the mass transfer characteristics of conventional reactors. Here the role of the membrane, is to increase the gas-liquid-solid interface through the high surface of the porous membrane acting as catalyst support. Akyurtlu et al. [79] and Harold and coworkers [80-81] used this concept first for hydrogenation reactions. Cini and Harold studied [81] ~-methylstyrene hydrogenation to cumene. The membrane consisted of a ~-A1203 layer deposited on an ~-A1203 tube and impregnated with Pd. Hydrogen passed through the active layer side whereas the liquid phase passed along the support side. The ceramic tube was completely filled with liquid and the reaction occurred within the catalytically active layer. The membrane reactor allowed for good temperature control and a well defined reactants/catalyst interface. Cini and Harold report significant rate enhancements when compared with a single pellet reactor. Torres et al. [82] used a similar type reactor for the nitrobenzene hydrogenation reaction. Their results suggest that depending on the conditions, the conversion can be either kinetically or mass transfer (hydrogen or nitrobenzene) controlled. Under diffusional limitations the best reactor configuration was the one where the volatile reactant was supplied in the reactor inner side. Van Swaaij and coworkers [5,6] proposed the use of nonpermselective macroporous membranes for gas phase reactions, which by their nature require strict stoichiometric conditions. Such reactions include selective catalytic reduction of NO by NH3, and the Clauss reaction. The principle for the CNMR operation is shown in Fig. 11.7. By creating a sharp reaction front within the membrane one avoids the slip by either reactant (NH3 or N O , SO 2 or H2S) on either side of the membrane. Small perturbations in the feed could be accommodated in principle by a shift in the reaction front within the membrane. The success of the CNMR concept depends, as one would expect, on the sharpness of the reaction front created within the membrane. For a non-instantaneous reaction the front created is rather diffuse (see Fig. 11.7) and there is, as a result, a reactant slip. A number of research groups [83-86] have used a rather related concept for carrying out various selectivity limited reactions in a membrane reactor. The concept is illustrated schematically in the top part of Fig. 11.8 which is from a study by Harold and coworkers [84]. It applies to partial oxidation reactions where the desired reaction product can further react with oxygen to produce an undesirable total oxidation product. In some instances it makes better sense (in terms of maximizing the reactor yield) to feed the two reactants separately on either side of the membrane rather that to co-feed them on either side. This is shown in the bottom part of Fig. 11.8 which shows the yield to the desired product as a function of the Thiele modulus as the degree of feed segregation
544
11 -- CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
THE CNMR
2H2S+
stal gas
SO 2 ~
--~ S 8 -!- 2H20 3
stagnant
t
/ / ~ ' ~ . , . , ~ ~
gasfiim -~::~
Ei2S
0'.0
~
9 0'.2 " 0'.4
0'.6
-----
2
0'.8
1.0
Iocadon(~..,)
stagnant
gasfilm
t
stagnant gasfilm
H20
U.UU t
0.0
~
0.2
0.4 0.6 location (x/L)
0.8
1.0
Fig. 11.7. Calculated mole fraction profiles for the CNMR. The upper figure is under conditions for which the reaction is instantaneous; in the lower figure the reaction is non-instantaneous and the front created is rather diffuse. Reproduced from Sloot et al. [5] with permission.
changes (higher values of CAindicate a higher degree of segregation with CA 0 corresponding to the completely segregated feed and CA ~ ~ to the total co-feed). For large values of the Thiele modulus the segregated feed clearly shows a better performance. The PBMR equivalent of the concept of Fig. 11.8, where a catalytically inactive membrane in contact with a catalyst bed is utilized has been applied to direct CH4 activation for the production of C2+ hydrocarbons [85,86].
11 ~ CURI~NT DEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
545
A + v B B ~ VR1R A + VR2 B ~ VpP Feed" A, diluimt
,~CTI:VE'
' Feed: B, diluent
Active layer side
Support side
bulkstreamlA n A t ~ / / / I / ~ /~J i B~pR bullkstream Emue
EITIuent
Reaction Structure:
Example:
A + B ..... >
R
A + R ..... >
P
A: oxygen B: ethylene R: acetaldehyde P: carbon oxides, water
.8
-x_ ;>., -
.
~
.
.
.
.
.
.
.
.
0.4 ,
>" ~,r
.
0.2
"" ,?.J #..,
"
.
1
0.0 "
-0.4 -0.8
.I
i
i0
I00
Thiele ~lodulus (Rcn. 2 Basis), ~,_ Fig. 11.8.Theuseof membranereactorsin partial oxidationreactions.Upper figure representsa
schematicof theconcept.Lowerfiguregivesthecalculatedyieldasaftmctionof theThielemodulus. Reproduced from Harold et al. [84] with permission.
Some of the major challenges facing the area of catalytic m e m b r a n e reactors are of a more conventional nature. We have briefly discussed the need for standardization in the m e m b r a n e preparation techniques and for better quality control; the preparation of inexpensive good quality macroporous supports still remains a challenge as are reliable chemically inert and cost effective seals (an excellent discussion on the latter topic can be found in the review paper by Saracco and Specchia [14]). For highly exothermic (endothermic) reactions
546
11 m CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
providing the required heating (cooling) poses some interesting technical questions, especially for multi-tubular porous membranes. These are some, but not necessarily all, the technical hurdles that loom in the horizon we face as we move away from the laboratory and bench-scale studies towards process development units (PDU) and commercial-scale installations. Before we proceed further with our discussion focusing on dense, solid oxide type membranes we would like to mention a class of porous membranes, which though not strictly inorganic in nature, they themselves show good potential for high temperature membrane reactor applications. These are carbon membranes [87,88]. Carbon membranes are, of course, not a recent development. Koresh and Softer [89-92], for example, prepared carbon membranes with molecular sieve type properties more than ten years ago. Mesoporous and macroporous carbon membranes are available commercially by GFT mostly for various liquid and gas phase separations. The first known application of carbon membranes in catalytic membrane reactors was the one reported recently at the First International Workshop on Catalytic Membranes [93] for Fischer-Tropsch synthesis. There the carbon membranes (in the form of hollow fibres) were used as a means to provide a high surface area catalytic support for the reaction. Carbon membranes are very exciting materials, since by a relatively simple modification of the preparation technique one can prepare materials, which span the full range from molecular sieves to macroporous membranes. Carbon membranes are not free of limitations however. They cannot be utilized, for example, in oxidative atmospheres; caution should be, furthermore, exercised even under reductive atmospheres in the presence of various metal impurities. Nevertheless carbon membranes, in our opinion, show great potential for use in catalytic membrane reactors especially for the CNMR type applications we have already discussed. 11.4 SOLID OXIDE M E M B R A N E S
Another class of dense inorganic membranes that have been used in membrane reactor applications are solid oxide type membranes. These materials (solid oxide electrolytes) are also finding widespread application in the area of fuel cells and as electrochemical oxygen pumps and sensors. Due to their importance they have received significant attention and their catalytic and electrochemical applications have been widely reviewed [94-98]. Solid materials are known which conduct a variety of cationic/anionic species [14,98]. For the purposes of the application of such materials in catalytic membrane reactor applications, however, only 0 2- and H + conducting materials are of direct relevance. The earlier solid oxide electrolytes were solid solutions of divalent or trivalent metal oxides ( Y 2 0 3 , YB203 o n C a O ) in oxides of tetravalent metals having a fluorite type structure A408 like ZrO2, T h O 2 or C e O 2 [97]. The introduction of
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
547
aliovalent cations in the lattice creates oxygen vacancies. Yitria and Calcia stabilized ZrO 2 (YSZ or CSZ) in particular have found most of the applications. Pyrochlore type electrolytes with an A2B207 structure and Bi20 3 based systems have received lesser attention. Solid oxide membrane reactors have been operated in three different modes. Their primary use so far has been as solid oxide fuel cells (SOFC), where the chemical energy of the reaction is converted into electrical energy (with the co-production in some instances of useful chemicals [99,100]). The reverse of this case is when the solid oxide membrane reactors are operated as electrochemical oxygen pumps, whereas electrical energy is expended to enhance the oxygen transport towards (or away from) the anodic membrane surface. With the development of perovskite type membranes, which exhibit good ionic and electronic conductivities, reports are appearing with a greater frequency where solid oxide membrane reactors are operated in the conventional mode without the need for imposing an external electric voltage. ABO3 type perovskites either with a 3:3 (like LaA103) or 2:4 (like SrTiO3) structure can accommodate various metals in the A, B position. Their conductivity can vary from totally electronic (metallic-like) like in LaNiO3 to ionic as in doped LaA103 [97]. Materials like BaCe0.9Gd0.102.95 and BaTh0.9Gd0.102.95 have good protonic conductivity at low temperatures (~10-~ s/cm -1 at 600~ Acceptor doped perovskites have received the greatest attention. Cation substitution in the A or B position significantly increases anion disorder. Perovskites with a structure like Lao.2Sro.sCOl_yFeyO3_xor Lao.6Sro.4Coo.sBo.203-x(B=Fe, Ni, Cu) have been shown by Yamazoe and coworkers [101-104] to have high ionic conductivities (~1 s / c m -1 at 800~ but also very highelectronic conductivities (,-'100 s/cm-1). For such high ionic conductivities gas-phase exchange seems to often be the controlling mechanism [97]. Hydrogen and CH 4 (and most recently CH3OH) SOFCs have been the subject of intensive investigations since the early sixties and a complete review of the area certainly goes beyond the scope of this chapter. Hydrogen SOFCs have received the lion's share of attention. CH4 (and more recently CHBOH) SOFCs have shown promise but concerns still remain with carbon deposition and low catalytic activity. Solid oxide membrane reactors operating as electrochemical oxygen pumps have been the subject of several laboratory investigations. Reactions studied include CH4 reforming and partial oxidation to synthesis gas; CH4 direct activation for the production of C2 and higher hydrocarbons and partial oxidation to CH3OH and formaldehyde, various decomposition reactions like those of CO2 and H20, NO and SO2. The latter two decomposition reactions have received attention by a number of research groups in the USA due to their importance in environmental applications. The earlier work was done by Mason and coworkers [105,106] and Giir and Huggins [107] who used various stabilized zirconias to study the decomposition of NO. YSZ electrolytes were
548
11 m CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTICMEMBRANE REACTORS
utilized by Ceramatech for NO and SO2 decomposition [108]. Most recently Helipump Corporation and Research Cotrell [109] have reported on the use of yitria stabilized ceria solid electrolytes for the SO2 and NOx reduction from a simulated flue gas from coal combustion containing up to 2% 02. Good rates of NOx and SO2 reduction were reported in the presence of a strodium ruthenate perovskite electrocatalyst. Fly-ash which is present in coal combustion flue gas seemed to have little effect on the solid electrolyte performance. An assortment of other reactions like various hydrogenation reactions, the production of HCN from CH 4 and NH3, etc. have also been studied [98]. Problems resulting from the estimated cost of the required electric energy and difficulties in scaling up the laboratory reactors [110] have hindered further commercialization. For a number of chemical reactions it has been shown recently that by cofeeding oxygen on the anodic side the rate of the electrochemical reaction far exceeds the rate of electrochemical oxygen pumping [111]. This effect called NEMCA (non-faradaic electrochemical modification of catalytic activity) shows promise in reducing the associated electrical energy costs. A significant recent effort in this area is a collaborative study by Amoco and the Argonne National Laboratory utilizing solid oxide type membranes [112113]. The newly developed membranes show improved mechanical and thermal characteristics and are reported to remain stable for over 21 days at 900~ under CH4 partial oxidation conditions. The membrane used was tubular in shape. A CH4/Ar mixture was allowed to flow in the tubeside which was packed with a Rh based catalyst. Air was the source of oxygen on the outside
o~" 9 6 "6
92
"~ =o
84
80
:>
=o 0
* H 2 SelJ2 9CO Sel.
7
~k 0 2 Perm.
6
9CH4 Conv.
76 = _0.o 72 e
~-
>
60
e
E
1
o
"~ r -5 E '-
68 ~ _ 64
o t--
4o
~ 3
5
7
9 11 13 Time (days)
15
17
19
21
3
I::::
>, x 0
Fig. 11.9. Methane conversion, CO mid H2 selectivities and 02 permeation in a solid oxide m e m b r a n e reactor. R e p r o d u c e d from Balachandran et al. [113] with permission.
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
549
of the membrane. Figure 11.9 shows the CH4 conversion and CO/H2 selectivities during a 21 day run. They all remain in the 90+% range throughout the whole run. The use of solid oxide membranes in partial oxidation reactions aims to avoid the complete oxidation of the desired partial oxidation products. When compared to similar efforts using microporous membranes solid oxide membrane reactors, are at a disadvantage (except for reactions that take place at high temperatures) because oxygen transport through the oxide lattice is generally low when compared with the permeability of porous materials. The synthesis of non-porous ceramics with good oxygen permeability and selectivity at lower temperatures is not a simple task. The application of membrane reactors to partial oxidation is often complicated by the fact that the desired product is often more reactive with oxygen than the reactant itself as was observed by Julbe et al. [114,115] for the methane oxidative coupling using lanthanum oxychloride membranes.
11.5 THEORETICAL C O N S I D E R A T I O N S
The modelling and simulation of catalytic membrane reactors has attracted the interest of many investigators over the last ten years. Most studies have focused on particular membrane reactor systems aiming to simulate their performance in terms of attainable yield and selectivities. The considerable body of modelling work in this area has been reviewed by Tsotsis et al. [13]. The authors of this paper presenteda discussion of the pre-1993 modelling literature in terms of a general mathematical model of a PBCMR, which is shown schematically in Fig. 11.10, with catalytic beds present both in the inner and outer membrane regions. The model takes into account mass and energy balances in the tubeside
~"
OUTER TUBE
z IS
~MBRANE
TUBESm~
CATALYST BED
SHELLSIDE CATALYST BED
Fig. 11.10. Schematic of m e m b r a n e reactor for PBCMR model. R e p r o d u c e d from Tsotsis et al. [13] with permission.
550
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
and shellside and in the membrane itself and accounts for the existence of pressure drops in the shellside and tubeside. The membrane is considered to consist of a single permselective layer either dense or mesoporous following a Knudsen type diffusion mechanism, an assumption utilized by most pre-1993 modelling investigations. There are a number of modelling efforts, however, which cannot be discussed in the context of the generalized model of Tsotsis et al. [13]. Van Swaaij and coworkers, for example, have modeled the behaviour of CNMR reactors using the Dusty Gas Model description of transport [5,6]. They have shown that when bulk diffusion and convective flows must be taken into account the Dusty Gas Model provides a more accurate description of transport through the membrane. The earlier studies of the group modelling the application of CNMR to reactions requiring strict stoichiometric ratios have been reviewed in detail by Tsotsis et al. [13]. More recent efforts by the same group deal with the application of the CNMR Dusty Gas Model to the combustion of CO and hydrocarbons. Membrane reactors utilizing multilayered membranes have been modelled in recent studies by Becker et al. [116] and Tayakout et al. [117,118]. In contrast to prior efforts these models account for mass transport both through the mesoporous permselective layer and the underlying macroporous support layer(s). Both are isothermal models. Becker et al. [116], however, in their analysis utilized the experimentally measured temperature gradients along the reactor length in the calculation of reaction constants and transport coefficients. Both models assume dilute reactant mixtures and, therefore, neglect complications resulting from changes in the number of moles due to the reaction. The reaction studied was ethylbenzene dehydrogenation in the model of Becker et al. [116] and cyclohexane dehydrogenation in the model of Tayakout et al. [117,118]. A schematic of the reactor analyzed by both groups is shown in Fig. 11.11 (in the model of Becker et al. [116] there is no catalyst bed in region 4). At steady-state Becker et al. [116] write the following set of equations.
On the tubeside (region 1)
3C~ 1 3 ! 3Ct] rl U T --~- = D t r -~r r ---~-r ) p b O~i O<_r<_r i
or
(11.1)
O
where C~ is the concentration of species i in region I (ethylbenzene, styrene and hydrogen in the Becker et al. model; cyclohexane, benzene and hydrogen in the Tayakout et al. model). D~ is an effective bed radial diffusivity, Pb the reactor bed density (g/cm3), c~i the stoichiometric coefficient and r 1 the reaction rate (gmol/g s) in region 1. UT the superficial fluid velocity is considered constant in the model of Becker et al. [116] (an assumption which is relaxed in the model of Tayakout et al. [117,118], see discussion to follow).
11
-
-
CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
shell side
4
membrane
2 ~, 1
OO._.O % 0 0 0 0 0 0
O O U O
--- ,.,
0
0
0
0 0
i
i
I
O 0 F%00C)~o0
- -/'hii
J i
O 0 " ~ u 0 " - ' 0 0 ...........
iiltt-i-
iil i ii ! iiill
oo o oooO~
oo
oooo
Z O .............................................................................
O
l -I'-~. . . . . . . . . ]~
feedside
551
catalyst packed bed
~
k
active layer
Fig. 11.11. S c h e m a t i c of the m u l t i l a y e r e d m e m b r a n e reactor.
In the reactive membrane (region 2) 1 3 ( 3C21= r2 D2 r ~ r ~ j (xi pm r i <__
(11.2)
r <_r m
with Pm being the membrane density. Becker et al. [116] utilize a Fickian description of transport both in the mesoporous and macroporous (see below) membrane regions with D~ (D 3) being an overall effective transport coefficient. In the macroporous support (region 3) D 3 1 3 I 3C31= r ~ r r--~-r J 0
(11.3)
rm <_r < rs On the shellside of the reactor (region 4) 3C 4 1 3 [ OC4/ Us - - ~ - D 4 r -~r r --~-r ) rs <-r < re
and
(11.4) 0 < Z _
Equations (11.1)-(11.4) are complemented by a set of corresponding initial and boundary conditions.
11 -- CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
552
A t Z - 0 (reactor inlet)
c
=cb
(11.5)
c4-c4 o
(11.6)
At r - 0 (symmetry condition) -0
3r
At r
-
re
(11.7)
(wall no-flux condition) 4
(11.8)
3r
In addition to Eqs. (11.5)-(11.8), one must account for the continuity of fluxes and concentrations at the interfaces between the various regions. Becker et al. [116] first converted the system of Eqs. (11.1)-(11.8) into an equivalent system of dimensionless equations by defining appropriate dimensionless variables like C~/Clo, r/ri, z / L , etc. By discretizing the first and second order derivatives the dimensionless equations were reduced to a mixed system of first order ordinary differential and algebraic equations which were solved by the DASSL numerical package [116]. The model was used to fit experimental data with the ethylbenzene dehydrogenation reaction. The agreement between experimental data and theory was good (see Fig. 11.12). As previously mentioned for fitting the data of Fig. 11.12, Becker et al. [116] incorporated in their model the experimentally measured temperature gradients along the reactor length.
100
Qs 0.75 cm=/s 80
0 *~ i-
t
80
> C 0 0
9 Experimental --
0.2
o.,
o's
o's
Simulated
1?o
Or (cm31s)
conversion conversion
1.'2
~.,
1;,
Fig. 11.12. C o n v e r s i o n vs. v o l u m e t r i c tubeside flow rate for a PBMR. Qs is the shellside flow rate. R e p r o d u c e d f r o m Becker et al. [116] w i t h permission.
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
553
The model of Tayakout et al. [117,118] in addition accounts for the possibility of axial dispersion effects in the tubeside and shellside. The inclusion of axial dispersion effects in regions (1) and (4) necessitates a different set of initial conditions at Z = 0 and a companion set of conditions at Z = L. The effect of pressure drop through the catalytic bed could be included in this type of model using Ergun's equation. A nonisothermal membrane reactor for the partial oxidation of C H 4 t o synthesis gas utilizing a multilayered membrane was recently presented by Tsai et al. [119]. The membrane is considered to consist of three layers. A macroporous support layer, a dense perovskite film permeable only to oxygen on the top of which a porous catalytic layer is placed. In modelling such a reactor Tsai et al. [119] distinguish five regions. These are the tubseside and shellside of the membrane and the three membrane layers. For each region Tsai et al. write the appropriate mass and energy balance equations. For the tubeside and shellside regions the mass balances are similar to Eqs. (11.1) and (11.4) above. For the macroporous support region the equation is the same with Eq. (11.3) and for the catalytic layer the same with Eq. (11.2). The dense perovskite layer allows only oxygen to transport through and Tsai et al. [119] opt for a Fickian description of transport with an activated diffusion coefficient. In cylindrical coordinates they write the following equation from the flux of 02 No2r (km~ s) through the dense perovskite layer.
NO2"
A exp I~T] T rln
r/~b)
In (P~II~/
(11.9)
)
In the above equation r is the radial position within the perovskite layer, r0 and rb the outer and inner radii and Po, ro and Po2,rb, the corresponding oxygen partial pressures at these positions. The preexponential factor A and the activation energy E were fitted to the experimental data of Teraoka et al. [101]. The energy balance equations in every region account for energy changes due to the flow and diffusional transport of the various species and the energetic effects associated with the reaction. Tsai et al. [119] describe C H 4 partial oxidation to syngas as the direct outcome of the total oxidation of C H 4 coupled with CO2 and steam reforming. The model of Tsai et al. is of direct relevance to the experimental study of Balachandran and coworkers [112,113] that we have previously discussed. The same group [120] have also recently presented a modelling study of the application of solid oxide membrane reactors in the area of environmentally benign processes. With the emergence of solid oxide membranes and their use in membrane reactor applications, a number of models have appeared recently to
554
11 -- CURRENT DEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANE REACTORS
model such reactors. Wang and Lin [121] and Lu et al. [122] presented models of m e m b r a n e reactors utilizing solid oxide membranes for C H 4 activation; so did Nozaki and Fujimoto [123] to model their data of selective oxidative methane coupling in a m e m b r a n e reactor utilizing a PbO oxide m e m b r a n e impregnated on a porous silica-alumina tube. In addition to the studies of Becker et al. [116], Tayakout et al. [117,118] and Tsai et al. [119], a n u m b e r of other groups have presented models which take into account the possibility of external mass transport limitations [124-126]. Such effects could become of concern in industrial m e m b r a n e reactors utilizing larger size membranes. Several other recent modelling m e m b r a n e reactor studies are also w o r t h discussing, Varma and coworkers [127] have analyzed the effect that nonuniform catalyst distribution on the m e m b r a n e itself (for CMR and PBCMR applications) and in the catalyst bed (for PBMR applications) has on m e m b r a n e reactor performance. Conventional m e m b r a n e reactor models were utilized by a n u m b e r of groups to model their experimental data. Shu and co-workers [33]
0
membrane not important
membrane
important
r,.
o ~D A V r
In
-I-
0
,r
tern
0
r" 0
e...,
u)
0 r
t_
(1) o > 0
L)
C) qr--
o
1 10 100 P e r m e a t i o n / R e a c t i o n rate (H)
1000
Fig. 11.13. Cyclohexane conversion vs. the (permeation/reaction rate) ratio. Curves 1 and 2 for m e s o p o r o u s membranes with Knudsen separation factors. Curves 3 and 4 for microporous m e m b r a n e s with a separation factor of 100. Curves 5 and 6 for membranes permeable only to hydrogen. O d d (even) n u m b e r e d curves correspond to an hlert sweep gas rate of I (10) times the cyclohexane flow. The temperature is 477 K, Pfeed= 100 kPa. From Harold et al. [130] with permission.
11 - - CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
555
modeled the data in their studies of CH 4 steam reforming in a catalytic membrane reactor using a Pd-Ag membrane deposited on porous stainless steel. Similar models were used by Gobina and Hughes [128] to model alkane dehydrogenation in a catalytic membrane reactor using a Pd membrane. Gokhale et al. [129] presented a model for isobutane dehydrogenation in a membrane reactor using a porous membrane. Some recent models have also appeared discussing the operation of three phase catalytic membrane reactors by Torres et al. [82]. The models which represent extension of prior models by Akyurtlu et al. [79] and Cini and Harold [80] are numerically analyzed and appear to simulate well the experimental results of the nitrobenzene hydrogenation reaction in a three phase catalytic membrane reactor. Some of the important findings of the modelling studies so far are nicely summarized in Fig. 11.13 which is from a paper by Harold et al. [130]. This figure is the result of a modelling investigation of a PBMR for the cyclohexane dehydrogenation reaction but the main findings, are generally applicable. Curves (1) to (4) correspond to meso/microporous [131,132] membranes and curves 5 and 6 to microporous membranes which are only permeable to hydrogen. Note that for porous membranes the conversion passes through a maxim u m the result of the detrimental influence of reactant losses; this is not true, however, for the perfect (i.e., only permeable to hydrogen) membranes. Such membranes (i.e., dense metal) on the other hand have problems of their own which we have already discussed. As one would expect the higher the permselectivity of the membrane to hydrogen the better the performance of membrane reactors especially in the region of higher (permeation/reaction rate) ratios.
11.6 EMERGING APPLICATIONS
Though most of the applications of membrane reactors that we have discussed so far involve the production of useful chemicals some of the most interesting applications currently under consideration relate to energy production, transfer and utilization. High temperature membrane reactors are, for example, being looked upon in the context of Chemical Energy Transmisssion Systems (CETS). Such systems utilize reversible catalytic reactions associated with large energy effects, typically hydrogenation/dehydrogenation type reactions. During the endothermic reaction (dehydrogenation) the heat of the reaction is provided by an energy source (like solar, geothermal, etc). During the exothermic reaction the heat (which in the endothermic part of the cycle is stored in the form of chemical energy) is released back to the user. To increase the energy efficiency of the process the hydrogenation reactor pressure must be increased with
556
11 -- CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS T
t
iIi I F c
1
~
Rdh
/i I ~3
9
Rh
HE2
,
~
1
F6
j
I -
I
/ CHE1
M4
,r
I
HE5
j
'
7
12
,
"
t C3
i
J
Fig. 11.14. Process flow sheet of cyclohexane/benzene heat pump using hydrogen permeable membranes: Reh and R4rdehydrogenation and hydrogenation reactors; C, compressors; T, turbine; HE, heat exchangers; CHE, counter-current heat exchangers; P, liquid pump; M, hydrogen membranes. Reproduced from Cacciola et al. [133] with permission.
respect to the dehydrogenation reactor pressure. High temperature hydrogen permselective membranes offer many advantages here. Figure 11.14 is from the study of Cacciola et al. [133] and represents a conceptual design of a CETS system involving a high temperature membrane reactor utilizing the cyclohexane dehydrogenation reaction. The process utilizes four hydrogen permselective membrane modules and a condensation/evaporation unit in addition to the two reactors, compressors and associated other hardware. The dehydrogenation reactor containing the high temperature membrane increases the yield of the dehydrogenation reaction while simultaneously allowing the removal of product hydrogen. To further increase the coefficient of performance (COP) Cacciola et al. [133] propose the use of three additional membrane modules. The mixture of benzene, unreacted cyclohexane and the remainder of hydrogen exiting the dehydrogenation reactor Rdh are cooled down in CHE1 and condensed in HE4
11 - - C U R I ~ N T D E V E L O P M E N T S A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
557
releasing energy; hydrogen is separated in HE4 by means of the membrane module M4 and mixed at point 14 with the hydrogen coming form Rdh. This hydrogen is then compressed to the level of its partial pressure in the hydrogenation reactor Rh and cooled down in the heat exchanger HE1 releasing additional energy; it is then mixed into the excess hydrogen circulating in the loop 16-9-10-19, which includes the dehydrogenation reactor Rh. Note that Cacciola et al. [133] envision Rh to be a conventional reactor with the membrane module M3 acting as a simple mixing device of the hydrogen with the cyclohexane and benzene vapour from the evaporator HE5 and heat exchanger CHE2; in fact M3 and Rh could be combined into a single unit as is true with Rdh and M2. The mixture exiting Rh is passed through the membrane module to partially remove the hydrogen, the retentate (consisting primarily of cyclohexane and some benzene and hydrogen) is cooled in CHE2 expanded in turbine T and heated to the dehydrogenation temperature Tdh to complete the cycle. The role of the membrane reactor is obvious. The membrane module M1 is important because it decreases the total amount of hydrogen flowing through Rdh; this in turn decreases the ratio of reactor pressures Ph/Pah, conversions (Xdh, Xh) and temperatures (Tdh, Th) remaining the same. M3 is also important from an energy efficiency point of view, since it allows one to keep the pressure at 17 at the level of the hydrogen partial pressure at Rh (rather than the total pressure Ph) thus reducing the load on compressor C2. It should be obvious from the above discussion that though the ability of the high temperature membrane reactor to increase the reactor yield is important equally important is the membrane's ability to efficiently separate hydrogen at high temperatures and to sustain a transmembrane pressure drop. A somewhat related concept is the Chemically Recuperated Gas Turbine (CRGT), see Fig. 11.15 [63,134]. Here C H 4 is being reformed in a catalytic membrane reactor. The purpose of the reforming, however, is not to completely convert C H 4 to H 2. Rather the goal is to convert C H 4 to a degree sufficient to create a fuel mixture with better flaming characteristics which would then allow higher steam to fuel ratios lower burner temperature and reduced NOx production. The success of this concept is strongly dependent on its ability to utilize waste heat, thus the need for efficient membrane reactor reformers. Several of the membrane reactor efforts under development are motivated by environmental regulations and factors. We have already briefly described the CRGT concept and the solid oxide membrane reactors for SO2 and NOx removal. With the growing awareness and public fear for the risk of global warming and climate change have come calls for the reduction of greenhouse gases, particularly CO2 from stationary power sources. This has led to the initiation of various research efforts for the development of high temperature membranes for H 2 - C O 2 separation in the context of integrated coal gasification combined cycle (IGCC) and molten carbonate fuel cells (IGMCFC) technology.
558
11 m CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
B )_ Air
650~ _
' to' ~
u2o C H 4.
[
,,
c
C H 4 + H 2 0 ---> C O + 3 H 2
Energy of Combustion (BTU/ft3)
1000
300
999
Fig. 11.15. T h e C R G T c o n c e p t .
Figure 11.16a, from a recent study by Jansen et al. [135], is a conceptual flow sheet of an IGCC power plant with CO2 removal; Fig. 11.16b is the corresponding flowsheet for an IGMCFC system. The plant of Fig. 11.16a consists of four basic subsystems; the gasifier, a high temperature cleaning system (HTG), a subsystem for CO2 removal (which includes a CO shift reactor, a high temperature hydrogen permselective membrane module and a CO2 compression/liquefaction unit) and the gas turbine together with the Heat Recovery Steam Generation System (HRSG). The study by Jansen et al. [135] considers a conventional Shell entrained bed gasifier operating at 30 bar, 1500~ and utilizing 95% pure oxygen. Though, to the best of our knowledge, not currently under development, the potential for adopting solid oxide membrane technology to gasification processes should be investigated. The conventional shift reactor of Fig. 11.16a can be replaced by a membrane shift reactor, as Jansen et al. themselves suggest. The basic difference between the systems in Figs. 11.16a and 11.16b is that the hydrogen burning turbine is replaced by a molten carbonate fuel cell system (MCFC). In the process of Fig. 11.16b the shift reactor and the membrane unit are placed after the MCFC unit. Jansen et al. [135] estimate energy efficiencies from the ICGG and IGMCFC systems with CO2 removal (assuming a 95% hydrogen recovery in the membrane modules) of 37.5 and 47.5% as opposed to 46.4 and 53.1% without CO2 removal. The energy efficiency losses (and the incremental additional per kWh costs) are small enough to warrant further study for the development of more efficient membranes.
11 - - C U R R E N T D E V E L O P M E N T S A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
559
(a)
I I !
! I
I
I
_~
, I
I
!
| . . . . .
!
(b)
~_~o._~~~~., I~,,, i
,
i
.
J
.2
.
I
'
. . . .
f~ i _ 7 " " } I
~l,
~ ......... -' i . . . . .
I
/
I ! 1~, , "-~ i'- j'~~ ~ ~~i",
'1
,, -I
-'
~
t
~-~!--'~ ~ , '
I ~ 1 r162
A
C
-
_
~:. ~
Fig. 11.16.Upper figure (a) is a simplified flowsheet of an IGCCpower plant ,withCO2 removal; lower figure (b) the same for an IGMCFCplant. Reproduced from Jansen et al. [135]with permission. An interesting application of catalytic m e m b r a n e reactors [14,136] relates to the production of tritium which together with deuterium will be the fuel for the fusion reactors of the future. Tritium is p r o d u c e d by means of a nuclear reaction between neutrons and lithium atoms in a breeder reactor. The tritium thus p r o d u c e d m u s t be further purified to reach the purity levels that are required in the fusion reactor. For the extraction and purification process Basile and
560
11 - - C U R R E N T D E V E L O P M E N T S A N D FUTURE R E S E A R C H IN CATALYTIC M E M B R A N E REACTORS
coworkers [136] propose the use of two membrane reactors. In the first reactor using a hydrogen permselective membrane, tritium is removed from a He stream with the aid of an oxidation reaction. The water thus produced is subsequently decomposed in a membrane shift reactor to release the tritium. The use of membrane reactors in homogeneous catalysis is attracting attention and will most likely be a very active research area in the future. Here the porous membrane can act as a solid matrix containing an active catalyst or as a separation barrier for reactants and products for the reactions taking place in the liquid phase. Several such applications have begun to appear using organic membranes [137-139]. It is possible that inorganic membranes will provide some advantages in this area over the organic ones, currently utilized. Multi-phase catalytic reactions have attracted some attention but the area has not in our opinion been fully exploited. Previous studies have demonstrated that the yields obtained with the catalytic membrane reactors are often better than the yields obtained with more conventional reactors. Future research in this area must involve reactions with more immediate industrial applications. Examples of such reactions could be the hydrogenation reactions studied by Gryaznov and co-workers with dense metallic membranes which we discussed earlier. New materials like zeolite membrane could offer some advantages here with their enhanced regio- or chemioselectivity. 11.7 C O N C L U D I N G R E M A R K S
As we hope has become apparent to the reader of this chapter catalytic membrane reactors have been fully demonstrated at the laboratory scale, the most important systems studied being dehydrogenation and oxidation reaction. Though some small industrial installations already exist the concept of coupling catalytic reaction and separation by membranes has yet to find widespread industrial application. The reasons for this and ways to overcome the existing barriers to commercialization are discussed in a recent paper by Saracco et al. [15]. It seems to us that the most important factor hindering further progress are the membranes themselves. Their synthesis and production methods must be improved. Membranes must not only have good permselectivity and flux, they must also be thermally stable, long-term reliable and affordable. Many of the activities in this area give reason for hope. A number of research groups are working on the direct synthesis of microporous solids, until now unavailable on a commercial scale (silica or alumina); others are focusing on the modification of the existing commercial membranes by the addition of very thin metallic or other films producing membranes with good mechanical strength and permeability combined with high selectivity provided by the thin film. Both approaches are very attractive and additional research in this area is certainly worthwhile. One must not forget, however, that by far the overriding
11 m CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
561
factor determining the commercial viability of any new membrane will be its cost and long term reliability under industrial operating conditions and temperature cycling. The synthesis of new thermally stable microporous materials with molecular sieving characteristics like zeolites shows a good potential for gas phase reactions. The availability of good, reliable membranes will not, of course, eliminate the need for optimal reactor design and process analysis, necessary to determine the type of membrane to be used and the optimal operating conditions. As was discussed previously, some reactions do not need permselective membranes. Other process parameters like the reactor configuration or the amount of sweep gas utilized can affect dramatically the observed performance. The use of inorganic membranes in catalytic reactors is providing opporttmity for good fundamental applied research in catalysis and chemical reaction engineering. The coupling of reaction and separation in one unit is an attractive concept but also a complex technological problem. Though for the moment widespread industrial application of the concept is not yet a reality the potential is certainly there. The reactor system more easily amenable to high temperatures industrial application is the PBMR, because in this reactor configuration the membrane does not require extensive regeneration and activation steps. Even for the PBMR applications, however the development of crack-free, uniform, microporous and high temperature resistant membranes is a key step for future developments. Three-phase catalytic membrane reactor systems, in our opinion, show significant promise, for near term application to hydrogenation reactions for fine chemicals synthesis. These reactions generally require mild operating conditions which will place less stringent requirements on the available and future commercial membranes.
Acknowledgements Professor Tsotsis wishes to acknowledge the support of the National Science Foundation. The assistance of Karen Woo and Ravi Kumar in typing the manuscript are also gratefully acknowledged.
REFERENCES 1. 2. 3. 4.
G. Belfort, Membranes and bioreactors: A technical challenge in biotechnology. Biotechnol. Bioeng., 33 (1989) 1047. H.N. Chang and S. Furusaki, Membrane bioreactors: present and prospects. Adv. Biochem. Eng. Bioeng., 44 (1991) 27. E. Drioli, G. Iorio and G. Katapano, in: M.C. Porter (Ed.), Handbook of Industrial Membrane Technology, Noyes, 1990, p. 401. P.R.Brookesand A.G. Livingston,Biologicaldetoxificationof 3-chloronitrobenzene manufacture wastewater in an extractivemembrane bioreactor. Water Res., 28 (1994) 1347.
562
11~ CURRENTDEVELOPMENTSANDFUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
5. H.J. Sloot, G.F. Versteeg and W.P.M. van Swaaij, A nonpermselective membrane reactor for chemical processes normally requiring stoichiometric feed rates of reactants. Chem. Eng. Sci., 45 (1990) 2415. 6. H.J. Sloot, C.A. Smolders, W.P.M. van Swaaij and G.F. Versteeg, High temperature membrane reactor for catalytic gas-solid reactions. AIChE J., 38 (1992) 887. 7. J.N. Armor, Catalysis with permselective inorganic membranes. Appl. Catal., 49 (1989) 1. 8. J.N. Armor, Challenges in membrane catalysis. Chemtech, 22 (1992) 557. 9. S. Ilias and R. Govind, Development of high temperature membranes for membrane reactor: an overview. AIChE Syrup. Ser., 85, 268 (1989) 18. 10. N. Itoh, Dehydrogenation by membrane reactors. Sekiyu Gakkaishi, 33 (1990) 136. 11. H.P. Hsieh, Inorganic membrane reactors. Catal. Rev. Sci. Eng., 33 (1991) 1. 12. V.T. Zaspalis and A.J. Burggraaf, Inorganic membrane reactors to enhance the productivity of processes, in: R.R. Bhave (Ed.), Inorganic Membranes: Synthesis, Characteristics and Applications. Reinhold, New York, 1991, Ch. 7. 13. T.T. Tsotsis, R.G. Minet, A.M. Champagnie and P.K.T. Liu, Catalytic membrane reactors. in: E.R. Becker and C.J. Pereira (Eds.), Computer Aided Design of Catalysts. Dekker, New York, 1993, pp. 471-551. 14. G. Saracco and V. Specchia, Catalytic inorganic membrane reactors: Present experience and future opportunities. Catal. Rev. Sci. Eng., 36 (1994) 305. 15. G. Saracco, G.F. Versteeg and W.P.M. van Swaaij, Current hurdles to the success of high temperature membrane reactors. J. Membr. Sci., 105 (1994) 95. 16. J. Zaman and A. Chakma, Inorganic membrane reactors. J. Membr. Sci., 92 (1994) 1. 17. V.M. Gryaznov and A.N. Karavanov, Hydrogenation and dehydrogenation of organic compounds on membrane catalysts (review). Khim.-Farm. Zh., 13(7) (1979) 74. 18. V.M. Gryaznov, Hydrogen permeable palladium membrane catalysts. Plat. Met. Rev., 30 (1986) 68. 19. J. Shu, B.P.A. Grandjean, A. van Neste and S. Kaliaguine, Catalytic palladium-based membrane reactors: A Review. Can. ]. Chem. Eng., 69 (1991) 1036. 20. V.M. Gryaznov, A.N. Karavanov, T.M. Belosljudova, A.M. Ermolaev, A.P. Maganjunk and I.K. Sarycheva, U.S. Patent 4,388,479 (1983). 21. J.E. Philport, Hydrogen diffusion technology. Commercial applications of palladium membrane. Platinum Met. Rev., 29 (1985) 12. 22. V.M. Gryaznov, O.S. Serebryannikova, M.Yu. Serov, M.M. Ermilova, A.N. Karavanov, A.P. Mishchenko and N.P. Orekhova, Preparation and catalysis over palladium composite membranes. Appl. Catal., 96 (1993) 15. 23. Z.Y. Li, H. Maeda, K. Kusakabe, S. Morooka, H. Anzai and S. Akiyama, Preparation of palladium-silver alloy membranes for hydrogen separation by the spray pyrolysis method. J. Membr. Sci., 78 (1993) 247. 24. S. Uemiya, Y. Kude, K. Sugino, N. Sato, T. Matsuda and E. Kikuchi, A palladium/porous-glass composite membrane for hydrogen separation. Chem. Lett. 1687 (1988). 25. S. Uemiya, T. Matsuda and E. Kikuchi, Hydrogen permeable palladium-silver alloy membrane supported on porous ceramics. J. Membr. Sci., 56 (1991) 325. 26. D.J. Edlund and W.A. Pledger, A study of long-term stability of composite metal membranes. Paper presented at the 3rd International Conference on Inorganic Membranes,
July 10-14, 1994, Worcester, MA, USA. 27. J.K. Ali, E.J. Newson and D.W.T. Rippin, Deactivation and regeneration of Pd-Ag membranes for dehydrogenation reactions. J. Membr. Sci., 89 (1994) 171. 28. A.F.Y. A1-Shammary, I.T. Caga, J.M. Winterbottom, A.Y. Tata and I.R. Harris, Palla-
11~ CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
29. 30. 31.
32.
563
dium-based diffusion membranes as catalysts in ethylene hydrogenation. J. Chem. Tech. Biotechnol., 52 (1991) 571. S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi, Steam reforming of methane in a hydrogen permeable membrane reactor. Appl. Catal., 67 (1991) 223. B.A. Raich and H.C. Foley, Super equilibrium conversion in palladium membrane reactors. Kinetic sensitivity and time dependence. Appl. Cat.A: General, 129 (1995) 167. A.M. Adris, J.R. Grace, C.J. Lim and S.S.E.H. Elnashaie, Fluidized bed reaction system for steam/hydrocarbon gas reforming to produce hydrogen. U.S. Patent Appl. 07965011, Canad. Patent Appl. 2081170 (1992). S. Laegsgaard and P.E. Jorgensen, H. Nielsen and P. Lehrmann, Stream reforming of methane in a hydrogen-permeable membrane reactor. Paper presented at the 1st Interna-
tional Workshop on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. 33. J. Shu, B.P.A. Grandjean and S. Kaliaguine, Asymmetric Pd-Ag/SS catalytic membranes for CH4 steam reforming. Appl. Catal. A, 119 (1994) 305. 34. E. Kikuchi, Palladium/ceramic membranes for selective hydrogen permeation and their application to membrane reactor. Paper presented at the 1st International Workshop
on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. 35. H.M. Adris, C.J. Lim and J.R. Grace, The fluidized bed membrane reactor system: A pilot-scale experimental study. Paper presented at ISCRE 13, September 1994, Baltimore,
MD, USA. 36. Y. Ohta, M. Gondaira, K. Kobayashi, Y. Fujimoto and K. Kuroda, Study of the performance of a steam reformer of town gas equipped with Pd membranes. Paper presented at
the 1994 AIChE Annual Meeting, November 1994, San Francisco, CA, USA. 37. K. Jarosch, J.D. Hazlett and H.I. de Lasa, A novel catalytic reactor for steam reforming of CH4. Paper presented at the 1994 AIChE Annual Meeting, November 1994, San Francisco,
CA, USA. 38. D. Eng and M. Stoukides, Catalytic and electrocatalytic methane oxidation with solid oxide membranes. Catal. Rev. Sci. Eng., 33 (1991) 375. 39. T. Kameyama, M. Dokiya, M. Fujishige, H. Yokokawa and K. Fukuda, Possibility for effctive production of hydrogen from hydrogen sulfide by means of a porous vycor glass membrane. Ind. Eng. Chem. Fundam., 20 (1981) 97. 40. T. Kameyama, M. Dokiya, M. Fujishige, H. Yokokawa and K. Fukuda, Production of hydrogen from hydrogen sulfide by means of selective diffusion membrane. J. Hydrogen Energy, 8 (1983) 5. 41. Y.H. Ma, W.R. Moser, S. Pien and A.B. Shelekhin, Experimental study of H2S decomposition in membrane reactor. Paper presented at the 3rd International Conference on Inorganic
Membranes, July 10-14, 1994, Worcester, MA, USA. 42. J.J. Hammel, W.P. Marshall, W.J. Robertson and H.W. Barch, Porous-siliceous containing gas enriching hollow fibers and process of manufacture and use. Eur. Pat. Appl. 248,391 (1987). 43. M. Hassan, P.M. Thoen, J.D. Way and A.C. Dillon, Separation of gas mixtures using hollow film silica membranes. Paper presented at the 3rd International Conference on
Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA. 44. G.R. Gavalas, C.E. Megiris and S.W. Nam, Deposition of H2-permselective SiO2 Films. Chem. Eng. Sci., 44 (1989) 1829. 45. C.E. Megiris and J.H.E. Glezer, Preparation of silicon dioxide films by low-pressure chemical vapor deposition on dense and porous alumina substrates. Chem. Eng. Sci., 3925 (1992) 47.
564
11~ CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
46. T. Ioannides and G.R. Gavalas, Catalytic isobutane dehydrogenation in a dense silica membrane reactor. J. Membr. Sci., 77 (1993) 207. 47. G.R. Gavalas and S. Jiang, Use of carbon barriers in chemical vapor deposition of silica membranes. Paper presented at the 1994 AIChE Annual Meeting, November 1994, San Francisco, CA, USA. 48. M. Tsapatsis and G.R. Gavalas, Structure and aging characteristics of H2 permselective SiO2-Vycor membranes. ]. Membr. Sci., 87 (1994) 281. 49. T. Kameyama, K. Fukuda, M. Dokiya and Y. Kotera, Separation of hydrogen from hydrogen sulfide and hydrogen mixtures. Japan Patent 78,99,078 (1978). 50. M. Dokiya, J. Kameyama and K. Fukuda, Separation of hydrogen from hydrogen sulfide. Japan Patent 78,130,291 (1978). 51. A.D. Davidson and M. Salim, Initial studies of a novel ceramic membrane reactor. Br. Ceram. Proc., 119 (1988) 43. 52. N. Nourbakhsh, T.T. Tsotsis and I.A. Webster, Model planar alumina catalyst preparation and aging under hydrotreating conditions. Appl. Catal., 50 (1989) 65. 53. N. Nourbakhsh, Anodic Alumina Films: Preparation, Characterization and Investigation of Reaction and Transport Properties. Ph.D. Thesis, USC, May 1990. 54. A.F.M. Leenaars, K. Keizer and A.J. Burggraaf, The preparation and characterization of alumina membranes with ultrafine pores. J. Mater. Sci., 19 (1984) 1077. 55. R.J.R. Uhlhorn, M.H.B.J. Huis In't Veld, K. Keizer and A.J. Burggraaf, Synthesis of nonsupported and supported ~'-alumina membranes without defects. J. Mater. Sci., 27 (1992) 527. 56. A. Julbe, C. Guizard, A. Larbot, L. Cot and A.G. Fendler, The sol-gel approach to prepare candidate microporous inorganic membranes for membranes reactors. J. Membr. Sci., 77 (1993) 137. 57. J. Randon, A. Julbe, P. David, K. Jaafari and S. Elmaleh, Computer simulation of inorganic membrane morphology, 2. Effect of infiltration at the membrane support interface. ]. Colloid Interface Sci., 161 (1993) 384. 58. C.L. Lin, D.L. Flowers and P.K.T. Liu, Characterization of ceramic membranes. II. Modified commercial membranes with pore size under 40 ~. J. Membr. Sci., 92 (1994) 45. 59. J. Luyten, J. Cooymans, R. Persons, R. Leysen and J. Sleurs, A New Ceramic Support Material for Gas Separative Membranes. Paper presented at the 3rd International Conference on Inorganic Membranes, July, 1994, Worcester, MA, USA. 60. P.K.T. Liu, Private communication, 1995. 61. J. Deng and J. Wu, Formaldehyde production by catalytic dehydrogenation of methanol in inorganic membrane reactors. Appl. Catal. A, 109 (1994) 63. 62. M. Chai, M. Machida, K. Eguchi and H. Arai, Promotion of hydrogen permeation on metal dispersed alumina membranes and its application to a membrane reactor for methane steam reforming. Appl. Catal. A, 110 (1994) 239. 63. R.G. Minet, S.P. Vasileiadis and T.T. Tsotsis, Experimental studies of a ceramic membrane reactor for the steam/methane reaction at moderate temperatures 400-700~ in: G.A. Huff and D.A. Scarpiello (Eds.), Proceedings of the Symposium on Natural Gas Upgrading, 37 (1992) 245. 64. V.T. Zaspalis, W. van Praag, K. Keizer and J.R.H. Ross, Synthesis and characterization of primary alumina, titania and binary membranes. J. Mater. Sci., 27 (1992) 1023. 65. K.N.P. Kumar, K. Keizer and A.J. Burggraaf, Stabilization of the porous texture of nano structured titania by avoiding a phase transformation. J. Mater. Sci. Lett., 59 (1994) 13. 66. Y.S. Lin, C.H. Chang and R. Gopalan, Improvement of thermal stability of porous
11-- CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
565
nanostructured ceramic membranes. I\ &EC Res., 33 (1994) 860. 67. C.J. Brinker, R. Sehgal, N.K. Raman, S. Prakash, S.L. Hietala and S. Wallace, Sol-gel strategies for amorphous inorganic membranes exhibiting sieving characteristics. Paper
68. 69.
70.
71.
72. 73.
74. 75.
76.
77.
78.
79. 80. 81. 82.
presented at the 3rd International Congress on Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA. W.O. Haag and J.G. Tsikoyiannis, Membrane composed of pure molecular-sieve. U.S. Patent 5,019,263 (1991). W.J.W. Bakker, G. Zheng, F. Kapteijn, M. Makkee, J.A. Moulijn, E.R. Geuss and H. van Bekkum, in: M.P.C. Weijnen and A.H.H. Drinkenburg (Eds.), Single and Multicomponent Transport through Supported MFI Zeolite Membranes. Kluwer, Dordrecht, 1993, p. 425. F. Kapteijn, W.J.W. Bakker, G. Zheng, J.A. Moulijn and H. van Bekkum, Permeation and separation behavior of a silicalite MFI membrane. Paper presented at the 1st International Workshop on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. K. Kusakabe, A. Murata and S. Morooka,, Permeation of inorganic gases thoughts zeolite membranes at high temperatures. Paper presented at the 3rd International Conference on Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA. M.D. Jia, B. Chen, R.D. Noble and J.L. Falconer, Ceramic-zeolite composite for the separation of vapor/gas mixtures. J. Membr. Sci., 90 (1994) 1. S. Xiang and Y.H. Ma, Formation and characterization of zeolite membranes from sols. Paper presented at the 3rd International Conference on Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA. J. Wang, Y. Wang, S. Fan and X. Shi, Permeation and gas permeabilities of zeolite membranes. Microporous Mater., 4 (1995) 213. P. Meriaudeau, A. Thangaraj and C. Naccache, Preparation and characterization of silicalite molecular sieve membrane over supported porous sintered glass. Paper presented at the 1st International Workshop of Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. D. Uzio, J. Peureux, A.G. Fendler, J.-A. Dalmon and J. Ramsay, Formation and pore structure of zeolite membranes, in: J. Rouquerol, F. Rodriguez-Reinoso, K.S.W. Sing and K.K. Unger (Eds.), Characterization of Porous Solids III. Studies in Surface Science and Catalysis, Vol. 87. Elsevier, Amsterdam and New York, 1994, p. 411. Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Transport properties of alkanes through ceramic thin zeolite MFI membranes. J. Membr. Sci.,113 (1996) 293. D. Cazanave, A.G. Fendler, J. Sanchez, R. Loutaty and J.-A. Dalmon, Control of transport properties with a microporous membrane reactor to enhance yields in dehydrogenation reactions. Paper presented at the 1st International Workshop on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. J.F.Akyurtlu, A. Akyurtlu and C.E. Hamrin, A study on the performance of the catalytic porous-wall three-phase reactor. Chem. Eng. Comm., 66 (1988) 169. M.P. Harold, P. Cini and B. Patenaude, The catalytically impregnated tube: an alternative multiphase reactor. AIChE Symp. Ser., 268 (1989) 26. P. Cini and M.P. Harold, Experimental study of the tubular multiphase catalyst. AIChE l., 37, 7 (1991) 997-1008. M. Torres, J. Sanchez, J.-A. Dalmon, B. Bernauer and J. Lieto, Modeling and simulation of a three-phase membrane reactor for nitrobenzene hydrogenation. Ind. Eng. Chem. Res., 33, 10 (1994) 2421.
566
11 ~ CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
83. A. Pantazidis, J. A. Dalmon and Mirodatos, Oxidative dehydrogenation of propane in catalytic membrane reactor. Paper presented at the 1st International Workshop on Catalytic
Membranes, September 1994, Lyon-Villeurbanne, France. 84. M.P. Harold, V.T. Zaspalis, K. Keizer and A.J. BurggraaL Intermediate product yield enhancement with a catalytic inorganic membrane, I. Analytical model for the case of isothermal and differential operation. Chem. Eng. Sci., 48 (1993) 2705. 85. D. Lafarga, J. Santamaria and M. Menendez, Methane oxidative coupling using porous ceramic membrane reactors. I. Reactor development. Chem. Eng. Sci., 49 (1994) 2005. 86. J. Coronas, M. Menendez and J. Santamaria, Methane oxidative coupling using porous ceramic membrane reactors. II. Reaction studies. Chem. Eng. Sci., 12 (1994) 2015. 87. Y.D. Chen and R.T. Yang, Preparation of carbon molecular sieve membrane and diffusion of binary mixtures in the membrane. Ind. Eng. Chem. Res., 33 (1994) 646. 88. M.B. Rao and S. Sircar, Nanoporous carbon membranes for separation of gas mixtures by selective surface flow. J. Membr. Sci., 85 (1993) 253. 89. J.E. Koresh and A. Softer, Study of molecular sieve carbons. Part I. Pore structure, gradual pore opening and mechanism of molecular sieving. J. Chem. Soc. Faraday Trans. /, 76 (1980) 2457. 90. J.E. Koresh and A. Softer, Molecular sieve carbon permselective membrane. Sep. Sci. Technol., 18 (1983) 723. 91. J.E. Koresh and A. Softer, Mechanism of permeation through molecular sieve carbon. ]. Chem. Soc. Faraday Trans. I, 82 (1986) 2057. 92. J.E. Koresh and A. Softer, The carbon molecular-sieve membranes, general properties and the permeability of CH4/H2 mixture. Sep. Sci. Technol., 22 (1987) 972. 93. V.M. Linkov, R.D. Sanderson, E.P. Jacobs, A.L. Lapidus and A.J. Krylova, Hollow fibre carbon membranes in Fischer Tropsch synthesis. Paper presented at the 1st International
Workshop on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. 94. T.H. Etsell and S.N. Flengas, The Electrical properties of solid oxide electrolytes. Chem. Rev., 70 (1970) 339. 95. A. Clearfield, Role of ion exchange in solid state chemistry. Chem. Rev., 88 (1988) 125. 96. P.J. Gellings, H.J.A. Koopmans and A.J. Burgraaf, Electrocatalytic phenomena in gas phase reactions in solid electrolyte electrochemical cells. Appl. Catal., 39 (1988) 1. 97. B.C.H. Steele, Oxygen ion conductors and their technological applications. Mater. Sci. Eng., B13 (1992) 79. 98. D. Eng and M. Stoukides, Catalytic and electrocatalytic methane oxidation with solid oxide membranes. Catal. Rev. Sci. Eng., 33 (1991) 375. 99. M. Stoukides, Application of solid electrolytes in heterogeneous catalysis. Ind. Eng. Chem. Res., 27 (1988) 1745. 100. S.H. Langer, Electrogenerative systems. Potential uses include clean-up of flue gases from coal fired stationary power plants. Plat. Met. Rev., 36 (1992) 202. 101. Y. Teraoka, H. M. Zhang, S. Furukawa and N. Yamazoe, Oxygen permeation through perovskite type oxides. Chem. Lett. (1985) 1743. 102. Y. Teraoka, H. M. Zhang, K. Okamffoto and N. Yamazoe, Mixed ionic electronic conductivity of Lal-xSrxCOl-yFeyO3-~perovskite-type oxides. Mater. Res. Bull., 33 (1988) 51. 103. Y. Teraoka, T. Nobunaga and N. Yamazoe, Effect of cation substitution on the oxygen semipermeability of perovskite type oxides. Chem. Left. (1988) 503. 104. Y. Teraoka, N. Nobunaga, K. Okamoto, N. Miura and N. Yamazoe, Influence of constituent metal cations in substituted LaCoO3 on mixed conductivity and oxygen permeability. Solid State Ion., 48 (1991) 207.
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC M E M B R A N E REACTORS
567
105. S. Pancharatnam, R.A. Huggins and D.M. Mason, Catalytic decomposition of nitric oxide on zirconia by electrolytic removal of oxygen. J. Electrochem. Soc., 122 (1975) 869. 106. D.M. Mason, Method and apparatus for catalytic dissociation of NO. US Patent 4,253,925, March 3, 1975. 107. T.M. Gur and R.A. Huggins, Decomposition of nitric oxide on zirconia in a solid state electrochemical cell. J. Electrochem. Soc., 126 (1979) 1067. 108. A.V. Joshi, An alternative method for the removal of oxides of nitrogen and sulfur from combustion processes. Final Report, US DOE contract DE-AC22-85PC81003, 1986. 109. J.W. Cook, L.P. Cornell, M. Keyrani, M. Neyman and D. Helfritch, Simultaneous particulate NOx, SOx removal from flue-gas by all solid-state eletrochemical technology. Final Report, US DOE Contract, DE-AC22-87PC79856. 110. D.J. Clark, R.W. Losey and J.W. Suitor, Separation of oxygen by using zirconia solid electrolyte membranes. Gas Sep. Purif., 6 (1992) 201. 111. C.G. Vayenas, S. Bebelis, I. V. Yentekakis, P. Tsiakaras and H. Karasali, Non-faradaic electrochemical modification of catalytic activity. Platinum Metals Rev., 34 (1990) 122. 112. U. Balachandran, J.T. Dusek, R.L. Mieville, R.B. Poeppel, M.S. Kleefisch, S. Pei, T.P. Kobylinski, C.A. Udovich and A.C. Bose, Dense ceramic membranes for partial oxygenation of methane. Paper presented at the 3rd International Congress on Inorganic Mem-
branes, July 10-14, 1994, Worcester, MA, USA. 113. U. Balachandran, J.T. Dusek, S.M. Sweeney, R.B. Poeffel, R.L. Mievile, M.S. Kleefisch, S. Pei, T.P. Kobylinski, C.A. Udovich and A.C. Bose, Methane to syngas via ceramic membranes. Am. Ceramic. Soc. Bull., 74 (1995) 71. 114. A. Chanaud, A. Julbe, C. Larbot, L. Guizard. H. Cot, A. Borges, A.G. Fendler and C. Mirodatos, Catalytic membrane reactor for the oxidative coupling of methane, Part 1: Preparation and characterization of LaOC1 membranes. Paper presented at the 1st Inter-
national Workshop on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. 115. H. Borges, A.G. Fendler, C. Mirodatos, P. Chanaud and A. Julbe, Catalytic membrane reactor for the oxidative coupling of methane, Part 2: Catalytic properties of LaOC1 membranes. Paper presented at the the 1st International Workshop on Catalytic Membranes,
September 1994, Lyon-Villeurbanne, France. 116. Y.L. Becker, A.G. Dixon, W.R. Moser and Y.H. Ma, Modelling of ethylbenzene dehydrogenation in a catalytic membrane reactor. J. Membr. Sci., 77 (1993) 197. 117. M. Tayakout, B. Bernauer, Y. Toure and J. Sanchez, Modeling and simulation of a catalytic membrane reactor. J. Simul. Pract. Theory, 2 (1995) 205. 118. M. Tayakout, J. Sanchez, D. Uzio and J.A. Dalmon, Catalytic membranes for cyclohexane dehydrogenation reaction. Paper presented at the 1st International Workshop on Cata-
lytic Membranes, September 1994, Lyon-Villeurbanne, France. 119. C.Y. Tsai, Y.H. Ma, W.R. Moser and A.G. Dixon, Simulation of nonisothermal catalytic membrane reactor of CH4 partial oxidation to syngas. Paper presented at the 3rd Interna-
tional Congress on Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA. 120. A.G. Dixon, W.R. Moser and Y.H. Ma, Waste reduction and recovery using 02 permeable membrane reactors. Ind. Eng. Chem. Res., 33 (1994) 3015. 121. W. Wang and Y. S. Lin, A theoretical analysis of oxidative coupling of CH4 in a tubular dense membrane reactor. Paper presented at the 3rd International Congress on Inorganic
Membranes, July 10-14, 1994, Worcester, MA, USA. 122. Y. Lu, A. Ramachandra, Y.H. Ma, W.R. Moser and A.G. Dixon, Reactor modeling of the oxidative coupling of methane in membrane reactors. Paper presented at the 3rd Interna-
tional Congress on Inorganic Membranes, July 10-14, 1994, Worcester, MA, USA.
568
11m CURRENTDEVELOPMENTSAND FUTURERESEARCHIN CATALYTICMEMBRANEREACTORS
123. T. Nozaki and K. Fujimoto, Oxide ion transport for selective oxidative coupling of methane with new membrane reactor. AIChE J., 40 (1994) 870. 124. N. Itoh, W. C. Xu and K. Haraya, Radial mixing diffusion of hydrogen in a packed-bed type of Pd membrane reactor. Ind. Eng.Chem. Res., 33 (1994) 197. 125. V.A. Papavassiliou, J.A. McHenry, E.W. Corcoran, H.W. Deckman and J.H. Meldon, High flux asymmetric catalytic membrane reactors. Optimization of operating conditions. Paper presented at the 1st International Workshop on Catalytic Membranes, September
1994, Lyon- Villeurbanne, France. 126. A.B. Bindjouli, Z. Dehouche, B. Bernauer and J. Lieto, Numerical simulation of catalytic inert membrane reactor. Computers Chem. Eng., 18 (Suppl.) (1994) 5337-5341. 127. A. Varma, K.L. Yeung, J. Szegner and G. Cao, Nonuniform catalyst distribution for catalytic membrane reactors. Paper presented at the 3rd International Congress on Inorganic
Membranes, July 10-14, 1994, Worcester, MA, USA. 128. E. Gobina and R. Hughes, Equilibrium-shift in alkane dehydrogenation using a high temperature catalytic membrane reactor. Paper presented at the 1st International Workshop
on Catalytic Membranes, September 1994, Lyon-Villeurbanne, France. 129. Y.V. Gokhale, R.D. Noble and J.L. Falconer, Analysis of a membrane-enclosed catalytic reactor for butane dehydrogenation. J. Membr. Sci., 77 (1993) 197. 130. M.P. Harold, C. Lee, A.J. Burggraaf, K. Keizer, V.T. Zaspalis, and R.S.A. de Lange, Catalysis with inorganic membranes. MRS Bull., XIX (4) (1994) 34. 131. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterization of supported microporous ceramic membranes prepared by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57. 132. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Permeation and separation studies sol-gel modified ceramic membranes. Microporous Mat., 4 (1995) 169. 133. G. Cacciola, Y.I. Aristov, G. Restuccia and V.N. Parmon, Influence of hydrogen-permeable membranes upon the efficiency of the high temperature chemical heat pumps based on cyclohexane dehydrogenation, benzene hydrogenation reactions. Int. J. Hydrogen Energy, 18 (1993) 673. 134. R.G. Minet and T.T. Tsotsis, Catalytic membrane steam/hydrocarbon reformer. U.S. Patent 4,981,676, January 1,1991. 135. D. Jansen, A.B.J. Oudhuis and H.M. van Veen, CO2 reduction potential of future coal gasification based power generation technologies. Energy Conserv. Manage., 33 (1992) 365. 136. A. Basile, V. Violante and E. Drioli, The membrane integrated system in the fusion reactor fuel cycle. Paper presented at the 1st International Workshop on Catalytic Mem-
branes, September 1994, Lyon-Villeurbanne, France. 137. J. Feldman and M. Orchin, Membrane-supported rhodium hydroformylation catalysts. J. Mol. Catal., 63 (1990) 213. 138. J. Kim and R. Datta, Supported liquid-phase catalytic membrane reactor-separator for homogeneous catalysis. AIChE J., 37 11 (1991) 165. 139. S. Chen, H. Fan and Y.-K. Kao, A membrane reactor with two dispersion-free interfaces for homogeneous catalytic reactions. Chem. Eng. J., 49 (1992) 35. 140. R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas transport and separation with ceramic membranes, Part II. Synthesis and properties of microporous membranes. J. Membr. Sci., 66 (1992) 271.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.I. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 12
Transport and fouling phenomena in liquid phase
separation with inorganic an d hybrid membranes
Christian Guizard and Gilbert Rios Laboratoire des Mat4riaux et Proc4d4s Membranaires (UMR 5635 CNRS-ENSCM-UMII), Ecole Nationale Sup6rieure de Chimie, 8, rue de I'Ecole Normale, 34053 Montpellier, France
12.1 INTRODUCTION Regardless of the type of the membrane used for liquid phase separation, three main classes of phenomena must be distinguished concerning solvent and solute transport. They develop: near the front face of the filtering element (polarization); inside the membrane structure (internal mass transport); at interfaces (fouling). If hydrodynamics and related mass transfer kinetics mainly control the first two, thermodynamic equilibrium also plays a main part in the third. The occurrence of secondary phenomena, such as solute modifications due to surrounding effects, must also be mentioned due to possible induced interferences. In this chapter, pressure driven processes involving porous inorganic or hybrid membranes are particularly examined. As recently shown by Bhave [1], differences with traditional organic elements mainly result from the structure and intrinsic properties of materials, either regarding flow (with ceramic membranes, the transport occurs through the intergranular spaces within the top layer, porous sublayers and support, while across polymeric barriers it develops -
-
-
570
12 m TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
through the continuous network of openings from one face to the other) or interactions (unlike polymers, metallic oxides present electric charges and thus surface phenomena that strongly depend upon pH and ionic strength of solutions). Basic phenomena are firstly recalled in a short presentation which is not intended to give a thorough description as this has already been done elsewhere [2], but which aims to provide a comprehensive study of classical routes for modelling. An unified approach of fouling during the separation process itself or even during membrane synthesis is also proposed, which gives a new opening on formed-in-place membranes. Solvent and solute interactions with membrane material during microfiltration and ultrafiltration are then described and related to process performance, i.e. permeability and rejection. It is worth recalling that molecules in the colloidal range, roughly between one and one hundred nanometers, are separated by ultrafiltration, while microfiltration retains larger particles. As recently shown, small molecules with sizes lower than about one nanometer and ions can be rejected by nanofiltration membranes. In the following, size exclusion, Donnan effect or combined mechanisms operating in such situations are examined. Finally some more prospective aspects are hinted at, either concerning coupled processes with at least one membrane separation step, or new membranes for facilitated transport or selective separation. Specific advantages of inorganic or hybrid material will be underlined.
12.2 BASIC P H E N O M E N A IN PRESSURE D R I V E N PROCESSES
12.2.1 Modelling of Hydrodynamics and Mass Transport Following the classical scheme, free transport of solutes and solvent in the boundary layer at the liquid-membrane interface and hindered transport of substances in the porous structure of the membrane material are described successively. Concentration polarization: Convective transport and retention of solutes by the membrane results in an accumulation of species at wall. Local concentrations, Cw, are higher than in the bulk, CB, and a back-diffusion from near the wall into the bulk liquid phase takes place. This is the so-called "concentration polarization" phenomenon (Fig. 12.1). A simple mass balance leads to the classical equation: (C w - Cp)/(C
b
-
C p ) --- exp
(Jv/k)
(12.1)
with k = D/~), the mass transfer coefficient defined as the ratio of solute diffusivity to boundary layer thickness.
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
571
membrane ~_~ --~ gel layer Cw
= Cg
t t Tt
permeate
CB Cp
boundary layer Ap
bulk feed
8
x
Fig. 12.1. Concentration polarization in pressure driven m e m b r a n e processes. Concentration profile and gel formation at the steady state.
From this relation, it can be easily inferred that the phenomenon will be strongly dependent on: membrane morphology and interactions through permeate flux, Jv; fluid properties (viscosity, diffusivity) and flow conditions through k. Regarding the definition of the observed rejection TRobsbetween feed and permeate and the real rejection TR defined from concentration at walls, the following equations are used: Cp Cb Cp Cw
TRobs = 1 = -
(12.2)
TR = 1
(12.3)
The real rejection can be calculated from the observed rejection using Eqs. (12.1) to (12.3). rr = 1 +
1 - Trobs
exp
8
(12.4)
Trobs
As a general rule, concentration polarization will be all the more important as membrane permeability and liquid viscosity will be higher and solute diffusivity will be lower. This is the reason why its effects, ordinarily negligible for nanofiltration, appear to be of major importance during ultrafiltration. Under gel polarization conditions, i.e. in those cases where Cw reaches a maximum gel value Cg (as an example with ultrafiltration of concentrated protein solutions),
572
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
the above analysis even indicates that the driven pressure Ap can no longer influence the permeate flux Jv: operating performance is essentially controlled by mass transfer parameters. On the assumption that permeate flow does not strongly modify hydrodynamics at walls, k is ordinarily estimated using classical dimensionless correlation for non-porous walls, such as the L6v6que equation for laminar flow or the Deissler equation for turbulent flow. More recently, in order to account for specific disturbances that take place at pore entrance when high permeate fluxes prevail, new specific equations have been proposed [3]. Membrane transport: There are two fundamentally different ways of modelling permeable solid materials. In the first the membrane is regarded as a heterogeneous structure through the pores of which solvent and solutes flow in, while in the second the membrane is considered as a homogeneous phase forming a molecular mixture with other species. Heterogeneous models are the natural choice for aggregated inorganic porous materials, while homogeneous models are most usefully considered with dense top polymeric layers of organic-inorganic hybrid membranes. Starting from the broad point of view of Maxwell and Stefan [4], it may be assumed that in a steady flow the thermodynamic forces acting on solute or solvent are counterbalanced by the frictional forces due to other species. This leads to the following equation
(d ln a i / d z )
+ (1/p~.dp/dz) + 1/@~.d@/dz = ~..~Xj[(Vj -- Vi)/Dij ] - ( v i / D i m )
(12.5)
with ai the activity of i and Xj the mole fraction of j at z, p and cI), the local pressure and electric potential, vj and vi the species velocities, Dij and Dim the Maxwell-Stefan diffusivities representative of the friction between species i and j or i and m the membrane, p~ and ~ are two constants in pressure and electric driving forces given by p~ - R T / V i
(12.6)
~ - RT/F zi
(12.7)
with R the gas constant, T the absolute temperature, F the Faraday constant, zi and Vi the charge number and partial molar volume of i. Fluxes with respect to the membrane are equal to li = Ci" (vi + Jv)
(12.8)
with Ci the local concentration, and Jv the viscous flow velocity in heterogeneous media, a linear function of the driven pressure, Ap, according to Darcy's law.
12-- TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
573
Various treatments of Eqs. (12.5) to (12.8) are proposed in the literature for those cases where charge effects are negligible (microfiltration, ultrafiltration or nanofiltration of neutral solutes), and low concentrated solutions are considered, i.e. with negligible non-idealities and activities assimilated to concentrations. Most of the time, membrane morphology is just considered through simple parameters such as effective pore size, tortuosity accounting for the effect of fouling on measurable transport properties. A well-known model thus obtained is due to Kedem and Katchalsky [5]: J v - ap. ( A F - (~ AI-[) Is = co" ( C w - C p ) +
(12.9)
(1 - o ) .
Jv" [ ( C w - C p ) / (ln C w / Cp)]
(12.10)
When the rejection, TR, is high, wall and permeate concentrations are better linked through
ZR - (Cw- Cp) / Cw= (~-(1 -F) / (1 - G F)
(12.11)
F = exp [- (1 - o ) . Iv / co]
(12.12)
These equations relate the fluxes for the solvent, Jv, and for the considered solute, Js, to three parameters: Lp the membrane permeability to the solvent, co the membrane permeability to the solute and c~ the reflection coefficient that measures the selectivity of the filtering element. This dimensionless number varies from 0 (pure convective transport, roughly the situation in microfiltration) to I (pure diffusive transport, preferably encountered in reverse osmosis). On the assumption that solute molecules are rigid and spherical, transport parameters may be linked to the membrane morphology, i.e. the mean pore radius or the ratio of the open surface area to the pore length, through hindrance and friction factors respectively accounting for the pore entrance and wall effects [6]. The one-dimensional flux equations (12.9) and (12.10) implicitly assume that local components of flux in the plane of the membrane can be ignored relative to fluxes in the z direction. If we consider z* as a dimensionless axis perpendicular to the faces of the membrane whose values range from zero to unity, a local form of Eqs. (12.9) and (12.10) can be rewritten for solute i: Jv =
-
Kvl~ th (dp/dz*
-
r~RTdCi/dz*)
Ji = (1 - (~)CiJv- (P/h)dCi/dz*
(12.13) (12.14)
Alternative definitions for the permeability coefficients have been utilized in Eqs. (12.13) and (12.14) for the purpose of explicitly representing the dependence on membrane thickness h and viscosity ~t. They are related to their counterparts in Eqs. (12.9) and (12.10): ap = K v/~l,h
(12.15)
574
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
(12.16)
m = P/h
Solution of Eq. (12.14) yields a concentration profile for the solute w i t h i n the membrane: 1 - exp [-Pe(1 - z*)]
C i ( z * ) -- C i ( 1 ) Ci(0 ) -
Ci(1)
(12.17)
1 - exp [-Pe]
in w h i c h Pe represents the Peclet number: Pe = (1 - c ~ ) J v h / P
(12.18)
The m a g n i t u d e of the Peclet n u m b e r indicates the importance of the convective relative to the diffuse process for solute transport. The solute concentration profiles for representative values of Pe are illustrated in Fig. 12.2 according to B u n g a y [7]. W h e n diffusion is d o m i n a n t (Pe ---0) the concentration varies nearly linearly in z*. For large absolute values of the Peclet n u m b e r , diffusion is significant only in a thin zone adjacent to the low pressure face of the m e m b r a n e in w h i c h the concentration profile is very steep. For micro- and ultrafiltration m e m b r a n e s , the solute concentration varies little from the value at high pressure face. For nanofiltration the Peclet n u m b e r can v a r y considerably d e p e n d ing on m e m b r a n e characteristic: almost dense or p o r o u s m e m b r a n e s .
+10
0.8 +3
C~(z*) - C~(])
0.6
+1 0
Ci(O) - C i ( | )
0.4
-3
--1
0.2
0
0
0.2
0.4
0.6
0.8
1.0
Z9
Fig. 12.2. Steady-state solute concentration profiles hi simultmleous diffusion and convection across a membrmle of uniform properties. Numbers adjacent to profiles hldicate values of the Peclet number whose sign depends upon the direction of the volumetric flux relative to the external solution concentration difference [7].
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
575
If electric forces influence solute retention, as an example with nanofiltration of electrolyte solutions on membranes presenting a high charge density, references in literature suggest the use of a peculiar form of Eqs. (12.5) to (12.8), the so-called Nernst-Planck equation: ]i = -Pi(dCi/dz + Ci I r
" d ~ / d z ) + J v ci. (1 -(~i)
(12.19)
in which Pi and Ji are the permeability and flux for ion i. Using appropriate boot-strap relations (electroneutrality of solutions or zero electric current conditions) and boundary conditions (at the surface between the membrane and. external solution the assumption is made of Donnan equilibrium) integration can be carried out across the membrane. But ordinarily it is not easy to get the effective charge density and related electrostatic potential of the material [8]. If for very dilute solutions of neutral rigid spherical species, the hydrodynamic theory for the hindered transport previously developed may be considered as at a mature stage, in sharp contrast is the situation for charged a n d / o r flexible solutes in more concentrated solutions where solute-pore and solutesolute interactions remain largely unevaluated. Other important gaps in the present knowledge concerns the prediction of solute adsorption in the pores and its effect on measurable transport properties [9]. 12.2.2 Fouling
There are two different levels where fouling phenomena and related effects may interfere with performance of composite inorganic or hybrid membranes. The first and the more classically reported in literature is the one of the separation process itself, which through various interactions between solution and material (adsorption, surface deposits, pore plugging) generally leads to reduced fluxes and increased retentions. The second, much more less described by authors but of the same nature and with analogous effects, concerns membrane preparation, and the possible interactions between deposited layers. Theses two aspects are linked up with the so-called formed-in-place membranes, obtained by deposition of species onto a ceramic support through cross-flow filtration. In what follows, they will be described in a unified approach. Membrane synthesis
The formation of ceramic membranes for microfiltration, ultrafiltration or nanofiltration by association of various granular layers is now a common procedure [10]. Each layer is characterized by its thickness, h, its porosity, r and its mean pore diameter, dp. These parameters are controlled by the particle size, d, and the synthesis method. Each layer induces a resistance which may be predicted through the classical Carman-Kozeny model:
576
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION I
Fig. 12.3. Simulation of multilayer deposits with particle diameter ratios of 2.5 (I, non-interpenetrated layers) and 4 (II, interpenetrated layers) [11].
R = Ap/(Jv" ~t) = 180 h(1 - E)2/(E 3. d 2) = (80 h / E ) . dp 2
(12.20)
It has been shown recently [11] that the experimental hydraulic resistance of such a composite structure can be much larger than the theoretical resistance obtained by simply summing the resistances of the different layers. As an interpretation, the existence of highly resistant transition boundary layers due to infiltration between adjacent media of drastically different particle sizes has been suggested (Fig. 12.3). In order to check this theory, infiltrated and noninfiltrated TiO2 membranes deposited on o~-alumina support have been prepared. Scanning electron micrograph of their interfaces is shown in Fig. 12.4. From them, complementary resistances have been measured. TABLE 12.1 Experimental and predicted hydraulic resistances (m -1 unit) for a m e m b r a n e with 0.2 ~tm m e a n pore diameter deposited on a s u p p o r t with 0.8 ~tm pore diameter [11] Membrane
Model
Support
3.6 x 10 l~
3.6 x 10 l~
Membrane
7.0 x 10 l~
7.2 x 10 l~
Noninfiltrated
Infiltrated
Noninfiltrated
Infiltrated
Filtration element
12 x 1010
17 x 1010
12.6 • 1010
15 x 1010
Infiltration zone
1.4 x 10 l~
6.4 x 10 l~
1.8 x 10 l~
4.2 x 10 l~
12 n TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
577
Fig. 12.4. Cross-section microscopy images showing interpenetrated (a) and non-interpenetrated (b) ceramic porous layers.
A computer model has been also developed to predict the porosity at the layer interface, as a function of the relative sizes of granules. From Table 12.1, it follows that measured and calculated resistances are comparable. This clearly proves the effect of infiltration on performance and shows that to increase the permeability of filtration elements new routes for synthesis limiting the extent of infiltrated zones are needed.
Membrane separation When a solution is filtered through a m e m b r a n e which can partially or completely retain one or more of the solutes present, the observed flux strongly decreases with time at first, and then under steady state conditions reaches a level much less than the flux of pure solvent. In addition, the retention of solutes is generally increased. This phenomenon corresponds to on-line membrane fouling. A clear distinction must be m a d e between m e m b r a n e fouling and concentration polarization. As already explained, concentration polarization is the develo p m e n t of a concentration gradient of the retained components near the m e m brane. It is a function of the h y d r o d y n a m i c conditions and is independent of the
578
12 m T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
physical properties of the membrane. The membrane pore size and porosity are not directly affected by concentration polarization. Fouling on the other hand is the deposition of material on the membrane, leading to a change in the membrane behaviour. Fouling is the coupling of deposited material to the membrane through the intermediate step of concentration polarization. From this point of view, fouling includes gel formation. Fouling must also be distinguished from membrane compaction, which is the compression of the membrane structure under the transmembrane pressure, causing a decrease in membrane permeability. If with purely inorganic membrane compaction may be neglected, it is not necessarily the same with organo-inorganic or hybrid membrane under high enough Ap as for nanofiltration. With hybrid filtering elements, swelling of polymeric layers may also induce flux decrease that must not be confused with fouling. Generally on-line fouling of membranes is extremely complicated and in several aspects is not fully understood. Authors used to consider three successive stages in flux decline. As an example Aimar et al. [12] for the UF of cheese whey with Carbosep M4 membranes (Fig. 12.5) mention a very short initial step (less than 1 mn) corresponding to reversible concentration polarization build up, and then two distinct features: a sharp decrease during the first hour, followed by a slow decrease over several hours. From a comparison of models, they suggest that the sharp decrease could be due to either protein adsorption or particle deposition, and that the longer term decline is related to further
Flux
Stage 1 - Flux loss due to concentration p o l a r i s a t i o n
Stage 2 - Flux loss due to protein deposition
Stage 3 - Flux loss d u e to p a r t i c l e d e p o s i t i o n c c o n s o l i d a t i o n of the f o u l i n g m a r e r i a l
Time Fig. 12.5.Variousstagesof flux decline durh~gprotehl filtration with an ultrafiltration membrane (Carbosep M4) [12].
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
579
convective deposition of species or fouling layer consolidation. Considering solute retention, a continuous increase may be observed as flux decreases. It is likely that both surface and internal fouling, to a greater or lesser extent, occur simultaneously, the predominant mechanism being a function of the experimental conditions, as well as of membrane and feed material properties [13]. In several recent works, the basic mechanisms of protein deposition on microfiltration c~-alumina membranes were thoroughly investigated. Vetier et al. [14] studied the processing of milk using a 0.2 gm Membralox membrane. Starting from scanning electron microscopy and physicochemical analysis,. authors show that at first there is an important surface adsorption phenomenon in the form of a thin film of casein and salts on which other micelles are then deposited. These may be connected by phosphocalcic bonds which form a porous layer that is largely responsible for the retention of serum proteins. It was observed that there was almost no penetration of casein micelles in the top layer, and that the cross-flow velocity also helped to reduce the thickness of the fouling layer. In another study carried out with gelatin solutions and several membranes with pore diameters between 0.2 and 2.0 gm, Freund and Rios [15] gave a further insight into the influence of pore diameter, fluid velocity, concentration and driven pressure as regards the extent and localization (surface deposition or pore plugging) of fouling. A clear distinction is made between irreversible fouling phenomena that are taken into account through a complementary term Rc increasing the hydraulic membrane resistance, and reversible phenomena probably due to highly concentrated layers of mobile macromolecules in the vicinity of interfaces that are associated to an abatement pressure term A~, as indicated in the following equation: Jv = (Ap- AFI)/~t. (Rm + Rc)
(12.21)
Rm is the hydraulic resistance of the clean membrane that is deduced from water
flux, while An and Rc are estimated through water and solution fluxes of the prefouled element. Another work conducted by Matsumoto et al. [16] also compared the resistances due to internal and surface fouling with ovalbumin. From all these data, it may be concluded that: with low velocities (laminar), and with 0.2 gm or to a lesser extent 0.8 gm membrane, surface layer fouling dominates; - with larger pore sizes or higher velocities, pore plugging prevails. As a whole, the main extent of fouling when protein are treated with alumina membranes is underlined, and it is proved that adsorption turns the initial microfiltration process into a true formed-in-place or dynamic ultrafilter, capable of effectively retaining species with a size ten or one hundred times lower than initial pore size. Because all inorganic materials (zirconia, titania, glass) do not present the same charge a n d / o r hydrophobic/hydrophilic characteristics, interactions -
580
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
may strongly vary. Thus surface modification can constitute a route of the highest interest for enhancing or preventing solute interactions with membranes. Another main consequence of previous findings is that, with an asymmetric element composed from different material, performance will not be only dependent on solute-membrane, but also on solute-support interactions. It is also worth keeping in mind that fouling mechanisms are influenced by modifications in solute configuration due to charge or high concentration effects (as an example, the aggregation of proteins in the concentrated polarization layer which then in turn will induce specific interactions with the membrane material) as well as by specific solute-solute interactions. Recently, Grund et al. [17] suggested that with bovine serum albumin in the presence of a sufficient amount of fatty acid protein-protein interactions may have the greatest effect on the permeability of the protein deposits and hence on the permeate fluxes, while at a sufficient reduced level of fatty acid these interactions appear to play a far lesser role than do protein-membrane interactions. Similar effects may be observed with pH, ionic strength, temperature as shown elsewhere [13]. From a practical view point, this highlights the main role of pretreatment and choice of working conditions as regards cross-flow filtration performance. Most of the time, fouling can be removed by cleaning solutions or other appropriate means, and membrane performance restored. But in same cases also, foulant behaves into the membrane in such a way that the filtering layer cannot be completely regenerated: this type of fouling, that has received much less attention, determines the lifetime of the membrane.
Formed-in-place membranes It has been previously shown that dynamically formed or formed-in-place ultrafiltration membranes naturally set down when biological molecules such as proteins tangentially flow along a ceramic microfiltration membrane. This new membranes directly result from infiltration/deposition of macromolecules, i.e. fouling mechanisms, that develop on the original filtering element. In a similar way, when organic or inorganic polyelectrolytes are added to pressurized salt solutions held in contact with inorganic porous supports (ceramic, stainless steel), salt-filtering layers are formed that present reverse osmosis and/or ion exchange properties with moderate salt rejection as compared to polymeric cellulose acetate but high flux often an order of magnitude higher [1]. Such membranes possess the desirable properties for applications requiring higher temperatures. Studies concluded that hydrous zirconium oxide/polyacrylic acid dual-layers membranes had the best performance characteristics. Recently, Negrel et al. [18] proposed a new route to get nanofiltration performance starting from a formed-in-place membrane of gelatin deposited on a
581
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION 100 -
PEG 6OOO
v
Q
PEG 2000
lrll
PEG 6o0
tO O ...=,,
rr
2
5
10
15
20
40
Time( min )
Fig. 12.6. PEG rejection versus time for a formed in place membrane made of tanned gelatin [18].
0.2 ~tm alumina support from SCT-US Filter, and then applying various chemical (tanning) a n d / o r physical (thermal) treatments to the macromolecular deposit. Separation performance were characterized using PEG (polyethylene glycol) as tracer molecules (Fig. 12.6). With the new membrane, high fluxes were obtained probably due to the fact that protein deposit are thinner than top layers of commercial nanofilters. It is worth noting that the idea, if not the means, of this new generation of organic-inorganic composite membrane is not very far from the one proposed by Guizard et al. [19] of impregnating organic polymers such as polyphosphazenes or heteropolysiloxanes in mesoporous inorganic supports of SiO2, A120 3, ZrO 2 o r T i O 2 oxides. This also bears witness of the interest of a global view of fouling, integrating at the same time synthesis and on-line separation problems.
12.2.3 Specific Aspects Attached to Ceramic Membranes Major developments of membrane processes using ceramic membranes have been aimed at microfiltration or ultrafiltration applications. Up to now the most important applications for these membranes are found in aqueous media for the separation of particles, bacteria, colloids, macromolecules. Recently, ceramic nanofilters based on sol-gel derived microporous materials have been described [20]. They extend separation capability of ceramic membranes to ions and organics.
582
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Behind the general parameters (viscosity, transmembrane pressure, temperature, flow velocity) which can influence cross-flow filtration with ceramic membranes two aspects must be considered to be more specific of this sort of membrane. One is related to the geometry (tubular multichannel or honeycomb) found for the major part of commercially available membranes, the other is the amphoteric behaviour of metal oxides used in the preparation of these ceramic membranes.
Membrane geometry The tubular or monolith geometry of ceramic membranes proceeds from inherent constraints in ceramic material processing. Due to the mechanical properties of sintered ceramic materials which can be s u m m e d up in a high compressive resistance and a brittle character, cylindrical shapes offer the best compromise between pressure resistance and cross-flow filtration adaptability. Typical geometries for ceramic membranes are shown in Fig. 12.7. While microfiltration is operated at low transmembrane pressure (-~ I bar), higher pressures are required for ultrafiltration (up to 10 bar) and nanofiltration (10-20 bar). Here also pressure vessels with pressure resistant sealing are easier to design for ceramic tubes than for ceramic plates. In micro- and ultrafiltration, high cross-flow velocity (2-8 m / s ) is generally recommended to minimize concentration polarization and induced fouling. Thanks to their abrasive resistance, tubular or multichannel ceramic membranes allow to attain a high wall shear rate with particles charged liquids without prefiltration. Figure 12.8 shows the details of a membrane module commercialized by Tech-Sep (Groupe Rhone-Poulenc) in which a number of tubular or multichannel ceramic membranes are arranged in a parallel way. In this case tubular membranes are 1.2 m long, 10 m m in diameter and up to 252 tubes can be inserted in one module. The problem encottntered with the first generation of tubular ceramic membranes was that they required a high pumping energy and a high volume/surface
honeycomb
multichannel
Fig. 12.7. Different types of geometries found in commercialceramicmembranes.
583
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE S E P A R A T I O N
Fig. 12.8. Example of a module obtained by assembling a number of multichannel membranes in a stainless steel pressure vessel (from Tech-Sep, Groupe Rhone Poulenc).
ratio. This p r o b l e m has been in major part o v e r c o m e b y the s e c o n d g e n e r a t i o n of ceramic m e m b r a n e s exhibiting a m o n o l i t h geometry. The characteristics of the m a i n m o n o l i t h ceramic m e m b r a n e s c o m m e r c i a l l y available are r e p o r t e d in Table 12.2. A better v o l u m e / s u r f a c e ratio results from these n e w g e o m e t r i e s while intrinsic p r o p e r t i e s of ceramic micro- a n d ultrafiltration m e m b r a n e s are p r e s e r v e d . C o m p a r e d to organic p o l y m e r m e m b r a n e s , m u l t i c h a n n e l ceramic m e m b r a n e s exhibit i n t e r m e d i a t e h y d r o d y n a m i c p e r f o r m a n c e s b e t w e e n those of t u b u l a r a n d h o l l o w fibres p o l y m e r m e m b r a n e s . Liquid flow in a p i p e can be TABLE 12.2 Characteristics of commercially available monolith ceramic membranes Mmlufacturer
Trade name
Membrane Support material material
Membrane Geometry of Chalmel pore membrane inside diameter element diameter
CERAMEM
metal oxides
cordierite 0.050.5 ~tm
honeycomb monolith
CERASIV
metal oxides
ocA1203
monolith
5 nm1.2 ~m
1.8 mm 3 mm
4.2 mm 6 mm 8 mm
TECH-SEP
KERASEPTM
TiO2 ZrO2
US FILTER
MEMBRALOX| metal oxides
czA1203- 2 n m TiO2 1/.tm
monolith
czA1203
monolith
4 nm1.4 ~m
4.5 mm 2.5 mm 6 mm 4 mm
584
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
described according to the following parameters: the kinematic viscosity ~t = rl/p, the hydraulic diameter d h and the feed flow velocity v. Laminar or turbulent flow in the pipe can be achieved for specific values of v and d h according to Reynolds number R e :
dhV 9Vdh
Re - ~
- ~
(12.22)
For undisturbed flow through a straight pipe, the change from laminar to turbulent flow occurs at a Reynolds number of about 2000. The behaviour of fluid flow in a porous channel with suction is different than that in a non-porous-walled channel. Mellis et al. [21] showed that in a porous stainless steel microfiltration tube, axial, ReF, and wall, Rew, Reynolds numbers influence axial pressure drop and related friction factor. Wall suction has its maximum effect on the axial pressure drop at intermediate flow rates (1,000 < R e F < 15,000). Moreover the value of R e F for transition to turbulent flow increases with R e w . The conclusion is that wall flux has a feedback effect on itself through the axial pressure drop and, hence, the transmembrane pressure drop. More recently, Nakao et al. [3] proposed to refer to axial and radial Reynolds numbers to improve mass transfer prediction at wall for high flux operations. Regarding the influence of membrane geometry on Reynolds number one can see that with monolith channels of a few mm in diameter, a turbulent flow can be attained with reasonable pumping energy. This is an important point in the economics of processes based on the utilization of ceramic multichannel membranes compared to tubular geometry. Membrane
material
Several characteristics of ceramic membranes, resulting from metal oxide grains sintering, are worth underlining: a composite structure made of a porous ceramic multilayer structure, - a pore geometry resulting from sintering of packed particles, an amphoteric behaviour of the ceramic surface in contact with water. The volume flux through the different porous layers cannot correctly be described by the basic Hagen-Poiseuille equation more specifically suited for transport through membranes consisting of a number of parallel pores. As above-mentioned volume flux in inorganic membranes can best be described by the Carman-Kozeny relationship, Eq. (12.20), taking into account the possible interpenetrating of the different layers. The amphoteric behaviour of metal oxides in contact with water has thoroughly been described by many authors [22-24]. This basic property results in charged surfaces depending on pH condition. In a first approximation, connected porosity in ceramic membranes can be represented by an array of -
-
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
585
a) aquo complex formation HO- H I ,~
H
7,, ... 9 :... 9 .
. 9
.
:
M 9 . "
.
M .
." . " .
.
""
:
': . . . . .
oI
M "':"
H
/o\ "
M "
:"]"
:. . . . . . . . .
solid surface
M :.i
"
.."
. . . .
oI
M ...
. . . .
7.:::"
:i':.
i:'::.::%.
" ....
"
b) amphoteric behavior ii~i:i~i:ili:il~i!il ::::::::::::::::::::::::::::::: .::...:::::::.::: ..:::::::::::.:::
.::: -: :.:..:
M - OH2+ + O H
// ~
.
iiiiiiiii!iiiiii::M - - OH + H20 ?:~i:i!i::i!i!:i;::i
M - O + H30 +
............
Fig. 12.9. M e c h a n i s m of c h a r g e d surface f o r m a t i o n d u e to a m p h o t e r i c b e h a v i o u r of m e t a l oxides.
microcapillaries. This approximation allows to apply the so-called microcapillary model for transport of electrolytes [25]. The consequences of applying pressure and potential gradients across capillaries (or porous media) filled with electrolytes are based on phenomena such as electro-osmosis, streaming potential and electroviscous retardation effects accompanying flow of electrolyte solutions. This has a direct influence on permeability and selectivity of microand ultrafiltration ceramic membranes and in a more pronounced manner in the case of ceramic nanofilters for which the effective thickness of diffuse electrical layer formed on pore walls is of the same order of magnitude than pore radius. A short description of the formation of this diffuse electrical layer on metal oxide surface is given hereafter. The occurrence of neutral and charged (+) surfaces at the oxide-solution interface has been attributed to the formation of metal aquo complexes as shown schematically in Fig. 12.9. One can see from Fig. 12.9 that the oxide surface can be negatively or positively charged as a function of pH. There exists a pH value defined as the zero point of charge (zpc) for which there is no charge on the surface. The negative surface charge originates from acidic dissociation of the surface hydroxyl groups (at pH > zpc) and increases with increasing pH. Cations present in the solution adsorb at the surface according to the double layer theory. At pH < zpc the positive surface charge is explained on the basis of proton addition to the neutral aquo complex, together with or without the replacement of the surface hydroxyl groups by anions present. It results from the adsorption of ions on charged oxide surfaces the formation of a double layer as shown in Fig. 12.10. It is generally accepted that the electrical double layer around charged oxide surfaces consists of two parts [22]:
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
586
t3"0 ~
~d
(3"0+(~13+(3'd-'0 [
| bulk liquid
@
iiiii!!iiiiiii i:!!i'~i~i2!!!?!:!i!'~!~
Q,
i:ii~gii;iii!ii!ii!i !:iiiiiii!iii-%ii ::::::::::::::::::::::::: ::::::::::::::::::::::::::: ::::::::::::::::::::::::::: iiiii~ili i:!!i:ii::ii!iiiii
~ii!il;!!ii!i!i!iii
i.l.....-.......,..... !;i~i!ii~i;....li~i~i :::::::::::::::::::::::::::
structured water
N~--q S t e r n l a y e r
~
Gouylayer
Fig. 12.10. Schematic representation of double layer formation in presence of electrolytes. - an extended diffuse outer part or Gouy layer, the potential distribution of which is well described by the Poisson-Boltzman equation, - an inner part or Stern layer of a few molecular layers thickness, reflecting the specific properties of the counterions and the nature of the surface. The potential profile through the double layer of an oxide surface in presence of electrolytes, Fig. 12.11, can be described by successively accounting for the resulting charges at the surface of the oxide, in the plane of specifically adsorbed ions and at the interface between Stern and Gouy layers. The above description of double layer formation on oxide surfaces in presence of electrolyte solution can be applied to pore wall in ceramic membranes. When the porous volume of a ceramic membrane is filled with an electrolyte solution pore walls are uniformly charged and surrounded by a diffuse electrical layer with an effective thickness which is given by the Debye length ~,D = 1/~c. ~c is the reciprocal of the Debye length and is given as:
_ _F 2Z i2CiO/
f
~c ~eRT
1/2
(12.23)
with e the permittivity of the solvent, zi the ion valence and C ~ the ion concentration at the pore axis r = 0. As shown in Fig. 12.11, the charge density G0 of the native surface of the pore walls is partially compensated (r~0 > r~) or overcompensated ( I c~01 > I c~l ) by a charge density r~ of specifically adsorbed counter ions in the Stern-layer. The resulting charge density (~d at the shear surface between Stern-layer and diffuse Gouy-layer results in an electrical potential ~)(r) in a perpendicular direction r
587
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION (3"O
(~B
O'B
(3"O
G'd
(Td
4)o
4)o
i Diffuse
DifFuse layer
~d
,, ~ 1 Stern-layer
i
.. r
specific adsorption =:> cyo > CYl~
layer
.
~
v
~d I "
super ,er-equivalent :=> I~01> )ecific adsorption soeciJ
I ,1
Fig. 12.11. Representation of potential evolution in a perpendicular direction to the oxide surface when counter ions are specifically adsorbed on the surface.
to the pore wall. This potential, called zeta potential, (~d, c a n be related both to the characteristics of metal oxide membrane and to the feed ionic strength. According to the space charge model (SC), when a solution is flowing in the porous structure under a pressure gradient, pore wall is reduced to the Sternsurface between the static and the mobile portions of solution. The pore radius equivalent to the Stern-surface is called hydrodynamic radius rh with rh = rp- l
(12.24)
where rp is the original pore radius and I usually taken as a counter-ion diameter. So that when a pressure gradient, Ap, acts through the membrane, the solution close to the pore wall stays immobile while the rest of it moves along the pore. This movement leads to the appearance of an electrical potential drop from one side to the other side of the membrane, A~g.This electrical potential results from an electrical field which develops because the flux of the counter-ions is greater than that of the coions into the membrane pores. This electrical field generates an electrokinematic flow of the counter-ions that is opposed to the previous one thereby satisfying the constraint that there is no net current flow through the membrane. The combination of ~)(r) and A~geffects corresponds to a dynamical contribution to the total electric potential profile, according to the space charge model which was originally proposed by Osterle et al. [26-28].
588
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
O(r,z) = (ziF/RT) ~(r) + ~ ( z )
(12.25)
The z-dependent potential ~(z), in zero current conditions, is related to streaming potential Vp. In the case of ceramic membranes ~)d and Vp are the two quantifies which influence both retention of charged solutes and volume flux. ~)d is related to ionic strength and pH of the feed solution; Vp is related to ionic strength but also to the transmembrane pressuredriven flow resulting from the pressure gradient Ap applied to the membrane. As a whole some general rules can be pointed out concerning the effect of ~)d and Vp on membrane behaviour. The spatial extent of the double layer in the radial direction of the pores is characterized by the Debye-length so as a high ionic strength leads to a short Debye-length and a weak electric effect on transport. The ionic strength is related to ions concentration in the feed solution but also to the valence of co- and counter-ions. Multivalent counter-ions which adsorb in the double layer have a more marked effect than monovalent counterions in diminishing the spatial extent of the double layer and the resulting zeta potential. On the contrary, due to their higher electric charge, multivalent coions are more rejected than monovalent coions. Taking into account the distribution of charges in the radial direction of pores, volume flux in the axial direction can be described by the addition of two opposite contributions: the convective-diffusive flow and the back electrokinematic flow due to streaming potential. The electrokinematic flow for a porous ceramic membrane can be related to the following parameters: - -
-
4?s - (ziF/R T)(~d
(12.26)
a dimensionless zeta potential dependent upon ceramic surface characteristics and pH of feed solution; rE = r p / ~ D
(12.27)
a dimensionless length also called electrokinetic radius accounting for the ionic strength of ionic feed solution; Ls = Lp/Lo
(12.28)
a dimensionless hydraulic permeability resulting from the variation of rE and ~)s. L0 is the pure water permeability for the membrane with the same pore radius and a zero surface charge density while Lpis the membrane permeability in presence of a counter-electrokinematic flow. Due to Donnan exclusion principle [29] charged membranes can reject inorganic salts even though they have pores much larger than the salts and this ion rejection is known to decrease with increasing feed ionic strength. The example of 1.1 electrolyte filtration through different pore sizes at a pH far from the
121 TRANSPORTANDFOULINGPHENOMENAIN LIQUIDPHASESEPARATION Dolman b--x~ effect
Streaming
bound~ ~ layer
membrane
boundary layer
,-.......
1
589
rp~l nanofiltration
n+@
@ C )~~, .' -.@ . . . - @ : [ :@ " . .~ _ @ ~ @ @
o
Stem-layer
.........| ..........~ . ~ / / / / / / / / ~ ~ rp>l ultrafiltration
............. .| ........ | .i__~ |174 @.@.@. o.:..,~-...,.,,--@.@__| @ !]..........@ G | =.=..-:::,.,7.=.i-.-. | |
e -~ ~ ~ ~ ~ 1 | .......G . . . . . e . . . . . |
|
|174g4:::~---"-:::G-~ | |174
G _ N ~ ...@ ~ @ ...
..... |
G
| |
|
o
rp>> 1 microfiltration
e
convective flow ~ -.-.:r,~--- electrokinematic flow Fig. 12.12. Influence of zeta-potential (Stem-layer thiclaless l) and Streaming-potential (electrokinematic flow) on ion rejection and volume flux for porous ceramic membranes exhibiting negatively charged pore walls. Cases of micropores (nanofiltration), mesopores (ultrafiltration) and macropores (microfiltratio11).
isoelectric point (high zeta-potential for ceramic m e m b r a n e materials) is given in Fig. 12.12. According to the above described dimensionless p a r a m e t e r s the occurrence of electrokinematic flow (Ls < 1) is expected for rE < 1. Usually electrokinematic effect is likely to occur for nanofiltration and ultrafiltration
590
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
membranes when deci- to centimolar electrolyte solution are used. Nevertheless this effect can be expected for microfiltration membranes in the case of very low ionic strength resulting in a Debye-length which were calculated to reach several tens of nm [30-32].
12.3 R E C E N T D E V E L O P M E N T S ULTRAFILTRATION
IN MICROFILTRATION
WITH CERAMIC
AND
MEMBRANES
Ceramic membranes were first applied to microfiltration processes. Several authors published a comprehensive description of basic transport phenomena involved in ceramic macroporous structures [1,33]. Lately improvements in ceramic membrane processing led to commercial ultrafiltration membranes exhibiting a mesoporous structure with transport phenomena close to those encountered in microfiltration. As described in Section 12.2, the major limitation in membrane performances for micro- and ultrafiltration processes is caused by concentration polarization and fouling. Methods that help to reduce concentration polarization and fouling can be classified into three categories: (i) chemical cleaning methods including strong acid and basic solutions or oxidizing agents due to high chemical resistance of ceramic membranes; (ii) physical methods such as backflushing and the use of turbulence promoters; (iii) hydrodynamic methods related to module design. In fact two aspects have been more specifically investigated in recent years concerning cross-flow filtration systems based on ceramic membranes: - the hydrodynamic of microfiltration and ultrafiltration systems and its influence on membrane performance in terms of fouling reduction and permeability enhancement; - the influence of membrane material (metal oxides in most cases) on selectivity and permeability.
12.3.1 Hydrodynamic of Micro- and Ultrafiltration systems In a review on cross-flow microfiltration Belfort et al. [34] outlined the importance of module design and hydrodynamic operating conditions in order to improve performances of cross-flow filtration using macroporous membranes. The authors suggest that unsteady flow conditions can be even more effective in disturbing the flux-limiting effects of concentration polarization and fouling [35]. Various approaches to inducing instabilities in bulk flow across a membrane surface include designing membrane surfaces with organized roughness, pulsation of axial and lateral flow, and the use of curvilinear flow under conditions that promote instabilities or vortices. A number of these devices shown in Figs. 12.14 and 12.15 can be adapted to ceramic membranes.
12 ~ TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
591
mlalltltJm~mmltl~,.-..~ r
porous wall
parabolic flow
with pulsation
Fig. 12.13. Effect of pulsations on flow profile in a s m o o t h - w a l l e d duct.
Permeate flux enhancement by pressure and flow pulsations has been investigated by many authors [36-38]. The effect of flow oscillations in a smoothwalled duct is shown in Fig. 12.13. P u l s a t e flows were applied to mineral microfiltrations membranes during apple juice filtration [36] illustrating the advantage of this method to enhance permeability compared to steady flow regime. With carefully chosen pulsations permeate flux increased up to 45% at I Hz pulsation frequency. Moreover well defined pulsations decreased the hydraulic power dissipated in the retentate per unit volume by up to 30%. In an other work on cross-flow filtration of plasma from blood [37] permeate flux increase was also observed when pressure and flow pulsations at I Hz are superimposed on the retentate. Pulsate flow can also be used to good advantage in rough walled ducts and those with inserts. Simulation of cross-flow filtration for baffled tubular channels and pulsate flow were reported by Wang et al. [38]. Wall and central baffles, in a similar way as in Fig. 12.14, with and without pulsations have been considered. Reynolds numbers up to 200 have been used in simulation that is notably lower than values used to obtain turbulent flow in smooth porouswalled channels. Concentration polarization effects have been included in calculation. The addition of pulsations improved the fluxes, the relative improvements being greater for the wall baffles. However the absolute values of the predicted fluxes were found greater for the central baffles. It has been suggested in the literature [34] that filtration devices producing Taylor or Dean vortices can help depolarization of the solute build up on membranes. This seems to be an attractive way because of excellent bulk fluid mixing, high wall shear rates and weakly decoupled cross-flow with transmembrane flux. Unfortunately there are some severe limitations on a technical and economical point of view with such devices. Build up and scale up of these modules are expensive with difficulties in repairing and changing membranes. A good compromise between economic and technical constraints has been described by Charpin et al. [39]. It consists in the preparation of mineral (metal
592
12 ~ TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
(a)
~
"
K
,
~
protuberance
~.x~-:-_--_--_--_--:__-_~_, ~ , ~
.....................
~ - ,.:...~,,<..,.,.:.:.:,.:.::;.~!~, . !;~...;.. : :..;?r..:..:.':...-.'..-:....>:.?!:.-.:.2.;;::~;~;~"~~:7:~!:':".:;:~:?.: : "'.i~!~":.~":r>; I porous wail i
.....~-.~......................... . ] porous wall [,. . . . . . ~
~ (b)
------
| , s ,,. /.~,;:.
x
~
~
J
.
~'''--'--''~
I inserts
~
i ;.,;~ ;,:,,4.s. ;~,.~:,f.~.~,,:..,,,:.~.
:L',,:,:i:
~::., ',','.::',.
:;~',.:
..~,;i'.4.is.~i.:s:~.::Ss:.~:./:;:'.,'z.:
~ . . - . 5;~.: 9
t!,.!~.-.'~.-;_.:.;C::'_,:!.:_-_.,;..,7.~.:.:_:~.,::,."._~ :-_".~i..................... ;; ~::"F.":,'.'~:':":~"~:"Y~"~)~/......?;';"~:1
}porous wa. !
. . . .
Fig. 12.14. M e t h o d s for haducing flow instabilities: (a) plachag objects ( p r o t u b e r a n c e s , baffles) o n t o the m e m b r a n e surface to f o r m a r o u g h surface, (b) plachag objects hato the flow charmel a w a y f r o m the m e m b r a n e surface. flow x
,'."
.~'~'.'/:'/'
:,,..;.~,.~~.,~.~z , .: ". .",'..:..
"~ ~. " '.'," ' / .~,~,',
, .
/---
tubular helicoidal shape
,j
~.~
,
,~?,~..~'
Fig. 12.15. Schematic representation of a tube with ma il~ler surface exhibiting an helicoidal profile.
or ceramic) m e m b r a n e s exhibiting an inner helicoidal structure as s h o w n in Fig. 12.15. These helicoidal shaped tubes can be sealed in m o d u l e s in the same w a y than with classical tubes or monoliths.
593
12 m T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
Rotating disc systems have also been described as efficient devices to overcome flux limitation due to matter deposit on membrane surface during crossflow filtration [40]. The problem of erosion of a macroscopic particle solid deposit on a rotating disc membrane has been quantified by Aubert et al. [41]. The influence of the transmembrane pressure, the thickness of the initial deposit and the pore size on the critical shear stress have been investigated and described by empirical fits. It results from this study that fouling is more efficiently eliminated at high Ap and large pore size.
12.3.2 Influence of Membrane Material on Permeability and Solute Rejection The influence of metal oxide derived membrane material with regard to permeability and solute rejection was first reported by Vernon Ballou et al. [42,43] in the early 70s concerning mesoporous glass membranes. Filtration of sodium chloride and urea was studied with porous glass membranes in closeend capillary form, to determine the effect of pressure, temperature and concentration variations on lifetime rejection and flux characteristics. In this work experiments were considered as hyperfiltration (reverse osmosis) due to the high pressure applied to the membranes, 40 to 120 atm. In fact, results reproduced in Table 12.3 show that these membranes do not behave as hyperfiltration membranes but as membranes with intermediate performances between ultra- and nanofiltration in which surface charge effect of metal oxide material plays an important role in solute rejection. Rejection data for NaC1 were explained according to a low-capacity ion exchange mechanism. The ion exchange mechanism put forward in this work is not consistent with the porous structure of the membranes and the high transmembrane pressure used in the filtration experiments. Ion exchange TABLE 12.3 Rejection of NaC1 (58.5 g mole-1) and Urea (56 g mole-1) ushlg mesoporous glass membranes over a range of solute concentration, from Ref. [42] Solute
Feed concentration g/1
R tool/1
NaC1 NaC1
0.47 1.30
0.008 0.022
0.86 0.68
NaC1 Urea Urea Urea
9.11 1.74 3.80 11.32
0.156 0.029 0.063 0.189
0.42 0.41 0.38 0.37
594
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
mechanism is better related to the working conditions described by Singh and Singh [44] for zirconium phosphate membranes. Regarding the work of Vernon Ballou [43] pore sizes were calculated from nitrogen isotherm data. Narrow pore volume distributions of unused glass membranes were found between 1.9 and 2.2 nm pore radius. One can see that NaC1 rejection cannot be explained by an hyperfiltration mechanism based on selective diffusion of water through the membrane. Moreover the decrease of NaC1 rejection when ionic strength of the feed solution increases is in favour of mass transport through an array of metal oxide microcapillaries filled with electrolytes. Unfortunately these glass membranes were not stable over a long period of time due to solubility of silica. Interpretation of the results were altered by a loss of rejection and an increase of permeability mainly caused by broadening of pore size distribution with time. Since then, the evidence of pH effect on cross-flow micro- or ultrafiltration using ceramic membranes has been given in the literature [45-48]. Hoogland et al. [45] showed that permeability of a Ceraflo (Norton) (x-alumina membrane towards pure water and mineral slurries was dependent of pH. For pure water the maximum of permeability was found in a pH range near the isoelectric point of the membrane (low pH) while the membrane resistance was highest far from the isoelectric point (high pH) when the charge of the membrane is strongly negative. At high pH, this effect can be explained by the flow through a charged porous barrier which leads to electro-osmosis phenomenon and an effective loss of permeability. Fluxes measured with mineral slurries (silica particles) were also dependent on pH. Higher permeabilities for the membrane were found at low and high pH while flux decline was maximum at intermediates pH. In this case two phenomena due to pH act simultaneously. One is related to the alumina membrane, the other to silica particles. At low pH near the isoelectric point of the particles, there is formation of large-size flocs generated by aggregation of weakly charged particles. These flocs prevent penetration of the individual particles inside the porosity and are easily removed from the membrane surface by the effect of cross-flow. At high pH both the membrane and the particles exhibits negative charges which lead to repulsive forces at the membrane-solution interface and depolarization of the membrane. At intermediate pH polarization and membrane resistance are maximum. One important parameter, the ionic strength of filtered solutions, was not investigated in this work. The effects of pH and ionic strength on the performance of an (x-alumina microfiltration membrane from U.S. Filter was evaluated by Nazzal and Wiesner [46]. Concerning pH effect on flux, results obtained in this work perceptibly differ from the previous one. Here the membrane operated at a significantly higher permeation rate at a pH well below the isoelectric point of the membrane. This variance can be explained considering the isoelectric point of the membrane was found at pH = 8.3 in this case while it was at pH = 3.5 in the
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
595
previous work. It can be observed that in both cases m a x i m u m of flux for pure water was measured at p H 3 to 4. Results concerning filtration studies with 0.2 ~tm titanium dioxide membranes supported on stainless steel or ceramic porous tubes were recently reported by Porter et al. [47,48]. Solutions containing sodium nitrate alone and in the presence of anionic, direct and acid dyes were filtered with adjusted solution pH. Electrolyte rejections and colour rejections were measured at p H values from 4 to 10. They showed that the charged membrane was responsible for ion rejection at low ionic concentration while rejection decreased to near 0% as the salt concentration was raised to 5000 ppm..These results are consistent with long range forces associated to Debye-length which can reach several hundred Angstroms in the solution for very low ionic concentrations. 12.4 NANOFILTRATION
WITH CERAMIC
MEMBRANES
In the early 1970s, several authors described separation membrane processes with intermediate performances between reverse osmosis (RO) and ultrafiltration (UF). Typically retention for these membranes was in the range of 50-70% for sodium chloride while it was in the 90% for organics. In the 1980s a suggested definition for these membranes was based on a molecular weight cutoff of 1000. Then "nanofiltration" was considered a suitable name for such a process which rejects molecules in the nanometer range. Presently basic properties of nanofiltration membranes can clearly be defined compared with ultrafiltration or reverse osmosis membranes: - a molecular weight cutoff of less than 1000 (membranes with MW cutoff of 1000 and above are considered UF membranes), - a lower transmembrane pressure and a higher flux than for RO, - a mixed mass transport mechanism involving convective and diffusive fluxes for both solutes and solvent, - in most cases membrane charged either positively or negatively due to their materials, - a marked influence of Donnan mechanism in the case of an aqueous feed solution containing mixed electrolytes. It results from these basic properties that nanofiltration offers unique performances for the separation of salts and organics. A negative salt rejection has been evidenced in these membranes which can be explained with reference to the above-mentioned capillary model in which the structure of nanofiltration membranes is represented by a bundle of charged capillaries with a pore radius in the nanoscale. In practice, this negative salt rejection effect can be usefully exploited in industrial desalting-concentration processes of molecules exhibiting molecular weight of less than 1000. In fact nanofiltration membranes are finding increased applicability in various fields but their transport mechanism
596
12 E TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
is not yet well understood. Up to now a number of published papers deal with the description of transport properties of organic nanofiltration membranes [49-51]. On the other hand, few data are available in the literature concerning ceramic nanofilters. In the following, recent results concerning separation properties of ceramic nanofilters are presented showing that some of these basic properties are relevant to describe mass transport and solute rejection observed with microporous ceramic membranes. Ceramic nanofilters are a new class of ceramic membranes which obey the basic properties of nanofiltration membranes. Some similarities can be noted between organic and inorganic NF membranes behaviour; however specificities exist with ceramic membranes due to the amphoteric properties of metal oxides in water media. Basically the structure of ceramic nanofilters can be described according to concepts developed for nanophase materials. The active layer is made of a supported microporous layer with a thickness in the micron range. This microporous structure which results from sintering of ceramic grains of less than 10 nm in size leads to membrane materials with a high surface area. Metal oxides already used for the preparation of micro- and ultrafiltration membranes can also be used for nanofilters. Microporous y-alumina, titania, zirconia and silica supported layers have been described by Julbe et al. [20] with suitable characteristics for nanofiltration. However, regarding industrial applications of these membranes for aqueous filtration, zirconia and titania are preferred to silica or y-alumina because of their stability in large pH and temperature ranges. The main characteristics of nanofiltration membranes made of oxide ceramics is that they exhibit a microporous structure with charged pore walls depending on pH and ionic strength of feed solutions. Three main cases are distinguished in the discussion of mechanisms involved in permeation and separation processes using microporous ceramic nanofilters: separation of neutral solutes in absence of electrolyte; - separation of pure electrolyte mixtures, - separation of solutions containing both organics (ionisable or not) and electrolytes; -
12.4.1 Separation of Neutral Solutes in Absence of Electrolytes When Donnan contribution can be neglected (case of neutral solutes), membrane cut-off can be determined based on respective sizes of model solutes and membrane pores. Mass transport can be described using both basic concepts of ultrafiltration and specific aspects of transport in micropores. Pure solvent flow can be described as a convective flow with a linear dependence to transmembrane pressure as shown in Fig. 12.16. With nanofiltration membranes a minimum value of pressure gradient has to be applied before to observe
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
597
Jv
l
increasing pore diameter
Ap Fig. 12.16.Schematicrepresentationof hydraulicpermeabilityversus transmembranepressure for a microporous membrane. a solvent flux through the membrane. This is due to the occurrence of important capillary forces in the case of micropores of less than 2 nm in diameter. In the presence of solutes with small molecular weights, concentration polarization is likely to occur but with much less effect than in the case of ultrafiltration as explained in Section 12.2.1. A theoretical model concerning separation of sucrose and raffinose by ultrafiltration membranes has been proposed by Baker et al. [53] which assumes transport of solvent and solute exclusively through pores. This model can apply to ceramic nanofilters as they exhibit a porous structure with a pore size distribution. The retention characteristics of a given membrane for a given solute is basically determined by its pore-size distribution. The partial volume flux jv through the pores which show no rejection to the solute can be expressed as a fraction of the total volume flux Jv. (12.29)
jv= f . Jv
The solute rejection is then given as a function of the total water flux, of the solute diffusion coefficient Ds and of the pore fraction el permeable to the solute:
fexp Z R -
100 1 --
~al"/ds ) h
(12.30)
f - 1+ e x p / l i D s ) It follows from Eq. (12.30) that as Jv goes to zero, the exponential term goes to unity and the rejection coefficient reduce to zero. On the contrary as Jv tends to become very large, the exponential term tends towards infinity, and the rejection coefficient approaches a specific limiting value for a given solute. The same evolution of the rejection coefficient with volume flow and indirectly with transmembrane pressure was predicted by Tremblay [54] using the finely porous model proposed by Merten [55] and modified by Mehdizadeh
598
12 u TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Separation factor
100
r ~,"" ~ ' ~
(%)
//
I . /
;
, /~/
80
9
60
-
.
I/.1'1"/
0
; :
e 0 0
r't "
40
Increasing Pressure Gradient
20
0
i 100
1000
I
10000
100000
Solute molecular weight Fig. 12.17. Evolution of separation factor versus solute molecular weight for different pressure gradients. Results obtained from mass transfer across micropores using radially averaged friction factors [54].
and Dickson [56] in which a radially friction factor b-1 has been included. This friction factor is directly related to the ratio ;~ of the solute radius to the pore diameter and represents the friction between a solute molecule translating along the centre line of a cylindrical pore and the pore wall.
1 P dCi 1
Ji- - ~ ~ d---~+ b Ci Jv = (1 -
~)2
b
(12.31) (12.32)
with b given by the Faxen equation b = 1 - 2.1044;~ + 2.089~ 3 - 0.948~5
(12.33)
It results from this approach that separation factors will depend on the ratio ;~ and on the operating pressure. As shown in Fig. 12.17 the influence of friction factor on separation factor is predominant at high pressure gradient.
12.4.2 Salt Rejection of Electrolyte Solutions Salt rejection of a single electrolyte by a nanofiltration membrane in the absence of Donnan contribution can be described by Eqs. (12.9) and (12.10) according to the work of Spiegler and Kedem [57]. With ceramic nanofilters the Donnan contribution has to be taken into account due to the amphoteric
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
599
behaviour of metal oxide surface resulting in membrane materials with charged pore wall. The extended Nernst-Planck equation (12.19) has been applied by Tsuru et al. [58-60] to predict ion rejection in the case of charged membranes for single and mixed electrolytes. This approach showed good agreement with mass transport description obtained from irreversible thermodynamics. The general tendencies for ion rejection are as follow" for a single electrolyte solution, rejection dependency on volume flux is the same as that of neutral solutes. Increasing the charge density in the membrane make rejection higher. Rejection of divalent coion electrolyte is expected to be higher than that of monovalent coion electrolyte, while divalent counter-ion rejection seems lower than that of monovalent counter-ion electrolyte; - for a mixed electrolyte solution, rejections are shown as strongly dependent upon the volume flux, mole fraction, and the ratio of the feed concentration to the membrane charge density. Mono- and divalent coions are suggested to be separated effectively, and the monovalent coion to show negative rejection under a certain condition. However, mono- and divalent counter-ions are not so effectively separated as coions under ordinary conditions. Recently Wang et al. [61] proposed the comparison of different models from the literature to describe electrolyte transport through nanofiltration membranes. The space charge (SC) model described in Section 12.2.3 was compared with the Toerell-Meyer-Sievers (TMS) model. The SC model assumes that ion concentration and electric potential have a distribution in the radial direction in the membrane, while the TMS model supposes that both of them held uniform. The evolution of ion rejections versus Peclet number (Pe) for a 1-1 electrolyte (KC1) were compared for the two models respectively with increasing charge density at constant pore radius (5 nm) and for decreasing pore radius at a constant charge density (3.336 C.m-2). In agreement with general expressions derived from linear, non-equilibrium thermodynamic theory [7] the rejection was found to increase with Pe. This is consistent with the fact that at small Pe number there is a dominant contribution of diffusion to electrolyte transport while contribution of convection is dominant at large Pe number. With decreasing pore radius, the rejections calculated from the two models tend to coincide and shows almost the same value for pore radius of I nm. This can be explained because an overlap of double layers into the pores due to a Debye-length equal or larger than the pore size. This overlap of double layer renders the distribution of concentration and electric potential uniform in agreement with the TMS model. According to definitions of electrokinetic radius rE and dimensionless hydraulic permeability Ls given by Eqs. (12.27) and (12.28) the authors calculated evolution of Ls versus rE. In Fig. 12.18, curves Ls =firE) drawn at different potential gradients show that a maximum effect of the electrical force is expected for rE --- 1 and high potential gradient. -
600
12
Ls
-
-
TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
0.95
0.9 " 1%(;) I 9I .047 [ b 1.094 [ c 1.234 I ~ 1.46e l e 1.936 f | 2.34 g 14.68
0.85 0.8 0.75 . . . . . . . . . . 0.1
J!
9
10
1 i"E
Fig. 12.18. Dimensionless water permeability as functions of electrokinetic radius at different potential gradient (q0) [61].
In the above-considered works the behaviour of electrically charged nanofilters towards electrolyte solution has been mainly regarded with respect to Donnan analysis to explain the coion rejection. Bardot et al. [62] looked at the effect of transmembrane transport kinetics on counter ion rejection through an alumina/polysulfone composite membrane in the case of electrolyte mixtures. The rejection phenomenon is based on a "decompensation" of the convective and electric flows of a given counter-ion as a consequence of the addition of counter-ions with a different mobility. It has been shown both theoretically and experimentally (for negative charged membranes) that the same physics accounts for not only the improvement of the retention of more mobile counterions upon addition of less mobile but also for a significant deterioration of the retention (down to the negative one) of less mobile counter ions upon addition of more mobile. Experimental correlations of the phenomenon with the ratio of mobilities of counterions, the concentration of starting electrolyte and transmembrane pressure difference (Pe number) have been in complete agreement with theoretical predictions. However the influence of the ceramic support versus pH of the feed solution, which can be of great influence on ion rejection, is not discussed. The evidence of electrokinetic salt rejection by a microporous inorganic material was given by Jacazio et al. [63] based on the model of Osterle [26-28]. Experiments were carried out on the salt rejecting properties of compacted clay through which saline solutions were forced under high pressures. In accordance with the model the performance of the porous material was shown to depend on three main parameters: the ratio of the Debye length to effective pore
12 -- TRANSPORTAND FOULING PHENOMENA IN LIQUIDPHASE SEPARATION
601
radius; a dimensionless wall potential related to the ~ potential; and a Peclet number based on the filtration velocity through the pore. Comparison between the experimentally determined and theoretically predicted rejections of potassium chloride in the case of effective pore radius in the range 1-2 nm were shown to be excellent. Regarding the rejection of salt mixtures with inorganic membranes AlamiYounssi et al. [64] investigated the performance of a y-alumina membrane. In aqueous media containing indifferent electrolytes such as NaC1, the point of zero charge (zpc) of the y-alumina is near 8.5; in the presence of divalent anions or cations which are able to form surface complexes respectively with the surface groups A1OH ~ or A 1 0 , the zpc of the material can be shifted, respectively, towards higher or lower pH values, pH values for feed electrolyte solutions were measured to be in the range 5-5.9, which means that the membrane is positively charged. Results are discussed only in terms of effective charge of the membrane and valence of the co- and counter-ions present in the feed solution. In this case the membrane is positively charged and the rejection obeys to the prediction of Tsuru concerning mixed electrolyte solutions. Measured rejections are reported in Table 12.4. Rejections were shown to depend on the charge of the ions and decrease in the order: (divalent cation, monoanion) > (monocation, monoanion) or (dication, dianion) > (monocation, dianion). Another work from Rios et al. [65] also deals with performance of a positively charged y-alumina membrane fed with single NaC1, MgC12, Na2SO4 and MgSO4 solutions at various concentrations (10-4 to 10-1 mole-l-I), or even with electroTABLE 12.4 Rejection with a y-alumina membrane of mixed electrolyte water solutions [64] Sodium and calcium nitrates [Ca 2+] feed (M)
0
10-3
10-2
10-2
10-2
[Na+] feed (M)
10-2
10-2
10-2
10-3
0
Rejection N O 3-- (%)
68
Rejection Ca 2+ (%)
75
75
93
96
90
95
95
96
63
38
47
Rejection Na + (%)
68
Potassium and sodium nitrates [Ca 2+] feed (M)
0
10-3
10-2
10-2
10-2
[K +] feed (M)
10-2
10-2
10-2
10-3
0
Rejection N O 3- (%)
55
Rejection Ca 2+ (%) Rejection N a + (%)
55
18
45
50
68
32
60
56
68
15
18
25
602
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
lyte mixtures (NaC1 + MgC12; Na2SO 4 + MgSO4) of constant counter ion concentration ( 1 0 -3 mole.l-I). Results that confirm the previous trends are explained using a new simplified model based on Eq. (12.19) that makes the assumption of Donnan effect at pore entrance. This model accounts for electrokinetic phenomena inside the pores and also considers differences in ion mobility. Zirconia nanofilters (partially stabilized or not) have been investigated by Guizard et al. [66,67] with respect to rejection performance towards model solutes. These membranes were synthesized by the sol-gel process using zirconium and magnesium alkoxide precursors, the later being used as stabilizer agent for the cubic zirconia phase. Pore diameter for these membranes is in the range I to 2 nm depending on preparation conditions. In agreement with data published in the literature a zpc near 7 has been evidenced. In this work special attention has been paid to the influence at one hand of pore size and specific surface area of the membranes, on the other hand of transmembrane pressure and ionic force of feed solutions. It has been shown that these parameters clearly relate to the dimensionless zeta-potential %, the electrokinetic radius rE and the dimensionless hydraulic permeability Ls resulting from the variation rE and %. The rejection versus pH of chloride and sulfate ions using a Na2SO4/NaC1 mixture (200 ppm) is shown in Fig. 12.19. At a pH < zpc chloride and sulfate must be regarded as counterions for the membrane while at pH > zpc they behave as coions. One can see that results are in good agreement with the prediction; sulfate are better rejected than chloride when they are coions of the
Rejection (%)
I00
',
,
'
!
'
'
'
I
,
,
-- Su,fate j
--~
80
60
Chl~
,
!
,
,
,
!
i
i
i
i
8
10
,
,
,
I
'--'~. iP."C~............i
40
................
. . . . . . . . . .
-
20
2
4
6
12
pH Fig. 12.19. Rejection of an electrolyte mixture Na2SO4/NaC1 (200 p p m ) b y a zirconia nanofilter. Effect of p H [66,67].
603
12 - - T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
Permeability (l/h.mZ.bar) 30
l..''
E zo
'
'
"
I
,
. . . .
i
I
'
"
"L.i
I
"~
'"
'"'
'
"
-,!
:!
.................... [............................ q """ - qm :
F - t.... ...................... - .............................. . ~ - ' , i i I . . . . I. . . . . . . . . .
'
. . :, . & . . . . . . . . . . . . . . . . . . . . . . . . . . ~ ............................... ! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
-10
9
, i
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii
0
, 0
$
10
15
20
P r e s s u r e (bar) Fig. 12.20.Permeability versus transmembrane pressure for two zirconia nanofilters with different microporous volumes: (a) 8x10 -2 cm3/g, (b) 6.6-10-2 cm3/g [66,67].
membrane. On the contrary both sulfates and chlorides are not rejected w h e n they are counterions of the membrane. Charge density at the pore wall is a key parameter for the description of electrolyte rejection by charged nanofiltration membranes. In ceramic nanofilters charge density can be related to the specific surface area measured on ceramic m e m b r a n e material. In the case of a negative charged m e m b r a n e material (pH > zpc), Figs. 12.20 and 12.21 show respectively the influence of transmembrane pressure on m e m b r a n e permeability and the sulfate rejection versus flux for two membranes exhibiting different specific surface area. Calculated hydraulic radius was almost the same for the two membranes (rh ----0.43 nm) so that permeability can be discussed in term of the Donnan effect and related streaming potential for electrically charged porous membranes assuming that m e m b r a n e thickness is the same in both cases. The effect of electrokinetic flow on m e m b r a n e permeability is shown in Fig. 12.20. When transmembrane pressure increases permeability increases at first and then decreases due to the opposite contribution of electrokinematic flow to convective flow. This can be explained by the occurrence of a non-negligible streaming potential for a transmembrane pressure higher than 3 bar. Moreover the electrokinematic flow effect was more marked for the m e m b r a n e with the higher surface area which is consistent with a higher charge density. If we consider now sulfate rejection by the two membranes, Fig. 12.21, a better rejection was obtained with the m e m b r a n e exhibiting the higher surface area and consistently the higher
604
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Rejection (%) 100
.....o ........ .....................i.....................i............... 6O
40
2O
0
0
$0
100
150
200
250
300
Permeate flux (l/h.m z) Fig. 12.21. Sulfate rejection versus volume flux through two zirconia nanofilters with different microporous volumes: (a) 8-10-2 cm3/g, (b) 6.6.10-2cm3/g [66,67]. charge density. However the evolution of sulfate rejection with flux is not totally explained by Tsuru calculation [59] on sulfate rejection through negatively charged polymer membranes. With these zirconia nanofilters a m a x i m u m in rejection versus flux was evidenced while an increase of rejection followed by a plateau has been described with polymer nanofiltration membranes.
12.4.3 Separation of Aqueous Ionized Molecule-Salt Solutions The case which consists of a mixture of a mono-monovalent salt like NaC1 and a multifunctional organic anion A n- containing n negatively charged groups per molecule in a sodium salt form has been described by Perry and Linder [50]. It has been assumed that the monovalent anion C1- is permeable through the m e m b r a n e and the organic anion A ~- is fully rejected. Accordingly a new expression for salt rejection was proposed: TR = 1 - [(1 - r~)~/(1 - r~F)]
(12.34)
F is defined as in Eq. (12.12) and [3 as [3- (1
+
n CAb/Csb) 1/2
(12.35)
with CAB and Csb respectively the concentration of the organic anion and the concentration of monovalent anion in the feed solution. When only pure salt is p r e s e n t C A B - - 0, [~ = 1, Eq. (12.34) becomes identical to Eq. (12.11).
12 -- TRANSPORTAND FOULINGPHENOMENA IN LIQUIDPHASE SEPARATION
605
Schirg and Widmer [52] published mathematical models for the calculation of retention and selectivity for nanofiltration of aqueous dye-salt solutions. A modification of Eqs. (12.11) and (12.12) has been proposed in which the integral salt permeability r could be described by the introduction of an exponential function c0 = (x C~
(12.36)
with (x a constant and 7 a coefficient for salt permeability dependence on concentration. Both calculations by Perry and Schirg have been performed to describe and to predict the rejection characteristics of organic nanofiltration membranes when ionic and chargedmolecular solute mixtures are used in the feed solution. Recently experiments were carried out with ceramic nanofilters [67] which showed that similar properties can be obtained. As an example, results concerning the rejection of a dye/electrolyte mixture at pH = 9 through a zirconia nanofilter are reported in Table 12.5. -As a general conclusion to this part dedicated to nanofiltration with ceramic membranes one can assume that the general behaviour of these membranes can be assimilated to the behaviour of electrically charged organic nanofiltration membranes. However some specificities exist with ceramic nanofilters due to a sintered metal oxide grains derived porous structure and an amphoteric character TABLE 12.5 Rejection of a mixture of an organic anion (bromocresol green) and salt anions (SO4, C1-) through a negatively charged zirconia nanofilter [67] Anion
Concentration (ppm)
Mw
Pressure ( A p ) (bar)
Rejection (%)
200
698
10 20
63 71
NaC1/Na2SO 4 mixture C1-
2000
58
SO~
2000
142
10 20 10 20
3 6 39 40
10 10 10
70 0 48
Bromocresol green A-
Bromocresol green/NaC1/Na2SO4 mixture A200 698 C12000 58 SO4
2000
142
606
12 - - T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
in water media. At this time few experimental data are available in view of an assessment of existing or new mathematical models well adapted to ceramic nanofilters. Further experiments with different categories of ceramic membrane material are needed to establish general principles of nanofiltration with ceramic membranes.
12.5 PROSPECTIVE ASPECTS
12.5.1 Organic-Inorganic Hybrid Membranes and Related Processes At the present time, organic-inorganic hybrid membranes do not exist at the commercial stage. However, recent results have shown the interest of these membranes in a non-limited list of applications such as gas separation, pervaporation, chemical and biological sensors, facilitated transport, ultra- and nanofiltration. The main interest of organic-inorganic membranes is that they can combine basic properties of both organic and inorganic membrane materials. Accordingly improved properties are expected from this new category of membrane. A short overview of recent works dedicated to these membranes is given hereafter which illustrates their potentiality in liquid phase separation. A first way to obtain an organic-inorganic hybrid membrane is to have a polymer material either deposited or grafted at the surface or embedded in the top-layer porosity of a ceramic support. For example Castro et al. [68] investigated the permeability behaviour of polyvinylpyrrolidone-modified porous silica membranes. The surface of 0.4 ~tm-pore-size silica membranes was modified with a covalently bonded polyvinylpyrrolidone brush layer. Hydraulic permeability measurements performed with six different solvents and both unmodified and modified membranes suggest that the permeability of the modified membrane is determined by the configuration of the terminally anchored polymer chains. In the modified ceramic-supported polymer membrane, the swelling of the polymer brush layer increased as the solvent power increased, resulting in a decrease in the pore radius and subsequently the permeability. In a previously mentioned study Bardot et al. [62] used nanofiltration membranes made by internal coating of porous tubular supports of R-alumina with sulfonated collodion followed by coagulation in an appropriate bath. More recently Sarrade et al. [69] have developed a hybrid nanofiltration membrane highly effective for separating non-charged solutes of molecular weight as low as 500-1000 Dalton in supercritical carbon dioxide medium. This is a combined organic-inorganic membrane that comprises a macroporous ~-alumina substrate (tubular or multichannel), an intermediate mesoporous inorganic titanium oxide layer (thickness: 1 ~tm) and a microporous Nation polymer top-layer (thickness: less than 0.1 ~tm). The overall performance and
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
607
TABLE 12.6 Transport parameter value of a Nation/titania hybrid membrane [69,70]
(~ co(m.s-1)
EG
PEG 200
PEG 400
PEG 600
PEG 1500
6.010-2 4.410-5
4.410-1 2.1 10-6
5.710-] 8.810-7
8.1 10-1 1.710-7
9.610-1 3.1 10-8
t r a n s p o r t m e c h a n i s m t h r o u g h this m e m b r a n e have been studied using ethylene glycol (EG) and various polyethylene glycols (PEG) as m o d e l solutes [70]. Starting from Eqs. (12.9) to (12.12), the m e m b r a n e permeabilities to water, Lp, and solutes, co, as well as the reflection coefficients, r~, were d e t e r m i n e d at first. These values are reported in Table 12.6 and in Fig. 12.22. Using the theory p r o p o s e d by Verniory et al. [6] to account for h i n d e r e d transport in pores, the m e a n pore radius was estimated from these parameters. It is w o r t h noting that the m e a n value of 0.6 n m calculated for the m e m b r a n e is consistent w i t h the pore d i m e n s i o n (0.8 nm) directly m e a s u r e d using the biliquid p e r m p o r o m e t r y [71]. It has been s h o w n that, regardless of the size of the solute molecule, convective transport is always more i m p o r t a n t than diffusive transport. In 1,0
0,8 x
Membrane TN
0,6
0,4 m Membrane A 0,2
0,0
0,0
0,2
0,4
rs
0,6
0,8
1,0
(nm)
Fig. 12.22. Variation of reflection coefficient (~versus equivalent radius of model solutes rs for an alumina nanofilter (A) [65] and a Nation/titania hybrid membrane (TN) [70].
608
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
accordance with nanofiltration behaviour this membrane also exhibited effective separation of ionic species. Organic-inorganic polymer at the molecular level are also of interest as shown in the following examples. A new concept in nanofiltration has been proposed by Guizard et al. [72] based on a hybrid polymer (cyclic polyphosphazene) supported on a zirconia ultrafiltration membrane. Excellent chemical and temperature resistances were obtained for these membranes due to intrinsic properties of polyphosphazenes as well as a high rejection of small organics and a good selectivity concerning multivalent versus monovalent ions. The reflection coefficient r~ was markedly related to the transmembrane pressure leading to adjustable working conditions. Another example is in an alternative way to selective transport of metal ions through liquid membranes, such as transport mediated by crown-ether and other macrocyclic ligands which has been extensively investigated during the last twenty years. No practical separation processes arose during this period mainly because liquid membranes suffer poor stability and thus short lifetimes: the membrane degradation is essentially due to the loss of carrier by dissolution in the aqueous phase and by emulsion formation at the membrane interfaces. Consequently the recent developments in facilitated transport membrane processes are focused on new membrane systems with improved lifetimes. One of these systems is based on the carrier grafting onto a solid membrane matrix. Grafting of benzo-15-crown-5 in a heteropolysiloxane membrane was investigated by Lacan et al. [73] in view of facilitated transport of alkaline ions. Very stable membranes over several months were obtained without loss of carrier during transport experiments. It has been demonstrated that covalently bound carriers allow facilitated transport of K § ions versus Li + ions to take place with high diffusion rates, high facilitation factor and good selectivity. These membranes open a new way in the application of facilitated transport to practical separation processes.
12.5.2 Coupled Membrane Processes Inorganic membranes, and to a less extent hybrid membranes, possess a high degree of resistance to chemical and abrasion degradation as well as tolerate a wide range of pH and temperature values. All these properties make them very useful for coupling with other processes and open up new fields of applications. In what follows, some examples of such integrated processes involving at least one membrane separation stage are presented.
Membrane bioreactor The idea of coupling membrane separation with bioreaction is not a new one. A lot of works published in the literature bear witness to this fact. But most of
609
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
the time organic membranes and aqueous phase reactions are considered [74]. For areas of biotechnology or food engineering, a main advantage of inorganic materials is that they may be repeatedly autoclavable and are very stable against microbiological attack [32]. With them, reactions in pure organic solvent may be also successfully faced. As an example, the enantiomeric resolution of menthol (+) into methyl (-) laurate, through a biological catalysis method involving a lipase from Candida rugosa and lauric acid as substrate, was recently investigated using n-heptane as the solvent medium [75]. A zirconia membrane with a pore diameter of about 4 nm was chosen to'retain the biocatalyst. This lets the substrate and product molecules pass. The transmembrane pressure was selected so as to get a space time leading to an optimum reaction yield. At the reactor outside, menthol (+) and methyl (-) laurate were separated from permeate, and then menthol (-) was regenerated from ester.
Nanofiltration plus supercritical fluid extraction Supercritical fluid extraction is used to recover small organic solutes with molecular weight below 1500 daltons. In a state of continuity between vapour and liquid, supercritical fluids exhibit intermediate transport properties with lower viscosities than liquid and higher diffusivities than gases. Because of its
(~
A I CO 2
I I I I
I I
I
P > 74 bars T > 31 "C
Valve
P < 60 bars
exc6ange r I ,l
' ~ ' " ~ - " " " ~
I I
!
_II
I
I
I
T Extract
.4 -
/
Mixt
extracts
Fig. 12.23. Nanofiltration/supercritical fluid extraction coupled processes.
"
610
12 m TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
attractive critical temperature/pressure conditions, stability, low cost and nontoxicity, CO2 is today the most widely used SCF. As previously indicated, nanofiltration now provides new ceramic and hybrid membranes with cut-off in the range of 300 to 1000 daltons. On principle, nanofiltration plus supercritical extraction aims to both enhance the selectivity of extraction and lower energy consumption. A schematic view of the process is given in Fig. 12.23. Regarding selectivity, it may be thought that membrane sieving effect will induce a separation of supercritical fluid mixture into fractions respectively containing high (retentate) and low (permeate) molecular weight solutes. From an economic viewpoint, a substantial energy saving may be expected due to the fact that only permeate flow that just represents a small part of total CO2 will be submitted to a strong pressure reduction from extract recovery, while low soluble heavy compounds will be continuously deposited from retentate by means of a small pressure and/or temperature effect. Experiments have proved that silica [76], or titane-nafion membranes [77] were able to endure supercritical fluid conditions with no alteration. With y-alumina membrane, fouling strongly develops probably due to chemisorption. Working with model molecules such as ethylene glycol and polyethylene glycols (PEG 200-400-600), the process capability to extract and separate various size solutes has also been checked [78].
Ultrafiltration plus electrophoresis It is worth recalling that the flux and the selectivity of ultrafiltration may be improved when treating electrically charged solutes ~ as an example, alkaline gelatin molecules (pI = 4.7) processed at pH = 6.0 present a negative charge by superposing upon the driven pressure an electric field which acts on the retained solute to control concentration polarization. This is the so-called "electro-ultrafiltration process". In the past various works have underlined the influence on performance of such parameters as pressure, fluid velocity, electric field strength or starting conditions particularly with ceramic membranes [79]. With membranes cylindrically shaped, and for instance when processing a negatively charged solutes, a classical setting diagram consists both in installing a stainless steel wire as anode through the centre of the membrane and in closely surrounding the outside of the supporting tube by a cathode made of a large mesh stainless wire lattice (Fig. 12.24). Because the supporting tube is placed along with the membrane itself in the electric field acting area, disadvantages may result from the use of this traditional set-up: excessive energy consumption, parasite and uncontrolled effects (such as electro-osmosis fluxes). A new concept has been recently proposed with inorganic membranes to overcome some of these difficulties. It consists in designing electronic conductive membranes in which the original feature is the possibility of using the
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Membrane support
/
611
Central electrode
,/
~
ctive membrane External electrode for non-conductive membrane
Retentate
t' b Electric,potential
I
Fig. 12.24. Schematic view of electro-ultrafiltration using conductive inorganic membranes.
active layer both as a filter and as an electrode. RuO2-TiO2 membranes coated on an alumina support belong to this class. With them, performance may be enhanced by applying the electric field only inside the filtration module, specially when the membrane is used as the anode, a classical way to work RuO2-TiO 2 electrodes [80,81].
Cross-flow filtration with mobile turbulence promoters It is well known that pumping of the fluid has a major effect on flux in the mass transfer controlled region for UF/MF process. Indeed agitation and mixing of the fluid near the membrane surface sweep away the accumulated solutes, thus reducing the thickness of boundary layers. This is the simplest and most effective method of controlling the effect of concentration polarization.
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
612 J (I/h.m2)
Fluidized bed (Stainless steel beads 3mm in diameter and 7.9 in density)
50-
40
30 20
E m p t y robe a
i --
10
s = 0,42
i
,_..._._....p..--I
/k
2O
E = 0,6
,
I
~
30
~ = 0,7
~
,'k
I 4O
,
I 50
u(~)
s = 0,8
Minimum fluidization Fig. 12.25. Permeate flux versus tangential fluid velocity with a gelatine solution (10 g 1-1, Ap = 1.5 bar) ushlg a tubular a l u m i n a m e m b r a n e filled with a fluidized bed.
The magnitude of the effect of flow rate on the mass transfer coefficient will depend on whether the flow is turbulent or laminar, as well as on rheological properties of the fluid, the key factor being the shear stress at wall. Another less c o m m o n method to effect permeate flux increases is through the introduction of turbulence promoters in the flow conduit. Up to now more attention has been given to fixed promoters due to damage that ordinarily results from the movements of free agents at the very fragile surface of traditional organic membranes [82]. On the contrary, ceramic membranes (alumina as an example) are resistant enough to endure the continuous bombardment of fltfidized particles [83] or even the friction of transported solids [84]. With such devices, high permeate fluxes may be obtained (Fig. 12.25) with no sharp decrease in solute rejection (Fig. 12.26), even at tangential fluid velocity as low as a few ten centimetres per second. The analysis of mass transfer coefficients and hydraulic resistances showed that moving particles insure a significant reduction of the mass transfer boundary layer, as well as a continuous mechanical erosion of the deposit at wall. Polarization is strongly modified. From a practical viewpoint, low retentate velocities may offer some interesting developments in those cases where fragile molecules are to be treated, or long enough residence times are needed. Solid particles could also be used as catalyst (enzymatic supports as an example) for heterogeneous reactions, adsorbent for coupled MF/adsorption processes. As shown in Figs. 12.25 and 12.26 even the existence of optimum working conditions for fluidized bed devices at an intermediate bed porosity could be turned to advantage to elaborate new permeate flux control strategies.
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUIDPHASE SEPARATION
613
Ti
100
Fl/dJze, d bed
80
70
Empty tube
0
,
I
10
,
t
20
,,,
t
30
,
t
40
,
,
,
J U(cm/s)
50
Fig. 12.26. Gelatin rejection versus tangential fluid velocity usin~ a tubular alumina m e m b r a n e filled with a fluidized bed (feed solution 10 g 1-, Ap = 1.5 bar).
12.6 CONCLUSION Different aspects of liquid phase separation using inorganic membranes should be emphasized compared with organic membrane behaviour. The first characteristic of inorganic membranes designed for liquid filtration is that they exhibit a non-deformable porous structure with pore size adapted to three main processes: macropores for microfiltration, mesopores for ultrafiltration and micropores for nanofiltration. Modelling of mass transfer across these membranes is related to basic phenomena involved in liquid flow through porous media. The Darcy law applies to convective volume flux: through macro- and mesoporous membranes while a convection-diffusion mechanism better explains solvent flux in the case of microporous membranes. Due to pore shape resulting from packing and sintering of mineral particles the Carman-Kozeny model which includes specific surface area and tortuosity provides a better description of the permeability coefficient than the Hagen-Poiseuille law. The second characteristic of inorganic membranes used in liquid phase separation is that most are made of ceramic oxides. If solute rejection basically originates in size effects related to pore dimension, specific properties are attached to ceramic membrane material. The amphoteric behaviour of metal oxide surfaces is certainly the most important one as membranes can exhibit negative or positive charge density depending on the pH of feeding solutions. Two parameters, zeta-potential and streaming potential, greatly influence rejection and permeability of electrolyte solutions all the more as membranes exhibit small pore size and large specific surface area.
614
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
Fouling, responsible for flux decline, is also an i m p o r t a n t p a r a m e t e r to deal w i t h in the description of transport m e c h a n i s m s w i t h inorganic m e m b r a n e s . Three m a i n causes have been identified as i m p o r t a n t contributions to fouling of inorganic m e m b r a n e s . It has been suggested that the formation of ceramic m e m b r a n e s can be a first cause of flux decline as far as association of adjacent g r a n u l a r layers results in highly resistant b o u n d a r y layers. A second p h e n o m e n o n responsible for flux decline is the on-line m e m b r a n e fouling w h i c h is a function of the h y d r o d y n a m i c conditions and is i n d e p e n d e n t of the physical properties of the m e m b r a n e . Finally interaction b e t w e e n m e m b r a n e material a n d molecules or macromolecules can result in the formation of a d y n a m i c layer on the original filtering element. This layer can be r e g a r d e d as a formed-inplace m e m b r a n e w i t h specific separation properties and it is responsible for an additional resistance to the v o l u m e flux.
REFERENCES
1. R. Bhave, Inorganic Membranes Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991. 2. M. Mulder, Basic Principles of Membrane Technology. Kluwer, Dordrecht, Boston, London, 1991. 3. S. Nakao, T. Anazawa, T. Tsuru and S. Kimura, Influence of high permeation on concentration polarization in ultrafiltration. Presented at ISMMP'94, 5-10 April 94, Hangzhou, P.R. China. 4. J.A. Wesselingh and R. Krishna, Mass Transfer. Ellis Horwood, New York and London, 1990. 5. O. Kedem and A. Katchalsky, Thermodynamic analysis of the permeability of biological membranes to non-electrolytes. Biochim. Biophys. Acta, 27 (1958) 229. 6. A. Vemiory, R. Dubois, P. Decoodi and J.P. Gassee, Measurement of the permeability of biological membranes. J. Gen. Physiol., 62 (1973) 489. 7. P.M. Bungay, Transport principles m Porous membranes, in: P.M. Bungay, H.K. Lonsdale, M.N. de Pinho (Eds), Synthetic Membranes: Science, Engineering and Applications. NATO ASI Series, Series C: Mathematical and Physical Sciences, Vol. 181, 1986. p. 109. 8. T. Tsuru, M. Urairi, S. Nakao and S. Kimura, Negative rejection of anions in the loose reverse osmosis separation of mono- and divalent ion mixtures. Desalination, 81 (1991) 219. 9. W.M.Deen, Hindered transport of large molecules in liquid -filled pores. AIChE J., 33 (9) (1987) 1409 10. C. Guizard, A. Julbe, A. Larbot and L. Cot, Ceramic membrane processing, in: B.I. Lee and E.A.J. Pope (Eds.), Chemical Processing of Ceramics. Marcel Dekker, New York, 1994, p. 501. 11. J. Randon, A. Julbe, P. David, K. Jaafari and S. Elmaleh, Computer simulation of inorganic membrane morphology: 2. Effect of infiltration at the membrane support interface. J. Colloid Interface Sci., 161 (1993) 384.
12-- TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
615
12. P. Aimar, C. Taddei, J.P. Lafaille and V. Sanchez, Mass transfer limitations during ultrafiltration of cheese whey with inorganic membranes. J. Membr. Sci., 38 (1988) 203 13. A.D. Marshall, P.A. Munro and G. Tr~igardh, The effect of protein fouling in microfiltration and ultrafiltration on permeate flux, protein retention and selectivity: A literature review. Desalination, 91 (1993) 65. 14. C. Vetier, M. Bennasar and B. Tarodo de la Fuente, Study of the fouling of a mineral microfiltration membrane using scanning electron microscopy and physicochemical analysis in the processing of milk. ]. Dairy Res., 55 (1988) 381. 15. P.Freund and G.M. Rios, M6canismes de colmatage et performances compar6es de quatre membranes d'alumine: cas de l'ultrafiltration de g61atine. Can. J. Chem. Eng., 70 (2) (1992) 250. 16. Y. Matsumoto, S. Nakao and S. Kimura, Cross-flow filtration of solutions of polymers using ceramic microfiltration. Int. Chem. Eng., 28 (1988) 677. 17. G. Grund, C.W. Robinson and B.R. Glick, Protein type effects on steady-state crossflow membrane ultrafiltration fluxes and protein transmission. J. Membr. Sci., 70 (1992) 177. 18. J.L. Negrel, G.M. Rios and J.L. Cuq, Design and performance of new formed-in-place proteinic membranes for nanofiltration. Presented at ISMMP'94, 5-10 April 1994, Hang-
zhou, P.R. China. 19. C. Guizard, A. Larbot and L. Cot, A new generation of membranes based on organicinorganic polymers. Proceedings of the First International Conference on Inorganic Membranes, Montpellier, France, 1989, p. 55. 20. A. Julbe, C. Guizard, A. Larbot, L. Cot and A. Giroir-Fendler, The sol-gel approach to prepare candidate microporous inorganic membranes for membrane reactors. J. Membr. Sci., 77 (1994) 137. 21. R. Mellis, W.N. Gill and G. Belford, Fluid dynamics in a tubular membrane: theory and experiment. Chem. Eng. Commun., 122 (1993) 103. 22. S.M. Ahmed, Studies of he double layer at oxide solution interface. ]. Phys. Chem., 73 (1969) 3546. 23. J. Lyklema, The structure of the electrical double layer on porous surfaces. J. Electroanal. Chem., 18 (1968) 341. 24. J.W. Perram, R.J. Hunter and H.J.L. Wright, The oxide-solution interface. Aust. J. Chem., 27 (1974) 461. 25. S. Levine, J.R. Marrott, G. Neale and N. Epstein, Theory of electrokinetic flow in fine cylindrycal capillaries at high zeta-potentials. J. Colloid Interface Sci., 52 (1975) 136. 26. F.A. Morrison, JR and J.F. Osterle, Electrokinetic energy conversion in ultrafine capillaries. J. Chem. Phys., 43 (6) (1965) 2111. 27. R.J. Gross and J.F. Osterle, Membrane transport characteritics of ultrafine capillaries. J. Chem. Phys., 49 (1) (1968) 228. 28. J.C. Fair and J.F. Osterle, Reverse electrodialysis in charged capillary membranes. J. Chem. Phys., 54 (8) (1971) 3307. 29. H.K. Lonsdale and W. Pusch, Donnan-membrane effects in hyperfiltration of ternary systems. J. Chem. Soc., Faraday Trans. 1, 71 (3) (1975) 501. 30. A.W. Adamson, Physical Chemistry of Surfaces. Wiley, New York, 1988, 272 pp. 31. J. Israelachvili, Intermolecular and Surface Forces. Academic Press, London, 1985, 180 pp. 32. S. Saksena and A.L. Zydney, Pore size distribution effects on electrokinetic phenomena in semipermeable membranes. ]. Membr. Sci., 105 (1995) 203. 33. G.M. Rios, B. Tarodo de la Fuente, M. Bennasar and C. Guizard, Cross-flow filtration of biological fluids on inorganic membranes: a first state of the art, in: S.Thorne (Ed.),
616
34. 35. 36. 37.
38. 39.
40. 41.
42.
43. 44. 45.
46. 47. 48.
12 n TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
Developments in Food Preservation 5. Elsevier, London, New York, 1989, p. 131. G. Belfort, R.H. Davis and A.L. Zydney, The behavior of suspensions and macromolecular solutions in crossflow microfiltration. J. Membr. Sci., 96 (1994) 1. H.B. Winzeler and G. Belfort, Enhanced performance for pressure-driven membrane processes: the argument for fluid instabilities. J. Membr. Sci., 80 (1993) 35. B.B. Gupta, P. Blanpain and M.Y. Jaffrin, Permeate flux enhancement by pressure and flow pulsations in microfiltration with mineral membranes. J. Membr. Sci., 70 (1992) 257. M.Y. Jaffrin, L.H. Ding and J.M. Laurent, Kinetics of concentration polarization formation in crossflow filtration of plasma from blood: experimental results. J. Membr. Sci., 72 (1992) 267. Y. Wang, J.A. Howell, R.W. Field and D. Wu, Simulation of crossflow filtration for baffled tubular channels and pulsatile flow. J. Membr. Sci., 95 (1994) 243. J. Charpin, P. Bergez, F. Valin, H. Bamier, A. Maurel and J.M. Martinet, Inorganic membranes: preparation, characterization, specific applications. Mater. Sci. Monogr., 38C (1987) 2211. J. Murkes and C.-G. Carlsson, Crossflow Filtration, Theory and Practice. Wiley, New York, NY, 1988. M.-C. Aubert, M.-P. Elluard and H. Barnier, Shear stress induced erosion of filrtration cake studied by a flat rotating disk method. Determination of the critical shear stress of erosion. J. Membr. Sci., 84 (1993) 229. E. Vernon Ballou, T. Wydeven and M.I. Leban, Solute rejection by porous glass membranes. I. Hyperfiltration of sodium chloride and urea feed solutions. Environ. Sci. Technol., 5 (1971) 1032. E. Vernon Ballou and T. Wydeven, Solute rejection by porous glass membranes. II. Pore size distributions and membrane permeabilities. J. Colloid Interface Sci., 41 (1972) 198. K. Singh and A. Singh, Membrane potential and solute separation studies on zirconium phosphate membranes. ]. Membr. Sci., 82 (1993) 141. M.R. Hoogland, A.G. Fane and C.J.D. Fell, The effect of pH on the cross-flow filtration of mineral slurries using ceramic membranes. Proceedings of the First International Conference on Inorganic Membranes, Montpellier, France, 1989, p. 153. F.F. Nazzal and M. Wiesner, pH and ionic strength on the performance of ceramic membranes in water filtration. J. Membr. Sci., 93 (1994) 91. J.J. Porter and R.S. Porter, Filtration studies of selected anionic dyes using asymmetric titanium dioxide membranes on porous stainless tubes. J. Membr. Sci., 101 (1995) 67. J.J. Porter and S. Zhuang, Microfiltration of sodium nitrate and direct red 2 dye using asymmetric titanium dioxide membranes on porous ceramic tubes, to be published in J. Membr. Sci.
P. Eriksson, Nanofiltration extends the range of membrane filtration. Environ. Progr., 7 (1) (1988) 58. 50. M. Perry and C. Linder, Intermediate reverse osmosis ultrafiltration membranes for concentration and desalting of low molecular weight organic solutes. Desalination, 71 (1989) 233. 51. R. Rautenbach and A. Gr6schl, Separation potential of nanofiltration membranes. Desalination, 77 (1990) 73. 52. P. Schirg and F. Widmer, Characterisation of nanofiltration membranes for the separation of aqueous dye-salt solutions. Desalination, 89 (1992) 89. 53. R.W. Baker, F.R. Eirich and H. Strathmann, Low pressure ultrafiltration of sucrose and raffinose. I. Phys. Chem., 76 (2) (1972) 238. 49.
12 u TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
617
54. A.Y. Tremblay, Finely porous models and radially averaged friction factor, J. Appl. Polym. Sci., 45 (1) (1992) 159. 55. U. Merten, in: U. Marten (Ed.), Desalination by Reverse Osmosis. M.I.T. Press, Cambridge, MA, 1966, p. 15. 56. H. Medizadeh and J.M. Dickson, ]. Appl. Polym. Sci., 42 (1991) 1143. 57. K.S. Spiegler and O. Kedem, Thermodynamics of hyperfiltration (reverse osmosis): criteria for efficient membranes. Desalination, 1 (1966) 311. 58. T. Tsuru, S.-I. Nakao and S. Kimura, Effective charge density and pore structure of charged ultrafiltration membranes. J. Chem. Eng. Jpn., 23 (5) (1990) 64. 59. T. Tsuru, S.-I. Nakao and S. Kimura, Calculation of ion rejection by extended NernstPlanck equation with charged reverse osmosis membranes for single and mixed electrolyte solutions. J. Chem. Eng. Jpn., 24 (4) (1991) 511. 60. T. Tsuru, M. Urairi, S.-I. Nakao and S. Kimura, Reverse osmosis of single and mixed electrolytes with charged membranes: experiments and analysis. J. Chem. Eng. Jpn., 24 (4) (1991) 518. 61. X.L.Wang, T. Tsuru, S.-I. Nakao and S. Kimura, Electrolyte transport through nanofiltration membranes by the Space Charge Model and the comparison with TeorellMeyer-Sievers model. J. Membr. Sci., 103 (1995)117. 62. C. Bardot, E. Gaubert and A.E. Yaroshchuck, Unusual mutual influence of electrolytes during pressure driven transport of their mixtures across charged porous membranes. J. Membr. Sci., 103 (1995) 11. 63. G. Jacazio, R.F. Probstein, A.A. Sonin and D. Yung, Electrokinetic salt rejection in hyperfiltration through porous materials. Theory and experiment. J. Phys. Chem., 76 (26) (1972) 4015. 64. S. Alami-Younssi, A. Larbot, M. Persin, J. Sarrazin and L. Cot, Rejection of mineral salts on a gamma alumina nanofiltration membrane. Application to environmental process. J. Membr. Sci., 102 (1995)123. 65. G.M. Rios, R. Jouli6, S. Sarrade and M. Carles, Investigation of ion separation by microporous nanofiltration membranes, AIChE J., to be published. 66. C. Guizard, C. Mouchet, R. Vacassy, A. Julbe and A. Ayral, Zirconia nanofiltration membranes: I. Mechanism of pore formation and static characterization., in preparation. 67. C. Guizard, C. Mouchet, R. Vacassy, X. Bouisson and V. Thoraval, Zirconia nanofiltration membranes: II. Performance of the membranes, dynamic characterization with model solutes, in preparation. 68. R.P. Castro, Y. Cohen and H.G. Monbouquette, The permeability behavior of polyvinylpyrrolidone-modified porous silica membranes. J. Membr. Sci., 84 (1993) 151. 69. S. Sarrade, C. Bardot, M. Carles, R. Soria, S. Cominotti and R. Gillot, Elaboration of new multilayer membrane for nanofiltration. Proceedings of the 6th World Filtration Congress,
18-21 May 1993, Nagoya, Japan. 70. S.Sarrade, G.M. Rios and M.Carles, Dynamic characterisation and transport mechanisms of two inorganic membranes for nanofiltration. J. Membr. Sci., 114 (1996) 81. 71. M.G. Liu, R.Ben Aim and M. Mietton-Peuchot, Characterization of inorganic membranes by permporometry method: importance of non-equilibrium phenomena. Key Eng. Mater., 61/62 (1991) 603. 72. C. Guizard, A. Boy6, A. Larbot and L. Cot, A new concept in nanofiltration based on a composite organic-inorganic membrane. Rec. Progr. Gdn. Procddds, 6 (22) (1992) 27. 73. P. Lacan, C. Guizard, P. Le Gall, D. Wettling and L. Cot, Facilitated transport of ions
618
12-- TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION through fixed-site carrier membranes derived from hybrid organic-inorganic materials.
J. Membr. Sci., 100 (1995) 99. 74.
M.Cheryan and M.A. Mehaia, Membrane bioreactors, in: W. Courtney McGregor (Ed.),
Membrane Separations in Biotechnology. Marcel Dekker Inc., New York, Basel, Vol. 10, 1986, p. 255. G.M. Rios, F. Lambert and J.C. Jallageas, Essais de mise en oeuvre d'un r6acteur d'ultrafiltration pour la catalyse biologique en milieu solvant: cas de l'est6rification enantios61ective par lipase du menthol. Entropie, 63 (1991) 31. 76. K. Nakamura, T. Hoshino, A. Morita, M. Hattori and R. Okamoto, Membrane separation of supercritical fluid mixture, in: T. Yano, R. Matsuno and K. Nakamura (Eds.), Developments in Food Engineering, 2. Blackie, London, New York, 1994, 820 pp. 77. S. Sarrade, C. Perre, M. Carles, R. Veyre and G.M. Rios, Nanofiltration coupled with supercritical carbon dioxide. Interest and preliminary studies. ICOM'93, 30 August-3 75.
Sept. 1993, Heildelberg, Germany. 78.
79. 80. 81.
82.
83. 84.
S. Sarrade, G.M. Rios, C. Perre and M. Carles, Performance of nanofiltration under supercritical fluid conditions, in: Y.H. Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, MA, p. 129. G.M. Rios, H. Rakotoarisoa and B. Tarodo de la Fuente, Basic transport mechanisms of ultrafiltration in the presence of an electric field. J. Membr. Sci., 38 (1988) 147. C. Guizard, N. Idrissi, A. Larbot and L. Cot, An electronic conductive membrane from sol-gel process. Br. Ceram. Proc., 38 (1986) 263. C. Guizard, F. Legault, N. Idrissi, A. Larbot, L. Cot and C. Gavach, Electronically conductive mineral membranes designed for electro-ultrafiltration. J. Membr. Sci., 41 (1988) 147. M.J. Van Der Waal, P.M. Van Der Velden, J. Koning, C.A. Smolders and W.P.M. Van Swaay, Use of fluidised beds as turbulence promotors in tubular membrane systems. Desalination, 22 (1977) 465. G.M. Rios, H. Rakotoarisoa and B. Tarodo de la Fuente, Basic transport mechanisms of ultrafiltration in the presence of fluidized particles. J. Membr. Sci., 34 (1987) 331. F. Clavaguera, E. Rjimati, S. Elmaleh and A. Grasmick, Intensification of microfiltration by a circulating bed, in Proceedings of ICIM 2. Key Eng. Mater., 61/62 (1991) 569.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 13
Applications of ceramic membranes in liquid filtration C.A.M. Siskens (Formerly: Hoogovens Industrial Ceramics BV) Ministry of Transport, Public Works and Water Management, Road and Hydraulic Engineering Division, Delft, The Netherlands
13.1 INTRODUCTION
The number of applications of ceramic membranes is immense and ever increasing. Many references on the use of ceramic membranes can be found in the proceedings of the two International Conferences on Inorganic Membranes [1,2] as well as in the excellent book of Bhave [3]; others [4-10] highlight developments since 1988. Except for a single reference, e.g. [11,12], direct data on the extent of installed ceramic membranes is rare. BCC's 1994 study "Inorganic membranes: markets, technologies, players" [13] estimates the inorganic membranes to grow to about 15% of the total separation membrane/module sales. This means that, worldwide in 2003, the sales in inorganic membranes are estimated at US$ 228 million, of which 69% is in ceramic membranes. These figures constitute an adjustment to earlier expectations [14], stating US$ 363 million in 2000. Clearly, great care should be exercised in using these values as data on market volumes are rather incongruous. Furthermore, detailed knowledge about commercial applications seems to be restricted: in the description of tests many publications deal with the potential of inorganic membranes rather than with ongoing industrial applications. Moreover, many market oriented publications are 'lost' in journals which are not abstracted in major data-bases. Both factors diminish the insight in the real scope of ceramic membranes. The limited nature of industrial application of ceramic membranes can be inferred too from the rather short description in some new textbooks like those of Gasper [15], Ripperger [16] and Ho and Sirkar [17].
620
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
This reticence contrasts with the often quoted, many advantages of ceramic membranes: chemical and thermal stability, narrow pore size distribution, high porosity, high flux, mechanical strength (enabling back flushing), micro biological resistance, long lifetime, etc. [3-11,14]. In practice it turns out that the points of chemical and thermal stability are successful, even under seemingly moderate circumstances, as they permit the ceramic membrane to be cleaned much more thoroughly (and harsher!) than polymer membranes. This constitutes an extra advantage because it substantially adds to the economy of use: higher average flux, lower cleaning frequency, longer lifetime of the membranes. Despite all these advantages a real breakthrough has not been accomplished, and this can hardly be attributed to the disadvantages of higher price of production and brittleness. More probably for this still rather new product, the success of application depends on other marketing factors such as elemental applications research, engineering development and guidance of the customer. The application of a technology is mainly governed by its costs versus its benefits. The economical place of any separation process then depends on the type and value of the materials to be treated. Based on this line of thought, the applications in this chapter are classified into three groups, viz.: treatment of wastes: cleaning of waste streams to enable their disposal, regeneration: enabling the recycling/reuse of material, processing: treatment of process streams. It appears that in certain cases the category 'wastes' coincides with the category 'regeneration', e.g. in cases where the permeate water is of sufficient quality to be reused, or where a retentate may serve as a raw material for another process. The material presented in this chapter is based mainly on open literature dealing with the use of commercially available ceramic membranes and on technical data as acquired in Hoogovens Industrial Ceramics BV (HIC), the author's former company. This review is not meant to be exhaustive, but reflects examples of the use that ceramic membranes have found in certain industrial applications. 13.2
TREATMENT
OF
WASTES
13.2.1 Wastes of Oily Emulsions
13.2.1.1 Compressor-condensate In oil lubricated compressors a condensate is formed [18]. Such condensate is an oil in water emulsion with typical oil contents of 0.5%. In various countries discharge of more than 10 ppm oil is prohibited, e.g. in Austria since the beginning of 1992. In Germany, for new investments for waste water cleaning, the use of chemicals to break emulsions is forbidden. Furthermore, incineration of oily waste streams requires the highest oil content possible.
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
621
With ceramic membranes (typical pore size 0.2 ~tm) this legal limit of 10 p p m is easily obtained [18]. Average membrane flux is 150 1/m 2 h; values are reported between 125 and 6001/m 2 h [19]. The interval between cleanings or the maintenance interval can amount to 1000 h. By combination of the membrane unit with a static separator the concentration factor can be as high as 180, the oil content in the final concentrate can amount to over 90%. The pay-back time for such an installation would be less than two years at a cost of NLG 9 4 / m 3 effluent, and related to the present costs of treatment [18].
13.2.1.2 Centralised Treatment of Industrial Emulsions In many industrialised countries oily wastes are collected and treated in commercial or public emulsion treatment centres. The supply of oil emulsions varies very considerably in type of oils, concentration, contamination with other materials, etc. Following coarse pre-filtration and decantation, oil/water emulsions can be treated very successfully with ceramic membranes. The concentrate is returned to the decanter, and microfiltered again after removal of the free oil, until all oil is removed. The extracted water can be fed into a biological treatment plant, or discharged directly, depending on the composition of the original emulsion a n d / o r local regulations. In a typical example [20] 6.4 m 2 of 0.2 ~trn membranes are used in a pilot scale operation, yielding average fluxes between 100 and 1251/m 2 h in a installation working at 55~ The concentration factor for the membrane installation varies between 6 and 12. Due to the extreme fouling nature of the feed, periodic cleaning is compulsory, but can be restricted to once a week. The system has been in operation since August 1992. The pay-back time is less than 3 years. A flux of 2001/m 2 h with the Ceramesh metal/ceramic composite membrane (0.2 ~tm pore size) on a metal working emulsion is reported by Cowieson and Gallagher [21]; similar data are given for Carbosep membranes [22]. Y6ksel et al. [23,24] describe the use of organic demulsifiers (ternary and quaternary polyamines) to enhance the breaking of oily emulsions. This method is particularly suitable when the composition of the oily waste water is fairly constant but it entails extra costs and maintenance. Ceramic membranes perform much better than polymer membranes because the latter get blocked by the polyamines.
13.2.1.3 Bilge Water Treatment Bilge water is the waste stream of (salt) water, fuel, oil, fats, detergents and others as found in the engine room of ships. The oil content of such water can be as high as 50%, the further composition cannot be quantified. Discharge of bilge water is a serious pollution item. Separation systems based on differences in density are not able to reach the discharge limit of 15 ppm as set by the new
622
13 - - A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
regulations of the International Maritime Organisation, especially not under seagoing circumstances. Bilge water can be treated with HIC's ceramic membranes on shore or on board of the ship. In all cases a permeate with less than 15 ppm is reached with 0.2 ~tm membranes, even in the presence of detergents. Fluxes vary between 50 and 100 1/m 2 h. The treatment of bilge water and emulsions resembles that of the treatment of oil field brines and produced water. Chen et al. [25], using ferric chloride and other chemicals to enhance the performance of Membralox 0.2, 0.5 and 0.8 ~tm membranes, describe permeate fluxes between 1400 and 3400 1/m 2 h. Without pretreatment however severe fouling occurred as well as break-through of oil. Zaidi et al. [26] report about the continuation of this work. They quote fluxes between 800 and 12001/m 2 h , but also mention substantial lower fluxes in long term pilot tests using 0.8 ~tm membranes. In addition they indicate a drop in permeate flux caused by conditions of low pH, the presence of sea water, corrosion inhibitor, oil slugs or flow variations.
13.2.1.4 Vegetable Waste Water In the production of olive oil large amounts of the so-called alpechine (Spanish) or margine (French) are produced as waste (vegetable waste water). Depending on the extraction process 1.2-1.7 m 3 of waste water is produced per ton of olives. The treatment of this stream is becoming important as discharge into surface water or as an agricultural fertiliser is no longer acceptable. Alpechine is characterised by a low pH, a low content of nutrients and a high content of low-biodegradable organics, thwarting aerobic treatment or anaerobic-aerobic treatment. In comparison with these methods and in comparison with evaporation a treatment consisting of pre-filtering followed by microfiltration with ceramic membranes and a polishing step (ultrafiltration plus reverse osmosis) produces very good results. The microfiltration step of fresh alpechine on 0.2 ~tm HIC membranes attains fluxes from 90-125 1/m 2 h at temperatures between 30 and 50~ In the polishing step COD was reduced to approx. 1700 mg/1. This method constitutes an important economical advantage over other methods: e.g. Mendia [27], describing different methods of treatment, the use of evaporation [28,29] or biological treatment [30] and earlier experiments with polymer membranes [31] with a combination of UF, RO and adsorption. 13.2.2 W a s t e s B a s e d on S e m i - s o l i d s
13.2.2.1 Fish Factory Effluent In fish processing plants a large quantity of water is used as cleaning and transport medium. The water becomes polluted with fats, proteins, bones and
13 - - A P P L I C A T I O N S O F C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
623
blood. The aim of microfiltration with ceramic membranes is to minimalize the sludge production on the one hand and to produce a dischargeable water on the other hand [32]. Over biological treatments this has the advantage of winning back part of the valuable material instead of transforming it into a sludge which has to be disposed of in landfills. Filtrating with 0.2 ~tm HIC ceramic membranes a COD reduction of over 60% was reached, at a content of suspended solids in the permeate of less than 10 ppm. Permeate flux was at a level of 150 1/m 2 h at a process temperature of 25~ Comparable results were found by Quemeneur and Jaouen [32].
13.2.2.2 Manure The disposal of pig manure poses problems comparable to those of the vegetable waste water. Pre-filtering is even more important here because of the very coarse nature of some of the manure components. As manure from pigs has solids contents as low as 5 to 11%, the main goal is the reduction of the amount to be transported from areas with a manure-surplus to regions with a fertiliser shortage. Very important too, but in the economical sense, is the possibility for disposal of the concentrate of the separation process. Local factors like fertilising limits and the nearby availability of fields that can be fertilised are decisive. Test results: starting from pig manure with a solids content of ca. 11% a vibrating screen separates this into a feed stream for the microfiltration containing 6% solids. On this feed, HIC's ceramic 0.1 ~tm membranes reach average fluxes of 80-100 1/m 2 h at filtration temperatures of 80~ The concentration factor can range between 2.5 and 3. Operating costs are below the DEM 2 0 / m 3 quoted by Meindersma [24]. The combined concentrate of pre-filter and MF is about 55% of the original volume and contains approximately 20% solids; the clear permeate of the MF contains approximately 2% solids, typically dissolved substances.
13.3 R E G E N E R A T I O N
13.3.1 Recycling of Solids from Suspensions 13.3.1.1 Ceramics Industry A good example of the filtration of hard, abrasive materials is the application of ceramic membranes in the cleaning of waste water of the ceramic industry [33]. Waste water in this industry typically contains clay, sand, glazes, etc. The use of microfiltration allows for the return of solids to the production
624
13 E A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
process and the recycling of water. Starting from a feed stream with 3-7% solids (density 1.02-1.05 g / c m 3) the use of 0.2 ~tm membranes permits concentration to a concentration of 27% solids (density 1.2 g / c m 3) at a flux of 220 1/m 2 h. To achieve higher levels of solids concentration consecutive stages would be necessary. Similar flux values are reported by Hoogland et al. [34] in the filtration of SiO2-slurries with Norton Ceraflo membranes (0.2 ~tm). They clearly show that at a pH value of about 7, flux is at a minimum. However this minimum flux is much less time-dependent than flux at higher or lower pH: at pH = 7 the flux decreases from 390 to 225 1/m 2 h in the first hour, whereas at pH = 2.5 the flux decreases from 1630 to 550 1/m 2 h in the same time. 13.3.1.2 Paint and Ink
In the paint and ink producing and applying industries three different water uses can be distinguished [35]: cleaning operations in the production process of solvent-based paints and inks, carrier and solvent in water-based paints and inks, - water-curtains to catch over-spray from water-based paints in spray booths. The waste water resulting from cleaning paint production equipment contains a high quantity of pigments and solvents. In order to reduce the waste water volume as well as to recover the pigments, tests were performed with ceramic microfiltration. The applied ceramic membrane [35] has pores of 0.2 ~tm and forms a barrier for the pigments. The waste water is recirculated across the membrane until a sufficient concentration is reached (12% dry matter). This enables treatment in a filter press, for reuse of the pigments in the production process. The water fraction permeates through the membrane (flux: 100-250 l / ( m 2 h). The water is colourless and contains no pigments. The system is compact, reliable, and can be fully automated. Moreover, the waste water can be treated batch by batch without any problems. The specific operational costs are about NLG 30-35/m 3, which is considerably lower than the waste processing costs, but also lower than the costs of alternative methods of treatment (including electro-flotation). The increasing demand for solvent-free paints led to the introduction of water-based paints. Both in the production and in the application of this type of paint quite often a water-paint waste mixture results. Two examples are: leftovers, diluted with (cleaning) water and water used in water curtains in spray booths, which becomes increasingly contaminated with paint. Using ceramic membranes with pore sizes of 0.1 micron [35], it is possible to concentrate the paint particles to a very high degree (35-65% dry matter) and at the same time produce an effluent containing less than 0.1% paint. This can be -
-
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
625
reused or directly discharged into open waters or treated further with reverse osmosis before discharge. The permeate flux through the membranes varies from 50 to 250 1/m 2 h, dependent mainly on paint-type and concentration. Temperature ranges between 20 and 50~ Average cleaning interval is 4 weeks. The specific overall treatment costs are in the same range as in the above mentioned example, NLG 30-35/m 3. As in case of the solvent based paints this compares very favourably with the costs of other treatment possibilities and even seem to be lower than costs with polymeric membranes [36].
13.3.2 Lifetime Extension of Cleaning Baths 13.3.2.1 Alkaline Degreasing Baths Degreasing baths remove oil and other pollutants from the surface of metal components before this surface is treated. In due time the contents (1-2 m 3) of the degreasing bath become polluted and have to be exchanged for fresh cleaning solutions. This poses several problems: oil and dirt have to be separated from the discarded bath, the bath has to be neutralised before disposal, - changing the bath incurs high costs. Most often these actions are performed by specialised firms. Typical lifetimes of degreasing baths amount to 1-2 weeks. By drawing a continuous stream from the degreasing bath, and circulating it over a microfiltration system and concentration tank, oil and dirt can be retained in the concentration tank. Microfiltration with HIC's 0.2 ~tm membranes yields average fluxes of 250 1/m 2 h at temperatures from 40 to 70~ the pH lies between 9 and 11. The permeate contains less than 100 ppm oil. This treatment extends the life time of the degreasing bath up to five times; pay back time is less than two years. Similar data are reported of Carbosep, Le Carbone Lorraine and Atech membranes [19,22,37,38]. In these processes the retention of the detergents has to be monitored in order to keep their concentration in the degreasing bath at the correct level. -
-
13.3.2.2 Industrial Washing Operations In industrial cleaning of laundry, wool, leather, feathers, etc. large amounts of water and detergents are used. In the washing process this solution becomes polluted with fats, proteins, metals, etc., causing a high chemical oxygen demand (COD) and metal content. The use of ceramic membranes for laundry is necessitated in those cases where there is a risk for chemical contamination of the laundry. In the case of wool, leather, feathers it is the typical processing of
626
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
fats and proteins which necessitates the ceramic membranes. By microfiltration, followed by reverse osmosis, it is possible to regenerate the washing fluid so it can either be reused or discharged. Microfiltration with 0.2 ~tm ceramic membranes (HIC) yields average fluxes of 125-150 1/m 2 h at temperatures from 40 to 70~ Suspended solids and concentration of hydrocarbons are both reduced to less than 10 ppm in the permeate. By RO the COD is reduced to below 100 mg/1. Cleaning interval for the microfiltration installation is once a week. The feed to the microfiltration system has to be filtered over 100 ~tm screens to prevent clogging of the equipment. Menjeaud [39], treating 7 m3/h washing water from a laundry for the printing and mechanical industry, achieves permeate fluxes decreasing from 250 to 501/m 2 h as the concentration factor increases from 2 to 25. The concentrate has such a high COD that it can be used as combustible. An important factor in this use of membranes is the detergents/surfactants retention and fouling of the membrane. Although Akay and Wakeman's review [40] only deals with polymeric membranes, it thoroughly describes the various parameters influencing the behaviour of surfactants. Y/iksel et al. [23] indicate that ceramic membranes are much less prone to fouling by surfactants than polysulfone membranes. Maleriat and Schlumpf [41] show the dependence of the retention of a detergent (dodecyl benzene sulfonate) on its concentration. At values below the critical micelle concentration the retention is low, above it, retention increases with concentration. This behaviour is further complicated by temperature dependence: with increasing temperature fluxes increase and retention decreases.
13.3.3 Recycling in Chemical Processes 13.3.3.1 Cleaning of Organic and Inorganic Reagents A typical example of the application of ceramic membranes in chemical industry is the cleaning of mono ethanol amine. Mono ethanol amine (MEA) is used for the absorption of H2S from acid gasses but is polluted during this process by various organic compounds. Filtration of the MEA over 0.2 ~tm HIC ceramic membranes at an average flux of 32 1 / m 2 h produces a clean, transparent yellow liquid, free of solids. Filtration temperature is 37~ pH is about 11.5. Tests lasted successfully for over 700 h. Another example is the filtration of TiO2 from a waste stream in the so-called sulphuric acid process [42]. Using Le Carbone Lorraine membranes (0.2 ~tm) a stable average flux of 250 1/m 2 h is reached at 5 bar transmembrane pressure and 30~ A plant of almost 300 m 2 is laid out on a flux of 2001/m 2 h. Cleaning is performed with HF (2%), I hour a day.
13 m A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN L I Q U I D F I L T R A T I O N
627
13.3.3.2 Galvanic Baths In one example, parts from a nickelling bath are rinsed [43] with water. This rinsing water contains approximately 150 ppm nickel which is precipitated as Ni(OH)2 by addition of NaOH. The slurry resulting from this treatment passes a sedimentation tank and a filter press. This lowers the overall nickel content of the filtrate to ca. 3 ppm. A further treatment with ceramic membranes of 0.2 ~tm lowers the nickel concentration in the permeate to 0.2 ppm, which is well below the Dutch legal discharge limit of 0.5 ppm. The retentate is fed back to the sedimentation tank. The system of 2.4 m 2 treats about 800 1/h. Over polymer membranes the ceramic membranes show the advantage of a much longer interval between cleaning, viz., once per week instead of every day. 13.4 P R O C E S S I N G
13.4.1 Treatment of Liquid Products 13.4.1.1 Fruit Juices The application of ceramic membranes in the production of fruit juices is a well established technique [3,6,12,14,44-50]. A very wide range of fruit juices is designated (apple, pear, peach, orange, grapefruit, pineapple, kiwi fruit, strawberry, cranberry, carrot, beet); the clarification of apple juice seems to be the main application [6,14,44-49]. Merin and Daufin [44] present a review of older data, economical restraints being the limiting fluxes of the membranes and the impact of the short operating season in the production of apple juice. Gillot et al. [47], using ZrO2 0.1 ~tm Membralox membranes, clarify apple juice at fluxes between 200 and 250 1/m 2 h, concentration factor 10. Baur et al. [ 4 8 ] u s e 6 8 m 2 of 0.2 ~tm Membralox membranes and achieve fluxes of 100 to 150 1/m 2 h at temperatures ranging from 40 to 60~ concentration reaches 70 ~ Brix. Gupta et al. [49] report fluxes of 100 1/m 2 h using Norton Ceraflo and Le Carbone Lorraine membranes of 0.2 ~tm. They introduce pulsations in the circulation flow in order to enhance these fluxes.
13.4.1.2 Beer Brewing Yeast rests in fermenting cellars in beer breweries typically have a composition of 90% beer and 10% solids, mainly yeast. The amount of this waste material is 2-3% of the annual output. It can be sold as cattle feed or discharged. In a system with 4 m 2 0.4 ~tm ceramic microfiltration membranes, beer recovery amounts to 42-62%; the concentrate contains 23% solid matter [51]. Fluxes in
628
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
this recovery process are about 40 1/m 2 h at a temperature of 15~ Production runs vary from 8 to 16 h. The recovered beer can be blended back into the fermenting or lager cellar (in amounts up to 5%). In the treatment of tank bottoms [44,45,48,50,52] Finnigan et al. [52] reach long-term fluxes of approximately 20 1/m 2 h; Baur et al. [48] report average fluxes of 18 1/m 2 h (cleaning included) treating 500 h / d a y with 72 m 2 0.2 ~tm Membralox membranes. According to them circulation velocity and transmembrane pressure should be adapted to the yeast concentration because the secondary membrane layer completely governs the filtration process. Loss of this layer leads to blocking of the membrane and the temporarily passing of unwanted components.
13.4.1.3 Beer and Wine Clarification Publications on clarification of alcoholic drinks like wine and beer deal mainly with the treatment of wine. Advantages of ceramic membranes over classical methods are the reduction of operating costs (reduction of filter aids, less loss of product) and a better clarification. Ceramic membranes last longer and can be back flushed. According to Castelas and Serrano [53] microfiltration with pore sizes over 0.4 ~tm does not influence the wine, whereas pore sizes of 0.25 ~tm and lower disturb the organoleptic characteristics of the wine. However the complete removal of bacteria can only be achieved by 0.2 ~tm. Fouling of the membranes (Membralox) with coarser pore sizes, limits fluxes to 40-601/m 2 h, 0.2 ~tm is less affected and retains a flux of 85 1/m 2 h. Red wines seem to have a stronger tendency to fouling than white wines. Bauer [46] reports a decrease in fluxes from 275 1/m 2 h bar down to 11 1/m 2 h.bar for red wine and down to 1101/m 2 h bar for white wine, using Le Carbone Lorraine membranes. Similarly, but less extreme, Horgnies [50], in a very detailed description of commercial systems with Millipore Ceraflo 0.2 ~tm membranes, finds fluxes of 80-100 1/m 2 h for red wine and 100-120 1/m 2 h for dry white wine. Baur et al. [48] treating wine, rich in colloids, with 3.6 m 2 0.2 ~tm Membralox membranes report fluxes around 40 1/m 2 h. Belleville et al. [54] give a full description of the chemical nature of the fouling species. The use of enzymes to enhance the filterability might also be effective for raising the economy in the application of ceramic membranes [55]. Another means of achieving economical operation could be flux enhancement by the application of pulsating flow as outlined by Jaffrin et al. [56,57]. The simultaneous introduction of pulses, and a rise in circulation velocity from 3 to 4.37 m / s and of the transmembrane pressure from I to 4.5 bar, increases the flux from 35 1/m 2 h to 50 1/m 2 h.
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
629
In the clarification of beer by cross-flow microfiltration the paper by Tr~igardh and Wahlgren [58] seems to be one of the sporadic examples of this application. Here the use of 0.5 ~tm membranes (Membralox) is necessary to maintain the taste of the beer; 0.2 ~tm shows an unacceptable retention of proteins and colour. Bacteria were retained by the 0.5 ~tm membrane. An editorial in "Filtration and Separation" [59] highlights the reluctance of the beer brewers to change over from kieselguhr filtration to membranes, but judges that the examples of fruit juice and wine production show good market prospects. 13.4.1.4 Potable Water
Drinking water is a major necessity of life: many membrane processes have been developed to produce it a n d / o r enhance its quality. Filtration [16] aims at the removal of: - suspended particles, precipitates caused by water hardness a n d / o r salts, micro-organisms: algae, bacteria, fungi. Examples of the use of ceramic membranes in the production of potable water are quite numerous [42,60-65]. An interesting review is presented by Pou6t et al. [60] of some 15 installations working with ceramic membranes for the production of drinking water. Sizes of these installations, installed in France between 1984 and 1990, vary from 5 to 100 m3/h. Moncorg6 and Pascal [61] and Bauer et al. [42] describe the use of the carbon/carbon composite membranes of Le Carbone Lorraine in the filtration of drinking water. With 0.2 ~tm membranes the fluxes range between 1000 and 2000 1/m 2 h at trans-membrane pressures from I to 2 bars. The use of Kerasep membranes [65] (Rhone-Poulenc's alumina/alumina membranes, 0.2 ~tm pore size) leads to fluxes of 600-1200 1/m 2 h at 2 bar transmembrane pressure. Micro-organisms form a very important source of fouling: various authors [66,67] report a strong decline in flux, even as much as 70% of the original value, in the presence of micro organisms. Moulin et al. [64] use a coagulant and ozone to enhance the flux of their 0.2 ~tm ceramic membrane (Membralox). Using ozone and a concentration of 45 ppm coagulant, flux is approximately 15001/m 2h, with 110 ppm coagulant flux increases to 20001/m 2 h. The ozone treatment decomposes the organic material, so fouling is kept at a minimum. The combination of electro-coagulation, flotation and microfiltration is applied by Pou6t et al. [60]. In this case Membralox 1P19-40 membranes are used: a zirconia/alumina composite membrane of 50 or 100 nm pore size. With 100 nm and the use of electro-coagulation fluxes stabilise on 250-350 1/m 2 h. In the treatment of river water Mietton Peuchot and Ben Aim [68] use polyaluminium chloride as flocculant, raising the flux of Membralox 0.2 ~tm membranes from 200 towards 800 1/m 2 h. -
-
630
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
13.4.2 Treatment of Semi-solid Products 13.4.2.1 Proteins Most of the processing of proteins with ceramic membranes is in the field of the dairy industry; some work with other proteins will be presented in Section 13.4.3. Merin and Daufin [44] and Bhave [3] present a comprehensive review of the field, the main use of ceramic membranes being protein concentration by microor ultra-filtration and bacteria removal by microfiltration. For the latter the Bactocatch process, as described by Gillot et al. [47], Merin and Daufin [44] and Bhave [3] forms an important example. At an average flux of 700 1/m 2 h 99.7% of the bacteria are withheld without retaining the proteins. The production of casein [12,69] is a good example of the processing of proteins with ceramic membranes. Surel and Famelart [69] delineate this process: using either 0.1 or 0.2 ~tm Membralox membranes fluxes are 54 1/m 2 h in the absence of calcium, decreasing to 31 1/m 2 h after addition of up to I g/1 of calcium. Retention of both 0~- and [~-casein by the 0.1 ~tm membrane is better, and is further enhanced by the addition of calcium. The very serious problem of fouling by proteins is corroborated by many publications [41,70,71]. Various parameters influencing the fouling behaviour have been studied. Clark et al. [70] discuss the influence of protein concentration, trans-membrane pressure, cross flow velocity and pH. For pore sizes of 0.1 ~tm (Membralox membranes), filtering bovine serum albumin, the flux has a minimum at the pH of the protein isoelectric point. Dumon and Barnier [71] show that the amount of protein adsorption depends on previous adsorption. Contacting with citrate or phosphate lowers a subsequent protein adsorption; contacting with nitrate increases the protein adsorption. Rios et al. [72] show that with small pores (< 0.2 ~tm) protein fouling remains on the outside of the membrane, whereas with the larger pore sizes the pores become blocked by the intrusion of protein into these pores. Cleaning after fouling by proteins is an important issue for the economical application of membranes. Kerherve et al. [73], Gillot et al. [47] and Daufin et al. [74] describe cleaning cycles, necessary to re-establish initial values of fluxes. Efficient cleaning was achieved by means of NaOC1 (with or without an acid step comprising HNO3) and by means of NaOH, 'reinforced' with complexing agents and surfactants.
13.4.2.2 Whey As whey contains many nutrient compounds (lactose, proteins, minerals and some fat) its use as starting material for the manufacture of various specialty products expands.
1 3 - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
631
Microfiltration of whey prior to ultrafiltration in the production of whey protein concentrates (WPC) was reported among others by Maubois et al. [75], van der Horst [76], and Wnuk et al. [77]. The microfiltration step also prevents fouling of the UF-membranes (either polymeric membranes or ceramic membrane): e.g. Daufin et al. [78] by phosphates and calcium. Ceramic membranes yield higher fluxes (up to 691/m 2 h) and better separation then polymeric membranes, resulting in WPCs with lower fat contents [76]. Experiments at HIC with its 0.5 ~tm ceramic membranes showed fluxes of well over 200 1/m 2 h. Special importance however is required for any aggregation step preceding the microfiltration: both Gesan et al. [79] and Daufin et al. [78] emphasise the influence of the controlled aggregation by the addition of calcium. Gesan et al. [79] describe the performance of 57 m 2 Carbosep M14 membranes in the defatting of rennet whey, stressing the point that the performance has to be improved by a better control of fouling. Daufin et al. [78] show that through this microfiltration step the UF-step (Carbosep M5, 10,000 D) performs very good, yielding fluxes of 50 to 120 1/m 2 h, even with very high protein contents. Surel and Famelart [69], in a study with 0.1 and 0.2 ~tm Membralox-membranes, show that an addition of calcium lowers the MF permeate flux from 54 1/m 2 h to 20 1/m 2 h at a velocity of 6 m/s. This flux is quite dependent on velocity: 35 1/m 2 h at 4 m/s, 541/m 2 h at 6 m / s and 631/m 2 h at 7.3 m/s. Analogous to this processing, Korolczuk and Mahaut [80] report the necessity to use ceramic membranes for the filtration of acid-coagulated milk in order to produce UF-fresh cheeses with good taste. Typical fluxes, using Carbosep M1 (cut-off 50,000 D) increase from 10 to 20 1/m 2 h at 40~ with decreasing concentration factor.
13.4.2.3 Sugars Punidadas et al. [81] describe detailed experiments on the refining of raw cane sugar. Use of 0.2 m 2 SCT membranes with pore sizes between 0.1 ~tm and 0.5 ~tm effectively removed almost 100% of the solutions turbidity and 50% of its colour. Average flux is 38 1/m 2 h; the higher values are reached at higher operating temperature (90~ Interestingly, it is shown that working with the smaller pore sizes requires use of high tangential velocity from the very start of the process, whereas pore sizes of 0.5 ~tm and larger perform better after building a secondary membrane layer by first applying a modest velocity. The microfiltration process on its own is not sufficient for the complete purification of thecane sugar, however it prevents the ion exchangers from fouling and poisoning [82]. One of the side streams in the production of glucose from corn starch contains a high concentration of glucose. This very sticky suspension is heavily
632
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
contaminated with fats, proteins, fibres and other insoluble material and usually is used for animal feed. Filtration of this suspension with 0.2 ~tm ceramic membranes (HIC) at 45-60~ and pH = 4.5, produced a clear permeate at an average flux of 1901 / m 2 h.
13.4.2.4 Paper and Pulp The manufacturing process of paper and pulp consumes enormous amounts of water. Reduction of these streams by recirculation of the process water in the plants is of great environmental benefit. Typical waste waters are the so-called white water and wash water of paper recycling plants, containing ink. Tests with ceramic microfiltration membranes (HIC) show fluxes in the range of 100-200 1/m 2 h; pre-filtering of coarser components is indispensable. Treatment of the MF-permeate with ultrafiltration, or direct ultrafiltration, removes most of the high-molecular weight components which interfere with the paper-making process [83]. It was shown that the brightness of the produced paper increased by the use of Carbosep M5 (10,000 D) or other ultrafiltration membranes [83]. Due to the fouling of polysulphone membranes in bleach plant effluents Afonso and Pinho [84] studied the use of Carbosep membranes for the combination of ultra-filtration (10,000 D) and microfiltration (0.14 ~tm). The introduction of microfiltration preceding ultrafiltration improves the performance of the latter, regarding limiting fluxes at given feed circulation velocities.
13.4.3 Biotechnology Speaking about biotechnology the topics of the use of membrane reactors and the filtration of yeast, enzymes and proteins are discussed most often. Sometimes it is difficult to discern biotechnology from applications in more established industries like dairy, etc. Besides that, in many papers biotechnology is mentioned in a rather general sense [6,11,85-87], perhaps indicating the freshness of these processes a n d / o r some reluctance in communicating details about the application. In the filtration of fermentation broths, lysed yeast [21,88-91] microfiltration is used to separate the yeast cells a n d / o r cell fragments. For the Ceramesh ceramic/metal composite membrane of 0.2 ~tm pore size a flux of 60 1/m 2 h is reported [21] for lysed yeast, at a temperature of about 55~ and a solids concentration of up to 16-17%. The same magnitude of flux and solids concentration are obtained with whole yeast suspensions. Using Kubota membranes in the range of 50 nm to 0.8 ~tm Narukami et al. [88] choose 0.8 ~tm for their work with fermentation broth. They report a stable flux of 20 1/m 2 h using suction (0.8 bar) on the permeate side as driving force, whereas the flux decreases as a
13 m A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
633
function of time to 12.5 1/m 2 h when one uses 0.8 bar pressure in the common way of operation. This difference is attributed to the compaction of the cake layer on the membrane. Chang et al. [89] filtrate a alcohol-distillery waste with 50 nm and 0.4 ~tm membranes from TIA (France). Although the 0.4 ~tm membrane has a higher initial flux, the flux of the 50 nm membrane is always higher in the long run. At a concentration factor of 2, the flux for the 50 nm pore size amounts to 2151/m 2 h, for the 0.4 ~tm pore size it is 1851/m 2 h. The dependence of flux on concentration factor also depends on the type of raw material for the distillation process. A further example of the separation of bio-mass is given by Maebashi [90] in the filtration of the sediment of soy sauce production. The flux of a rotating TOTO 0.1 ~tm disk membrane ranges from 25 to 8 1/m 2 h as the concentration factor increases from 15 to 200, at a transmembrane pressure of 2 bar. Imasaka et al. [91] study the effect of gas-liquid two-phase crossflow filtration of bakers yeast. They employ tubular TOTO 0.2 and 0.5 ~tm membranes. They show that this way of operation significantly reduces the specific energy of the separation process. The separation of proteins and enzymes is performed with ultrafiltration membranes. Branger et al. [93] use Carbosep M1 and M4 (40,000 and 20,000 Dalton respectively) for the separation of enzyme hydrolysates. The fluxes with these membranes compare favourably with polymeric membranes: 37-102 1/m 2 h vs. 7-41 1/m 2 h. The use of Schott's porous glass membranes (pore sizes from 10 to 90 nm) in the separation of proteins with molecular weights from 14,400 to 450,000 is illustrated by Langer and Schnabel [85] who show a decrease in retention with increasing pore size for different proteins. Due to the chemical nature of the membrane material, it lends itself to surface modifications, including the coupling of enzymes or the attachment of micro-organisms. The separation of proteins can be improved by chemical modification of the membrane surface [94]. Coating a Carbosep M5 membrane (10,000 D) with quaternized polyvinylimidazole raises the retention of tetracycline from 25% towards 76%. Unfortunately the flux declines at the same time from 32 1/m 2 h to 7.6 1/m 2 h. The porous ceramic membrane can be used to either separate biologically reacting material in reactors, or carry catalysts, microbes or enzymes to influence the desired reactions. An overview of the Japanese efforts for the establishment of membrane reactors in the "Aqua Renaissance '90 Project" are summarised by Kimura [95]; a very recent review was written by Zaman and Chakma [96]. The preparation of microporous membranes (pore diameters smaller than 2 nm) for the application in membrane reactors is described by Keizer et al. [97] and Julbe et al. [98], however without detailing the membrane reactor itself.
634
13 - - A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
Various Japanese researchers s h o w the possibilities of i m m o b i l i s i n g an enz y m e or yeast o n / i n a ceramic m e m b r a n e [99-101]. In the first example [99], the e n z y m e is b o u n d by Nakajima et al. to the ceramic surface of the TOTO 50 n m m e m b r a n e by activating it first w i t h a silane-glutaraldehyde technique. Invertase is then b o u n d to this activated surface and converts 100% of the 10-50 w t % sucrose in the feed solution. Alternatively glucose-isomerase yields a fructose ratio of 42% in a 45 wt% glucose feed at a residence time of 1000 s. The p r o d u c t i v i t y of such systems is 10-fold higher than in conventional columns in w h i c h the e n z y m e is immobilised in beads. H o r i t s u [100] immobilises yeast cells on the surface of a ceramic carrier by the different charge of carrier and cells. Using this set-up, soy sauce, beer a n d sake are p r o d u c e d w i t h fermentation times m u c h shorter, up to 10 times, than in conventional processing. Kawase et al. [101] s t u d y this immobilisation beh a v i o u r by m e a s u r i n g zeta-potentials and find a neat correlation of this potential w i t h the n u m b e r of a d s o r b e d cells.
REFERENCES
1. L. Cot and J. Charpin (Eds.), Proceedings of the 1st International Conference on Inorganic Membranes, 3-6 July 1989, Montpellier, France. 2. A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Proceedings of the 2nd International Conference on Inorganic Membranes, 1-4 July 1991, Montpellier, France. Trans Tech Publications, Z/irich, 1991. 3. R.R. Bhave, Inorganic Membranes: Synthesis, Characteristics, and Applications. Van Nostrand Reinhold, New York, 1991. 4. H.P. Hsieh, Inorganic Membranes. AIChE Symposium Series, Vol. 84, 1988, lp. 18. 5. L. Cot, Ch. Guizard and A. Larbot, Novel ceramic material for liquid separation process: present and prospective applications in microfiltration and ultrafiltration. Industr. Ceram., 8 (1988) 143--148. 6. J. Charpin and P. Rigny, Inorganic membranes for separative techniques: from uranium isotope separation to non-nuclear fields, in Ref. [1], pp. 1-16. 7. K.K.Chan and A.M. Brownstein, Ceramic membranes m Growth prospects and opportunities. Am. Ceram. Soc. Bull., 70 (1991) 703-707. 8. J. Charpin, A.J. Burggraaf and L. Cot, A survey of ceramic membranes for separations in liquid and gaseous media, in: P. Vincenzini (Ed.), Ceramics Today Tomorrow's Ceramics. Materials Science Monographs, Vol. 66D. Elsevier, New York, 1991, pp. D2623-2638. 9. S.I.Nakao, Current status of inorganic membrane in Japan, in Ref. [2], pp. 219-228. 10. L.M. Sheppard, Porous ceramics: processing and applications, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 3-23. 11. G. Cueille and M. Ferreira, Place of mineral membranes in the processes for bio-industry and the food-industry, in: Ref. [1], pp. 303-309. 12. D. Bonneau, G.E. Brinkman and J. Guibaud, Nouveaux produits et d6veloppements dans la gamme Membralox, in: Ref. [2], pp. 313-318.
13-- APPLICATIONSOFCERAMICMEMBRANESIN LIQUIDFILTRATION
635
13. Business Communications Company: Inorganic membranes: markets, technologies, players; February 1994; press release. 14. A. Crull, Prospects for the inorganic membrane business, in: Ref. [2], pp. 279-288. 15. H. Gasper, Handbuch der industriellen Fest/Fliissig-Filtration. H~ithig Buch Verlag, Heidelberg, 1990. 16. S. Ripperger, Mikrofiltration mit Membranen: Grundlagen, Verfahren, Anwendungen. VCH Verlagsgesellschaft, Weinheim, Germany, 1992. 17. W.S.W. Ho and K.K. Sirkar, Membrane Handbook. Van Nostrand Reinhold, New York, 1992. 18. J.W. Wouters, Goedkope olie/waterscheiding met keramische membranen. Polytechn. Tijdschr., May (1994) 46--47. 19. T.A. Peters, Aufbereitung von Abwasser und Prozet~fliissigkeiten durch Ultrafiltration und Crossflow-Mikrofiltration mit atech-SiC-Membraben. Filtrieren Separieren, 7 (1993) 312-316. 20. C. De Haan and J.W. Wouters, Emulsieverwerking met keramische membranen. MilieuMarkt, Jan./Feb. (1994) 20-21. 21. D.R. Cowieson and P.M. Gallagher, Recent developments in formation and applications of metal mesh/ceramic composite membranes, in: Ref. [2], pp. 23-32. 22. Anon., Carbosep-Ultrafiltration zur Emulsionsspaltung. Filtrieren Separieren, 5 (1991) 387-388. 23. L. Yfiksel, R. Sedelies, U. Mfiller and B. Spei, A new concept in the economical treatment of waste emulsions: chemical emulsion breaking and membrane filtration, in: Proceedings 5th World Filtration Congress, Nice, France, 1990, pp. 286-292. 24. G.W. Meindersma, Membraanprocessen in de praktijk. PT-Procestechniek, 44 (1989) 44-47. 25. A.S.C. Chen, J.T. Flynn, R.G. Cook and A.L. Casaday, Removal of oil, grease, and suspended solids from produced water with ceramic crossflow microfiltration. SPE Prod. Eng., (1991) 131-136. 26. A. Zaidi, K. Simms and S. Kok, The use of micro/ultrafiltration for the removal of oil and suspended solids from oilfield brines. Water Sci. Tech., 25 (1992) 163-176. 27. L. Mendia, P. Carbone, G. d'Antonio and M. Mendia, Treatment of olive oil waste waters. Water Sci. Tech., 18 (1986) 125-136. 28. M.C. Annesini, A.R. Giona, F. Gironi and F. Pochetti, Distillation process for vegetable water treatment. Effluent Water Treat. J., 24 (1984) 450-455. 29. G. DiGiacomo, V. Brandani and G. DelRe, Evaporation of olive oil mill vegetation waters. Desalination, 81 (1991) 249-259. 30. S.G. Velioglu, K. Curi and S.R. Camlilar, Activated sludge treatability of olive oil-bearing wastewater. Water Res., 26 (1992) 1415-1420. 31. P. Canepa, N. Marignetti, U. Rognoni and S. Calgari, Olive mills wastewater treatment by combined membrane processes. Water Res., 22 (1988) 1491-1494. 32. F. Quemeneur and P. Jaouen, Potentialities of inorganic membranes in fish processing i n d u s t r i e s - economic aspects, in: Ref. [2], pp. 585-588. 33. Netzsch Filtrationstechnik GmbH, Waste water avoidance in the ceramic industry by means of microfiltration. Interceram, 42 (1993) 44-45. 34. M.R. Hoogland, A.G. Fane and C.J.D. Fell, The effect of pH on the crossflow filtration of mineral slurries using ceramic membranes, in Ref. [1], pp. 153-162. 35. W.F.J.M. Nooijen and F. Muilwijk, Paint/water separation by ceramic microfiltration. Filtration Separation, 31 (1994) 227-229.
636
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
36. E. VanKooij, Spoelwater van verfindustrie bruikbaar als grondstof, Proces Technologie, 3 (1993) 20-23. 37. Anon., Crossing the micron flow, Filtration Separation, 30 (1993) 19-20. 38. M. Hansmann, Dynamisch filtrieren, Industrie-Anzeiger, 39 (1993) 19-21. 39. C. Menjeaud, Treatment and regeneration of waste water produced by industrial laundry with inorganic membranes, in: Ref. [2], pp. 589-592. 40. G. Akay and R.J. Wakeman, Ultrafiltration and microfiltration of surfactant dispersions An evaluation of published research, Trans. AIChE (Chem. Eng. Res. Des.), 71 (1993) 411-420. 41. J.P. Maleriat and J.P. Schlumpf, Interactions solute-membrane lors de l'ultrafiltration d'un tensio-actif, in: Ref. [2], pp. 482-484. 42. J.M. Bauer, J. Elyassini, G. Moncorge, T. Nodari and E. Totino, New developments and applications of carbon membranes, in: Ref. [2], pp. 207-212. 43. Novem, Demonstratieprojecten energiebesparing: projectresultaat 412, 1993. 44. U. Merin and G. Daufin, Separation processes using inorganic membranes in the food industry, in: Ref. [1], pp. 271-281. 45. J. Guibaud, Some applications of Membralox | ceramic membranes, in ref. [1], pp. 343-348. 46. J.M. Bauer, Utilisation de composites carbone-carbone dans la fabrication de membranes minerales de microfiltration et d'ultrafiltration, in Ref. [1], pp. 361-366. 47. J. Gillot, R. Soria and D. Garcera, Recent developments in the Membralox ceramic membranes, in: Ref. [1], pp. 379-381. 48. W. Baur, L. Gottkehaskamp and D. Oechsle, Die Anwendung von keramischen Membranen bei der Querstrom-Filtration in der Getr~inkeindustrie. Filtrieren Separieren, 6 (1992) 141-147. 49. B.B. Gupta, P. Blanpain and M.Y. Jaffrin, Permeate flux enhancement by pressure and flow pulsations in microfiltration with mineral membranes. J. Membr. Sci., 70 (1992) 257-266. 50. M.C. Horgnies, D6veloppement et application de la filtration tangentielle sur c6ramique en agro-alimentaire. Liquides Mag. (1994) 49-53. 51. Anton Steinecker Maschinenfabrik GmbH, Freising (BRD), Crossfow-micro-filtration with ceramic membranes, Company Product Bulletin, 1993. 52. T. Finnigan, R. Shackleton and P. Skudder, Using ceramic microfiltration for the filtration of beer and recoevry of extract. Filtration Separation, 26 (1989) 198-200. 53. B. Castelas and M. Serrano, Utilisation des membranes dans le traitement du vin, in: Ref. [1], pp. 283-290. 54. M.P. Belleville, J.M. Brillouet, B. Tarodo de la Fuente and M. Moutounet, Cross-flow microfiltration of a red wine on a alumina membrane: investigation on fouling colloids, in: Ref. [2], pp. 477-480. 55. R. Urlaub, Wirkungsspezifisch und gut steuerbar: Enzyme in der Weinbereitung. Die Erniihrungsindustrie, 1992 (1992) 6-10. 56. M.Y. Jaffrin, R. Ben Amar and B.B. Gupta, Membrane fouling control in cross flow filtration of wine with mineral membranes, in: International Technical Conference on Membrane Separation Processes, Brighton, UK, 24-26 May 1989, Paper E2. 57. M.Y. Jaffrin, B.B. Gupta and P. Paullier, Energy saving pulsatile mode cross flow filtration. J. Membr. Sci., 86 (1994) 281-290. 58. G. Tr~igardh and P.E. Wahlgren, Removal of bacteria from beer using crossflow microfiltration, in: Ref. [1], pp. 291-295.
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
637
59. Editorial, Beer filtration- Drinktec Interbrau. Filtration Separation, 30 (1993) 521. 60. M.F. Pou~t, F. Persin and M. Rumeau, Etude du couplage electrocoagulation-flottation pour limiter le colmatage des membranes en traitement des eaux, in: Ref. [2], pp. 237-242. 61. M.F. Pou~t, F. Persin, M. Gros and M. Rumeau, Etude de pretraitements avant ultra ou microfiltration tangentielle, in: Ref. [2], pp. 549-552. 62. G. Moncorg6 and G. Pascal, Neueste Entwicklung an Kohlefaser-KohlenstoffMikrofiltrationsmembranen und deren Anwendung. Filtrieren Separieren, 6 (1992) 156--160. 63. C. Moulin and M. Rumeau, Potabilisation d'eau par microfiltration tangentielle sur membrane minerale, in: Ref. [1], pp. 515-518. 64. C. Moulin, M.M. Bourbigot and M. Faivre, Interest of the ozone/coagulant combination for the potabilization of surface waters by crossflow microfiltration on mineral membranes, in: Ref. [2], pp. 229-236. 65. Rhone-Poulenc, Der Einsatz von Kerasep-Membranen in der Wasser- und Abwasserbehandlung. Filtrieren Separieren, 7 (1993) 267-268. 66. F. Duclert and M. Rumeau, Microfiltration d'eau sur membrane minerale; influence de la qualit6 de l'eau, in: Ref. [1], pp. 493-496. 67. S. Elmaleh and W. Naceur, Transport of water through an inorganic composite membrane. J. Membr. Sci., 66 (1992) 227-234. 68. M. Mietton Peuchot and R. Ben Aim, Improvement of crossflow microfiltration performances with flocculation. J. Membr. Sci., 68 (1992) 241-248. 69. O. Surel and M.H. Famelart, Microfiltration of sodium caseinate on ceramic membranes, in: Ref. [2], pp. 509-512. 70. W.M. Clark, A. Bansal, M. Sontakke and Y.H. Ma, Protein adsorption and fouling in ceramic ultrafiltration membranes. J. Membr. Sci., 55 (1991) 21-38. 71. S.Dumon and H. Barnier, Ultrafiltration of protein solutions on ZrO2 membranes. The influence of surface chemistry and solution chemistry on adsorption, J. Membr. Sci., 74 (1992) 289-302. 72, G.M. Rios and P. Freund, Basic studies on transport and fouling phenomena during protein UF and EUF on alumina membranes, in: Ref. [1], pp. 171-176. 73. F.L. Kerherve, S. Rio, U. Merin, J.P. Labb6, A. Qu6merais, F. Michel and G. Daufin, Nettoyage de membranes d'ultrafiltration de lactoserum et de lait, in: Ref. [1], pp. 419-423. 74. G. Daufin, U. Merin, F.L. Kerherve, J.P. Labb6, A. Qu6merais and Ch. Bousser, Effinciency of cleaning agents for an inorganic membrane after milk ultrafiltration, in: Ref. [2], pp. 553-556. 75. J.L. Maubois, J. Pierre, J. Fauquant and M. Piot, Industrial fractionation of main whey proteins. IDF Bull., 212 (1987) 154-159. 76. Van der Horst, Microfiltration in whey processing, in: Ref. [1], pp. 297-302. 77. R. Wnuk, N. Stroh and H. Chmiel, Inorganic membranes in the food and biotechnology industries; A study on fouling inorganic membranes, in: Ref. [1], pp. 479-482. 78. G. Daufin, J.P. Labb6, A. Quemerais, F. Michel, J. Fauquant and J.F. Radenac, Optimizing pH for improved defatted whey ultrafiltration using an inorganic membrane, in: Ref. [2], pp. 557-560. 79. G. Gesan, U. Merin, G. Daufin and J.J. Maugas, Performance of an industrial microfiltration plant for defatting rennet whey, in: Ref. [2], pp. 307-312. 80. J. Korolczuk, and M. Mahaut, Rheological properties of UF-fresh cheeses, in: Ref. [2], pp. 491-494.
638 81.
82.
83.
84. 85. 86.
87. 88.
89.
90.
91. 92. 93. 94.
95. 96. 97.
98.
99. 100.
13 ~ A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S I N LIQUID F I L T R A T I O N
P. Punidadas, M. Decloux and G. Trystram, Microfiltration tangentielle sur membrane min6rale en c6ramique. Application au traitement du sucre roux. Ind. Alim. Agricol., (1990) 615-623. M. Decloux, E.B. Messaoud and M.L. Lameloise, Etude du couplage microfiltration tangentielle/6change d'ions en raffinerie de sucre de canne. Ind. Alim. Agricol., (1992) 495-502. J. Nuortila-Jokinen, T. Uusluoto and M. Nystr6m, Removal of disturbing substances by ultrafiltration of make-up waters in the pulp and paper industry. Paper Timber, 76 (1994) 256-261. M.D. Afonso and M.N. Pinho, Membrane separation processes in pulp and paper production. Filtration Separation, 28 (1991) 42-44. P. Langer and R. Schnabel, Porous glass UF-membranes in biotechnology, in: Ref. [1], pp. 249-255. V.A. Lyalin and V.D. Alpem, Filtration tangentielle sur les membranes inorganiques: comment augmenter son rendement en biotechnologie et industrie alimentaire, in: Ref. [2], pp. 123-130. Anon., Membranes: Projet Eureka pour Tech-Sep. Informations Chimie (1992) 137-139. Y. Narukami, A. Kayawake, M. Shioyama, Y. Okamoto, K. Tokushima and M. Yamagata, Ceramic membrane filtration of methane fermentation broth, in: Ref. [1], pp. 267-270. I.S. Chang, K.H. Choo, C.H. Lee, U.H. Pek, U.C. Koh, S.W. Kim and J.H. Koh, Application of ceramic membrane as a pretreatent in anaerobic digestion of alcohol-distillery wastes, J. Membr. Sci., 90 (1994) 131-139. N. Maebashi, Ceramic membranes and application to the recovery of soy sauce, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 81-87. T. Imasaka, N. Kanekuni, H. So and S. Yoshino, Gas-liquid two-phase cross-flow filtration by ceramic modules. Kagaku Kogaku Ronbunshu, 15 (1989) 638-644. W.M. Clark, A. Bansal, M. Sontakke and Y.H. Ma, Protein adsorption and fouling of ceramic membranes during ultrafiltration, in: Ref. [1], pp. 415-418. J.L. Branger, R. Audinos, J. Noguera and M. Chignac, Ultrafiltration concentration of enzyme hydrolysates by mineral membranes, in: Ref. [1], pp. 243-248. B. Chaufer, M. Rollin, A. Grangeon and J. Dulieu, Tetracycline removal or concentration with an inorganic ultrafiltration membrane modified by a quatemarized polyvinylimidazole coating, in: Ref. [2], pp. 249-254. S. Kimura, Japan's Aqua Renaissance '90 Project. Water Sci. Tech., 23 (1991) 1573-1582. J. Zaman and A. Chakma, Inorganic membrane reactors, J. Membr. Sci., 92 (1994) 1-28. K. Keizer, V.T. Zaspalis and A.J. Burggraaf, Passive and catalytically active membranes for affecting chemical reactions, in: P. Vincenzini (Ed.), Ceramics T o d a y - - Tomorrow's Ceramics. Materials Science Monographs, Vol. 66D, Elsevier, New York, 1991, pp. D2511-2524. A. Julbe, C. Guizard, A. Larbot, L. Cot and A. Giroir-Fendler, The sol-gel approach to prepare candidate microporous inorganic membranes for membrane reactors, J. Membr. Sci., 77 (1993) 137-153. M. Nakajima, N. Jimbo, H. Nabetani and A. Watanabe, Use of ceramic membrane for enzyme reactors, in: Ref. [1], pp. 257-266. H. Horitsu, A new approach that uses bioreactors with inorganic carriers (ceramic) in
13 -- APPLICATIONSOF CERAMICMEMBRANESIN LIQUIDFILTRATION
639
the production of fermented foods and beverages, in: in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 381-389. 101. M. Kawase, Y. Kamiya and M. Kaneno, Porous ceramic carrier for bioreactor, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 391-400.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 14
Feasibility of the application of porous inorganic gas separation m e m b r a n e s in s o m e large-scale chemical processes Henk M. van Veen, Maarten Bracht, Edwin Hamoen and Peter T. Alderliesten Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands
14.1 INTRODUCTION During the last decade there has been intensified activity in research and development of ceramic membranes for gas separation applications. In several studies it is said that the market for these membranes will expand very rapidly in the near future [1-3]. This market growth will be due to advantages such as high permeation and membrane stability as compared with other membrane separation technologies. During the first years of inorganic membrane development, R&D was mainly focused on the membrane as the product, and research was driven by materials development and materials scientists. Research was carried out by universities, while research institutes and especially (end-user) industry were hardly involved. The main reason for this was that a lot of fundamental knowledge was needed before these membranes could be implemented in the foreseen market.
642
14
-
-
APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
Nowadays somewhat more attention is paid to the application and use of these membranes in processes and under process conditions. Industry is now getting more involved in R&D. In several review articles, membrane development and possibilities of inorganic membranes in gas separation applications [4-8] and especially in membrane reactor applications [9-16] have been summarised. In most of the literature the use of inorganic membranes for gas separation and reactor applications are considered to be very promising. However, research is still strongly focused on the membrane as a material and much less on the membrane process. As a result, insufficient data, especially on testing under realistic circumstances, are available on the real possibilities of inorganic membranes in large-scale processes. There is a need for such data and extensive technical and economic evaluations of membranes in different possible applications should be made, preferably using a multidisciplinary approach. Aspects such as chemical engineering and mechanical engineering are as important as materials engineering to introduce inorganic gas separation membranes into commercial processes. Furthermore, in order to introduce these membranes into the market successfully all aspects starting from fundamental material development to marketing strategies must be considered, depending, of course, on the state of development. If these aspects are taken into account it will become clear that the introduction of inorganic membranes in petrochemical and energy production processes is more difficult than first expected. The aim of this chapter is to show that a multidisciplinary approach, focusing on materials, processes and modelling as depicted in Fig. 14.1, is needed to judge the techno-economic feasibility of inorganic membranes in large-scale processes. This will be done by discussing examples of the potential use of porous inorganic membranes in three different membrane reactor applications.
Materials
/
'l
Processes
Modelling
Fig. 14.1. Disciplines to be c o n s i d e r e d .
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
643
In all three, hydrogen separation will take place: the dehydrogenation of propane to propylene, the dehydrogenation of ethylbenzene to styrene, and the water-gas shift reaction. Membrane characteristics such as permeation, selectivity and separation factor are given throughout this chapter. The definitions for these characteristics are given in the appendix.
14.2 B A C K G R O U N D I N F O R M A T I O N
14.2.1 Materials
Hydrogen selective inorganic membranes can be mesoporous (2 nm < pore diameter < 50 nm; ceramic, glass or Carbon) microporous (pore diameter < 2 nm; ceramic, carbon or zeolite) or dense (ceramic or metal). These membranes can be used from ambient temperatures up to about 600~ for mesoporous materials, up to about 500~ for microporous inorganic membranes and up to about 800~ for dense inorganic membranes [14-16]. These temperatures are only a rough indication, because of the different materials which can be used and the test conditions at which the membranes have to operate. Typical characteristics of both porous and dense inorganic membranes are given in Table 14.1. Only applications with porous ceramic membranes will be dealt with in this chapter. TABLE 14.1 Typical characteristics of inorganic gas separation membranes Membrane system
Pore diameter + (thickness)
Temp. (~
Gas mixture 1
Permeation 2 ( m o l / m 2 s Pa)
Permselectivity 3
Mesoporous alumina [17]
4 nm (3 btm)
25 250 475
H2/C3H8
6 x 10-5 4 x 10-5 3 x 10-5
3.0 3.7 4.0
Mesop~rous glass: Vycor .... (Toshiba) [18]
4.5 n m (300 l.tm)
20
H2/N2 and H2/CO2
7.4 x 10-8
Knudsen
Mesoporous carbon [19]
several nm (12-18 btm)
400
H 2 /C O-C O2- 2 x 10-6 H2S
3.5:CO* 4.5:CO2"
Microporous SiO2 on A120 3 by polymeric sols [20]
appr. 10 ~, (100 nm)
25 100 200
H2/C3I--I6
14 62 156
7 x 10-7 10 x 10-7 11 x 10-7
(continued)
644
14 -- APPLICATIONOF POROUSINORGANICGAS SEPARATIONMEMBRANES
TABLE 14.1 (continuation) Membrane system
Pore diameter + (thickness)
Temp. (~
Gas mixture 1
Permeation 2 (mol/m 2 s Pa)
Permselectivity 3
Microporous SiO2 on A1203 by polymeric sols + CVD [21]
appr. I nm (5 ~tm)
50 250
H2/N2 H2/N2
1.8 x 10-8 5.4 x 10-8
44 200
Hollow fibre microporous glass (PPG) [22]
4-8/~ (5 ~In)
204 260 316 371
H2/CO H2/CO H2/CO H2/CO
8.7 x 10-9 11.0 x 10-9 10.9 x 10-9 10.2 x 10-9
325* 205* 147" 101"
Microporous carbon molsieve; hollow fibre [23]
appr. 5/~ (6 ~xn)
20 20 200 500
O2/N2 He/N2 H2/CH4 H2/CH4
6 x 10-8 3 x 10-7 1.07 x 10-7 1.16 x 10-7
8 20 57* 35*
Silicalite on ceramic disc [24]
appr. 4.5/~ (5 ~lm)
20 20 20 20
H2/N2 H2/n-C4H10 N2/n-C4H10 N2/i-C4H10
2.3 x 10-7 2.3 x 10-7 2 x 10-7 2 x 10-7
3.1 146 15 55
Dense SiO2 by CVD dense (5 lirn) modification of microporous silica on alumina [25]
50 250 50 250 270
H2/N2 H2/N2 H2/CI-I4 H2/CH4 H2/N2
3 x 10-9 2.2 x 10-8 3 x 10-9 2.2 x 10-8
26 250 17.5 166 47*
Metal: Pd alloy on ceramic [26]
dense (6-8 ~tm)
440
H2/N2
1.6 x 10-6
>1000
Metal: P t / P d (80/20) on alumina [27]
dense (?)
100 200 300
H2/N2
1.2 x 10-7 3.6 x 10-7 7.5 x 10-7
5.6 37 200
1 The fastest permeating compound is mentioned first. 2 Permeation of fastest permeating compound. 3 If marked * then these figures are real separation factors. 9Separation factor is a function of the process variables and process circumstances.
I n o r g a n i c g a s s e p a r a t i o n m e m b r a n e s n o r m a l l y c o n s i s t of a s u b s t r a t e , o n w h i c h o n e or m o r e i n t e r m e d i a t e l a y e r s a n d a t o p l a y e r or g a s s e p a r a t i o n l a y e r h a v i n g K n u d s e n d i f f u s i o n s e l e c t i v i t y ( p o r e s of a b o u t 4 n m i n d i a m e t e r ) is a p p l i e d . T h e s e m e m b r a n e s w e r e d e v e l o p e d d u r i n g t h e l a s t t e n y e a r s i n m a i n l y flat a n d t u b u l a r c o n f i g u r a t i o n a n d b y u s i n g a l u m i n a as t h e b a s e m a t e r i a l . T h e y a r e
14 m APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
645
now available on a semi-commercial scale. Because of the low price of organic membranes and the rather low selectivity of inorganic Knudsen diffusion membranes it will be hard to find commercial applications for these membranes, unless they can be used under conditions where organic membranes would not be able to operate, e.g. high temperature or chemically harsh applications. The Knudsen diffusion gas separation layer can be modified by e.g. sol-gel, cvd, or crystallisation techniques to enhance the selectivity, but this decreases the permeation. Silica is the material mainly used for modification. However, data on reproducibility and stability are still scarce. The large scale use of high selective inorganic membranes and these membranes at high temperatures, up to at least 600~ will probably last another 5-10 years. On a laboratory scale (maximum membrane surface area of about 50 cm 2) these high selective membranes are now available, although stability can be a problem in certain atmospheres. Only a few years ago it was recognised that research and development should also be focused on high temperature gas tight sealing, membrane systems/modules and decreasing of costs by e.g. the increase of membrane surface area to volume ratio. On a laboratory scale membrane sealing technology is now available up to temperatures of about 600~ [28,29]. Some work has been reported on the increase of membrane surface area to volume ratio for ceramic gas separation membrane systems [30]. However, difficulties are foreseen in scaling up and controlling this technology [30,31].
14.2.2 Membrane Reactors
Besides the application of inorganic membranes in stand-alone gas separation units, attention is focused on more process-integrated applications. In such configurations the separation function of the membrane can be used to shift the equilibrium of a chemical reaction by selective removal of one or more components on the product side of the reaction in a so-called membrane reactor. Four basic catalytic membrane reactor configurations, when the membrane and reactor are in the same physical unit, can be distinguished [32]: - a catalytic membrane reactor (CMR), in which the membrane is permselective to one or more components and is catalytically active; a catalytic non-permselective membrane reactor, where the membrane acts as the catalyst, but is not selective to any of the components; a packed bed or fluidized bed membrane reactor (PBMR or FBMR), in which the selective membrane is surrounded by a packed bed or fluidized bed of catalyst particles; - a packed bed or fluidized bed catalytic membrane reactor (PBCMR or FBCMR), in which the selective and catalytic active membrane is also surrounded by a packed bed or fluidized bed of catalyst particles. -
-
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
646
CO2
CATALYST 9
FUEL
9
9
9
9
9
CO
"i
~
9
o
9 'qlml~._o .,@.
@
9
H20
~e -
9
9
i
,O
q
9
9
9
9
9
9
9
9 9
9
9
9
9
9 9
9
9
9
9
9
9
@
9
9
9
9 9
9
9
9
%%'.~% .:. ';;, : i ~ " . "" = ~ ' ~ ~ : : , '
9
+
2
9
H2
9
21 9
~.
'!
9
9
9
9
9
9
9
9
o
9 9
9
9
9 9
9
9
9
9
9
O tOtO~I[O
9
9
9o . ;,"~
4 0
9
9
9
RICH STREAM
9
9
9 9
9
~'-.- -io9
9
9
9
~'~ C O
E E~T~ IN ~M
21 9 9
9
REACTION
9
+
9
H2
9
"'''''''!*-'"
(
RICH STREAM
9
9
9
9 9
9
9 9
9
9
9
9
_L_J 9
GAS
9 9
9
q 9B ' O 9 jl oil
e
, , ~ :,., , . - . - .9..'%". :~. - "o . . . . , ,9 9"j
f
R E A C T O R SHELL/ F i g . 14.2. S c h e m a t i c o f t h e w a t e r - g a s
shift membrane
reactor.
A schematic of a PBMR, in this case for the water-gas shift reaction, is given in Fig. 14.2. Of course the catalytic reactor and the membrane unit can also be separated from each other, but can still be used to enhance the yield of a catalytic process, as will be s h o w n in Section 14.3 (see also Fig. 14.5).
14.2.3 Membrane Process Modelling In the various feasibility studies presented in this chapter, models of membrane separation and membrane reactor systems play an important role. Models are being used for various reasons: not only because there is a lack of experimental data, or the calculations concern non-existing, ficfive membranes, they are also used to conveniently represent available data. In the various studies, different types of models have been used. However, the basis of all the models used is the same and will be discussed here. In a membrane permeator unit two important phenomena are encountered: transmembrane transport and flow around the membrane. In a membrane reactor a third p h e n o m e n o n is of importance: chemical reaction In the feasibility studies relatively simple models have been used because these concern mostly a rough estimate of the possibilities. For high selective and microporous types of membranes, permeation through the membrane is ass u m e d to occur only via diffusion which obeys Fick's law. In the case of the Knudsen diffusion membranes the contribution of the non-separating viscous flow through the membrane is also accounted for. The basis of the flow models are ordinary differential mass balances for each component on either side of the membrane. The mechanism for permeation is substituted in the mass balances. When reaction occurs the kinetic expression is also added to the balances. The chemical reaction is assumed to take place in
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
647
close proximity to the membrane surface (passive membrane reactor system), and not in the membrane pore itself (active membrane reactor system). The latter case will lead to a completely different and more complicated description of the phenomena. Further steady-state conditions are assumed, so the p a r a m e ters are time-independent. The model as used here is treated extensively in Refs. [29,57]. Further information can be found in Refs. [61,68-71]. In the membrane reactor m a n y parameters influence the performance of the system. By making the model equations (mass balances)dimensionless [61], parameters are grouped so that a few dimensionless groups appear which describe the process. The physical meaning and their definitions are given in Table 14.2. A kinetic expression of the power law type for the reaction rate is assumed. The P e number is an important parameter which has an influence on the performance of the membrane process. Permeation and surface area are coupled via the P e number. In the equation QH2 is the permeation of the fastest permeating component (usually H 2 in this study). In membrane gas separation processes P e is usually between 0.1 and 1.0. For new applications P e = 0.5 can be taken as a first guess. The actual performance of the systems depends on m a n y more parameters than the P e number only, i.e. membrane selectivity, pressure drop, sweep gas flow to feed gas flow ratio, composition of the feed. TABLE 14.2 Dimensionless numbers and their meaning Parameter
Meaning
Definition
Peclet number
Ratio of total feed rate and maximum possible transmembrane flow rate
Pe =
Ratioof maximum conversion and the total feed rate (dimensionless residence time)
Da =
Ratio of permeation of H 2 and component i (permselectivity)
Si= Qi-
Damk6hler number
Si
~)
Ratio of permeate and feed side pressure
~ot AmQH2Pf V(Pf)~ k ~ot(RT) ~
QH 2
pP
Molar ratio of total sweep gas flow and total F~ot feed flow Y=
G
648
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
For the purpose of two of the studies described in this chapter a membrane separation model based on the characteristics given above has been implemented in the flow sheeting package ASPEN PLUS TM. This package allows the use of self-made user sub-routines and is therefore suitable for the implementation of the membrane model. The advantage of the use of the flow sheeting package is that the sensitivity of the total system performance to changes in membrane parameters can be determined quickly and that optimum process configurations can be found more easily. 14.3 GAS
SEPARATION
APPLICATIONS
FOR INORGANIC
MEMBRANES
In this section some examples of inorganic gas separation membranes in membrane reactor applications will be discussed. A first indication of the technical and economic feasibility of these membranes in dehydrogenation reactions and in the water-gas shift reaction will be given.
14.3.1 Dehydrogenation of Propane This section is written in close cooperation with Kinetics Technology International B.V. in Zoetermeer and Holland Industrial Ceramics in Velsen-Noord, The Netherlands [33]. 14.3.1.1 Introduction
Steam cracker plants based on naphtha a n d / o r gas-oil feedstocks are the major source of locally produced propylene in Europe and the Far East. In the United States approximately 90% of propylene comes from steam crackers and refinery operations. The balance comes from catalytic dehydrogenation units. The growth rate of propylene use is expected to be 3-4% worldwide. With the more conventional sources of propylene such as steam cracker operations and refinery operations, it is not possible to supply sufficient propylene for this growing demand. However, at the price levels of mid 1993 the economics of propane dehydrogenation are not very attractive. In recent decades various processes have been developed for catalytic dehydrogenation of propane to propylene [34-37]. These processes can be divided into two groups: - processes with an adiabatic reactor concept, and - processes with an isothermal reactor concept. Current commercial processes for catalytic dehydrogenation of propane to propylene are based on adiabatic reactor systems. Typical examples are: - the Catofin process (Lummus/Air Products); - the Oleflex process (UOP)
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
649
the fluidized bed dehydrogenation process (FBD) (Snamprogetti/Yarsintez) Recent developments in catalytic dehydrogenation have led to nearly commercial processes, using an isothermal reactor concept. Examples are: - the STAR process (Phillips) [35,37]; - the LINDE process (Linde/BASF) [36]. The potential benefits which can be achieved by using ceramic membranes in comparison to conventional propane dehydrogenation processes such as Oleflex and Catofin will be discussed here. -
14.3.1.2 Thermodynamics of propane dehydrogenation Besides several side reactions, the following main endothermic reactions are of importance in the dehydrogenatiort of propane to propylene" C3H8 ~ C3H6 + H2
(14.6)
C3H 8 ~
(14.7)
C2H 4 + CH 4
By selectively removing hydrogen from the reaction mixture, the reaction can be shifted beyond the original thermodynamic equilibrium. In this way reaction limitations can be overcome and the propylene yield enhanced. In Fig. 14.3 the equilibrium conversion for these two reactions as a function of temperature is given. From this figure it is concluded that: reaction (7) is more
60 C 3 H 8 - > C2H4 + CH4
50 O
-9 40
o
30
X 2o
H6 + H2
m 10 0 200
300
400
500
600
Temperature ~ Fig. 14.3. Thermodynamic equilibrium at I bar.
700
650
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
675 ~ 4 O
"~ 3 650 ~
o 2
625 ~
0
100
200
300
400
500
Residence time (ms) Fig. 14.4. Influence of temperature on thermal cracking.
favoured from a t h e r m o d y n a m i c point of view; and for high propylene yields a high temperature is needed. Another important reaction which can take place at high temperatures is thermal cracking, which sets an upper limit to the reaction temperature. Therefore, a high-selective catalyst is necessary which only promotes the dehydrogenation and not the cracking reaction. The upper limit temperature of the cracking reaction has been determined by thermodynamic calculations using the p r o g r a m m e SPYRO | The results are given in Fig. 14.4 which shows that a temperature above 625-650~ leads to important thermal cracking reactions, which reduces the selectivity towards propylene, but also leads to increased coke formation, which deactivates the catalyst.
14.3.1.3 Adiabatic reactor concepts; reactor modelling evaluation In this modelling study only a packed bed m e m b r a n e reactor has been dealt with, because the regeneration of the catalyst and m e m b r a n e can be done separately, and also it will be easier to match the catalyst and m e m b r a n e surface necessary. Both Catofin and Oleflex use an adiabatic reactor concept. The Oleflex process uses four reactor beds in series, which as such is more suitable for addition of a ceramic m e m b r a n e separation unit than the Catofin process which uses a parallel reactor system. A comparison between the Oleflex process as a base case and an Oleflex process equipped with ceramic membranes is m a d e for the following cases:
14
-
-
651
APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
FeedPI"IReact~~---~modul 111.~IMembrane e l~,mo.du! e~ 2 = ~~Permeate....
IMembranemodul ~11 "=IReact~ .....e4,' .3,k,,
= IMembrane~1= !Reach~ 31~'=
,~Permeate
~Permeate
Fig. 14.5. Process flow diagram includhlg a membrane module after each reactor.
1. A 'Knudsen diffusion selective' m e m b r a n e after the first, second and third reactor (see Fig. 14.5). The permeation of the pure gases is inversely proportional to the square root of the molecular masses. 2. A 'Knudsen diffusion selective' m e m b r a n e after the third reactor only. 3. 'Ideal' membranes, which remove all the hydrogen formed in the reaction, after the first, second and third reactor. 4. 'Ideal' membranes, which remove all the h y d r o g e n formed in the reaction, after the first, second and third reactor and with increased outlet temperature. In cases 1 to 4 part of the reactor effluent is split off by the m e m b r a n e as permeate. The retentate stream, depleted in hydrogen, is then fed to the next reactor. After the fourth reactor m e m b r a n e permeate and reactor effluent are mixed again to be treated further in the d o w n s t r e a m section of the process. Permeation characteristics of 'Knudsen diffusion' membranes, consisting of a support and two consecutive layers, have been used to calculate the performance of the ceramic m e m b r a n e reactor, see also Section 14.2.1 [17,31]. The pore size of the separation layer of these membranes is 4 n m in diameter [31,38]. Ideal m e m b r a n e s which remove all the hydrogen formed do not exist (possible Pd-based membranes will come close to the required characteristics), but are used as a basis for calculating the m a x i m u m possible increase in conversion and selectivity. Two semi-quantitative models describing the reactor and m e m b r a n e performance were used to evaluate the overall performance. The reactor was modelled using the flow-sheeting package PRO II. A membrane model was used which describes both the transport through the membranes and transport along the membrane. These models are described in Ref. [33]. Based on the Oleflex process the following boundary conditions were chosen for the calculations: Plant capacity: Pressure after first reactor: Pressure after second reactor: Pressure after third reactor: Permeate pressure: Residence time in reactor: Permeate flow:
150,000 t / y e a r propylene 1.7 bar 1.5 bar 1.3 bar 1.1 bar 0.5 s 10% of feed flow
The results of the calculations are given in Table 14.3.
652
14-- APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES
TABLE 14.3 Adiabatic Oleflex based reactor performance (all figures on weight bases)
Conversion (%) Selectivity (%) Yield (%) Tin (~ Tout(~ Membrane area (m2)
Base case
Case 1 Case 2 Case 3 Case 4 Knudsen + Knudsen + Ideal + reactors Ideal + reactors reactors 1,2,3 reactor 3 1,2,3, constant T 1,2,3,higher temp.
47.0 73.9 34.7 650 595 -
42.2 74.0 31.2 650 594 475
46.1 74.0 34.2 650 595 313
49.4 75.7 37.4 650 583 -*
54.0 74.5 40.3 685 595 -*
* Membrane area not calculated because no estimation of the permeation for ideal membranes has been made.
F r o m Table 14.3 it is clear that in process configurations w i t h K n u d s e n diffusion selective m e m b r a n e s a d r o p in yield is obtained, as c o m p a r e d w i t h the base case. A p p a r e n t l y , the use of K n u d s e n diffusion m e m b r a n e s u n d e r the chosen conditions in these configurations is not attractive d u e to the relatively large a m o u n t of p r o p a n e p e r m e a t i n g t h r o u g h the membrane. With 'ideal' m e m b r a n e s (Cases 3 and 4) positive effects are observed. In Case 3, w i t h the same heat input as in the base case, the increase in yield is limited. H o w e v e r , in Case 4, w i t h a higher inlet temperature, higher yields are obtained and m a x i m u m profit of the m e m b r a n e is made. F r o m this w e can conclude that: m e m b r a n e s w i t h a selectivity higher than K n u d s e n diffusion are needed; the process conditions s h o u l d be changed in order to increase the m e m b r a n e separation performance, and the d e h y d r o g e n a tion reaction kinetics seem fast e n o u g h to react on the h y d r o g e n removal, at the chosen residence time of 0.5 s. In order to increase the m e m b r a n e separation performance there are t w o possibilities: (1) increase the m e m b r a n e permselectivity, to values higher than for K n u d sen diffusion; (2) increase the driving force for separation across the m e m b r a n e (a higher d r i v i n g force for separation means a m e m b r a n e process w i t h higher separation factors, at the same m e m b r a n e permselectivity) by: increasing the feed pressure, increasing the a m o u n t of h y d r o g e n in the feed; using a sweep gas at the p e r m e a t e side; reducing the permeate pressure. Since the driving force for h y d r o g e n transport is low, a m o d e r a t e increase in m e m b r a n e selectivity (to a permselectivity of 10 for H 2 vs C3H 8 a n d C3H6) has s h o w n to have h a r d l y any influence on the performance. Furthermore, the
14 m APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
653
membrane surface needed will increase because it is assumed that the increase in selectivity is obtained by a decrease in permeability of all the components, except hydrogen whose permeability is constant. An increase in membrane surface means an increase in costs. An increase in feed gas pressure is not attractive since the reaction conversion drops significantly with an increase in the feed gas pressure. As can be expected, the use of extra hydrogen in the feed has a negative effect on the conversion. Potential sweep gas candidates are steam and propane. Unfortunately steam will permeate in reverse through the membrane (when the membrane has a rather low selectivity) and deactivate the catalyst in the next reactor. Propane as a sweep leads to a significant change in the hydrogen and propane quantities in the retentate. The propane recycle which is required to use propane as a sweep gas, leads to a significant increase in utility consumption. Finally, we have calculated the effect of permeate pressure reduction. In Table 14.4 the results for the base case and Case 2 are compared with a new case 5, in which Knudsen diffusion membranes have been used only after the third reactor and in a process having a permeate pressure of 0.3 bar. Permeate pressure reduction leads only to a marginal improvement in yield. The only possibility of using inorganic membranes in an adiabatic reactor concept for dehydrogenation of propane is to use membranes with a selectivity much higher than Knudsen diffusion, in combination with a reduced permeate pressure. In this case, hardly any reactant will be lost through the membrane and the driving force for hydrogen transport will be high enough. Results of calculations for this combination will be reported in future. Possibilities for the use of inorganic membranes in an isothermal concept may lead more easily to a technically feasible process, because extra heat for propane conversion is available. Detailed flow sheeting calculations for the TABLE 14.4. Adiabatic Oleflex based reactor performance (all figures on weight bases) Base case
Case 2 K n u d s e n + reactor 3
Case 5 K n u d s e n + reactor 3
Conversion (%)
47.0
46.1
46.6
Selectivity (%)
73.9
74.0
74.0
Yield (%)
34.7
34.2
34.5
Tin (~
650
650
650 593
Tout (~ M e m b r a n e area (m 2)
595 -
595 313
Feed pressure (bar)
-
1.3
1.3
Perm. pressure (bar)
-
1.1
0.3
60
654
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N MEMBRANES
integrated process are not yet available. However, to obtain a first indication of the economic feasibility of this concept, laboratory-scale membrane data are being used for performance estimation, see Section 14.3.1.4. In general, membranes to be applied should be stable under working conditions. Also, coke formation on the membranes should not lead to dramatic reduction of permeation and selectivity and regeneration with steam should not be a problem [39]. Another technical constraint can be the connection of membranes to the metal housing.
14.3.1.4 Isothermal reactor concepts; economic evaluation An isothermal reactor concept incorporating a ceramic membrane is more attractive compared to an adiabatic reactor concept from a thermodynamic point of view. In this concept we assumed a reactor with reactor tubes located in a direct-fired heater and operated in a cyclic way to remove coke formed on the catalyst. Parallel bed and heaters have been assumed [35-37]. On behalf of KTI an experimental programme on these reactor concepts has been started at the University of Southern California (USC). Some of the experimental results, concerning the use of Knudsen diffusion membranes are available in the literature [32,40]. These data have been used to calculate the economics of an isothermal propane dehydrogenation membrane reactor concept and are compared with the commercial Oleflex and Catofin processes, based on an adiabatic concept. The experimental circumstances of these lab-scale experiments, especially residence time, pressures and gas composition are not the same as in commercial, large-scale processes. However, we do not expect these differences to have a great influence on the results of the work presented here. Two process flow diagrams have been developed for a ceramic membrane reactor process: - the CMRL process: a process based on the commercial Oleflex process with a low propane conversion and Knudsen diffusion membranes - the CMRH process: a process based on the commercial Catofin process with a high propane conversion and Knudsen diffusion membranes The operating characteristics of these processes are given in Table 14.5. The design capacity of the plant is 150,000 M T / y e a r polymer grade propylene, which is equivalent to a production of 18,750 k g / h (8,000 h/year). The basis of the economic evaluation is the comparison of operating and investment costs for a membrane reactor with those for a conventional dehydrogenation plant. The return on investment (ROI) and the propylene production costs of the different processes have been calculated. The results are summarised in Table 14.6. Details of the calculations are reported in Ref. [33]. In the calculations a propane price of 130 $/tonne and a propylene price of 330 $ / t o n n e has been assumed [33].
14-- APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES
655
TABLE 14.5 Operating characteristics
Reactor type Conversion (%) Selectivity (wt%) H2/feed ratio Pressure (bar) Inlet temp. (~ Outlet temp. (~ LHSV (h-1) Reactor volume (m3)
Oleflex [34]
Catofin
CMRL [32,33,40] CMRH [32,33,40]
adiabatic 35.0 77.0 0.8 1.6-1.1 625 520-580 2.5 55.7
adiabatic 54.9 69.4 0.0 0.5 650 520-600 1.2 82.0
isothermal 38.0 89.0 0.2 1.15 560 560 2.5 44.4
isothermal 53.4 78.6 0.2 1.15 580 580 2.5 36.1
TABLE 14.6 Production cost breakdown (in US$)
Propane feed Co-product credit Utilities, catalyst, chemicals Fixed expenses Full production costs Dep recia tion Accounting production costs Selling price Overall margin ROI resulting from overall margin (%)
Catofin
Oleflex
CMRL
CMRH
203 -55.8 60.3 41.1 248.7 66.2 314.9 330 15.1 1.5
183.0 -47.3 66.8 42.7 245.2 69.2 314.4 330 15.6 1.4
171.6 -28.8 67.1 37.6 247.5 59.8 307.3 330 22.7 2.4
177.9 --44.3 59.3 37.2 230.1 59.0 289.1 330 40.9 4.4
F r o m Table 14.6 it can be seen that Catofin a n d Oleflex give a b o u t the s a m e ROI. This ROI is not v e r y attractive. The C M R L gives a n ROI of a b o u t 2.4%. The C M R H case gives an absolute increase in ROI of 3% p o i n t s c o m p a r e d to c o m m e r c i a l adiabatic processes. A sensitivity analysis of the ROI on b o t h the feed costs a n d the p r o d u c t v a l u e s is p e r f o r m e d . For the Oleflex a n d C M R H case these results are s u m m a rised in Figs. 14.6 a n d 14.7 w h i c h indicate that the ROI of s u c h a p r o p a n e d e h y d r o g e n a t i o n unit is not attractive w h e n the price difference b e t w e e n prop a n e a n d p r o p y l e n e is less t h a n a b o u t 250-300 $ / t o n n e . At m i d 1993, price levels of 330 $ / t o n n e p r o p y l e n e a n d 130 $ / t o n n e p r o p a n e , the process is n o t e c o n o m i c a l l y viable. Historical price levels s h o w t h a t a price difference of 300
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
656
11
$ / ton
,' . . . .
10~-
110
9P
120
8
130 140
6-
5~ 4~ 3r 2 ~10 -1
C3H8 Price I
'
-2
,
-3 ~~ "i
.5 [,
,
9
.
0
.
|
9
290 300 3~0 a20 330 340 3so 3;0 370 380 390 C3H6 Price [$1tonj
Fig. 14.6. Influence of propane/propylene price on Oleflex return on investment.
$/ton
15 14 13 12 11 10 9 8 7
110 120 13o 1 o
5 4 3 2 1 0 -1 -2 290
300
310
320
330
340
350
360
370
380
390
C 3 H 6 Price IS/toni
Fig. 14.7. h l f l u e n c e o f p r o p a n e / p r o p y l e n e
price on CMRH return on investment.
$ / t o n n e has not been encountered d u r i n g the last 3 years. It is concluded that a ceramic m e m b r a n e reactor based on K n u d s e n diffusion m e m b r a n e s can give i m p r o v e m e n t s in an isothermal reactor concept a l t h o u g h the difference in price level b e t w e e n feedstock and p r o d u c t is too small to give an economically viable process.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
657
Following the results of the adiabatic reactor concept it is expected that high selective membranes will further improve the economics. However, it should be recognised that the process conditions in an isothermal concept are more severe than in an adiabatic concept. In particular, decoking conditions can be a problem in using high selective membranes. Detailed calculations on the isothermal membrane reactor concept are being performed and will be reported in future.
14.3.1.5 General conclusions propane dehydrogenation The selectivity of Knudsen diffusion membranes is not high enough to give a technically and economically feasible ceramic membrane reactor process for the dehydrogenation of propane to propylene based upon an adiabatic reactor concept. Measures such as an increased driving force or a moderately increased selectivity do not lead to positive results, because the driving force for hydrogen separation under the chosen process conditions is not high enough. Probably the only possibility is the combination of a high driving force (sweep gas or low permeate pressure) and a very high selective membrane. The use of ceramic membranes in an isothermal reactor concept shows better prospects. This process, in combination with high selective membranes and the necessary membrane boundary conditions are being studied, and the results will be reported in future. Propane and propylene prices are the main actors in the introduction of a dehydrogenation process in general, thus also for processes based upon membrane reactors. At a price difference (propylene-propane) of 300 $/tonne or less membrane based dehydrogenation processes will hardly be economic feasible.
14.3.2 Dehydrogenation of Ethylbenzene to Styrene 14.3.2.1 Introduction Next to ethylene, propylene and vinylchloride, styrene is one of the most important monomers for the production of plastics. The worldwide demand for styrene in 1992 was 18.2 million tonnes and is expected to grow annually with 3-5% to 23.9 million tons in 2000 [42]. Recent production statistics show an annual production of about 1.3 million tons of styrene in the Netherlands. Approximately 75% of this is produced at DOW Benelux in Terneuzen by catalytic adiabatic dehydrogenation of ethylbenzene [42]. The conversion of the endothermic reaction by which styrene is produced from ethylbenzene is mainly limited by temperature and thermodynamic equilibrium. The conversion to styrene increases with temperature, decreases with pressure and with dilution of an inert component like steam.
658
14 APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES -
H
-
H
H\
I
~
I
/H
./O~C\
H " - - C . - - C "--H
I H -
+
Fig. 14.8. The dehydrogenation of ethylbenzene to styrene.
When producing styrene from ethylbenzene several reactions besides the main reaction take place. Six reactions are of importance; these include the production of toluene, benzene, ethylene and methane and the thermal cracking of ethylbenzene (coking) [43]. This last reaction is the main reason for the upper temperature limit of 630~ On the other hand, high temperatures favour the dehydrogenation reaction, so the process takes place between approximately 570 and 630~ The dehydrogenation reaction is presented in Fig. 14.8. As with the dehydrogenation of propane, removing hydrogen from the reaction mixture may shift the conversion beyond the reaction equilibrium to the product side, obtaining higher selectivities to and yields of styrene. In the literature several experiments and some modelling results are presented about the possibilities of membrane reactors in the dehydrogenation of ethylbenzene. The results vary from a small increase in yield and selectivity [39,44] to very large increases in yield up to 20% [45--49]. In this study the feasibility of implementing ceramic membranes on an industrial scale in the styrene production process is treated. Therefore, a model has been set up in the flowsheeting package ASPEN PLUSTM,which describes a styrene process production plant. Some modelling has been done with different types of membrane reactors in different reactor section configurations to investigate the influence on the performance of the production of styrene.
14.3.2.2 Conventional process description This work focuses on the reactor section of the styrene production process because it is the most promising part for the implementation of membranes. The reactor section of this process is shown in Fig. 14.9 [50]. The process uses two radial reactors in series with one preheater and one interstage heater. Steam is used as an energy carrier (adiabatic reactor) and diluent [43,50,51]. Reactor temperatures and pressures are 570--630~ and 1.5 bar, respectively. Total hydrocarbon mass flow (96 wt% ethylbenzene) is 95,000 kg/h. The steam/hydrocarbon ratio is 2. Typical conversion, selectivity and yield numbers are 71, 92 and 66%, respectively. Definitions are given in the appendix. Reaction equations and kinetics are taken from literature [43,51].
14 - - APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
659 Raw styrene
Ethylbenzene
A
F e e ~ @ ~
C = Heat exchanger
B Fig. 14.9. Reactor section of the styrene production process.
14.3.2.3 Implementation of membranes The packed bed ceramic membrane reactor configuration (PBMR) has been chosen as the reactor set-up (see Section 14.2.2). In the PBMR configuration three possible sub-configurations can be envisioned for a specific sweep gas in combination with a hydrogen or oxygen selective membrane for the dehydrogenation of ethylbenzene. These sub-configurations are shown in Fig. 14.10. In sub-configuration (A) hydrogen will permeate through the hydrogen selective membrane tube under the influence of a pressure difference over the membrane and it will be carried away with an inert sweep gas (steam). The partial pressure of hydrogen in the reaction mixture will decrease and the equilibrium will shift to the product side. In sub-configuration (B) the permeated hydrogen will be swept away with air. Hydrogen will be burned and the heat generated by this exothermic reaction flows through the membrane to the reaction mixture. In this way the reactor will get an isothermic character and therefore higher conversions. The third sub-configuration (C) uses oxygen permeable membranes instead of hydrogen permeable membranes. Again air is used as an oxygen source in Catalyst
Sub-conflgumtlon A
Sub-configumtlon B .
,
Steam
~,\
'
Fer
o
~ o
Steam___~,.
o
~
/(
' o o o~
9
/
o o P~du~ o
"~'12
i
o
-~ -~
Air
Fe(~l
Air
.
.
.
.
.
.
.
.
.
.
'
' '
i... o
~'-
o o
~
~
"o o o
o o P[oduct o
o'-P
~LI2 ill
Sul~conflguratlon C
Air
Feed I ~ ~ 9
o
o" o~
_oo~o~ Oo
Air
02
i
/
i
i
i
i
i
/ ' Wall Permeate side
Fig. 14.10. Membrane reactor sub-configurations.
....
660
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
the annular space of the reactor. Now oxygen will permeate through the membrane into the reaction mixture were it will burn the hydrogen formed. In this way there will be less loss of heat compared to the second case. Sub-configuration (C) is a principally different process: oxidative dehydrogenation. The most important disadvantage of the last two configurations is that not only will hydrogen be burned, but also hydrocarbons such as styrene and ethylbenzene. It is assumed that in sub-configuration (B) the membranes do not have an infinite hydrogen selectivity. For this reason we have chosen to focus our investigation on the first reactor sub-configuration (A). Although, especially sub-configuration (C), the oxidative dehydrogenation process seems very promising if a catalyst active only for H 2 oxidation (and not CxHy oxidation) is developed. By implementing the membrane reactor in the process according to sub-configuration (A), the conventional process is changed as little as possible. Again several configurations for the reactor section are possible. The first is to implement membranes in the first reactor (see Fig. 14.9) and leave the rest of the process intact. The second possibility is to implement the membranes in the second reactor and leave the first intact. Another option is to leave both reactors intact and to implement membranes between the two reactors. Because of the expected high costs of ceramic membranes, implementation in both reactors will probably be too expensive in relation to the possible advantages. For modelling the styrene process in ASPEN PLUS TM, several assumptions have been made: the radial flow reactors are estimated by plug flow reactors; all reactions are catalytic and only the main reaction is reversible; under these conditions coking is negligible; - the pressure drop in the reactors is negligible; heat transfer through the membrane is posed ideal; the sweep gas flows concurrently with the reactant gas; a small part of the steam which is originally used as carrier gas and heat carrier, is now used as sweep gas. The total amount of steam used, stays the same; the standard pressure at the permeate side of the membrane reactor is 0.1 bar. In modelling we used (if possible) permeations and selectivities based upon real measurements on different types of membranes [17,26,27,29,31,38] (see also Section 14.2.1). These membranes are: - Knudsen diffusion membranes, as developed by ECN; - microporous, highly selective membranes, as developed by ECN; and - palladium membranes [26,27]. We also formulated a non-existing, hypothetical membrane to see what would be ultimately possible. The performance of these membranes is in -
-
-
-
-
-
-
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
661
TABLE 14.7 Permeations and permselectivities of the membranes used Knudsen
H2 Ethylbenzene Styrene Toluene Benzene H20
Microporous
Palladium
Hypothetical
Q*
S*
Q*
S*
Q*
S*
Q*
S*
13.10-6 1.8.10-6 1.9.10-6 2.0.10-6 2.2.10-6 4.6.10-6
1 7 6.9 6.5 5.9 2.8
1.10-6 2.10-8 2.10-8 2.10-8 2.10-8 1.10-7.
1 50 50 50 50 10
0.8.10-6 5.10-9 5.10-9 5.10-9 5.10-9 8.10-9
1 160 160 160 160 100
1.10-6 2.10-9 2.10-9 2.10-9 2.10-9 1.10-8
1 500 500 500 500 100
*Q = Permeation (mol/m 2 s Pa) and S = permselectivities (PermH2/Permx). principle the same as of m i c r o p o r o u s membranes. The only difference is that the permselectivity of h y d r o g e n in relation to the other c o m p o n e n t s is ten times higher. The p e r m e a t i o n of h y d r o g e n however, remains the same [29,31]. In Table 14.7 the p e r m e a t i o n s and permselectivities are given for the chosen m e m b r a n e types. These data have been used to perform the modelling. In this investigation Pe = 0.5 is taken as a s t a n d a r d for calculations (see Section 14.2.3). No further optimisation t o w a r d s m e m b r a n e surface area has been carried out.
14.3.2.4 Results With Pe = 0.5, it has been calculated that u n d e r the chosen conditions in all configurations of the reactor section a m e m b r a n e surface area of a p p r o x i m a t e l y 43,000 m 2 is required for m i c r o p o r o u s and p a l l a d i u m m e m b r a n e s and 3,300 m 2 for K n u d s e n diffusion membranes.
Sub-configuration (A) Results of the i m p l e m e n t a t i o n of all four types of m e m b r a n e s in only the first reactor (PBMR) are given in Table 14.8. I m p l e m e n t a t i o n of these m e m b r a n e s decreases the performance of the reactor because: a part of the steam that is used for dilution and energy carrier in the conventional m o d e l is n o w used as sweep gas; less dilution and e n e r g y i n p u t has a negative effect on the d e h y d r o g e n a t i o n ; in this early stage not m u c h h y d r o g e n has been f o r m e d that can be t r a n s p o r t e d t h r o u g h the membrane; and - the high partial pressure of ethylbenzene will enhance the p e r m e a t i o n of this reactant t h r o u g h the m e m b r a n e w i t h the consequence that there is less ethylbenzene left to react to styrene. -
-
662
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
TABLE 14.8
Results of the simulations with sub-configuration (A) Membrane
Implementation of membranes in the first reactor Yield (%)
Conversion (%)
Selectivity (%)
N o membrane
43
46
95
Knudsen Microporous Palladium Hypothetical
38
40
94
40
42
95
40
42
95
40
43
95
Yield of styrene 3.67
=,,
,,,.
,,,,
.,,
..
..
..,
.=
=,,,
,,=
,4.. 0.65 >-
0.63 0.1
0,3
0.5
0.7
0.9
1.1
P e r m e a t e p r e s s u r e (bar]
',,-~Convent~nal
reactor ~ - Memloran~reactor,
}
Selectivity to styrene 0.93.=,
>,
0.93
..,= ~ O,925
0.92
,.
.
.
_--
.
.
~
.
.
~
.
.
.
,-
.
.
.,
.
.
~
.
.
.
-.
.
.
~
~
I 0.1
0.3
0.5
0.7
0,9
,1
P e r m e a t e pressure (bar] ': - ~ - Conventzonal reactor ~
Mernl0ranereactor
']
Fig. 14.11. Yield and selectivity as a function of the permeate pressure. For m e m b r a n e s i m p l e m e n t e d in the second reactor only the results of microporous m e m b r a n e s will be discussed in detail, because p a l l a d i u m m e m b r a n e s gave almost the same results and the performance is better than that of K n u d s e n diffusion membranes. The yield and selectivities have been calculated at different permeate pressures and are plotted in Fig. 14.11. The results for the conventional reactor are obtained w i t h o u t a m e m b r a n e i m p l e m e n t e d in the process.
14 B APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION M E M B R A N E S
663
With decreasing permeate pressures, the yield increases to the same level as in the conventional reactor and the selectivity increases to a higher level. In this stage of the process the reaction approaches the equilibrium closer than in the first reactor, so the negative effect of permeating ethylbenzene is less and the positive effect of hydrogen permeation is larger. The increase in yield, with decreasing permeate pressure is due to the suppression of the hydrogenation reaction in which styrene reacts to ethylbenzene. The increase in selectivity to styrene is explained by the suppression of the side reactions to, e.g., toluene and benzene, due to the lower partial pressure of hydrogen. Thus, implementation of membranes leads to the expected effects, but these are too small to compensate the negative effect of less steam in the reactant gas (partly used in this case as sweep gas). The results from simulations with a membrane unit placed between the two conventional reactors are comparable to the above-mentioned results, because the average partial pressures are almost the same. The simulation results are not as promising as expected beforehand and not as good as those reported in literature [45-49]. To find an explanation for our results, we first simulated the implementation of membranes having a permselectivity which is thousand times higher than that for microporous membranes, but which has the same hydrogen permeability. To create an even more ideal environment for extremely selective removal of hydrogen from the reactant gas, the permeate pressure has been set to 0.005 bar. Although the hydrogen partial pressure in the reactant gas was negligible (meaning that the hydrogen transport through the membrane was as large as possible), the increase in yield and selectivity was no more than 2 percentage points. However, a large increase was expected under these conditions. From this it can be concluded that the application of membranes under these circumstances will not lead to an improvement in the performance of the process. In order to explain this, the influence of the kinetics of the main reaction on the performance of the membrane reactor has been studied, for microporous membranes implemented in the second reactor. The reaction rate of the main reaction is successively multiplied by a factor 2 and 10, and as a consequence the reaction equilibrium is reached much faster. Under these circumstances increases are found in both yield and selectivity for the conventional dehydrogenation reactor without membranes. The results of the calculations are presented in Table 14.9 in which the differences in yield and conversion are given in percentage pointswith respect to the conventional case. The higher yields and conversions for the PBMR compared to the conventional reactor are due to the fact that the conversion is no longer limited by the kinetics, as in the previous cases, but by the permeation of hydrogen. It can be concluded that, if the reaction rate of the dehydrogenation process is faster than in current processes, implementation of microporous membranes
664
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
TABLE 14.9 Yield and selectivity in a PBMR as a function of reaction kinetics, compared to those in a conventional reactor Reaction rate
Conventional
2 times faster
10 times faster
Yield
_+0 % point
+5 % point
+12 % point
Selectivity
+ 1 % point
+1.5 % point
+2 % point
will give a significant improvement in yield and selectivity. The ratio of permeation and reaction rate is very important when selecting membranes for this application. This is comparable to that reported for the cyclohexane dehydrogenation process [52]. 14.3.2.5 Discussion
The results found in this study are less promising then those reported in literature [45-49]. There are several reasons for this difference. In some publications experiments have been reported in which process conditions a n d / o r feed compositions have been used that are not realistic or feasible on an industrial scale but do have a large impact on the performance of the membrane reactor. Also, when results are reported from modelling this process, incorrect assumptions were sometimes made, e.g. side-reactions which have a large influence on the performance of this process have been neglected [47]. In other publications a very large heat input is taken, which leads to a more or less isothermal reactor, and as a consequence to higher conversions [45,46,48]. Even if implementation of membranes on technical basis is possible, one of the major problems of implementing ceramic membranes in this styrene production process is the enormous membrane surface area required, which does not appear economically viable. Other membrane reactor configurations with a higher surface area to volume ratio may reduce the required module volume and simplify the installation. The biggest problem, however, is that these large membrane surface areas lead to very high costs. We assumed that in about ten years installation of one square metre of microporous gas separation membranes will cost about as much as $ 5,700. In the case treated, a surface area of 43,000 m 2 of microporous membranes is required, which will cost a total of 250 million dollars. The possible gain is 1% in selectivity, which gives an annual profit of 4.5 million dollars. Even when the reaction rate of the main reaction is enhanced by a factor ten, the possible profit of 9 million dollars does not lead to an interesting ROI. It should be realized that extra costs for, e.g., process integration are not yet included.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
665
14.3.2.6 Conclusions
In the present concept of styrene dehydrogenation implementation of inorganic membranes is not feasible. Application of Knudsen diffusion membranes with a low permselectivity to hydrogen leads to a considerable permeation of ethylbenzene and thus, to lower yields. Microporous and palladium membranes give better results, but worse than a conventional case, because the conversion is limited by reaction kinetics. The ratio of permeation rate to reaction rate is very important in selecting membranes in a membrane reactor process in which equilibrium shift is foreseen. From the simulations with higher reaction rates it follows that implementation of ceramic membranes can lead to higher yields and selectivities. However, even under these conditions the profit from extra styrene yield does not compensate the costs of the membranes. For profitable implementation of inorganic membranes, a high-selective membrane with a higher permeability than the membranes now available is necessary, in combination with higher reaction rates. 14.3.3 Water-Gas Shift Membrane Reactor 14.3.3.1 Introduction
The water-gas shift (WGS) reaction is an important reaction in many commercial processes where hydrogen has to be generated or where CO must be converted. In the WGS reaction carbon monoxide together with steam is converted to carbon dioxide and hydrogen. The reaction is a reversible chemical reaction, usually assisted by a catalyst (see Eq. (14.8)). CO + H20 ~
CO 2 +
H2
AH = -41.1 k J / m o l
(14.8)
Hydrogen is a very important gas in many areas of industry [50,53]. Currently, hydrocarbons are the main source for large-scale production of hydrogen [42,54]. Most of the hydrogen for industrial purposes is produced from natural gas and oil. The contribution of coal is limited at present but is expected to grow considerably in the future. In addition, biomass is expected to be a growing hydrogen source in the future. All these processes a n d / o r feedstocks produce a gas mixture containing mainly hydrogen and carbon monoxide (syn gas). The hydrogen and carbon monoxide concentration will vary depending on the feedstock and process. The WGS process is being used to adjust the syn gas compositions for further syntheses or to enhance the hydrogen content for hydrogen production. Coal particularly is thought to play a major role in future world energy supplies and possibly also as a feedstock supplier for the chemical industry [54].
666
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
Through integrated coal gasification combined cycle (IGCC) power plants, coal can be converted in a clean and efficient way into electricity with syn gas as an intermediate product. In future IGCC options, syn gas can be partially converted into secondary gaseous products and be partially used to generate power. Hydrogen is an obvious secondary product for such a system [55]. A possible problem in future energy generation from coal is the emission of large quantities of CO2. The rapidly increasing concentration of greenhouse gases in the atmosphere has already triggered the development of clean coal technologies for power generation worldwide. In the long term even the introduction of the highly efficient IGCC systems might not be sufficient to ensure the use of coal for power production, and further measures to decrease the emission of greenhouse gases, CO2 in particular, might be necessary. The options to do so in an IGCC system also compare favourably with other large-scale coal-based combined cycles. The generation of hydrogen (WGS process) also plays an important role here [56,57]. The attainable conversion with the WGS reaction depends on how the chemical equilibrium is set. The equilibrium constant Kp decreases as the temperature increases. This implies that the CO conversion decreases with increasing temperature. In many cases a high hydrogen yield is the objective of the WGS application. Increased hydrogen yield and reduced carbon monoxide content can be obtained in several ways. In principle it is desirable to carry out the reaction at low temperatures. This can be achieved by: (1) cooling during the reaction by heat transfer, or inert gas addition; 2) intercooling through execution of the reaction in several steps (reactors). Other methods to increase the level of conversion are: (3) increasing the steam to carbon monoxide ratio; (4) forcing equilibrium displacement to the product side; this should be achieved by continuous removal of either hydrogen or carbon dioxide directly at the place where it is formed. Generally, in a conventional WGS system a two-step shift is used to obtain high CO conversion rates. In the first high-temperature shift reactor the major part of the CO is converted at high activity, whereas in the second shift reactor the rest of the CO (closely up to the thermodynamic equilibrium) is converted at low temperature and also low activity. Steam to carbon monoxide ratios above the stoichiometric ratio (higher than 2) are generally being used to attain the desired carbon monoxide conversion, but also to suppress carbon formation on certain catalysts. Different types of catalysts exhibit shift activity. Commercially available high temperature (HT) FeCr and low temperature (LT) CuZn are generally employed. Among some interesting new developments is the Co/Mo based catalyst. This type is completely insensitive to sulphur and certain formulations are claimed to possess good activity at both high and low temperatures. The
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
667
steam/carbon ratio is set by equilibrium considerations and carbon formation suppression. Of the methods to increase the CO conversion mentioned, the first three possibilities are accompanied by severe penalties with respect to energy use, exploitation and investment costs. The fourth possibility seems less affected by such drawbacks and is therefore preferred [57]. Inorganic gas separation membranes with their unique properties can be used to selectively remove hydrogen in a membrane reactor. A schematic of the combination of membranes and the WGS reaction has already been shown in Fig. 14.2 (Section 14.2.2). With such a PMBR reactor it is possible to enhance the CO conversion of the reaction and concurrently separate hydrogen from the reaction mixture, and furthermore have a separate CO2 rich stream. The membrane reactor replaces two unit operations, has an enhanced hydrogen yield and will save steam, and therefore has the potential of energy efficiency improvement. The hydrogen produced can either be sold as an end-product or consumed directly as feed stock in down-stream hydrogen consuming processes, in e.g. the petrochemical industry. A first step to explore the potential of a WGS inorganic membrane reactor is to assess its technical and economic feasibility. The potential and exact lay-out of such a reactor is thought to be strongly dependent on the upstream raw gas production and gas treatment processes and the respective downstream processes. Therefore the techno-economic feasibility of the application can only be judged after a detailed investigation of the performance of the reactor against the background of the specific detailed characteristics of the process chains envisaged. Some typical applications of the water-gas shift membrane reactor that are currently being foreseen have been very briefly mentioned in the Introduction. One specific application will be dealt with in greater detail to illustrate the assessment of the feasibility of the reactor system.
14.3.3.2 WGS membrane reactor for C02 emission control The potential of the WGS membrane reactor in CO2 control in IGCC installations has been studied in greater detail [57]. The possibilities of the reactor and demands set for the membranes have been determined by carefully assessing the process integration options, by experimental membrane characterisation and by using a membrane reactor model.
Process integration Various possible process flow schemes have been proposed. The C O 2 r e moval generally takes place from the coal gas [56]. Conventional approaches generally consist of a separate multistage water-gas shift (WGS) conversion of the fuel gas, followed by a low temperature CO2 removal process. Hydrogen is
668
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
l
Steam
IVlain c x ) m l o o r e ~ s :
Coal
P '
k Steam turbine
................................................................
, I~
Air
.I' ..............
y
Fig. 14.12. L a y o u t of an IGCC with CO2 control using a WGS m e m b r a n e reactor.
the only fuel component left after the WGS conversion and is fed to the gas turbine to convert to water only. However, the conversion of the WGS reaction is limited by its chemical equilibrium and the low temperature CO2 removal makes an additional cooling step necessary. Application of the membrane reactor can enhance the equilibrium production of H 2 from fuel gas and establish a separation between H 2 and CO2 at an elevated temperature. In Fig. 14.12 the layout of an IGCC with CO2 emission control is shown. The layout is similar to an ordinary IGCC except that the gas leaving the gas cleaning section is now fed to the membrane reactor. Nitrogen from the air separation unit is available as sweep gas for the reactor. Unconverted CO and unseparated hydrogen are catalytically burnt and the heat is utilized in the steam turbine. In this scheme CO2 is recovered separately from the other components and is ready available for e.g. disposal or re-use. The availability of sweep gas (02 blown gasifier), the low requirements of the purity of the product streams, as well as the high pressure of the feed gas leaving the gasifier makes this process very favourable for the application of membranes as a separation technique. Hydrogen from the membrane reactor is converted in a gas turbine with a high efficiency. The process efficiency will increase when the hydrogen production (CO conversion) and recovery (on the permeate side) from the membrane reactor is raised. CO2 abatement increases with increasing recovery of carbon components on the retentate side of the membrane. The performance of the reactor can be measured in terms of these three parameters. The boundary conditions for the membrane reactor in the total system depends upon final
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
669
performance of, amongst others, the membrane reactor itself. For an initial insight, it is desirable to carry out sensitivity analyses with the membrane reactor. For this purpose the boundary conditions around the membrane reactor will be estimated in the first instance. Ceramic membranes Inorganic membrane development is still in progress [57] (see also Section 14.2.2). Microporous silica membranes have been developed at several universities and research institutes. Membrane selectivities of 15 and 20 for the separation of H 2 f r o m C O 2 have been reported. Even higher selectivities for H 2 arid CO, C H 4 and N 2 have been measured [20,57]. Most measurements reported in the literature have been performed on a laboratory scale. However, it has been shown that it is possible to upscale these microporous ceramic membranes to, at least, bench scale [31,57]. With other membranes such as noble (Pd) metal membranes and dense ceramic membranes very high and almost infinite selectivities for hydrogen are possible [58]. The permeation of these membranes is generally smaller than the permeation of microporous membranes. Microporous carbon membranes have been developed [59] but their possibilities in high temperature hydrogen separation are still unclear, although it is believed that there are opportunities. Scaling-up of these membranes seems possible from a technical point of view. All these membrane types are potentially suitable for application in the WGS membrane reactor concept, provided their endurance is sufficient. Results and discussion The initial parameters used for the membrane reactor sensitivity analysis are shown in Table 14.10. These parameters are a first guess of the boundary conditions of the total process. The conversion in the reactor is plotted in Fig. 14.13 against the Da number which can be regarded as a dimensionless residence time. From this plot it follows that the conversion in the membrane reactor equipped with high selective membranes can exceed the values possible with an ordinary plug flow reactor. From the graph it is clear that the conversion increases with increasing TABLE 14.10 Initial s i m u l a t i o n p a r a m e t e r s
Tf pf
= 623 K = 36 b a r
pP
= 21 b a r
Y
= 1.035
H20/CO
= 1.5 (after s t e a m injection)
670
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
1 O 0
"~
--= ........
=---: .......................
9. . . . . .
~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
~......................
.:. . . . . . . . . . . . . . . . . .
95
i_...a
~9
90
~
85 80
tt
S ~ = 15
0
5
10
15
20
Da
~~Plug
flow reactor ~ M e m b r a n e
reactor[
Fig. 14.13. Conversion vs. Da. ,..., 1oo ~- 95 ~ 90 ~ 0 o
85 80
~
75
~
70
"~
65 60 55 50, 0.1
D a - 12 S i - 15 0.3
0.5
0.7
0.9
1.1
Pe
{~CO
conversion ~
H2 recovery -=a,-C recovery I
Fig. 14.14. Conversion and recovery vs. Pe. Da as w o u l d be expected. Conversion in an o r d i n a r y reactor reaches a certain m a x i m u m d u e to the establishment of the chemical equilibrium. In a m e m b r a n e reactor conversion keeps increasing as a result of continuous h y d r o g e n permeation. This is an i m p o r t a n t aspect of the m e m b r a n e reactor, because the steam excess can be r e d u c e d which leads to a favourable e c o n o m y of the process. The influence of the Peclet n u m b e r is s h o w n in Fig. 14.14. Pe is reciprocally p r o p o r t i o n a l to the m e m b r a n e surface. Decreasing the Pe n u m b e r increases h y d r o g e n recovery and as a consequence the CO conversion. W h e n m o r e m e m b r a n e surface is available, also more carbon dioxide and carbon m o n o x i d e p e r m e a t e s t h r o u g h the m e m b r a n e and the carbon recovery decreases. A n i m p o r t a n t question for the application of m e m b r a n e s is w h a t the desired selectivity for the m e m b r a n e s has to be. The influence of the m e m b r a n e selec-
14 - - A P P L I C A T I O N
OF POROUS INORGANIC
GAS SEPARATION
671
MEMBRANES
,..., 100 ......a
90
o
80
~
70
.e
60
Da = ! 2 Pe =0.4
/ /
~" 50
! i i
4o 0
10
20
30
40
50
Si l. ~" CO conversion, ~
H2 recovery ~
C recovery t
Fig. 14.15. Conversion, recovery vs. selectivity 9 tivity on the performance of the reactor is s h o w n in Fig. 14.15. CO conversion a n d C recovery increase w i t h rising selectivity. The h y d r o g e n recovery is m o r e or less constant as a result of the Pe n u m b e r that is held constant for the various simulations. CO conversion rises slightly due to the decreased losses of reactants w i t h higher selectivity. Carbon recovery increases strongly with rising selectivity. T h o u g h the performance of the reactor increases w i t h rising selectivity, the incremental increase in conversion and especially the C-recovery decreases with g r o w i n g selectivity. The gain in performance is marginal w h e n the selectivity surpasses 40. The current m e a s u r e d H 2/CO2 permselectivity of scaled up m e m b r a n e s is 15 [57]. With this value a good recovery of both H 2 and CO2 is possible as appears from Fig. 14.15. For a p o w e r plant including a m e m b r a n e reactor with m e m b r a n e s w i t h a selectivity of 15 the efficiency of the total system has been d e t e r m i n e d t h r o u g h flow sheet calculations. In these calculations the requirements and the d e m a n d s of the m e m b r a n e reactor and the rest of the system m u s t match, so one or m o r e iterative calculations is necessary to optimise the total system. The results of the calculations after optimisation are presented in Table 14.11 in w h i c h three TABLE 14.11 Results of power plant efficiency calculations Process
Process efficiency (%)
IGCC with membrane reactor for CO2 removal IGCC with conventional CO2 removal IGCC without CO2 removal
42.8 40.5 46.7
672
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
systems are compared. Obviously a system without CO2 removal exhibits the highest efficiency. However, the efficiency penalty induced is significantly lower when CO2 is removed using the membrane reactor system instead of removal with a conventional technique. The conventional technique used here is a wet scrubbing process in which CO2 is absorbed on a physical sorbent. J
_
14.3.3.4 Full-scale process considerations
When the membranes are used on an industrial scale, a considerable amount of surface area will be necessary to process the gas stream involved. A typical surface area necessary is 1500 m 2 for a 300 MWe class power plant. For ceramic membranes this is a rather large surface area. Considering that permselectivity is already good for this application, it seems reasonable to direct research towards enlargement of the permeation or explore module concepts with a high surface area to volume ratio (e.g. monolytic systems) next to selectivity improvement. When membranes are produced in a tubular geometry, which seems the most feasible currently, all membranes have to be sealed separately. This favours tubes with large diameters to reduce the number of seals. On the other hand, the smaller the tube diameter, the higher the specific surface area attainable in a module. High pressure, high temperature membrane sealing is an important aspect of the full scale module and this hurdle has been taken for laboratory and bench scale [16,28,31,57]. The membranes can be sealed gas-tight to a stainless steel tube by a special joining technique. Experiments will be carried out initially for the so-called passive reactor concept in which a high selective membrane is surrounded by catalyst. Dead end tube configuration, in which only one end of the membrane tube is connected and the other end is closed [14], seems favourable since it needs one ceramic to metal joint less than two-side connected tubes. A drawback of this option is the large force that will act upon the dead end side of the membrane when the process works with a considerable pressure drop as in this application. These aspects show that it is important to realise for which application the membranes are being developed and to consider scaling up in an early stage. 14.3.3.5 Conclusion
Through membrane reactor model calculations it has been shown that membranes can enhance the conversion of a WGS membrane reactor and concurrently separate hydrogen from carbon dioxide. This system can be used to control the release of CO2 to the atmosphere from a IGCC power plant. Through process
14 w A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
673
flow sheet calculations it has been shown that the efficiency of CO2 control using the membrane reactor is significantly higher than when a conventional technique (i.e. wet washing with a sorbent) is applied. When selectivity of the membranes can be increased, it does not seem to be necessary to surpass approximately 40 for the process under consideration, because the gain in reactor performance seems marginal. Enlargement of the permeation is an important aspect on the other hand, so that the total surface area necessary for the full scale application can be reduced. This example shows that knowledge of the demands and requirements of the application are also very important in the development of membrane material.
14.4 CONCLUSIONS Three examples of the use and feasibility of inorganic membranes in reactor applications have been discussed. Although several references give a very positive indication on the technical possibilities of the use of inorganic membranes in reactor applications, it has been shown that measurements under realistic conditions and calculations involving the complete process can show the opposite. A multidisciplinary approach is needed to study the feasibility of inorganic membranes in (membrane reactor) applications. A combination of membrane and materials know-how and an insight into application opportunities and process economics is necessary to discuss the techno-economic feasibility of inorganic membranes. Furthermore, measurements and calculations should be performed on a realistic basis and scale in order to obtain reliable data on the performance of the membranes. As yet, insufficient realistic data on longterm membrane stability and coke formation on the membranes are available. In general the membrane reactor examples show that knowledge of the demands and requirements of each foreseen membrane application is very important in the choice and development of the membrane materials. Application of ceramic membranes can improve the return on investment in the propane dehydrogenation process. Probably the only possibility for a technically and economically feasible propane dehydrogenation process, able to enhance the ROI enough to make the investment worthwhile, is the combination of a high driving force (sweep gas or low permeate pressure) and a very high selective membrane. The isothermal reactor concept shows better prospects than the adiabatic concept. At a price difference smaller than 300 $/tonne between propylene and propane the propane dehydrogenation process based upon membranes will hardly be economically viable. The present concept of implementation of ceramic membranes in the styrene process is not feasible, because:
674
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
- the permselectivity to hydrogen of Knudsen membranes is too low, which leads to a significant loss of ethylbenzene from the reaction side which can no longer take part in the reaction; - the lower permeation of high selective microporous and palladium membranes leads to very high surface areas, which makes the application unattractive; - the reaction rate of the dehydrogenation of ethylbenzene to styrene is to low: the kinetics and not the permeation are the limiting step in the membrane reactor. From simulations with higher reaction rates it has been shown that implementation of ceramic membranes can lead to higher yields. However, even under these conditions the profit from the extra styrene yield does not compensate for the costs of the membranes. For profitable implementation of inorganic membranes, a high-selective membrane with a higher permeability than the membranes now available is necessary, in combination with higher reaction rates. It has been shown that membranes can enhance the conversion of a watergas shift membrane reactor and concurrently separate hydrogen from carbon dioxide. The efficiency of CO2 control using the membrane reactor with a H 2 / C O 2 selectivity of 15 is significantly higher compared to a conventional technique (i.e. wet washing with a sorbent). It is not necessary to exceed a selectivity of approximately 40 for H 2 / C O 2 for the process under consideration, because further increase in reactor performance seems marginal. Enlargement of the permeation is an important aspect on the other hand, so that the total surface area necessary for the full-scale application can be reduced. In all three applications discussed, the stability of the membranes in these high temperature processes and the design of suitable modules still needs much research and development.
Acknowledgements The research work described here has been funded in part by the Dutch Organization for Energy and Environment (NOVEM), the Dutch Ministry of Economic Affairs (EZ), and the Commission of the European Union. Kinetics Technology International BV is gratefully acknowledged for their helpful discussions and calculations. The authors would also like to express their thanks to Prof. R. Pruschek, Dr. G. Oeljeklaus and R. Kloster of the University of Essen, G. Haupt of Siemens AG Power Generation (KWU), Dr. H. van den Berg of Dow Benelux BV, and Dr. L. van der Ham of the University of Twente. K. Hemmes, G. Leendertse and E. Delnoij are gratefully thanked for their help in modelling and setting up the membrane models.
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
List of Symbols and Abbreviations
A Da F k
G
P Pe
Q R Si T V Y
membrane surface area (m 2) Damk6hler number flow rate (mol/s) reaction rate coefficient (mol/m3) 1-~ (l/s) equilibrium constant pressure (Pa) Peclet number permeation (mol/msPa) gas constant (J/molK) permselectivity (ratio H2/component i permeation) temperature (K) volume reactor (m 3) molar ratio sweep flow vs. feed flow sum of powers in power law expression ratio of permeate and feed side pressure
Superscripts
? m P s
Subscripts i
tot
feed membrane permeate sweep
component i total
Abbreviations
CMR CMRL CMRH FBCMR FBMR HT IGCC LHSV LT PBCMR PBMR ROI WGS
catalytic membrane reactor catalytic membrane reactor low conversion catalytic membrane reactor high conversion fluidized bed catalytic membrane reactor fluidized bed membrane reactor high temperature integrated coal gasification combined cycle liquid hourly space velocity (h-1) low temperature packed bed catalytic membrane reactor packed bed membrane reactor return on investment water-gas shift
675
676
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
APPENDIX
The p e r m e a t i o n is defined as the (pure gas) flow ( m o l / s ) t h r o u g h the m e m brane per surface area and per pressure difference over the m e m b r a n e . The permselectivity is defined as the ratio of the p e r m e a t i o n of p u r e gases. The separation factor is defined as" y 1-y
1- x x
(14.9)
in w h i c h y = concentration fastest p e r m e a t i n g c o m p o n e n t on p e r m e a t e side; x = concentration fastest p e r m e a t i n g c o m p o n e n t on feed side. The conversion a n d selectivity are given for the p r o p a n e d e h y d r o g e n a t i o n reaction. For the ethylbenzene d e h y d r o g e n a t i o n and w a t e r - g a s shift reaction the same definitions can be used for the respective r e a c t a n t s / p r o d u c t s . The conversion is defined as: conversion =
mass flow propane i n - mass flow propane out mass flow propane in (14.10)
The selectivity can be expressed on a molar basis or mass basis: select. (mol) =
tool flow propylene o u t - mol flow propylene in
select. (mass) =
mol flow propane i n - mol flow propane out (14.11) mass flow propylene o u t - mass flow propylene in mass flow propane i n - mass flow propane out (14.12)
The yield is defined as c o n v e r s i o n , selectivity, on mol or mass basis.
REFERENCES
1. 2. 3.
4.
5.
A. Crull, Prospects for the inorganic membrane business, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.) Proceedings ICIM2-91. Key Eng. Mater., 61/62 (1991) 279-288. Business Communications Co. Inc., Inorganic Membranes: Markets, Technologies, Players. BCC Report No. RGB-112R, Jan. 1994. H.E. Johnson and B.L. Schulman, Assessment of the Potential for Refinery Applications of Inorganic Membrane Technology ~ An Identification and Screening Analysis. DOE/FE61680-H3, Final report, May 1993. H.P. Hsieh, Inorganic Membranes. Am. Inst. Chem. Eng. Symposium Series, No. 261, Vol. 84, 1989, pp. 1-18. Ceramic Membranes ~ New Horizon in Membrane Separation. Emerging Technologies no.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C G A S S E P A R A T I O N M E M B R A N E S
677
36, Technical Insights, Inc., 1989. 6. B.Z. Egan, Using Inorganic Membranes to Separate Gases: R&D Status Review. Report ORNL/TM-11345, Nov. 1989. 7. A.J. Burggraaf, Key points in understanding and development of ceramic membranes, in: Y.H. Ma (Ed.), Proceedings of the 4th International Conference on Inorganic Membranes, July 10-14, Worcester, MA, USA, 1994, pp. 1-16. 8. A.J. Burggraaf, K. Keizer, R.S.A. de Lange, Z. Vroon and V.T. Zaspalis, Ceramic membranes for separation and reactions, in: R. Ballmoos et al. (Eds.), Proceedings of the 9th International Conference on Zeolites, Montreal, 1993, pp. 47-70. 9. J.N. Armor, Catalysis with permselective inorganic membranes. Appl. Catal., 49 (1989) 1-25. 10. H.P. Hsieh, Inorganic Membrane Reactors - A Review. Am. Inst. Chem. Eng. Symposium Series, No. 268, Vol.85, 1989, pp. 53-67. 11. H.P. Hsieh, Inorganic membrane reactors. Catal. Rev. Sci. Eng., 33 (1991) 1-70. 12. J. Shu, B.P.A. Grandjean, A. van Neste and S. Kaliaguine, Catalytic palladium-based membrane reactors: a review. Can. J. Chem. Eng., 69, (1991) 1036. 13. K.R. Westerterp, Multifunctional reactors. Chem. Eng. Sci., 47 (1992) 2195. 14. G. Saracco and V. Specchia, Catalytic inorganic membrane reactors: present experiences and future opportunities. Catal. Rev. Sci. Eng., 36 (1994) 305. 15. J. Zaman and A. Chakma, Review: inorganic membrane reactors. J. Membr. Sci., 92 (1994) 1. 16. G. Saracco, G.F. Versteeg and W.P.M. van Swaaij, Current Hurdles to the Success of High-Temperature Membrane Reactors. J. Membr. Sci., 95 (1994) 105. 17. H.M. van Veen, J.P.B.M. Tol, C.W.R. Engelen and H.J. Veringa, High temperature gas separation with alumina membranes, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) pp. 593-598. 18. Y. Shindo, K. Obata, T. Hakuta, H. Yoshitome, N. Todo, and J. Kato, Permeation of hydrogen through a porous vycor glass membrane. Adv. Hydrogen Energy, 2 (1981) 325. 19. A.S. Damle and S.K. Gangwal, Catalytic carbon membranes for hydrogen production, in: Proceedings of the lOth Annual Gasification and Gas Stream Cleanup Systems Contractors Review Meeting, Morgantown, 1990, pp. 322-329. 20. R.S.A de Lange, Microporous Sol-Gel Derived Ceramic Membranes for gas Separation Synthesis, Gas Transport and Separation Properties. Thesis, Twente University, 1993. 21. S. Kitao, H. Kameda, and M. Asaeda, Gas Separation by thin porous silica membrane of ultra fine pores at high temperature. Membrane (Maku), 15 (1990) 222. 22. D.L. Roberts, I.C. Abraham, Y. Blum and J.D. Way, Gas Separations Using Ceramic Membranes m Final Report. Report: DOE/MC/25204-3133, May 1992. 23. (a) J.E. Koresh and A. Softer, Molecular sieve carbon permselective membrane. Part I: Presentation of a new device for gas mixture separation. Sep. Sci. Technol., 18 (1983) 723. (b) J.E. Koresh and A. Softer, The carbon molecular sieve membranes. General properties and the permeability of CH4/H2 mixture. Sep. Sci. Technol., 22 (1987) 973. 24. M.D. Jia, K.V. Peinemann and R.D. Behling, Ceramic Zeolite Composite Membranes. Preparation, Characterization and gas Permeation. J. Membrane Sci., 82 (1993) 15. 25. S. Kitao and M. Asaeda, Gas separation performance of thin porous silica membrane prepared by sol-gel and cvd methods, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) 267-272. 26. Johnson Matthey Technology Centre, Thin Supported Pd-alloy Membranes for Hydrogen Purification. Report: ETSU F/02 / 00034 / REP, 1995.
678
14-- APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
27. M. Konno, M. Shindo, S. Sugawara and S. Saito, A composite palladium and porous aluminium oxide membrane for hydrogen separation. J. Membr. Sci., 37 (1988) 193. 28. (a) F.M. Velterop and K. Keizer, Development of high Temperature Resistant Module for Ceramic Membranes, in A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) 391-394. (b) F.M. Velterop, Ceramic membranes m technological implementation at high temperatures. Chem. Plants Process., Nov. (1993) 66. 29. M. Bracht, A. Bos, P.P.A.C. Pex, H.M. van Veen and P.T. Alderliesten, Water-gas shift membrane reactor, in: W.R. Bowen et al. (Eds.), Proceedings of Euromembrane "95, Bath. Vol. 2, 1995, pp. 425-430. 30. R.L. Goldsmith, Low-Cost Ceramic Membranes and Supports for Gas Separation. Report of Ceramem Corp, Waltham, USA. 31. P.P.A.C. Pex et al., Materials aspects of microporous inorganic gas separation membrane manufacturing, in: W.R. Bowen et al. (Eds.), Proceedings of Euromembrane "95, Bath, Vol. 1, 1995, pp. 295-300. 32. Z.D. Ziaka, R.G. Minet and T.T. Tsotsis, A high temperature catalytic membrane reactor for propane dehydrogenation. ]. Membr. Sci., 77 (1993) 221. 33. Technical and economic evaluation study for the use of ceramic membrane reactors for the dehydrogenation of propane to propylene. Confidential NOVEM Report No. 33105/0090, by KTI, ECN and HIC, Jan. 1994. 34. (a) P.R. Cottrell et al., UOP Technology Conference, 1992. (b) G.M. Horneday et al., Advances in Petroleum Chemistry and Refining, Vol. 4, Ch. 10, 1961, 482 pp. 35. P.R. Sarthy et al., Hydrocarbon Process., 43 (2) 1993. 36. H. Bolt et al, Dehydrogenation process for propane and isobutane. Presented at the AIChE Spring Meeting, Houston, Texas, April 7-11, 1991. 37. R.D. Dunn et al., The Phillips steam active reforming (STAR) process for the dehydrogenation of C3, C4 and C5 paraffins. Presented at the 1992 NPRA meeting, New Orleans, LA, March 22-24, 1992. 38. H.M. van Veen, R.A. Terpstra, J.P.B.M. Tol and H.J. Veringa, Three layer ceramic alumina membranes for high temperature gas separation applications, in: Proceedings of the First International Conference on Inorganic Membranes, Montpellier, 1989, pp. 329-335. 39. G.R. Gallaher, T.E. Gerdes and P.K.T. Liu, Experimental evaluation of dehydrogenations using catalytic membrane processes. Sep. Sci. Technol., 28, (1993) 309. 40. Z.D. Ziaka, R.G. Minet and T.T. Tsotsis, Propane dehydrogenation in a packed-bed membrane reactor. AIChE J., 39 (1993) 526. 41. P. Trambouze, Energy saving in the petrochemical industry, in: P.A. Pilavachi (Ed.), Energy Efficiency in Process Technology. Elsever Applied Science, London, 1993, pp. 61-72. 42. A. Chauvel and G. Lefebvre, Petrochemical Processes 1. IFP Publications, Paris, 1989, pp. 361-372. 43. J.G.P. Sheel and C.M. Crowe, Simulation and optimization of an existing ethylbenzene dehydrogenation reactor. Can. J. Chem. Eng., 47 (1969) 183. 44. J.C.S. Wu and P.K.T. Liu, Mathematical analysis on catalytic dehydrogenation of ethylbenzene using ceramic membranes. Ind. Eng. Chem. Res., 31 (1992) 322. 45. B.K. Abdalla and S.S.E.H. Elnashaie, A membrane reactor for the production of styrene from ethylbenzene. J. Membr. Sci., 85 (1993) 229. 46. B.K. Abdalla and S.S.E.H. Elnashaie, Catalytic dehydrogenation of ethylbenzene to
14 w APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
679
styrene in membrane reactors. AIChE J., 40 (1994) 2055. 47. Y.L. Becker, A.G. Dixon, W.R. Moser and Y.H. Ma, Modelling of ethylbenzene dehydrogenation in a catalytic membrane reactor. J. Membr. Sci., 77 (1993) 233. 48. F. Tiscareno-Lechuga and C.G. Hill Jr., Experimental studies of the non-oxidative dehydrogenation using a membrane reactor. Appl. Cat. A: Gen., 96 (1993) 33. 49. J.C.S. Wu et al., Dehydrogenation of ethylbenzene to styrene using commercial ceramic membranes as reactors. Sep. Sci. Technol., 25 (1990) 1489. 50. Ullmans Encyclopedia of Industrial Chemistry, 5th edn., 1993. 51. G.F. Froment and K.B. Bischoff, Chemical Reactor Analysis and Design. Elsevier, New York, 1990, pp. 501-508. 52. M.P. Harold, C. Lee, A.J. Burggraaf, K. Keizer, V.T. Zaspalis, and R.S.A. de Lange, Catalysis with inorganic membranes. MRS Bulle., April (1994) 34-39. 53. Petrochemical Handbook 1993. Hydrocarbon Process., 72 (1993) 162. 54. M. Steinberg, H.C. Cheng, Modern and Perspective Technologies for hydrogen production from fossil fuels. Int. J. Hydrogen Energy, 14, (1989) 797. 55. G.A. Mills, Coproduction of Hydrogen and Electricity: Catalytic Applications. ACS Div. Fuel Chem., 39 (1994) 1162. 56. P. Riemer, Carbon Dioxide Capture from Power Stations. IEA Greenhouse Gas R&D Programme, 1993. 57. An Attractive Option for C02 Control in IGCC Systems: Water-gas Shift with Integrated Hydrogen~carbon Dioxide Separation (WIHYS) Process. Final Report, CEC project number JOU2 CT92-0158. P.T. Alderliesten and M. Bracht (eds.). Partners: Netherlands Energy Research Foundation, Univ. of Essen (Germany), Univ. of Limerick (Ireland), National Technical Univ. of Athens (Greece) and Siemens AG (Germany), 1996. 58. H.M. van Veen, Applicatiestudie keramische membranen voor g a s s c h e i d i n g - Onderzoek en ontwikkelingen op het gebied van hoge temperatuur gasscheiding met anorganische membranen: status tot november 1993 (Application Study Ceramic Membranes for gas S e p a r a t i o n - R&D Aimed at high Temperature gas Separation with Inorganic Membranes - - State of the art to November 1993, in Dutch), Confidential ECN report, Report number ECN-CX-94-056, June 1994. 59. Microporous Carbon Membranes for Gas Separations. Brite Euram Project BREU-CT920568. Partners: BP Research and Eng. Centre (England), NRC Democritos (Greece), SCT (France) and Imperial College (England). 60. R.L. Keiski, T. Salmi, V.J. Pohjola, Development and verification of a simulation model for a non isothermal water-gas shift reactor. Chem. Eng. J., 48 (1992) 17. 61. C.R.F. Lund and L.A. Bernstein, Membrane reactors for catalytic series and series-parallel reactions. J. Membr. Sci., 77 (1993) 155. 62. G.R. Gavalas, Hydrogen Separation by Ceramic Membranes in Coal Gasification. Final report, DOE/MC/26365-3423, 1993. 63. E. Kikuchi, S Uemiya, N. Sato, H. Inoue, H. Ando and T. Matsuda, Membrane reactor using microporous glass-supported thin film of palladium; application to the water-gas shift reaction. Chem. Lett., 3 (1989) 489. 64. S. Uemiya, N. Sato, H. Ando and E. Kikuchi, The water-gas shift reaction assisted by a palladium membrane reactor. Ind. Eng. Chem. Res., 30 (1991) 585. 65. D.J. Edlund, W.A. Pledger, B.M. Johnson, D.T. Friesen, presented at the 5th Annual Meeting of the NAMS, Lexington, KE, 1992, paper 11F. 66. L. Lloyd, D.E. Riddler, M.V. Twigg, The water-gas shift reaction, in: M.V. Twigg (Ed.), Catalyst Handbook. Wolfe Publishing, 1989, pp. 283-338.
680 67.
14m APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
R. Wurster and W. Zittel, Energy Technologies to Reduce CO2 Emissions in Europe: Prospects, Competition, Synergy. Conference proceedings, IEA-OECD, Paris, 1994. 68 J. E. Hogsett, W.H. Mazur, Estimate membrane system area. Hydrocarbon Proc., 62 (1983) 52. 69. A.S. Kovvali, S. Vemury, K.R. Krovvidi and A.A. Khan, Models and analyses of membrane gas permeators. ]. Membr. Sci., 73 (1992) 1. 70. Y. Shindo, T. Hakuta, H. Yoshitome, H. Inoue, Calculation methods for multicomponent gas separation by permeation. Sep. Sci. Technol., 20 (1985) 445. 71. R. Rautenbach and W. Dahm, Simplified calculation of gas-permeation hollow fiber modules for the separation of binary mixtures. J. Membr. Science, 26 (1986) 319.
Subject Index abrasive materials, 623 acetylacetone (acacH), 243 acidity, effects of, 304 activated gas permeation, 378 activation energy, 393 additives effect on microstructure, 290 effect on stress, 290 adsorption competitive, 381 multicomponent, 40 t-curve, 48 adsorption isotherms determination of, 43 types of, 36 ageing, 303, 308 agglomerate forming, 153 aggregates, 154 alkalinity, 315 y-alumina, 240 film, 321 membranes, 233, 289, 296-297 supporting membrane, 300 supports, 306, 309 amphiphilic systems, 245 amphoteric behaviour of metal oxides, 584 anodic alumina membranes, 539 apparent activation energy, 347, 385 applications, 641 architecture, 21 asymmetric structures, 23, 27 atomic force microscopy, 307 BaAl~2019, 235 back diffusion, 364 BaTiO3 membrane, 236
beer brewing, 627 bending strength, 136 BET isotherms, 40, BICUVOX, 470 bilge water, 621 binary silica-metaloxide, 305 binder polymer, 206 binders, 122, 203, 214 blocking electrodes, 491 boehmite, 240 Bosanquet equation, 358 Brownian agglomeration, 161 Brownian movement, 160 brownmillerite, 502 structure, 499 bubble point, 99, 134 burst pressure, 136 cake filtration, 154 calcia-stabilized zirconia, 465 capillary colloidal filtration, 183 capillary condensate permeation, 352 capillary condensation, 350, 373 capillary pressure, 272, 352 capillary suction, 351 carbon membranes, 312, 354, 546 carbon supports, 321 carbonate formation, 512 Carman-Kozeny, 187, 212, 575 catalysts, 666 cationic surfactant, 249 ceramic membranes applications of, 619 m commercially available, 620 ceramic nanofilters, 240, 596 ceramic paste preparation, 121 681
682 ceramics industry, 623 ceria-doped ZrO2--Y203, 475 CETS, 555, 556 CH3OH dehydrogenation, 540 CH4 reforming, 537 Chapman-Enskog relation, 359 characterisation, 67 charge density, 603 charge disproportionation, 484, 493 charge-pH diagram, 233 chemical diffusion coefficient, 451,492 chemical energy transmission systems (CETS), 555 chemical modification of the membrane surface, 633 chemical processes, recycling in, 626 chemical resistance, 33 chemical stability of y-alumina membranes, 234 chemical vapour deposition, 310, 375 chemically recuperated gas turbine (CRGT), 557 chemisorption, 36 clarification, 628 cluster-cluster model, 302 clusters, 230, 238 CMR, 540 configuration, 531 CNMR model, 550 CO2 removal, 558 coagulant, 629 coal gasification combined cycle, 666 coating flow dynamics, 190 coating thickness, 189, 190 coke formation, 654 colloidal filtration, 210 colloidal particles, 232 colloidal processing, 142, 163 colloidal ceramic processing, 150 colloidal sols, 232 colloidal stability, 210 colloidal suspensions, 229
SUBJECTINDEX commercially available membranes, 31 compact layer, 151 contact time, 151 withdrawal speed, 151 competitive adsorption, 381 competitive Langmuir adsorption, 387 compressor-condensate, 620 concentrated suspension, 173, 175 concentration factor, 623 concentration polarization, 570 condensation rate, 299 condensation reactions, 301 configurational diffusion coefficient, 388 connectivity, 72 contact angle, 196, 198 advancing, 198 receding, 198 contact time, 212 continuous stirred tank reactor, 504 convective flow, 603 conversion, 651, 670 correlation factor, 492 coupled membrane processes, 608 coupling effects, 355 cracking, 208, 279 cracking phenomena, 294 criteria for the selection of materials, 510 critical cracking thickness (CCT), 177 critical point, 272, 275 critical stress, 275,276, 280 critical thickness, 296 cross-flow microfiltration, 590 crystallisation, 315 CVD techniques, 538 Damk6hler number, 647 Darcy's law, 158 dead end pores, 335 deagglomeration, 214 Deborah number, 161 Debye length, 586 Debye-H~ckel screening length, 456
SUBJECTINDEX decomposition, 508 defects, 178-181 defect chemistry, 472 deflocculants, 123, 131 degreasing baths, 625 dehydrogenation, 550 of ethylbenzene, 643, 657 of propane, 643, 648 dense ceramic membranes, 435 dense inorganic membranes, 643 density, 70 density states, 488 depolarization, 591 detergents/surfactants retention, 626 dewetting, 155, 181, 190 diffusion coefficients, 359, 390 chemical, 451, 492 intrinsic, 384 tracer, 491 dip-coating, 183 Donnan effect, 603 Donnan exclusion, 588 drainage, 190 drying, 175 characteristics of membranes, 287 Constant rate period, 272 first falling rate period, 273 forces, 153 front, 274, 294 process, 271 rate, 274 second falling rate period, 274 stress, 276, 288 zone, 292 dual-phase composites, 470 dual-phase membrane, 438 Dubinin-Raduschkevich adsorption, 388 equation, 43, 52 dust, 181 Dusty Gas Model, 355, 359
683 edge-effects, 504 electrical double layer, 585 electro-osmosis, 594 electro-ultrafiltration, 610 electrokinematic flow, 588, 603 electrokinetic radius, 588 electroless plating, 538 electrolytic domain, 464 electronic conductivity, 492 electronic stoichiometry, 493 electrophoresis, 610 electrostatic interaction, 164 ellipsometry, 94 emulsion treatment centres, 621 emulsions, 621 enzymes, 632 equations of state, adsorption isotherms from, 41 erbia-stabilized bismuth oxide, 467 ethylbenzene, 657-658 extended defects, 495 extrusion, 119 facilitated transport, 608 FBMR configuration, 531 fermentation broths, 632 film coating, 189, 215, 262 film formation, 260 film thickness, 190 fish processing, 622 flat supports, 120 flow pulsations, 591 flow sheeting, 648 fluid flow measurements, 102 formed-in-place membranes, 580 fouling, 61, 575, 622 fractal concept, 238 fractal dimension, 299, 301 fractal geometry, 72 fracture, 280 fruit juices, 627
684 galvanic baths, 627 gas adsorption, 78 gas permeability, 103 gel structure, 154 Gibbs-Thomson equation, 273 Gouy layer, 586 grain boundaries, 508 grain boundary diffusivity, 508 gyration radius, 301 Henry constant, 385 Henry's law, 38, 346 high temperature NMR, 500 hollow fibres, 29 Horvath-Kawazoe equation, 55 hybrid installations, 11 hybrid membranes, 606 hydraulic permeability, 588 hydraulic pore radius RH, 246 hydraulic radius, 51 hydraulic resistance, 576 hydrodynamic model, 349 hydrodynamic of micro- and ultrafiltration systems, 590 hydrodynamics, 570 hydrogen recovery, 670 hydrolysis, 301 hydrostatic pressure difference, 351 image analysis, 77 immersion calorimetry, 84 immobilising an enzyme or yeast, 634 impurity phase, 467 initial layer formation, 260 ink, 624 ink-bottle pores, 50 inorganic membranes, 642 interaction forces, 162 interconnectivity, 26 intergrowth, 497 intergrowth structures, 511 intrinsic diffusion coefficient, 384
SUBJECTINDEX ionic conductivity, 454 ionic pre-exponential term, 496 isoelectric point, 594 isotopic exchange, 459, 468 Kelvin equation, 50, 350 kinetic demixing, 511 kinetics, 658, 663 Knudsen contribution, 342 Knudsen diffusion, 357 Knudsen equation, 338 Knudsen number, 337 Knudsen permeabilities, 343 Knudsen permeation of mixtures, 357 LaA111018, 235 lamellar systems, 248 Langmuir adsorption constant, 384 Langmuir isotherm, 36 lanthanum oxychloride porous thin film, 242 large micropores, permeation in, 387 late transition metal-containing perovskites, 492 layer thickness, 212, 215, 306 layered structures, 141 m dip-coating, 142 m porous, 142 substrate, 141 support structure, 141 suspensions, 142 withdrawal coating, 142 liquid adsorption, 61 liquid crystal phase, 249 liquid displacement techniques, 99 liquid permeability, 102 lubricants, 123 macromolecules, 156 macropores, 71 macroporous support, 119 manure, 623
SUBJECTINDEX market penetration, 10 market situation, 2 mass fractal dimension, 238 mass fractal, porosity of, 299 mass transport, 570 maximum packing, 171 Maxwell-Stefan equations, 386, 572 mean velocity, 338 membrane applications, prospects for, 12 membrane architecture, 335 membrane bioreactor, 608 membrane compaction, 578 membrane concepts, 436 membrane cut-off, 596 membrane fouling, 577 membrane geometry, 582 membrane reactors, 11, 633, 645, 658, 667, 673 applications, 642 membrane separation, 577 membrane thickness, 8 characteristic, 456 membrane transport, 572 membranes from RuO2-TiO2, 235 membranes, types of, 21 mercury porosimetry, 78 mesopores, 71 mesoporous alumina membranes, 539 mesoporous inorganic membranes, 643 mesoporous structures, 229 mesoporous textures, 248 metal alkoxides, 237 metal salts, 232 microdomains, 498, 501 micropore filling, 58 micropore size distribution, 53, 57 micropore volume filling, 58 micropores, 71 microporous ceramic membranes, 669 microporous inorganic membranes, 643 microporous material, 231 microporous membranes, 16, 298, 555
685 ---- highly selective, 660 obtained by chemical vapour deposition, 310 microporous silica membranes, 57, 253 microporous structure, 240 microporous top layers, 239, 240 microporous volume, 248 microscopy, 74 microstructural development, 275 microstructure, 162 MIEC membrane, 437 migration enthalpy, 495 mixed conduction, 438 mixed-conducting oxide membranes, 435 mixin g, 123 mobile turbulence promoters, 611 modelling, 549, 555, 650, 661 modelling equations, 464 modelling membrane processes, 646 modification technologies, 15 modified membranes, 354 modified structures, 26 molecular flow, 348 multi-valent dopants, 472 multichannel monolithic elements, 29 multiphase reactors, 542 multiple step coating, 267 nanofiltration, 595 nanophase ceramics, 240 NEMCA, 548 Nernst-Einstein, 458, 490 Nernst-Planck equation, 575 non-ionic surfactants, 246 non-Newtonian, 215 nonstoichiometry, 483 nuclear magnetic resonance, 87 nucleation, 315 observed rejection, 571 oily emulsions, 620 operating costs, 623
686 ordered microporosity, 249 ordered porous texture, 249 ordering local, 501 vacancy, 495, 497 organic additives, 268 oxidative coupling of methane, 507 oxygen desorption, 488 oxygen flux, 464--466 difficulties in measuring, 504 oxygen permeability data, 440 oxygen permeation, factors controlling, 448 oxygen pumps, 438, 469 oxygen transport equations, 489 packed bed membrane reactor, 650 packed structures, 275 paint, 624 palladium membranes, 660 paper and pulp, 632 partial conductivity, 454 partial electronic conductivity, 463 partial molar entropy, 487 partial oxidation, 543, 549, 553 particle compact, 173 particle packing, 152 advancing contact angle, 198 agglomerate forming, 153 aggregates, 154 binder polymer, 206 binders, 203, 214 Brownian agglomeration, 161 Brownian movement, 160 cake filtration, 154 capillary colloidal filtration, 183 Carman-Kozeny, 187, 212 coating flow dynamics, 190 coating thickness, 189, 190 colloidal filtration, 210 colloidal processing, 163 colloidal stability, 210
SUBJECTINDEX colloidally stable, 210 compact, 207 m concentrated suspension, 173, 175 contact angle, 196 contact time, 212 cracking, 208 critical cracking thickness (CCT), 177 Darcy's law, 158 deagglomeration, 214 Deborah number, 161 m defects, 178-181 dewetting, 155, 181, 190 dip-coating, 183 drainage, 190 drying, 175 drying forces, 153 dust, 181 dynamic contact angle, 198 electrostatic interaction, 164 film thickness, 190 film-coating, 189, 215 gel structure, 154 interaction forces, 162 layer thickness, 212, 215 macromolecules, 156 maximum packing, 171 microstructure, 162 non-Newtonian, 215 particle compact, 173 Peclet number, 169 pinholes, 178 polyelectrolytes, 203, 214 polymer solutions, 205 polymer thickener, 214 polymeric interaction, 166 pore diameter, 207 porosity, 207 random packing, 171 receding contact angle, 198 rheological properties, 158 rheology, 156, 171,215 shear erosion, 212
SUBJECTINDEX shear flow field, 160 shear induced agglomeration, 161 shear induced diffusion, 188 sintering, 175 sintering stress, 176 slip-casting, 183 sol, 158 spreading parameter, 197 stability of liquid coatings, 200 stability ratio W, 162 structure, 207 suspension, 158, 169, 171 m thickeners, 203 thickening polymer, 157 Van der Waals attraction, 163 viscosity, 212 wetting, 154, 190 work of adhesion, 197 work of cohesion, 197 work of wetting, 197 PBCMR, 532, 540 model, 549 PBMR, 531, 540 Pd-alloy membrane reactors, 534 Peclet number, 169, 574, 599, 601, 647, 670 penetration rate, 299 peptization, 229, 233 percolation, 72, 494 threshold, 470 permeability, 334, 594 coefficient, 59, 360, 390, 573 maximum, 392 measurements, 502 permeance, 333 permeation, 8, 333, 338, 645, 660 permporometry, 104 permselectivity, 365 perovskite -brownmillerite two-phase region, 503 membranes, long-term stability of, 511 space, 497 m stability, 488
687 m structure, 482 phase transformations, 282 physical adsorption, 35 physisorption, 78 pinholes, 27, 178 plasticizers, 122, 132 Poiseuille-type law, 338 polyelectrolytes, 203, 214 polymer solutions, 205 polymer thickener, 214 polymeric gels, 230, 248 polymeric interaction, 166 polymeric sols, 237 polymeric specimens, 301 pores m blocking, 373 characteristics, 335 clogging, 261 diameter, 207 growth, 297 hydraulic radius, 71 narrowing, 311 shapes, 23, 72 size, 71,308 size distribution, 49, 293 types of, 25 porosity, 70, 207, 461,307 porous structure, 67 porous substrates, 150 coating technique, 150 colloidal ceramic processing, 150 m dispersion technology, 150 positron lifetime spectroscopy, 97 potable water, 629 pre-filtering, 622 precursor chemistry, 300 preferential sorption, 369 Present-De Bethune model, 361 process integration, 664, 667 propylene, 648 protein deposition, 579 proteins, 630
688 proton conduction, 512 pulsate flows, 591 radiation scattering, 91 random packing, 171 random point defect chemistry, 490 real rejection, 571 reflection coefficient, 573, 608 reflection conditions, 362 rejection, 603 measurements, 98 of salt mixtures, 601 relaxation experiments, 492 relaxation methods, 461 reliability, 6 reliability factor, 5 repairing defects, 311 return on investment, 654, 673 Reynolds number, 584 rheological properties, 158 rheology, 156, 171,215 rotating disc, 593 roughness, 29 roughness effects, 270 saddle point, 495 Saito-Foley equation, 55 salt rejection, 598, 604 sealing, 645, 672 of the membrane, 32 selectivity, 393, 651, 660 selectivity coefficients, 58 separation factor, 8, 364, 366, 390 shape factor, 339 shape selectivity, 389 shear erosion, 212 shear flow field, 160 shear induced agglomeration, 161 shear induced diffusion, 188 silica, 375 silica materials, 248, 253 silica membranes, 236, 301,306, 309
SUBJECTINDEX synthesis route of, 303 silica microporous membranes, 300 silica supported membranes, 239 silica-titania, 375 silica-titania microporous membranes, 300 silica-titania/zirconia membranes, 306 silicalite layer thickness, 321 sintering, 175, 281 mechanisms, 282 stress, 176 SiO2 membrane, 60 SIO2/A1203, 239 SiO2/TiO2, 239 SiO2/TiO2 membrane, 60 SiO2/ZrO2, 239 size exclusion, 380 slip casting, 183, 264 slip coefficient, 345 slip flow, 345 slurry Preparation, 131 small polaron mechanism, 473, 493 sol, 158 sol-gel process, 227, 539 solid oxide electrolytes, 462 solid oxide membranes, 546 sorption, 390 space charge (SC) model, 599 specific surface area, 71,248, 603 spreading parameter, 197 stability, 6, 15 of liquid coatings, 200 ratio W, 162 stage cut, 367 stainless steel supports, 317 static lattice simulation, 495 stereology, 74 Stern layer, 586 sticking probability, 299 streaming potential, 588, 603 stress diagram, 284
SUBJECTINDEX during calcination, 291, 294 in supported films, 279 levels, 296 measurements, 283 model, 292 relaxation, 293, 294, 296 structures, modified, 26 styrene, 657 sugars, 631 supercritical fluid extraction, 609 supports, 5, 27 support system, 143 particulate materials, 143 requirements, 146 quality of, 27, 270 technology, 15 surface acoustic waves, 96 surface area, 7 determinations, 46 surface diffusion, 345, 347 surface exchange kinetics, 506 surface homogeneity, 148 surface modification, 506 surface oxygen exchange, 455 surface processes, 9 surface roughness, 146, 461 layer thickness, 147 support requirements, 146-150 surfactant molecules, 246 suspension,. 158, 169, 171 syn gas, 665 tank bottoms, treatment of, 628 tape casting, 119, 130, 133 tape drying, 133, 134 technical and economic feasibility, 648 techno-economic feasibility of inorganic membranes, 673 template agents, 245, 251 template approach, 310 template effect, 231 template ligands, 302
689 template molecules, 300, 313 template removal, 319 terbia-doped ZrO2-Y203, 478 thermal stability, 297 of y-alumina membranes, 235 thermodynamic factor, 383 thermoporometry, 84 thickeners, 203 thickening polymer, 157 thickness, characteristic, 458 three-phase boundary, 471 titania -alumina composite membranes, 236 -doped ZrO2-Y203, 477 membranes, 236, 289, 298 microporous layers, 243 -zirconia ultrafiltration membranes, 235 Toerell-Meyer-Sievers (TMS) model, 599 tortuosity, 26, 72 definition of, 341 tracer diffusion coefficient, 491 transition flow, 339 transition region, 341 trapping of defects, 453 tubular configurations, 120 ultrafiltration, 590 ultrasonic methods, 95 vacancy diffusion coefficient, 491 vacancy ordering, 495, 497 Van der Waals attraction, 163 vegetable waste water, 622 viscosity, 212, 340, 354 viscous flow, 341 Poiseuille, 334 Vycor glass membranes, 537 Wagner equation, 449, 490 warping, 278 washing operations, 625 water-gas shift reaction, 643, 665
690 wetting, 154, 190 whey, 630 Wicke-Callenbach cell, 363 work of adhesion, 197 work of cohesion, 197 work of wetting, 197 yeast, 632 zeolite chemistry, 312 zeolite layers, 320
SUBJECTINDEX particle sizes, 319 thicknesses, 319 zeolite membranes, 313, 376 synthesis, 317 zeolite precursor solution, 317 zeolite-type membranes, 542. zero point of charge (zpc), 585 zeta potential, 587 zirconia, 241, 243, 298 membranes, 290 ZSM5 zeolites, 377
