Past and Present in Denox Catalysis: From Molecular Modelling to Chemical Engineering

Studies in Surface Science and Catalysis 171

PAST AND PRESENT IN DeNOx CATALYSIS

This page intentionally left blank

Studies in Surface Science and Catalysis 171 Advisory Editors: B. Delmon and J.T. Yates Series Editor: G. Centi Vol. 171

PAST AND PRESENT IN DeNOX CATALYSIS From Molecular Modelling to Chemical Engineering Edited by P. Granger University of Lille, Unité de Catalyse et de Chimie du Solide, France

V.I. Pˆarvulescu University of Bucharest, Department of Chemical Technology and Catalysis, Romania

Amsterdam • Boston • Heidelberg Paris • San Diego • San Francisco

• •

London • New York • Oxford Singapore • Sydney • Tokyo

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands Linacre House, Jordan Hill, Oxford OX2 8DP, UK First edition 2007 Copyright © 2007 Elsevier 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 Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-444-53058-5 ISSN: 0167-2991 For information on all Elsevier publications visit our website at books.elsevier.com Printed and bound in The Netherlands 07 08 09 10 11

10 9 8 7 6 5 4 3 2 1

Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org

CONTENTS

Preface 1

Introduction: State of the art in the development of catalytic processes for the selective catalytic reduction of NOx into N2 G. Centi and S. Perathoner

PART 1. A molecular view of reactions involved over DeNOx catalysts – Mechanisms and kinetics 2

3

4

1

25

DFT modeling and spectroscopic investigations into molecular aspects of DeNOx catalysis P. Pietrzyk and Z. Sojka

27

Surface science studies of the mechanism of NOx conversion: correlations between kinetics in vacuum versus under catalytic conditions F. Zaera

67

General features of in situ and operando spectroscopic investigation in the particular case of DeNOx reactions P. Bazin, O. Marie and M. Daturi

97

5

A three-function model reaction for designing DeNOx catalysts G. Djéga-Mariadassou, M. Berger, O. Gorce, J. W. Park, H. Pernot, C. Potvin, C. Thomas and P. Da Costa

6

Identification of the reaction networks of the NOx storage/reduction in lean NOx trap systems P. Forzatti, L. Castoldi, L. Lietti, I. Nova and E. Tronconi

PART 2. Novel developments and future trends to ensure continuous restrictive standard regulations 7

vii

Current tasks and challenges for exhaust after-treatment research: an industrial viewpoint J. M. Trichard

145

175

209 211

vi

8

9

10

11

12

Contents

The role of cerium-based oxides used as oxygen storage materials in DeNOx catalysis X. Courtois, N. Bion, P. Marécot and D. Duprez

235

Aspects of catalyst development for mobile urea-SCR systems – from vanadia-titania catalysts to metal-exchanged zeolites O. Kröcher

261

The formation of N2 O during NOx conversion: fundamental approach and practical developments P. Granger, J. P. Dacquin, F. Dhainaut and C. Dujardin

291

Design of experiments combined with high-throughput experimentation for the optimization of DeNOx catalysts R. Vijay and J. Lauterbach

325

Plasma-assisted NOx abatement processes: a new promising technology for lean conditions M. Ma˘ gureanu and V. I. Pârvulescu

361

Index

397

PREFACE

Presently, there is a general consensus that heterogeneous catalytic processes play an important role in environmental issues regarding their high selectivity towards the removal of undesired side products, such as atmospheric pollutants, in comparison with that obtained from non-catalysed processes. However, such a benefit could be disputed in the future with the implementation of severe restrictions on standard emission of those atmospheric pollutants, particularly nitric oxide, which is a very challenging aspect. Nowadays, the pressure to develop more efficient, environmental-friendly propulsion systems for light and heavy vehicles has dramatically increased. In this particular context, lean-burn engines, integrating both low consumption and low CO2 emission, is a suitable technology. Such systems are already recognised for their greenhouse gas behaviour. However, other atmospheric pollutants are also under concern in those running conditions. Further, technological development towards meeting future standard regulations regarding NOx is an important issue, because the current three-way technology using near stoichiometric conditions will be incompetent to meet the changing regulations in Europe, United States and Japan. Unfortunately, alternative technologies such as the selective reduction of NOx and the NOx storage reduction catalyst suffer from strong limitations regarding selective conversion of NOx into nitrogen at low temperature. In this challenging technological context, car and catalyst manufacturers are stimulated to develop noble-metal-free catalysts, which probably represent an important breakthrough both in the conceptual ideas and original systems developed by academia in order to fulfil future requirements. Considering the above-mentioned aspects, this book offers in its introduction an overview of the state of the art in the field of DeNOx catalysis. The comparison of critical aspects reported in the literature both for stationary and mobile sources could help to focus on novel orientations and new technological developments. The first part of this book deals with fundamental aspects at the molecular level. A better understanding of the reactions involved in unsteady-state conditions is a prerequisite step for improving the performances of the existing processes or developing new ones. The development of powerful in situ spectroscopic techniques is of fundamental interest in kinetic modelling. Correlations between spectroscopic and kinetic data with those obtained from theoretical calculations are also reported. Some illustrations are provided to emphasise the fact that these comparisons may help in determining the nature of the catalytically active sites and building predictive tools for simulations under running conditions. The second part of this book illustrates different practical approaches and aspects related to catalyst

viii

Preface

preparation and the development of alternative technologies. This part is strongly related to industrial considerations, and offers some suggestions for future developments in the field. Pascal Granger University of Lille, France Unité de Catalyse et de Chimie du Solide UMR CNRS 8181 59655 Villeneuve d’Ascq, France Vasile Pˆarvulescu University of Bucharest, Romania Department of Chemical Technology and Catalysis, 4-12 Regina Elisabeta Bvd. Bucharest 030016, Romania

Chapter 1

INTRODUCTION: STATE OF THE ART IN THE DEVELOPMENT OF CATALYTIC PROCESSES FOR THE SELECTIVE CATALYTIC REDUCTION OF NOx INTO N2 G. Centi∗ and S. Perathoner Dipartimento di Chimica Industriale ed Ingegneria dei Materiali, Università di Messina, Italy ∗

Corresponding author: Dipartimento di Chimica Industriale ed Ingegneria dei Materiali, Università di Messina, Italy. E-mail: [email protected]

Abstract The selective reduction of NOx to nitrogen (DeNOx ) is a known technology for eliminating this pollutant from both stationary and mobile sources, but it is still a relevant target in catalysis research and an open problem to meet the future exhaust emission regulations, in particular, for emissions from diesel engines. This chapter will introduce the topics, critically analyzing the state-of-the-art and the different strategies involved in reducing NOx emissions. More specific discussions, particularly on the reaction mechanism and new technical developments, have been made in the following chapters. The effort is to identify the pros and cons in the research in the sector and to outline the areas where there is a need to intensify the research effort.

1. INTRODUCTION AND GENERAL ASPECTS Nitrogen oxides (NOx ), together with CO, sulphur oxides and volatile organic compounds are the primary air pollutants. The main oxides of nitrogen present in the atmosphere are nitric oxide (NO), nitrogen dioxide (NO2 ) and nitrous oxide (N2 O). Nitrous oxide occurs in much smaller quantities than the other two, but it is a powerful greenhouse gas. The major anthropogenic source of nitrogen oxide is fuel combustion, especially in motor vehicles. Home heaters, cookers and gas stoves can also produce substantial amounts of NO in indoor settings. Once emitted, NO (an odourless, colourless gas) oxidizes to NO2 (reddish-orange-brown non-flammable gas with a characteristic pungent odour), especially photochemically, although some NO2 is released directly from the source. NO2 plays a major role in the chemical reactions which generate photochemical smog and ground-level ozone, as well as contributes to the acid rain effect. Nitrogen dioxide is a strong oxidizing agent, which reacts in the air to form corrosive nitric acid, as well as Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

2

Past and Present in DeNOx Catalysis

toxic organic nitrates. It plays a major role in the formation of particulate matter (PM) and haze (nitric acid and nitrate aerosols). Since NO2 is a traffic-related pollutant, emissions are generally higher in urban areas than in rural areas. Annual mean concentrations of nitrogen dioxide in urban areas are generally in the range 10–45 ppb, and lower in rural areas. Levels vary significantly throughout the day, with peaks generally occurring twice daily as a consequence of rush hour traffic. Concentrations can be as high as 200 ppb. In significant concentrations, NO2 is highly toxic, causing serious lung damage with a delayed effect. NO2 is a respiratory irritant which may worsen the symptoms of existing respiratory illness. Other health effects of too high nitrogen dioxide concentration in the air include shortness of breath and chest pains. These harmful effects of nitrogen oxides being known from several years, regulations in their emissions have been progressively introduced in most of the countries worldwide. Therefore, new technologies have been introduced to either limit their formation or convert them to N2 . Among these technologies, the selective catalytic reduction (SCR) was the one which was most successfully developed. Selective catalytic reduction indicates the selective conversion of nitrogen oxides (NO, NO2 or N2 O) to nitrogen (N2 ) in the presence of gaseous oxygen and a reducing agent, either inorganic (ammonia principally) or organic (saturated or unsaturated hydrocarbons (HC), oxygenated HCs such as methanol or nitrogen-containing chemicals such as urea). Often the reaction is indicated also as DeNOx which, however, includes also other techniques to remove nitrogen oxides from gaseous emissions, such as decomposition and absorption, even though these technologies were never fully developed on a commercial scale, except for N2 O catalytic decomposition in some industrial emissions (mainly adipic acid production). The SCR of NOx using ammonia (NH3 ) as the reducing gas was patented earlier in the U.S. by Engelhard Corporation in 1957. The original catalysts, employing platinum or platinum group metals, were unsatisfactory because of the need to operate in a temperature range in which explosive ammonium nitrate forms. Other base metal catalysts were found to have low activity. Research done in Japan in the 1960s in response to severe environmental regulations in that country led to the development of vanadium/titanium oxides catalysts which have proved successful and which still constitute the active elements of current catalysts for NOx abatement from stationary sources (power plants, etc.) [1]. The SCR concept was first used to indicate the abatement of NOx emissions from stationary sources using ammonia as the selective reducing agent (SCR-NH3 ) [1–10], but now it is also used to indicate the selective reduction of NOx with HCs (SCR-HC) [10–20] or urea (SCR-urea) [21–23] in the exhaust emissions from lean burn or gasoline engines as well as to indicate the selective reduction of N2 O with HCs in the tail gas from nitric acid plants [24]. While the SCR process with ammonia for the reduction of NOx emissions from stationary sources is a well-established commercial process to treat emissions of industrial and utility plants (gas-, oil- and coal-fired applications), industrial and municipal waste incinerators, chemical plants (HNO3 tail gases, FCC regenerators, facilities for the manufacture of explosives) and in the glass, steel and cement industries, the use of SCR with HCs or urea [3,7,8–11,24–31] for both stationary and mobile sources) has not yet been commercially established. However, probably SCR with ammonia will soon be introduced on heavy-duty diesel-powered commercial vehicles.

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

3

For light-duty diesel cars, an alternative option which is well-effective, but quite sensitive to sulphur deactivation, is based on the concept of NOx storage–reduction (NOx -SR) [32–35]. In this case, NOx accumulates over the catalyst and is periodically reduced by short, transient excursions from lean (excess O2 ) to rich (excess reductants, i.e. HC, CO and H2 ) conditions. The NOx -storage–reduction process cannot be properly considered as a type of SCR process, because the specific stage of NOx reduction occurs in conditions of oxygen deficiency, similar to three-way catalysts (TWC). TWC being well established and already at the 4th commercial generation, we will not discuss about TWC here, focusing discussion on the NOx to N2 conversion in the presence of oxygen. Advances in TWC have been discussed in several reviews [36–38]. Although there is still a continuing debate, NOx -SR appears to be the preferable solution for light-duty diesel vehicles, being more compact, not requiring an additional tank for the reductant (urea) and avoiding the possibility of formation of harmful by-products in traces (nitriles, HCN, etc.) under conditions of malfunctioning of the catalysts. A recent development is to realize on these catalysts the conversion of both PM and NOx (diesel particulate-NOx reduction – DPNR – system) [39]. This possibility will make the monolith quite compact, an important issue, for use especially in light-duty diesel cars. The DPNR system takes advantage of the oxidation of NO to NO2 over the NOx -SR catalyst and the higher activity of NO2 in oxidizing the PM. Of increasing relevance is also the reduction of NOx in fluid catalytic cracking (FCC) refinery process, due to the new stricter environmental legislation. In alternative to the post-treatment of the emissions, it is possible to add catalytic additives to the FCC catalyst. The advantage is that it is a simple and cost-effective method which can be applied to existing FCC units without modifications. Iliopoulou et al. [40] reported, for example, the use of Ir-based materials (Ir supported on Ce-promoted alumina) for the reduction of NO by CO in the presence of oxygen, under conditions close to those of FCC reaction. Up to 40% oxygen excess in the feed (e.g. 0.7% vol. in the feed in the conditions they used), the activity in NO reduction and at the same time CO oxidation is quite high, but decreases increasing oxygen excess to 100%. An area of recent increasing interest is the use of non-thermal plasma (NTP) to eliminate NOx at low-reaction temperature both from stationary and mobile sources [40,41]. Pulsed electrical discharges, such as dielectrical barrier or corona discharges, are typically used to generate NTP which oxidizes NO to NO2 and N2 O5 , and at time generates O atoms or OH radicals which oxidize the HCs present in the emissions mainly to aldehydes. These may then react over a catalyst with NOx to give N2 and CO2 . In the presence of PM, the NOx also converts them. Thus, NTP has the main function of activating NOx and HC and provides a lower temperature pathway for their conversion. While this is not particularly relevant for stationary emissions, apart from few cases or when the main goal is the low-temperature elimination of volatile organic compounds, it is quite important for auto-exhaust emissions where a large part of the testing cycles for the emissions occurs at low temperature. For diesel engines, the mean temperature of the exhaust is below 200 C for over 70% of the testing cycles. As a consequence, there is a recent, relevant interest for analyzing the application of NTP to especially the treatment of diesel emissions. An alternative use of NTP is to generate H2 (by reforming a part of the fuel) to be added to the auto-exhaust to promote low-temperature activity in HC-SCR. Quite interesting results have been recently shown from Ag−Al2 O3 catalysts whose performances

4

Past and Present in DeNOx Catalysis

in HC-SCR are significantly promoted by the addition of low concentrations of H2 . Although it is claimed that this solution would be preferable over alternative solutions, particularly NOx -SR, for light-duty diesel [20,29], since it eliminates complex engine management related to the usage of NOx -SR with an additional improvement of fuel efficiency, even if current results under practical conditions do not fully support this claim. A wide range of catalytic materials have been investigated for the selective catalytic reduction of NOx . For stationary emissions, NH3 -SCR using vanadium-tungsten oxides supported on titania is the most used method; however, when there is a simultaneous emission of NO and NO2 (in tail gas from nitric acid plants), copper-based zeolites or analogous systems have been proven to be preferable [31b]. In fact, there are two main reactions for NH3 -SCR: 4NH3 + 2NO + 2NO2 → 4N2 + 6H2 O

(1)

2NH3 + 2NO2 → NH4 NO3 + N2 + H2 O

(2)

The ‘fast SCR reaction’, which involves both NO and NO2 , exhibits a reaction rate at least 10 times higher than that of the well-known standard SCR reaction with pure NO: 4NH3 + 4NO + O2 → 4 N2 + 6H2 O

(3)

and dominates at temperatures above 200 C. At lower temperatures, the ‘ammonium nitrate route’ (Eqn. 2) becomes increasingly important. Under extreme conditions, e.g. a powder catalyst at ∼140 C, the ammonium nitrate route may be responsible for the whole NOx conversion observed [42]. This reaction leads to the formation of ammonium nitrate within the pores of the catalyst and a temporary deactivation. For a typical monolithic sample, the lower threshold temperature at which no degradation of catalyst activity with time is observed is ∼180 C. Using Cu-zeolite catalysts may be instead possible to operate at a temperature of ∼150 C which is relevant for economics of the nitric acid tail gas treatment process. Various other classes of catalysts have been investigated for NH3 -SCR, in particular, metal-containing clays and layered materials [43–45] supported on active carbon [46] and micro- and meso-porous materials [31b,47,48], the latter also especially investigated for HC-SCR [25,31b,48–53]. However, while for NH3 -SCR, either for stationary or mobile applications, the performances under practical conditions of alternative catalysts to V-W-oxides supported on titania do not justify their commercial use; if not for special cases, the identification of a suitable catalyst, or combination of catalysts, for HC-SCR is still a matter of question. In general terms, supported noble metals are preferable for their low-temperature activity, centred typically ∼200 C. As commented before, lowtemperature activity is a critical issue. However, supported noble metals have a quite limited temperature window of operation. At higher reaction temperatures (>300 C), micro- or meso-porous materials and/or oxides containing transition metals are preferable. The performances are considerably dependent on the type of reductant, besides the characteristics of the catalyst and the type of transition metal. Although all possible combinations have been explored, including the usage of high-throughput methods, identification of a suitable catalyst formulation active in HC-SCR under practical conditions, especially to decrease by more than

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

5

50% the NOx concentration during testing cycles of diesel engines, have been not yet identified. When initially discovered [14], the possibility of using zeolites containing transition metals (in particular Co-zeolites) for a new process of eliminating NOx from stationary emissions using methane as reductant (CH4 -SCR) has stimulated a large research interest; however, this process in not economical because of the higher temperatures required and the lower catalyst productivity (besides some stability problems of the catalyst). From this point of view, the use of propane instead of methane would be preferable [31b], because of its ease in storage, significantly higher catalyst activity and improved catalyst stability owing to lower temperatures of operations. However, NH3 -SCR would still be preferable for stationary emissions. A particular case is the simultaneous abatement of NOx and SOx , often indicated as combined DeNOx −deSOx , in particular, for power plants using coal or high sulphur content oil. With respect to the separate flue gas desulphurization (FGD) and DeNOx processes, the combined removal of SO2 and NOx in a single operation using a supported oxide sorbent/catalyst has the potential for reducing the cost of environmental control [54,55]. The catalyst should be active in NH3 -SCR and act at the same time as regenerable sorbent for SOx . This means that its NOx activity should be maintained in the presence of large concentrations of SO2 in the feed and a progressive sulphation of the catalyst. Copper supported on alumina was proven to possess such characteristics, different from various other catalysts including the classical active in NH3 -SCR [54]. Worth noting the reaction mechanism of NOx to N2 conversion in the presence of O2 and NH3 is different with respect to that present in V−TiO2 catalysts, for example Reference [31b]. In general terms, the ideal reaction of NO to N2 conversion is by the catalytic decomposition of NO [56]. It is well known that several metal particles, particularly noble metals, are active in this reaction even at low temperature, but the problem is that oxygen upon dissociation remain stuck to the surface of the catalyst and there is a progressive fast deactivation. Therefore, large interest was stimulated from the discovery that Cu ions exchanged into the FAU and MFI micro-porous matrix exhibit unique and stable activity among metal ions exchanged into zeolites in NO decomposition [52,57], in particular, the ‘over-exchanged’ Cu-MFI (Cu2+ /Al > 0.5). However, in the presence of O2 and water vapour, the activity is too low to be of practical interest and the significant effort over the last 15 years has not significantly changed the situation. Worth noting the mechanism of NO decomposition over Cu-MFI or analogous catalysts is different from that over metal particles, and involves nitrosyl or dinitrosyl as intermediate species, and probably as a key step the formation of nitrite−nitrosyl species [56]. Bioprocesses for the removal of nitrogen oxides from polluted air are an interesting alternative [58], but current reaction rates are still too low for large-scale applications. Advanced biological processes for the removal of NOx from flue gases are based on the catalytic activity of either eukaryotes or prokaryotes, e.g. nitrification, denitrification, the use of microalgae and a combined physicochemical and biological process (BioDeNOx ). Finally, the abatement of NOx pollution by using sorbing catalytic materials [59,60] must also be cited. Several solid sorbents for NOx removal (metal oxides, spinels, perovskites, double-layered cuprates, zeolites, carbonaceous materials, heteropolyacids and supported heteropolyacids) have been tested. The results are interesting, but not competitive to actual technologies. To mention that the use of sorbing materials allows

6

Past and Present in DeNOx Catalysis

producing a stream concentrated in NOx which should then be sent to a technology for their abatement. We may thus conclude after this short overview on DeNOx technologies that NH3 SCR using catalysts based on V-W-oxides supported on titania is a well-established technique for stationary sources of power plants and incinerators, while for other relevant sources of NOx , such as nitric acid tail gases, where emissions are characterized from a lower temperature and the presence of large amounts of NO2 , alternative catalysts based on transition metal containing microporous materials are possible. Also, for the combined DeNOx −deSOx , alternative catalysts would be necessary, because they should operate in the presence of large amounts of SOx . Similarly, there is a need to develop new/improved catalysts for the elimination of NOx in FCC emissions, again due to the different characteristics of the feed with respect to emissions from power plants. On the contrary, there is no need to develop generic ‘more active’ NH3 -SCR catalysts, e.g. not in relation to specific different feed conditions as commented above, because ‘activity’ in NH3 -SCR for power plants is mainly dictated from the need to limit SO2 oxidation, as discussed in more detail later. However, still a large part of literature on NH3 -SCR focus on this issue of improving activity, while eventually more appropriate would be to develop catalysts having an improved ratio between the rate of NO conversion and that of SO2 oxidation, but at the very high space velocities typical to commercial operations where mass-transfer limitations are relevant. The V-W-O/TiO2 based catalysts being so effective and stable (typical lifetime is over 10 years), there are few incentives for their substitution, although there is still a space for their improvement, mainly from technological side, in relation to improve the resistance to poisoning, the mechanical stability, etc., especially for operation under severe reaction conditions. For similar motivations, there are limited incentives to develop an alternative SCR process for stationary sources based on methane (CH4 -SCR) or other HCs, or based on NTP technologies, if not for specific, better applications. The situation is instead quite different for mobile sources, and in particular for diesel engine emissions. The catalytic removal of NOx under lean conditions, e.g. when O2 during the combustion is in excess with respect to the stoichiometric one (diesel and lean-burn engines, natural gas or LPG-powered engines), is still a relevant target in catalysis research and an open problem to meet future exhaust emission regulations. Due to the increasing fuel cost and the better fuel economy of novel generation diesel engines, there is a continuous growth in the diesel car (light-duty) market which has already reached over 60% of the share in Europe. In addition, diesel engines are preferred in almost all heavy-duty applications worldwide. Starting from the mid-1980s, when the first emission regulations for automotive diesel engines were issued, several engine modifications have been developed, in order to reduce pollutant species production. While CO and HC emissions are manageable through the use of a simple oxidation catalyst, NOx and particulate are harder task to abate. Optimization of the engine’s combustion towards low NOx emission (e.g. by exhaust gas recirculation) and soot emission has to face a trade-off behaviour. The reduction in NOx and particulates requires a carefully defined balance of the two competing targets (Figure 1.1). A reduction in NOx leads to an increase in the emission of particulates (point ‘A’ in Figure 1.1) and vice versa (point ‘B’ in Figure 1.1). Besides the optimized combustion through engine design, the engine is tuning to operating point B, to further reduce fuel consumption.

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2 0.10

220

EURO3

US2004

7

Without EGR 0.08

0.06

Particulate filter

Particulates, g/kWh

210

NOx reduction

0.04

0.02

Trade-off

Particulate reduction Japan 2005 = Euro V (2008)

EURO5

deNOx system US2007 US2010

200

Fuel consumption, g/kWh

A

With EGR

B

EURO4

190 0

2

4

6

8

10

12

NOx, g/kWh

Figure 1.1. General trend of the NOx and particulate emissions in Europe, Japan and the U.S. for light- and medium-duty engines (ESC test cycle) and effect of engine tuning on NOx /particulate emissions and fuel consumption. EGR: exhaust gas recirculation. ESC test cycle: European stationary cycle (http://www.dieselnet.com/standards/cycles/esc.html).

The necessary compromises between NOx and particulate emissions make advanced after-treatment technologies a must to meet the present and, most important, future regulations (Figure 1.1). In fact, Euro IV regulations, at present in force, have been accomplished by all car manufacturers’ thanks to the extensive use of after-treatment devices, in close synergy with engine management strategy. However, future regulations will necessarily need to introduce a DeNOx system to meet emission targets. As mentioned earlier, the most promising technologies for NOx emission aftertreatment from diesel engines are mainly the following: • NOx -SR traps: NOx adsorbers (traps) are the latest control technologies being developed from Toyota (and later from the main companies in the field) for partial lean burn gasoline engines and for diesel engines. The adsorbers, which are incorporated into the catalyst washcoat, chemically bind nitrogen oxides during lean engine operation. After the adsorber capacity is saturated, the system is regenerated, and released NOx is catalytically reduced, during a period of rich engine operation. • Selective catalytic reduction: In the SCR process, NOx reacts with ammonia, which is injected into the flue gas stream before the catalyst. Different SCR catalyst systems based on platinum, vanadium oxide or zeolites have different operating temperature windows and must be carefully selected for a particular SCR process. As shown earlier, NH3 -SCR has been used for years in industrial processes, in stationary diesel engine applications as well as in marine engines. Urea-SCR technology, using urea as the ammonia precursor, is being adapted for mobile diesel engines. In fact, urea is just a precursor which is hydrolyzed on site (on a first catalyst layer) to generate ammonia and CO2 . On-board storage of urea as well as the distribution

8

Past and Present in DeNOx Catalysis

infrastructure (tank refillers) is simpler and safer using an aqueous solution of urea than in the case of ammonia (compressed tanks must be used, and ammonia itself is a toxic chemical), compensating for the higher costs related to vaporization and mixing of urea as well as the necessity of an additional catalyst layer for the hydrolysis of urea. Another case in which urea instead of ammonia could be preferable as the selective reducing agent is when urea is already available, such as in boiler services for hothouses. A specific case of mobile sources, where nearly the same conventional system used for stationary sources (SCR-NH3 ) can be applied, regards the reduction of NO in the exhaust gas from marine engines and gas turbines. The first marine installation was in 1989 on a 37 000 tdw deep-sea bulk carrier [61]. Many more installations exist today, but this still remains a niche application. Several aspects regarding the reaction mechanism and the new trend in research for NH3 -SCR in stationary and mobile sources, as well as for NOx -SR, will be discussed in the following chapters of this book or in recent reviews [1–20,62–68]. Even so there are some points which deserve comments, in particular, issues and questions which should be clarified regarding future prospects and new directions of research in this field. However, some areas of future research need to be highlighted: (1) noble metalbased formulations which do not form N2 O, (2) novel catalyst formulations which decompose/reduce N2 O below 300 C, (3) on-board routes to form oxygenated reductants, (4) NTP technologies, (5) maintain catalyst within peak operating temperature window and (6) techniques for storing NOx emissions during cool exhaust conditions followed by re-injection of the stored NOx when the catalyst has achieved light-off conditions. There is already an active research on these topics, but a further intensification would be necessary.

2. NH3 -SCR, STATIONARY SOURCES The process has been commercially implemented in Japan since 1977 [1] and a decade later in the U.S., Germany and Austria. The catalysts are based on a support material (titanium oxide in the anatase form), the active components (oxides of vanadium, tungsten and, in some cases, of molybdenum) and modifiers, dopants and additives to improve the performance, especially stability. The catalyst is then deposited over a structured support based on a ceramic or metallic honeycomb and plate-type structure on which a washcoat is then deposited. The honeycomb form usually is an extruded ceramic with the catalyst either incorporated throughout the structure (homogeneous) or coated on the substrate. In the plate geometry, the support material is generally coated with the catalyst. When processing flue gas containing dust, the reactors are typically vertical, with down flow of flue gas. The catalyst is typically arranged in a series of two to four beds, or layers (Figure 1.2). For better catalyst utilization, it is common to use three or four layers, with provisions for an additional layer which is not initially installed. NOx , which consists primarily of NO with less amounts of NO2 in power plant emissions, is converted to N2 by reaction with NH3 over the catalyst in the presence of oxygen (Eqn. 3). A small fraction of the SO2 , produced in the boiler by oxidation

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

9

SCR trays and monoliths

Economizer SCR Reactor

ESP Boiler Air Heater

Figure 1.2. Schematic flow diagram of SCR process and of the trays and monoliths of the SCR reactor.

of sulphur in the fuel, is oxidized to sulphur trioxide (SO3 ) over the SCR catalyst. This causes problems of corrosion and deposit of ammonium sulphate ((NH4 2 SO4 ) and ammonium bisulphate (NH4 HSO4 ), downstream the reactor. Unreacted NH3 in the flue gas downstream the SCR reactor is referred to as NH3 slip. It is essential to hold the NH3 slip below 5 ppm, preferably 2–3 ppm, to minimize the formation of (NH4 2 SO4 and NH4 HSO4 , which can cause plugging and corrosion of downstream equipment. In order to avoid the ammonia slip, and to limit the direct oxidation of NH3 to N2 , the NH3 /NO ratio in the feed is typically maintained below the stoichiometric values, e.g. between 0.90 and 0.95. Catalyst cost constitutes 15–20% of the capital cost of an SCR unit; therefore, it is essential to operate at temperatures as high as possible to maximize space velocity and thus minimize catalyst volume. At the same time, it is necessary to minimize the rate of oxidation of SO2 to SO3 , which is more temperature sensitive than the SCR reaction. The optimum operating temperature for the SCR process using titanium and vanadium oxide catalysts is about 380–480 C. Most installations use an economizer bypass to provide flue gas to the reactors at the desired temperature during periods when flue gas temperatures are low, such as low-load operation. SCR systems can be installed in three possible locations in a power plant: (1) Hot side, high dust: upstream of the air preheater (APH) and electrostatic precipitator (ESP). (2) Hot side, low dust: upstream of the APH and downstream of the ESP. (3) Cold side, low dust: downstream of the APH and ESP. The first two locations are preferred because it eliminates the need to reheat the flue gas to the reaction temperature, thereby minimizing loss of thermal efficiency. The advantages of tail-end configurations include clean flue gas (small trace element concentrations, low SO2 concentrations), more space to accommodate any changes in

10

Past and Present in DeNOx Catalysis

the power cycle or fuel, at least a 50% decrease in catalyst volume and hence cost as the result of clean flue gas, increase in the catalyst life and ammonia bypass. At the tail-end location, however, the flue gas temperature drops below 150 C (∼130–140 C) making it necessary to reheat the flue gas back to the SCR operating temperature, resulting in a penalty in the overall thermal efficiency of the power plant. Therefore, when SO2 is not present, a low-temperature SCR system that operates at about 150 C and avoids reheating the stack gas could be used. Shell [69] and Rhône-Poulenc [70] have developed V-based supported catalysts that operate in the temperature range of 140–250 C while providing over 90% NOx conversion. The higher performances at low temperature are essentially due to high vanadium loading (∼10% with respect to less than 1% in conventional catalysts), but they can be used only when SO2 is absent, because these catalysts are also very effective in SO2 oxidation. Cu-zeolites can also be active in this temperature range [31b], but they are very sensitive to deactivation by SO2 . Therefore, there is still some space and industrial interest to develop new catalysts active in this low-temperature region and not sensitive to SO2 deactivation and not active in SO2 oxidation. In fact, over the last two decades, extensive research has been conducted on low-temperature SCR, but none has been fully convincing under practical situations. Lot of researches have been done, for example, to develop low-temperature NH3 -SCR catalysts based on vanadium oxide, copper oxide or other transition metal oxides supported on activated carbon and carbon fibres [46,71–75], but never resulted in catalysts having the long stability (over 10 years) required. As indicated earlier, the superior activity of a new type of catalyst with respect to ‘conventional’ SCR catalysts is often claimed in the literature. Long et al. [76], for example, claimed the superiority of pillared clay catalysts for selective catalytic reduction of nitrogen oxides to control power plant emissions by comparing Ce-FeTiO2 -PILC (pillared interlayer clay) with a V2 O5 -WO3 /TiO2 catalyst. There are many more examples of this type of claim in the literature. While it is usually worthwhile to improve the rate of reaction, this is not a specific issue of the ‘conventional’ V2 O5 WO3 /TiO2 type of catalyst for power-plant applications. Catalyst activity can be easily improved by increasing the vanadium loading which, however, is instead maintained low (typically below 1% wt.) to limit the oxidation of SO2 to SO3 , a source of problems downstream from the reactor such as plugging (formation of ammonium sulphate) or corrosion. Depending on the SO2 content in the feed, the catalyst activity can be tuned by changing the V2 O5 and WO3 contents in the catalyst to maximize the activity in the reduction of NO, while minimizing SO2 oxidation activity. The ratio between these two reaction rates is the key factor for the choice of the catalyst and not for the activity itself in the conversion of NO. Another important parameter is the selectivity in conversion of NO with respect to ammonia, being always present as a side reaction of ammonia combustion, the minimization of which is a key factor for both process economics and NO efficiency. Related to this aspect, another important parameter is the efficiency of reduction of NO when the ammonia present is less than the stoichiometric value. A final relevant parameter is the temperature window in which the conversion of NO is at the maximum value. In fact, due to the presence of the side combustion of ammonia, the conversion of NO typically decreases at temperatures above the maximum. The presence of a sharp or broader maximum is related to the rate of reaction of ammonia

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

11

with NO (or NO2 ) to give N2 (SCR) with respect to the reaction with O2 to give either N2 or NOx (ammonia oxidation). A broader maximum is helpful in applications where there are fluctuations in the temperature of the emissions. Resistance to deactivation is also an issue, because commercial SCR catalysts have a lifetime of over 10 years. Therefore, the claim of new ‘more active’ catalysts is correct only when all these aspects are considered, but it should also be taken into account that there are not main motivations and incentives from the commercial point of view to develop new NH3 -SCR catalysts. However, some specific cases may be interesting to develop new catalysts, especially when considering different characteristics of the feed and the reaction temperature of operations. There are four main applications of the SCR-NH3 process for the reduction of NO in the emissions: (1) power plants, (2) gas turbines, (3) waste incineration and (4) nitric acid plants. Although, often, specific distinction is not made between these cases and the same catalysts are assumed to be applicable in all cases, there are significant differences in terms of composition of the emissions and space velocities. A specific difference between the first three cases (combustion) and the latter (nitric acid plants) regards NO/NO2 ratio which is typically close to 20 for combustion processes and close to 1 for the nitric acid plants. Furthermore, no SO2 is present in the latter case. While vanadia- on titania-based catalysts can be used for both the classes of applications, there are other types of catalysts such as those based on copper [31b], which show good performances in case of mixtures of NO/NO2 (nitric acid plants), while performances are worse when applied to emissions from catalytic processes. The presence of NO2 in the feed (in nitric acid plant emissions, or when the feed is pre-treated by NTP) induces a significant change in the reactivity and in the reaction mechanism, as shown in the comparisons of Eqns (1) and (2) with Eqn. (3). The ‘fast SCR reaction’, e.g. in the presence of NO2 , exhibits a reaction rate at least 10 times higher than that of the well-known standard SCR reaction with pure NO and dominates at temperatures above 200 C. At lower temperatures, the ‘ammonium nitrate route’ becomes increasingly important. At temperatures in the range of 140–180 C (typical, however, for SCR-NH3 in mobile applications), the ammonium nitrate route may be responsible for the whole NOx conversion observed. Ammonium nitrate decomposition may lead to N2 O. Avila et al. [77] also observed that the NO conversion over a V2 O5 -WO3 /TiO2 catalyst shows a maximum for a NO2 /NO ratio close to one for the presence of an additional reaction pathway involving the formation of ammonium complexes which react with adsorbed NO and NO2 [77], differently from the mechanism occurring in the absence of NO2 involving weak adsorption of NO over the catalyst [62]. Although often it is considered that a single reaction mechanism occurs in the selective reduction of NO by ammonia, data show that instead different mechanisms are possible and that too depending on the type of catalyst and reaction conditions (feed composition, reaction temperature) – one mechanism may prevail over the others [31b]. However, not considering this aspect and making extrapolation regarding the reaction mechanism from one catalyst to another or to different reaction conditions may lead to erroneous conclusions. In addition, it is important to consider all possible opportunities to develop new kinds of catalysts, for example, for the combined removal of NOx and N2 O from nitric acid plant emissions [25].

12

Past and Present in DeNOx Catalysis

There are three different classes of commercial catalysts for the SCR-NH3 process (noble metals, metal oxides and zeolites), but the metal oxide catalysts are those used in almost all the cases. The largest part of commercial applications uses monolithtype catalysts based on V2 O5 -WO3 or V2 O5 -MoO3 oxides supported on TiO2 in the anatase crystalline form. The choice of these catalysts is mainly associated to their higher resistance to deactivation by SO2 and their limited rate of SO2 oxidation with respect to the rate of NO reduction. This derives from the fact that TiO2 is only weakly and reversibly sulphated in the presence of SO2 and O2 , and that the vanadia is wellspread over the titania surface. The vanadia content is generally kept low and is reduced below 1% wt. in the presence of high SO2 concentrations in the feed. WO3 or MoO3 (approximately 10 and 6% wt., respectively) are typically added to increase the surface acidity, the activity and the thermal stability of the catalyst, and to limit the rate of SO2 oxidation. MoO3 also improves catalyst stability in the presence of poisons such as As. The catalysts are further improved by the addition of silico-aluminate and fibreglass to increase their mechanical resistance and strength. The active phase is coated over ceramic (typically cordierite) or sometimes metallic monoliths (thin foils coated with an oxide washcoat) which make it possible to reduce pressure drop, improve resistance to attrition, and lower the plugging rate due to fly ash. The density of the cells as well as geometry of the monolith depends on the type of application (amount of fly ash, SO2 concentration, location of the SCR unit in the plant). Proper design of the monolith is very important to maximize the performance and reduce deactivation. Usually, honeycomb monoliths have a wall thickness of 0.5–0.6 mm and a channel width of 3–4 mm in the case of low dust gas and greater wall thickness (>1 mm) and larger channels (about 7 mm) for flue gas with high dust content. Increasing cell density (i.e. decreasing the wall thickness and size of the channels) increases the external geometric surface area and therefore monolith productivity. The thickness of the active phase coating is usually maintained low, because under the typical high space velocities of common applications (linear velocities in the range of 4–10 m/s), the reaction of NO conversion is controlled by interphase gas diffusion, while the control is kinetic for the oxidation of SO2 . In other words, the conversion of SO2 to SO3 depends on the catalyst volume and can be reduced by decreasing the thickness of the active phase layer (as well as providing a more uniform thickness at the edges of the square-shaped channel), while the reduction of NO depends mainly on the external geometric surface area and therefore is little influenced by the reduction of the catalyst layer thickness, being the internal effectiveness factor very low [3,78]. A simple pseudo-homogeneous 1D model of the reactor based on these assumptions provides a good description of the results [78]. The effects of inter- and intra-phase mass transfer limitations are lumped into an effective pseudo-first-order rate constant. However, more complex models which take explicitly into account (1) the surface roughness and turbulence effects at the monolith entrance, for example, using CFD (computer fluid dynamic) models (therefore, determining local fluidodynamic conditions with respect to the use of the same gas–solid mass transfer coefficient along all the monolith walls) and (2) intra-particle diffusion (using the effective porosity distribution and considering the lack of homogeneity of the washcoat thickness especially in the square-shaped channels) would be preferable for better design of catalysts (optimal porosity, use of additives such as fibres which increase the surface roughness, optimal catalyst thickness profile, etc.) and to improve performances (reduce ammonia slip and

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

13

rate of SO2 oxidation, improve productivity). Therefore, even though SCR-NH3 over V-(W, Mo)/TiO2 -based monolith catalysts is a well-studied process, there is still room for improvement, especially in the area of understanding the interactions between catalyst and reactor engineering aspects. Analysis of the dynamics of SCR catalysts is also very important. It has been shown that surface heterogeneity must be considered to describe transient kinetics of NH3 adsorption–desorption and that the rate of NO conversion does not depend on the ammonia surface coverage above a critical value [79]. There is probably a reservoir of adsorbed species which may migrate during the catalytic reaction to the active vanadium sites. It was also noted in these studies that ammonia desorption is a much slower process than ammonia adsorption, the rate of the latter being comparable to that of the surface reaction. In the SO2 oxidation on the same catalysts, it was also noted in transient experiments [80] that the build up/depletion of sulphates at the catalyst surface is rate controlling in SO2 oxidation. Analysis and modelling of the dynamic behaviour of the catalyst is useful to closely describe the performance during start up, shut down and load variation of stationary applications, and of critical relevance for SCR-NH3 of mobile diesel engine emissions. Use of dynamic models for exhaust transients has not been extensively reported in the literature for the design of improved catalysts, although it is a very valuable method. On the contrary, as will be discussed later, use of this tool to derive mechanistic implications is much less convincing. Other metal oxide catalysts studied for the SCR-NH3 reaction include iron, copper, chromium and manganese oxides supported on various oxides, introduced into zeolite cavities or added to pillared-type clays. Copper catalysts and copper−nickel catalysts, in particular, show some advantages when NO−NO2 mixtures are present in the feed and SO2 is absent [31b], such as in the case of nitric acid plant tail emissions. The mechanism of NO reduction over copper- and manganese-based catalysts is different from that over vanadia−titania based catalysts. Scheme 1.1 reports the proposed mechanism of SCR-NH3 over Cu−alumina catalysts [31b].

NO

NH3 O2

Cu-O NO

NH2

Cu-O

Cu OH

NH3 H+

NO

Cu OH NO

NH4+

–H2O

NH2

O

NH (N) Cu O

N2 + H2O

Cu-O O

NO2– NH4+ Cu-O

NO3– NH4+ O

Cu-O

N2O + H2O

Scheme 1.1. Proposed mechanism for the NO-NH3 reaction over Cu-based catalysts in the presence of oxygen [31b].

14

Past and Present in DeNOx Catalysis

The reaction occurs either via formation of an amide (NH2 ) species which then reacts with NO to form (1) a nitrosamide intermediate, similar to that proposed for vanadia−titania catalysts [81] (however, in the latter catalysts, the NO does not strongly chemisorb or oxidize to NO2 differently from copper-based catalysts), or (2) via oxidation of NO to NO2 or nitrite species which then react with chemisorbed ammonia. Formation of nitrate (NO3 − ) species is also possible (as detected by FTIR, for example), leading to the formation of ammonium nitrate that is responsible for the formation of N2 O. It is also evident from the data that the relative importance of the different pathways of reaction (which are always present) depends on the reaction conditions (feed composition, reaction temperature). Therefore, a single pathway and mechanism of reaction do not exist, but rather, multiple pathways coexist and their rates of reaction depend on the experimental conditions. This concept is valid more in general also for other type of catalysts. It may be noted that generally little attention has been given in the literature to consider that no single reaction mechanism occurs over a specific catalyst and that multiple pathways of reaction are possible. Even though the SCR-NH3 process may seem to be well established, there is still need to focus attention on investigation of the reaction mechanism(s).

3. NH3 -SCR, MOBILES SOURCES The previous section has evidenced that NH3 -SCR technology has been used successfully for more than two decades, to reduce NOx emissions from power stations fired by coal, oil and gas, from marine vessels and stationary diesel engines. NH3 -SCR technology for high-duty diesel (HDD) vehicles has also been developed to the commercialization stage and is already available as an option in the series production of several European truck-manufacturing companies starting from 2001. For mobile source applications, the preferred reductant source is aqueous urea, which rapidly hydrolyses to produce ammonia in the exhaust stream. SCR for heavy-duty vehicles reduces NOx emissions by ∼80%, HC emissions by ∼90% and PM emissions by ∼40% in the EU test cycles, using current diesel fuel (<350 ppm sulphur) [27]. Fleet tests with SCR technology show excellent NOx reduction performance for more than 500 000 km of truck operation. This experience is based on over 6 000 000 km of accumulated commercial fleet operation [82]. The combination of SCR with a pre-oxidation catalyst, a hydrolysis catalyst and an oxidation catalyst enables higher NOx reduction under low-load and low-temperature conditions [83]. Apart from the hydrolysis step, the SCR-urea process is equivalent to that of stationary sources, and in fact the key idea behind the development of SCR-urea for diesel powered cars was the necessity to have a catalyst (1) active in the presence of O2 , (2) active at very high space velocities (∼500.000 per hour based on the washcoat of a monolith) and low reaction temperatures (the temperature of the emissions in the typical diesel cycles used in testing are in the range of 120–200 C for over half of the time of the testing cycle), and (3) resistant to sulphur and phosphorus deactivation. V-TiO2 -based catalysts for SCR-NH3 have these characteristics and for this reason their applications have also been developed for mobile sources.

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

15

However, there are additional aspects (not present in typical SCR-NH3 applications for stationary sources) which make the problem more complex: (1) Very fast transients in the concentration of NO (and CO and HCs) which also require a fast parallel dosing of the reducing agent (in addition, current engines do not have a NOx sensor, and commercially available NOx sensors do not have a sufficiently fast response to closely follow the transients in NOx concentration). (2) The necessity of having a post catalyst layer which can eliminate slipping ammonia (in addition, since CO and HC also must be eliminated, current catalytic SCRurea systems applied to diesel engine emissions are composed typically of five catalytic layers, making the size of the catalytic converter quite large and therefore applicable essentially only to heavy-duty trucks and buses). (3) The necessity of completely avoiding the possible formation of harmful byproducts, even in sub-ppm concentrations (especially at low temperatures during cold starts and when the catalyst is deactivated due to either long mileage or deposition of PM). There are two main possible options for NH3 -SCR for mobile sources which are summarized in Figure 1.3. The classical (configuration A) considers a first layer of a catalyst for the hydrolysis of urea and mixing of NH3 , then a NH3 -SCR catalyst for the removal of NOx and finally an oxidation catalysts, all in the monolith form, for the (partial) removal of HC, CO and PM, together with the elimination of the possible ammonia slip. In the second configuration,(B in Figure 1.3), first there is a CRT® module (Johnson Matthey Continuously Regenerating Technology – Diesel Particulate Filter) which removes HC, CO and PM and oxidizes NO to NO2 , thus making possible the following NH3 -SCR catalyst layer to improve NOx conversion especially at low temperature (fast SCR commented earlier). CRT® uses NO2 produced by a specially formulated catalyst (based on high loading supported noble metals) to burn soot collected by the filter at typical operating temperatures of diesel engine exhaust (∼250 C). The main limitation of the second option is that it requires the use of ultra-low sulphur fuel

(a)

(b) Urea

Urea

Urea hydrolysis catalyst

Urea hydrolysis catalyst

Engine

Engine SCR Removal of NOX

Oxidation cat

CRT

Removal of NH3 slip if required removes CO and some HC’s and PM SOF

Removal of HC, CO, PM NO NO2

SCR Removal of NOX

Oxidation cat Removal of NH3 slip if required

Figure 1.3. Possible options for NH3 -SCR for mobile sources: (A) conventional and (B) coupled to a CRT® catalyst (developed by Johnson Matthey) which oxidizes NO to NO2 , besides to remove HC, CO and PM.

16

Past and Present in DeNOx Catalysis

(<50 ppm S) for maximum emission reduction and filter regeneration, being the catalyst quite sensible to sulphur deactivation. The development of an SCR system for vehicle applications requires precise calibration of the amount of urea injected as a function of the quantity of NOx emitted by the engine, exhaust temperature and catalyst characteristics. Although model simulations can help in the control, it is necessary to use specific NOx sensors which, however, still have problems of sensitivity and transient response. Installing a clean-up catalyst for ammonia would provide more latitude and obtain higher NOx conversion ratios without re-emission of ammonia into the atmosphere.

4. NON-THERMAL PLASMA Non-thermal plasma has received considerable attention in recent years as a potential method to reduce NOx and PM emissions in diesel exhaust as well as NOx and cold start HCs in lean gasoline exhaust [84]. The plasma derives from the application of a high electrical field that causes ionization of the molecules and generation of hot electrons and radicals. The radicals determine the chemical oxidation of the PM converting it into water and carbon dioxide. This reactivity can be engineered to perform chemical reactions under low-temperature conditions (NTP) that allow destruction of pollutants in the gas. The process is highly efficient as the electrons in the plasma carry almost all of the energy and these electrons are used to produce reactive radicals. The result is that a wide range of chemical reactions can be stimulated with relatively low energy input. The reactions that occur to auto-exhaust emissions when exposed to plasma include oxidation of HCs, carbon monoxide, and partially diesel PM also. Nitric oxide (NO) can be oxidized by plasma to NO2 . Plasma alone, due to its oxidizing character, is not a viable NOx control method. However, combinations of plasma with catalysts, referred to as ‘plasma-assisted catalysts’ or simply ‘plasma catalysts’, have been suggested for NOx reduction. The plasma is believed to show potential to improve catalyst selectivity and removal efficiency. Current ‘state-of-the-art’ plasma catalysts have efficiencies comparable to those of active DeNOx systems, removing about 50% of NOx at a fuel economy penalty of less than 5% [85]. Non-thermal plasmas can be produced in a number of ways, including a variety of electrical corona discharges, radio frequency discharges, microwave discharges and electron beams. The most common NTP technologies for emission reduction in engine exhaust streams are the following. • Corona discharge is the simplest type of plasma generator. A feature of the corona discharge, which differentiates it from the other discharges, is that no dielectric is involved. Instead, an electron avalanche is initiated from a sharp metallic surface where the radius of curvature is small. The electric field has to be pulsed in order to prevent the plasma from going into the thermal mode and forming an arc. The electric field in corona reactors is about 50 kV/cm. • Dielectric barrier discharges (DBD) is the case in which a dielectric material is placed on the electrode surfaces. Similar to corona discharges, small-scale electron streamers are formed. In the DBD mode of operation, the threshold electric field

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

17

is relatively low and low-energy electrons are formed. DBD reactors can be also configured as concentric cylinders or as arrays of alternating positive and negative dielectric-coated electrodes. The parallel-plate architecture, however, features the most simple and cost-effective design. • Microwave resonating waveguides are a theoretically simple and effective way to generate cold plasma, working around the issues traditionally bound to other known techniques. Plasma is generated in this case by a resonating electromagnetic wave, whereas electric field locally overcomes gas dielectric rigidity, an electric discharge happens starting up a stable plasma field. Another possibility for the plasma device is the generation of NO2 for the enhancement of NH3 -SCR at low temperature, (the so-called fast SCR reaction) which occurs if there is a 1:1 NO:NO2 stoichiometry over V-TiO2 type catalysts and with variable stoichiometry over Fe-beta zeolite. Being able to ‘switch on’ and tune NO2 production over a limited temperature range will help to avoid NO2 slip issues, that can be an issue for oxidation catalysts. Also, if tuned correctly, plasma can do the NO oxidation without in turn doing SO2 oxidation and so generate NO2 without making sulphates (and associated particulates). Together with the fast oxidation (at low temperatures) of NO to NO2 , the plasma causes the partial HC oxidation (using propylene, the formation of CO, CO2 , acetaldehyde and formaldehyde was observed). Both the effects cause a large promotion in activity of the downstream catalyst [86]. For example, a -alumina catalyst which is essentially inactive in the SCR of NO with propene at temperatures ∼200 C allows the conversion of NO of about 80% (in the presence of NTP). Formation of aldehydes follows the trend of NO concentration suggesting their role in the reaction mechanism. Metal oxides such as alumina, zirconia or metal-containing zeolites (Ba/Y, for example) have been used [84–87], but a systematic screening of the catalysts to be used together with NTP was not carried out. Therefore, considerable improvements may still be expected. Problems observed were regarding the catalyst stability and the tendency to form organic deposits on the catalyst surface. From the application point of view, the main technical problem is the energy required to maintain the plasma (typically, 20–60 J/L). However, significant improvements are also expected in this regard. A further aspect which deserves attention is the formation of toxic by-products such as CH2 O and HCN within the NTP plumes. Nevertheless, this route appears promising, especially when a better integrated design between the NTP device and the catalyst becomes available. NTP-promoted catalytic removal of NOx has also been tested as a new NH3 -SCR process [88,89]. V2 O5 /TiO2 [88] and V2 O5 −WO3 /TiO2 monolithic catalysts [89] have also been evaluated. At low-reaction temperatures (∼150 C), in the condition without plasma treatment, the removal of NOx was minimal, conversions of NO of about 70% could be obtained. Similar effects were observed for selective catalytic reduction in gas mixtures containing equal amounts of NO and NO2 , indicating that the role of plasma is not only that of promoting the conversion of NO to NO2 . The conversion of NO to NO2 is a key aspect of HC-SCR [17], while it may or may not be an important factor in NH3 -SCR, depending on the type of catalyst. Copper-based catalysts are significantly promoted using NO2 /NO mixtures [31b], while the effect is less important on vanadiaor titania-type catalysts under non-NTP conditions. Both the classes of catalysts have been found to be highly active under NTP conditions and a synergistic combination of

18

Past and Present in DeNOx Catalysis

NTP and both NH3 - and HC-SCR was observed to be present under real diesel engine exhaust conditions [90]. Low-temperature activity promotion is an issue in mobile (diesel) applications, but may not be a critical issue in several stationary applications, apart from those where the temperature of the emissions to be treated is below 200 C (for example, when a retrofitting SCR process must be located downstream from secondary exchangers, or in the tail gas of expanders in a nitric acid plant). In the latter cases, a plasmacatalytic process [91] could be interesting. In the other cases, the use of NTP together with the SCR catalyst is not economically viable. However, the synergetic combination of plasma and catalysts has been shown to significantly promote the conversion of hazardous chemicals such as dioxins [92]. Although this field has not yet been explored, it may be considered as a new plasmacatalytic SCR process for the combined elimination of NOx , CO and dioxins in the emissions from incinerators.

5. NOx STORAGE–REDUCTION CATALYST The NOx storage–reduction (NRS) catalyst, often called as NOx trap, offers efficiency ratios comparable to those obtained by the SCR, but without the disadvantage of having to carry another reducer on board. In principle, it operates by alternating two stages [93]: • Engine runs normally on a lean mixture. During this stage, the nitrogen oxides (after being oxidized to NO2 ) are stored in nitrate form on an adsorbant mass. • The engine is run on a rich mixture. The NOx released are then reduced by the reducers (CO, HC) present in the exhaust gases. To release and reduce the nitrogen oxides, the air/fuel ratio must exceed or be equal to 1, which is unusual for a diesel engine. This is done by tuning the engine (air flow, injection phasing and duration, EGR ratio). Therefore, a relevant priority is to optimize these alternations (lean versus rich) to obtain the best trade-off between NOx emissions and over fuel consumption. The conditions under which regeneration occurs (fuel/air ratio, tuning) affect the duration and efficiency of release. For instance, it takes less time when the mixture is richer (>1.15) and is better when tuning adjustments which favour CO emissions over HCs. Regeneration also causes a substantial rise in the temperature of the trap, which could exceed its storage window and thereby limit the efficiency of the subsequent storage phase. The impact of these parameters, on both storage and release of NOx , shows that the best NOx /consumption trade-off is obtained when regeneration occurs at high levels of richness. By optimizing the system as a whole, it is possible to obtain reduction efficiencies of about 80% for over diesel fuel consumption of 2–5% [94]. To avoid discharge of CO and HCs, which can happen when running a richer fuel mixture, an oxidation catalyst is installed downstream from the trap to treat these emissions. The NOx trap also requires the use of a low-sulphur fuel (<10 ppm). In the presence of sulphur, the trap becomes progressively saturated with sulphates, more stable than nitrates, which quickly reduces its efficiency. In addition, it requires periodic desulphation by running a richer fuel mixture at high temperature. In order to enhance the durability of the NSR catalyst, sulphur poisoning should be suppressed by the acceleration of sulphur desorption. To achieve this, the acidity of the

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

19

support needs to be optimized. A combination of TiO2 and -Al2 O3 could be used to maintain the amount of NOx storage and to minimize the amount of SOx deposition. To enhance the removal of sulphate, a hexagonal cell monolithic substrate was developed [93b], which resulted in a uniform catalytic wash-coat thickness, and in-situ hydrogen generation on the catalyst was promoted by adding Rh/ZrO2 . The Rh/ZrO2 -added catalyst had a high activity for hydrogen generation via steam reforming [93b]. Unlike SCR, the NOx trap works closely with the engine in managing rich-mixture running. In order to optimize strategies, any impact on engine operation and its durability (lubricant dilution, thermal stresses, etc.) should be taken into account, along with the thermal aging of the trap. Rich mixture running, especially at high engine loads, generates smoke emissions that would necessitate the use of a particulate filter. In NSR catalysts, the Ba−Pt interface plays an important role in the storage of NOx , which occurs by the formation of Ba(NO3 2 . Recent results [95] using scanning transmission microscopy (STM) and a model catalyst formed by deposition of a Ba thin films on Pt(111) showed that, at room temperature, a film of Ba was formed with few individual Ba atoms, which were locally ordered. Upon annealing, particles are produced, of which atomic resolution is achieved with an atomic spacing consistent with the (111) plane of Ba. Upon admission of oxygen, the step edges appear to be pinned, which was an attribute to the formation of the oxide (BaO) or the peroxide (BaO2 ). Furthermore, from the step edge, an additional layer begins to grow on the Ba film that was suggested to be BaO2 . This layer was meta-stable and reduced in size before disappearing when the oxygen pressure was removed. Moreover, when the surface was exposed to a mixture of O2 and NO at 300 C, a number of bright adsorbates appear and additional growth was observed at the step edges. This was attributed to the formation of the nitrate species (Ba(NO3 2 ). In the case of excess oxygen (O2 /NO ratio >1.0), this reaction was accelerated. When NO and O2 were removed from the system, these adsorbates and additional growth at the step edges disappeared. These observations of reactions on the Ba thin film using STM suggest that the NOx storage reactions involving NO and O2 to form the nitrates occur via the following pathways [96]: Ba + O2 → BaO + O → BaO + 1/2O2 → BaO2 at step edges or defects on the terrace

(4)

BaO2 + NO → BaO2 −NO → BaO−NO2 bright adsorbates

(5)

BaO2 + 2NO2 → BaNO3 2 formation on nitrate at step edges

(6)

There are also different hypotheses on the reaction mechanism, as will be discussed in the following chapters. This is still an open area of research and a further understanding will certainly lead to the development of improved catalysts. There are, in particular, three main areas in which further development is necessary: (1) improve the lowtemperature activity, e.g. below 250 C, (2) improve resistance by deactivation by sulphur and (3) improve the hydrothermal stability. Hydrotalcite-based materials [31a,97] offer interesting opportunities in this direction. A recent area of development is to use these NRS catalysts for the combined treatment of NOx and particulates. The combination of the NOx trap and of a particulate filter depends on various factors: (1) the selection of the best suited filter technology

20

Past and Present in DeNOx Catalysis

(CRT, catalytic filter, regeneration with fuel-borne additive), (2) the constrains on the respective position of each system, (3) the modalities to combine the engine control strategies needed to ensure that both the systems operate properly (nitrate storage– reduction, desulphation of the NOx trap, regeneration of the particulate filter) and (4) the strategy to optimize the overall NOx /particulates/fuel consumption trade-off. Therefore, many factors have to be considered and the optimization of the NOx -SR catalyst is also in relation to these factors. The optimum solution will depend on the target application: base engine emissions, exhaust temperatures, conditions of use (city or highway driving). Similarly, the SCR and the particulate filter can be configured into one system. Here again, respective positioning will depend on respective requirements (efficiency window, regeneration strategies and durability). The ‘ultimate’ emissions control solution is a four-way catalysis eliminating the four pollutants simultaneously (CO, HC, NOx and PM). This involves applying a CO, HC and NOx adsorber catalyst to a filtering medium to treat these emissions. As far as passenger cars are concerned, Toyota has brought out its DPNR system for light-duty vehicles. This is a diesel particulate filter onto which a NOx -SR catalyst is applied. Similar applications are under development for heavy-duty vehicles. Four-way concepts based on SCR technology are also under study.

6. CONCLUSIONS The selective reduction of NOx to nitrogen (DeNOx ) is a known technology for eliminating this pollutant from both stationary and mobile sources, but is still a relevant target in catalysis research and an open problem to meet future exhaust emission regulations, in particular, for emissions from diesel engines. Besides, to offer a general overview of the DeNOx technologies and options, we have discussed in this chapter an analysis of the main open problems and directions of research. Due to the increased research effort and different commercial interests, this analysis will not be exhaustive and reports a personal of the pros and cons in the research in the sector and of the areas where there is a need to intensify the research effort. Further aspects will be discussed in the following chapters of the book. In conclusion, not withstanding the relevant ‘past’ in the sector of DeNOx technologies, there is still a ‘future’, new stringent limits and the availability of novel catalysts as well open new perspectives of research. From this point of view, the interest in developing new NTP technologies, especially to address the emissions during the lower temperature part of the cycles, has and will stimulate new scientific challenges which need to intensify the fundamental research in this relevant area.

REFERENCES [1] Nojiri, N., Sakai, Y. and Watanabe, Y. (1995) Catal. Rev.-Sci. Eng., 37, 145. [2] Yeh, S., Rubin, E.S., Taylor, M.R. et al. (2005) J. Air & Waste Manage. Assoc., 55, 1827. [3] (a) Forzatti, P. (2000) Catal. Today, 62, 51. (b) Forzatti, P., Lietti, L. and Tronconi, E. (2003) Nitrogen oxides removal – industrial, In Encyclopedia of Catalysis, Horváth I.T. (Chief ed.), J. Wiley & Son: Chichester, UK. (c) Forzatti, P. (2001) Appl. Catal. A Gen., 222, 221. (d) Forzatti, P. and Lietti, L. (1996) Heter. Chem. Rev., 3, 33.

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2 [4] [5] [6] [7] [8]

[9] [10] [11]

[12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22]

[23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39]

21

Heck, R.M. (1999) Catal. Today, 53, 519. Nakajima, F. and Hamada, I. (1996) Catal. Today, 29, 109. Topsøe, N.Y. (1997) CATTECH, 1, 125. Janssen, F.J. (1999) Environmental catalysis – stationary sources. In Environmental Catalysis (G. Ertl, H. Knözinger and J. Weitkamp, eds) Wiley-VCH: Weinheim, chapter 2, pp. 119. Centi, G. and Perathoner, S. (2006) In Metal Oxides: Chemistry and Applications (J.L.G. Fierro, ed.) chapter 21, pp. 661, Chemical Industries, vol. 108, CRC Press LLC: Boca Raton, FL, USA. Srivastava, R.K., Hall, R.E., Khan, S. et al. (2005) J. Air & Waste Manage. Assoc., 55, 1367. Curtin, T. (2005) Selective catalytic reduction of NOx . In Environmental Catalysis (V.H. Grassian, ed.) CRC Press LLC: Boca Raton, Fla, USA, pp. 197. (a) Burch, R., Breen, J.P. and Meunier, F.C. (2002) Appl. Catal. B Env., 39, 283. (b) Burch, R., Breen, J.P. and Meunier, F.C. (2003) Nitrogen oxides removal with supported metals and oxides. In Encyclopedia of Catalysis, Horváth I.T. (Chief ed.) J. Wiley & Son: Chichester UK. Amiridis, M.D., Zhang, T. and Farrauto, R.J. (1996) Appl. Catal. B Env., 10, 203. Hamada, H. (1994) Catal. Today, 22, 21. Li, Y. and Armor, J.N. (1992) Appl. Catal. B Env., 1, L31. Fokema, M.D. and Ying, J.Y. (2001) Catal. Rev. Sci. & Eng., 43, 1. Shimizu, K., Satsuma, A. and Hattori, T. (2000) Catal. Surveys from Jpn, 4, 115. Misono, M. (1998) CATTECH, 2, 183. Kim, M.H. and Nam, I.-S. (2001) Korean J. Chem. l Eng., 18, 725. Kung, M.C. and Kung, H.H. (2004) Topics in Catal., 28, 105. He, H. and Yu, Y. (2005) Catal. Today, 100, 37. Saito, S., Shinozaki, R., Suzuki, A. et al. (2003) SAE Papers, SP-1801 (Emissions: Advanced Catalyst and Substrates, Measurement and Testing, and Diesel Gaseous Emissions) p. 209. (a) Koebel, M., Elsener, M. and Kleemann, M. (2000) Catal. Today, 59, 335. (b) Koebel, M., Elsener, M. and Madia, G. (2001) SAE Papers, SP-1641 (Diesel Emission Control Systems) p. 101. (a) van Helden, R., Verbeek, R., Willems, F. et al. (2004) SAE Papers, SP-1861(Diesel Exhaust Emission Control Modeling) p. 25. (b) Ibidem, SP-1835 (Diesel Emissions) p. 275. van den Brink, R.W., Booneveld, S., Verhaak, M.J.F.M. et al. (2002) Catal. Today, 75, 227. Delahay, G., Berthomieu, D., Goursot, A. et al. (2003) Surfactant Science Series (Interfacial Applications in Environmental Engineering), 108, 1. Johnson, T.V. (2003) SAE Papers, SP-1754 (Diesel Exhaust Emissions Control) p. 1. Bosteels, D. and Searles, R.A. (2002) Platinum Metals Rev., 46, 27. Wen, B. and He, M. (2000) J. Environmental Sciences (China), 12, 310. Burch, R. (2004) Catal. Rev. Science and Eng., 46, 271. Koenig, A., Held, W. and Richter, T. (2004) Topics in Catal., 28, 99. (a) Centi, G., Arena, G.E. and Perathoner, S. (2003) J. Catal., 216, 443. (b) Centi, G. and Perathoner, S. (1995) Appl. Catal. A General, 132, 179. (a) Matsumoto, S. (2004) Catal. Today, 90, 183. (b) Takeuchi, M. and Matsumoto, S. (2004) Topics in Catal., 28, 151. Kanazawa, T. (2004) Catal. Today, 96, 171. Broqvist, P., Gronbeck, H., Fridell, E. et al. (2004) Catal.Today, 96, 71. Gill, L.J., Blakeman, P.G. and Twigg, M.V. (2004) Topics in Catal., 28, 157. (a) Di Monte, R. and Kaspar, J. (2004) Topics in Catal., 28, 47. (b) Kaspar, J. and Fornasiero, P. (2003) J. Solid State Chem., 171, 19. Granger, P., Dujardin, C., Paul, J.-F. et al. (2005) J. Mol. Catal. A Chem., 228, 241. Sugiura, M., Ozawa, M., Suda, A. et al. (2005) Bull. Chem. Soc. Jpn, 78, 752. Suzuki, J. and Matsumoto, S. (2004) Topics in Catal., 28, 171.

22

Past and Present in DeNOx Catalysis

[40] (a) Iliopoulou, E.F., Efthimiadis, E.A., Nalbandian, L. et al. (2005) Appl. Catal. B Env., 60, 277. (b) Iliopoulou, E.F., Efhimiadis, E.A., Lappas, A.A. et al. (2004) Ind. Eng. Chem. Res., 43, 7476. [41] (a) Kim, H.-H. (2004) Plasma Process Polym, 1, 91S. (b) Pasquiers (2004) Eur. Phys. J. Appl. Phys., 28, 319. [42] Koebel, M., Elsener, M. and Madia, G. (2001) IEC Research, 40, 52. [43] (a) Serwicka, E.M. (2001) Polish J. Chem., 75, 307. (b) Serwicka, E.M. and Krzysztof, B. (2004) Catal. Today, 90, 85. [44] De Stefanis, A. and Tomlinson, A.A.G. (2006) Catal. Today, 114, 126. [45] Vaccari, A. (1999) Appl. Clay Science, 14, 161. [46] Yoshikawa, M., Yasutake, A. and Mochida, I. (1998) Appl. Catal. A General, 173, 239. [47] Larsen, S.C. (2005) In Applications of Zeolites in Environmental Catalysis (V.H. Grassian, ed.) CRC Press LLC, Boca Raton, Fla, USA, pp. 269. [48] Chen, H.-Y., Sun, Q., Wen, B. et al. (2004) Catal. Today, 96, 1. [49] Brosius, R. and Martens, J.A. (2004) Topics in Catal., 28, 119. [50] Choi, B.-C. and Foster, D.E. (2005) J. Ind. Eng. Chem.y (Seoul, Rep. Korea), 11, 1. [51] Liu, Z., Hao, J., Fu, L. et al. (2004) Chin. Science Bull., 49, 2231. [52] Yahiro, H. and Iwamoto, M. (2001) Appl. Catal. A General, 222, 163. [53] Wichterlova, B. (2004) Topics in Catal., 28, 131. [54] (a) Centi, G., Passarini, N., Perathoner, S. et al. (1992) IEC Research, 31, 1963. (b) Centi, G. and Perathoner, S. (1997) IEC Research, 36, 2945. (c) Centi, G., Riva, A., Passarini, N. et al. (1990) Chem. Eng. Sci., 45, 2679. [55] (a) Yeh, J.T., Demski, R.J., Strakey, J.P. et al. (1985) Environ. Prog., 4, 223. (b) Yeh, J.T., Ma, W.T., Pennline, H.W. et al. (1992) Chem. Eng. Comm., 114, 65. [56] Garin, F. (2001) Appl. Catal. A General, 222, 183. [57] (a) Iwamoto, M., Furukawa, H., Mine, Y. et al. (1986) J. Chem. Soc., Chem. Commun., 12, 1272. (b) Iwamoto, M., Yahiro, H., Tanda, K. et al. (1991) J. Phys. Chem., 95, 3727. [58] Jin, Y., Veiga, M.C. and Kennes, C. (2005) J. Chem. Techn. and Biotechn., 80, 483. [59] Gomez-Garcia, M.A., Pitchon, V. and Kiennemann, A. (2005) Environ. Intern., 31, 445. [60] (a) Machida, M. (2000) Catalysis, 15, 73. (b) Arai, H. and Machida, M. (1994) Catal. Today, 22, 97. [61] Jensen, A.B. (2000) Scandinavian Shipping Gazette, 37–38, B82. [62] (a) Busca, G., Larrubia, M.A., Arrighi, L. et al. (2005) Catal. Today, 107–108, 139. (b) Busca, G., Lietti, L., Ramis, G. et al. (1998) Appl. Cataly. B Env., 18, 1. [63] Sadykov, V.A., Lunin, V.V., Matyshak, V.A. et al. (2003) Kinetics and Catalysis, 44, 379. [64] Satsuma, A., Shichi, A. and Hattori, T. (2003) CATTECH, 7, 42. [65] Captain, D.K., Mihut, C., Dumesic, J.A. et al. (2002) Catalysis Letters, 83, 109. [66] Liu, D.-J. and Robota, H.J. (1999) J. Phys. Chem. B, 103, 2755. [67] Iwamoto, M. and Yahiro, H. (2003) In Handbook of Zeolite Science and Technology (S.M. Auerbach, K.A. Carrado and P.K. Dutta, eds) Marcel Dekker, Inc., New York, NY, USA, pp. 951. [68] (a) Hums, E. (1998) Kinetics and Catalysis, 39, 603. (b) Hums, E. (1998) Catal. Today, 42, 25. [69] van der Grift, C.J.G., Woldhuis, A.F. and Maaskant, O.L. (1996) Catal. Today, 27, 23. [70] Visciglio, V., Blanchard, G. and Surantyn, R. (1996) Catal. Lett., 40, 39. [71] Kasaoka, S., Sasaoka, E. and Iwasaki, H. (1989) Bull. Chem. Soc. Jpn., 62, 1226. [72] Singoredjo, L., Slagt, M., van Wees, J. et al. (1990) Catal. Today, 7, 157. [73] Grzybek, T. and Papp, H. (1992) Appl. Catal. B Environ., 1, 271. [74] Pasel, J., Käßner, P., Montanari, B. et al. (1998) Appl. Catal. B Environ., 18, 199. [75] (a) Zhu, Z., Liu, Z., Liu, S. et al. (1999) Appl. Catal. B Environ., 23, L229. (b) Zhu, Z., Liu, Z., Niu, H. et al. (1999) J. Catal., 187, 245.

Development of Catalytic Processes for the Selective Catalytic Reduction of NOx into N2

23

[76] Long, R.Q., Yang, R.T. and Zammit, K.D. (2000) J. Air & Waste Manage. Assoc., 50, 436. [77] Avila, P., Barthelemy, C., Bahamonde, A. et al. (1993) Atmospheric Environment, Part A: General Topics, 27A, 443. [78] Tronconi, E., Forzatti, P., Gomez Martin, J.P. et al. (1992) Chem. Eng. Science, 47, 2401. [79] Lietti, L., Nova, I., Camurri, S. et al. (1997) AIChE J., 43, 2559. [80] Tronconi, E., Cavanna, A., Orsenigo, C. et al. (1999) IEC Research, 38, 2593. [81] Ramis, G., Busca, G., Bregani, F. et al. (1990) Appl. Catal., 64, 259. [82] (a) Fritz, N., Mathes, W., Zürbig, J. et al. (1999) SAE Paper, 1999-01-0111. (b) Miller, W., Klein, J., Mueller, R. et al. (2000) SAE Paper 2000-01-0190. [83] Gieshoff, J., Schäfer-Sindlinger, A., Spurk, P.C. et al. (2000) SAE Paper 2000-01-0189. [84] Miessner, H., Francke, K.-P. and Rudolph, R. (2002) Appl. Catal. B Env., 36, 53. [85] Hoard, J. (2000) J. SAE Diesel Engine TOPTEC, September, Indianapolis, IN. [86] (a) Chen, Z. and Mathur, V.K. (2003) IEC Research, 42, 6682. (b) Kwak, J.H., Szanyi, J. and Peden, C.H.F. (2003) J. Catal., 220, 291. [87] Yoon, S., Panov, A.G., Tonkyn, R.G. et al. (2002) Catal. Today, 72, 243. [88] Mok, Y.S., Koh, D.J., Kim, K.T. et al. (2003) IEC Research, 42, 2960. [89] Broer, S. and Hammer, T. (2000) Appl. Catal. B Env., 28, 101. [90] Miessner, H., Francke, K.-P., Rudolph, R. et al. (2002) Catal. Today, 75, 325. [91] Francke, K.-P., Miessner, H. and Rudolph, R. (2000) Catal. Today, 59, 411. [92] Hayashi, Y. (2003) Oyo Butsuri, 72, 448. [93] (a) Miyoshi, N., Matsumoto, S., Katoh, K. et al. (1995) SAE Technical Paper 950809. (b) Matsumoto, S., Ikeda, Y., Suzuki, H. et al. (2000) Appl. Catal. B Environ., 25, 115. [94] Martin, B. (2005) IFP – Panorama. [95] Matsumoto, S. (2006) Plenary Lecture at TOCAT5, Tokyo, July. [96] Ishii, M., Matsumoto, S., Stone, P. et al. (2002) EUCHEM Conference on Environmental Catalysis, 28 November–1 December, Göteborg, Sweden. [97] (a) Centi, G., Fornasari, G., Gobbi, G. et al. (2002) Catal. Today, 73, 287. (b) Fornasari, G., Livi, M., Trifirò, F. et al. (2002) Catal. Today, 75, 421.

This page intentionally left blank

PART 1

A molecular view of reactions involved over DeNOx catalysts – Mechanisms and kinetics

This page intentionally left blank

Chapter 2

DFT MODELING AND SPECTROSCOPIC INVESTIGATIONS INTO MOLECULAR ASPECTS OF DeNOx CATALYSIS P. Pietrzyka and Z. Sojkaab∗ a

Faculty of Chemistry, Jagiellonian University, Ingardena 3, 30-060 Krakow, Poland b

Regional Laboratory for Physicochemical Analyses and Structural Research, Jagiellonian University, Ingardena 3, 30-060 Krakow, Poland ∗

Corresponding author: Faculty of Chemistry & Regional Laboratory for Physicochemical Analyses and Structural Research, Jagiellonian University, Ingardena 3, 30-060 Krakow, Poland. Tel.: +48 12 663 22 95, Fax.: +48 12 634 05 15, E-mail: [email protected]

Abstract Even though the nitric oxide is thermodynamically unstable toward decomposition (2NO → N2 + O2 , r G = −413 kcal/mol) and disproportionation (4NO → 2NO2 + N2 , r G = −578 kcal/mol; 4NO → O2 + 2N2 O, r G = −327 kcal/mol; 3NO → NO2 + N2 O, r G = −246 kcal/mol), none of these reactions can occur to any appreciable extent without a catalyst, due to the quite large kinetic stability of NO [1]. However, despite of being thermodynamically favored, catalytic abatement of NOx still remains a challenging problem from both the fundamental and the practical viewpoints [2,3]. Indeed, a number of mechanistically important issues concerning the active site design, the nature of the key intermediates, as well as detailed molecular understanding of the reaction route have not been definitely resolved as yet. The wide scientific interest in the transition-metal ion (TMI)-exchanged zeolites stems from their unique nitrosyl chemistry and remarkable activity in direct decomposition of NOx , which is combined with the advantage of exhibiting a relatively simple structure. This makes such materials convenient model systems for basic mechanistic studies and quantum chemical modeling [4–10]. In this context, the interfacial chemistry of nitrogen oxides (NO, NO2 , and N2 O) with intrazeolite TMIs acting as the active sites of DeNOx reaction is a subject of the fundamental importance for elucidation of the reaction elementary processes such as coordination, charge and spin density redistribution, accompanying the N−N and O−O bond making. A noteworthy feature of those processes is a dramatic spin change on passing from reactants to products, which along with the orbital symmetry barrier creates principal molecular constraints for efficient decomposition of nitrogen oxides in a conceivably simple concerted way. The aim of this contribution is to provide a molecular insight through density functional theory modeling corroborated with spectroscopic investigations (both in static and flow regimes) into the binding and activation of nitrogen oxides on various TMIs of different

Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

28

Past and Present in DeNOx Catalysis electron configuration (d5 –d10 ) and spin multiplicity, which are heterogenized in siliceous and alumino-siliceous frameworks. In this chapter, we will illustrate a variety of mechanistic aspects of DeNOx process using selected examples coming from our laboratory and from literature, with emphasis on the CuZSM-5 system.

1. COMPUTATIONAL SPECTROSCOPY AND MOLECULAR MODELING Molecular quantum chemical modeling of heterogeneous catalytic systems is a rapidly developing field, and several easy-to-use quantum chemical packages are nowadays available for wide applications [11–16]. However, despite of the progress made, for the large non-molecular catalytic systems often containing transition-metal ions (TMIs), such calculations are computationally quite demanding. Various approximations and truncation schemes are then commonly used to achieve sensible results at reasonable computational costs. One of the most fruitful strategies of molecular modeling in catalysis is based on the feedback between the experiment and computations. However, for complex heterogeneous systems, the connection between the spectrum and the structure is often unclear. As a result, a great deal of important molecular information about the investigated active sites and intermediate species is available only with the help of complementary quantum chemical calculations. The computational spectroscopy approach is based on the advanced computer analysis and simulation of experimental spectra, allowing for reliable extraction of their parameters, joined with the density functional theory (DFT) calculations of these properties. Owing to a progress in the relativistic DFT methods, for large TMI containing non-molecular systems often nearly quantitative reproduction of various spectroscopic parameters is made possible at reasonable computing costs. Using the double perturbation approximation [17], spectral parameters can be expressed in terms of mixed second derivatives of the total energy (E) with respect to the corresponding variables: the electric field (E) and the nuclear coordinates (qi ) for IR, and the magnetic field (B) along with the electron (S) and nuclear (I) spins for electron spin resonance (ESR). In the case of IR active transitions, their frequencies are given by ij = 2 E/qi qj , whereas the intensities can be assessed from the equation Ii ∼ 2 E/Eqi 2 . Commonly used generalized gradient approximation methods usually give intensities accurate within ±10% only, and the double-harmonic approximation introduces further uncertainty of about ±10% [18]. The spin-restricted and spin-unrestricted zeroth-order regular approximation (ZORA) [19] and the scalar relativistic Pauli [20] method are used for g tensor calculation, expressed in general double perturbation equation as gij = 2 −1 e  E/Bi Sj B=S=0 . The non-relativistic computations of hyperfine coupling constants (N Aij = 2 E/INi Sj I=S=0  are based on the standard spin density formulation. Usually the value of the Iasym /Isym ratio is used for estimating the angle between geminal ligands in the {M(XY)2 }n complexes, based on the approximate formula of Cotton [21]: Iasym /Isym = tg 2  /2

(1)

However, this equation is valid only in the case when the dipole moment of the adsorbed XY molecule and the direction of the bond, defined by the arrangement of the

DFT Modeling and Spectroscopic Investigations

29

Spectra Structure 1

+

Cu + 2NOg

NO EPR and IR spectroscopy

νxy A

exp N, ij

g ijexp

∂ 2E ∂IN,i ∂Sj 2 1 ∂ E gij = βe ∂Bi ∂Sj 2

AN,ij =

min F [Y(B), y (B), w ]

νij =

DFT calculations E = E [ρ(r )]

∂E ∂qi ∂qj

η1{2CuNO}11 1 3

12

NO trans - η { CuN2O2} trans - η1{1CuN2O2}12

cis- η1{1CuN2O2}12

Parameters

Reaction modeling Computational spectroscopy

function catalytic tests TPSR, KIE k , E a, TOF

Reactivity studies

Figure 2.1. Schematic presentation of the research methodology combining spectroscopic investigations, reactivity studies, and DFT molecular modeling adopted in this work.

M−X−Y atoms, are collinear. For two angularly bound NO ligands, the actual N−M−N angle, obviously, cannot be estimated using Eqn. 1 as sometimes assumed. In our method of exploring the mechanism of catalytic reactions, we combine molecular modeling with computational spectroscopy and reactivity investigations using functional model systems capable of mimicking the key properties of real catalysts (Figure 2.1). Since for the heterogeneous systems, the structure of the active sites and surface intermediates is not distinctly defined, therefore judicious selection of appropriate models and adequate calculation schemes are of great importance for obtaining sensible results. Spectral parameters of the postulated active sites, reaction intermediates, and spectators, inferred from the reactivity and spectroscopic studies, are calculated and directly compared with available experimental data (magnetic EPR parameters, IR frequencies, and intensities). Such approach provides not only a quantitative bridge between the molecular structure of the investigated species and their spectroscopic fingerprints, but it can also be used for guiding molecular modeling of the investigated catalytic reactions, providing useful reference frame justifying the adequacy of both the adopted model and the method as well [22,23].

2. MODEL SYSTEMS CONTAINING TRANSITION METAL IONS The active sites in our investigations are constituted by mononuclear surface transitionmetal complexes (corresponding to low TMI loadings). Empirical models of such sites

30

Past and Present in DeNOx Catalysis

in the zeolite or silica hosts can be prepared in a well-defined state by grafting or ionic exchange methods, as described elsewhere in more detail [24,25]. The molecular modeling was carried out within the cluster model approach. The following first series TMI of different electron configuration and spin multiplicity encaged in the ZSM-5 zeolite or molecularly dispersed on the silica surface were chosen for the DFT investigations: Fe3+ , Mn2+ , Cr+ (d5 , 6 S), Fe2+ (d6 , 5 D), Co2+ (d7 , 4 F), Ni2+ (d8 , 3 F), Ni+ , Cu2+ (d9 , 2 D), and Cu+ , Zn2+ (d10 , 1 S). Due to the large number of ions under study, only selected exchangeable positions within MFI and SiO2 frameworks were considered explicitly. Following Sauer et al. [26], we use the notation Yn, where Y = M (for the main channels), Z (for the zig-zag channels), and I (for the channel intersections) indicates the location of the hosting site within the MFI framework, whereas n defines a number of the T atoms used for building the cluster. The list of clusters cut-off from the ZSM-5 structure applied in this study is shown in Table 2.1 along with the positions of the Al atoms. Their spatial placement and structural relation with the MFI framework are shown in Figure 2.2a. The M5 site was mimicked with a [Si4 AlO5 (OH)10 ]− cluster having the Al atom in the T1 position. This site is able to accommodate monovalent cations, such as Cu(I) [26,27], while for divalent TMI the M7 and Z6 hosts are chosen [27,28]. The Al placing within the cluster corresponds to the energetically most stable positions [26]. The Z6 site has been claimed to be the most occupied in the case of cobalt ions [29]. The channels intersections in the ZSM-5 zeolite are represented by I2 and I10 clusters (Table 2.1). The I2 cluster, despite its small size and relatively high symmetry, has been successfully used for revealing molecular details of NO and NO2 bonding to the Cu(I) sites in ZSM-5 [30–32], whereas the I10 cluster gives a unique opportunity for probing the Table 2.1. Selected TMI hosting sites within the ZSM-5 framework, their notation, and the corresponding cluster epitomes Localization Main channel (M) Zig-zag channel (Z) Channel intersection (I)

Cluster type

Position of the Al atom

Cluster composition

M5 M7 Z6 I2 I10

T1 T1 (T7) T7 (T10) T6 T6

[Si4 AlO5 (OH)10 ]2− [Si6 AlO8 (OH)12 ]2− [Si5 AlO6 (OH)12 ]2− [Si2 AlO2 (OH)8 ]− [Si9 AlO10 (OH)6 H14 ]−

(a)

(b) I2

M7

3T 2T

I10

5T

6T

M5

Z6

4T

Figure 2.2. Definition of the hosting sites and their localization within (a) the ZSM-5 framework and (b) on the surface of amorphous silica.

DFT Modeling and Spectroscopic Investigations

31

influence of the supramolecular interactions of the adsorbate with the zeolite framework, since it represents a whole aperture of the main channel. The structure of the optimized copper(I) intrazeolite complexes in various hosting sites, shown in Figure 2.3, may serve as an illustration of the TMI accommodation in the clusters specified above. The site I2 (Figure 2.3a) exhibits a symmetrical bidentate coordination with the average Cu−Oz bond distance dCu−O  = 2029 Å and the Oz −Cu−Oz angle of 93 . Similar coordination, but with slightly longer Cu−Oz bonds (dCu−O  = 2062 Å), is observed for the more robust I10 site (Figure 2.3f). In the case of the M5 site formed by a 10-membered ring, the copper ion exhibits a nearly planar asymmetric trigonal bonding geometry making two longer links with Oz {Al,Si} oxygens proximal to aluminum atom (dCu−O  = 2105 Å), and one slightly shorter bond with Oz {Si}2 oxygen situated in front of Al (Figure 2.3b). The copper center Z6 (Figure 2.3c) is defined by a slightly deformed 12-membered ring situated on the walls of the sinusoidal channels. The copper ions in a distorted trigonal pyramid arrangement exhibit two shorter Cu−Oz bonds (dCu−O  = 2078 Å) and a longer one (dCu−O = 2329 Å).

(a)

(b) 2.034

2.024

(c)

2.115

2.065 2.043 2.329

2.145

(d)

2.040

(e) 2.046 2.184 2.487

2.018 2.180 2.156

(f)

(g) 1.820

1.990 2.020

2.075 2.048

Figure 2.3. Optimized geometries [DMol, Becke1988 exchange and Perdew–Wang correlation, (BP), double numerical plus polarization functions basis set (DNP)] of copper(I) complexes encaged in ZSM-5 zeolite or grafted on silica: (a) {CuI }I2 complex, (b) {CuI }M5 complex, (c) {CuI }Z6 complex, (d) {CuI }M75T complex, (e) {CuI }M76T complex, (f) {CuI }I10 complex, and (g) {CuI }silT5 complex. Bond lengths are given in Å.

32

Past and Present in DeNOx Catalysis

In the case of the M7 site (composed of two conjoint 5T rings with three common T atoms) localized on the walls of the main channels, two possible bonding schemes CuI M75T and CuI -M76T are shown in Figure 2.3d, e. These two variants of the intrazeolite complexes correspond to the different arrangements of copper, which is accommodated either within one of the two joined 5T rings or within the atop T6 ring [33]. Those results reflect the well-known speciation of copper in ZSM-5 imposed by the rigid framework of the zeolite functioning as a macroligand, and remain in a good agreement with experimental XAFS, IR, UV-vis and photoluminescence data [25]. Table 2.2 contains mean values of the Cu−Oz bond lengths (dCu−O , copper valence index [11] (VCu ) and partial charge (QCu ), one-electron energies of the HOMO and LUMO (HOMO , LUMO ) levels along with the bonding energy of CuI to the hosting cluster. From the comparison of the results, it can be inferred that copper ions exchanged in the ZSM-5 zeolites assumes a bidentate (sites I2 and I1) or tridentate coordination (sites M5, Z6, and M7). These two groups differ also in the molecular properties (Table 2.2). The I-centers are characterized by lower values of the valence index and greater partial charges, QCu , in comparison to the M and Z centers, which is associated with the deeper laying HOMO and LUMO levels. In the M5, Z6, and M7 sites CuI ions exhibit more covalent character, and the frontier orbitals have less negative energies. As a result, the chemical hardness of the I-centers, located at the channel intersections, is smaller than those located on the walls of the ZSM-5 zeolite. Hosting sites of amorphous silica are shown in Figure 2.2b. For modeling surface complexes of FeIII , NiII , and CuI/II supported on SiO2 , a silT5 cluster of the [Si5 O8 H8 ]2− stoichiometry in a chair conformation (Figure 2.3g) was adopted [34]. This model has been successfully used for description of the active sites in {Ni}SiO2 [35] and {Cu}SiO2 catalysts [36]. For further analysis of the whole d5 –d10 TMI series, due to the large number of the structures, only some selected cluster are explicitly taken into account. A survey of the optimized geometries of the intrazeolitic 3dn complexes hosted in various zeolite and silica derived sites in the ground state is shown in Figure 2.4, whereas their structural and electronic properties are summarized in Table 2.3. Mean values of the metal−oxygen bond lengths (dM−Oz , valence indexes (VM − FVM ) corresponding to the cumulated metal−zeolite bond order, partial metal charge and spin density (QM and M ), oneelectron energy levels of the frontier orbitals (HOMO , LUMO ), as well as the adiabatic energy distance between the highest and the first lowest spin states (EHS→LS  for a given electron configuration are also listed. Table 2.2. Selected structural and geometrical parameters for various CuI centers in ZSM-5 Center I2 M5 Z6 M75T M76T I10

dCu−O  (Å)

VCu

QCu (au)

2029 2084 2161 2184 2173 2062

1083 1364 1321 1305 1317 1105

0402 0312 0319 0327 0333 0416

HOMO /LUMO (eV) −5626/−3444 −5183/−2327 −5417/−2648 −5648/−2603 −4988/−2506 −5679/−3613

Ebond (kcal/mol) −160 −167 −169 −157 −161 −162

DFT Modeling and Spectroscopic Investigations

33

Valence state

+1

+2

+3

d5

d6

d7

d8

d9

d 10

{2NiI}Z6

{1CuI}M5

{2CuII}Z6

{1ZnII}Z6

Electron configuration {6CrI}M5 {6MnII}Z6

{5FeII}Z6

{4CoII}Z6

{1NiII}Z6

{6FeIII}siIT5

Figure 2.4. Survey of the structures of TMI hosted in various zeolite- and silica-derived clusters as a function of the electron configuration obtained by VWN/DNP method (after [37]). Table 2.3. Comparison of the selected molecular parameters for intrazeolitic complexes of TMI TMI Cr+ 3 Cr+ 3 6 Mn2+ 4 Mn2+ 2 Mn2+ 6 Fe3+ 4 4 Fe3+ 4 2 Fe3+ 4 5 Fe2+ 3 Fe2+ 4 Co2+ 2 Co2+ 3 Ni2+ 1 Ni2+ 2 Ni+ 2 Cu2+ 1 Cu+ 3 1 Zn2+ 6

Electron dM−Oz  configuration (Å) d5

4

1

d5

d5

d6 d7 d8 d9 d10

2152 2059 2162 2026 1996 1798 1858 1875 1889 1921 2064 1994 2050 1960 1994 2063 2017 1974

VM – FVM 1 1392 1968

3259

2454 3940 2500 2420 2355 2175 1442 1842

QM (au) M (au) HOMO /LUMO (eV) 0309 0339 0474 0318 0247 0535 0311 0290 0313 0186 0346 0253 0203 0176 0057 0490 0299 0577

4.831 −3228/−2008 −3326/−2352 4.759 −5947/−4224 −4144/−3900 −4889/−4688 3.950 −6526/−6192 −5529/−5094 −5724/−5653 3.510 −4828/−4612 −4411/−4267 2.551 −5982/−5790 −5451/−5449 1.343 −5570/−5177 0 −5775/−5261 0.879 −3925/−3876 0.484 −6457/−6245 0 −4919/−2011 0 −6523/−2957

Difference value of total valence and free valence indices Adiabatic energy distance between the high (HS) and low spin (LS) 3 Calculated for M5 site 4 Calculated for silT5 site 2

EHS→LS 2 (kcal/mol) −238 −271

−143

−232 −32 ∼0

34

Past and Present in DeNOx Catalysis (a)

(b)

(c)

Figure 2.5. Kohn–Sham (a) HOMO orbital and (b) spin density contours calculated for the {2 CuII }Z6 complex, along with (c) spin density contour for the {3 Cu−O}M5 site.

According to our calculation results, both electron and spin state influence the structure of a given surface TMI. In general, the coordination number decreases upon passing to higher spin multiplicity, whereas the effective charge on the encaged metal (the QM value) distinctly increases. A trigonal coordination is characteristic of d5 and d10 states, while tetragonal one was found for the d6−9 configuration. Due to the weak ligand field generated by oxide frameworks, the intrazeolite TMI exhibit a high-spin state. Indeed, the values of EHS→LS clearly confirm a weak character of the alumino-siliceous and siliceous macroligands. They rapidly decrease with the decrease of the spin multiplicity so that for the d8 state the spin triplet and the singlet become nearly degenerate. Lower oxidation states of the same cations are associated with reduction of the coordination number from four to three (e.g., CuII and CuI or NiII and NiI ) and decreasing of the valence index. Moreover, they exhibit partially covalent nature, as inferred from the QM values much lower than for the corresponding isolated ions. Bond orders change from 0.95 for {6 FeIII }5T to 0.28–0.22 for {6 CrI }M5, the valence index (VM – FVM  can be therefore identified with the bonding strength of TMI to the hosting site, and in consequence with the degree of local distortion of the hosting framework caused by the coordination [38]. Molecular nature of the TMI interaction with the surface can be deduced from the pronounced mixed metal-lattice oxygen character of the Kohn–Sham orbitals responsible for the bonding, which is well illustrated by the HOMO contour in the case of the {2 CuII }Z6 complex (Figure 2.5a). The lowering of the TMI partial charge upon coordination, already described for other zeolites [39], indicates that substantial zeolite-to-metal charge transfer and reversed metal-to-zeolite spin transfer take place (Figure 2.5a, b), reaching the level up to 50% for copper. Thus, the zeolite lattice can act as a reservoir of electron density, making the hosted cations considerably softer, which is important for their DeNOx reactivity (vide Section 4).

3. GENERIC SCENARIO OF deNOx PROCESSES The backbone of the DeNOx process over mononuclear TMI encaged in zeolites can be epitomized in the form of three interconnected cycles associated with the formation of the N2 and O2 reaction products (Figure 2.6), inferred from the steady state and transient rate data combined with spectroscopic evidence for surface species and

DFT Modeling and Spectroscopic Investigations NO

on {M-NO3}Z

{M-O2}Z O2

NO

{M}Z {M-O}Z {M-NO}Z NO

O2

NO2 {M-NO2}Z

{M-O2}Z

{M-(NO)2}Z

N2O

N2O NO

{M-O}Z

{M}Z

N2

n

{M-N2O}Z

N bond forma N– ti o

N2

{M-N2O2}Z

O bond forma O– ti

35

Figure 2.6. Generic mechanism of NO decomposition over mononuclear intrazeolite transition metal ions.

active sites [4,5,40,41]. The energetic feasibility of NO decomposition on isolated sites has been confirmed in numerous computational studies [6,42–44]. The principal mechanistic events include N−N bond formation stage, where the coordinated NO reactant is transformed into the N2 O semi-product via dinitrosyl ({M−(NO)2 }Z) or dinitrogen dioxide ({M−N2 O2 }Z) intermediates, depending on the nature of TMI (vide infra). Simultaneously, the primary {M}Z active sites are converted into the secondary {M−O}Z active sites involved in the dioxygen formation cycle [5]. The mononitrosyl complexes are usually postulated to be the key intermediate species of this step [2,5,41], whereas the mechanistic role of dinitrosyls and dinitrogen dioxide is more indistinct as yet. Decomposition of N2 O into the constituent elements can occur either via oxygen transfer to the metal-oxo centers, restoring the initial active sites upon O2 release, or to the intact primary centers {M}Z. This latter step is supposed to require a molecular transport of nitrogen in the form of N2 O between the different active sites [5]. Oxidation of gas phase NO on the metal-oxo centers leads to NO2 , which is the central semi-product of the O−O bond formation cycle, involving nitrates as the key intermediates [5,25]. In the proposed mechanism, NO2 acts as an oxygen carrier allowing for efficient kinetic communication between distant active centers, without requiring direct diffusion of oxygen to produce the key nitrates [40]. Thermal decomposition of the surface {M−NO3 }Z species into O2 and NO restores the initial {M}Z sites. As a result, during the whole process the active site alternates between the primary {M}Z and the secondary {M−O}Z centers, which drive both the N2 and O2 production cycles. While alternative scenarios involving dinuclear active sites also exist [5,40,45], we focus on the better defined model involving the {M}Z/{M−O}Z couple, and by assuming these entities to be the active species, we develop a consistent mechanistic description of DeNOx process catalyzed by mononuclear TMIs.

36

Past and Present in DeNOx Catalysis

4. NATURE OF THE ACTIVE SITES Establishing the molecular mechanism by which surface TMI afford NO decomposition is a fundamental problem for rational design of the optimal active centers. Apparently, there is a missing link between the electronic and magnetic structure of the {M}Z and {M−O}Z active sites and their DeNOx reactivity, which needs further elucidation. Formally, these two centers are mutually related by shuttling of an extra-lattice oxygen (ELO) atom. This is concomitant with the change in the spin state and appearance of constraints related with the intersystem crossing, implying a TSR (two-state reactivity) mechanism, discusses elsewhere [46]. Our specific focus is on the influence of the electronic and spin configuration of both active sites and on the reactant ligation. Thus, taking again the intrazeolite cuprous site as an example this process can be written as

CuI Z + O3 P → CuI −O3 PZ ↔ CuII −O− Z

(2)

entailing a change in the spin state from singlet to the triplet one, with the triplet oxygen atomic state 3 P chosen to conserve electron spin. Figure 2.7 shows an approximate state correlation diagram of such process along with the contours of the relevant MO of both copper and copper-oxo sites. Conceptually, the reaction pathway for such process is in part on the singlet surface and in part on the triplet one, with the energy barrier determined by the intersection locus, which is closer to {Cu−O}Z than {Cu}Z. The probability of transferring from one surface to the other, gauged by the Landau–Zener coefficient [47], restrains the rate of this process and can be enhanced using spin catalysis [48]. Repartition of the spin density within the copper-oxo unit obtained from DFT calculations (BP/DNP) with the largest fraction localized on the oxygen moiety (Figure 2.5c) indicates that the {CuI −O(3 P)}Z ↔ {CuII −O− }Z resonance spin equilibrium is noticeably shifted to the left, so the ELO can indeed be regarded approximately as a bound O(3 P) atom.

4s

{3CuI}M5

T S 2.3 eV

{1Cu-O}M5

3d electron manifold

0.4 eV

T

0.6 eV

{3Cu-O}M5

{1CuI}M5

S

Figure 2.7. Simplified energy diagram of copper and copper-oxo sites (BP/DNP), showing relative positions of the triplet and singlet states, along with the relevant Kohn–Sham orbital contours.

DFT Modeling and Spectroscopic Investigations

37

The chemical function of zeolite lattice plays an important latent role in rendering this process more facile. The CuI (3d10  ion binds to the zeolite framework basically by the electrostatic forces [49]. These attractive interactions are counterbalanced by exchange repulsion between the closed shell of CuI and the non-bonding electron pairs on the zeolite oxygen atoms, which push the 3d electron manifold up in energy, so that they can interact with the frontier NO orbitals more effectively. A particular feature of the encaged copper site is an energetic proximity of the 4s and 3d orbitals controlling the excitation energy into the triplet state, brought about by the interaction with the zeolite framework. It is reduced from 2.47 eV in the bare state to 193 eV for zeolite embedded state [49]. This allows for partial 3d-4s hybridization [50], which is a prerequisite for opening of the d10 shell of the monovalent copper, facilitating binding of the reactants and the necessary spin changes, reducing at the same time the repulsion between CuI and the zeolite framework. The strength of the metal−ELO bonding is crucial for being catalytically functional. In the early stages of DeNOx process, a strong bond favors oxygen transfer form the intermediary N2 O to the {M}Z center, but in the later stages, it makes the abstraction of oxygen to yield gas phase O2 clearly more difficult (Figure 2.6). For early TMI, the electrons occupy the bonding  and  orbitals, localized largely on the ELO moiety, making the M−O bond quite strong. In the case of late TMI, with the increasing number of electrons, their antibonding counterparts having the most electron density on the metal are also occupied. Filling those orbitals progressively weakens the M−O bond, promoting desorption of bound dioxygen (Figure 2.6), and restitution of the initial {M}Z active sites. Thus, an apposite balance of the M−O bonding energy provides a thermodynamic rational for the well established DeNOx activity of the Cu, Co, Fe, or Ru and inertness of the Sc, Ti, or V ions.

5. INTERFACIAL COORDINATION CHEMISTRY OF NITRIC OXIDE The catalytic implications of the alternative coordination modes of nitric oxide in TMI complexes were first noted by Collman already in the late sixties [51]. Therefore, to understand the reactivity of bound NO, one has to consider first the molecular and electronic structures of surface TMI nitrosyl complexes and the interrelationship of nitrosyl bonding modes and coordination geometries. Furthermore, since the molecular 2* orbitals (which along with 5 MO are primarily responsible for M−NO bonding) are similar in energy to the metal 3d orbitals, their mutual overlap is sizeable, and any assignment of electrons in the corresponding complex to one fragment or to the other is quite arbitrary. This dichotomy has been formally overcome by using the m {MNO}n notation in which n is the number of electrons on the metal and the 2* orbitals of the NO ligand and m indicates its hapticity.

5.1. Binding and bending of NO As already mentioned, the chemical reactivity of the nitrosyls is strongly related to the type of bonding, because it controls the rehybridization of NO-based orbitals and associated significant redistribution of the electron and spin densities within the M−NO moiety.

38

Past and Present in DeNOx Catalysis

This has been well documented both experimentally (mainly by IR and EPR spectroscopies) and theoretically by means of molecular DFT modeling [3,22,24,41,49,52].

5.1.1. Mononitrosyl complexes Traditionally, the bonding in the M−NO moiety is described by assigning formal oxidation states for the metal and the ligand moieties. In such approach, two extreme modes of bonding vary from linear 1 complexes, often associated with the sp hybridization of the ligand, to bent complexes related with the spn (1 < n ≤ 2) hybridization, and further through kinked conformation to 2 coordination [53]. The process culminates in oxidative addition in which the N−O bond is broken. However, such a simple picture may be misleading, and it is very difficult to predict a priori the preferential geometries for the whole range of complexes and correlate them with the formal charge assignment. Thus, any in-depth understanding of the nature of the M−NO bond would require a more adequate quantum chemical approach. Our DFT calculations revealed that coordination of nitric oxide to the series of intrazeolite TMI leads to the formation of the bent 1 {MNO}n adducts of various spin states exhibiting generally the Cs microsymmetry with mirror plane defined by the M−N−O moiety. Optimized structures of some representative mononitrosyl complexes are depicted in Figure 2.8, and their selected geometric parameters and molecular properties are listed in Table 2.4. For the sake of comparison, the calculated molecular properties of gas-phase NO are also given: dN−O = 1169 Å, N−O bond order bN−O = 2145, harmonic frequency (VWN/DNP) NO = 1904 cm−1 (experimental NO = 1876 cm−1 , experimental harmonic NO = 1904 cm−1 [54]).

1.145 1.161

130°

1.992

2.144

{4FeNO}7Z6

1.973

2.042

1.952

{2NiNO}9Z6 1.167

1.7 55

1.181 151°

2.035

{1CuNO}10Z6

2.158

2.012

1.773

1.954

45

1.155 123° 1.983

1.8

1.693

168°

1.956

121°

1.991 2.065

{2CuNO}11M5

2.023

1.985

{2ZnNO}11Z6

Figure 2.8. Optimized geometries (DMol, VWN/DNP) of some selected mononitrosyl complexes. All bond lengths are given in Å, and angles, in degrees.

Adduct

dM−N (Å)

M−NO  )

QM

QNO

M

NO

dN−O (Å)

bN−O

pN−O

{3 CrNO}6 {3 MnNO}6 {3 FeNO}6

1.651 1.640 1.662

171 167 153

+0.099 −0143 −0227

−0280 +0.045 +0.073

2.470 2.449 1.894

−0514 −0509 −0325

1208 1171 1173

1752 1947 1913

0021 0042 0056

−4073/−3092 −6453/−4885 −6515/−5654

{4 FeNO}7 {1 CoNO}8 {2 NiNO}9

1.693 1.658 1.773

168 141 123

−0126 −0141 −0049

+0.118 +0.138 +0.134

3.209 0 0.775

−0576 0 −0110

1161 1164 1155

2031 1995 2141

0034 0059 0053

−6287/−5575 −5994/−5356 −6291/−5653

{1 NiNO}10 {1 CuNO}10

1.601 1.845

165 130

+0.073 −0031

+0.065 +0.249

0 0

0 0

1166 1145

1950 2200

0068 0008

−5726/−3956 −6252/−5689

{2 CuNO}11 {2 ZnNO}11

1.755 2.158

151 122

+0.082 −0083

−0020 0.171

0.120 0.056

0.932 0.919

1181 1162

2097 2374

0020 0028

−4551/−4295 −5994/−5681

HOMO /LUMO (eV)

DFT Modeling and Spectroscopic Investigations

Table 2.4. Comparison of the molecular properties for the mononitrosyl complexes of selected TMIs encaged in ZSM-5 zeolite (VWN/DNP)

dN−O – bond distance; QM , QNO – partial charge changes on metal and NO ligand; M , NO – spin density on metal and NO ligand; M−NO – M−N–O bending angle; bN−O – NO bond order; pN−O – polarization of NO bond.

39

40

Past and Present in DeNOx Catalysis

In the ground state, the coordination number (CN) equals to 4 or 5 for all complexes except of {CuNO}11 for which CN = 3. The coordination polyhedron resembles distorted tetrahedron or trigonal pyramid with NO in the apical position. The NO coordination is an isodesmic process (i.e., preserving the initial value of the metal CN) with exception of the {1 ZnII }Z6 site, where an associative adsorption takes place. The strongest bending is observed in the case of {2 ZnNO}11 and {2 NiNO}9 species, reaching practically the sp2 -limiting value of 120 , whereas the nearly linear {3 CrNO}6 complex assumes the sp-extreme structure (Table 2.4). From the inspection of the data in Table 2.4, it is clear that NO changes its original molecular character after adsorption. In general, coordination of nitric oxide leads to a pronounced redistribution of the electron and spin densities, accompanied by modification of the N−O bond order and its polarization. Thus, in the case of the {MNO}7−10 and {ZnNO}11 species, slender shortening of the N−O bond is observed, whereas for the {MNO}6 and {CuNO}11 complexes it is distinctly elongated. Interestingly, polarization of the bound nitric oxide assumes its extreme values in the complexes of the same formal electron count ({NiNO}10 and {CuNO}10  exhibiting however different valence. The most pronounced changes were observed for the spin and electron density allocation within the {MNO} unit. Analysis of the results shows that participation of the zeolite framework in the electron density repartition upon NO bonding is clearly dictated by the electron count. Thus, for the {MNO}67 complexes the zeolite framework is acting as an electron donor, for the {CoNO}8 one the charge is redistributed essentially within the Co−NO unit, whereas in the case of {MNO}9−11 the hosting site is acting as an acceptor of the electron density. The largest involvement of the framework in charge redistribution was observed for the {CuNO}10 complex, which is associated with highly covalent nature of the CuII −Oz bond (vide infra). Dramatic variations are also observed in the spin density redistribution in the {MNO}n complexes. For the open-shell metals, the original spin of the intact nitric oxide is lost, and the NO ligand exhibits a negative spin density (due to polarization by the paramagnetic metal center). Consequently, the latter assumes the highest values for complexes of the highest spin multiplicity. Different behavior is observed for the closed-shell metals, where the original spin is essentially retained on the ligand ({CuNO}11 and {ZnNO}11 . It is worth noting that changes in the spin and the electron densities are not mutually correlated. The spin density distribution and the structural characteristics of the investigated mononitrosyls provide a useful rational for assignment of the NO-adducts into three generic classes: (1) ligand-centered radical complexes (1 {CuNO}11 , 1 {ZnNO}11 , (2) metal-centered radical complexes (1 {Co(NO)2 }9 , 1 {NiNO}9 , 1 {FeNO}7 , 1 {FeNO}6 , and (3) diamagnetic 1 {CoNO}8 , 1 {CuNO}10 , 1 {NiNO}10 complexes, which has an important impact on their chemical reactivity. The relationship between two selected activation parameters, namely the change in the N−O bond length (dN−O  and the NO partial charge (QNO  acquired upon adsorption are shown in Figure 2.9. It is apparent that the N−O distance is correlated with the charge flow, which can be assigned to antibonding character of the partially occupied NO orbital. Its population results in the weakening of the N−O bond and the consequent elongation. The areas I and II in Figure 2.9 correspond to oxidation of NO ligand and shrinking (I) or minute elongation (II) of the internuclear distance. There are only two cations

DFT Modeling and Spectroscopic Investigations

41

0.3 0.2

Zn2+

Cu2+

Q NO

0.1 0

Mn2+

Co2+

Fe3+

Ni2+

I

II

Ni+

–0.1

III

Cu+

Fe2+

–0.2 –0.3 –0.03

Cr+

–0.015

0

0.015

0.03

0.045

ΔdN–O /A

Figure 2.9. Correlation between change of the N−O bond length (dN−O  and the partial charge (QNO  accumulated on the NO ligand upon interaction with TMI encaged in ZSM-5.

with the semi-occupied d shell (Cr+ ) and with the fully occupied d shell (Cu+ ) that are located in the region III. Their characteristic feature is a specific localization of the 3d orbitals in proximity to the NO orbitals, due to the low-valent state of both metals, which is further reduced by the interaction with the hosting site (vide supra). Although Figure 2.9 apparently accounts for the dramatic difference in the reactivity of the CuI and CuII cations, the weakening of the N−O bond, despite of being claimed a simple measure of the NO activation [33], is actually not related to reaction coordinate of the DeNOx process. Indeed, following the mechanistic scenario (described in Section 3), the reaction is not initiated via N−O bond braking but through N−N bond making. A more in-depth insight into the NO ligation can be obtained from the analysis of the DOS structure and from the molecular orbital correlation diagrams. In the ground state the odd electron of NO molecule is characterized by a pair of the antibonding orbitals of the 2pxy (N,O) origin, which degeneracy is removed upon angular coordination. In the gas-phase the one-electron energy levels, the -SOMO and the -LUMO, are situated at SOMO = −4761 eV, LUMO = −3722 eV, respectively. The electron density redistribution within the 2* and 3d orbitals is governed by the equilibration of the chemical potential of the encaged metal and the ligand. Thus, at first glance, the observed changes in their valence states can be rationalized in terms of the relative positions of the NO frontier orbitals with respect to the Fermi level of the complex. As an example, they are shown in Figure 2.10 for the CoII , CuII , and CuI centers. In the case of cobalt, the metal-based orbitals overlap strongly with the framework oxygen orbitals. The Fermi level is dominated by the partially filled 3d states and is located below the 2* level of NO, implying a net transfer of the electron density form the ligand to the metal, observed experimentally [55]. The small contribution of oxygen to DOS in the vicinity of the Fermi level would imply that the influence of framework is indirect for such centers, and that the charge is distributed almost exclusively within the Co−NO moiety (Table 2.4). For copper(II) the contributions from both the metal and the lattice are more equilibrated in accordance with the highly covalent Cu−Oz bonding, revealed in EPR investigations [39]. Again the structure of DOS favors partial oxidation of the NO ligand upon coordination. But in this case there is a considerable direct involvement of the zeolite framework in this redox process, as it can be inferred form the corresponding changes in the QCu and QNO values (Table 2.4).

42

Past and Present in DeNOx Catalysis (b)

(a)

EF

EF

NO(g)

NO(g)

× 0.5

–10.00

× 0.5

–5.00

0.00

5.00

–10.00

–5.00

Energy/eV

0.00

5.00

Energy/eV

(c) EF NO(g)

× 0.5

–10.00

–5.00

0.00

5.00

Energy/eV

) along with atomic framework oxygen 2p ( ) and metal Figure 2.10. Plot of the total DOS ( 3d ( ) projected DOS, calculated using spin-unrestricted VWN/DNP for the (a) {4 CoII }Z6, (b) {2 CuII }Z6, and (c) {1 CuI }M5 complexes. EF denotes Fermi level.

In the case of the mostly investigated monovalent copper, the 3d states located below the Fermi level are quite shrunk and exhibit small contribution from the Oz ligands. Their close proximity to 2* levels of NO implies strong mutual interaction between Cu and NO. Indeed, for the corresponding copper(I) nitrosyl complex this region of DOS is dominated by the contribution of the NO ligand (55% due to nitrogen 2p and 34% due to oxygen 2p). The rest is distributed over copper-based 4s, 3dz2 , 3dxz , and 3dyz orbitals (11% in total). More detailed insight into the interaction of NO with the {CuI }M5 site can be inferred from the simplified Kohn–Sham orbital correlation diagram shown in Figure 2.11. The essential NO orbitals were reduced to the low-lying lone pair n (En = −11252 eV) and the 2* levels (E∗ = −4860 eV) situated close to the 3d manifold of copper (E3d = −5840 ÷ −5183 eV). The coordination of NO (2  to the {CuI }M5 (1 S) can be analyzed in terms of two mutually connected effects related to the interaction of the copper center with the hosting site M5 and with the NO ligand. The first one involves 3d-4s hybridization between the empty 4s level (E4s = −2327 eV) and two occupied 3d orbitals, 3dz2 (−5569 eV) and 3dyz (−5718 eV), as already mentioned. The second one is a symmetry allowed binding effect, arising from the overlap of the resultant copper (4s + 3dz2 + 3dyz ) hybrid with the in-plane NO 2 y and the lone pair n orbitals, leading to the formation of the SOMO (SOMO = −5031 eV). Interaction between the 3dxz and the 2 x orbitals, orthogonal to the symmetry plane, gives rise to the LUMO combination (LUMO = −4708 eV). Both orbitals are antibonding and show a mixed 2–3d character with dominant contribution from NO ligand. The resultant

DFT Modeling and Spectroscopic Investigations

43

(b) –2.0

4s

–4.0 LUMO –4.5 3d electron manifold

3dx 2–y 2 –5.0

E /eV

3dxy 3dyz 3dxz 3dz 2 –6.0

SOMO

π2p * x ,y (a)

–7.0

σ –12.0 {CuI}M5

η1–N {2CuNO}11M5

{NO}1

Figure 2.11. Kohn–Sham molecular correlation diagram of principal spin orbitals for the 1 -N {2 CuNO}11 M5 complex (BP/DNP). Bending of the adduct results in a splitting of both 2* levels giving rise to (a) SOMO and (b) LUMO (after [32]).

NO Cu

Figure 2.12. BP/DNP calculated spin density contour of the 1 -N {2 CuNO}11 M5 complex and for the NO molecule, showing two possible routes of NO attachment: an inner-sphere attack at the metal center (slim arrow) and an outer-sphere attack at the NO ligand (bold arrow) (after [75]).

spin density repartition, N = 0581, O = 0351, and Cu = 012, shown in Figure 2.12, allows for formal description of the magnetic properties of the copper nitrosyl adduct in terms of the [CuI −2 • N = O)]+ magnetophore [22,56]. As it is discussed below, the bent geometry with retention of the spin on the NO ligand has important mechanistic consequences for the reactivity of the copper(I) nitrosyl species.

44

Past and Present in DeNOx Catalysis

5.1.2. Dinitrosyl complexes Depending on the difference in adsorption energies (see Section 5.4) dinitrosyl complexes are formed either concomitantly or subsequently with the mononitrosyl complexes. Those processes have been widely investigated for selected TMIs and can be followed easily by IR technique [57]. The appearance of a characteristic doublet due to the collective antisymmetric and symmetric vibrations of the M(NO)2 moiety growing at the expanse of the NO valence band is usually taken as a confirmation of the dinitrosyl formation. As discussed below in more detail, they play important role in the inner-sphere route of the N−N bond making (see Section 6.2.1). Structural characteristics of the calculated {M(NO)2 }n Z6 complexes are summarized in Table 2.5. Except of the {2 Fe(NO)2 }7 and {1 Fe(NO)2 }8 complexes (with one NO molecule highly bent and the second almost straight), the NO ligands are nearly equivalent. The intranuclear N−O distance is longer for the {2 Cr(NO)2 }7 and {1 Cu(NO)2 }12 adducts than in the gas-phase NO, contrary to the remaining dinitrosyls with the NO bond lengths shorter. The changes in the coordination number are also irregular. For the {2 Cr(NO)2 }7 and {1 Cu(NO)2 }12 complexes, an associative addition of NO ligand is observed (the coordination number increases from 4 to 5 and from 3 to 4, respectively), while for the rest of the dinitrosyls adsorption of the second molecule follows an isodesmic pathway. Formation of dinitrosyl adducts is oxidative for Fe, Co, and Ni (QM = −03 ÷ −01), whereas in the case of copper and chromium it is reductive (QM ∼ 01), similarly to the corresponding mononitrosyls. The highest bending of the M−N−O angle ( ∼ 123 ) was observed for 1 {Ni(NO)2 }10 with both NO ligands coordinated formally as the nitrosonium species (vide infra). Apart from the {2 Fe(NO)2 }7 , none of the investigated complexes possess a distinct cis-coplanar structure with one linear and one bent NO moiety, which is required by the MO symmetry restrictions for the concerted decomposition of two NO molecules into N2 and O2 [53]. In the case of paramagnetic dinitrosyls, the spin density distribution indicates their metal-centered radical character, with a small fraction of the negative spin density on both NO ligands. Table 2.5. Comparison of molecular properties of the dinitrosyl complexes 1 {M(NO)2 }n of selected TMIs encaged within ZSM-5 zeolite 2

M−NO1   M−NO2   N−M−N   dN1−O1 Å dM−N1 Å dN2−O2 Å dM−N2 Å QM (au) QNO1 (au) QNO1 (au) M NO1 (au) NO2 (au)

CrNO2 7 176 171 89 1188 1690 1185 1689 009 −013 −011 141 −022 −023

2

FeNO2 7 170 123 92 1155 1665 1166 1855 −031 018 014 097 −005 −025

1

FeNO2 8 145 173 110 1174 1672 1161 1621 −013 004 014 0 0 0

2

CoNO2 9 149 134 97 1155 1698 1166 1728 −012 018 014 108 −018 −008

1

NiNO2 10 123 125 97 1164 1766 1159 1773 −006 017 019 0 0 0

1

CuNO2 12 136 126 64 1184 1852 1171 2092 008 −003 004 0 0 0

DFT Modeling and Spectroscopic Investigations

45

5.2. Spectroscopic diagnostic features of nitrosyl complexes Surface nitrosyl complexes of TMI have been thoroughly investigated by the computational spectroscopy [22,23,32,33,36,49], and their molecular structure has been ascertained by a remarkable agreement between the theory and experiment of both vibrational (oscillation frequencies and intensities) and magnetic (g and A tensors) parameters. The calculated NO values for the examined mononitrosyls along with the experimental frequencies are listed in Table 2.6. Analogous collation of the IR data for dinitrosyl species is shown in Table 2.7. In the case of paramagnetic complexes their experimental magnetic parameters are determined by computer simulation of the powder spectra [59]. Together with the corresponding calculated values, obtained using a relativistic spin-unrestricted ZORA approach, they all are collected in Table 2.8. Although the spectroscopic parameters prove to be diagnostic for simple finger-print identification of the corresponding species, more attentive analysis of the data contained in Tables 2.7 and 2.8 indicate that there is no correlation between the NO values and the M−N−O bond angles, for both the mono- and the dinitrosyl complexes. It is then incorrect to attempt assignments of the MNO geometries based on the observed N−O Table 2.6. Experimental and DFT (VWN/DNP) calculated valence NO frequencies along with their intensities of the selected mononitrosyl complexes in ZSM-5 zeolite NO cm−1 (int. km/mol)

{3 CrNO}6

{3 MnNO}6

{3 FeNO}6

{4 FeNO}7

{1 CoNO}8

DFT Experimental

1727 (480)

1890 (424) 1894

1852 (887)

1920 (509) 1876

1863 (428) 1857

{2 NiNO}9

{1 NiNO}10

{1 CuNO}10

{2 CuNO}11

{2 ZnNO}11

1899 (382) 1906–1895

1820 (380) 1815–1807

1863 (458)

DFT Experimental

1854 (811) 1895

Experimental data are quoted after Iwamoto et al. [58]

Table 2.7. Comparison of DFT calculated and experimental stretching frequencies for the selected dinitrosyl complexes  exp (cm−1 

 DFT (cm−1 (int. km/mol)

Isym /Iasym 1

Symmetric Antisymmetric

1900 1817

1938 (721) 1800 (966)

0.75

Repulso NiII (NO)2 silT5

Symmetric Antisymmetric

1882 1841

1856 (620) 1807 (783)

0.79

1

Attracto CuI (NO)2 Z6

Symmetric Antisymmetric

1825 1730

1799 (328) 1683 (568)

0.58

1

Repulso 1 CuI (NO)2 Z6

Symmetric Antisymmetric

1805 (193) 1703 (1043)

0.19

Structure

Vibration mode

Attracto 2 CoII (NO)2 Z6

1

Ratio of the intensities of the symmetric (Isym  to antisymmetric (Iasym  bands

46

Table 2.8. Calculated and experimental magnetic parameters for selected nitrosyl surface complexes hosted in the ZSM-5 framework or silica surface M

g Tensor

System

A tensor (10−4 cm−1

N

A tensor (10−4 cm−1 

gxx

gyy

gzz

aiso

Txx

Tyy

Tzz

aiso

Txx

Tyy

Tzz

117 127 123

191 179 157

−98 −93 −83

−93 −86 −74

{CuI NO}M5 {CuI NO}Z6 Experimental

2011 2008 1999

2019 2012 2003

1904 1881 1889

1587 1978 1585

−41 −52 −90

−190 −219 −136

231 271 226

CoII ZSM-5

{CoII (NO)2 }Z6 Experimental

2073 2088

2105 2188

2037 2083

−13

71

92

−162

NiII SiO2

{NiII NO}silT5 Experimental

2046 2156

2053 2184

2241 2369

FeII ZSM-5

Experimental

∼4E/D < 1

Past and Present in DeNOx Catalysis

CuI ZSM-5

DFT Modeling and Spectroscopic Investigations

47

stretching frequencies. Furthermore, it is not always possible to find simple relationship between NO and the electronic population on the NO ligand, since other factors such as the nature of the metal, the influence of other ligands, and even the overall charge of the complex are important. Thus, computation spectroscopy is apparently the only method for reliable structure resolution.

5.2.1. Conformation assignment from IR data Dinitrosyl complexes are rarely linear assuming usually either an attracto or a repulso conformation [60], exemplified for instance with the {CuI (NO)2 }M7 and {NiII (NO)2 }silT5 cluster in Figure 2.13a, b, respectively. In the repulso conformation, the NO ligands are bent outwards, giving rise to a molecular arrangement with the proximal nitrogen and distal oxygen atoms, whereas in the attracto form the NO ligands are bent inwards, bringing both oxygens close to and the nitrogen atoms away from each other. The attracto geometry has been suggested for dinitrosyl complexes of 3dn TMIs with good -acceptor ligands [60]. For such complexes the value of the N−M−N angle has been found to be lower than 130 . On the contrary, the repulso adducts are characteristic of 4dn and 5dn complexes of TMIs with poor -acceptor ligands. The N−M−N angle is usually greater than 130 in such a case [60]. A straightforward distinction between both conformers is based on the comparison of the intensity ratio of the antisymmetric and symmetric bands. For the particular case of the {CuI (NO)2 } complex, the attracto conformer is favored by 77 kcal/mol over the repulso one. Vibrational analysis raveled the presence of two associated normal modes, a symmetric mode (sym ) of lower intensity and an antisymmetric one (asym ) of higher intensity, which are well separated and not contaminated by other vibrations. The calculated sym and asym values for copper in various hosting sites of the ZSM-5 zeolite show rather small variance (sym = 15 cm−1 and asym = 10 cm−1 ). Thus, taking into account the experimental linewidth of 10 cm−1 and 13 cm−1 , respectively, the band positions of the symmetric and antisymmetric components are virtually almost insensitive to the molecular environment of the metal center and the assumed conformation. They are therefore not structurally diagnostic. For distinguishing between the attracto and repulso conformers of the {CuI (NO)2 }ZSM-5 complex, the IR peak intensities are much more useful. Indeed, from the inspection of Table 2.7 it is clear that the relative intensities of the symmetric and

1.184 136°

1.171

125°

134°

1.166

1.833

8 72

34

2.0

1.795

1.164

1.

126°

(c)

1.159

6

(b) 1. 76

(a)

1.155 2.043

2.114

Figure 2.13. Optimized geometries (DMol, VWN/DNP) of three selected dinitrosyl complexes, (a) {1 Cu(NO)2 }12 M7, (b) {1 Ni(NO)2 }10 silT5, and (c) {2 Co(NO)2 }9 Z6. All bond lengths are given in Å, and angles, in degrees (after [71,75]).

48

Past and Present in DeNOx Catalysis

825 O

1257

O

N

N Cu

attracto 0 1850 1950

1750

1650

1550

repulso 0 1950 1850

1750

1650

1550

ν/cm–1

ν/cm–1

Figure 2.14. Density functional theory calculated (BP/DNP) IR spectral profiles of the N−O bond stretching region for the 1 -N {1 Cu(NO)2 }12 I2 complex in the attracto and repulso conformation (after [75]).

antisymmetric stretchings are markedly altered, whereas the band positions do not vary much with the conformation change. The spectral profiles calculated for the attracto and repulso conformers of the {CuI (NO)2 }I2 complex presented in Figure 2.14 substantiate this premise in a clear-cut way. As shown in Table 2.7, the calculated intensity of the antisymmetric band is five times greater for the repulso form, whereas for the attracto conformer the intensities of both bands are more equilibrated. Comparison of the calculated (Figure 2.14) and experimental IR spectra (Isym /Iasym = 058) shows that the copper dinitrosyls encaged in ZSM-5 exhibit the attracto conformation.

5.3. Magnetic pathways of NO coordination Because of the paramagnetic nature of NO (2 1/2 ) and the high-spin configuration of the investigated TMI hosted in ZSM-5 (Table 2.3), coordination of nitric oxide entails not only changes in the electron distribution but also involves dramatic repartition of the spin density within the M−NO unit. The ensuing nitrosyl complexes can either be diamagnetic ({1 CoNO}8 , {1 NiNO}10 , {1 CuNO}10 , or paramagnetic ({3 CrNO}6 , {3 MnNO}6 , {3 FeNO}6 , {4 FeNO}7 , {2 NiNO}9 , ({2 Co(NO)2 }9 , {2 CuNO}11 , {2 ZnNO}11 . Form the collation of the experimental and computational data we have identified the following prototype magnetic pathways of NO adsorption: (1) Spin addition: 1 CuI M5 + 2 NO → 1 2 CuNO11 M5

(3)

S+D → D (2) Spin coupling: 5 FeII Z6 + 2 NO → 1 4 FeNO7 Z6

(4)

Qi + D → Q (3) Spin coupling joint with internal spin pairing:

4 CoII Z6 + 22 NO → 1 2 CoNO2 9 Z6 Q+D+D → D

(5)

DFT Modeling and Spectroscopic Investigations

49 g=2

g=4

Co2+ – (NO)2, S = 1 2 NO-zeolite

Fe2+ – NO, S = 3 2 Cu+ – NO, S = 1 2

1400

3000

2200

3800

B/G

Figure 2.15. X-band EPR spectra recorded after NO adsorption (1 ÷ 5 torr) onto CoII ZSM-5, FeII ZSM-5 (after [64]), and CuI ZSM-5 (after [41]) zeolites.

where S = singlet, D = doublet, Q = quartet, Qi = spin quintet. Following the Wigner and Witmer rules [61], the number of the allowed spin state combinations increases with the multiplicity. Thus, for TMI with several unpaired electrons, spin conservation does not impose serious limitations, since there are many possible alternatives available [62]. The resultant nitrosyl complexes have been identified using EPR techniques (Figure 2.15) and characterized by computational spectroscopy. Simple spin addition occurs when only one reactant (the ligand) is paramagnetic. A good example of such process is the adsorption of NO on the closed-shell CuI ZSM-5 zeolite [41]. Spin coupling is characteristic of two paramagnetic reactants, when the odd electrons of the NO ligand and the metal center combine. This route has been observed for reaction of NO with 5 Fe(II) hosted in ZSM-5 [64] or silicalite [65]. Because the internal spin pairing energy is apparently too large (Table 2.3), no concomitant spin pairing of the remaining add electrons of the iron center is observed. Spin coupling joint with the internal spin pairing (spin crossing) involves, besides the metal−ligand spin coupling, also pairing of the odd 3d-electrons of the metal, due to enhanced ligand field. Formation of cobalt(II) dinitrosyls follows this pathway [66], owing to quite small spin pairing energy (see Table 2.3). An exclusive internal spin pairing can be illustrated by coordination of a strong diamagnetic ligand such as carbon monoxide to high spin TMI [67].

5.4. Adsorption energetics Formation of the mono- and dinitrosyl complexes is a thermodynamically favorable process, which distinctly depends on the electronic configuration of the metal center. The adsorption energy, defined as Eads = Eaddukt − EMZSM−5 + ENO , is shown in the form of a histogram in Figure 2.16. Formation of mononitrosyl complexes is exothermic

50

Past and Present in DeNOx Catalysis 0

Eads(kcal/mol)

–20 –40

NO

–60

N2O2 (NO)2

–80

6

5

}Z

II

{1 Zn

uI }M {1 C

u II }Z 6 {2 C

i I}I Z6 {3 N

6 }Z

II

o II }Z 6 {4 C

5 {5 Fe

}s ilT

III

n II }Z 6

{6 Fe

{6 M

{6 C

r }I M

5

–100

Figure 2.16. Calculated (BP/DNP) energies of the formation of the mononitrosyl, dinitrosyl, and dinitrogen dioxygen complexes of selected TMI encaged in ZSM-5 zeolite.

by 20 ÷ 22 kcal/mol for both CuI and CuII ions, it is somewhat higher (29 kcal/mol) for FeIII and MnII with semi-occupied orbitals and increases to 35 kcal/mol for CoII and NiII , reaching 50 ÷ 53 kcal/mol for low valent CrI and FeII ions. The lowest affinity to NO is observed in the case of ZnII (6 kcal/mol), where the ligand–metal interaction exhibits a weak electrostatic character. For the other mononitrosyl complexes, the bonding has a mixed iono-covalent character. The adsorption energies are apparently greater for the d5−8 metals in contrast to the d9 and d10 configuration. Generally, this tendency is in line with the changes in the M−NO bond lengths and the bond orders (Tables 2.4 and 2.5). However, there is a remarkable variance for the d5 configuration and between the FeII and FeIII centers, already noted elsewhere [68]. The energy of the formation of the dinitrosyl complexes (i.e., adsorption of the second NO molecule) is with the exception of Cr+ distinctly smaller (5 ÷ 20 kcal/mol) that those observed for the mononitrosyl counterparts, as expected. In few cases (MnII , CuII , and ZnII , the dinitrosyls turn out to be even unstable, which is in line with IR data [69,70]. The results show that for the CrI and NiII complexes, both forms can coexist at elevated temperatures, whereas for other TMI only mononitrosyl complexes are persistent, which has an important implication for DeNOx mechanism.

6. INSIGHTS INTO N−N AND O−O BONDS FORMATION MOLECULAR EVENTS The specific goal of the mechanistic studies of DeNOx reaction is to identify the key intermediates involved in the N−N and O−O bonds making, discriminate them from spectator species and ascertain the sequence and conditions of their appearance. To clarify the role of the mono- and dinitrosyl complexes as intermediates or spectators of the principal mechanistic reaction steps, it is necessary to develop a more in-depth insight into the structure–reactivity relationships for both adducts, and to understand the possible ways of attaching the second NO molecule to the mononitrosyl complex.

DFT Modeling and Spectroscopic Investigations

51

6.1. Nitroside and nitrosonium pathwys of NO activation The binding of nitric oxide to a transition metal center imparts unique catalytic chemistry both to the metal and the nitrosyl ligand itself [53,60]. Owing to its particular electronic configuration with the odd electron on the 2* antibonding orbital, the chemical reactivity of the coordinated NO is chiefly featured by the simple redox processes that involve removal of the unpaired electron from the 2* MO (formation of a nitrosonium cation NO+  or addition of an electron to this orbital (leading to a nitroside anion NO− . The nitrosonium ligands exhibit a tendency to nucleophilic attack at the N atom, while the nitroside ligands undergo an electrophilic attack [53]. The resultant two generic pathways along which nitric oxide can be decomposed or selectively reduced are then primarily driven by the particular bonding mode of the NO reactant and the charge displacement within the metal−ligand moiety. Taking into account the electron density relocation (Table 2.4) two routes of NO adsorption can be distinguished. Thus, the nitric oxide coordinates to the monovalent Cr, Ni, and Cu ions in an oxidative way (QM > 0), whereas for the rest of the TMIs in a reductive way (QM < 0). Although this classification is based on the rather simplified criteria, it is well substantiated by experimental observations. Examples of reductive adsorption are provided by interaction of NO with NiII SiO2 and NiII ZSM-5, leading at T > 200 K to a NiI −NO+ adduct identified by the characteristic EPR signal [71]. At elevated temperatures, similar reduction takes place for CoII ZSM-5 [63], whereas in the case of CuI ZSM-5 part of the monovalent copper is oxidized by NO to Cu2+ , as it can readily be inferred from IR and EPR spectra [72,73]. This point is discussed in more detail elsewhere [4,57]. Based on the collation of the theoretical and experimental results, two distinct NO activation pathways can then be proposed. The oxidative adsorption, giving rise to the bound NO− species, defines a nitroside pathway of activation, while the reductive adsorption leading to NO+ species is tantamount with a nitrosonium pathway of activation (Figure 2.17). In the latter case the M−N−O moiety is highly bent (130–140 ), the N−O bond length shortened and its polarization increased by about three times in comparison to free NO molecule. Such changes indicate that the NO acquires electrophilic properties upon the coordination, and is characterized by NO > 1850 cm−1 [53]. Examples are given by the 1 {CoNO}8 , 1 {NiNO}9 , and 1 {1 CuNO}10 complexes. In contrast, for the NO− species the N−O bond is elongated, only slightly polarized, and the stretching frequency, NO , decreases below 1850 cm−1 . Such changes indicate that the activation consists in redistribution of the electron and spin densities within the M−NO unit, which accumulates on the nitrogen atom. Among the first series TMI, the oxidative adsorption is less common and includes only the 1 {1 CuNO}11 and 1 {3 CrNO}6 adducts. The mechanistic implications of the electronic structures for both type of the nitrosyl complexes are discussed in the next section.

6.2. Dichotomy of the NO attack For making the N−N bonds, one obviously needs two NO molecules brought into an intimate contact. Conceivably, there are two possible mechanistic issues for attachment

52

Past and Present in DeNOx Catalysis NO activation through coordination Redistribution of electron density within the M-NO unit

NOδ–

NOδ+

nitroside pathway

nitrosonium pathway

OXIDATIVE ADSORPTION

REDUCTIVE ADSORPTION

NO becomes nucleophilic

NO becomes electrophilic

NO bond elongated slightly polarized

NO bond shortened polarized

adducts less bent (150 – 170°)

adducts significantly bent (130 – 140°)

{3CrNO}6

ν = 1727 cm–1

{1CoNO}8

ν = 1863 cm–1

{2CuNO}11

ν = 1820 cm–1

{1CuNO}10

ν = 1899 cm–1

{4FeNO}7

ν = 1920 cm–1

Reductive adsorption, aduct less bent

Figure 2.17. Diagram showing nitroside and nitrosonium ways of NO activation upon coordination to the transition-metal ions in zeolites along with basic characteristics of the NO+ and NO− species.

of the second NO molecule to the mononitrosyl intermediate. The attack can be directed either at the metal (this is equivalent to an inner-sphere pathway) or at the NO ligand (implying an outer-sphere pathway). The spin density repartition within the {M−NO}n magnetophore appears to be the principal factor determining direction of the second NO molecule addition, because both reactants are paramagnetic. The inner-sphere attack leading to the formation of the dinitrosyl complexes is observed for the metal-centered radical complexes such as {CrNO}7 , where the spin density is mainly localized on the metal moiety (Figure 2.18a). The outer-sphere route entailing formation of a dinitrogen dioxide intermediate (N2 O2  is characteristic of the ligand-centered radical complexes (Section 6.2.2.), such as {ZnNO}11 , with the preponderant spin density on the ligand (Figure 2.18b). (a)

(b)

Figure 2.18. Spin density contours (VWN/DNP) of the (a) 1 -N {3 CrNO}6 M5 and (b) 1 N {2 ZnNO}11 Z6 complexes.

DFT Modeling and Spectroscopic Investigations

53

Thus, for paramagnetic complexes the reactivity patterns promoted either by the metal center or by the ligand (equivalent to an inner- vs. an outer-sphere pathway), are essentially triggered by the spin density distribution.

6.2.1. Inner-sphere nitrosonium route The inner-sphere nitrosonium route can be well illustrated by spectroscopic (EPR and IR) investigations into NO decompositin over tripodal NiII grafted on silica (Figure 2.19) supported by DFT modeling. The observed reactivity can be rationalized in terms of the surface disproportionation reaction accompanied by alternation of the nickel oxidation states. The multi-step reaction is initiated by concomitant formation of the paramagnetic mononitrosyl ({Ni3c −NO• }9 /SiO2 (gz1 = 188, gz2 = 186, and gxy = 197 and  = 1867 cm−1 ) and diamagnetic dinitrosyl species {Ni2c −(NO)2 }10 /SiO2 , (sym = 1882 cm−1 and asym = 1841 cm−1 ). NiII 3c /SiO2 + NO•g → Ni3c −NO• 9 /SiO2 + NO•g → Ni2c −NO2 10 /SiO2 NiII 3c /SiO2 + 2NO•g → Ni2c −NO2 10 /SiO2

(6)

The dinitrosyl complex exhibits the repulso conformation with both Ni−N−O moieties bent outwardly, with the angle = 1251 and the ON−Ni−NO angle = 97 Ni

O

NO 3

2

II

N

9

{ NiNO} /silT5

{ Ni }/silT5

(a)

20 mT

0 min 0.5 min 17 min 107 min 200 min g = 2.00

(b)

(c)

I/I0 Ni

Ni

1

0.5 NO

NO

t /min

40

20

0

100

200 t /min

Figure 2.19. Interaction of NiII SiO2 with NO along with the calculated structures and spin density contours of the {3 NiII } silT5 and {2 NiNO}9 silT5 complexes. (a) Evolution of the X-band EPR signal (recorded at 77 K) obtained upon contacting 2 torr of NO with NiII /SiO2 at 77 K and subsequent warming to 296 K for increasing periods of time and the corresponding plots of the intensity variation of the {NiII −NO• }9 /SiO2 (•) and {NiI −NO+ }9 /SiO2 () species as a function of the elapsing time for (b) 373 K, and (c) 296 K, (adopted from [71]).

54

Past and Present in DeNOx Catalysis

(Figure 2.13b). The calculated harmonic frequencies of such species (Tables 2.6 and 2.7), in fair agreement with the experimental values, confirm assignment of its structure. The ligand-centered radical {Ni3c −NO• }9 /SiO2 is unstable and gradually transforms into a metal-centered radical {NiI3c −NO+ }9 /SiO2 (gz = 2369, gy = 2184, gx = 2156) through inner-sphere ligand-to-metal electron transfer,

Ni3c −NO• 9 /SiO2 → NiI 3c −NO+ 9 /SiO2

(7)

as it can be inferred from the evolution of the corresponding EPR spectra (Figure 2.19a). The plot of their integral intensities as a function of the time for 373 and 296 K (Figure 2.19b, c) clearly indicate that {NiI3c −NO+ }9 /SiO2 complex is formed at the expense of the {Ni3c −NO• }9 /SiO2 one. The apparent activation energy for ligand to metal electron transfer is found to be Ea = 10 ± 3 kJ/mol. The geometry of the {NiI3c −NO+ }9 /SiO2 complex is shown in Figure 2.19, along with the associated spin density contour. The latter is mainly confined to the metal center (61%) with sizable part spread onto the matrix ligands (37%), whereas only mere 2% is localized on the NO moiety, accounting well for the formulation of the {NiI3c −NO+ }9 /SiO2 complex in terms of the metal-centered radical. Since the spin density is largely on the nickel part, therefore, formation the spin-paired dinitrosyl {Ni2c −(NO)2 }10 /SiO2 complexes is preferred upon addition of the second NO molecule. The resultant dinitrosyl complex assumes a repulso conformation, and its thermal decomposition gives rise to the N2 O formation via an inner-sphere route, transforming the primary nickel active sites into the secondary nickel-oxo sites [71] as shown in Figure 2.20.

NiI3c −NO+ 9 /SiO2 + NOg → Ni2c −NO2 10 /SiO2 → Ni3c −O8 /SiO2 + N2 O (8) Similar chemistry has been proposed for other intrazeolite metal sites [60,72,74]. The spin density distribution within the {Ni3c −O}8 moiety (43% on Ni and 57% on O), being distinctly shifted towards the oxygen, renders this terminal atom a preferred site for subsequent NO attack to produce Ni-bound NO2

Ni3c O8 /SiO2 + NOg → Ni3c −NO2 9 /SiO2

(9)

The corroborative DFT calculations reveals that among four possible coordination modes (1 -N {Ni−NO2 }9 , 1 -O{NiNO2 }, 2 -O O{Ni−NO2 }9 , and 2 -N O{Ni−NO2 }9 , the monodentate nitrite-form 1 -N {Ni−NO2 }9 , shown in Figure 2.20, is the most stable one (Eads = −167 kcal/mol). The calculated magnetic parameters for such species are consistent with the spin density localized on the nickel center with a dx2 −y2 ground state. Thus, formally nickel is reduced back to NiI in this reaction step. Because of the apparent lability of the {Ni3c −NO2 }9 /SiO2 complex, arising from relatively small NO2 binding energy, the NO2 ligand gradually leaves the coordination sphere of the nickel and spillovers onto the silica matrix.

Ni3c −NO2 9 /SiO2 → NiII3c /SiO2 + NO2 /SiO2

(10)

The resultant steadily growing triplet EPR signal due to NO2 /SiO2 (with gz = 2002, gy = 1992, gx = 2004, N Az /g e = 632 mT, N Ay /g e = 478 mT, and

DFT Modeling and Spectroscopic Investigations

55

{2NiNO2}9silT5

5 min 12 h

2 days

NO N2O

Ni+ 4 days

{INi(NO)2}10silT5

{3Ni-O}8silT5

10 mT

{2NO2}1SiO2 Figure 2.20. Transformation of silica supported dinitrosyl complexes of nickel(II) leading to formation of nitrogen dioxide and its final stabilization on the support. The picture shows the molecular structure and the spin density contours calculated with BP/DNP method of the involved species, and evolution of the X-band EPR spectra of the {NiNO2 }SiO2 complex due to spillover of the ligand (adopted from [71]).

N Ax /g e = 515 mT) is shown in Figure 2.20. The unraveled full reversibility of the nickel redox state alternation, allowing for persistent binding and disproportionation of NO, is of vital importance for sustaining the catalytic cycle. This study has also provided a direct experimental evidence for NO2 migration across the surface, which has been postulated to be involved in the oxygen intersite molecular transport [74], allowing for the O−O bond formation (see Section 3).

6.2.2. Outer-sphere nitroside route Binding of NO to the {CuI NO}M5 complex can be used for illustrating the nitroside pathway. Although the energetic arguments are clearly in favor of the dinitrosyl formation (E = 182 kcal/mol) than the direct NO coupling to form a dinitrogen dioxide intermediate (E = 122 kcal/mol), the spin density contour calculated for the {CuI NO}M5 complex (Figure 2.12) shows that the spin, being largely retained on the NO ligand, is accumulated on the nitrogen atom (58%). This suggests that the attack of the paramagnetic NO molecule should be directed just at the ligand. Indeed, comparison of the free valence indexes for the Cu (FV = 0.015) and N centers (FV = 0.28) indicates that the nitrogen atom is expected to be a preferred site of the NO attack, favoring thereby the direct NO coupling. Furthermore, the DFT calculations show that it occurs with nearly zero activation energy, so that we might reasonably suppose that the dinitrosyls are most probably the spectators (vide infra), whereas the N2 O2 species produced by direct coupling are the actual transients in the N−N bond formation process.

56

Past and Present in DeNOx Catalysis

6.2.2.1. Mechanistic importance of the dinitrosyls’ conformation As mentioned in Section 6.2.1., for the inner-sphere mechanism of NO decomposition, dinitrosyl complexes play the role of a central intermediate of the N−N bond making process. The chemical reactivity of those complexes remarkably depends on their conformation mode. It is well established in the case of the {Cu(NO)2 }12 complex. The spatial proximity of both oxygen atoms in the attracto conformation would suggest that the O−O bond making may initiate the NO decomposition in a conceivably simplest way. However, the DFT results shows that once the oxygen−oxygen distance decreases, a positive charge develops on both terminal O-atoms, giving rise to a steadily increasing energy, due to the growing coulombic repulsion (Figure 2.21). Thus, apparently the attracto conformation is an essentially inert form, and the NO decomposition cannot be initiated by the O−O bond formation step. The alternative N−N bond making route entails prior transformation of the attracto to repulso conformation. An approximate energy profile of this process (Figure 2.22) has been obtained by systematic variation of the appropriate reaction coordinate, that

(b) 25

0.03

15

0.01

δO

ΔE/kcal × mol–1

(a)

5 1.7

2.1

2.5

–0.01 1.7

2.1

dO-O /Å

dO-O /Å

Figure 2.21. Changes in (a) total energy and (b) partial charge on the terminal oxygen atoms of the NO ligands, calculated for the stepwise decrease of the O−O distance in the attracto conformation of the copper(I) dinitrosyl complex (after [75]).

Repulso

TS

NO +

6.0 13.4 kcal/mol 7.7 6.0

Attracto

{CuIN2O2}I2

Spectator

Intermediate

Figure 2.22. Schematic potential energy profile for two types of the N−N bond formation mechanisms calculated for the I2 hosting site (BP/DNP) (after [75]).

DFT Modeling and Spectroscopic Investigations

57

was kept frozen during the optimization. As implied by Figure 2.22, the transition state of the attracto to repulso transformation is located 13.4 kcal/mol above the attracto level, and assumes the geometry with two nearly linear NO ligands. Although the subsequent inner-sphere coupling of both NO ligands is energetically less demanding, requiring 6.0 kcal/mol only, the alternative outer-sphere coupling of the ligated and gas-phase NO molecules is barrierless. Thus, the outer-sphere direct coupling should be kinetically favored over the more involved dinitrosyl pathway, preferred thermodynamically. 6.2.2.2. Experimental distinction between spectators and intermediates A more detailed analysis of the IR data provides additional arguments for the above conjectures [75]. The IR spectra recorded upon NO pulse adsorption at 303 K on the thermally preactivated CuZSM-5 reveal development of the mononitrosyl band ( = 1814 cm−1 ), but at higher coverage a doublet of the dinitrosyl complex grows up progressively (Figure 2.23a). The mono- and dinitrosyl complexes remain in a mutual equilibrium, that can be easily shifted by varying the number of NO doses, provided that the NOg pressure is kept small. On the contrary, at 423 K only the {CuI NO}ZSM5 adduct was present, regardless the number of NO pulses (Figure 2.23b), indicating that dinitrosyls are essentially unstable at such conditions, in accordance with a marked difference in their adsorption energies (Figure 2.16). Thus, from the IR results, we may infer that upon temperature increase, the dinitrosyls are merely converted into the mononitrosyl complexes, but without formation of any other product relevant for the catalytic cycle (see Section 3), provided that no gas-phase NO is present in substantial amounts. Since the copper dinitrosyl complexes are essentially unstable at the temperatures above 223 K, their involvement in the NO decomposition over CuZSM-5 as intermediates of the N−N bond formation step is unlikely, despite some earlier claims [72,76], in accordance with the structure–reactivity analysis discussed above. A different picture emerged from the comparison of IR measurements in static and flow conditions. The IR spectra recorded in the temperature range of 423–823 K during the catalytic reaction of NO with the CuZSM-5 are collected in Figure 2.24. The spectra are dominated by the broad and intense absorption bands centered at 1470, 1630 and 2224 cm−1 , assigned to NO2 -related species such as surface nitrates NO3 − [77,78],

(b)

1850

A = 0.05

A = 0.02

(a)

1800

1750

ν/cm–1

1700

1650

1850

p NO

1800

1750

1700

1650

ν/cm–1

Figure 2.23. IR spectra obtained after subsequent 10 mol pulse adsorption of NO on a thermally preactivated CuI ZSM-5 sample recorded at (a) 303 K and (b) 423 K (after [75]).

58

Past and Present in DeNOx Catalysis

A = 0.2

Cu+(NO)2

Cu+NO N2O a b c d e

2400

2000

1600

ν/cm–1

Figure 2.24. In situ IR spectra recorded at (a) 423 K, (b) 523 K, (c) 573 K, (d) 673 K, and (e) 823 K, during the reaction of 1% NO in He with CuZSM-5 at GHSV = 24,000 h−1 . In the insert, static spectra of the dinitrosyl (recorded at 303 K) and mononitrosyl complexes (recorded at 423 K) are shown (after [75]).

nitrites NO2 − [57,76], and to N2 O intermediary product, respectively. In comparison to static measurements recorded at the same temperature of 423 K (Figure 2.24), the band due to the {CuI NO}ZSM-5 adducts was quite weak, and no traces of the dinitrosyls could be distinguished. Such behavior indicates that once gas-phase NO is present, the {CuI NO}ZSM-5 species are readily consumed (their concentration drops distinctly below the thermal stability level, assessed in the static experiment), undergoing fast transformation into the intermediate products of the successive reaction steps such as N2 O and NO23 , involved in the O−O bond formation cycle (Section 3). Thus, it may be concluded that the transformation of nitric oxide into N2 O does not involve a dinitrosyl complex, but requires the presence of gaseous NO reactant, providing a tangible evidence for involvement of the outer-sphere mechanism.

6.3. Molecular description of the mechanistic steps of DeNOx process Having all the essential building blocks of the DeNOx mechanism well established and verified spectroscopically, quantum chemical modeling may be then used for providing a molecular rational for the observed structure–reactivity relationships. The first mechanistic cycle of the DeNOx reaction, where NO reacting with {CuI }Z center is transformed into N2 O, involves the following steps:

CuI Z + NO → CuI NOZ

(11)

CuI NOZ + NO → cis- CuI N2 O2 Z → trans- CuI N2 O2 Z

(12)

CuI N2 O2 Z → CuOI Z + N2 O

(13)

DFT Modeling and Spectroscopic Investigations

59

S 1Cu+ + 2NO g

9.4 η1{2CuON}11

35.1 kcal/mol

D NO

S

(1Cu-O)+ + 1N2Og

η1{2CuNO}11 NO η1{3Cu(NO)2}12

12.2 18.2

T

trans- η1{3CuN2O2}12 trans- η { CuN2O2} 1 1

cis- η1{1CuN2O2}12

12

(3Cu-O)+ 7.1

3.2

η1{1Cu(NO)2}12

1N O 2

Figure 2.25. Energy landscape (BP/DNP) for the {Cu}ZSM-5 + 2NO →{Cu−O}ZSM-5 + N2 O reaction, showing all associated spin and conformation isomers calculated for the M5 site. The values are given in kcal × mol−1 . The letters S, D and T indicate the singlet, doublet, and triplet states, respectively (after [75]).

The total energy change for the overall reaction of {CuI }M5+2NO→{CuO}M5+N2 O is equal to −473 kcal/mol. A more detailed reaction energetic landscape is shown in Figure 2.25, along with the proposed spectator species (drawn in gray), to provide a suitable integral chemical context. The crucial steps of the N−N bond knitting include exoergic formation of the mononitrosyl complex, its transformation into the {CuI N2 O2 }Z transient via an outer-sphere NO coupling, and finally decomposition of the bound N2 O2 species into N2 O and the copper-oxo site. The latter process is kinetically constrained by the intersystem crossing, because the spin singlet {1 CuN2 O2 }Z intermediate is transformed into the {3 CuO}Z center exhibiting a spin triplet ground state (see Section 4). It is worth noting that the N−N bond formation via N2 O2 intermediate has already been postulated on the basis of spectroscopic results to take place on metallic surfaces [79,80] and in biological systems containing copper [81,82,83] and iron [84] as well. The interaction of NO, N2 O, and NO2 reactants with the secondary copper-oxo sites leads to terminal O2 , restoring the initial active sites (Figure 2.26). The central NO2 semi-product is produced through oxidation of NO on the {3 CuO}Z centers, similarly to the nickel complexes described in Section 6.2.1.

CuOZ + NO → CuNO2 Z

(14)

At the reaction conditions the rather loosely bound NO2 can migrate and react with remaining {CuO}Z centers, producing surface nitrates (with 2 (O,O){CuNO3 }11 being the most stable one).

CuOZ + NO2 → CuNO3 Z

(15)

60

Past and Present in DeNOx Catalysis

1

N2g

1Cu+ + N O 2 g

(1Cu-O)+

(3Cu-O)+

{3CuO2} + 2NOg

5.2 14.7 η1-N{1CuN2O}10

NO

1Cu+

NO2

54.2 52.6 1Cu+ + NO

5.6

56.2

1N O 2

3O 2g

20.2

2g

{3CuO2} + 1N2g

21.8 η1{2CuNO2}12

η2(O,O ){2CuNO3}11

Figure 2.26. Energy landscape (BP/DNP) for the reaction of the copper-oxo species {Cu−O}ZSM5 with N2 O, NO2 , and NO molecules, including associated spin isomers calculated for the M5 site (the values are given in kcal × mol−1 .

Their decomposition into NO and O2 is apparently the most difficult step of the whole DeNOx process and requires elevated temperatures (Figure 2.24). Most likely it takes place in two following steps:

CuNO3 Z → CuO2 Z + NO → CuZ + O2

(16)

In parallel, dioxygen can also be formed by oxygen transfer from N2 O (produced in the N−N bond formation cycle) to the {CuO}Z centers, restoring again the primary active sites {CuI }Z.

CuOZ + N2 O → CuO2 Z + N2 → CuZ + O2

(17)

Due to relatively high affinity of dioxygen to monovalent copper (Eads = −202 kcal/mol), the reaction is expected to be poisoned by oxygen [4]. A particular feature of the whole process is the trade-off between the key intermediates of both mechanistic cycles. While the N−N bond formation (controlled by thermal stability of the mononitrosyl intermediate) is favored by lower temperatures, the O−O bond formation step (constrained by endothermic decomposition of the nitrate intermediate) is favored by higher temperatures. Indeed, as revealed by operando IR studies (Figure 2.24), at low temperatures nitrates accumulate on the surface, whereas at high temperatures the surfaces is essentially depleted of the mononitrosyl complexes. The optimal reaction temperature corresponds, therefore, to a subtle balance between the rate of formation of the {CuI NO}Z surface complex in the early stages, and the rate of decomposition of the {CuNO3 }Z complex in the late stages of the reaction.

DFT Modeling and Spectroscopic Investigations

61

7. SUMMARY In this short review we have shown that interaction of nitric oxide with TMI-exchanged zeolites provides a useful functional model for understanding basic molecular aspects of the interfacial coordination chemistry of nitric oxide as well as of its decomposition. Combination of quantum chemical modeling with computational spectroscopy appears to be a powerful bottom-up approach to DeNOx catalytic chemistry. The effect of spin on the structure of the active sites and adsorption complexes has been demonstrated. Various magnetic pathways of NO ligation and oxidative or reductive NO activation routes, depending on the electronic configuration and the spin state of the transition metal centers, have been distinguished. The spin density repartition within the TMI−NO moiety has been found to be a primary factor determining direction of second NO molecule attack. As a consequence, the N−N bond may be formed either via an innersphere process, with the formation of a dinitrosyl complex, or through an outer-sphere coupling, leading to a dimeric {MN2 O2 } intermediate. In the case of CuZSM-5, the latter mechanism is kinetically favored, whereas for Ni/SiO2 the dinitrosyl route prevails.

ACKNOWLEDGEMENT Financial support by the Committee for Scientific Research of Poland, KBN, grant no. 3 T09A 147 26 is acknowledged. The calculations were carried out with the computer facilities of CYFRONET-AGH, under grant no. KBN/SGI2800/UJ/018/2002. Piotr Pietrzyk is indebted to the Prime Minister of Poland for a Ph. D. Thesis Award and the Foundation for Polish Science for a stipend within the START program.

REFERENCES [1] Gwost, D. and Caulton, K.G. (1974) Oxidation of coordinated nitric oxide by free nitric oxide, Inorg. Chem., 13, 414. [2] Iwamoto, M., Yokoo, S., Sakai, K. et al. (1981) Catalytic decomposition of nitric oxide over copper(II)-exchanged Y-type zeolites, J. Chem. Soc. Faraday Trans., 77, 1629. [3] Turnes Palomino, G., Fisicaro, P., Bordiga, S. et al. (2000) Oxidation states of copper ions in ZSM-5 zeolites. A multitechnique investigation, J. Phys. Chem. B, 104, 4064. [4] Konduru, M.V. and Chuang, S.S.C. (2000) Dynamics of NO and N2 O decomposition over Cu-ZSM-5 under transient reducing and oxidizing conditions, J. Catal., 196, 271. [5] DaCosta, P., Moden, B., Meitzner, G.D. et al. (2002) Spectroscopic and chemical characterization of active and inactive Cu species in NO decomposition catalysts based on Cu-ZSM5, Phys. Chem. Chem. Phys., 4, 4590. [6] Bell, A.T. (1997) Experimental and theoretical studies of NO decomposition and reduction over metal-exchanged ZSM-5, Catal. Today, 38, 151. [7] van Santen, R.A. (2004) Molecular modelling in zeolite catalysis, Stud. Surf. Sci. Catal., 154, 52. [8] Sauer, J. and Sierka, M. (2000) Combining quantum mechanics and interatomic potential functions in Ab initio studies of extended systems, J. Comput. Chem., 21, 1470. [9] Elanany, M., Koyama, M., Kubo, M. et al. (2005) Periodic density functional investigation of Lewis acid sites in zeolites: Relative strength order as revealed from NH3 adsorption, Appl. Surf. Sci., 246, 96.

62

Past and Present in DeNOx Catalysis

[10] Bromley, S.T., Catlow, C.R.A. and Maschmeyer, Th. (2003) Computational modeling of active sites in heterogeneous catalysts, CatTech 7, 164. [11] Cerius2 (2001) Accelrys Inc., Users Guide. Accelrys Inc., San Diego. [12] Frish, M.J., Trucks, G. W., Schlegel, H. B. et al. (2003) Gaussian03, Revision B.05. Gaussian Inc., Pittsburgh, PA. [13] te Velde, G., Bickelhaupt, F.M., van Gisbergen, S.J.A. et al. (2001) Chemistry with ADF, J. Comput. Chem., 22, 931. [14] ADF2002.03, SCM, Theoretical Chemistry. Vrije Universiteit, Amsterdam, The Netherlands, http://www.scm.com. [15] Blaha, P., Schwarz, K., Dufek, P. et al. (1995) WIEN95: A Full Potential Linearized Augmented Plane Wave Package for Calculating Crystal Properties. Technical University, Vienna. [16] Berthomieu, D. and Delahay, G. Recent advances in CuI/II Y: Experiments and modeling (2006) Catal. Rev., 48, 269. [17] Amos, R.D. (1987) Molecular property derivatives, In Ab initio Methods in Quantum Chemistry – Part I (K.P. Lawley, ed.), pp. 99–153, Wiley, New York. [18] Koch, W. and Holthausen, M.C. (2000) A Chemist’s Guide to Density Functional Theory. Viley: VCH, Weinheim, Germany. [19] van Lenthe, E., Wormer, P.E.S. and van der Avoird, A. (1997) Density functional calculations of molecular g-tensors in the zero-order regular approximation for relativistic effects, J. Chem. Phys., 107, 2488. [20] Schreckenbach, G. and Ziegler, T. (1997) Calculation of the g-tensor of electron paramagnetic resonance spectroscopy using gauge-including atomic orbitals and density functional theory, J. Phys. Chem. A, 101, 3388. [21] Cotton, F.A. and Wilkinson, G. (1980) Advanced Inorganic Chemistry, Wiley, New York. [22] Pietrzyk, P. and Sojka, Z. (2005) Relativistic density functional calculations of EPR g tensor for 1{CuNO}11 species in discrete and zeolite-embedded states, J. Phys. Chem. A, 109, 10571. [23] Pietrzyk, P. (2005) Ph. D. Thesis. Jagiellonian University, Krakow. [24] Dyrek, K. and Che, M. (1997) EPR as a tool to investigate the transition metal chemistry on oxide surfaces, Chem. Rev., 97, 305. [25] Lamberti, C., Bordiga, S., Ricchiardi, G. et al. (1997) XAFS, IR, and UV-Vis study of the CuI environment in CuI -ZSM-5, J. Phys. Chem. B., 101, 344. [26] Nachtigallová, D., Nachtigall, P., Sierka, M. et al. (1999) Coordination and siting of Cu+ ions in ZSM-5: A combined quantum mechanics/interatomic potential function study, Phys. Chem. Chem. Phys., 1, 2019. [27] Rice, M.J., Chakraborty, A.K. and Bell, A.T. (2000) Theoretical studies of the coordination and stability of divalent cations in ZSM-5, J. Phys. Chem. B., 104, 9987. [28] Centi, G., Wichterlová, B. and Bell, A.T. (2001) Catalysis by Unique Metal Ion Structures in Solid Matrices. Kluwer Academic, Dordrecht. [29] Dedecek, J., Kaucký, D. and Wichterlová, B. (2000) Co2+ ion siting in pentasil-containing zeolites, part 3. Co2+ ion sites and their occupation in ZSM-5: A VIS diffuse reflectance spectroscopy study, Microporous Mesoporous Mater., 35–6, 483. [30] Rodrigez-Santiago, L., Sierka, M., Branchadell, V. et al. (1998) Coordination of Cu+ ions to zeolite frameworks strongly enhances their ability to bind NO2 : An ab initio density functional study, J. Am. Chem. Soc., 120, 1545. [31] Sierraalta, A., Añez, R. and Brussin, M.-R. (2002) Theoretical study of NO2 adsorption on a transition-metal zeolite model, J. Catal., 205, 107. [32] Pietrzyk, P., Piskorz, W., Sojka, Z. et al. (2003) Molecular structure, spin density distribution, and hyperfine coupling constants of the 1{CuNO}11 adduct in the ZSM-5 zeolite: DFT calculations and comparison with EPR data, J. Phys. Chem. B., 107, 6105.

DFT Modeling and Spectroscopic Investigations

63

[33] Broclawik, E., Datka, J., Gil, B. et al. (2002) Why Cu+ in ZSM-5 framework is active in DeNOx reaction – Quantum chemical calculations and IR studies, Catal. Today, 75, 353. [34] Garofalini, S.H. (1990) Molecular dynamics computer simulations of silica surface structure and adsorption of water molecules, J. Non-Cryst. Solids, 120, 1. [35] Garrot, J.M., Lepetit, C., Che, M. et al. (2001) Modeling of the tridentate amorphous silica ligand, J. Phys. Chem. A, 105, 9445. [36] Pietrzyk, P. (2005) Spectroscopy and computations of supported metal adducts. 1. DFT study of CO and NO adsorption and coadsorption on Cu/SiO2 , J. Phys. Chem. B, 109, 10291. [37] Pietrzyk, P., Zasada, F., Piskorz, W. et al. (2007) Computational spectroscopy and DFT investigations into nitrogen and oxygen bond breaking and bond making processes in model deNOx and deN2 O reactions, Catal. Today, 119, 219. [38] Broclawik, E., Datka, J., Gil, B. et al. (2000) T-O-T skeletal vibration in CuZSM-5 zeolite: IR study and quantum chemical modeling, Phys. Chem. Chem. Phys., 2, 401. [39] Berthomieu, D., Krishnamurty, S., Coq, B. et al. (2001) Theoretical modeling of a copper site in a Cu(II) – Y zeolite, J. Phys. Chem. B, 105, 1149. [40] Moden, B., da Costa, P., Fonte, B. et al. (2002) Kinetics and mechanism of steady-state catalytic NO decomposition reactions on Cu-ZSM5, J. Catal., 209, 75. [41] Sojka, Z., Che, M. and Giamello, E. (1997) EPR investigation of the electronic structure of mononuclear copper(I) nitric oxide adduct formed upon low-pressure adsorption of NO onto Cu/ZSM-5 zeolite, J. Phys. Chem. B, 101, 4831. [42] Schneider, W.F., Hass, K.C., Ramprasad, R. et al. (1998) Density functional theory study of transformations of nitrogen oxides catalyzed by Cu-exchanged zeolites, J. Phys. Chem. B, 102, 3692. [43] Tajima, N., Hashimoto, M., Toyama, F. et al. (1999) A theoretical study on the catalysis of Cu-exchanged zeolite for the decomposition of nitric oxide, Phys. Chem. Chem. Phys., 1, 3823. [44] Sengupta, D., Adams, J.B., Schneider, W.F. et al. (2001) Theoretical analysis of N2 O to N2 conversion during the catalytic decomposition of NO by Cu-zeolites, Catal. Lett., 74, 193. [45] Groothaert, M.H., van Bokhoven, J.A., Battiston, A.A. et al. (2003) Bis(-oxo)dicopper in Cu-ZSM-5 and its role in the decomposition of NO: A combined in situ XAFS, UV-visnear-IR, and kinetic study, J. Am. Chem. Soc., 125, 7629. [46] Shaik, S., Filatov, M., Schroder, D. et al. (1998) Electronic structure makes a difference: Cytochrome P-450 mediated hydroxylations of hydrocarbons as a two-state reactivity paradigm, Chem. Eur. J., 4, 193. [47] (a) Zener, C. (1933) Dissociation of the excited diatomic molecules by external perturbations, Proc. Roy. Soc. Lond. A, 140, 660. (b) Landau, L. (1932) Zur theorie der Energieubertragung, Physik. Zeits. Sowjetunion, 2, 46. (c) Stearn, A.E. and Eyring, H. (1935) Nonadiabatic reactions. The decomposition of N2 O, J. Chem. Phys. 3, 778. [48] Buchachenko, A.L. and Berdinsky, V.L. (2002) Electron spin catalysis, Chem. Rev., 102, 603. [49] Davidova, M., Nachtigallova, D. and Nachtigall, P. (2004) Nature of the Cu+ —NO bond in the gas phase and at different types of Cu+ sites in zeolite catalysts, J. Phys. Chem. B, 108, 13674. [50] Bauschlicher, C.W. Jr., Langhoff, S.R. and Partrige, H. (1991) The binding energies of Cu+ −(H2O)n and Cu+ −(NH3 )n (n = 1–4), J. Chem. Phys., 94, 2068. [51] Collman, J.P. (1968) Patterns of organometallic reactions related to homogeneous catalysis, Acc. Chem. Res., 1, 136. [52] Lunsford, J.H. (1975) The Catalytic Chemistry of Nitrogen Oxides. Plenum, New York. [53] Eisenberg, R. and Meyer, C.D. (1975) The coordination chemistry of nitric oxide, Acc. Chem. Res., 8, 26. [54] Herzberg, G. (1945) Molecular Spectra and Molecular Structure. II. Infrared and Raman Spectra of Polyatomic Molecules. Van Nostrand Reinhold, New York.

64

Past and Present in DeNOx Catalysis

[55] Hadjiivanov, K., Ivanova, E., Daturi, M. et al. (2003) Nitrosyl complexes on Co-ZSM-5: An FTIR spectroscopic study, Chem. Phys. Lett., 370, 712. [56] Pietrzyk, P. and Sojka, Z. (2005) EPR spectroscopy and DFT calculations of the g tensors of {VO}1 /ZSM-5, {CuNO}11 /ZSM-5 and {NaNO}1 /ZSM-5 intrazeolitic complexes, Stud. Surf. Sci. Catal., 158, 617. [57] Hadjiivanov, K. (2000) Identification of neutral and charged NxOy surface species by IR spectroscopy, Catal. Rev. Sci. Eng., 42, 71. [58] Iwamoto, M. and Yahiro, H. (2003) Zeolites in the Science and Technology of Nitrogen Monoxide Removal, pp. 951–88, In Handbook of Zeolite Science and Technology (S.M. Auerbach, K.A. Carrado, P.K. Dutta, eds). Marcel Dekker Inc, New York. [59] Spalek, T., Pietrzyk, P. and Sojka, Z. (2005) Application of the genetic algorithm joint with the Powell method to nonlinear least-squares fitting of powder EPR spectra, J. Chem. Inf. Model., 45, 18. [60] Richter-Addo, G.B. and Legzdins, P. (1992) Metal Nitrosyls. Oxford University Press, New York. [61] Wigner, E. and Witmer, E.E. (1928) On the structure of the spectra of two-atomic molecules according to quantum mechanics, Z. Physik., 51, 859. [62] Katakis, D. and Gordon, G. (1987) Mechanism of Inorganic Reactions. Wiley, New York. [63] Pietrzyk, P., Sojka, Z. (2007) Co2+ /Co0 redox couple revealed by EPR spectroscopy triggers preferential coordination of reactants during SCR of NOx with propene over cobalt-exchanged zeolites, Chem. Commun., 1930. [64] Volodin, A.M., Dubkov, K.A. and Lund, A. (2001) Direct ESR detection of S = 3/2 states for nitrosyl iron complexes in FeZSM-5 zeolites, Chem. Phys. Lett., 333, 41. [65] Fisicaro, P., Giamello, E., Berlier, G. et al. (2003) Paramagnetic nitrosyliron adducts in pentasilic zeolites: An EPR study, Res. Chem. Intermed. 29, 805. [66] Pietrzyk, P., Kukulska-Zajac, E. and Lorens, D. et al. (2004) The interaction of Co and No with Co sites in CoZSM-5 zeolite. FTIR, EPR and theoretical studies, Stud. Surf. Sci. Catal., 154 B, 1589. [67] Sojka, Z. and Witkowski, S. (2002) EPR study of CO adsorption onto CoZSM-5 zeolite: Evidence for spin crossover upon coordination and |z2 , 2 A1 ground state, Top. Catal., 18, 279. [68] Ford, P.C., Bourassa, J., Miranda, K. et al. (1998) Photochemistry of metal nitrosyl complexes. Delivery of nitric oxide to biological targets, Coord. Chem. Rev., 171, 185. [69] Aylor, A., Lobree, L., Reimer, J. et al. (1997) NO adsorption, desorption, and reduction by CH4 over Mn-ZSM-5, J. Catal., 170, 390. [70] Valyon, J. and Hall, W.K. (1993) Studies of the surface species formed from NO on copper zeolites, J. Phys. Chem., 97, 1204. [71] Sojka, Z., Pietrzyk, P., Martra, G. et al. (2006) EPR and DFT study of NO interaction with Ni/SiO2 catalyst: Insight into mechanistic steps of disproportionation process promoted by tripodal surface nickel complex, Catal. Today, 114, 154. [72] Giamello, E., Murphy, D., Magnacca, G. et al. (1992) The interaction of NO with copper ions in ZSM-5: an EPR and IR investigation, J. Catal., 136, 510. [73] Gil, B., Datka, J., Witkowski, S. et al. (2000) Copper redox chemistry in Cu/ZSM-5 zeolites: EPR, IR and DFT investigations, Stud. Surf. Sci. Catal., 130, 3249. [74] Moden, B., Da Costa, P., Fonfe, B. et al. (2002) Kinetics and mechanism of steady-state catalytic NO decomposition reactions on Cu-ZSM5, J. Catal., 209, 75. [75] Pietrzyk, P., Gil, B. and Sojka, Z. (2007) Combining computational and in situ spectroscopies joint with molecular modeling for determination of reaction intermediates of deNOx process – CuZSM-5 catalyst case study, Catal. Today. doi:10.1016/j.cattod.2006.09.033. [76] Spoto, G., Bordiga, S., Scarano, D. et al. (1992) Well defined CuI (NO), CuI (NO)2 and CuII (NO)X (X = O− and/or NO2 − ) complexes in CuI -ZSMS prepared by interaction of H-ZSM5 with gaseous CuCl, Catal. Lett., 13, 39.

DFT Modeling and Spectroscopic Investigations

65

[77] Henriques, C., Ribiero, M., Abreu, C. et al. (1998) An FT-IR study of NO adsorption over Cu-exchanged MFI catalysts: Effect of Si/Al ratio, copper loading and catalyst pre-treatment, Appl. Catal. B Environ., 16, 79. [78] Konduru, M.V. and Chuang, S.S.C. (2000) Dynamics of NO and N2 O decomposition over Cu-ZSM-5 under transient reducing and oxidizing conditions, J. Catal., 196, 271. [79] Bartke, T.U., Franchy, R. and Ibach, H. (1992) Adsorption and photoreactions of NO on Nb(110) at 20 K, Surf. Sci., 272, 299. [80] Ludviksson, A., Huang, C., Jänsch, H.J. et al. (1993) Isotopic studies of the reaction of NO on silver surfaces, Surf. Sci., 284, 328. [81] Ruggiero, C.E., Carrier, S.M. and Tolman W.B. (1994) Reductive disproportionation of NO mediated by copper complexes: Modeling N2 O generation by copper proteins and heterogeneous catalysts, Angew. Chem. Int. Ed., 33, 895. [82] Tolman W.B. (1995) Synthetic Modeling of the Interactions of Nitrogen Oxides with Copper Proteins: Copper Nitrosyl Complexes Relevant to Putative Denitrification Intermediates, Adv. Chem. Ser., 246, 195. [83] Ford, P.C. and Lorkovic, I.M. (2002) Mechanistic aspects of the reactions of nitric oxide with transition-metal complexes, Chem. Rev., 102, 993, and references therein. [84] Lin, R. and Farmer, P.J. (2001) O atom transfer from nitric oxide catalyzed by Fe(TPP), J. Am. Chem. Soc., 123, 1143.

This page intentionally left blank

Chapter 3

SURFACE SCIENCE STUDIES OF THE MECHANISM OF NOx CONVERSION: CORRELATIONS BETWEEN KINETICS IN VACUUM VERSUS UNDER CATALYTIC CONDITIONS F. Zaera∗ Department of Chemistry, University of California, Riverside, CA 92521, USA ∗

Corresponding author: Department of Chemistry University of California, Riverside, CA 92521, USA. E-mail: [email protected]

Abstract A review is provided on the contribution of modern surface-science studies to the understanding of the kinetics of DeNOx catalytic processes. A brief overview of the knowledge available on the adsorption of the nitrogen oxide reactants, with specific emphasis on NO, is provided first. A presentation of the measurements of NOx reduction kinetics carried out on well-characterized model system and on their implications on practical catalytic processes follows. Focus is placed on isothermal measurements using either molecular beams or atmospheric pressure environments. That discussion is then complemented with a review of the published research on the identification of the key reaction intermediates and on the determination of the nature of the active sites under realistic conditions. The link between surface-science studies and molecular computational modeling such as DFT calculations, and, more generally, the relevance of the studies performed under ultra-high vacuum to more realistic conditions, is also discussed.

1. INTRODUCTION With the advance of three-way catalysis for pollution control, used mainly in automobile catalytic conversion but also for the purification of gas exhausts from stationary sources, a need has arisen to develop a basic understanding of the reactions associated with the reduction of nitrogen oxides on transition metal catalytic surfaces [1,2]. That conversion is typically carried out by using rhodium-based catalysts [3], which makes the process quite expensive. Consequently, extensive effort has been placed on trying to minimize the amount of the metal needed and/or to replace it with an alternatively cheaper and more durable active phase. However, there is still ample room for improvement in this direction. By building a molecular-level picture of theprocesses involved, Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

68

Past and Present in DeNOx Catalysis

the surface-science community attempts to provide the chemical foundations for the development of a new generation of catalysts for NOx conversion. Perhaps, the greatest emphasis in this field has been on the characterization of the adsorption and thermal chemistry of the reactants, of NO in particular. However, since this is not the focus of the present chapter, only some highlights of that research are presented below, mainly to place the kinetic studies in proper context. Instead, the discussion focuses on the kinetic information that has derived from modern surface-science studies of DeNOx reactions. By and large, most of the results surveyed comprised experiments using single-crystal surfaces, although occasional reference is also made to reports using more realistic catalytic models. The bulk of the work in this area has been carried out on late transition metals, rhodium and palladium in particular, so these also occupy most of the following review. Finally, since nitrogen monoxide has been the most used molecule in this research, the majority of the examples cited below involve that reactant.

2. TEMPERATURE-PROGRAMMED DESORPTION STUDIES The most commonly used method to obtain kinetic information in ultrahigh vacuum (UHV), surface-science studies is temperature-programmed desorption [TPD, sometimes also called thermal desorption spectroscopy (TDS) or temperature-programmed reaction (TPR)] [4a]. In these experiments, a fixed amount of the reactant is first adsorbed on the surface at low temperature, and that temperature then is ramped (typically in linear fashion) while the desorbing products are detected using mass spectrometry. A wealth of information can be extracted from such experiments. First, the nature of the products of the reaction can be identified by their cracking patterns. Second, since the experiments are carried out in chambers with high pumping speeds, desorption rates can be derived directly from the mass spectrometry signals. Finally, the yields of different reaction products can be estimated by integration of the rate data. An analysis of the evolution of the rates as a function of surface temperature can provide information on activation barriers [5]. Much work has been published on the desorption and decomposition kinetics of nitrogen oxides on many surfaces using this technique. An example is the report of Niemantsverdriet et al. of their detailed study of the chemistry of NO on Rh(111) [6] and Rh(100) single crystals [7]. As in many other surfaces, both molecular desorption and N2 formation were observed. On Rh(111), it was determined that at 100 K NO adsorption is molecular, with a high initial sticking probability and via a mobile precursor state. Complementary static secondary ion mass spectrometry (SSIMS) studies identified two distinct NO adsorption states indicative of threefold adsorption at low coverages and bridge site occupation at higher surface concentrations. Three characteristic coverage regimes were identified in terms of the dissociation of the NO (Figure 3.1): (1) NO < 025 ML (monolayers), where NO dissociates completely at temperatures between 275 and 340 K, with effective pre-exponential factor and activation barrier of approximately NO = 1011 s−1 and ENO = 65 kJ/mol, respectively. The adsorbed nitrogen (Nads ) and oxygen (Oads ) atoms that result from this NO surface dissociation desorb as N2 and O2 , respectively, with desorption parameters EN2 = 118 ± 10 kJ/mol and N2 = 10101±10 s−1 for N2 in the zero coverage limit. (2) 025 < NO < 050 ML, where the desorption kinetics of N2 is strongly influenced by the presence of coadsorbed oxygen. Part of the NO desorbs molecularly, with an estimated desorption barrier of ENO =113 ± 10 kJ/mol

Surface Science Studies of the Mechanism of NOx Conversion 0.40 ML

N2

N2

TPD

0.15 ML

0.65 ML

N2

NO

NO

“θN”

TPSSIMS

69

“θNO”

“θN”

“θN”

“θNO” “θNO” 300

500

Temperature/K

700

300

500

Temperature/K

700

300

500

700

Temperature/K

Figure 3.1. NO (- - -) and N2 (—) temperature-programmed desorption (TPD) rates (top), and NO (- - -) and N (—) temperature-programmed static secondary ion mass spectrometry (TPSSIMS) ion intensity ratios (bottom) obtained during temperature-programmed reactions of NO on Rh(111) at low (left panel), medium (central panel) and high (right panel) initial NO coverages [79]. The NO TPD spectra have been scaled down by a factor 4 with respect to the N2 TPD spectra. This figure summarizes most of the key features of this NO/Rh(111) chemistry, in particular the coverage dependence of the dissociation of NO. (Reproduced with permission from the American Institute of Physics, Copyright 2003).

and a pre-exponential factor of NO =10135±10 s−1 . The dissociation of NO becomes progressively inhibited due to site blocking, with the onset shifting from 275 K at 0.25 ML to 400 K at 0.50 ML (coinciding with the temperature of NO desorption). In addition, the accumulation of nitrogen and oxygen atoms on the highly-covered surface causes a destabilization of the nitrogen atoms and leads to a second low-temperature desorption state for N2 . (3) Finally, at NO > 050 ML, NO dissociation is completely self-inhibited (presumably, because all the required surface sites are blocked). The initial desorption of the more weakly bound – perhaps bridged – NO does not generate the sites required for dissociation; these become available only after the desorption of the alleged triply-coordinated, more strongly bonded, NO. Similar qualitative behavior has been reported on other surface planes and on other metals. The role of the defects on this chemistry has been investigated by using highly stepped, and even kinked, surfaces. In one study, the dissociation of NO was studied on a number of Rh single-crystal surfaces consisting of (111) and (100) terraces [8]. In that work it was found that the kinetics of the thermal decomposition of NO and the subsequent formation of N2 are not dependent on the structure of the surface at saturation coverages. The presence of Nads and/or Oads do inhibit the dissociation of NO, and under an excess of Oads NO is formed via recombination of Nads and Oads atoms. The atomic nitrogen was found to be more stable on (100) terraces than on (111) terraces, but again, the steps seem to exert no effect on the N2 or NO TPD spectra. This is in contrast with most other studies, both experimental and theoretical, which do report some effect on the rates of NO dissociation by surface defects. For instance, in a comparative study on flat Pd(111) versus stepped Pd(112) surfaces, Yates and co-workers observed that the presence of

70

Past and Present in DeNOx Catalysis

steps on the Pd(112) plane enhances the thermal dissociation of NO, even though they determined that adsorption at the step sites is not a criterion for this decomposition [9]. Nozoye et al. reported that about half of the NO molecules on Pt(112) adsorb at the terrace sites (the remaining half adsorbing at the steps), and that about half of the NO molecules adsorbed at step sites then decompose at around 483 K, producing N2 [10]. Density functional theory calculations confirm the adsorption states of NO on Pt, and the fact that the dissociation of NO is the rate-limiting step in the formation of nitrogen [11]. One interesting aspect of the role of the nature of the metal used in defining the NO conversion is that this can be used to tune selectivity. For example, in their work on the conversion of NO on Pt(100), Koel et al. found that on pure (although reconstructed) Pt(100) approximately 25% of the adsorbed NO monolayer decomposes above 400 K to form N2 and O2 exclusively [12]. On the other hand, they found that alloying Sn into this Pt(100) surface not only reduces the saturation coverage and the adsorption energy of the molecularly-bound NO, but also, by removing all pure platinum two-fold bridge and four-fold hollow sites, opens a new partial reduction channel to N2 O. Based on vibrational data, it was determined that at low coverages NO adsorption takes place on the same sites and with similar geometries on both Pt(100) and Sn/Pt(100) surfaces. At monolayer saturation, however, an adsorbed N2 O forms on the Sn-containing surface together with either two modes of bent and linearly-bonded atop NO or a surface dinitrosyl species. The dinitrosyl species was proposed as a possible intermediate for the new N2 O formation pathway. Quite complex kinetic behavior has been identified on some surfaces. For instance, on Ir(100), the TPD data from NO-saturated surfaces display two N2 desorption peaks, one at 346 K from the decomposition of bridge-bonded NO, and a second at 475 K from the decomposition of atop-bonded NO molecules [13]. Interestingly, the first feature is quite narrow, indicating an autocatalytic process for which the parallel formation of N2 O appears to be the crucial step. An additional complication arises from the fact that this Ir(100) surface undergoes a (1×5) reconstruction, and that NO adsorbed on the metastable unreconstructed (1×1) phase leads to N2 desorption at lower temperatures. In another example, on the reconstructed hexagonal Pt(100) surface, when a mixed NO + CO adsorbed layer is heated, a so-called surface ‘explosion’ is observed where the reaction products (N2 , CO2 and N2 O) desorb simultaneously in the form of sharp peaks with half-widths of only 7 to 20 K. The shape of the TPD spectra suggests again an autocatalytic mechanism [14]. The reactions of NO with a number of reducing agents have also been explored by means of TPD experiments, by starting with surfaces co-dosed with both reactants. The majority of this work has involved the use of carbon monoxide. On Rh(111), it was determined that neither NO desorption nor its dissociation are affected in any significant way by the presence of CO [15]. On the other hand, repulsion between NO and CO molecules appears to destabilize the adsorbed CO, enhancing its desorption and diminishing CO2 production. The Rh(100) surface was found to be not only more active for the dissociation of NO, but also significantly more active and selective for the subsequent oxidation of CO by the resulting oxygen atoms [16]. Recombination of N atoms, on the other hand, proceeds at a much slower pace on Rh(100) than on Rh(111). In similar fashion, the reactivity of NO under oxidation conditions has been explored by starting with oxygen-pretreated surfaces. Vibrational spectroscopy experiments show that although oxygen-predosing of a Pt(211) surface does not result in the formation of

Surface Science Studies of the Mechanism of NOx Conversion

71

any new species, it does inhibit the formation of N2 and N2 O desorption seen on the clean Pt(211) [17]. This effect has been explained based on blocking of the NO dissociation sites by the pre-adsorbed O atoms, a common effect also reported in experiments with coadsorbed atomic nitrogen. Less, but still significant, information is available on the surface chemistry of other nitrogen oxides. In terms of N2 O, that molecule has been shown to be quite reactive on most metals: on Rh(110), for instance, it decomposes between 60 and 190 K, and results in N2 desorption [18]. NO2 is also fairly reactive, but tends to convert into a mixed layer of adsorbed NO and atomic oxygen [19]; on Pd(111), this happens at 180 K, and is partially inhibited at high coverages. Ultimately, though the chemistry of the catalytic reduction of nitrogen oxide emissions is in most cases controlled by the conversion of NO. Finally, some details of the dynamics of NO conversion reactions on surfaces have been recently probed by using angle-resolved TPD. For instance, in a study of the decomposition of N2 O on Rh(110), Matsushima and co-workers have identified four N2 peaks between 60 and 150 K originating from direct N2 O-dissociation, and a fifth feature at 160 K attributed to the desorption of N2 molecules adsorbed on the surface after previous decomposition (Figure 3.2) [20]. The appearance of each of these peaks –40° 8

β1

ΘN2O = 0.1

–20° 0°

(a) –60°

20°

40° cos θ

β1–N2

60°

0.5

7

β3

6

AR15 N signal/arb units

80°

–80°

β5 β4

AR

β2

θ = 0°

1.0

(b) –60° β2–N2

0.5

5 27°

60° 80°

1.0

(c)

4

cos16(θ – 67)

cos θ

–80°

cos100(θ – 27)

cos100(θ – 27)

–60° β3–N2

60°

0.5

45° 3

–60° β4–N2

85°

–60° 160

Ts/K

200

cos100(θ – 27) 0.5

cos θ

–80° [001]

60°

cos15(θ – 67)

80°

cos100(θ – 27)

β5–N2

240

80°

1.0

(e)

90° 0 120

cos θ

–80°

1

80

cos18(θ – 67) 1.0

(d)

67°

2

40

cos θ

–80°

0.5

60°

cos16(θ – 68)

80° [001]

Figure 3.2. Angle-resolved TPD spectra for 15 N2 from 15 N2 O adsorbed on an oxygen-modified Rh(110) [20] The left panels display the molecular nitrogen desorption traces at several angles. The five desorption components, sequentially named 1 to 5 , were deconvoluted using Gaussian peaks. The right panels report the resulting calculated angular distributions of each of those peaks, together with their deconvolution into two inclined and one cosine components. These data point to the complex dynamics associated with the decomposition of N2 O on the Rh(110) surface. (Figure provided by Professor Matsushima and reproduced with permission from Elsevier, Copyright 2005).

72

Past and Present in DeNOx Catalysis

was shown to be sensitive to annealing after oxygen adsorption, and also to the amount of N2 O exposure [21].

3. MOLECULAR BEAM STUDIES Although TPD is a versatile and useful technique widely available within the surfacescience community, it does have some limitations. For one, because the experiments are carried out under vacuum, they can only probe irreversible reactions; no readsorption of the desorbing products is possible. In addition, as the temperature is ramped during detection, the surface temperature and the reaction rates become coupled in a way difficult to separate or control. Of particular importance here is the fact that as the reactions proceed and the products desorb, the surface coverages decrease, so the rates at higher temperatures correspond to the new lower surface concentrations. In order to overcome this problem, isothermal kinetic experiments have been carried out using molecular beams [22,23]. These molecular beam techniques have been used to characterize the adsorption and decomposition of NO on Rh(111) [24], Rh(110) [25], Pd(111) [26], and Pd(110) [27]. On Rh(111), the adsorption of NO was shown to display high sticking probabilities, to be precursor-mediated at low temperatures, and to not be affected significantly by the presence of coadsorbed nitrogen and/or oxygen atoms at any temperature below 900 K [24]. Decomposition into Nads and Oads takes place at higher temperatures, and while the former eventually desorbs as N2 , the latter builds up on the surface and poisons it toward further reaction. The rate of molecular nitrogen production was found to be significant above 450 K, and to be limited by the slow diffusion of nitrogen atoms across the surface. A strong additional effect due to lateral repulsions between nitrogen and/or oxygen atoms was also inferred from the data. Similar behavior was seen on Rh(110), Pd(111), and Pd(110), except than in the latter a window of temperatures was identified where some nitrous oxide is produced [27]. However, even in that case N2 is produced with 100% selectivity above 600 K, possibly because of the low lifetime of molecular NO on the surface at those temperatures. A similar beam approach has also been used to study the adsorption of NO on structurally well-defined palladium model catalysts prepared under UHV conditions by evaporation of palladium metal on a flat thin layer of aluminum oxide [28]. With the aid of vibrational spectroscopy, it was found that atomic nitrogen and oxygen species are formed initially in the vicinity of edges and defects on the Pd particles, their preferred sites. At temperatures up to 300 K the mobility of these atomic species is suppressed, whereas at higher temperatures diffusion onto the (111) facets of the particles can occur. The presence of these atomic species was determined to control the ability of the surface to promote NO dissociation: the presence of strongly-bonded nitrogen in the vicinity of the edge or defect sites gives rise to an enhanced dissociation probability at high adsorbate coverages (Figure 3.3). The main focus of the molecular beam experiments has been to investigate the kinetic details of the catalytic reduction of NO in the presence of a reducing agent (most often CO) under isothermal steady-state conditions. This type of studies have been carried out on Rh(111) [29], Rh(110) [30], and Pd(111) [31] single-crystal surfaces. On Rh(111), we have reported systematic studies as a function of surface temperature, NO + CO

Surface Science Studies of the Mechanism of NOx Conversion NO Dissociation: Dissociation at edge/defect/(100) sites

73

NO, N, O Co-Adsorption: Preferential occupation of edge/defect sites by atomic O/N

Enhanced dissociation in the presence of atomic N

NO Depopulation of edge/ defect sites in the presence of O/N

Different influence of co-adsorbed O/N on NO dissociation barrier

(111) facet

NO Depopulation of edge/defect sites by co-addorbed N

N/O surface diffusion on to facets(>300 K), site blocking on facets NO Adsorption:

Particle edge

NO on (111) facets NO at edges/ defects/(100) facets

(100) facet

NO (on-top) at edges/defects

N

O

Pd

NO

Figure 3.3. Schematic representation of the adsorption, surface diffusion, and surface reaction steps identified by surface-science experiments on model supported-palladium catalysts [28]. Important conclusions from this work include the preferential dissociation of NO at the edges and defects of the Pd particles, the limited mobility of the resulting Nads and Oads species at low temperatures, and the enhancement in NO dissociation promoted by strongly-bonded nitrogen atoms in the vicinity of edge and defect sites at high adsorbate coverages. (Figure provided by Professor Libuda and reproduced with permission from the American Chemical Society, Copyright 2004).

beam composition, and total beam flux [32]. A maximum reaction rate was observed between 450 and 900 K, the exact temperature of this maximum being dependent on the NO:CO beam ratio. Indeed, a synergistic behavior was seen where lost in reactivity induced by increasing the CO concentration in the beam is partly compensated by a higher surface temperature. The data were consistent with the rate-limiting step of the overall NO reduction process being the production of molecular nitrogen, not the dissociation of NO. The NO + CO conversion rate was also found to be directly proportional to the coverage of atomic oxygen on the surface. The relation between reaction rates and nitrogen coverages, however, proved to be much more complex, since an inverse correlation between these two parameters was in fact seen in most cases. The buildup of a critical coverage of atomic nitrogen was found to be necessary to trigger the production of N2 . This critical coverage of strongly-held nitrogen was determined to not depend in any significant way on the composition of the beam, but to decrease with reaction temperature in all cases. Similar results were reported on Pd(111), except that the steady-state reaction rates attained in that case displayed a maximum at a fixed temperature of about 475 K nearly independently of beam composition [31]. In addition, a minor production of N2 O was detected on that surface. Transient kinetic experiments have also been carried out to complement the information deduced from the steady-state measurements [33]. Systematic variations were observed during the transition from the clean surface to the steady-state catalytic regime that correlate well with the overall reaction rates in the latter. Specifically, there is a time delay in the production of molecular nitrogen because of the need to buildup a threshold of atomic nitrogen coverage on the surface. This atomic nitrogen coverage, which could

74

Past and Present in DeNOx Catalysis

Fraction of 14N remaining on surface

be estimated both by the time delay in the transient and by TPD experiments after the reaction, was found to depend on the NO:CO ratio in the reaction mixture, increasing at a given temperature as the beam becomes richer in CO. Initial sticking coefficients were also determined for both NO and CO in NO + CO mixtures as a function of surface temperature and beam composition. In order to explore the origin of the delay in molecular nitrogen production and the nature of the surface atomic nitrogen inferred from the work reported above, additional isotope-labeling experiments were performed where 14 N-dosed surfaces were subsequently exposed to 15 NO + CO gas mixtures for varying times [34]. Postmortem TPD data from the resulting surfaces indicated a non-statistical distribution of isotopes in the resulting molecular nitrogen, the yield of the mixed 14 N15 N isotopomer being significantly lower than that expected on statistical grounds. Monte Carlo simulations explained the observed isotopic distributions in terms of the formation of islands with the nitrogen isotopes distributed in a layered structure, the 14 N atoms in a core surrounded by a 15 N outer shell [34a] (Figure 3.4). In addition, the direct measurement of nitrogen production rates in those experiments indicated that the replacement of surface 14 N by 15 N upon switching from 14 NO to 15 NO in the gas mixture occurs almost exclusively via the

1.4

NO + CO/Rh(111) Conversion Surface Nitrogen Isotopic Exchange

1.2

1:1 15NO:CO Beam on 14N-Covered Rh Steady State ΘN = 0.17 ML, T = 480 K k p-k p23-k p -k p4-k

1.0 14

N N

15

0.8

0.6 k = 0.13 s–1 p = 0.15 0.4

0.2

4

⎛ Ni ⎞ exp(–pikt) ⎠ Total

X14N =

Σ ⎝N i=0

50

100

0.0 0

150

200

250

550

Exposure time to 15NO + CO/s

Figure 3.4. Isotope-labeling molecular beam results pointing to the formation of nitrogen surface islands during NO reduction by CO on Rh(111) [34a]. The solid circles represent the fraction of 14 N from the total adsorbed nitrogen on the surface during the steady-state conversion, plotted as a function of the time elapsed after replacing the 14 NO + CO beam with an equivalent 15 NO + CO mixture. The original 14 N atoms are slowly replaced by new 15 N, as expected, but that follows complex kinetics only explained by a model that assumes the formation of surface nitrogen islands (inset and solid line). According to this model only the outer nitrogen atoms are available for reaction, and those are converted to molecular nitrogen not by direct recombination but rather via a reaction with new incoming NO molecules to form a N−NO intermediate. (Figure provided by Professor Zaera and reproduced with permission from the American Chemical Society, Copyright 2002).

Surface Science Studies of the Mechanism of NOx Conversion

75

formation of 14 N15 N [35]. This provides direct kinetic evidence for a mechanism for molecular nitrogen production involving the formation of a N−NO intermediate. There is, however, an small desorption of 14 N14 N immediately after switching from 14 NO + CO to 15 NO + CO mixtures, a fact that points to the role of neighboring adsorbates in facilitating the dissociation of the N−NO intermediates [36]. The results from experiments with mixed 14 NO + 15 NO + CO indicated additional reactions at the edges of previously deposited 14 N islands and the growth of new mixed 14 N + 15 N surface clusters [37]. In order to test the possibility of the formation of N2 O-type intermediates forming during the catalytic conversion of NO on Rh(111), kinetic molecular beam measurements were also carried out directly with N2 O, by itself and in mixtures with CO [38]. The decomposition of pure N2 O was determined to occur at temperatures as low as 120 K, to follow first order kinetics, and to lead to the stoichiometric production of N2 (g) and atomic adsorbed oxygen. However, two unusual observations derived from this work, namely: (1) lower rates for N2 O decomposition are seen at higher reaction temperatures; and (2) lower total nitrogen yields and final oxygen surface coverages accompany that behavior. It was determined that after the rhodium surface is rendered inactive by N2 O decomposition at high (520 K) temperatures, significant activity is still possible at lower (350 K) temperatures [39]. Monte Carlo simulations explain these observations by assuming that the surface sites required for the activation of adsorbed N2 O increase in size with increasing reaction temperature [39]. It was also determined that the N2 O conversion is poisoned by the oxygen atoms deposited as byproducts unless a reducing agent such as CO is used for their removal from the surface, in which case N2 O reduction can be carried out catalytically. Steady-state reaction rates were determined for different temperatures and N2 O:CO beam mixtures, and deemed to be controlled by the rate of oxygen removal, not by the decomposition of the N2 O. By using cross-correlation time-of-flight techniques, Matsushima and co-workers have investigated the dynamics of the desorbing products from steady-state NO + CO reactions on Pd(110) and Rh(110) [40]. They found that while the desorption of CO2 is sharply collimated along the normal surface on both surfaces, below 650 K N2 desorbs from Pd(110) narrowly around 40 off the normal surface in the plane along the [001] direction, and with a translational temperature of about 3600 K. The inclined desorption was assigned to the decomposition of a N2 O intermediate, which was independently shown to yield such highly peaked N2 desorption distributions [20]. More normal desorption was detected at higher temperatures and complete normal N2 desorption was seen from Rh(110), implying that N2 O formation may not be as prominent on these surfaces, but, as discussed in the preceding paragraphs, such an intermediate is central to the NO conversion on Rh(111). Combined, all these results point to the importance of the nature of the metal and the structure of the surface in determining the relative relevance of the competing elementary steps in these reactions. Additional dynamic information was obtained by the group of Heinzmann via the use of pulsed beams with rotationally oriented NO molecules [41]. State-selected NO molecules were focused onto a Pt(100) plane with their molecular axis oriented so that either preferential N-end or O-end collisions occur with the surface. The sticking probability of the NO was found to be larger for N-end collisions and to decrease with increasing translational energy, and a similar decrease was observed in the orientation-dependent scattering asymmetry. Also, on both Pt(100) [42] and Rh(100) [43] single-crystals predosed with CO it was found that the yield of the resulting CO2 depends strongly on the

76

Past and Present in DeNOx Catalysis

initial NO orientation [43]. A comparison of the observed steric effects for CO2 and NO led the authors to conclude that the CO2 reaction asymmetry may induce a NO trapping asymmetry. However, the very high value measured for that asymmetry at the beginning of the CO2 production could only be fully explained if an additional transition in the reaction mechanism, from a direct to an indirect surface reaction channel, is considered, presumably an orientation-dependent NO dissociation step. Finally, the effect that an oxidizing environment exerts on the NO + CO conversion reactions was studied by adding molecular oxygen to the gas mixture [44]. The addition of this O2 was found to inhibit the activity of the rhodium catalyst toward NO reduction in most cases, as expected. The reason for this behavior, however, was determined not to be the consumption of some of the CO in the mixture by the added O2 , but rather a poisoning of the adsorption of CO by adsorbed atomic oxygen. In fact, oxygen addition to the NO + CO mixtures reduces not only the rates of N2 production, but also those of CO2 formation. Moreover, NO was proven to always compete favorably against O2 for the consumption of CO. Optimum reaction rates for the production of N2 and CO2 are reached at temperatures around 500–600 K and under conditions leading to stoichiometric coverages of all reactants. An interesting consequence of this is the fact that with CO-rich mixtures the addition of oxygen sometimes actually facilitates, not poisons, NO reduction, presumably because it helps in the removal of the excess CO from the surface. A synergy was observed in terms of reaction rate maxima between temperature and beam composition, CO-richer mixtures requiring higher temperatures to reach comparable reaction rates. This is explained by a decrease in CO surface coverage because of the increase in desorption rate with temperature, a trend that also explains the gradual increase in poisoning of the NO-reducing activity of Rh by oxygen with increasing temperature. Alternations between oxygen-rich and oxygen-lean beams as a way to better manage NO reduction under net oxidizing atmospheres were shown to work well on this surface.

4. ISOTHERMAL KINETICS UNDER VACUUM Although the temperature-programmed desorption and molecular beam studies of nitrogen oxide reactions have been quite useful in the determination of kinetic parameters for the surface processes in this catalysis, both approaches suffer from the limitation of being capable to detect only desorbing molecules. In order to obtain information on the identity and temporal evolution of surface species, the use of other spectroscopies is required. Much work has been carried out on the characterization of stable species in surface-science studies of NOx molecules on metals, but a review of that work is beyond the scope of the present chapter. Much less has been done on the determination of the surface kinetics of reactions on those surfaces. The work performed by the group on Niemantsverdriet et al. using SSIMS discussed in Section 1 belongs to this category. Below we cite other examples of research in this area. These surface kinetics studies initially focused on the dissociation of NO. For instance, Comelli and co-workers reported on the kinetics of the isothermal decomposition of NO on Rh(110) at temperatures ranging from 198 to 240 K and NO coverages below NO ≤ 03 ML [45]. Auger electron spectroscopy (AES) lineshape analysis was used to measure the amount of undissociated NO as a function of time, and the resulting NO (t)

Surface Science Studies of the Mechanism of NOx Conversion

77

plots for the different reaction temperatures combined to show that the initial dissociation rate obeys first-order kinetics, with Ea = 4.9 ± 0.4 kcal/mol and  = 1 × 1025±05 s−1 . Compositional and structural changes occurring at various reaction temperatures were considered to explain the surprisingly low values obtained for these kinetic parameters. More recent studies using synchrotron radiation X-ray photoemission spectroscopy (XPS) identified, in addition to the expected atomic nitrogen and oxygen and molecular upright NO adsorbates, a new surface species characterized by an O 1s binding energy of 530.7 eV and a N 1s binding energy similar to that of the atomic nitrogen [46]. This state, seen only within a narrow range of coverages, was tentatively assigned to a ‘lying down’ NO bonding configuration. The effect of surface defects was probed by looking into the interaction of NO with a flat and two stepped Rh(111) surfaces, also using fast high resolution XPS [47]. From the O 1s intensity during uptake on the Rh(533) plane, it was determined that the NO initial sticking coefficient is temperature-independent between 330 and 490 K. In addition, the N 1s spectra revealed the consecutive appearance of two atomic nitrogen species attributed to adsorption on terrace and step sites, NT and NS respectively. It was seen that NT is the only species that forms in the initial stages of the adsorption, whereas the NS species develops only later on in the uptake (Figure 3.5). This finding indicates that in the low-coverage regime (total coverage ≤ 025 monolayers) NO dissociation occurs on terrace adsorption sites. On the other hand, the stepped surfaces conserve a high reactivity at higher coverages, and NO dissociation continues on the steps even after the terraces have been filled. The abrupt changes seen in the NO dissociation probability with increasing adsorbate coverage were explained in terms of the destabilization of the NT species by repulsive interactions and of subsequent compression of the adsorbed O and NO layers. The kinetics of the steady-state conversion of NO with H2 was also studied on Rh(533) [48]. Specifically, hydrogen-rich NO:H2 mixtures were investigated in the 10−6 mbar pressure range, where oscillatory behavior is often observed, and where the three reaction products, N2 , NH3 and H2 O, show a large divergence in formation rates on the two branches of the heating-cooling cycles. This hysteresis proved to be the most prominent between 450 and 500 K. Five nitrogen- and/or oxygen-containing surface species were identified at different stages of this hysteresis. It was established that the reactive surface is covered by an atomic nitrogen species, N(I), and the unreactive surface by atomic oxygen O and another atomic nitrogen species, N(II), attributable to either atomic nitrogen adsorbed at step sites or a NHx species. No NO was observed during the hysteresis loop. The results showed the presence of a very small amounts of oxygen which drastically destabilizes the first nitrogen species, N(I), and that higher oxygen coverages result in the formation of the N(II) species. Another approach to the determination of surface kinetics in these systems has been to combine molecular beams in the 10−2 –10−1 mbar pressure range with the use of the infrared chemiluminescence of the CO2 formed during steady-state NO + CO reactions. This methodology has been used to follow the kinetics of the reaction over Pd(110) and Pd(111) surfaces [49]. The activity of the NO + CO reaction on Pd(110) was determined to be much higher than on Pd(111), as expected based on the UHV work discussed in previous sections but in contrast with result from experiments under higher pressures. On the basis of the experimental data on the dependence of the reaction rate on CO and NO pressures, the coverages of NO, CO, N, and O were calculated under various flux conditions. Note that this approach relied on the detection of the evolution of gas-phase

78

Past and Present in DeNOx Catalysis

Total coverage [ML]

0.6

0.4

Rh(111) Rh(533) Rh(311)

0.2

NT coverage [ML]

0.0

0.15

0.10

0.05

Ns coverage [ML]

0.00

0.10

0.05

O coverage [ML]

0.00

0.4

0.4

0.2

0.2

0.0

0

1

2

3

4

5

6

8

0.0 10

Exposure [L]

Figure 3.5. Time evolution of the surface coverages of atomic oxygen (O) and atomic nitrogen at terrace (NT ) and step (NS ) sites, and of the total coverage of all the surface species (O + NT + NS ), during the isothermal uptake of NO on Rh(311), Rh(533), and Rh(111) single-crystal surfaces at 430 K, measured using a fast time-resolved X-ray photoelectron spectroscopy (XPS) setup [47]. Only dissociative adsorption takes place at this temperature. The coverage of the NT species reaches its maximum at exposures of about 0.7 L (vertical line), at which point the slope in the O uptake drops on Rh(111). The results strongly suggest that at total coverages below Total < 025 ML NO dissociation occurs only on terraces, but that the stepped surfaces conserve their reactivity in the high coverage regime (Total > 025 ML). (Figure provided by Professor Baraldi and reproduced with permission from the American Institute of Physics, Copyright 1999).

Surface Science Studies of the Mechanism of NOx Conversion

79

species, as in other molecular beam work. Nevertheless, additional information could be extracted in this case from the analysis of the IR emission spectra. In particular, the asymmetric vibrational temperature (TVAS) of the desorbing CO2 was seen to decrease and its bending vibrational temperature (TVB) to increase significantly with increasing surface temperature on both Pd(110) and Pd(111) crystals. On the other hand, the order of importance of the two vibrations reverses with temperature, going from the TVB being lower than the TVAS at low temperatures to the TVB being higher than the TVAS at higher temperatures. A higher TVAS compared to the TVB suggests a more bent activated complex for CO2 formation than in the opposite case, so this switch could be accompanied with a change in reaction dynamics. Also, the TVB of the CO2 produced by the NO + CO reaction on Pd(110) above 800 K is much higher than that originated by the CO + O2 reaction.

5. HIGH PRESSURE CELLS In order to better reproduce realistic conditions and close the so-called ‘pressure gap’, some studies on model single-crystal surfaces have been performed in-situ under atmospheric conditions. For instance, the room-temperature coadsorption of NO and CO on Rh(111) has been studied with scanning tunneling microscopy (STM) in the catalytically relevant range of ∼1 mbar [50,51]. The authors of that work reported that, for gas mixtures where NO is not in large excess, a mixed NO + CO coadsorbed layer with (2×2) structure is formed on the surface. The difference in binding energies between NO and CO on top sites was estimated from the measured surface and gas mole fractions of each species at about 1.5 kcal/mol, with CO being the more strongly-bonded species. This is in direct contradiction with molecular beam results at lower pressures (where NO was found to be the more stable adsorbed species, by ∼11.5 kcal/mol, instead [33]), implying that the total pressure of the reaction mixture, not just the relative CO and NO pressures, may play a significant role in determining the energetics of the reaction. However, recent in situ XPS measurements in the same Pascal pressure range have shown that even under those conditions NO readily displaces CO near room temperature [52]. It was also determined in that work that the CO molecules are displaced from threefold hollow sites when the NO partial pressure is below 30%, but from top sites only at higher NO pressures and after heating. The kinetics of the NO:CO exchange at top sites is very slow at room temperature, taking on the order of hours, and requires heating for complete exchange. Extensive work on the kinetics of NO reduction has been performed on rhodium and palladium single-crystal surfaces by using so-called ‘high pressure cells’, where the sample is transferred from the ultrahigh vacuum environment used for surface characterization to a microbatch reactor without exposure to the outside environment [53a]. In particular, Belton et al. reported that, during the reduction of NO by CO over a Rh(111) catalyst at pressures in the 1–100 mbar range, the production of CO2 , N2 O, and N2 all display the same apparent activation energy, ∼33.5 kcal/mol, in good agreement with previous catalytic work [54]. N2 O was found to be the primary N-containing product from this Rh(111) surface in the high NO pressure end, with a selectivity of about 70% regardless of the reaction temperature, reactant pressure, or NO conversion, and the CO2 , N2 O, and N2 production rates were all determined to exhibit near zero-order kinetics in

80

Past and Present in DeNOx Catalysis 500 100

(a) PCO = 4.0 Torr

CO2

PNO = 0.8 Torr

N2 N2O

TON (molecules/site-sec)

10

1

0.1 500 100

(b) PCO = 8.0 Torr

CO2

PNO = 8.0 Torr

N2O N2

10

1

0.1 1.4

1.5

1.6

1.7

1.8

1.9

2.0

1/ T (103) (K–1)

Figure 3.6. Example of the type of kinetic information available for the catalytic reduction of NO on rhodium single-crystal surfaces under atmospheric conditions. The data in this figure correspond to specific rates for CO2 , N2 O, and N2 formation over Rh(111) as a function of inverse temperature for two NO + CO mixtures: PNO = 06 mbar and PCO = 3 mbar (A), and PNO = PCO = 4 mbar (B) [55]. The selectivity of the reaction in this case proved to be approximately constant independent of surface temperature at high NO pressures, but to change significantly below PNO ∼ 1 mbar. This highlights the dangers of extrapolating data from experiments under vacuum to more realistic pressure conditions. (Reproduced with permission from the American Chemical Society, Copyright 1995).

both CO and NO pressure in that regime. On the other hand, for NO pressure around or below 0.6 mbar, the reaction selectivity was shown to depend strongly on reaction temperature, with N2 O being the major product below 635 K but N2 formation taking over above 635 K (Figure 3.6) [55]. This switchover in selectivity at elevated temperature has been repeatedly seen with low-load Rh/Al2 O3 catalysts, and may explain why it has been difficult to identify the production on N2 O in experiments under vacuum conditions. Complementary in-situ characterization of the surface species using infrared (IR) spectroscopy has provided information on the identity and coverage of the surface species involved in the NO catalytic reduction [56]. It was found that the changes observed in the surface coverages of NO and CO correlate well with the observed changes in N2 O selectivity mentioned above: below 635 K, where N2 O formation is favored, NO is the major adsorbate on the surface, whereas above 635 K, where N2 formation is preferred,

Surface Science Studies of the Mechanism of NOx Conversion

81

CO is the majority surface species. The IR data support a model in which the N2 O and N2 products are formed as adsorbed nitrogen atoms react with either adsorbed NO or N, respectively. In fact, in additional kinetic experiments using isotopic labeling, it was shown that gas-phase N2 O may not be an intermediate to N2 formation at 648 K, as reported at lower temperatures under vacuum [35]. The reaction intermediates that form during the reduction of NO on rhodium films under atmospheric pressures have also been characterized by surface-enhanced Raman spectroscopy (SERS) [57,58]. In the case of NO + CO mixtures [58], facile NO dissociation was observed even at room temperature. A change in activation temperature with temperature was observed, and explained by a change in rate-limiting step from N2 formation at low temperatures to NO dissociation at higher temperatures. Nitrogen atoms appear to inhibit CO2 production, but, surprisingly, the faster removal of the atomic nitrogen at higher temperatures is not accompanied by the expected increase in NO reduction rate, presumably because under those conditions the rhodium surface becomes oxidized. The structure sensitivity of the NO + CO reaction was tested by running kinetic experiments on different crystal faces of rhodium single-crystals. Under the 1–100 mbar pressures used in these studies, the NO + CO activity, as measured by the rate of NO loss, proved between 1.3 and 6.3 times faster over Rh(110) than over Rh(111) [59]. The (110) surface also exhibits a lower apparent activation energy than does the (111) surface, 27.2 vs. 34.8 kcal/mol, presumably because of a slightly more facile NO dissociation process on the more open (110) surface. A more complex behavior was seen on Rh(100), where the activation energy was determined to decrease by a factor of 2.3 in going from 528 to 700 K [60]. This suggests that the reaction kinetics in that case dominated by variations in NO coverage, going from a NO desorption rate-limiting step (needed to open up empty reaction sites) at low temperatures to NO dissociation-dominated kinetics at high temperatures. Also, the selectivities for N2 O vs. N2 production can vary significantly between the Rh(110) and Rh(111) surfaces, with the more open Rh(110) tending to make significantly less N2 O than Rh(111) under virtually all the conditions probed in these experiments (presumably because of the greater steady-state concentrations of adsorbed N atoms on the former surface, as confirmed by post-reaction XPS measurements). On Rh(100) the data are consistent with a mechanism where the formation of N2 occurs via the reaction of N atoms with adsorbed NO [61], so it is quite possible that N2 is in fact made from N2 O dissociation in all cases, and that the N2 vs. N2 O production selectivity is determined by the relative rates of N2 O dissociation versus N2 O desorption. Finally, a comparative study of the conversion of NO + CO mixtures over Rh(111) single crystals versus Al2 O3 -supported Rh catalysts concluded that the two exhibit substantially different activation energies and specific reaction rates [62]. This discrepancy could be rationalized by assuming that the dissociation of the NO occurs much more slowly on supported Rh than on Rh(111). The effect of oxidizing atmospheres on the reduction of NO over rhodium surfaces has been investigated by kinetic and IR characterization studies with NO + CO + O2 mixtures on Rh(111) [63]. Similar kinetics was observed in the absence of oxygen in the gas phase, and the same adsorbed species were detected on the surface as well. This result contrasts with that from the molecular beam work [44], where O2 inhibits the reaction, perhaps because of the different relative adsorption probabilities of the three gas-phase species in the two types of experiments. On the other hand, it was also determined that the consumption of O2 is rate limited by the NO + CO adsorption-desorption

82

Past and Present in DeNOx Catalysis

equilibrium controlling the vacant sites required for dissociative adsorption, and that, consequently, the oxidation of CO by adsorbed atomic oxygen is significantly slower than previously estimated; that agrees quite well with the molecular beam work. In the atmospheric pressure studies it appears that the NO conversion activity is dominated by the dissociation of adsorbed NO and the selectivity toward N2 O production by the NO surface coverage, as in the absence of gas-phase oxygen. On the basis that the nature of the metal is expected to affect its catalytic performance, the effect of alloying on the conversion of NO + CO mixtures was studied by using a Rh90 Pt10 (111) surface [64]. It was found that this Rh-Pt single-crystal alloy displays a catalytic activity quite similar to that of pure rhodium in terms of activation energies, reaction orders, products, and selectivities. The turnover numbers for the Rh90 Pt10 (111) alloy are slightly lower than those for Rh(111) when compared on a per surface atom basis, but virtually the same per surface Rh atom, suggesting that the primary effect of Pt is to dilute the Rh surface atom concentration. It was concluded that the Rh90 Pt10 (111) single-crystal mimics the behavior of supported Rh-Pt catalysts in that both show high selectivity for N2 O at low temperatures, low conversions or high NO:CO ratios, and low or zero N2 O production at high temperatures and high NO conversions. Much detailed characterization of the kinetics of NO reduction in atmospheric environments, that is, at pressures up to 240 mbar, has also been performed by Goodman and co-workers on palladium single crystals [65]. They have determined by using in-situ infrared spectroscopy that, under reaction temperatures below 500 K and pressure ratios PCO /PNO ≥ 15, the conversion of NO + CO mixtures on Pd(111) is accompanied by the formation of an isocyanate (−NCO) intermediate (Figure 3.7) [66]. The formation of

CO + NO/Pd(111) PCO+NO = 240 mbar CO:NO = 1.5

0.005 1922

PM-IRAS intensity (a.u.)

2256

1879

2119

1744

1568 625 K 600 K 550 K 500 K 450 K 400 K 350 K 300 K

2200

2000

1800

1600

Wavenumber (cm–1)

Figure 3.7. In-situ reflection-absorption infrared (RAIRS) spectra as a function of catalyst temperature from a Pd(111) single-crystal surface in the presence of a NO + CO gas mixture (240 mbar, PCO /PNO = 15) [66]. The data clearly show the appearance of an isocyanate-related band at 2256 cm−1 at temperatures above 500 K. In-situ spectroscopic experiments such as these have proven indispensable to detect and identify key reaction intermediates for the catalytic reduction of NO on metal surfaces. (Figure provided by Professor Goodman and reproduced with permission from the American Chemical Society, Copyright 2003).

Surface Science Studies of the Mechanism of NOx Conversion

83

this species requires total NO + CO pressures of at least 0.6 mbar, but once produced, it is stable within the entire temperature range studied (300–625 K) [67]. Both a substantial increase in activity for CO2 , N2 O, and N2 formation and an increase in N2 O selectivity are observed above 550 K, apparently the result of temperature-dependent changes in the NO :CO ratio. The apparent activation energy of the reaction was determined to be 54 ± 21 kJ/mol.

6. MODELING OF SUPPORTED CATALYSTS In addition to performing experiments under pressures similar to those encountered in real processes to bridge the ‘pressure gap’, surface scientists have also been increasing the level of complexity of the model surfaces they use to better mimic real supported catalysts, thus bridging the ‘materials gap’. A few groups, including those of Professors Freund and Henry, have extended this approach to address the catalytic reduction of NO. The former has published a fairly comprehensive review on the subject [23]. Here we will just highlight the information obtained on the reactivity of NO + CO mixtures on these model supported catalysts. The first molecular beam study on the NO + CO reaction on a supported catalyst system was reported by Valden et al. [68]. In their work, a catalyst consisting of highly dispersed (85%) palladium clusters supported on -alumina, prepared ex situ, was used. The authors determined that while CO adsorbs and desorbs molecularly on those catalysts throughout the whole coverage range probed, exposure to NO leads to a mixture of molecularly-bonded NO and atomic nitrogen, and oxygen. Reactivity could also be measured between CO from the gas phase and pre-adsorbed NO. Freund and co-workers have investigated the interaction of NO on better-defined model catalysts, consisting of palladium particles deposited under UHV on an ordered Al2 O3 film grown on NiAl(110), as briefly discussed in Section 2 [28,69]. They found that NO decomposes slowly even at low surface temperatures (100 K) on the pure Al2 O3 support, a reaction that starts at oxide defect sites, produces a variety of Nx Oy surface species, and involves strong structural transformations of the oxide film. However, in the presence of the palladium particles, that low-temperature decomposition channel is strongly suppressed, and NO dissociation occurs only on the Pd particles and only above 300 K. Henry and co-workers have carried out a more systematic study of the kinetics of the NO + CO reaction on a Pd/MgO(100) model catalysts as a function of NO and CO pressure, sample temperature, and particle size [70,71]. Three samples with particle mean sizes of 3, 7, and 16 nm, as determined by transmission electron microscopy (TEM), were studied. It was found that under steady-state NO dissociates between 450 and 720 K on all particles to produce CO2 , N2 , and a small amount of N2 O, the latter only below 570 K. The rate-limiting step for the reaction was determined to be the dissociation of NO at low temperatures and the adsorption of CO at high temperatures. The plot of the reaction rates measured as a function of temperature has a ‘volcano’ shape with a maximum activity that shifts toward higher temperatures with increasing CO pressures (Figure 3.8), as also seen on Rh(111) [32]. At first sight the turnover rate for CO2 production appeared to increase with decreasing particle size, but after correcting for NO and CO adsorption on the support it was concluded that the (111) facets found predominantly in the medium-sized particles are the more active for the

84

Past and Present in DeNOx Catalysis

0.012

pNOequiv = pCO = 5.0 × 10–8 Torr

d = 15.6 nm d = 6.9 nm

CO2 turnover rate [s–1]

0.010

d = 2.8 nm 0.008

0.006

0.004

0.002

0.000 150

200

250

300

350

400

450

Ts [°C] 0.005

pNOequiv = pCO = 5.0 × 10–8 Torr

d = 15.6 nm

Effective turnover rate [a.u.]

0.004

d = 6.9 nm d = 2.8 nm

0.003

0.002

0.001

0.000

–0.001 150

200

250

300

350

400

450

Ts [°C]

Figure 3.8. Kinetic data from molecular beam experiments with NO + CO mixtures on a Pd/MgO(100) model catalyst [70]. The upper panel displays raw steady-state CO2 production rates from the conversion of PCO = PNO = 375 × 10−8 mbar mixtures as a function of the sample temperature on three catalysts with different average particle size (2.8, 6.9, and 15.6 nm), while the bottom panel displays the effective steady-state NO consumption turnover rates estimated by accounting for the capture of molecules in the support. After this correction, which depends on particle size, the medium-sized particles appear to be the most active for the NO conversion. (Reproduced with permission from Elsevier, Copyright 2000).

Surface Science Studies of the Mechanism of NOx Conversion

85

reaction, and that the (100) surfaces more prevalent in the larger particles display lower reactivity. Again, this is consistent with ideas from experiments on single crystals, where the higher dissociation probability of NO on the (100) surface was shown to lead to a faster poisoning by atomic nitrogen. In addition to the steady-state experiments, kinetic information was also derived from transient experiments using modulated beams on the Pd/MgO system [72]. It was found that NO dissociation is highly probable on the clean surface and leads to the deposition of two types of adsorbed nitrogen atoms: a strongly and irreversibly-bonded species and a more loosely adsorbate capable of converting to N2 during reaction. The first type of nitrogen clearly poisons the catalyst, since it is not possible to restore the original NO dissociation rate by removal of the atomic oxygen via CO titrations. Also, the oxygen surface coverage decreases rapidly as a function of time, presumably because of its diffusion into the subsurface and bulk. Residence time measurements proved that the adsorption energy of NO is slightly larger than that of CO, the same as on rhodium [33]. Additional molecular beam experiments on the kinetics of the NO + CO reaction on small size-selected Pdn clusters (1 ≤ n ≤ 30) supported on a thin MgO(100) film have been reported by Heiz et al. [73,74]. Several noticeable results have been derived from these experiments. First, it was found that clusters smaller than Pd4 are inert, and that larger Pd particles display an activity for CO2 production that increases in a non-monotonic fashion with size. Furthermore, TPD experiments revealed that CO2 production occurs in two pathways around 145 and 300 K. The high-temperature peak appears to originate from the conventional mechanism where the dissociation of molecularly adsorbed NO is followed by CO + O recombination, but the low-temperature CO2 production may occur via a direct NO + CO reaction channel on the surface instead. Interestingly, for all reactive clusters, the reaction temperature of the main mechanism is at least 100 K lower than on either palladium single crystals or larger particles, and N2 desorption also occurs at a relatively low temperatures, between 400 and 450 K (Figure 3.9), thus preventing any significant surface poisoning.

7. COMPUTATIONAL SIMULATIONS Numerous quantum mechanic calculations have been carried out to better understand the bonding of nitrogen oxide on transition metal surfaces. For instance, the group of Sautet et al. have reported a comparative density-functional theory (DFT) study of the chemisorption and dissociation of NO molecules on the close-packed (111), the more open (100), and the stepped (511) surfaces of palladium and rhodium to estimate both energetics and kinetics of the reaction pathways [75]. The structure sensitivity of the adsorption was found to correlate well with catalytic activity, as estimated from the calculated dissociation rate constants at 300 K. The latter were found to agree with numerous experimental observations, with (111) facets rather inactive towards NO dissociation and stepped surfaces far more active, and to follow the sequence Rh(100) ≥ terraces in Rh(511) > steps in Rh(511) > steps in Pd(511) > Rh(111) > Pd(100) ≥ terraces in Pd (511) > Pd (111). The effect of the steps on activity was found to be clearly favorable on the Pd(511) surface but unfavorable on the Rh(511) surface, perhaps explaining the difference in activity between the two metals. The influence of

86

Past and Present in DeNOx Catalysis

(a)

(b)

(c) 0.8

Pd30 (0.28% ML)

Pd8 (0.50% ML) Pback (CO) = 5 × 10–7 mbar Peff (NO) = 1 × 10–4 mbar T = 632 K

T = 557 K

T = 450 K

0.6

# CO2/(Pd atom * s)

CO2+ ion signal [a.u.]

CO2+ ion signal [a.u.]

T = 632 K

T = 557 K

T = 450 K

Pd8 0.4

Pd30 0.2

–1 0

1

2

T = 423 K

T = 423 K

T = 251 K

T = 251 K

3

Time [s]

4

5

0.0 6

–1 0

1

2

3

Time [s]

4

5

6

200 300 400 500 600 700

Temperature [K]

Figure 3.9. Transient CO2 formation rates on Pd30 (a) and Pd8 (b) mass-selected clusters deposited on a MgO(100) film at different reaction temperatures [74]. In these experiments CO was dosed from the gas background while NO was dosed via a pulsed nozzle molecular beam source. The turnover frequencies (TOFs) calculated from the experiments displayed in (a) and (b) are displayed in the last panel (c). CO2 formation starts at lower temperatures but reaches lower maximum rates on the larger cluster. (Figure provided by Professor Heiz and reproduced with permission from Elsevier, Copyright 2005).

coadsorbates on the dissociation of nitrogen oxide was also tested by DFT calculations on the vicinal Rh(311) surface [76]. Additional calculations on the adsorption of N2 O on Rh(110) indicated surface binding in two alternative forms, an stable horizontal configuration and a less favorable tilted geometry bonded through the terminal N atom [77]. From these, the horizontal adsorption was found to be dissociative at low coverages, readily producing molecular nitrogen. The tilted species appears to be less reactive, but to nevertheless decompose as well. The observation of these two states of N2 O may explain the complex dynamics of its decomposition observed experimentally, as discussed in Section 2 [20]. In terms of the formation of other intermediates during reaction, it was estimated that CN formation is preferred over N2 production in surfaces covered with mixtures of C and N atoms, and that HCN may be formed from adsorbed CH and N species rather than by recombination of C and N atoms into CN followed by hydrogenation [78]. In a few instances, quantum mechanical calculations on the stability and reactivity of adsorbates have been combined with Monte Carlo simulations of dynamic or kinetic processes. In one example, both the ordering of NO on Rh(111) during adsorption and its TPD under UHV conditions were reproduced using a dynamic Monte Carlo model involving lateral interactions derived from DFT calculations and different adsorption

Surface Science Studies of the Mechanism of NOx Conversion

87

sites (top, fcc and hcp) [79]. A single-site model was able to reproduce the experimental TPD data, but only by using values for the lateral interactions at odds with those from the DFT calculations. A three-site model resolved this problem. More commonly, kinetic simulations have been carried out at a more empirical level using sets of proposed elementary steps and kinetic parameters, many extracted from experimental studies [80]. For instance, the dissociation of NO on Rh(100) surfaces has been modeled by kinetic Monte Carlo simulations that included nearest neighbor (NN) and next-nearest neighbor (NNN) interactions and zero-coverage kinetic parameters obtained from experiments [81]. The lateral interactions were estimated by fitting the model to the experimental data, and all found to be repulsive, with NN interactions involving atoms typically of the order of 20–30 kJ/mol and NN interactions between molecules and all NNN interactions smaller than 10 kJ/mol (all significantly smaller than those estimated from DFT calculations). The simulations also showed that in the initial stages of the NO uptake the molecules dissociate to nitrogen and oxygen atoms, which then form ordered c(2 × 2) islands on the surface and inhibit NO dissociation at higher coverages. Monte Carlo simulations have been also used to reproduce the dynamics of adsorbates associated with NOx reduction reactions. As mentioned above, complex desorption dynamics have been observed experimentally in some instances. For example, the N2 produced from decomposition of N2 O on Rh(110) leaves the surface in five peaks associated with both the N2 O dissociation events and the desorption of the adsorbed products. Monte Carlo simulations of those spectra was possible by using a model that takes into account both channels of N2 desorption and also N2 O−O lateral interactions to stabilize N2 O adsorption [18]. The extensive studies of the microkinetics of the NO + CO reaction on Rh catalysts have been reviewed by Zhdanov and Kasemo [80]. For Pd-based catalysts, kinetic models are more rare, but have been also used to successfully reproduce some molecular beam experiments [82]. Much of this modeling has been based on a kinetic scheme similar to that shown in Figure 3.10, where the initial step is assumed to be the adsorption of NO [23]. That can occur either on the support or on the active particles, so surface diffusion processes between the two regions are taken into account by defining capture zones, sometimes by introducing a scaling factor for the NO flux. Both NO dissociation and nitrogen recombination steps show strong coverage dependences often included by defining coverage-dependent activation energies, and are also typically considered structure sensitive. Supported particle systems are accounted for by allowing for the presence of several reaction sites, such as different crystallographic facets, steps, edges, or metal-support interfacial sites. The effect of adsorbed atomic oxygen produced from NO dissociation must be also considered, in particular its removal by CO and its diffusion into the bulk. Many of these factors are not completely understood, and must therefore be simulated by combining the use of known experimental observations and the use of adjustable parameters. Perhaps the most extensive computational study of the kinetics of NO reactions on Rh and Pd surfaces has been provided by the group of Zgrablich. Their initial simulations of the NO + CO reaction on Rh(111) corroborated the fact that the formation of N−NO intermediate is necessary for molecular nitrogen production [83]. They also concluded that an Eley-Rideal mechanism is necessary to sustain a steady-state catalytic regime. Further simulations based on a lattice-gas model tested the role of the formation of

88

Past and Present in DeNOx Catalysis NxOy(a) Support reactions

–N2O (a)

(NO)2(a)

* Strongly structuredependent step

NO (a, support ) NO (g)

Capture zone (fast)

(surface) oxide formation, subsurface/bulk diff.

NO (a) Low T (*)

O (bulk )

*

+ NO (a)

* O (a) + N (a)

*

+ N (a)

N2O (g)

N2 (g)

Capture zone

+ CO (a), –N (a) + CO (a)

CO (a, Support )

N2O (g) CO2 (g)

CO (g)

Figure 3.10. Schematic representation of the elementary steps used in microkinetic simulations of the reduction of NO on supported metal particles [23]. The mechanism represented here incorporates adsorption and desorption steps, surface reactions such as NO dimerization and dissociation and N2 , N2 O and CO2 formation, surface oxidation, and mobility of adsorbates. (Figure provided by Professor Libuda and reproduced with permission from Elsevier, Copyright 2005).

atomic nitrogen islands in the N2 formation mechanism [84a]. There it was shown how the range of conditions under which the reaction can occur is influenced by the mechanism for the formation of the N−NO intermediate, by the growth of N surface islands, and by side steps such as NO surface diffusion and NO, and CO desorption. Further developments included the addition of an increase in NO dissociation probability with number of neighboring vacant sites, and the blocking of NO dissociation due to the presence of neighboring coadsorbed NO and CO species [85]. The anomalies in the kinetics of N2 O decomposition observed experimentally were reproduced by assuming that the surface sites required for the activation of adsorbed N2 O increase in size with increasing reaction temperature (Figure 3.11) [39]. Even the temperature dependence of the NO + CO reaction has been reproduced by adding appropriate activation energies into the model [85a]. The same group has looked into the conversion of NO on palladium particles. The authors in that case started with a simple model involving only one type of reactive site, and used as many experimental parameters as possible [86]. That proved sufficient to obtain qualitative agreement with the set of experiments on Pd/MgO discussed above [72], and with the conclusion that the rate-limiting step is NO decomposition at low temperatures and CO adsorption at high temperatures. Both the temperature and pressure dependences of the CO2 production rate and the major features of the transient signals were correctly reproduced. In a more detailed simulation that included the contribution of different facets to the kinetics on Pd particles of different sizes, it was shown that the effects of CO and NO desorption are fundamental to the overall behavior

Surface Science Studies of the Mechanism of NOx Conversion

89

1.0

Experimental Simulation

ΘO /ΘO(335 K)

0.8

0.6

Temp. = 407° K

0.4

0.2

0.0 300

400

500

600

700

800

Temperature/K Temp. = 639° K

O N2O

Figure 3.11. Example of the success of Monte Carlo simulations in reproducing and explaining experimental facts about the kinetics of NOx catalytic reactions, in this case in connection with the isothermal decomposition of N2 O on Rh(111) [39]. The left plot shows the excellent agreement obtained between experiments and simulations on the final oxygen coverages reached as a function of temperature. The lower saturation coverages seen at higher temperatures, one of the surprising observations from the molecular beam kinetic measurements, is explained here with the help of the final surface snapshots obtained in the Monte Carlo simulations (right). It is concluded that significantly higher oxygen coverages are obtained at lower temperatures because smaller surface ensembles are needed for the reaction under those conditions. (Figure provided by Professor Zgrablich).

of the system [82]. At low temperatures, where NO desorption can be neglected because its activation energy is about 5 kcal/mol larger than that of CO desorption, the largest particles were revealed as the less active. On the other hand, at high temperatures, where both NO and CO desorption take place, the smallest particles were found to be the less active. Finally, a group from General Motors has explored the mechanistic importance of the N2 O + CO reaction as an intermediate step during the reduction of NO by CO on noble metal exhaust catalysts [87,88]. Quasi-linearization of the non-linear NO + CO reaction system by identifying a critical kinetic parameter revealed that, indeed, the rate of the N2 O + CO conversion as an intermediate step in the overall NO + CO conversion can be two to three orders of magnitude faster than the isolated N2 O + CO reaction. This suggests that the observed suppression of N2 O production at higher temperatures may be due to its fast reaction with adsorbed CO once produced, and that, contrary to the accepted wisdom, the formation of N2 O and its subsequent reaction with CO can make a major contribution to the kinetics of the reduction of NO by CO in three-way catalytic converters. The validity of the theoretical results was verified by both

90

Past and Present in DeNOx Catalysis

steady-state and isotopic transient experiments using a number of supported catalysts. The newer molecular beam [35] and Monte Carlo simulations [39] reported previously provide further support for these conclusions.

8. SUMMARY AND CONCLUDING REMARKS This review has highlighted the key contributions of modern surface science to the understanding of the kinetics and mechanism of nitrogen oxide reduction catalysis. As discussed above, the conversion of NO has been taken as the standard to represent other NOx , and CO has typically been used as the reducing agent in these studies. The bulk of the work has been carried out on rhodium and palladium surfaces, the most common transition metals used in three-way catalytic converters. Based on the collective work in this area, a general consensus has been developed on the main features of the NO catalytic reduction process. A mechanism has been advanced with relatively simple steps, including the reversible adsorption of reactants, the dissociation of nitrogen oxides on the surface, and the recombination of adsorbed species to produce the desirable products. In spite of the deceiving simplicity of this mechanism, however, a number of subtleties have also been identified by the surface-science studies reviewed in this chapter. In particular, since the kinetics of the overall NO reduction often displays non-Arrhenius behavior, it has been suggested that the rate-limiting step in that mechanism may change, from the dissociation of NO or the formation of N2 at low temperatures to the adsorption of CO (or any other reducing agent) at higher temperatures. Moreover, a NO:CO ∼1 stoichiometry of adsorbed species is assumed to explain the synergy seen experimentally between temperature and reaction mixture composition in connection with the optimum reaction rates. A third issue is that, although it has generally been assumed that nitrogen production occurs by recombination of two adsorbed nitrogen atoms, both molecular beam experiments and theoretical calculations suggest that it may occur via an initial association of NO with one adsorbed nitrogen atom to form a N−NO intermediate instead. This species may then decompose into N2 and atomic oxygen, or just simply desorb molecularly. Such a proposal can certainly explain why N2 O production is seen in some catalytic processes but not under vacuum. Other kinetic and dynamic complications observed in these studies include a strong coverage dependence of the reaction rates, presumably because of site blocking and/or interactions among neighboring adsorbates, the requirement of the formation of atomic nitrogen islands on the surface for the N2 production to proceed, and the fairly complex dynamics of both NO + CO and N2 O conversions manifested by the angular patterns displayed by the desorbing products. A few additional points have also been raised by specific surface-science work concerning the catalytic reduction of NO. For instance, it has been widely recognized that the reaction is sensitive to the structure of the catalytic surface. It was determined that rough surfaces such as (110), or even (100), planes enhance NO dissociation over flatter (111) surfaces, and also favor N2 desorption instead of N2 O production. On the other hand, NO dissociation leads to poisoning by the resulting atomic species, hence the faster reaction rates seen with medium-size vs. larger particles on model rhodium supported catalyst (the opposite appears to be true on palladium). Also, at least in the case of palladium, the formation of an isocyanate (−NCO) intermediate was identified

Surface Science Studies of the Mechanism of NOx Conversion

91

at high pressures, although its role in the mechanism of the NO conversion has not been fully determined. In terms of the presence of an oxidizing environment during reaction, it has been concluded that the main effect of any added oxygen is to slow down the reaction, but not by competing for surface sites, since NO adsorption is favored over O2 activation, or by the consumption of the reducing agent, but rather via poisoning of the surface. Also, surface oxygen can lead to surface metal oxide formation, or even to its diffusion into the bulk. The effects that these processes exert on the rate of NO reduction are not completely understood. These and other observations from the mechanistic studies of NO reduction on model catalysts by using modern surface-sensitive techniques have been corroborated by more conventional work using supported catalysts. For instance, recent studies on two Rh/SiO2 catalysts with 31 and 127 Å crystallite sizes concluded that the reaction of NO with CO is selective toward the formation of N2 O [89]. The authors of that research also found that although the selectivity toward N2 O production was similar on both catalysts, the catalysts with smaller crystallite size display lower turnover frequencies. In a separate investigation, the group of Goodman et al. contrasted the kinetics of the NO + CO reaction over single crystal, planar-supported, and conventional high-surface area Pd/Al2 O3 catalysts, and managed to explain the main trends in the behavior of the supported catalysts using the basic knowledge obtained from the model systems [90]. Although more research is needed to clarify some of the controversies generated by conflicting results from different groups, it could be stated that a reasonably complete picture is available on the reactions of nitrogen oxides on Rh and Pd surfaces. Three-way catalysts are still the main way exhaust gases are processed in gasolinepowered automobiles [1,91,92]. In these, rhodium is still used as the main component to promote the reduction of nitrogen oxides, although palladium is also employed in some countries as a viable substitute. The original catalysts have been improved over the years by the addition of oxides such as lanthana, ceria, and/or zirconia to increase thermal stability and oxygen-storing capability. Nevertheless, these catalysts are expensive, and need to be replaced with cheaper alternatives. There are also problems associated with durability, sulfur poisoning, and cold engine starts, to mention only a few. Different applications such as diesel engines and natural gas turbines post additional individual demands, and require different catalyst formulations. As a result of all these concerns and derivations, new approaches have been advanced in recent years for the control of nitrogen oxide emissions, including the implementation of adsorption traps such as barium oxides to operate under lean conditions, in cyclic designs in conjunction with reduction catalysis such as platinum under rich incursions [93,94], and the use of transition metal-exchanged zeolites for selective catalytic reduction (SCR) with or without fuel additives such as ammonia or urea [93,95]. The surface-science community has not yet properly addressed the fundamental chemistry behind these new technologies. Perhaps a more concerted encouragement from funding agencies and from industry may be required for this to happen.

ACKNOWLEDGMENTS Financial assistance for the preparation of this chapter has been provided in part by the US National Science Foundation and the US Department of Energy.

92

Past and Present in DeNOx Catalysis

REFERENCES [1] Belton, D. N. and Taylor, K. C. (1999) “Automobile exhaust emission control by catalysts”, Curr. Opin. Solid State. Mater. Sci., 4, 97. [2] Farrauto, R. J. and Heck, R. M. (2000) “Environmental catalysis into the 21st century”, Catal. Today, 55, 179. [3] Shelef, M. and Graham, G. W. (1994) “Why rhodium in automotive three-way catalysts?” Catal. Rev.-Sci. Eng., 36, 433. [4] Menzel, D. (1975) “Desorption phenomena” In: Interactions on Metal Surfaces (R. Gomer, ed.), Springer-Verlag, New York, pp. 102–43. [4a] Wilson, J., Guo, H., Morales, R. et al. (2007) “Kinetic measurements of hydrocarbon conversion reactions on model metal surfaces”, Phys. Chem. Chem. Phys., 9, 3830. [5] Habenschaden, E. and Küppers, J. (1984) “Evaluation of flash desorption spectra”, Surf. Sci., 138, L147. [6] Borg, H. J., Reijerse, J. F. C.-J. M., van Santen, R. A. et al. (1994) “The dissociation kinetics of NO on Rh(111) as studied by temperature programmed static secondary ion mass spectrometry and desorption”, J. Chem. Phys., 101, 10052. [7] Siokou, A., van Hardeveld, R. M. and Niemantsverdriet, J. W. (1998) “Surface reactions of nitrogen oxide on Rhodium(100), adsorption, dissociation and desorption”, Surf. Sci., 404, 110. [8] Janssen, N. M. H., Cholach, A. R., Ikai, M. et al. (1997) “The interaction of NO with stepped Rh surfaces”, Surf. Sci., 382, 201. [9] Ramsier, R. D., Gao, Q., Waltenburg, H. N. et al. (1994) “NO adsorption and thermal behavior on Pd surfaces. A detailed comparative study”, Surf. Sci., 320, 209. [10] Sugisawa, T., Shiraishi, J., Machihara, D. et al. (2001) “Adsorption and decomposition of NO on Pt (112)”, Appl. Surf. Sci., 169–70, 292. [11] Backus, E. H. G., Eichler, A., Grecea, M. L. et al. (2004) “Adsorption and dissociation of NO on stepped Pt (533)”, J. Chem. Phys., 121, 7946. [12] Panja, C. and Koel, B. E. (2000) “Influence of alloyed Sn on adsorption and reaction of NO on Pt(100) surfaces”, J. Phys. Chem. A, 104, 2486. [13] Khatua, S., Held, G. and King, D. A. (2005) “Model car-exhaust catalyst studied by TPD and TP-RAIRS: Surface reactions of NO on clean and O-covered Ir{100}”, Surf. Sci., 586, 1. [14] Matveev, A. V., Sametova, A. A. and Gorodetskii, V. V. (2004) “The study of nitric oxide adsorption and the mechanism of surface ‘explosions’ in the reaction of CO + NO on Pt(100) and Pd(110) single crystal surfaces”, Kinet. Catal., 45, 598. [15] Hopstaken, M. J. P., van Gennip, W. J. H. and Niemantsverdriet, J. W. (1999) “Reactions between NO and CO on rhodium(111): an elementary step approach”, Surf. Sci., 435, 69. [16] Hopstaken, M. J. P. and Niemantsverdriet, J. W. (2000) “Lateral interactions in the dissociation kinetics of NO on Rh(100)”, J. Phys. Chem. B, 104, 3058. [17] Mukerji, R. J., Bolina, A. S. and Brown, W. A. (2003) “An investigation of the effect of pre-dosed O atoms on the adsorption of NO on Pt{211}”, Surf. Sci., 547, 27. [18] Zhdanov, V. P. and Matsushima, T. (2005) “Simulation of thermal desorption spectra of N2 observed during N2 O decomposition on Rh(110)”, Surf. Sci., 583, 253. [19] Wickham, D. T., Banse, B. A. and Koel, B. E. (1991) “Adsorption of nitrogen dioxide and nitric oxide on Pd(111)”, Surf. Sci., 243, 83. [20] Matsushima, T., Imamura, K., Horino, H. et al. (2005) “Inclined N2 desorption in N2 O decomposition on Rh(110)”, Appl. Surf. Sci., 244, 141. [21] Liu, S., Horino, H., Kokalj, A. et al. (2004) “N2 Desorption in the Decomposition of Adsorbed N2 O on Rh(110)”, J. Phys. Chem. B, 108, 3828.

Surface Science Studies of the Mechanism of NOx Conversion

93

[22] Zaera, F. (2002) “Infrared and molecular beam studies of chemical reactions on solid surfaces”, Int. Rev. Phys. Chem., 21, 433. [23] Libuda, J. and Freund, H. J. (2005) “Molecular beam experiments on model catalysts”, Surf. Sci. Rep., 57, 157. [24] Aryafar, M. and Zaera, F. (1998) “Isothermal kinetic study of the decomposition of nitric oxide over Rh(111) surfaces”, J. Catal., 175, 316. [25] Li, Y. and Bowker, M. (1996) “The adsorption and decomposition of nitrous oxide on Rh(110) and Rh(111)”, Surf. Sci., 348, 67. [26] Thirunavukkarasu, K., Thirumoorthy, K., Libuda, J. et al. (2005) “Isothermal kinetic study of nitric oxide adsorption and decomposition on Pd(111) surfaces: Molecular beam experiments”, J. Phys. Chem. B, 109, 13283. [27] Sharpe, R. G. and Bowker, M. (1996) “The adsorption and decomposition of NO on Pd(110)”, Surf. Sci., 360, 21. [28] Johánek, V., Schauermann, S., Laurin, M. et al. (2004) “On the role of different adsorption and reaction sites on supported nanoparticles during a catalytic reaction: NO decomposition on a Pd/alumina model catalyst”, J. Phys. Chem. B, 108, 14244. [29] Zaera, F. and Gopinath, C. S. (2001) “Surface intermediates during the catalytic reduction of NO on rhodium catalysts: A kinetic inference”, J. Mol. Catal. A, 167, 23. [30] Bowker, M., Guo, Q., Li, Y. et al. (1995) “CO–NO reaction on Rh(110) studied using molecular beams”, J. Chem. Soc., Faraday Trans., 91, 3663. [31] Thirunavukkarasu, K., Thirumoorthy, K., Libuda, J. et al. (2005) “A molecular beam study of the NO + CO reaction on Pd(111) surfaces”, J. Phys. Chem. B, 109, 13272. [32] Gopinath, C. S. and Zaera, F. (1999) “A molecular beam study of the kinetics of the catalytic reduction of NO by CO on Rh(111) single crystal surfaces”, J. Catal., 186, 387. [33] Gopinath, C. S. and Zaera, F. (2000) “Transient kinetics during the isothermal reduction of NO by CO on Rh(111) as studied with effusive collimated molecular beams”, J. Phys. Chem. B, 104, 3194. [34] Zaera, F., Wehner, S., Gopinath, C. S. et al. (2001) “Evidence for the formation of nitrogen islands on rhodium surfaces”, J. Phys. Chem. B, 105, 7771. [34a] Zaera, F. (2002) “Kinetics of chemical reactions on solid surfaces: Deviations from conventional theory”, Acc. Chem. Res., 35, 129. [35] Zaera, F. and Gopinath, C. S. (2000) “Evidence for a N2 O intermediate in the catalytic reduction of NO to N2 on rhodium surfaces”, Chem. Phys. Lett., 332, 209. [36] Zaera, F. and Gopinath, C. S. (2003) “On the mechanism for the reduction of nitrogen monoxide on Rh(111) single-crystal surfaces”, Phys. Chem. Chem. Phys., 5, 646. [37] Zaera, F. and Gopinath, C. S. (2002) “Effect of coverage and temperature on the kinetics of nitrogen desorption from Rh(111) surfaces”, J. Chem. Phys., 116, 1128. [38] Wehner, S., Paffett, M. T. and Zaera, F. (2004) “Molecular beam studies of the kinetics of N2 O thermal conversion on Rh(111) single-crystal surfaces”, J. Phys. Chem. B, 108, 18683. [39] Uñac, R. O., Bustos, V., Wilson, J. et al. (2006) “Molecular beam measurements and Monte Carlo simulations of the kinetics of N2 O decomposition on Rh(111) single-crystal surfaces”, J. Chem. Phys., 125, 074705-1. [40] Rze´znicka, I. I., Ma, Y., Cao, G. et al. (2004) “Removal pathways of surface nitrogen in a steady-state NO + CO reaction on Pd(110) and Rh(110): Angular and velocity distribution studies”, J. Phys. Chem. B, 108, 14232. [41] Brandt, M., Mueller, H., Zagatta, G. et al. (1995) “Sticking and scattering of gas-phase oriented NO at Pt(100) as a function of the translational energy and of molecular orientation”, Surf. Sci., 331–3, 30. [42] Brandt, M., Zagatta, G., Böwering, N. et al. (1997) “Reaction of NO and CO on a Pt(100) surface as a function of the NO translational energy and of molecular orientation”, Surf. Sci., 385, 346.

94

Past and Present in DeNOx Catalysis

[43] Brandt, M., Müller, H., Zagatta, G. et al. (1996) “Reaction of NO and CO on a Rh(100) surface studied with gas-phase oriented NO”, Surf. Sci., 352–4, 290. [44] Gopinath, C. S. and Zaera, F. (2001) “NO + CO + O2 reaction kinetics on Rh(111): A molecular beam study”, J. Catal., 200, 270. [45] Comelli, G., Dhanak, V. R., Pangher, N. et al. (1994) “NO dissociation on Rh(110)”, Surf. Sci., 317, 117. [46] Lizzit, S., Baraldi, A., Cocco, D. et al. (1998) “NO adsorption on the Rh(110) surface: Kinetics and composition of the adlayer studied by fast XPS”, Surf. Sci., 410, 228. [47] Esch, F., Baraldi, A., Comelli, C. et al. (1999 ) “Atomic nitrogen on steps: A fast x-ray photoelectron spectroscopy study of the NO uptake on Rh(533), Rh(311), and Rh(111)”, J. Chem. Phys., 110, 4013. [48] Cobden, P. D., Nieuwenhuys, B. E., Esch, F. et al. (1998) “A fast X-ray photoelectron spectroscopy study of the NO-H2 reaction over Rh(533): Understanding hysteresis and oscillations in the reaction rate”, Surf. Sci., 416, 264. [49] Nakao, K., Ito, S. I., Tomishige, K. et al. (2005) “Infrared chemiluminescence study of CO2 formation in CO + NO reaction on Pd(110) and Pd(111) surfaces”, J. Phys. Chem. B, 109, 17579. [50] Rider, K. B., Hwang, K. S., Salmeron, M. et al. (2001) “Structure and dynamics of dense monolayers of NO adsorbed on Rh(111) in equilibrium with the gas phase in the Torr pressure range”, Phys. Rev. Lett., 86, 4330. [51] Rider, K. B., Hwang, K. S., Salmeron, M. et al. (2002) “High-pressure (1 Torr) scanning tunneling microscopy (STM) study of the coadsorption and exchange of CO and NO on the Rh(111) crystal face”, J. Am. Chem. Soc., 124, 5588. [52] Requejo, F. G., Hebenstreit, E. L. D., Ogletree, D. F.et al. (2004) “An in situ XPS study of site competition between CO and NO on Rh(111) in equilibrium with the gas phase”, J. Catal., 226, 83. [53] Blakely, D. W., Kozak, E. I., Sexton, B. A. et al. (1976) “New instrumentation and techniques to monitor chemical surface reactions on single crystals over a wide pressure range (10–8 –105 Torr) in the same apparatus”, J. Vac. Sci. Technol., 13, 1091. [53a] Zaera, F. (2001) “Probing catalytic reactions at surfaces”, Prog. Surf. Sci., 69, 1. [54] Belton, D. N. and Schmieg, S. J. (1993) “Oxidation of CO by NO over Rh(111)”, J. Catal., 144, 9. [55] Permana, H., Ng, K. Y. S., Peden, C. H. F. et al. (1995) “Effect of NO pressure on the reaction of NO and CO over Rh(111)”, J. Phys. Chem., 99, 16344. [56] Permana, H., Ng, K. Y. S., Peden, C. H. F. et al. (1996) “Adsorbed species and reaction rates for NO-CO over Rh(111)”, J. Catal., 164, 194. [57] Tolia, A. A., Williams, C. T., Weaver, M. J. et al. (1995) “Surface-enhanced Raman spectroscopy as an in situ real-time probe of catalytic mechanisms at high gas pressures. The NO-H2 reaction on rhodium”, Langmuir, 11, 3438. [58] Tolia, A. A., Williams, C. T., Takoudis, C. G. et al. (1995) “Surface-enhanced Raman spectroscopy as an in-situ real-time probe of catalytic mechanisms at high gas pressures. The CO-NO reaction on rhodium”, J. Phys. Chem., 99, 4599. [59] Peden, C. H. F., Belton, D. N. and Schmieg, S. J. (1995) “Structure sensitive selectivity of the NO–CO reaction over Rh(110) and Rh(111)”, J. Catal., 155, 204. [60] Herman, G. S., Peden, C. H. F., Schmieg, S. J. et al. (1999) “A comparison of the NO–CO Reaction over Rh(100), Rh(110) and Rh(111)”, Catal. Lett., 62, 131. [61] Peden, C. H. F., Goodman, D. W., Blair, D. S. et al. (1988) “Kinetics of CO oxidation by O2 or NO on Rh(111) and Rh(100) single crystals”, J. Phys. Chem., 92, 1563. [62] Oh, S. H., Fisher, G. B., Carpenter, J. E. et al. (1986) “Comparative kinetic studies of CO–O2 and CO–NO reactions over single crystal and supported rhodium catalysts”, J. Catal., 100, 360.

Surface Science Studies of the Mechanism of NOx Conversion

95

[63] Permana, H., Ng, K. Y. S., Peden, C. H. F. et al. (1997) “Adsorbed species and reaction rates for NO–CO–O2 over Rh(111)”, Catal. Lett., 47, 5. [64] Ng, K. Y. S., Belton, D. N., Schmieg, S. J. et al. (1994) “NO–CO activity and selectivity over a Pt10 Rh90 (111) alloy catalyst in the 10-torr pressure range”, J. Catal., 146, 394. [65] Ozensoy, E., Hess, C. and Goodman, D. W. (2004) “Understanding the catalytic conversion of automobile exhaust emissions using model catalysts: CO + NO reaction on Pd(111)”, Top. Catal., 28, 13. [66] Hess, C., Ozensoy, E. and Goodman, D. W. (2003) “Combined in situ infrared and kinetic study of the catalytic CO + NO reaction on Pd(111) at pressures up to 240 mbar”, J. Phys. Chem. B, 107, 2759. [67] Ozensoy, E., Hess, C. and Goodman, D. W. (2002) “Isocyanate formation in the catalytic reaction of CO + NO on Pd(111): An in situ infrared spectroscopic study at elevated pressures”, J. Am. Chem. Soc., 124, 8524. [68] Valden, M., Aaltonen, J., Kuusisto, E. et al. (1994) “Molecular beam studies of CO oxidation and CO-NO reactions on a supported Pd catalyst”, Surf. Sci., 307–9, 193. [69] Schauermann, S., Joh´anek, V., Laurin, M. et al. (2003) “Interaction of NO with alumina supported palladium model catalysts”, Phys. Chem. Chem. Phys., 5, 5139. [70] Piccolo, L. and Henry, C. R. (2000) “Reactivity of metal nanoclusters: Nitric oxide adsorption and CO + NO reaction on Pd/MgO model catalysts”, Appl. Surf. Sci., 162, 670. [71] Piccolo, L. and Henry, C. R. (2001) “NO–CO reaction kinetics on Pd/MgO model catalysts: Morphology and support effects”, J. Mol. Catal. A, 167, 181. [72] Prévot, G., Meerson, O., Piccolo, L. et al. (2002) “Reactivity of supported metal clusters: The reduction of NO by CO on a Pd/MgO(100) model catalyst”, J. Phys.: Condens. Matter, 14, 4251. [73] Wörz, A. S., Judai, K., Abbet, S. et al. (2003) “Cluster size-dependent mechanisms of the CO + NO reaction on small Pdn (n = 30) clusters on oxide surfaces”, J. Am. Chem. Soc., 125, 7964. [74] Judai, K., Abbet, S., Wörz, A. S. et al. (2004) “Low-temperature cluster catalysis”, J. Am. Chem. Soc., 126, 2732. [75] Loffreda, D., Simon, D. and Sautet, P. (2003) “Structure sensitivity for NO dissociation on palladium and rhodium surfaces”, J. Catal., 213, 211. [76] Inderwildi, O. R., Lebiedz, D., Deutschmann, O. et al. (2005) “Influence of coadsorbates on the NO dissociation on a rhodium(311) surface”, Chem. Phys. Chem., 6, 2513. [77] Kokalj, A. and Matsushima, T. (2005) “A density-functional theory study of the interaction of N2 O with Rh(110)”, J. Chem. Phys., 122, 1. [78] Ricart, J. M., Ample, F., Clotet, A. et al. (2005) “Structure and catalytic processes of N-containing species on Rh(111) from first principles”, J. Catal., 232, 179. [79] Hermse, C. G. M., Frechard, F., Van Bavel, A. P. et al. (2003) “Combining densityfunctional calculations with kinetic models: NO/Rh(111)”, J. Chem. Phys., 118, 7081. [80] Zhdanov, V. P. and Kasemo, B. (1997) “Mechanism and kinetics of the NO–CO reaction on Rh”, Surf. Sci. Rep., 29, 31. [81] Van Bavel, A. P., Hermse, C. G. M., Hopstaken, M. J. P. et al. (2004) “Quantifying lateral adsorbate interactions by kinetic Monte-Carlo simulations and density-functional theory: NO dissociation on Rh(100)”, Phys. Chem. Chem. Phys., 6, 1830. [82] Bustos, V., Uñac, R. O., Zgrablich, G. et al. (2003) “Lattice-gas study of the kinetics of the NO-CO catalytic reaction on Pd nanoclusters”, Phys. Chem. Chem. Phys., 5, 2906. [83] Bustos, V., Gopinath, C. S., Uñac, R. et al. (2001) “Lattice-gas study of the kinetics of catalytic conversion of NO–CO mixtures on rhodium surfaces”, J. Chem. Phys., 114, 10927. [84] Bustos, V., Uñac, R., Zaera, F. et al. (2003) “Lattice-gas study of the kinetics of the catalytic NO-CO reaction on rhodium surfaces. II. The effect of nitrogen surface islands”, J. Chem. Phys., 118, 9372.

96

Past and Present in DeNOx Catalysis

[84a] Zaera, F., Sales, J. L., Gargiulo, M. V. et al. (2007) “On the formation of nitrogen islands on rhodium surfaces”, J. Phys. Chem. C, 111, 7795. [85] Avalos, L. A., Bustos, V., Uñac, R. et al. (2005) “Toward a realistic model for the kinetics of the NO + CO reaction on rhodium surfaces”, J. Mol. Catal. A, 228, 89. [85a] Avalos, L. A., Bustos, V., Uñac, R. et al. (2006) “Dynamic Monte Carlo simulation of the NO + CO reaction on Rh(111)”, J. Phys. Chem. B, 110, 24964. [86] Prévot, G. and Henry, C. R. (2002) “Microkinetic modeling of the CO + NO reaction on Pd/MgO particles”, J. Phys. Chem. B, 106, 12191. [87] Cho, B. K. (1992) “Mechanistic importance of intermediate N2 O + CO reaction in overall NO + CO reaction system: I. Kinetic analysis", J. Catal., 138, 255. [88] Cho, B. K. (1994) “Mechanistic importance of intermediate N2 O + CO reaction in overall NO + CO reaction system: II. Further analysis and experimental observations”, J. Catal., 148, 697. [89] Araya, P., Gracia, F., Cortes, J. et al. (2002) “FTIR study of the reduction reaction of NO by CO over Rh/SiO2 catalysts with different crystallite size”, Appl. Catal. B, 38, 77. [90] Rainer, D. R., Vesecky, S. M., Koranne, M. et al. (1997) “The CO + NO reaction over Pd: A combined study using single-crystal, planar-model-supported, and high-surface-area Pd/Al2 O3 catalysts”, J. Catal., 167, 234. [91] Shelef, M. and McCabe, R. W. (2000) “Twenty-five years after introduction of automotive catalysts: what next?” Catal. Today, 62, 35. [92] Heck, R. M. and Farrauto, R. J. (2001) “Automobile exhaust catalysts”, Appl. Catal. A, 221, 443. [93] Burch, R. (2004) “Knowledge and know-how in emission control for mobile applications”, Catal. Rev. Sci. Eng., 46, 271. [94] Epling, W. S., Campbell, L. E., Yezerets, A. et al. (2004) “Overview of the fundamental reactions and degradation mechanisms of NOx storage/reduction catalysts”, Catal. Rev. Sci. Eng., 46, 163. [95] Amiridis, M. D., Zhang, T. J. and Farrauto, R. J. (1996) “Selective catalytic reduction of nitric oxide by hydrocarbons”, Appl. Catal. B, 10, 203.

Chapter 4

GENERAL FEATURES OF IN SITU AND OPERANDO SPECTROSCOPIC INVESTIGATION IN THE PARTICULAR CASE OF DeNOx REACTIONS P. Bazin, O. Marie and M. Daturi∗ Laboratoire Catalyse et Spectrochimie, UMR 6506 – CNRS/ENSICAEN/ Université de Caen,6, boulevard Maréchal Juin, 14050 Caen cedex, France ∗

Corresponding author: Laboratoire Catalyse et Spectrochimie, UMR 6506 – CNRS/ENSICAEN/ Université de Caen, 6, boulevard Maréchal Juin, 14050 Caen cedex, France. Tel.: +33 2 31 45 27 30, Fax.: +33 2 31 45 28 22, E-mail: [email protected]

Abstract As combinatorial chemistry has not been able to find the optimal formulation for NOx removal till now, the recent trend to improve the performance of catalysts is to better understand the reaction mechanisms looking at the molecular level the closest from the reaction place. In order to illuminate the active sites, the intermediate and/or spectator species, vibrational spectroscopies represent a unique tool as they involve rather low-energy radiations which very little affect the observed species. This is one of the reason why, among the other techniques, infrared and Raman carried out in operando conditions have been successful for the last five years. Several examples will be given about the progress in the comprehension of DeNOx mechanisms obtained via spectroscopic studies. The adsorption of specific probe molecules can, for example, show the presence and the characteristics of surface sites, as well as the modifications undertaken by the species interacting with them. The choice of the adsorption conditions is critical for a reliable interpretation of the investigated process. Actually, the operando approach is more and more required to get closer to the reaction phenomena when they are taking place. This implies that a continuous development of the in situ spectroscopic techniques (especially, infrared and Raman spectroscopies) is often coupled to have complementary information in a limited lapse of time. Time resolution remains a crucial point to obtain a realistic following of the chemical reaction and to evidence intermediate species. The actual limits in intermediate species detection and some methods to differentiate them from spectator species in DeNOx reactions will be highlighted.

1. INTRODUCTION In the last 15 years, an intense effort has been made to decrease the environmental impact of automotive exhaust gases. More and more severe regulations have pushed Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

98

Past and Present in DeNOx Catalysis

researchers to investigate novel ways for emission control, concerning the reduction of NOx species from industrial sources and especially from automotive engines working in lean conditions, which remains the biggest challenge. Any kind of process [selective catalytic reduction (SCR) by ammonia or hydrocarbons, thermal- or plasma-assisted decomposition, storage and reduction cycling, etc.] and materials (zeolitic compounds, oxides, noble metals, etc.) have been applied to such a problem without finding the optimum solution. Combinatorial chemistry has not been able to find out a formulation presenting the required performances for NOx removal. Therefore, it has been necessary to go back to a deep investigation of the different phenomena taking place during NOx reactions, to give a detailed description of the involved mechanisms, in order to design a proper catalyst for a chosen process. Spectroscopic techniques, carried out in in situ and operando conditions, obviously represent powerful tools for the description of the reactions and the catalysts in running conditions. In fact, the exigency of the scientist to look at the chemical process at a molecular level cannot only address the traditional kinetics modelling, where the reactor itself behaves as a black box. The use of spectroscopy allows monitoring the catalytic material under duty, directly revealing species and transformations, which can then support the hypothesis made for mathematical calculations applied to a kinetic model [1]. In this contribution, several examples will be given about the progress in the comprehension of DeNOx mechanisms obtained via spectroscopic studies. In general, several spectroscopic techniques have been applied to the study of NOx removal. X-ray photoelectron spectroscopy (XPS), electron paramagnetic resonance (EPR), nuclear magnetic resonance (NMR), extended X-ray absorption fine structure (EXAFS) and X-ray absorption near-edge structure (XANES) are currently used to determine the surface composition of the catalysts, with the aim to identify the cationic active sites, as well as their coordinative environment. XPS is mainly applied in extra situ conditions (before and after the reaction experiments) for these purposes [2–8]. Recently, few research teams have been able to conceive experimental apparatus allowing to work in quasi in situ XPS [9]. For example, Granger et al. have shown that using a catalytic chamber coupled to the XPS apparatus and transferring the sample directly in the analysis chamber under ultrahigh vacuum for XPS analysis permit to determine the oxidation state of an element in different chemical environments. They have applied this methodology to obtain correlations between catalytic and spectroscopic features during the reduction of NO by H2 in the presence of an excess of oxygen on prereduced Pd/LaCoO3 . According to this thermal treatment, they found an extensive reduction leading to the formation of Pd0 and Co0 particles in interaction with La2 O3 . Thanks to XPS measurements, they highlighted surface modifications explaining the selective NO reduction at high temperatures [10]. Again, Granger et al., coupling in situ XPS and IR, evidenced the influence of Ce additive on the catalytic performances of three-way bimetallic Pt−Rh/Al2 O3 in the CO + NO reaction and the effect of a hydrogen prereduction treatment on the Ce and metal oxidation state [11]. Some teams have even experienced in real in situ conditions, using a newly-developed high-pressure photoelectron spectrometer (HPPES), capable of operation under several hundreds Pa [12]. These studies carried out on a Rh(111) surface in CO and NO gas mixtures showed that NO efficiently displaces CO from hollow sites at room temperature. Solid state NMR experiments are currently performed in standard conditions (room temperature and atmosphere, or inert gas) as well [13,14]. More recently, some scientists

General Features of In Situ and Operando Spectroscopic Investigation

99

have realized in situ NMR probes for the study of catalytic reactions [15]. In particular, a continuous flow magic angle spinning NMR probe head has been developed, which simultaneously enables the observation of catalytic events occurring on catalyst surfaces and the online identification of the gas products in the output using mass spectrometry [16]. EXAFS and XANES, on the contrary, are more often carried out in chambers with a controlled atmosphere or even a reaction flow, thus giving results closer to the working conditions of the catalysts. For example, Hungría et al. studying Pd−Ni [3] or Pd−Cr [17] catalysts supported on Al2 O3 , (Ce,Zr)Ox /Al2 O3 and (Ce,Zr)Ox , examined with regard to their catalytic activity for CO oxidation and NO reduction under stoichiometric CO−O2 and CO−O2 −NO mixtures, showed that the effects induced by the presence of nickel on the catalytic activity strongly depend on the support used [3]. In the case of the Pd−Cr systems, they pointed out that the Pd−ceria interface appears significantly more active than Pd−ceria/zirconia in both pollutant elimination processes under stationary conditions, having a dominant impact on the improved light-off performance of ceria/alumina-supported systems. In the case of monometallic Pd, the larger Pd particle size of Pd/ceria/alumina may also exert a positive influence on the NO reduction process. For bimetallic Pd−Cr systems, the base metal may have an influence over Pd with important catalytic implications. The base metal strongly influenced the chemical state of zero-valent Pd for PdCr/ceria/alumina with formation of a Pd−Cr alloy in the ceria/alumina-supported sample, which seems to dominate the NO reduction behaviour at temperatures above 473 K. Such effects were not significant in the ceria−zirconia/alumina-supported system indicating a lower degree of interaction between both metals in the latter case [17]. Martinez-Arias et al., combining in situ XANES with EPR, Raman and diffuse-reflectance infrared fourier transform (DRIFT) on Pd/ceria−zirconia/alumina samples, confirmed that NO reduction is a structure-sensitive reaction, which should be dependent on noble metal size/morphology properties, particularly influencing NO dissociation and N recombination steps. In particular, they pointed out that Pd particle size and, to a limited extent, Ce concentration at the promoter surface appear as the key physico-chemical factors affected by thermal ageing and influencing catalytic activity for NO elimination [18,19]. Also, Lecomte et al., coupling EPR and catalytic tests, reported on the role of anionic O− 2 species in the CO + NO reaction obtained on a Pt−Rh/Al2 O3 −CeO2 catalyst. They revealed that ceria incorporation to noble metals improves the activity: noble metals in interaction with CeO2 could enhance the formation of O2 − from O− of CeO2 . The promotional effect of ceria disappears at low temperature due to deactivation via site blocking or surface reconstruction [20]. Brückner and coworkers have shown that combined EPR and UV–vis-DRS (Diffuse Reflectance Spectroscopy) in situ measurements are powerful tools for elucidating the nature of coexisting Fe species formed in the Fe-ZSM-5 materials investigated. The new structural data have been discussed with regard to the catalytic properties of the materials in the SCR of NO by isobutane or NH3 [21]. They also evidenced the participation of mononuclear iron ions in the N2 O−CO reaction and also supported the involvement of oligonuclear Fex 3+ Oy species. They show that the interaction of N2 O and CO and the reaction mechanism are iron site dependent: over isolated sites, the reduction of N2 O with CO occurs via coordinated CO species on Fe3+ ions, not involving change of oxidation site. The reaction over oligonuclear iron clusters proceeds via a redox Fe3+ /Fe2+ process via intermediate formation of O− radicals [22]. This ensemble of results collected combining in situ EPR,

100

Past and Present in DeNOx Catalysis

UV–vis and DRS measurements on Fe-ZSM-5 catalysts with variable amounts of isolated, oligomeric and heavily aggregated Fe3+ oxo sites, led to the development of a unified concept of the active Fe sites in these reactions, according to which isolated Fe sites catalyse both SCR reactions; while oligomeric sites, though also involved in the selective reduction path, limit the catalyst performance by causing the total oxidation of the reductant [23]. Raman is a suitable structure analysis technique alternative to X-ray diffraction (XRD) and particularly adapted for highly dispersed materials or to evidence the formation of thin layers at the surface of a solid [4,24]. Raman spectroscopy, being insensitive to water presence, is also adapted to in situ conditions. Mamede et al., using in situ Raman spectroscopy, in the temperature range 25–300 C, characterized the surface palladium modifications under various controlled atmospheres composed of NO and/or CO. They identified Pd−N vibrational mode of species resulting from the dissociation of NO, which is supposed to occur more readily on the supported Pd catalyst characterized by smaller particles, this reaction being structure sensitive. Additionally, they observed the development of PdO after extensive NO dissociation and its role in the rate limiting step of the CO + NO reactions at high temperature [25]. Bañares and coworkers underlined that Raman operando methodology combines both structural and catalytic measurements in a single experiment. Since the actual molecular structure of a catalyst depends on the specific environmental conditions, such combination is critical to assess structure– activity relationships at a molecular level reliably. Therefore, the characterization must provide information relevant to the catalytic process, i.e. about the phases directly interacting with the reactants. With this aim, the design of the Raman cell and the reaction conditions are critical to prevent any mass or heat transfer limitation or the contribution of non-catalytic reactions, e.g. gas phase reaction [26]. The applications of IR spectroscopy to the study of catalytic reactions are particularly broad and diffused among the catalysis community. Many specific examples of IR utility will be given in the paragraphs below. This fast overview shows that the aim of this methodology, combining catalyst analysis underflow and ‘classical’ online gas phase catalytic activity measurements, is to observe the catalyst under real working conditions. Operando spectroscopy is a methodology that combines the spectroscopic characterization of a catalytic material during reaction with the simultaneous measurement of catalytic activity/selectivity. The potential impact of operando spectroscopy on catalysis science lies in its ability to significantly assist in the establishment of fundamental molecular structure–activity/selectivity relationships for catalytic systems [27]. However, one cannot improvise oneself as an operando expert and deduce a reaction mechanism from a single operando experiment. It would be the same dream than hoping to beat a chess master without even knowing the chessboard and the way the pieces are disposed on it. To go on with the comparison, we can associate the chessboard with the catalyst surface (see Figure 4.1) and an inevitable step to develop a gaining strategy consists in the characterization of the ‘playground’. Of course, in reality, the situation is much more complicate: in the chessboard, the places on the board are fixed, the pieces are known and their movements are also known. It is for the players to use them with his own intellects and skills. The catalysts’ surfaces keep on changing at molecular level during the reaction; the intermediate species keep on changing as well. Nevertheless, this simplified image of the catalytic surface as a chessboard could represent, in our opinion, a little help for the reader. Depending

General Features of In Situ and Operando Spectroscopic Investigation

NO, NO2 CxHy CO, H2…

101 N2 CO2 H2O…

IR or Raman beam

O O

N

O

O

C Pt

Figure 4.1. Imaginary representation of a catalytic surface as a chess playground.

on the specific DeNOx application, the required catalyst properties are different and a very wide range of chemical functions will be present: acidic and basic sites, redox properties and oxygen mobility, dispersion of supported metal   In order to obtain the best information about each specific chemical function, direct analysis of the catalyst surface or appropriate probe molecule adsorption (the reactant itself whenever possible) followed by infrared or Raman have been developed for many decades. In the first part of our contribution, we will thus try to summarize the main data obtained with these vibrational techniques and yielding to a proper characterization of the catalyst surface and the way molecules can adsorb on it. The second part of this work will be dedicated to the start of the game: what are the pieces motions? How can the adsorbed molecules react on the surface and among all the playground, where does the real action take place? This is the so-called in situ approach for which techniques such as temperature-programmed surface reaction (TPSR) or transient analysis by pulse injection have been developed. The last part will consist in the description of a chess game in the real conditions of a tournament with time constraints, influences of the spectators in the audience and possible ‘poisoning’ of the actors. This is the so-called operando approach for which many technical improvements have been made to obtain the maximum of information. Detailed examples will be given concerning the developments allowing to hunt intermediates with very short lifetimes.

2. THE CHESSBOARD AND THE PIECES: CATALYST AND ADSORBED SPECIES CHARACTERIZATION 2.1. Catalyst acidity From a general point of view, in heterogeneous catalysis, acidity has a very important influence on both activity and selectivity. More specifically, in the context of NOx

102

Past and Present in DeNOx Catalysis

abatement, acidic materials such as V-Mo-W-Ti mixed oxides or zeolites (with or without transition metal loading) find applications in purifying stationary sources emissions. The last ones are even studied for potential applications in automotive exhaust gas control. A correct determination of acidic properties is thus a base to improve industrial processes. We will first describe some basic principles useful for a correct study of solid acidity using vibrational spectroscopies and will then detail some recent examples in line with DeNOx applications.

2.1.1. Basic principles for investigations on acidity A recent review by Busca [28] describes the bases of IR spectroscopic methods for the characterization of the surface acidity, of the Lewis type and of the Brønsted type, of solid, simple and mixed oxides. A systematization is proposed associating the surface acidity with the ionicity/covalency of the element−oxygen bond, mainly affected by the size and charge of the cation. In a following work, the results obtained for the characterization of the Lewis acid strength of more than 30 binary and ternary mixed oxides are interpreted on the basis of the different polarizing powers of the involved cations [29]. Recent progress on acidity characterization is reviewed elsewhere and described to be related to the broadening of the spectral range (investigation of overtones, combination bands and low-frequency modes) and to the adsorption of new non-traditional probe molecules for identification of acid sites [30]. Concerning the choice of the appropriate probe molecule, Lercher et al. [31], give special emphasis to the criteria that have to be met to arrive at a characterization of the solid that is useful for its catalytic application. Table 4.1 is presented as a guide to help selecting the right molecule for the right site. Among all the basic probe molecules described in the literature, pyridine C5 NH5 is probably most widely used to distinguish between Brønsted or Lewis acidity because the aromatic ring vibration modes so-called 8a and 19b are sensitive to the interaction. However, recent progress in the use of pyridine CH(D) vibrations for the study of Lewis acidity of metal oxides [32] makes us remind that, for complex molecules (for a non-linear molecule containing n atoms, the overall number of possible vibration modes is 3n – 6), the complete analysis of the whole spectrum is not straightforward and investigations are still required. More specifically, for the study of zeolites, techniques using adsorption of small and weakly basic molecules were developed. It has been revealed that N2 and CO are effective probe molecules to characterize both Brønsted acidity and Lewis acidity [33,34]. Wakabayashi et al. [33] also used O2 and rare gases to monitor the strong acid sites and discussed in detail the characteristics of the N2 probe in comparison with the CO probe. In addition, Hadjiivanov et al. [35] remind that when applied to materials with surface hydroxy groups, CO undergoes hydrogen bonding and information can also be collected on the proton acid strength.

2.1.2. Acidity of DeNOx catalysts: selected examples The SCR of NOx by NH3 is actually the best developed and widespread method for the NOx removal from stationary sources due to its efficiency and selectivity. The process

General Features of In Situ and Operando Spectroscopic Investigation

103

Table 4.1. Conceptional criteria for the selection of probe molecule to characterize solid acids (from ref. [31]) Lewis acid site Sorption complex Detection of complex formation

Electron pair donor acceptor Change in the wavenumber of the absorption maximum of B

Brønsted acid site Hydrogen bond Change in shape and absorption maximum of OH

Most exact methods of acid strength determination

Correlations between the changes in B and the heat of adsorption

Shifts of OH for a given probe molecule

Determination of concentration of acid sites Required spectral properties

From the intensity of B

From the intensity of OH

Shift of B must be significant compared to its half width

Absence of OH groups in the probe molecule

Frequently used molecules

Pyridine, ammonia, acetonitrile, benzonitrile, CO

Benzene, acetone, pyridine, substituted pyridines, amines, acetonitrile

Ion pair (hydrogen bonded) Disappearance of the catalysts OH . Appearance of  + B-H and/or + B-H Thermal stability of the hydrogenbonded ipc From the intensity of characteristic bands of the ipc The characteristic band has to be unequivocally attributed to the ipc Ammonia, pyridine and its derivatives

B = characteristic vibration of the probe molecule B. OH = OH stretching vibration of the catalyst. ipc = Ion pair complex

is based on the reaction between NO and NH3 to produce water and nitrogen according to the global equation: 4NO + 4NH3 + O2 = 4N2 + 6H2 O Commercial SCR catalysts are constituted by a high surface area TiO2 carrier (anatase) that supports the active components vanadium pentoxide and tungsten (or molybdenum) trioxide [36]. Vanadia is responsible for the activity of the catalyst in the reduction of NOx , but also for the undesired oxidation of SO2 to SO3 . The V2 O5 content was thus empirically kept low (0.3 ± 1.5%, w/w). It was recently evidenced by Raman spectroscopy [37] that for low vanadium loading, mainly superficial vanadate species are formed, whereas for high vanadium amount, the crystalline V2 O5 phase is also observed. The NOx SCR efficiency is reported to increase with the amount of polymeric vanadate species. In fact, these coordinated VOx surface species contain V−OH groups observed by infrared around 3655 cm−1 [38], which according to Topsøe act as Brønsted sites capable of binding NH3 during reaction [39]. The catalytic cycle represented in Scheme 4.1 was deduced from both Raman and infrared measurements.

104

Past and Present in DeNOx Catalysis O–…H N+H3

V5+

NH3

O V5+

V5+ O–…+H3N…H O V4+ NO

H2O

V5+ O H V5+ O–…+H3N N

O …H O V4+ O2

N2 + H2O

V5+ O–…+H3N N

H O V4+

O

Acid-base

Redox

Scheme 4.1. Catalytic cycle for SCR reaction over vanadia−titania catalyst (from ref. [39]).

WO3 is employed in larger amounts (10%, w/w): it acts as a promoter by enlarging the temperature window of the SCR reaction and imparts superior thermal stability and better mechanical properties to the catalysts [40]. Commercial catalysts containing MoO3 instead of WO3 are also used being less active, but more tolerant to As. An example of the characterization of their acidity by NH3 adsorption [41] is given on Figure 4.2 showing bands at 1606 cm−1 from NH3 coordinated over Lewis acid sites (mainly from the TiO2 support) and at 1445–1435 cm−1 typical of NH+ 4 formed from the acidic V−OH previously evidenced by Topsøe et al. [38]. Despite their rather good activity, vanadia-based mixed oxides are expected to be replaced because of their high deactivation upon ageing. One of the pioneer work giving an alternative to mixed oxides was reported by Richter et al. [42]. This paper dealing with the NOx SCR using ammonium zeolites was followed by others aiming at understanding the activities of the differently coordinated NH+ 4 species in YFAU (faujasite) zeolites [43] and GAPON (gallium phosphate) compounds [44]. Moreover, transition metal partially exchanged acidic zeolites are also of great interest for NOx SCR using hydrocarbons such as propane/propene (C3 H8 /C3 H6  [45,46] or methane (CH4 ) [47]. Thus, characterization of acidic zeolites is

0.60

Absorbance

0.48 0.36

a

0.24 0.12 0.00 2200

b

2000

1800

1600

1400

1200

Wavenumber (cm–1)

Figure 4.2. FTIR spectra of adsorbed species on catalyst after contact with ammonia (a) and outgassing at room temperature (b) (from ref. [41]).

General Features of In Situ and Operando Spectroscopic Investigation

105

of interest to better understand the reaction mechanisms. Zeolites are crystalline microporous silico-aluminates in which aluminium is constrained in a tetrahedral environment leading to framework oxygen atoms that are proton holder in the acidic form. The so-formed O−H bond gives rise to an active infrared vibration around 3600 cm−1 that is frequently used to characterize the Brønsted acidity. An attempt to rationalize the stretching frequencies of lattice hydroxyl groups in protonic-zeolites has been given [48]: the frequency of the unperturbed OH groups – vibrating in structural elements larger than eight-membered rings – is linearly correlated with the Sanderson intermediate electronegativity of the zeolites, which in its turn varies with the average chemical composition. This implicitly suggests that the crystal composition is homogeneous, and samples dealumined with, e.g. chelating agents, which show distinct Al gradients in the crystals, cannot obey such a linear relation. For homogeneous zeolites, the lower the aluminium content, the higher the intermediate Sanderson electronegativity and thus higher the hydroxyl acidic strength. Another mean for a direct characterization of zeolites thus consists in studying the structural (Al−O) vibration whose wavenumber is sensitive to the number of structural aluminium in the unit cell. Empirical relations are, for example, given for the YFAU [49] or the mordenite (MOR) [50] structure. Nevertheless, the characterization of zeolites acidity is not so evident and the infrared study of adsorbed molecules is very informative. The choice of the probe molecule is crucial to obtain an overall view of the acidity. Its size has to be small enough to interact with all available sites and to avoid confinement effects [51], but its basic strength has to be strong enough to interact even with the lower acidic sites. Ammonia (NH3 ) seems to be a good candidate for this, but due to the high polarity of the NH bonds, hydrogen bonding with basic entities governs the coordination of adsorbed species and direct conclusions about acidic strength are not straightforward [52]. That is why more often the adsorption of several probe molecules is required. For example, the faujasite (FAU) and the MOR structure are made of large and small cavities in which some acidic hydroxyls are out of reach of basic molecules such as pyridine. Coadsorbing the strongly basic trimethyl amine (TMA) and NH3 (see Figure 4.3), we were recently able 3637 TMA adsorption

3548 a b

c d

0.4

NH3 saturation

e f

g

h

i

3501 3750

3650

3550

3450

3350

Wavenumber (cm–1)

Figure 4.3. Infrared spectra of the HY sample upon TMA adsorption and NH3 saturation evidencing OH in the supercages at 3637 cm−1 , OH in the sodalite units at 3548 cm−1 and OH in the hexagonal prism at 3501 cm−1 [54].

106

Past and Present in DeNOx Catalysis

to give an infrared evidence of three distinct acidic hydroxyls in defect-free HY [53] and to characterize their respective acidic strength [54]. The combined use of these two molecules moreover helped us to better characterize the various coordinated NH+ 4 and determine the activity ranking between the ammonium species and coordinated ammonia over Lewis sites during NOx SCR [43]. Concerning the Lewis acidity in zeolites, some strong sites can be obtained during steaming leading to the formation of extraframework aluminium species. Soft Lewis sites may be naturally present (in the alkaline form) or generated upon ionic exchange with transition metals, which are necessary for NOx SCR with hydrocarbons. For overexchange level, some Lewis species may remain on the external surface and coadsorption of the bulky ortho-toluonitrile and CO was recently reported to identify the different Con+ species and their location in a Co−H−MFI zeolite [47]. The characterization of the transition metal containing catalysts will be further described as they are used for their redox properties.

2.2. Catalyst basicity Heterogeneous catalysis using basic solids has been much less studied than acidic catalysis, however, the recent legislation concerning the NOx level for automotive exhaust gases from diesel or lean burn gasoline engines switched on the light on basic materials. Effectively, the typical three-way catalysts (TWC) (based on ceria and ceria−zirconia solid solutions whose basic properties were already exploited) are unable to work properly out of stoichiometric conditions. As it seems rather difficult to reduce NOx species in the presence of excess air (lean period), it was firstly proposed as an alternative to fully oxidize the NOx and to trap them on the catalyst surface, their reduction being further achieved during short injections of hydrocarbons (rich period). This is the so-called NOx -trap concept developed by Toyota [55] and recently reviewed by Epling et al. [56]. The basic function of the catalyst must then be properly tuned to maintain nitrites or nitrates species adsorbed on the surface during lean conditions, but not in a too stable way in order to enable their reduction during the rich period. As for the acidity study, we will first describe some basic principles and will then focus on some recent examples in line with DeNOx applications.

2.2.1. Basic principles for investigations on basicity As an acid site is always associated to its conjugated basic site, reviews dealing with characterization of acidity by infrared also report interesting data about basicity [30,57–59] and even describe some typical probe molecules interactions: CH-acids such as chloroform Cl3 CH(D), acetylene C2 H2 and methylacetylene CH3 C2 H are shown to be potentially suitable probe molecules for basic properties using the H-bonding method [34]. All three molecules undergo Oz2 · · · H−C hydrogen bonding and the induced red-shift of the C−H stretching frequency permits a ranking of the base strength of a given series of materials. Many other probe molecules were tested for the specific study of the surface basicity of divided metal oxides [60] and Lavalley recently reviewed the infrared spectrometric studies of the surface basicity of metal oxides and zeolites using adsorbed probe molecules [61]. Results obtained from carbon monoxide (CO), carbon

General Features of In Situ and Operando Spectroscopic Investigation

107

dioxide (CO2 , sulphur dioxide (SO2 , pyrrole (C4 H5 N), chloroform (CHCl3 , acetonitrile (CH3 CN), alcanes, thiols, boric acid tri-Me-ether, ammonia (NH3  and pyridine (C5 H5 N) are discussed. As we already noticed in the case of the acidity study, the author reminds us that no probe can be used universally. CO2 for weakly basic metal oxides and for basic OH groups, CO for the characterization of highly basic structural defects on metal oxides activated at high temperature and pyrrole in the case of alkaline zeolites, appear to be quite suitable probes. Moreover, both NH3 and pyridine (generally used as probes for the measure of the acidity of catalysts) are also described to adsorb on basic oxides via dissociative chemisorption. More specific study dedicated to the Lewis basicity in zeolites was initiated by Barthomeuf who related the basic strength with the O charges calculated using the Sanderson equalization principle and insisted on the fact that interaction of adsorbed molecules with acid–base pairs should be dependent of the zeolite structure [62]. Alkaline zeolites are described as rather soft bases, their basicity increasing within the series Li < Na < K < Rb < Cs [63]. A quite exhaustive review of their basic properties is also due to Barthomeuf [64]. More recently, NO2 disproportionation on alkaline zeolites was used to generate nitrosonium (NO+ ) and nitrate ions whose infrared vibrations are shown to be very sensitive to the cation chemical hardness and to the basicity of zeolitic oxygen atoms [65,66]. The nature of the nitrites/nitrates species formed upon NO2 adsorption on basic Lewis oxygen is thus very informative of their basic strength. Hadjiivanov focussed on the identification of neutral and charged Nx Oy surface species by IR spectroscopy and Figures 4.4 and 4.5 are extracted from his review [67] essential for anyone interested in DeNOx catalysts.

I

II

O

N

O

O

M

M

Monodentate nitrito

VIII

O

N

O M

Chelating bidentate nitrito

O

O

M

M

Bridging bidentate nitrito

O M

Monodentate nitrato

X

M

Bridging monodentate nitrato

XI

O

O N

N O

O N

M

N

O

O

O M

Bridging monodentate nitrito

N

O N

IV

III

IX O

O

M Chelating bidentate nitrato

O

O

M

M

Bridging bidentate nitrato

V O

O N M Nitro

VI

VII O

O N

N

O M

Chelating nitro-nitrito

M

O M

Bridging nitro-nitrito

Figure 4.4. Possible structures of nitrites NO2 − species (I to VII) and nitrates NO3 − species (VIII to XI) (from ref. [67]).

108

Past and Present in DeNOx Catalysis NO2+

h

NO3– NO3–

(free-like)

NO3– NO3– N2O4 NO2 NO2– NO2– NO2– NO2– NO+ N2O3

i

a - v(N-N) (monodentate) b - v(N-O) c - v(N=O) (chelating) d - v(N=O) (bridged) e - v2(NO)2 f - v44(NO)3 g - v3(NO2) h - v44(NO2) (nitro-nitrito) i - v(NO3) (nitro)

i

i

i

i

i

i g

h h c

b g

h

g

h

(bidentate nitrito) (monodentate nitrito) d

b

c c

g

h f

(NO)2

a

NO N2O2–

c a

e b

NO– N2O

a

b

a

N2

2400

2200

2000

1800

1600

1400

1200

1000

Wavenumber/cm–1

Figure 4.5. An help to investigate the nature of surface Nx Oy species: N−N and NO stretching modes [67].

2.2.2. Basicity of DeNOx catalysts: selected examples Considering the NOx reduction exhausting from gasoline engines, the first generation process involved TWC based on CeO2 and CeZrO2 . They were so-called due to their ability to remove both NOx , CO and non-burnt hydrocarbons in a single step. Ceria and ceria−zirconia solid solutions surface properties were thus extensively studied [68–73], and we will here summarize their basic characteristics. Pyrrole adsorption on CeO2 leads to dissociative adsorption characterized by stretching ring vibrations at 1444 and 1367 cm−1 typical of the pyrrolate ions and (OH) vibration at 3628 cm−1 typical of surface hydroxyls formed upon proton transfer [68]. This complete dissociation of C4 H5 N is indicative of the high basicity of CeO2 surface O2− ions, but does not allow an investigation of its variation upon reducing ceria. CO2 was further adsorbed as it acts as a Lewis acid either towards O2− surface ions (carbonates so being produced) or towards residual basic OH surface species [hydrogen carbonates (HC) so being produced]. The study extended to Ce−Zr mixed oxides [69] indicates that HC are mainly observed for rich ceria compounds (see Figure 4.6) and that the intensity of carbonate species is directly proportional to the cerium content, as it was expected according to the basic properties of this element. Identification of the spectral features typical of each species arising from CO2 adsorption was clarified studying the split of the 3 band of carbonates and their thermal stability. Table 4.2 summarizes their band positions. Considering now the more basic catalysts required for the NOx -trap process, the typical formulation of the material originally described by Toyota is the following: Pt−Rh/Ba/Al2 O3 . The -Al2 O3 support basicity (characterized by CO2 adsorption [74]) is limited; nevertheless, it is proposed that alumina surface O2− could participate in the NOx storage process. However, Epling et al. [56] explain that even if the participation of alumina in the overall NOx -trapping mechanism cannot be ruled out, this catalyst

General Features of In Situ and Operando Spectroscopic Investigation HC

a

= 0.5 a.u. a

= 0.2 a.u. b

c

d

e f 3800

3700

3800

Wavenumber (cm–1)

109

A b s o r b a n c e

HC BC

HC

b

HC BC HC

c d

e f 2500

2000

1500

1000

Wavenumber (cm–1)

Figure 4.6. FTIR spectra of 1.3 kPa of CO2 adsorbed at room temperature over oxidized CZ-x/y samples where x and y stand, respectively, for the molar CeO2 amount and molar ZrO2 amount. (a) CZ-100/0, (b) CZ-80/20, (c) CZ-68/32, (d) CZ-50/50, (e) CZ-15/85, (f) CZ-0/100. HC stands for hydrogen carbonate and BC for bridged carbonate [69].

support component will play a less-significant role as an ultimate NOx trap in diesel engine exhaust gas, since the large quantities of H2 O which are present in hydrocarbon fuel combustion exhaust will lead to alumina surface hydroxyl groups high coverage, which hinder the NOx sorption process. The major contribution to the overall NOx adsorption capacity is thus provided by alkali and alkaline-earth components (i.e. Ba for the Toyota catalyst) due to their highly basic properties. A comparative study of barium- and potassium-based formulation [75] indicated upon NO2 adsorption the formation of both ionic- and covalent-like NO3 − species over Pt−Rh/Ba/Al2 O3 , whereas only very stable ionic potassium nitrates (sharp peak at 1373 cm−1 ) were detected over Pt/K/Mn/Ce−Al2 O3 . This is due to the higher basicity of the potassium sites, which furthermore enlarges the adsorbing temperature window and delays the nitrates release during the rich step impeaching NO sudden outlet. Alumina-supported indium oxide was recently reported to be an efficient catalyst for NOx reduction with ethylene in an oxygen-rich atmosphere [76], but even if the amphoteric In2 O3 oxide is described to be more basic (InO−H basic hydroxyl with an associated band at 3766 cm−1 ) than acidic, no direct link with the activity is given. Using ammonia as a powerful reducing agent is now even studied for automotive applications especially for trucks. NH3 can be provided by thermal hydrolysis of urea stored in a tank and the availability of such a strong reducing agent allows the NOx elimination at much lower reaction temperatures. This opened up the way to potential applications for zeolites, which were given up in automotive depollution due to their low thermal stabilities. One recent paper even reported the direct SCR of NO2 with urea in nanocrystalline NaY zeolite where external acidic defects sites are described to favour the urea decomposition into NH3 and the internal basic oxygens to provide the NOx storage sites [77].

110

Table 4.2. Band position (cm –1) of species arising from CO2 adsorption at room temperature [69] Samples

CO2 Linear Bent1 Hydrogen carbonates as ,

CZ-100/0-HS

s

3,

1

2353

1736

1377

1120

Bidentate carbonates

Monodentate carbonates

(OH)

(CO3 )

(OH)

(CO3 )

3617

1599

1218

823

1576 1290 1021

856

1504 1351 1045 1465 1353 1060

856

1219

836

1575 1304 1011

868

1510 1325 1060 1476

857

1219

834

1574 1306 1012

871

1514 1329 1059

857

3

3

1

(CO3 )

Polydentate carbonates

3

3

2

3

3

1

(CO3 )

1413 1025

CZ-80/20-HS

2352

1776

3617

1405 1060

CZ-68/32-HS

2352 1376

1775

3619

1598 1407 1059

Past and Present in DeNOx Catalysis

1377

1592

2354

1806(sh), 1774

3616

1376

1 59 5

1221

833

1565

1296

1025

872

1519

1336

1060

1449

1056

856

1223

86 1

1546

1305

1060

866

1520

1331

1060

1448

1042

857

1225

861

1553

1314

1063

862

1421 1060

CZ-15/85-HS

2355

1806(sh), 1775

3611

1376

1599 1425 1060

CZ-0/100-HS

2359

1806(sh), 1778

1374

1150

3608

1610 1450 1063

1

General Features of In Situ and Operando Spectroscopic Investigation

CZ-50/50-HS

These bands are assignable also to bridged carbonate species; sh = shoulder

111

112

Past and Present in DeNOx Catalysis

To finish with another trend for NOx removal consisting in NO direct decomposition, we would like to depict the infrared study of NO adsorption and decomposition over basic lanthanum oxide La2 O3 [78]. In this case, the basic oxygens are proposed to lead to NO2 − and NO3 − spectator species, whereas the active sites for effective NO decomposition are described as anion vacancies, which are often present in transition metal oxides. This last work makes the transition with the study of DeNOx catalysts from the point of view of their ability to transfer electrons, i.e. their redox properties.

2.3. Catalyst redox properties, oxygen mobility and supported metal characterization The aim of the DeNOx treatment is to transform NOx into the most inert and healthy form of nitrogen containing species, i.e. dinitrogen (N2 . From a basic chemical point of view, the NOx transformation into N2 appears as a redox reaction and therefore redox properties of catalysts are undoubtedly solicited. Indeed, the NOx -trap process is − described to proceed via the following steps: NO2gas → NO2 − ads → NO3 ads but NO2 is far from being the main NOx component in diesel exhaust (10%) and NO thus needs to be oxidized into NO2 . Concerning the three-way process, effective reducing agent of NO are reported to be partially oxidized hydrocarbon species (Cx Hy Oz ), which come from the mild oxidation of hydrocarbons by NO2 [79]. One requisite for a TWC efficiency is then its ability to oxidize NO into NO2 . The NO oxidation into NO2 is also required in order to improve the ammonia SCR NOx reduction reaction kinetic. Finally, for the NOx -trap process, the catalyst must also be efficient for the oxidation of CO and unburnt hydrocarbons during rich conditions. A proper characterization of the redox active sites is quite important as they are subject to poisoning by sulphur containing species whose elimination is energy demanding. We will try in the next part to summarize the main results obtained in this domain.

2.3.1. Basic principles for investigations on redox properties Transition metal oxides, rare earth oxides and various metal complexes deposited on their surface are typical phases of DeNOx catalysts that lead to redox properties. For each of these phases, complementary tools exist for a proper characterization of the metal coordination number, oxidation state or nuclearity. Among all the techniques such as EPR [80], UV–vis [81] and IR, Raman, transmission electron microscopy (TEM), X-ray absorption spectroscopy (XAS) and NMR, recently reviewed [82] for their application in the study of supported molecular metal complexes, Raman and IR spectroscopies are the only ones we will focus on. The major advantages offered by these spectroscopic techniques are that (1) they can detect XRD inactive amorphous surface metal oxide phases as well as crystalline nanophases and (2) they are able to collect information under various environmental conditions [83]. We will describe their contributions to the study of both the support (oxide) and the deposited phase (metal complex). Raman spectroscopy has been used for a long time in order to study supported and promoted metal catalysts and oxide catalysts [84] since many information can be obtained: (1) identification of different metal oxide phases; (2) structural transformations of metal oxide phases; (3) location of the supported oxide on the oxide substrate and

General Features of In Situ and Operando Spectroscopic Investigation

113

(4) structure of the supported metal oxide phase [85]. The molecular structures of the =O and bridging M−O−M surface metal oxide species are reflected in the terminal M= vibrations. The location of the surface metal oxide species on the oxide supports can be detected by directly monitoring the surface hydroxyls of the support that are being titrated. A review limited to supported metal oxide catalysts containing group V–VII transition metal oxides (e.g. V, Nb, Cr, Mo, W and Re) on several different oxide supports (alumina, titania, zirconia, niobia and silica) [86] was recently completed in order to give more insight in the molecular structure and reactivity of the group V metal oxides [87]. Paper more specifically dedicated to the identification and characterization of the highly isolated transition metal ion and oxide species in the framework of zeolites and on oxide supports by UV resonance Raman is given elsewhere [88]. Infrared spectroscopy has also been widely used in order to characterize the metallic centre in oxides and deposited metal complex, however even if direct investigations on metal–oxygen vibrations are reported [36,89–92] most of the studies related with catalysis are dealing with the adsorption of probe molecules. Among these molecules, N2 , methanol, NO and CO are frequently used and the last two ones, which are the most interesting in the frame of DeNOx application (being reactants), are also from far the most common in the literature. In 2002, two reviews concerning the infrared spectra of chemisorbed carbon monoxide as a characterization tool for the cationic sites of oxides [35] or more generally for metal heterogeneous catalysts [93] were published. A peculiar property of CO is that the slightly antibonding HOMO 5 orbital is occupied. This orbital is very important for the electron-donating properties of CO, because a decrease of electron occupation on it leads to stabilization of the entire molecule and thus to an increase [if compared to the  (CO) gas at 2143.5 cm−1 ] of the  (CO) wavenumber. On the contrary, the addition of one electron from a metal d orbital to one of the 2 ∗ LUMO orbitals (so-called -back donation) leads to a substantial decrease of the vibrational frequency of CO, i.e. to a weakening of the CO bond. As far as the redox properties are concerned and in the ideal case, the information to be expected using CO as a probe is as follows: • • • • • •

oxidation state of the cations on the surface. coordination state of these cations. position of the cations on flat planes or other surface structures. position of the supported active phase. surface phase analysis. existence of strong oxidizing agents on the surface.

For the study of mixed oxides, one should characterize the various sites. In this case, the first step is to characterize the CO adsorption at various equilibrium pressures at low temperature, followed by evacuation at increasing temperatures to obtain information about the stabilities of the various species. Although the C−O stretching frequency is the most informative parameter, the data determining the stabilities of the various species can be decisive for the assignment of the bands. Multiple carbonyls adsorbed on the same metal cation are possible, and in order to identify them isotopic mixtures should be used. Sometimes the polycarbonyls are very stable and in this case, if 12 CO is adsorbed first and then 13 CO introduced, mixed species may not form at ambient temperature. Concerning NO, Hadjiivanov [67] reported that the coordination of the NO molecule to a cationic site via the nitrogen atom is accompanied by a partial charge transfer from

114

Past and Present in DeNOx Catalysis

the 5 orbital and an increase in the bond order, just as in the case of CO. Formation of a -back bond, although not so easy as with CO, is also possible, and this results in a decrease in the N−O stretching modes. The different surface mononitrosyls absorb in a wide spectral range: 1966–1710 cm−1 . When only a  bond is formed, a frequency above that of gaseous NO (1876 cm−1 ) is expected, whereas with low-valent cations, rich in d-electrons, -back donation is possible and the N−O stretching modes can fall below 1876 cm−1 . Cations having no d-electrons produce mononitrosyls only; on the contrary, dimeric molecules are very often the principal adspecies on transition metal cations. This is the case of NO adsorption on V-, Cr-, Mo-, W-, Fe- and Co-containing oxide systems where the metal cations are not in their highest oxidation state. Thus, it is evident that a -back donation stabilizes dinitrosyls. Examples of complementary information obtained by CO and NO coadsorption are available [59] and further details will be given in the following part dealing with DeNOx catalysts.

2.3.2. Redox properties of DeNOx catalysts: selected samples Raman studies [87] revealed that the active surface sites present in pure V2 O5 are primarily redox sites. This is in complete agreement with the requirement of both acidic and redox properties of the V2 O5 /TiO2 NH3 SCR catalyst (see Scheme 4.1). Adsorption of both CO and NO is also informative for the determination of the oxidation state of vanadium on vanadia−titania catalysts. Although the carbonyl bands for V4+ −CO, V3+ −CO and Ti4+ −CO species almost coincide [94], the fact that NO forms dinitrosyls with Vn+ but not with Ti4+ allows the effective use of NO as a probe molecule [95]. Going on in the frame of NOx reduction by ammonia, TiO2 -supported MoO3 catalysts with increasing amount of MoO3 loading were characterized by both IR and Raman spectroscopy. Figure 4.7 reports the Raman spectra obtained in air using the pure powders [96]. All the spectra show bands at 396, 514, 637 cm−1 due to the Raman active fundamentals =O stretching mode of of anatase. A weak band near 950 cm−1 , characteristic of the M= molybdenyl species, is also well evident in all the samples. The catalysts with MoO3 loading of 8.7 and 11.3% (w/w) show additional broad bands near 820 and 980 cm−1 , associated with the presence of micro-crystalline MoO3 . Complementary EPR study confirmed the presence of Mo5+ molybdenyl species whose ability to transform into Mo6+ probably provides the required redox properties to the catalyst. Concerning the NOx reduction for lean burn engines, it has been reported that NO reduction to N2 could be achieved with hydrocarbons as reducing agent even in the presence of excess O2 using Cu−zeolite catalysts [97]. To improve the process a huge amount of basic researches on the copper electronic state mainly in the ZSM-5 structure was undertaken. Since CO complexes with Cu2+ are only stable at very low temperature, CO adsorption is specific to Cu+ sites. Its characteristic band at 2158 cm−1 provides quantitative results on integrating its molar extinction-coefficient [45]. Increasing the CO vapour pressure above 2 Torr, the vibrations of Cu+ (CO)2 dicarbonyls with associated s and a modes lying at 2178 and 2151 cm−1 are observed. Hadjiivanov et al. [98] describe the water effect (which is always present during DeNOx real conditions) and report that bands at 2158 and 2134 cm−1 may characterize CO bound to ‘dry’ and ‘wet’ Cu+ centres, respectively. However, a comparative study using both Cu-ZSM-5 and CuO/Al2 O3 allowed Praliaud et al. to propose the (CO) at 2123–2133 cm−1 to be due to nonisolated Cu+ species (Cu+ surrounded by Cu2+ ions) arising from the partial reduction

Intensity* 10–5 (u.a.)

Intensity* 10–3 (u.a.)

9.0

MoO3(8, 7)/ TiO2

7.2 5.4 3.6

107.1 100.8 94.5 88.2

1.8 81.9 1199 1104 1009 914

819

724

629

534

439

344

1250 1150 1050 950

Wavenumber (cm–1) MoO3(6)/ TiO2

20.8

Intensity* 10– 6 (u.a.)

Intensity* 10– 4 (u.a.)

68.8 60.2 51.6 43.0 34.4 25.8

1201 1106 1011 916 821

728

631 536

Wavenumber (cm–1)

850

750

650

550

450

350

534

439

344

Wavenumber (cm–1)

441

346

MoO3(11, 3)/ TiO2

19.5 18.2 16.9

General Features of In Situ and Operando Spectroscopic Investigation

113.4 MoO3(3, 1)/ TiO2

15.6 14.3 1199 1104 1009 914

819

724

629

Wavenumber (cm–1)

Figure 4.7. FT-Laser Raman spectra of MoO3 (3.1)/ TiO2 (a), MoO3 (6)/ TiO2 (b), MoO3 (8.7)/ TiO2 (c) and MoO3 (11)/ TiO2 (d) catalyst [96]. 115

116

Past and Present in DeNOx Catalysis

of bulk CuO. Whereas, the band at 2152–2157 cm−1 would characterize isolated Cu+ ions, which are described to be responsible for the high activity in NO reduction into N2 [99]. The almost perfect match between the positions of the Fermi level of Cu+ ZSM5 and NO molecule, together with proper symmetry of the interacting orbitals, should be responsible for a strong -back donation, which seems to be crucial in the process of NO activation [100]. Fine coating of Cu-ZSM-5 with nanoparticles of ceria can enhance the oxidation power of the catalyst as evidenced by a lower stability of the copper carbonyl complex in oxygen [101]. Concerning NO, its adsorption even at room temperature may lead to Cu+ oxidation into Cu2+ and, therefore, its use for the determination of the copper oxidation state distribution is rather difficult. However, the formation of Cu+ mono and dinitrosyls is observable for high temperature (770 K) vacuum activated Cu-ZSM-5, and Datka et al. [102] even report the possible existence of two distinct Cu+ −NO species at 1812 and 1825 cm−1 , associated to two distinct Cu+ sites differing in the density of oxygen packing. The higher the oxygen density, the higher the Cu+ site ability to activate the NO molecule. This idea previously proposed by Wichterlová et al. [103] could explain why Cu-zeolites are more active than mesoporous Cu−MCM−41. Going on with the characterization of Cu+ by NO, it has been shown that the zeolitic structure also influences the symmetry of the dinitrosyl species, and for Cu−MOR sample the Cu+ −NO (1813 cm−1 ) transformation into Cu+ (NO)2 leads to different spectral feature depending on the Cu+ location: a doublet with s and as at 1828 and 1730 cm−1 in the main channels and another doublet with s and as at 1870 and 1785 cm−1 in the constrained side-pockets [104]. For oxidized Cu-ZSM-5, the great reactivity upon NO adsorption leads to Cu2+ −NO nitrosyls and nitrates formation. Both the fast appearance of nitrates species and the relatively low frequency (1884–1868 cm−1 ) of the band assigned to NO stretching of nitrosyls formed on associated Cu2+ sites are tentatively explained by a rather high Cu loading, which would lead to the formation of polynuclear (−Cu−O−)n species whose extra-lattice oxygen possess a high oxidation power [98]. More recently, iron containing ZSM-5 were also studied for their potential application in DeNOx by hydrocarbons. Complementary study using EXAFS and NO adsorption followed by infrared spectroscopy allowed the identification of different Fe2+ species: some isolated ones leading to Fe2+ −NO, with corresponding (NO) at 1841 cm−1 , and some non-isolated ones (called iron oxo-nanoclusters with structure similar to the one of ferredoxin or high-potential iron protein) leading also to Fe2+ mononitrosyl, but characterized by a (NO) at 1880 cm−1 . These iron-oxo nanoclusters are reported to be the active sites for NOx SCR with propene [105]. Starting from roughly the same experimental results, oxygen bridged binuclear iron complex [HO−Fe−O−Fe−OH]2+ were also proposed as the active sites [106]. Fe3+ characterization in the Y structure was also previously studied using CO adsorption [107]. Now focusing on the TWC, the redox properties are known to emerge from the ability of the Ce cation oxidation number in ceria and ceria−zirconia to easily pass from 3+ to 4+ and reciprocally. Given a TWC catalyst, the surface state (reduced or oxidized) is available using methanol (CH3 OH) adsorption [71]. Its dissociation leads to cation coordinated methoxys and hydroxyls formation. Both the (CO) wavenumbers associated to methoxy species [72] and the (OH) associated to surface hydroxyls [73] depend on the cerium oxidation state. The Ce4+ /Ce3+ surface ratio is thus available from the quantitative study of the corresponding methoxy intensities. Figure 4.8 describes the consecutive adsorption of oxygen (O2 )-calibrated doses after methanol dissociation

General Features of In Situ and Operando Spectroscopic Investigation Oxidised Ce4+ sites: I

117

II-A

0.2 A b s o r b a n c e

n)

O2 addition

a) 1400

Reduced Ce3+ sites: I 1300

1200

II* 1100

1000

Wavenumber (cm–1)

Figure 4.8. (OC) bands for methoxy species adsorbed on CZ-50/50 sample reduced at 673 K, then reoxidized at room temperature by adding successive doses of O2 (spectra a–n) [72].

over prereduced cerium−zirconium mixed oxide, which enable the determination of the oxygen storage capacity of the sample. Another important point to study is the redox properties evolution upon increasing temperature. Reduction of surface Ce4+ into Ce3+ occurs for temperature above 423–473 K, which corresponds to the transformation of methoxy into formate species [70]. For temperature above 523 K a new band appears at 2120 cm−1 and is attributed to the 2 F5/2 →2 F7/2 electronic transition of Ce3+ ions in internal structural defects thus indicating the beginning of bulk ceria reduction. The disappearance of this characteristic electronic transition band upon NO adsorption at 773 K together with the formation of N2 were given as proofs for the NO dissociation on surface defects leading to ceria reoxidation [108]. Finally, for the last generation of NOx -trap catalysts the redox function is brought by supported noble metals (i.e. platinum for the original Toyota formulation). The noble metal role being to favour the NO to NO2 oxidation during lean period, as it was evidenced by the disappearance of the Raman line at 639 cm−1 typical of the B1g active mode of palladium oxide upon exposure of a Pd/Al2 O3 catalyst to NO at 473 K [109]. Noble metals are also described to be effective for the exothermic hydrocarbon combustion during fuel pulse injection and the so obtained reduced noble metals (metallic like) to be further active for the NOx decomposition [110]. One important characteristic of supported metal particles is their size, which is related to the metal dispersion defined as the fraction of metallic atoms present on the surface. It is reported that catalyst DeNOx activity increases with Pt dispersion [111], however other investigations suggest that the bigger the Pt particles, the better is the NO oxidation to NO2 [112]. An optimal value of noble metal dispersion must then be achieved and a proper measure of this value is then fundamental. Reliable dispersion values can be deduced from CO

118

Past and Present in DeNOx Catalysis

quantitative chemisorption at room temperature followed by Fourier transform infrared (FTIR) spectroscopy. This technique effectively allows to discriminate the adsorption on the platinum surface from that on the oxide support [113]. Now that we have described the catalytic ‘playground’ and the different places where the reaction can take place, we will concentrate on the molecules reactivity using the in situ approach.

3. WHAT ARE THE PIECES MOTIONS? THE IN SITU INFRARED APPROACH Our purpose in this part will be to determine how can the adsorbed molecules react on the surface and among all the playground, where does the real action take place. This kind of studies aiming at elucidating the various reaction paths required several technical developments. The most widespread vibrational spectroscopic methodology probably consists in the study of reaction mechanism using batch IR reactor cells. Some further improvements in the DeNOx description were provided by the transient analysis allowing to differentiate spectator species from intermediate ones. The spectroscopic approach of some more classical analysis techniques such as temperature-programmed reduction (TPR) enables the characterization of the surface upon heating and thus provides another interesting complementary tool. Any of these techniques are further described in detail focusing on typical examples, which led to a significant contribution in the understanding of DeNOx mechanisms.

3.1. Batch IR cell reaction All techniques dealing with the introduction of reactants inside an infrared reactor and the study of the evolution of typical adsorbed species with time will be reported in this part. Generally, the systems work in IR transmission mode and several cells are commonly used, both commercial and home made ones. A recent review gives a rather complete description of the main ones [114]. Prior to the IR experiment and whatever the used cell, the catalyst is generally heated in a vacuum at high temperature to remove adsorbed impurities. Subsequently, the cell is cooled to the relevant reaction temperature, and a spectrum of the activated sample is generally recorded. This spectrum is further used to calculate difference spectra enabling to evidence the adsorbed reactants characteristic IR signature (negative bands) and the adsorbed products ones (positive bands). It is believed that SCR by hydrocarbons is an important way for elimination of nitrogen oxide emissions from diesel and lean-burn engines. Gerlach et al. [115] studied by infrared in batch condition the mechanism of the reaction between nitrogen dioxide and propene over acidic mordenites. The aim of their work was to elucidate the relevance of adsorbed N-containing species for the DeNOx reaction to propose a mechanism. Infrared experiments showed that nitrosonium ions (NO+ ) are formed upon reaction between NO, NO2 and the Brønsted acid sites of H−MOR and that this species is highly reactive towards propene, forming propenal oxime at 120 C. At temperatures above 170 C, the propenal oxime is dehydrated to acrylonitrile. A mechanism is proposed to explain the acrylonitrile formation. The nitrile can further be hydrolysed to yield

General Features of In Situ and Operando Spectroscopic Investigation

119

adsorbed ammonium ions, which are known to be efficient in reducing nitrogen oxides to nitrogen. The dehydration of propenal oxime appears to be the rate determining step in nitrile formation. By in situ IR spectroscopy, assisted by isotopic substitution, Szanyi et al. also determined the precise nature of the species formed on a Na-Y, FAU zeolite upon NO and NO2 adsorption [116]. Kantcheva [117] focused on other catalysts and identified the nature of the NOx species formed on TiO2 and MnOx /TiO2 catalysts by in situ FTIR and their reactivity with decane. The adsorption of NO with oxygen at room temperature leads to the formation of various kinds of surface nitrates differing in their coordination mode (see first part). Kantcheva particularly observed that nitrates formed on the manganesecontaining samples are characterized by a significantly lower thermal stability than the ones formed on the pure titania support. The interaction of decane with nitrates species was further tested by heating at various temperatures coadsorbed nitrates and decane on the closed IR cell. A correlation between the thermal stability of the nitrates and their reactivity towards the reducing agent is then proposed: bidentate nitrates on titania present no reactivity with adsorbed decane. On the contrary, monodentate and bridged nitrates formed on the manganese containing catalysts (having lower thermal stability) are able to oxidize adsorbed hydrocarbons at temperatures as low as 373 K into formic acid and isocyanate ions. The mechanism proposed for the interaction between the surface nitrates and the adsorbed decane describes that NO3 − and NCO species are intermediates for the dinitrogen formation. Ethanol has also been shown to be an efficient reducing agent for NOx SCR over alumina-supported silver catalyst Ag/-Al2 O3 [118,119]. Yeom et al. [120] proposed a mechanism for this reaction using batch IR cell technique. They identified various pathways with different intermediates in function of the reaction temperature. The proposed mechanism is summarized in Figure 4.9. At high temperature (320 C), ethanol principally reacts with oxygen to form acetaldehyde. The subsequent SCR mechanism is very similar to that identified for NOx reduction with acetaldehyde over BaNa/Y: surface acetate ions are formed and react with NO2 to yield nitromethane. With Ag/-Al2 O3 there is evidence that the aci-anion of nitromethane is an intermediate in the DeNOx process. As with BaNa/Y, ammonia is formed by hydrolysis of NCO− ions. This ammonia forms ammonium nitrite through reaction with NO + NO2 + H2 O. This compound is thermally unstable above 100 C and swiftly decomposes to N2 and H2 O. At 200 C, the dominant pathway for oxidation of ethanol is the reaction with NO2 . Ethyl nitrite, 200°C EtOH + 2NO + N2O + H2O 4CH3CH2OH

8NO2

4HNO3 + 4EtONO 3CH3HC O

Surface nitrate

3CH3COO–

320°C CH3CH2OH

O2

CH3HC

O

NO2

?

CH3COO–

NO2

CH2

NO2–

NCO– N2

NH3 NH4NO2

Figure 4.9. Mechanism for NOx reduction with ethanol on Ag/-Al2 O3 proposed in ref. [120].

120

Past and Present in DeNOx Catalysis

which is produced in this reaction, decomposes into various products, including N2 O and acetaldehyde. In principle, acetaldehyde is desirable because its adsorption leads to surface acetate; however, at 200 C, contrarily to the BaNa/Y system, these surface acetate ions are rather unreactive. In addition, N2 O is a relatively undesirable decomposition product, because it is not reduced to N2 under reaction conditions. Thus, nitrogen is sequestered in a ‘greenhouse gas’ rather than being converted to environmentally benign N2 . In an actual NOx SCR process with ethanol, most of the acetaldehyde could, in principle, be produced from ethanol oxidation by O2 , because the O2 concentration is in excess. Thus, a plausible efficiency for ethanol as a NOx reductant would consist in using it with a catalyst that efficiently convert ethanol to acetaldehyde via direct oxidation with O2 at low temperatures, and then takes advantage of the known efficiency of adsorbed acetaldehyde over BaNa/Y for NOx reduction. Other authors also determined by FTIR that organic nitrocompounds are formed as primary products of the NOx CH4 -SCR reaction on ZSM-5-based catalysts [121– 124]. They preadsorbed nitromethane on the sample placed in the IR cell and followed by IR its transformation into other intermediates under O2 and NO versus time at different temperatures. For Cu- and Co-ZSM-5, it was shown that around 300 C adsorbed nitromethane is easily converted into isocyanates and then melamine via polymerization of the former species. Both species easily interact with molecular oxygen, while no reaction with NO is observed and the reactivity depends on the temperature and the nature of the transition metal cation. Szanyi et al., using alternate heating and cooling cycles in a batch IR/reactor-cell, coupled with mass spectrometer (MS), temperature-programmed desorption (TPD) and time-resolved XRD techniques, investigated the efficiency of H2 and CO in the reduction of stored NOx on Pt/Al2 O3 and Pt/BaO/Al2 O3 catalysts. Both reducing agents were found to effectively reduce stored NOx , H2 being a much more effective reductant than CO. Moreover, it was pointed out that H2 more effectively reduced the surface nitrates than the bulk nitrates on Pt/BaO/Al2 O3 . In the reduction with CO (in the absence of H2 O), the formation of isocyanates bound to the oxide components of the catalysts was observed. In the absence of water, these species accumulated on the catalyst surface and exhibited high thermal stability. The surface-bound NCO species reacted readily with H2 O to form the hydrolysis products: NH3 and CO2 . The NH3 thus produced then could react with NOx to form an intermediate that eventually decomposes to give N2 and H2 O. Removing NOx from the Pt/BaO/Al2 O3 catalyst was very effective when both H2 and CO were used as reducing agents; the rate of removal was lower than that in the presence of only H2 but higher than that in presence of only CO. At low temperatures, H2 was the primary reducing agent. At high temperatures, NCO could react directly with NOx to form N2 and CO2 , whereas at intermediate temperatures, water, formed in the reduction with H2 , could hydrolyse NCO to form the intermediate that eventually decomposes to N2 and H2 O [125].

3.2. Transient-mode catalytic reaction In order to get closer from the catalytic conditions (for example in DeNOx reaction exhaust gases are to be treated) investigators developed reactor cells allowing the infrared study of catalysts underflow. The principle of transient technique is then to introduce

General Features of In Situ and Operando Spectroscopic Investigation

121

a perturbation on a continuous flow in order to briefly modify the steady state and discriminate spectator and intermediate species.

3.2.1. Transient pulse reaction The use of IR pulse technique was reported for the first time around the year 2000 in order to study a catalytic reaction by transient mode [126–131]. A little amount of reactant can be quickly added on the continuous flow using an injection loop and then introduce a transient perturbation to the system. Figure 4.10 illustrates the experimental system used for transient pulse reaction. It generally consists in (1) the gas flow system with mass flow controllers, (2) the six-ports valve with the injection loop, (3) the in situ IR reactor cell with self-supporting catalyst wafer, (4) the analysis section with a FTIR spectrometer for recording spectra of adsorbed species and (5) a quadruple MS for the gas analysis of reactants and products. Freysz et al. [126] have used two complementary techniques for gas products analysis: FTIR using a gas microcell (multireflexion cell with an inner volume of 88 l) and mass spectrometry. The dead volumes are minimized in the whole system in order to allow time-resolved analysis in particular inside the IR reactor cell [132]. The time resolution of both surface analysis and gas analysis is around 1–2 s. This methodology was applied to study the reduction of NO by CO on a silica-supported platinum catalyst by the addition of very small CO pulses (20 l) in a continuous NO flow [126]. The simultaneous qualitative and quantitative analysis and the correlation of both surface species and gas phase evolution during the transient-pulse catalytic reaction indicated that a direct reaction between CO and NO (adsorbed or not) must be rejected. When a pulse of CO is sent, two distinct successive periods are observed (Figure 4.11). The first period concerns the reduction of Pt by CO. The diminution of the NO/Pt concentration at the beginning of the pulse does not correspond to a significant reduction of NO (no significant production of N2 /N2 O), but to a desorption of nitrogen monoxide: the amount of observed gas NO increases during this period. The adsorbed CO displaces NO from the surface but does not react with it. The concomitant production of CO2 can then only be justified by a reaction between CO and the initially adsorbed oxygen: CO reduces the Injection loop

Six-port valve Continuous flow

IR cell with catalyst disk

Gas analysers (MS, IRgas, GC,…)

Figure 4.10. Scheme of the experimental system used for transient pulse reaction [126].

122

Past and Present in DeNOx Catalysis (a)

(b)

Tim e

0.02 0.00

N2O CO 2500

2000

1500

0.04 0.02 0.00

1000

NO/Pt 110 100 90 80 70 60 50 40 30 20 10

( s)

( s)

110 100 90 80 70 60 50 40 30 20 10

CO/Ptox CO/Ptred

Tim e

CO2 Absorbance

Absorbance

CO2 NO NO2(traces)

2200 2000 1500 1600

1400

Wavenumbers (cm–1)

Wavenumbers (cm–1)

(c) 10

CO2

CO

Arbitrary units

0

Gas

1

0.2

N2O 0.0 0.5

NOx 0.0

NO/Pt 0 4 2

Surface

2

CO/Ptred 20

40

60

80

100

Time (s)

Figure 4.11. Pulse of 20 l of CO in a 1000-ppm NO in He continuous flow on a Pt/SiO2 catalyst; T = 498 K. (a) IR spectra of the gas phase (one spectrum per 2 s). (b) IR spectra of the surface species (one spectrum per 2 s). (c) Correlation between the surface species and the gas phase IR analysis [126].

platinum surface. The second period corresponds to the reduction of NO. At the end of the pulse, CO progressively disappears from the surface, whereas NO readsorbs on the Pt sites. The important NO consumption does not correspond only to this re-adsorption, but mainly to an important N2 /N2 O production (i.e. to NOx reduction). Remaining COlin /Ptred undergoes a total oxidation into CO2 , because no more CO is observed in the gas phase and no more carbonaceous species are observed on the surface (if we except the initial COlin /Ptox spectator species). This second CO2 production corresponds to the N2 /N2 O formation, and NO is the only possible oxidant: COlin /Ptred is the origin of NO reduction. At the end of the pulse, the NO/Pt concentration is higher than at the beginning. In fact,

General Features of In Situ and Operando Spectroscopic Investigation

123

adsorbed nitrosyl species now occupy Pt sites, which were unavailable before the pulse as other species were already adsorbed (Oads or nitrates). The surface then slowly returns to its initial equilibrium. Chuang and coworkers [128] also used the transient IR technique coupled with site poisoning and CO site promotion. They aimed at distinguishing between active and spectator species among the IR-observed adsorbates during NO decomposition over Cu-ZSM-5. They evaluated the variation of adsorbate and product concentration with time during the CO pulse (added into NO in a He flow) in order to determine the NO decomposition reaction mechanisms. The O2 formation/desorption pathway proceeds via two routes: (1) autoreduction of Cu2+ to Cu+ followed by desorption of O2 and (2) decomposition of Cu2+ (NO3 − 2 to N2 , O2 , N2 O and NO. According to the authors, Cu+ (NO) and Cu2+ (NO3 − 2 act as active adsorbates while NO+ seems to behave as a spectator adsorbate during the NO decomposition to N2 and O2 . SiH4 and H2 O severely poison O2 formation but do not inhibit NO dissociation and N2 formation. Inhibition of adsorbed O migration is the dominant deactivation pathway for O2 formation/desorption rather than the blockage of Cu+ and Cu2+ sites. Site promotion by CO addition enhances NO conversion and O2 formation, suggesting the ability of a small amount of CO to promote NO decomposition. Bion and coworkers [133] studied the NOx SCR in excess of oxygen using ethanol as reducing agent on silver/alumina. They underlined the role of isocyanate species using the pulse technique. The comparison of the model CO + NO reaction on bare alumina and on silver/alumina allowed them to give evidence of a strong relationship between the presence of silver and isocyanate (NCO) species on a catalyst. They further investigated in more details the NCO groups (formation, localization and reactivity) in order to propose a pathway for the NOx SCR into N2 . Three elemental sequences are suggested: (1) the formation of silver cyanide and its transformation into Al3+ NCO (Figure 4.12), (2) the isocyanates hydrolysis into ammonia and (3) the reaction of the latter species with NO in the presence of oxygen giving rise to nitrogen.

3.2.2. Transient reaction technique

N16O

C18O 16O 18

18 N 16O OC

Ag O

N C O

Haneda et al. [134,135] studied the formation and reaction of adsorbed species in NO reduction by propene over Ga2 O3 −Al2 O3 . IR transient reaction technique was employed to examine the reactivity and dynamic behaviour of surface species. The catalyst was first exposed to either C3 H6 /O2 /Ar or NO/O2 /Ar at 623 K for a long time to form and accumulate the surface species. The catalyst was further purged with pure Ar and the reaction gas then switched to various gas mixtures. Changes in the intensity of IR bands were measured with time on stream. The main surface species detected by IR during

O C

Al

Al O

Al O

O

Al

Al O

N

O

O Al

Al O

O

O

Al

Al O

O

Al O

O

Figure 4.12. Formation mechanism of NCO groups in a CO + NO reaction on Ag/Al2 O3 catalysts proposed by [133].

124

Past and Present in DeNOx Catalysis

the steady-state reaction under NO/C3 H6 /O2 /Ar flow were nitrate, acetate, formate, isocyanate and cyanide (−CN). Transient reactions revealed that the selective reduction of NO with propene in the used temperature range could start with the formation of adsorbed NOx species (NO2 − and NO3 − ) resulting from NO oxidation. The subsequent reaction of these adsorbed NOx species with propene and/or propene-derived species then generate cyanide (−CN) and isocyanate (−NCO) species. The −NCO species are then hydrolysed to −NH compounds when reacting with the water present in the stream and these −NH compounds finally react with adsorbed NOx species and/or NO2 to produce N2 . Acetate and formate species formed by mild oxidation of C3 H6 are reported to be spectator species in the reaction. Prinetto et al. studied the storage of NO2 and NO/O2 in the presence and in the absence of CO2 at 350 C on Pt-Ba/Al2 O3 lean NOx -trap catalysts and on the reference Ba/Al2 O3 system, by coupling in situ FTIR spectroscopy and transient response method. Experiments were performed by admitting NO2 /CO2 or NO/O2 /CO2 mixture (CO2 /NOx ratios in the range 1–7) or, in alternative, by admitting NO2 or NO/O2 on catalysts previously saturated by flowing CO2 . The picture obtained upon admission of NO2 /CO2 mixtures or of NO2 on the catalyst surface previously saturated with CO2 strictly parallels the results collected in the case of NO2 adsorption in the absence of CO2 . The collected results indicate that, also in the presence of high amounts of CO2 , Pt−Ba/Al2 O3 catalysts are able to efficiently perform NOx storage through a pathway that involves the formation of surface nitrate species (nitrate route) according to a disproportion reaction. In the case of the storage of NO and O2 , two pathways operating simultaneously are proposed: the ‘nitrite route’, that involves the initial formation of surface nitrite and their subsequent evolution to nitrates, and the ‘nitrate route’, that involves NO oxidation to NO2 over Pt and its subsequent adsorption on Ba phase in the form of nitrates. In the presence of CO2 , if ‘nitrite route’ is in some extent inhibited due to the competition between NO and CO2 for the surface oxygen sites of the Ba phase, ‘nitrate route’ can proceed as in the absence of CO2 . It turns out that also in this case Pt−Ba/Al2 O3 catalysts are able to efficiently perform NOx storage [136].

3.2.3. Stopped-flow technique Another way to work in transient conditions is to stop suddenly (or conversely to instantaneously introduce) one of the reactants, in order to destabilize the system and to enhance the concentration of labile species. With this method, for example, Poignant et al. studied the DeNOx reaction mechanism on a H−Cu-ZSM-5 catalyst, using propane or propene as reducing agents. The introduction of 2000 ppm of hydrocarbon in a flow of NO (2000 ppm) + 5% O2 allowed to evidence the formation of acrylonitrile, which behaved as an intermediate. Its reactivity with NO+ species constituted a fundamental point to describe a detailed SCR mechanism for NO removal on zeolitic compounds [137]. The main inconvenient of this methodology is that the results cannot be considered stricto sensu as obtained in operando conditions, because the system was perturbed from the steady state to reveal hidden species. It could be even hypothesized that such compounds are uniquely due to the particular test conditions and not to the real reaction pathway. A method to discard such kind of criticism is to maintain the chemical steady state of the reaction, while introducing a perturbation via a sudden exchange of one

General Features of In Situ and Operando Spectroscopic Investigation

125

of the reactant by an isotopic labelled molecule. This is the principle of the steadystate isotopic transient kinetic analysis (SSITKA) technique, which has already given interesting results for determining intermediate species in the case of water–gas-shift reactions [138]. Curiously, in our knowledge, this method has not yet been applied to the study of DeNOx reaction, probably due to the high cost of isotopic products when used under continuous flow. In the future, a wider application of SSITKA technique to NOx abatement domain has to be encouraged to resolve open discussions on different reaction mechanisms.

3.3. IR temperature-programmed reaction Temperature-programmed reduction and oxidation (TPR/TPO) are widespread techniques for catalysts characterization. Coupling TPR/TPO analysis with IR spectroscopy allows the identification of species adsorbed on the catalyst surface. The operating mode generally used is the following: the sample is introduced into a high-temperature IR cell and pretreated at high temperature in an inert gas flow (with or without oxygen) to remove adsorbed species (water, carbonates, nitrates, etc.). The sample is then cooled to the desired temperature (generally room temperature) and the inert gas flow is switched to reduction or oxidation flow. Generally hydrogen or oxygen are, respectively, used but other agents might be used (CO, HC, NH3 , NO, NO2 , etc.). The sample is finally progressively heated under the chosen atmosphere. IR spectra of the surface are collected upon increasing temperature (a rate of 10 K/min is normally used) and simultaneously the effluent gases from the IR cell are monitored with a MS. Complementary gas analysers might be used (IR gas analyser, GC, etc.). Long and Yang [139] studied the SCR of NO to N2 with NH3 on Fe-exchanged ZSM-5 using TPSR technique. They concluded that the SCR reaction requires two kinds of sites: the Brønsted acid sites for ammonia adsorption and the metal ion sites (i.e. Fe3+ ions) for NO oxidation to NO2 . The proposed reaction mechanism is reported on Figure 4.13. According to the authors, the NO reduction involves −1 NO2 and NO2 (NH+ 4 2 as intermediates detected by an IR band near 1602 cm . An alternative to the NH3 -SCR is the NOx reduction by hydrocarbons over metal exchanged zeolites (Cu, Co, Fe, Ni, etc.). Different reaction mechanisms have been proposed in the literature and several intermediates are suggested to take part in the whole NOx reduction process leading to the formation of nitrogen. Among all the described intermediates species, nitrites [140], nitrates [141–146], isocyanates [146–153], cyanides [140,154], ammonia [146,149,151], nitro [146,148,155],

2 NH3

2 H+

+

2 NH4

Fast reaction +

NO2(NH4)2 NO

1/2 O2 Fe3+

NO

…

NO2

Slow reaction

2 N2 3 H2O 2 H+

Fast reaction

Figure 4.13. Reaction scheme of SCR of NO with ammonia on Fe-ZSM-5 [139].

126

Past and Present in DeNOx Catalysis

organic nitro [146,149,156], nitrito [146,150,151,157] and nitroso [146,158,159] compounds are the most frequent ones. A lot of these species are reported in the mechanism proposed by Mosqueda et al. for the reduction of NO with propene and propane on Ni-exchanged mordenite [146]. The study was investigated by mean of in situ infrared temperature-programmed reduction. The IR TPR experiments permitted the observation of the overall intermediate species. The IR bands assignment is based on a careful literature review but, according to the authors, the complexity of the spectra sometimes obtained made some assignment occasionally speculative. The first step of the proposed mechanism consists in the formation of nitrates and nitro compounds on the surface of the catalyst from the interaction of NO and O2 . Their transformation into organic nitrito and nitro compounds then follows in the presence of hydrocarbons. Whatever the nature of the reducing agent (C3 H8 or C3 H6  the same principal reaction mechanism is concluded to occur because C3 H8 dehydrogenates into C3 H6 . Nitrosonium ions are also formed and react with hydrocarbons to form nitroso compounds. These intermediates are however not observed during TPR because they are unstable species. Organic nitro, nitrito and nitroso compounds are further transformed to surface oximes. These species rearrange subsequently to isocyanates. Finally, isocyanates react with NO2 in gas phase and give rise to N2 . Three main isocyanate species are observed: isocyanates associated with Ni (∼2230 and ∼2195 cm−1 ) and isocyanate with Al (∼2270 cm−1 ). However, the first one is only observed when the reaction temperature increases and alkanes and alkenes are absent from the gas flow. In the presence of reactive gases, its absence is attributed to its high reactivity leading to immediate N2 formation. The other isocyanates species are reported to be less efficient reaction intermediates. According to the authors, oxygenated carbon species, e.g. formaldehyde, formate, acetate, carboxylate and acrolein might be involved in the formation of isocyanates. Finally, the presence of water leads to the readily isocyanates hydrolysis giving rise to amines and/or ammonia, which actually reduce NO2 to N2 . The TPSR technique has also been used by Konduru and Chuang [160] in order to investigate N2 O and NO decomposition pathways on Cu-ZSM-5. The infrared monitoring of the adsorbed species during the sample heating under NO showed that the Cu+ (NO) intensity parallels the rate of N2 O formation. This TPSR result allowed the authors to suggest that NO adsorbed on Cu+ acts as precursor for N2 O formation. Shen and Kawi studied the reaction mechanism for the SCR of nitric oxide (NO) over Pt/Si-MCM-41 catalyst by XPS and in situ FTIR spectroscopies as well as TPR. The XPS results indicated that some Pt species are still reduced during the reaction of NO-SCR in the presence of excess oxygen. FTIR results indicated that CO is one of the reaction intermediates taking part in the selective reduction of NO. Although carboxylates species are believed not to participate in the HC-SCR reaction, CO was suggested to facilitate the reduction of NO: CO might react with NO to form NCO, a possible intermediate for the selective reduction of NOx . Even if the authors proposed a pathway for the catalytic process, the reaction mechanism for the whole NOx elimination over Pt/Si-MCM-41 is believed to be very complicated and still unclear [161]. Szanyi et al., using again in situ IR-TPD coupled skills, studied the effect of acid sites on the catalytic activities of a series of H+ -modified Na-Y zeolites in the nonthermal plasma assisted NOx reduction reaction using a simulated diesel engine exhaust gas mixture. The acid sites were formed by NH+ 4 ion exchange and subsequent heat treatment of a NaY zeolite. The catalytic activities of these H+ - modified NaY zeolites

General Features of In Situ and Operando Spectroscopic Investigation

127

significantly increased with the number of acid sites. This NOx conversion increase was correlated with the decrease in the amount of unreacted NO2 . The increase in the number of acid sites did not change the NO level, it stayed constant. TPD following NO2 adsorption showed the appearance of a high-temperature desorption peak at 453 K in addition to the main desorption feature of 343 K observed for the base Na-Y. The results of both the IR and TPD experiments revealed the formation of crotonaldehyde, resulting from condensation reaction of adsorbed acetaldehyde. Strong adsorptions of both NOx and hydrocarbon species enhanced by the introduction of Brønsted acidic sites into the Na-Y catalyst were proposed to be responsible for the higher catalytic activity of the zeolites in the protonated form in comparison to the base material: the increased stabilities of both the reductant and the NOx result in higher concentrations of reactants at the actual reaction temperature, thereby resulting in higher rates of NOx reduction [162]. Using the same methodology, completed by TEM and energy dispersive X-ray spectroscopy, this team deeply investigated NOx storage reduction (NSR) Ba-oxide-based catalysts, looking for the modification of the original BaO/Al2 O3 formulation after thermal ageing. Interestingly, being able to distinguish between surface and bulk type phases, they could reveal the Ba limit concentration and the limit treatment temperature for the unwished BaAl2 O4 phase formation [163]. The amount of papers dealing with TPSR technique used to complete studies based on infrared experiments is increasing [164–166] and this methodology will probably be further developed.

4. LET’S PLAY NOW! (THE OPERANDO APPROACH) The spectroscopic study of a catalyst under duty is related with the concept of making a movie of the chemical process taking place during the reaction. As well pointed out by Weckhuysen and coworkers [167], this implies the ability to master time resolution, space resolution and working condition parameters during a catalytic reaction. It means that to really see a chemical step during its development, it is necessary (1) to have a time-scale adapted to the speed of the chemical reaction; (2) to be able to discriminate the contribution of the geometric parameters and inhomogeneities in a real, tailored catalyst; (3) to reproduce as close as possible the real working conditions (temperature, pressure, reacting agents composition, space velocity, etc). This is actually far to be achieved and one single technique is not able to furnish all the information necessary to draw the entire catalytic process. For such a reason more and more scientists are operating in spectroscopic skills coupling and development, in order to get closer to the real conditions when observing a catalytic reaction [26,167,168]. Meanwhile, it must be underlined that the borderline between in situ and operando conditions is hard to be defined and it can be considered as a concept continuously evolving with the progress of the techniques. Few years ago, I.E. Wachs [169] and H. Topsøe [170] reported a state of the art of different operando spectroscopic techniques. We will try to give here some examples of operando investigations in the domain of NOx removal choosing among the reports which represent, in our opinion, genuine efforts to reproduce real catalytic conditions in a spectroscopic reactor-cell.

128

Past and Present in DeNOx Catalysis

4.1. Operando studies for NOx removal from stationary sources DeNOx of waste gases from stationary sources can be efficiently achieved by using the so-called SCR process [171]. Nowadays, the large majority of the current industrial NOx removal is carried out by this technology, using ammonia injected in the waste gases as the reducing agent, due to the high efficiency of the reaction stoichiometry. Nevertheless, this process is not exempt from problems concerning ammonia slip and reaction temperature control. Alternatively, hydrocarbons can also be used as reducing agents (and they are preferred, when possible), but applying a technology, which is closer to that of vehicles exhaust control [172]. Therefore, a big research effort is being made to understand all the reaction parameters in such kind of heterogeneous catalysis. Montanari et al., for example, studied a Co−H-MFI sample through FT-IR spectroscopy of in situ adsorption and coadsorption of probe molecules [o-toluonitrile (oTN), CO and NO] and CH4 -SCR process tests under IR operando conditions. The oTN adsorption and the oTN and NO coadsorption showed that both Co2+ and Co3+ species are present on the catalyst surface. Co3+ species are located inside the zeolitic channels while Co2+ ions are distributed both at the external and at the internal surfaces. The operando study showed the activity of Co3+ sites in the reaction. The existence of three parallel reactions, CH4 -SCR, CH4 total oxidation and NO to NO2 oxidation, was also confirmed. Isocyanate species and nitrate-like species appear to be intermediates of CH4 SCR and NO oxidation, respectively. A mechanism for CH4 -SCR has been proposed. On the contrary, Co2+ substitutional sites, very evident and predominant in the catalyst, which are very hardly reducible, seemed not to play a key role in the SCR process [173]. The reaction mechanism of SCR of NOx with decane on acid and iron-exchanged MFI-type zeolite was also investigated by operando FTIR spectroscopy and a special reactor cell enabling the use of water-containing gas mixtures. Brosius et al. found that water has a dramatic influence on the reaction pathway, while the formation of organic nitro and nitrite compounds does not proceed via chemisorbed states of NOx , as observed in SCR with dry gases [174]. Romero Sarria et al., using an IR operando system coupled with MS, gave important conclusions about the catalytic behaviours of zeolitic and GAPON compounds applied to ammonia SCR. The parallel following of both the gas phase and the surface during the reaction in real conditions allowed establishing all the parameters influencing the catalyst functionalities, giving therefore the possibility to conceive the necessary modifications in the formulation of the catalyst to improve its performances. In particular, they demonstrated the feasibility of an efficient and very selective SCR of NOx on zeolitic compounds presaturated with ammonium and/or ammonia. Additionally, they pointed out the necessity of the presence of NO2 in the reaction stream to oxidize the adsorbed NH3 species, proposing a way to obtain an optimal NO/NO2 ratio, allowing to reach a satisfactory SCR efficiency at low temperatures [175]. Groothaert et al., using operando UV–vis spectroscopy combined with online GC analysis [176] and operando X-ray absorption fine structure (XAFS) [177], presented the first experimental evidence for the formation of the bis( -oxo)dicopper core in CuZSM-5 and for its key role of intermediate in the sustained high activity of Cu-ZSM-5 in the direct decomposition of NO into N2 and O2 . In particular, monitoring the catalytic conversion of NO and N2 O above 673 K, they found that the bis( -oxo)dicopper core is formed by the O abstraction of the intermediate N2 O (Figure 4.14). Subsequently,

General Features of In Situ and Operando Spectroscopic Investigation

129

(b)

(a)

1 N2 O2 N2O

0.4

Kubelka-Munk

Reactor outlet composition (mol%)

0.5

0.3 0.2

.8 773 K .6 .4 .2

0.1 0

523 K

0.0 523

573

623

673

T [K]

723

773

35000

30000

25000

20000

15000

Wavenumber (cm–1)

Figure 4.14. (a) Catalytic activity of Cu-ZSM-5 (Si/Al = 31; Cu/Al = 0.58) for the decomposition of NO (1 mol% in He; 900 h−1 GHSV) as function of the temperature. (b) Corresponding operando UV–vis spectra at temperature intervals of 50 K [176].

bis( -oxo)dicopper fulfils the roles of O2 production and release, guaranteeing the selfreduction of the catalytic site. Studying the NO decomposition as a function of the O2 content in the feed, they proposed that O2 release from bis( -oxo)dicopper is rate limiting in the NO decomposition at 773 K. A reaction cycle for the NO decomposition was proposed, involving the intermediates: Cu+    Cu+ , [CuOCu]2+ and [Cu2  -O)2 ]2+ . The use of coupled spectroscopic techniques is in fact a powerful method to obtain information on both the active centres of the catalytic material and the reaction intermediates: Schwidder et al., for example, investigated by XAFS, EPR and UV–vis spectroscopy Fe-ZSM-5 catalysts prepared via chemical vapour deposition (CVD) of FeCl3 and via mechanochemical treatment of a FeCl3 /H-ZSM-5 mixture in the SCR of NO with ammonia and with different hydrocarbons. In the material prepared via the CVD route, extensive clustering was found by EPR and UV–vis, which was not detected by XAFS, probably due to a disordered cluster structure. In the mechanochemically prepared material, the predominant iron species were mononuclear Fe3+ ions although a clustered minority phase was detected by UV–vis and EPR. In the SCR with hydrocarbons (1000 ppm NO, 1000 ppm reductant, 2% O2 at 30 000 h−1 , the catalyst prepared by the mechanochemical route was superior to the CVD-derived catalyst at almost all temperatures while it was inferior with the ammonia reductant (1000 ppm NO, 1000 ppm NH3 , 2% O2 at 750 000 h−1 . However, in terms of SCR rates related to Fe content, the mechanochemically prepared catalyst was superior in all cases. It was suggested that hydrocarbon SCR was favoured by atomic or oligomeric dispersion of the Fe3+ ions, while SCR with ammonia appeared to be catalysed also by FeOx aggregates [178]. An interesting result should be highlighted reporting the work of Lercher et al. concerning the simulation of the catalytic process intended to reduce NO and CO emitted during the regeneration of spent fluid catalytic cracking catalysts. It was found that the use of Ce-promoted alumina carrier resulted in the formation of larger supported Ir crystallites. Coexistence of reduced and oxidized Ir sites seemed important for the

130

Past and Present in DeNOx Catalysis

activity of the Ir-based catalysts. IR studies suggested that NO reduction with CO proceeds via the dissociative adsorption of NO on the noble metal. NCO species are formed on Ir and migrate subsequently to the alumina support. Isocyanates on the metal or the support react with NO leading to the formation of N2 and CO2 . In the presence of Ce, an additional path of NO reduction via NO2 formation was judged possible, while dissociative adsorption of CO on the metal is additionally enhanced. The presence of Ce modifies the type of active Ir phase and the reaction mechanism. By this means, CO oxidation is increased, while an alternative reaction route retains DeNOx performance [179].

4.2. Operando studies for NOx removal from automotive sources Delahay et al., using operando DRIFT analysis coupled with catalytic tests for diesel engines waste treatment, found that the oxidation of NO in NO2 and the formation of nitrile intermediates are key steps of the SCR of NO by n-decane on Fe-ZSM-5 in general agreement with most of previous studies on the SCR of NO by hydrocarbons on TMIzeolites. Additionally, they pointed out that the deactivation is due to polymerization of conjugated nitrile intermediates promoted by the strong Fe Lewis sites. By the way, they remarked that the DRIFT analysis of the HC-SCR with long chain hydrocarbons on Fe-ZSM-5 leads to higher conversions than steady state experiments in classical catalytic tests, due to shorter exposure time to SCR mixture or/and strong adsorption of n-decane, showing once more that the experimental conditions (i.e. the level of operando compatibility with the real conditions) are crucial for the right interpretation of the results [180]. This latter point was stressed by some of us in a recent report studying NOx storage and reduction on commercial LSR (lean storage-reduction) catalysts, in order to catch valuable information about the behaviour of typical NOx storage materials in real application conditions. Nature, thermal stability and relative amounts of the surface species formed on a commercial catalyst upon NO and O2 adsorption in the presence and in the absence of water were analysed using a novel system consisting of a quartz infrared reactor. Operando IR plus MS measurements showed that carbonates present in the fresh catalyst are removed by replacement with barium nitrate species after the first nitration of the material. Nitrate species coordinated to different barium sites are the predominant surface species under dry and wet conditions. The difference in the species stabilities suggested that barium sites possess different basicity and, therefore, that they are able to stabilize nitrates at different temperatures. At temperatures below 523 K, nitrite species were observed. The presence of water at mild temperatures in the reactant flow makes unavailable for NOx adsorption the alumina sites [181]. A deeper investigation on the NOx storage mechanism was undertaken on a newly developed NOx -trap Pt/K/Mn/Al2 O3 −CeO2 catalyst. Using the FTIR operando technique, the synchronous analysis of gaseous and adsorbed phases all along the experiments gave highly valuable information in order to establish consecutive steps of NOx adsorption on this material. It has, therefore, been established that this reaction proceeds initially through NO oxidation, NO2 so produced being subsequently adsorbed on potassium sites in the form of nitrite species. The last step of this mechanism is thus the oxidation of nitrites to nitrates, probably involving NO2 oxidizing molecules. Two

General Features of In Situ and Operando Spectroscopic Investigation

131

different periods have also been determined for the NOx storage process, and linked to the progressive adsorption of NOx all along the catalytic reactor, as well as to the presence of two types of adsorbing sites, i.e. surface and bulk potassium sites. The first period is characterized by full NOx adsorption, which corresponds to surface nitration from the front to the rear part of the catalyst, and which is accompanied by carbonates decomposition. During the second period, NO and NO2 appear in the exhaust stream to reach a steady level, NO2 diffusion into the bulk being the main phenomenon that occurs. These results are precious for the engineering of the vehicle waste gas treatment system! (see Figure 4.15). Note that for commercial applications, lean exposure should be interrupted at the end of the first period, since NOx released will not be permitted. Furthermore, a strong thermal dependency has been observed for the NOx storage process, which has been primarily correlated to the NO oxidation rate. In fact, it has been demonstrated that the higher the NO2 concentration in the lean flow, the stronger the progressive adsorption of NOx , since in this case NOx gaseous species have to diffuse to a limited extent to be completely stored, this being also favourable for nitrite oxidation. On the other hand, when NO2 concentration is low, it infers that NOx have to diffuse farther on the catalytic surface to be fully consumed, the full NOx storage period being thus shorter at lower than at higher temperatures, and nitrite oxidation rate being also lower. Finally, the excellent properties of the Pt/K/Mn/Al2 O3 −CeO2 catalyst towards its long full NOx storage period and the high level of potassium sites implicated in NOx adsorption have been related to the particular structure of this material (a Hollandite structure with potassium atoms in the tunnels), which slow down the diffusion of NOx species, and reinforce their progressive storage [182]. Lesage et al. also investigated the NOx reduction process after storage on a classical Ba-based catalyst, using a IR + MS operando set-up [132]. The alternate exposure to Initial stages of NOx storage NO + O2

NO2

NO + O2 NO2–

–

NO3

NO2

NO2

NO2

On the front part of the catalyst, NO2 formed is adsorbed as a nitrite, whereas NO diffuse farther on an area where the equilibrium between NO and NO2 is shifted toward oxidation. The lower the reaction temperature, the slower is the NO oxidation, and farther it has to diffuse to be fully oxidized and adsorbed.

Late stages of NOx storage NO2–

NO3–

NO + O2 NO3–

NO3–

NO, NO2 NO2–

The catalytic surface being saturated to a large extent by NO2 adsorbed species. NO and NO2 are detected in the exhaust stream. The higher the reaction temperature, the more efficient the progressive adsorption of NO2 and the oxidation of nitrites into nitrates.

Figure 4.15. Schematic representation of the progressive adsorption of NOx on a catalytic bed in a reactor flow [182].

132

Past and Present in DeNOx Catalysis

rich and lean streams close to the real reaction mixture, allowed the identification of some steps in the reduction pathway, active sites, intermediate species and by-products for NOx -trap reaction. In particular, they differentiated the role of the support and the noble metal in the mechanism. Moreover they pointed out the reactivity of isocyanate adspecies and ammonia among the detected species, as well as their connection with the N2 selectivity. Interesting results concerning sulphur poisoning and regeneration of a NSR catalysts were found by Anderson et al. using FTIR or combined XRD/XAS, both with simultaneous gas phase analysis by MS and chemiluminescence. Pt/Ba/Al2 O3 catalysts pretreated in SO2 showed significant, but not total, loss of NOx storage capacity with barium sulphate being formed if exposure occurs in the presence of air. Bulk sulphate formation was not so readily formed when SO2 treatment took place competitively (in the presence of NOx , although again the result was a suppression of NOx adsorption. Even when SOx uptake was restricted to the baria surface, displacement and suppression of carbonate formation occurred. It was ascertained that Pt−S species are not favoured when oxidants are present in the gas stream and that bulk phase Pt reduction takes place before propene activation can occur [183]. It was also ascertained that in the presence of Pt the order of regeneration efficiency was H2 > CO > propene, while in contrast hydrogen was least effective in the absence of Pt. The presence of oxygen during the rich regeneration period appeared to be significant as this allows formation of water, which has the potential to hydrolyse S residues on the noble metal. The use of hydrogen at low temperatures to release and reduce stored NOx has the potential to limit sulphate build-up and thus limit the extent to which deactivation due to growth of barium sulphate particles takes place and to avoid thermal degradation of the catalyst, which accompanies high-temperature desulphation [184].

4.3. Checkmate to the intermediates As pointed out above, the detection of intermediates is the crucial step when determining a reaction mechanism by spectroscopic techniques. This is due to both the low concentration and the short lifetime of such species, intrinsically depending on their quality of intermediates, i.e. species representing the transition state of the reactants transforming into products. Nevertheless, when such compounds precede a slow step in the reaction pathway, they can cumulate on the surface of the catalytic material and their residential time increases to a level allowing their detection at the second time scale. In such a case, intermediates are mixed with spectator species (i.e. species being present during the reaction without playing a major role in the principal reaction path); consequently, they must be discriminated by several methods. The main point to be verified is that the compound claimed as an intermediate is following the kinetics of the main reaction mechanism. This can be obtained comparing the reaction itself under the real stream with the reactivity of the intermediate species taken as main reacting agent. We should stress that in such a case the experiment is realized in conditions, which cannot be considered as strictly operando, since the reacting flow presents no longer the composition expected in real conditions. Different examples of intermediate investigation have been cited above. In all the cases the most debated and still undetermined point is the formation of the N−N bond leading

General Features of In Situ and Operando Spectroscopic Investigation

133

to the final step of the NOx SCR. The majority of the scientists agrees on the hypothesis that this should be achieved via the formation of an intermediate cyano-compound. We should recall that to verify this point Szanyi et al. realized the adsorption of HCN and the reaction of HCN with NO2 over Na-, and Ba-Y, FAU zeolite catalysts using in situ FTIR and TPD/TPR spectroscopies. The results of this investigation strongly suggest that HCN/CN can be an important intermediate in the overall NOx reduction on these zeolite catalysts. Adsorbed CN- and NCO- ionic species formed in the HCN + NO2 reaction are believed to be the important surface intermediates. Their reaction with ionic NOx species (in particular with NO+ ) can lead to the formation of the N−N bond [185]. On the contrary, the experimental data did not seem to support a mechanism that would involve the formation of NHx -containing reaction intermediates for the N−N bond formation. However, under significantly different reaction conditions, their participation in the overall N2 formation process cannot be ruled out [186]. Again, as already highlighted, the detection of the intermediates being mainly realized in in situ conditions more than in operando conditions, the doubt persist about the extrapolation of such kind of results for the description of the ‘real’ reaction. Therefore, the only way to directly observe an intermediate using a spectroscopic method is to enhance the time resolved analysis properties of the technique. The time resolution of modern Fourier transform IR spectrometers is nearly 0.01 s, insufficient to detect the majority of the intermediate species. Step scan is a technique giving FTIR spectroscopy unprecedented time resolution, up to a few nanoseconds. The only condition is that the observed perturbation is easily and indefinitely reproducible, which is not the case, generally speaking, in heterogeneous catalysis. Recently, Seguin et al. have shown that combining a femtosecond pulsed laser (used to trigger the reaction on the catalyst) and step scan IR spectroscopy at 33 ns (Figure 4.16), the detection of very short lived intermediates is possible in catalysis domains as general as DeNOx chemistry on supported metals on alumina [187], directly observing a bridged cyano-intermediated, which had been only postulated in a previous study [133].

IR KBr window 1000

Vacuum line

0

Sample Oven

800 nm 4 mJ

–1000 0 100 200 300 400

–2000

Dichroic mirror

int

4

Optical filter

µse8 c

Po

6

IR MCT detector

10

500 12

Figure 4.16. Experimental setting of the combined femtosecond pulsed laser and step scan IR spectrometer (left) and modifications of the infrared interferogram after the laser pulse (right) [187].

134

Past and Present in DeNOx Catalysis

5. CONCLUSIONS Actually, the operando approach is more and more required to get closer to the reaction phenomena when they are taking place. This implies a continuous development of the in situ spectroscopic techniques (especially infrared and Raman spectroscopies), often coupled to have complementary information in a limited lapse of time. In the domain of DeNOx processes, the determination of the reaction mechanisms is fundamental to achieve a correct formulation and tailoring of effective catalysts. In this view, a classical kinetic approach maintains its fundamental role to establish the reaction pathway, helping to distinguish, among the observed species, those being really important for the chemistry of the system. Nowadays, the massive use of spectroscopic operando methods of investigation, possibly coupled with micro-kinetics calculations, seems the only possible solution to overcome the difficulties connected with the chemistry of NOx removal. In this frame, time resolution remains a crucial point to obtain a realistic following of the chemical reaction and to evidence intermediate species.

REFERENCES [1] Meunier, F. and Daturi, M. (2006) Recent developments in in situ and operando spectroscopy for catalysis – Preface, Catal. Today, 113, 1. [2] Birgersson, H., Boutonnet, M., Klingstedt, F. et al. (2006) An investigation of a new regeneration method of commercial aged three-way catalysts, Appl. Catal. B Environ., 65, 93. [3] Hungría, A.B., Fernández-García, M., Anderson, J.A. et al. (2005) The effect of Ni in Pd−Ni/(Ce,Zr)Ox /Al2 O3 catalysts used for stoichiometric CO and NO elimination. Part 2: Catalytic activity and in situ spectroscopic studies, J. Catal., 235, 262. [4] Larese, C., Lopez Granados, M., Mariscal, R. et al. (2005) The effect of calcination temperature on the oxygen storage and release properties of CeO2 and Ce−Zr−O metal oxides modified by phosphorus incorporation, Appl. Catal. B Environ., 59, 13. [5] Yamazaki, K., Takahashi, N., Shinjoh, H. et al. (2004) The performance of NOx storagereduction catalyst containing Fe-compound after thermal aging, Appl. Catal. B Environ., 53, 1. [6] Larese, C., Cabello Galisteo, F., López Granados, M. et al. (2004) Effects of calcination temperature on the stability of CePO4 detected in vehicle-aged commercial three-way catalysts, Appl. Catal. B Environ., 48, 113. [7] Suhonen, S., Polvinen, R., Valden, M. et al. (2002) Surface oxides on supported Rh particles: thermal decomposition of Rh oxide under high vacuum conditions, Appl. Surf. Sci., 200, 48. [8] Kobayashi, T., Yamada, T. and Kayano, K. (2001) Effect of basic metal additives on NOx reduction property of Pd-based three-way catalyst, Appl. Catal. B Environ., 30, 287. [9] Suárez, S., Martín, J.A., Yates, M. et al. (2005) N2 O formation in the selective catalytic reduction of NOx with NH3 at low temperature on CuO-supported monolithic catalysts, J. Catal., 229, 227. [10] Twagirashema, I., Engelmann-Pirez, M., Frere, M. et al. (2007) An in situ study of the NO + H2 + O2 reaction on Pd/LaCoO3 based catalysts, Catal. Today, 119, 100. [11] Granger, P., Delannoy, L., Lecomte, J.J. et al. (2002) Kinetics of the CO + NO Reaction over Bimetallic Platinum–Rhodium on Alumina: Effect of Ceria Incorporation into Noble Metals, J. Catal., 207, 202. [12] Requejo, F.G., Hebenstreit, E.L.D., Ogletreeb, D.F. et al. (2004) An in situ XPS study of site competition between CO and NO on Rh(111) in equilibrium with the gas phase, J. Catal., 226, 83.

General Features of In Situ and Operando Spectroscopic Investigation

135

[13] Rokosz, M.J., Chen, A.E., Lowe-Ma, C.K. et al. (2001) Characterization of phosphoruspoisoned automotive exhaust catalysts, Appl. Catal. B Environ., 33, 205. [14] Kim, D.H., Woo, S.I. and Yang, O.-B. (2000) Effect of pH in a sol-gel synthesis on the physicochemical properties of Pd–alumina three-way catalyst, Appl. Catal. B Environ., 26, 285. [15] Hunger, M. and Horvath, T.J. (1995) A new mas NMR probe for in-situ investigations of hydrocarbon conversion on solid catalysts under continuous-flow conditions, Chem. Soc. Chem. Comm., 14, 1423. [16] Sundaramurthy, V., Cognec, J.-P., Thomas, K. et al. (2006) A continuous gas flow MAS NMR probe for operando studies of hydrocarbon conversion on heterogeneous catalysts, C.R. Chimie 9, 459. [17] Iglesias-Juez, A., Martinez-Arias, A., Hungria, A.B. et al. (2004) Influence of the nature of the Ce-promoter on the behavior of Pd and Pd–Cr TWC systems, Appl. Catal. A Gen., 259, 207. [18] Martinez-Arias, A., Fernández-Garcia, M., Iglesias-Juez, A. et al. (2001) New Pd/Cex Zr1−x O2 /Al2 O3 three-way catalysts prepared by microemulsion Part 2. In situ analysis of CO oxidation and NO reduction under stoichiometric CO + NO + O2 , Appl. Catal. B Environ., 31, 51. [19] Iglesias-Juez, A., Martínez-Arias, A. and Fernández-García, M. (2004) Metal-promoter interface in Pd/(Ce,Zr)Ox /Al2 O3 catalysts: Effect of thermal aging, J. Catal., 221, 148. [20] Lecomte, J.J., Granger, P., Leclercq, L. et al. (1999) An EPR investigation on the reactivity of oxygen from ceria modified bimetallic Pt-Rh:Al2 O3 catalysts in the CO + NO reaction, Colloids Surf. A Physicochem. Eng. Asp., 158, 241. [21] Santhosh Kumar, M., Schwidder, M., Grünert, W. et al. (2004) On the nature of different iron sites and their catalytic role in Fe-ZSM-5 DeNOx catalysts: New insights by a combined EPR and UV/VIS spectroscopic approach, J. Catal., 227, 384. [22] Pérez-Ramírez, J., Santhosh Kumar, M. and Brückner, A. (2004) Reduction of N2 O with CO over FeMFI zeolites: Influence of the preparation method on the iron species and catalytic behavior, J. Catal., 223, 13. [23] Schwidder, M., Santhosh Kumar, M., Brückner, A. et al. (2005) Active sites for NO reduction over Fe-ZSM-5 catalysts, Chem. Commun., 6, 805. [24] Larese, C., Cabello Galisteo, F., López Granados, M. et al. (2004) Effects of the CePO4 on the oxygen storage and release properties of CeO2 and Ce08 Zr02 O2 solid solution, J. Catal., 226, 443. [25] Mamede, A.-S., Leclercq, G., Payen, E. et al. (2003) In situ Raman characterisation of surface modifications during NO transformation over automotive Pd-based exhaust catalysts, J. Mol. Struct., 651–653, 353. [26] Guerrero-Pérez, M.O. and Bañares, M.A. (2006) From conventional in situ to operando studies in Raman spectroscopy, Catal. Today, 113, 48. [27] Bañares, M.A. (2005) Operando methodology: combination of in situ spectroscopy and simultaneous activity measurements under catalytic reaction conditions, Catal. Today, 100, 71. [28] Busca, G. (1999) The surface acidity of solid oxides and its characterization by IR spectroscopic methods. An attempt at systematisation, Phys. Chem. Chem. Phys., 1, 723. [29] Busca, G. (1998) Spectroscopic characterization of the acid properties of metal oxide catalysts, Catal. Today, 41, 191. [30] Kustov, L.M. (1997) New trends in IR-spectroscopic characterization of acid and basic sites in zeolites and oxide catalysts, Top. Catal., 4, 131. [31] Lercher, J.A., Gruendling, C. and Eder-Mirth, G. (1996) Infrared studies of the surface acidity of oxides and zeolites using adsorbed probe molecules, Catal. Today, 27, 353. [32] Travert, A., Vimont, A., Sahibed-Dine, A. et al. (2006) Use of pyridine CH(D) vibrations for the study of Lewis acidity of metal oxides, Appl. Catal. A, 307, 98.

136

Past and Present in DeNOx Catalysis

[33] Wakabayashi, F. and Domen, K. (1997) A new method for characterizing solid surface acidity – an infrared spectroscopic method using probe molecules such as N2 and rare gases, Catalysis Surveys from Japan 1 181. [34] Knozinger, H. and Huber, S. (1998) IR spectroscopy of small and weakly interacting molecular probes for acidic and basic zeolites, J. Chem. Soc., Faraday Trans., 94, 2047. [35] Hadjiivanov, K.I. and Vayssilov, G.N. (2002) Characterization of Oxide Surfaces and Zeolites by Carbon Monoxide as an IR Probe Molecule, Adv. Catal., 47, 307. [36] Nova, I., Lietti, L., Casagrande, L. et al. (1998) Characterization and reactivity of TiO2supported MoO3 de-NOx SCR catalysts, Appl. Catal. B, 17, 245. [37] Hoang-Van, C., Zegaoui, O. and Pichat, P. (1998) Vanadia–titania aerogel deNOx catalysts, J. Non Cryst. Solids, 225, 157. [38] Topsøe, N.Y., Anstrom, M. and Dumesic, J.A. (2001) Raman, FTIR and theoretical evidence for dynamic structural rearrangements of vanadia/titania deNOx catalysts, Catal. Lett., 76, 11. [39] Topsøe, N.Y. (2006) In situ FTIR: A versatile tool for the study of industrial catalysts, Catal. Today, 113, 58. [40] Madia, G., Elsener, M., Koebel, M. et al. (2002) Thermal stability of vanadia-tungsta-titania catalysts in the SCR process, Appl. Catal. B, 39, 181. [41] Larrubia, M.A., Ramis, G. and Busca, G. (2000) An FT-IR study of the adsorption of urea and ammonia over V2 O5 -MoO3 -TiO2 SCR catalysts, Appl. Catal. B, 27, L145. [42] Richter, R., Eckelt, R., Parlitz, B. et al. (1998) Low-temperature conversion of NOx to N2 by zeolite-fixed ammonium ions, Appl. Catal. B, 15, 129. [43] Romero Sarria, F., Saussey, J., Gallas, J.-P. et al. (2005) In situ and operando IR study of adsorption sites for NH4 + active species in NOx -SCR via NH3 using a Y zeolite, Stud. Surf. Sci. Catal., 158, 821. [44] Romero Sarria, F., Marie, O., Bazin, P. et al. (2006) FT-IR operando study on selective catalytic reduction of NOx species by ammonia: A comparison between zeolitic and GAPON compounds, Catal. Today, 113, 87. [45] Pieplu, T., Poignant, F., Vallet, A. et al. (1995) Oxidation state of copper during the reduction of NOx with propane on H-Cu-ZSM-5 in excess oxygen, Stud. Surf. Sci. Catal., 96, 619. [46] Poignant, F., Saussey, J., Lavalley, J.C. et al. (1996) In situ FT-IR study of NH3 formation during the reduction of NOx with propane on H-Cu-ZSM-5 in excess oxygen, Catal. Today, 29, 93. [47] Montanari, T., Marie, O., Daturi, M. et al. (2005) Cobalt on and in zeolites and silica–alumina: Spectroscopic characterization and reactivity, Catal. Today, 110, 339. [48] Jacobs, P.A. and Mortier, W.J. (1982) An attempt to rationalize stretching frequencies of lattice hydroxyl groups in hydrogen-zeolites, Zeolites, 2, 226. [49] Fichtner-Schmittler, H., Lohse, U., Engelhardt, G. et al. (1984) Unit-cell constants of zeolites stabilized by dealumination – determination of al-content from lattice-parameters, Cryst. Res. Technol., 19, K1. [50] Goovaerts, F., Vansant, E.F., De Hulsters, P. et al. (1989) Structural vibrations of acidleached mordenites – determination of structural aluminum by wave-number and intensity analysis, J. Chem. Soc., Faraday Trans., 85, 3687. [51] Marie, O., Thibault-Starzyk, F. and Lavalley, J.C. (2000) Confirmation of the strongest nitriles-hydroxy groups interaction in the side pockets of mordenite zeolites, Phys. Chem. Chem. Phys., 2, 5341. [52] Zecchina, A., Marchese, L., Bordiga, S. et al. (1997) Vibrational spectroscopy of NH4 + ions in zeolitic materials: An IR study, J. Phys. Chem. B, 101, 10128. [53] Romero Sarria, F., Marie, O., Saussey, J. et al. (2005) Infrared evidence of three distinct acidic hydroxyls in defects free HY faujasite, J. Phys. Chem. B, 109, 1660. [54] Romero Sarria, F., Blasin-Aubé, V., Saussey, J. et al. (2006) Trimethylamine as a Probe Molecule To Differentiate Acid Sites in Y-FAU Zeolite: FTIR Study, J. Phys. Chem. B, 110, 13130.

General Features of In Situ and Operando Spectroscopic Investigation

137

[55] Kato, K., Nohira, H., Nakanishi, K. et al. (1993) H. Eur. Patent applications 0573672 A1 to Toyota. [56] Epling, W., Campbell, L.E., Yezerets, A. et al. (2004) Overview of the fundamental reactions and degradation mechanisms of NOx storage/reduction catalysts, Catal. Rev, 46, 163. [57] Busca, G. (2006) The surface acidity and basicity of solid oxides and zeolites, Chemical Industries (Boca Raton, FL, United States), 108 (Metal Oxides), 247. [58] Coluccia, S., Marchese, L. and Martra, G. (1999) Characterisation of microporous and mesoporous materials by the adsorption of molecular probes: FTIR and UV–Vis studies, Microporous Mesoporous Mater., 30, 43. [59] Coluccia, S., Marchese, L. and Martra, G. (2000) Characterisation of zeolitic catalysts by the adsorption of molecular probes, Res. Chem. Intermed., 26, 1. [60] Lavalley, J.C. (1991) Use of probe molecules for the characterization of the surface basicity of divided metal oxides, Trends Phys. Chem., 2, 305. [61] Lavalley, J.C. (1996) Infrared spectrometric studies of the surface basicity of metal oxides and zeolites using adsorbed probe molecules, Catal. Today, 27, 377. [62] Barthomeuf, D. (1984) Conjugate acid-base pairs in zeolites, J. Phys. Chem., 88, 42. [63] Barthomeuf, D., Coudurier, G. and Vedrine, J.C. (1988) Basicity and basic catalytic properties of zeolites, Mater. Chem. Phys., 18, 553. [64] Barthomeuf, D. (1996) Basic zeolites: Characterization and uses in adsorption and catalysis, Catal. Rev. Sci. Eng., 38, 521. [65] Marie, O., Malicki, N., Pommier, C. et al. (2005) NO2 disproportionation for the IR characterisation of basic zeolites, Chem. Commun., 8, 1049. [66] Thibault-Starzyk, F., Marie, O., Malicki, N. et al. (2005) NO+ : Infrared probe for basic zeolites, Stud. Surf. Sci. Catal., 158, 663. [67] Hadjiivanov, K.I. (2000) Identification of Neutral and Charged Nx Oy Surface Species by IR Spectroscopy, Catal. Rev. Sci. Eng., 42, 71. [68] Binet, C., Daturi, M. and Lavalley, J.C. (1999) IR study of polycrystalline ceria properties in oxidised and reduced states, Catal. Today, 50, 207. [69] Daturi, M., Binet, C., Lavalley, J.C. et al. (2000) Surface FTIR investigations on Cex Zr1−x O2 system, Surf. Interface Anal., 30, 273. [70] Finocchio, E., Daturi, M., Binet, C. et al. (1999) Thermal Evolution of the Adsorbed Methoxy Species on Cex Zr1−x O2 Solid Solution Samples: A FT-IR study, Catal. Today, 52, 53. [71] Fally, F., Perrichon, V., Vidal, H. et al. (2000) Modification of the oxygen storage capacity of CeO2 −ZrO2 mixed oxides after redox cycling aging, Catal. Today, 59, 373. [72] Binet, C. and Daturi, M. (2001) Methanol as an IR probe to study the reduction process in ceria-zirconia mixed compounds, Catal. Today, 70, 155. [73] Daturi, M., Finocchio, E., Binet, C. et al. (1999) Study of bulk and surface reduction by hydrogen of Cex Zr1−x O2 mixed oxides followed by FTIR spectroscopy and magnetic balance, J. Phys. Chem. B, 103, 4884. [74] Tethwisch, D.G. and Dumesic, J.A. (1986) Effect of metal-oxygen bond strength on properties of oxides. 1. Infrared spectroscopy of adsorbed carbon monoxide and carbon dioxide, Langmuir, 2, 73. [75] Lesage, T.,Verrier, C., Bazin, P. et al. (2004) Comparison between a Pt-Rh/Ba/Al2 O3 and a newly formulated NOx -trap catalysts under alternate lean-rich flows, Top. Catal., 30, 31. [76] Perdigon-Melon, J.A., Gervasini, A. and Auroux, A. (2005) Study of the influence of the In2 O3 loading on -alumina for the development of de-NOx catalysts, J. Catal., 234, 421. [77] Li, G., Jones, C.A., Grassian, V.H. et al. (2005) Selective catalytic reduction of NO2 with urea in nanocrystalline NaY zeolite, J. Catal., 234, 401. [78] Klingenberg, B. and Vannice, M.A. (1999) NO adsorption and decomposition on La2 O3 studied by DRIFTS, Appl. Catal. B, 21, 19. [79] Djega-Mariadassou, G. and Boudart, M. (2003) Classical kinetics of catalytic reactions, J. Catal., 216, 89.

138

Past and Present in DeNOx Catalysis

[80] Dyrek, K. and Che, M. (1997) EPR as a tool to investigate the transition metal chemistry on oxide surfaces, Chem. Rev., 97, 305. [81] Frache, A., Gianotti, E. and Marchese, L. (2003) Spectroscopic characterisation of microporous aluminophosphate materials with potential application in environmental catalysis, Catal. Today, 77, 371. [82] Fierro-Gonzalez, J.C., Kuba, S., Hao, Y. et al. (2006) Oxide- and zeolite-supported molecular metal complexes and clusters: physical characterization and determination of structure, bonding, and metal oxidation state, J. Phys. Chem. B, 110, 13326. [83] Wachs, I.E. (2005) Recent conceptual advances in the catalysis science of mixed metal oxide catalytic materials, Catal. Today, 100, 79. [84] Bartlett, J.R. and Cooney, R.P. (1987) Adv. Spectrosc, 14, 187. [85] Wachs, I.E., Hardcastle, F.D. and Chan, S.S. (1986) Spectroscopy, 1, 30. [86] Wachs, I.E. (1996) Raman and IR studies of surface metal oxide species on oxide supports: Supported metal oxide catalysts, Catal. Today, 27, 437. [87] Wachs, I.E., Briand, L.E., Jehng, J.-M. et al. (2000) Molecular structure and reactivity of the group V metal oxides, Catal. Today, 57, 323. [88] Li, C. (2001) Stud. Surf. Sci. Catal., 132, 677. [89] Sobalik, Z., Tvaruzkova, Z. and Wichterlova, B. (1998) Skeletal T-O-T vibrations as a tool for characterization of divalent cation complexation in ferrierite, J. Phys. Chem. B, 102, 1077. [90] Capek, L., Kreibich, V., Dedecek, J. et al. (2005) Analysis of Fe species in zeolites by UV–VIS–NIR, IR spectra and voltammetry. Effect of preparation, Fe loading and zeolite type, Microporous Mesoporous Mater., 80, 279. [91] Chang, T.H. and Leu, F.C. (1999) Reactivity of titanosilicates-support catalysts for the selective catalytic reduction of NO with NH3 , Appl. Catal. A, 180, 123. [92] El-Malki, E.M., van Santen, R.A. and Sachtler, W.M.H. (2000) Active sites in Fe/MFI catalysts for NOx reduction and oscillating N2 O decomposition, J. Catal., 196 212. [93] Marchi, A. (2002) J. Trends Appl. Spectrosc., 4, 239. [94] Concepcion, P., Reddy, B.M. and Knozinger, H. (1999) FTIR study of low-temperature CO adsorption on pure Al2 O3 -TiO2 and V/Al2 O3 -TiO2 catalysts, Phys. Chem. Chem. Phys., 1, 3031. [95] Hadjiivanov, K., Concepcion, P. and Knozinger, H. (2000) Analysis of Oxidation States of Vanadium in Vanadia–Titania Catalysts by the IR Spectra of Adsorbed NO, Top. Catal., 11/12, 123. [96] Nova, I., Lietti, L., Casagrande, L. et al. (1998) Characterization and reactivity of TiO2 supported MoO3 de-NOx SCR catalysts, Appl. Catal. B, 17, 245. [97] Iwamoto, M., Mizuno, N. and Yahiro, H. (1991) Removal of nitrogen monoxide over copper ion-exchanged zeolite catalysts, Sekiyu Gakkaishi, 34, 375. [98] Hadjiivanov, K., Klissurski, D., Ramis, G. et al. (1996) Fourier Transform IR Study of NOx Adsorption on a CuZSM-5 DeNOx Catalyst, Appl. Catal. B, 7, 251. [99] Praliaud, H., Mikhailenko, S., Chajar, Z. et al. (1998) Surface and bulk properties of Cu−ZSM-5 and Cu/Al2 O3 solids during redox treatments. Correlation with the selective reduction of nitric oxide by hydrocarbons, Appl. Catal. B, 16, 359. [100] Broclawik, E., Datka, J., Gil, B. et al. (2002) Why Cu+ in ZSM-5 framework is active in deNOx reaction—quantum chemical calculations and IR studies, Catal. Today, 75, 353. [101] Castagnola, M.J., Neylon, M.K. and Marshall, C.L. (2004) Coated bifunctional catalysts for NOx SCR with C3 H6 : Part II. In situ spectroscopic characterization, Catal. Today, 96, 61. [102] Datka, J., Kozyra, P. and Kukulska-Zajac, E. (2004) IR studies and DFT calculations concerning the status of Cu+ ions in CuZSM-5 and CuMCM-41, Catal. Today, 90, 109. [103] Wichterlová, B., Sobalík, Z. and Dedecek, J. (1997) Cu ion siting in high silica zeolites. Spectroscopy and redox properties, Catal. Today, 38, 199.

General Features of In Situ and Operando Spectroscopic Investigation

139

[104] Llabres i Xamena, F.X., Fisicaro, P., Berlier, G. et al. (2003) Thermal reduction of Cu2+ -mordenite and Re-oxidation upon interaction with H2 O, O2 , and NO, J. Phys. Chem. B, 107, 7036. [105] Joyner, R. and Stockenhuber, M. (1999) Preparation, characterization, and performance of Fe-ZSM-5 catalysts, J. Phys. Chem. B, 103, 5963. [106] Chen, H.Y., El-Malki, El-M., Wang, X. et al. (2000) Identification of active sites and adsorption complexes in Fe/MFI catalysts for NOx reduction, J. Mol. Cat. A, 162, 159. [107] Novakova, J., Kubelkova, L., Wichterlova, B. et al. (1982) Properties of different Fe (III) species introduced into NH4 -Y zeolites, Zeolites, 2, 17. [108] Daturi, M., Bion, N., Saussey, J. et al. (2001) Evidences of a lacunar mechanism for deNOx activity in ceria-based catalysts, Phys. Chem. Chem. Phys., 3, 252. [109] Le Bourdon, G., Adar, F., Moreau, M. et al. (2003) In situ characterization by Raman and IR vibrational spectroscopies on a single instrument: deNOx reaction over a Pd/-Al2 O3 catalyst, Phys. Chem. Chem. Phys., 5, 4441. [110] Burch, R., Breen, J.P. and Meunier, F.C. (2002) A review of the selective reduction of NOx with hydrocarbons under lean-burn conditions with non-zeolitic oxides and platinum group metal catalysts, Appl. Catal. B, 39, 283. [111] Takahashi, N., Shinjoh, H., Iijima, T. et al. (1996) The new concept 3-way catalyst for automotive lean-burn engine: NOx storage and reduction catalyst, Catal. Today, 27, 63. [112] Olsson, L. and Fridell, E. (2002) The influence of pt oxide formation and pt dispersion on the reactions NO2  NO + 1/2 O2 over Pt/Al2 O3 and Pt/BaO/Al2 O3 , J. Catal., 210, 340. [113] Perrichon, V., Retailleau, L., Bazin, P. et al. (2004) Metal dispersion of CeO2 −ZrO2 supported platinum catalysts measured by H2 or CO chemisorption, Appl. Catal. A, 260, 1. [114] Ryczkowski, J. (2001) IR spectroscopy in catalysis, Catal. Today, 68, 263. [115] Gerlach, T., Schütze, F.-W. and Baerns, M. (1999) An FTIR study on the mechanism of the reaction between nitrogen dioxide and propene over acidic mordenites, J. Catal., 185, 131. [116] Szanyi, J., Kwak, J.H., Moline, R.A. et al. (2003) The adsorption of NO2 and the NO + O2 reaction on Na-Y,FAU: An in situ FTIR investigation, Phys. Chem. Chem. Phys., 5, 4045. [117] Kantcheva, M. (2001) Identification, Stability, and Reactivity of NOx Species Adsorbed on Titania-Supported Manganese Catalysts, J. Catal., 204, 479. [118] Kass, M.D., Thomas, J.F., Lewis, S.A. et al. (2003) Soc. Automot. Eng. (SAE) 01–3244. [119] Noto, T., Murayama, T., Tosaka, S. et al. (2001) Soc. Automot. Eng. (SAE), 01–1935. [120] Hoon Yeom, Y., Li, M., Sachtler, W.M.H. et al. (2006) A study of the mechanism for NOx reduction with ethanol on -alumina supported silver, J. Catal., 238, 100. [121] Sadykov, V.A., Paukshtis, E.A., Beloshapkin, S.A. et al. (1996) In: Proceedings of 8th International Symposium on Heterog. Catalysis, Vatrna, 5–9 October, 1996, pp. 347. [122] Sadykov, V.A., Beloshapkin, S.A., Paukshtis, E.A. et al. (1997) Pol. J. Environ. Stud., 6, 21. [123] Beloshapkin, S.A., Matyshak, V.A., Paukshtis, E.A. et al. (1999) IR studies of the transformations of nitrogen-containing organic intermediates during selective reduction of nitrogen oxides by hydrocarbons, React. Kinet. Catal. Lett., 66, 297. [124] Beloshapkin, S.A., Paukshtis, E.A. and Sadykov, V.A. (2000) FTIRS study of the nature and reactivity of the products of nitromethane transformation on the surface of copper- and cobalt-exchanged ZSM-5 zeolites as related to the mechanism of NOx CH4 -SCR, J. Mol. Catal. A Chem., 158, 355. [125] Szailer, T., Kwak, J.H., Kim, D.H. et al. (2006) Reduction of stored NOx on Pt/Al2 O3 and Pt/BaO/Al2 O3 catalysts with H2 and CO, J. Catal., 239, 51.

140

Past and Present in DeNOx Catalysis

[126] Freysz, J.L., Saussey, J., Lavalley, J.C. et al. (2001) In situ FT-IR Study of the NO + CO reaction on a silica-supported platinum catalyst at atmospheric pressure using a new pulse technique, J. Catal., 197, 131. [127] Konduru, M.V. and Chuang, S.S.C. (2000) Dynamics of NO and N2 O Decomposition over Cu−ZSM-5 under Transient Reducing and Oxidizing Conditions, J. Catal., 196, 271. [128] Konduru, M.V. and Chuang, S.S.C. (1999) Active and spectator adsorbates during NO decomposition over Cu−ZSM-5: Transient IR, site-poisoning, and site-promotion studies, J. Catal., 187, 436. [129] Almusaiteer, K.A., Chuang, S.S.C. and Tan, C.-D. (2000) Reactivity of Rh+ (CO)2 during the NO−CO and CO−O2 Reactions over Rh/Al2 O3 , J. Catal., 189, 247. [130] Chi, Y. and Chuang, S.S.C. (2000) The effect of oxygen concentration on the reduction of NO with propylene over CuO/-Al2 O3 , Catal. Today, 62, 303. [131] Chi, Y. and Chuang, S.S.C. (2000) Infrared study of NO adsorption and reduction with C3 H6 in the presence of O2 over CuO/Al2 O3 , J. Catal., 190, 75. [132] Lesage, T., Verrier, C., Bazin, P. et al. (2003) Studying the NOx -trap mechanism over a Pt-Rh/Ba/Al2 O3 catalyst by operando FT-IR spectroscopy, Phys. Chem. Chem. Phys., 5, 4435. [133] Bion, N., Saussey, J., Haneda, M. et al. (2003) Study by in situ FTIR spectroscopy of the SCR of NOx by ethanol on Ag/Al2 O3 – Evidence of the role of isocyanate species, J. Catal., 217, 47. [134] Haneda, M., Joubert, E., Menezo, J.C. et al. (2001) Reaction intermediates in the selective reduction of NO with propene over Ga2 O3 −Al2 O3 and In2 O3 −Al2 O3 catalysts, J. Mol. Catal. A Chem., 175, 179. [135] Haneda, M., Bion, N., Daturi, M. et al. (2002) In situ FT-IR study of the selective reduction of NO with propene over Ga2 O3 –Al2 O3 , J. Catal., 206, 114. [136] Frola, F., Prinetto, F., Ghiotti, G. et al. (2007) Combined in situ FT-IR and TRM analysis of the NOx storage properties of Pt-Ba/Al2 O3 LNT catalysts, Catal. Today 126, 81–89. [137] Poignant, F., Freysz, J.L., Daturi, M. et al. (2001) NO SCR by propane in oxygen excess on CuH-ZSM-5; formation of Cu+ NC species via acrylonitrile intermediate, Catal. Today, 70, 197. [138] Tibiletti, D., Goguet, A., Reid, D. et al. (2006) On the need to use steady-state or operando techniques to investigate reaction mechanisms: An in situ DRIFTS and SSITKA-based study example, Catal. Today, 113, 94. [139] Long, R.Q. and Yang, R.T. (2002) Reaction mechanism of selective catalytic reduction of NO with NH3 over Fe-ZSM-5 catalyst, J. Catal., 207, 224. [140] Lobree, L.J., Aylor, A.W., Reimer, J.A. et al. (1997) Role of cyanide species in the reduction of NO by CH4 over Co−ZSM-5, J. Catal., 169, 188. [141] Shimizu, K.I., Kawabata, H., Maeshima, H. et al. (2000) Intermediates in the selective reduction of NO by propene over Cu-Al2 O3 catalysts: Transient in-Situ FTIR study, J. Phys. Chem. B, 104, 2885. [142] Hadjiivanov, K., Knözinger, H., Tsyntsarski, B. et al. (1999) Effect of water on the reduction of NOx with propane on Fe-ZSM-5. An FTIR mechanistic study, Catal. Lett., 62, 35. [143] Campa, M.C., Pietrogiacomi, D., Tuti, S. et al. (1998) The selective catalytic reduction of NOx with CH4 on Mn-ZSM5: A comparison with Co-ZSM5 and Cu-ZSM5, Appl. Catal. B, 18, 151. [144] Satsuma, A., Enjoji, T., Shimizu, K.I. et al. (1998) Reactivity of surface nitrate species in the selective reduction of NO with propene over Na-H-mordenite as investigated by dynamic FTIR spectroscopy, J. Chem. Soc., Faraday Trans., 94, 301. [145] Ishihara, T., Kagawa, M., Hadama, F. et al. (1997) Copper ion-exchanged SAPO-34 as a thermostable catalyst for selective reduction of NO with C3 H6 , J. Catal., 169, 93.

General Features of In Situ and Operando Spectroscopic Investigation

141

[146] Mosqueda-Jiménez, B.I., Jentys, A., Seshan, K. et al. (2003) On the surface reactions during NO reduction with propene and propane on Ni-exchanged mordenite, Appl. Catal. B, 46, 189. [147] Lobree, L.J., Hwang, I.C., Reimer, J.A. et al. (1999) An in situ infrared study of NO reduction by C3 H8 over Fe-ZSM-5, Catal. Lett., 63, 233. [148] Hwang, I.C., Kim, D.H. and Woo, S.I. (1996) The existence of dual Cu site involved in the selective catalytic reduction of NO with propene on Cu/ZSM-5, Catal. Lett., 42, 177. [149] Meunier, F.C., Zuzaniuk, V., Breen, J.P. et al. (2000) Mechanistic differences in the selective reduction of NO by propene over cobalt and silver-promoted alumina catalysts: A kinetic and in situ DRIFTS study, Catal. Today, 59, 287. [150] Park, S.K., Park, Y.K., Park, S.E. et al. (2000) Comparison of selective catalytic reduction of NO with C3 H6 and C3 H8 over Cu(II)-ZSM-5 and Co(II)-ZSM-5, Phys. Chem. Chem. Phys., 2, 5500. [151] Meunier, F.C., Breen, J.P., Zuzaniuk, V. et al. (1999) Mechanistic aspects of the selective reduction of NO by propene over alumina and silver-alumina catalysts, J. Catal., 187, 493. [152] Captain, D.K. and Amiridis, M.D. (1999) In Situ FTIR studies of the selective catalytic reduction of NO by C3 H6 over Pt/Al2 O3 , J. Catal., 184, 377. [153] Sumiya, S., He, H., Abe, A. et al. (1998) Formation and reactivity of isocyanate (NCO) species on Ag/Al2 O3 , J. Chem. Soc., Faraday Trans., 94, 2217. [154] Bell, A.T. (1997) Experimental and theoretical studies of NO decomposition and reduction over metal-exchanged ZSM-5, Catal. Today, 38, 151. [155] Hwang, I.C., Kim, D.H. and Woo, S.I. (1998) Role of oxygen on NOx SCR catalyzed over Cu/ZSM-5 studied by FTIR, TPD, XPS and micropulse reaction, Catal. Today, 44, 47. [156] Hayes, N.W., Joyner, R.W. and Shpiro, E.S. (1996) Infrared spectroscopy studies of the mechanism of the selective reduction of NOx over Cu-ZSM-5 catalysts, Appl. Catal. B, 8, 343. [157] Chen, H.Y., Voskoboinikov, T. and Sachtler, W.M.H. (1998) Reduction of NOx over Fe/ZSM-5 catalysts: Adsorption complexes and their reactivity toward hydrocarbons, J. Catal., 180, 171. [158] Poignant, F., Freysz, J.L., Daturi, M. et al. (2001) NO SCR by propane in oxygen excess on CuH-ZSM-5; formation of Cu+ NC species via acrylonitrile intermediate, Catal. Today, 70, 197. [159] Gerlach, T., Schutze, F.-W. and Baerns, M. (1999) An FTIR study on the mechanism of the reaction between nitrogen dioxide and propene over acidic mordenites, J. Catal., 185, 431. [160] Konduru, M.V. and Chuang, S.S.C. (2000) Dynamics of NO and N2 O decomposition over Cu−ZSM-5 under transient reducing and oxidizing conditions, J. Catal., 196, 271. [161] Shen, S.-C. and Kawi, S. (2003) Mechanism of selective catalytic reduction of NO in the presence of excess O2 over Pt/Si-MCM-41 catalyst, J. Catal., 213, 241. [162] Kwak, J.H., Peden, C.H.F. and Szanyi, J. (2006) Non-thermal plasma-assisted NOx reduction over Na-Y zeolites: The promotional effect of acid sites, Catal. Lett., 109, 1. [163] Szailer, T., Kwak, J.H., Kim, D.H. et al. (2006) Effects of Ba loading and calcination temperature on BaAl2 O4 formation for BaO/Al2 O3 NOx storage and reduction catalysts, Catal. Today, 114, 86. [164] Kondarides, D.I., Chafik, T. and Verykios, X.E. (2000) Catalytic reduction of NO by CO over rhodium catalysts. 2. effect of oxygen on the nature, population, and reactivity of surface species formed under reaction conditions, J. Catal., 191, 147. [165] Halkides, T.I., Kondarides, D.I. and Verykios, X.E. (2002) Mechanistic study of the reduction of NO by C3 H6 in the presence of oxygen over Rh/TiO2 catalysts, Catal. Today, 73, 213. [166] Westerberg, B. and Fridell, E. (2001) A transient FTIR study of species formed during NOx storage in the Pt/BaO/Al2 O3 system, J. Mol. Catal. A Chem., 165, 249.

142

Past and Present in DeNOx Catalysis

[167] Tinnemans, S.J., Mesu, J.G., Kervinen, K. et al. (2006) Combining operando techniques in one spectroscopic-reaction cell: New opportunities for elucidating the active site and related reaction mechanism in catalysis, Catal. Today, 113, 3. [168] Brückner, A. and Kondratenko, E. (2006) Simultaneous operando EPR/UV–vis/laser–Raman spectroscopy – A powerful tool for monitoring transition metal oxide catalysts during reaction, Catal. Today, 113, 16. [169] Wachs, I.E. (2003) Letter to the Editor; International Congress on Operando Spectroscopy: Fundamental and technical aspects of spectroscopy of catalysts under working conditions Lunteren, The Netherlands (March 2–6, 2003), Catal. Commun., 4, 567. [170] Topsøe, H. (2003) Developments in operando studies and in situ characterization of heterogeneous catalysts, J. Catal., 216, 155. [171] Busca, G., Lietti, L., Ramis, G. et al. (1998) Chemical and mechanistic aspects of the selective catalytic reduction of NOx by ammonia over oxide catalysts: A review, Appl. Catal. B Environ., 18, 1. [172] Armor, J.N. (1995) Catalytic reduction of nitrogen oxides with methane in the presence of excess oxygen: A review, Catal. Today, 26, 147. [173] Montanari, T., Marie, O., Daturi, M. et al. (2006) Searching for the active sites of Co-HMFI catalyst for the selective catalytic reduction of NO by methane: A FT-IR in situ and operando study, Appl. Catal. B Environ., 71, 216. [174] Brosius, R., Bazin, P., Thibault-Starzyk, F. et al., (2005) Operando FTIR study of reaction pathways of selective catalytic reduction of NOx with decane in the presence of water on iron-exchanged MFI-type zeolite, J. Catal., 234, 191. [175] Romero Sarria, F., Marie, O., Bazin, P. et al. (2006) FT-IR operando study on selective catalytic reduction of NOx species by ammonia: A comparison between zeolitic and GAPON compounds, Catal. Today, 113, 87. [176] Groothaert, M.H., Lievens, K., Leeman, H. et al. (2003) An operando optical fiber UV– vis spectroscopic study of the catalytic decomposition of NO and N2 O over Cu-ZSM-5, J. Catal., 220, 500. [177] Groothaert, M.H., van Bokhoven, J.A., Battiston, A.A. et al. (2003) Bis( -oxo)dicopper in Cu-ZSM-5 and its role in the decomposition of NO: A combined in Situ XAFS, UV-VisNear-IR, and kinetic study, J. Am. Chem. Soc., 125, 7629. [178] Schwidder, M., Heinrich, F., Kumar, M.S. et al. (2004) Fe-ZSM-5 catalysts for the selective reduction of NO: Influence of preparation route on structure and catalytic activity, Stud. Surf. Sci. Catal., 154, 2484. [179] Iliopoulou, E.F., Efthimiadis, E.A., Nalbandian, L. et al. (2005) Ir-based additives for NO reduction and CO oxidation in the FCC regenerator: Evaluation, characterization and mechanistic studies, Appl. Catal. B Environ., 60, 277. [180] Delahay, G., Guzman-Vargas, A. and Coq, B. (2007) Deactivation of a Fe-ZSM-5 catalyst during the selective catalytic reduction of NO by n-decane: An operando DRIFT study, Appl. Catal. B Environ., 70, 45. [181] Mosqueda-Jiménez, B.I., Lahougue, A., Bazin, P. et al. (2007) Operando systems for the evaluation of the catalytic performance of NOx storage and reduction materials, Catal. Today, 119, 73. [182] Lesage, T., Saussey, J., Malo, S. et al. (2006) Operando FTIR study of NOx storage over a Pt/K/Mn/Al2 O3 -CeO2 catalyst, Appl. Catal. B Environ., 72, 166. [183] Anderson, J.A., Liu, Z. and Fernandez Garcia, M. (2006) Use of in situ FT-IR and XAS/XRD to study SO2 poisoning over model Pt/Ba/Al2 O3 NOx storage and reduction (NSR) catalysts, Catal. Today, 113, 25. [184] Liu, Z. and Anderson, J.A. (2004) Influence of reductant on the regeneration of SO2 -poisoned Pt/Ba/Al2 O3 NOx storage and reduction catalyst, J. Catal., 228, 243.

General Features of In Situ and Operando Spectroscopic Investigation

143

[185] Szanyi, J., Kwak, J.H. and Peden, C.H.F. (2005) The catalytic chemistry of HCN + NO2 over Na- and Ba-Y,FAU: An in situ FTIR and TPD/TPR study, J. Phys. Chem. B, 109, 1481. [186] Szanyi, J., Kwak, J.H., Moline, R.A. et al. (2004) Adsorption, coadsorption, and reaction of acetaldehyde and NO2 on Na-Y,FAU: An in situ FTIR investigation, J. Phys. Chem. B, 108, 17050. [187] Seguin, E., Thibault-Starzyk, F. and Arnolds, H. (2006) Coupling step scan IR and pulsed laser for operando at 33ns in catalysis, proceedings of operando – II: Second international congress on operando spectroscopy, Toledo (Spain), 23–27 April, 2006.

This page intentionally left blank

Chapter 5

A THREE-FUNCTION MODEL REACTION FOR DESIGNING DeNOx CATALYSTS G. Djéga-Mariadassoua∗ , M. Bergerb , O. Gorcec , J. W. Parkd , H. Pernota , C. Potvina , C. Thomasa and P. Da Costaa a

b

Laboratoire Réactivité de Surface, CNRS UMR 7609, Université Pierre et Marie Curie, 4 Place Jussieu, Case 178, 75252 Paris Cedex 05, France

Gaz de France, Direction de la Recherche, 361 av. Président Wilson, B.P. 33, 93211 La Plaine Saint Denis, Cedex, France c

Renault sas, Centre Technique de Lardy, 1 allée Cornuel, 91510 Lardy, France d

On leave for Korea

∗

Corresponding author: Laboratoire Réactivité de Surface, CNRS UMR 7609, Université Pierre et Marie Curie, 4 Place Jussieu, Case 178, 75252 Paris Cedex 05, France. Tel.: +33 (0)1 44 27 36 24, Fax.: +33 (0)1 44 27 60 33, E-mail: [email protected] ([email protected])

Abstract A three-function catalyst model for hydrocarbon SCR of NOx is described, based on experimental evidence for each function, during temperature-programmed surface reactions (TPSR). The release of N2 occurs within function 3. It involves the dissociation of NO (via a dinitrosyl-adsorbed intermediate), followed by subsequent formation of N2 and scavenging of the adsorbed oxygen species left from NO dissociation. The removal of adsorbed oxygen is due to the total oxidation of an activated reductant (Cx Hy Oz . This reaction corresponds to ‘a supported homogeneous catalytic process’ involving a surface transition metal complex. The corresponding catalytic sequence of elementary steps occurs in the coordinative sphere of the metal cation. A function 2 has to turn over simultaneously to function 3. It has to deliver the active reductant species Cx Hy Oz to function 3, at the temperature where function-3 cycle turns over. Function 2 is the mild oxidation of HC (or any initial oxygenate) by NO2 , through organic nitrogen-containing intermediates (RNOx . The very important feature is that these RNOx species are quite thermally instable: they decompose with temperature to Cx Hy Oz + NO (not to N2 , according to the following global equation: HC or Cx Hy Oz  + NO2 = Cx Hy Oz + NO It is therefore obvious that functions 2 and 3 have to turn over simultaneously. Nevertheless, at the molecular level, these two functions are not in the same catalytic cycle.

Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

146

Past and Present in DeNOx Catalysis

A function 1 has also to turn over simultaneously with functions 2 and 3, as it has to provide NO2 to function 2. The oxidation of NO to NO2 is therefore the first function of any efficient catalyst. The concept of ‘simultaneous turn over’ between catalytic functions in multi-functional catalysis is widely accepted (for instance, in bi-functional metal/acid transformation of alkanes), and this aspect of the proposed mechanism is a ‘normal’ behaviour in steady state. Keywords Three-function catalyst; oxygenates; dinitrogen formation; supported homogeneous catalysis; metal cation active sites.

1. INTRODUCTION There can be some apparent contradiction in the general concept of NO reduction to N2 , in the presence of an excess of oxygen, according to the following global equation: 2 NO + Reductant + O2‘excess = N2 + CO/CO2 + H2 O

(1)

In lean conditions, reductants are found to be either hydrocarbons (HCs) or, in some industrial processes or academic studies, some more active oxygenates Cx Hy Oz . It should be recognized that the global reaction (1) is confusing, as nitrogen is globally reduced in oxidizing atmosphere, its formal oxidation state going from ‘+II’ (in NO) to ‘0’ (in N2 . Nevertheless, at a molecular level, there are several assumptions explaining this contradiction. In any case, the key of the problem still remains: How is dinitrogen (N2 ) formed? This question has to be answered to completely understand the DeNOx process, and design the final efficient catalyst, according to the nature of reductant and the experimental conditions (more particularly, temperature window). Three main approaches can be found in literature to describe the N−N bond formation: (1) An intermediate organic nitroso compound ‘RNOx ’ is formed, leading to N2 during its decomposition [1–5]. The mechanistic studies by Sachtler and co-workers [1–4] for the reduction of NOx by light alkanes over Fe/ZSM-5 involved adsorbed RNOx species which further react with gas-phase NOx to produce N2 , through the decomposition of diazo compounds [2,4]. (2) A ‘N−N’ bond pre-exists in RNOx compound, which then decomposes, leading to N2 . Over zeolite-supported Ag and Cu catalysts, the formation of nitrosonium and N -nitroso-N -alkylhydroxylamate intermediates have been assumed by Martens et al. [6], Brosius and Martens [8] and Kharas [7], respectively, to explain the N−N bonding before the release of molecular N2 . In fact, the authors [6 and 7] also explain the formation of N2 on a mechanism that involves organo-nitrogen (RNOx  intermediates, but the pairing of nitrogen atoms can involve different reaction pathways, through the formation of unstable diazonium compounds or by the transformation of the initially formed organo-nitrogen molecules to ammonia, and reduction of NOx via an ammonia-SCR process [8]. So, they did not seem to have evidences for RNOx compounds with pre-existing N−N bonds (as shown

Three-Function Model Reaction

147

in the Figure 5 of [8]). The proposed intermediates can have two atoms of N, but they are generally not directly linked. Finally, Iwamoto and Takeda [9] assumed that the decomposition of RNOx species leads to the formation of oxygenated compounds Cx Hy Oz which could be the intermediates of the global DeNOx process. This last assumption corresponds to a fundamental step of the model described in this chapter. (3) A dinitrosyl (NO)2 species forms, whose dissociation leads to N2 ; the remaining adsorbed oxygen species have to be scavenged by the activated reductant, Cx Hy Oz to recover the free active site, permitting the catalytic cycle to turn over (Figure 5.1, function 3) [10]. At a molecular scale, a three-function model can be defined with no direct interaction between the so-called ‘reductant’ and NO. In this case, the apparent contradiction to the global equation (1) can be ruled out, and the DeNOx reaction by itself occurs owing to the third function of the model (Figure 5.1).

2. GLOBAL PRESENTATION OF THE THREE-FUNCTION MODEL [10–13] 2.1. General concepts and definitions The present model deals with a supported transition metal cation which is highly dispersed, at the molecular scale, on an oxide, or exchanged in a zeolite. In the case of zeolite-supported cations, the formation of different metal species in metal/zeolite catalysts (metal oxides, metal oxocations, besides cationic species) has been considered by different authors who have suggested these species to play key roles in SCR catalysis [14,15]. This supported cation can also be considered as located at a metal oxide/support interface. The important feature is the formation of a coordinatively unsaturated site (cus), permitting the reaction to occur in the coordinative sphere of the metal cation. The cus is a metal cationic site that is able to present at least three vacancies permitting, in the DeNOx process, to insert ligands such as NO, CO, H2 O, and any olefin or Cx Hy Oz species that is able to behave like ligands in its coordinative environment. A cus can be located on kinks, ledges or corners of crystals [16]; in such a location, they are unsaturated. This situation is quite comparable to an exchanged cation in a zeolite, as studied by Iizuka and Lundsford [17] or to a transition metal complex in solution, as studied by Hendriksen et al. [18] for NO reduction in the presence of CO. The first pathway proposed by Iizuka and Lunsford [17] considers the reduction of nitric oxide by CO over rhodium/Y-zeolite. It leads only to N2 O as follows: I RhI CO2 + NO → ← Rh CO2 NO I RhI CO2 NO + NO → ← Rh CO-V  + N2 O + CO2

where (V ) is an oxygen vacancy in the coordinative sphere of the metal.

148

Past and Present in DeNOx Catalysis

Complexes such as RhI (CO)(NO) and RhI (NO)2 were also considered. In solution chemistry, at low temperature, Hendriksen et al. [18] assumed that the following sequence can occur, leading again to N2 O formation. III − RhI CO2 + 2 NO → ← Rh CONO 2 + CO I RhIII CONO− 2 → ← Rh + N2 O + CO2

Cataluna et al. [19] also proposed the formation of N2 O over ceria alone as follows: 4+ 4+ 2− 2Ce3+ -V  + 2 NO → ← Ce -ONNO -Ce 4+ 4+ 2− Ce4+ −ONNO2− −Ce4+ → ← Ce −O  + Ce V  + N2 O

where (V ) stands again for an oxygen vacancy of ceria. Let us note that in these three examples, the reaction leads to N2 O and not to N2 due to either low-temperature experiments or low catalyst activity. As can be seen, the catalytic process over a zeolite-supported cation, or an oxidesupported cation, can be considered as a supported homogeneous catalysis, as far as adsorbed reactants and products behave like reactive ligands. The model developed for lean DeNOx catalysts over supported cations (function 3), as well as this supported homogeneous catalysis approach, is also suitable for stoichiometric mixture (TWC) comprising CO and H2 as reductants over supported transition metal cations [20–22].

2.2. Metal active sites We shall mainly consider, in the present chapter, non-precious transition metals, but the model can be extended to precious metals presenting an oxidation state higher than zero [10,11], such as Rhx+ , Pdx+ , Ptx+ and Irx+ . The model also applies to some oxides alone, such as ceria (CeO2 ) [19] or mixed oxides such as ceria-zirconia (CeZrO2 ) able to present redox properties and oxygen vacancies during catalytic reactions.

2.3. The three-function catalytic system To be active in DeNOx reaction, a catalyst has to present two functions to assist function 3, in which the N−N bonding and N2 releases occur (Figure 5.1). This figure presents the three-function design of a DeNOx catalyst [12,13]. According to this model, the oxygenated intermediates (CH3 OH, HCHO) produced from initial HC (CH4 ) suffer total oxidation in function 3. One of the important targets of DeNOx reaction is that the catalyst has to produce by itself these oxygenates; they correspond to a mild oxidation of HC to Cx Hy Oz (function 2). The direct (faster, parallel or successive) total oxidation of HC to [CO/CO2 + H2 O] has to be avoided in function 3. The existence of these Cx Hy Oz is in agreement with Iwamoto and Tanaka’s conclusions [9]. If the total oxidation occurs in function 3, then the system will lose a great part of reductant. Let us note that the NO ‘reduction’ by oxygenates (Cx Hy Oz  is stoichiometric.

Three-Function Model Reaction

NO

149 CH3OH CO2 HCHO 2 NO N2 H2O

NO2 NO2 CH2

Function 1

Function 2 Function 3

Support support 1: NO + O2 → NO2 NO Oxidation

2: NO2 + HC => Cx HyOz HC Partial oxidation

3: 2 NO + Mx+ + Cx Hy Oz => N2 + xCO2 + y/2H2O + Mx+ DeNOx

Figure 5.1. The three-function model for designing DeNOx catalysts in the presence of methane as reductant [12].

So, we do not need a very high quantity of oxygenates. This can be experimentally demonstrated in the absence of the total HC oxidation reaction, by removing dioxygen in the feed. Furthermore, lower the reaction temperature, lower will be the total oxidation of the reductant. In contrast, lower the reaction temperature, higher will be the production of N2 O, escaping from the catalytic cycle of third function. As always, there is some compromise, and all the work consists in adjusting these parameters.

2.4. Global approach of function 3 Function 3 will be first globally described through two main catalytic stages [10–12]. (1) Stage 1 of function 3: dissociation of NOads to Nads and Oads It occurs via two adjacent adsorbed NO molecules, leading to an adsorbed ‘dinitrosyl’ species. These last two co-adsorbed NO species made the two N−O bonds weaker, and the successive two N−O bond scissions led to N2 . According to a general kinetic model [12], the N2 O intermediate can desorb before dissociating to N2 , if the desorption rate constant, kdes , is higher than the reaction (dissociation) rate constant, kreaction , as presented in the following set of rate constants (Figure 5.2): These rate constants are representative of the ‘rake mechanism’ of Germain [24]. Let us note that there is no need for any ‘organic nitroso intermediate’ for N2 formation. Therefore, two adsorbed oxygen species ‘Oads ’ (ex-NO) are remaining, strongly adsorbed k desorption k formation

Intermediategas phase Adsorbed intermediate

k reaction

Product

Figure 5.2. Set of rate constants for the catalytic activity of a reaction intermediate [13].

150

Past and Present in DeNOx Catalysis

on the catalytic site where function 3 is occurring, inhibiting any further NO adsorption– dissociation. (2) Stage 2 of function 3 If the two Oads species are not scavenged, then the reaction will stop. This is the case, for instance, of NO decomposition on Cu/ZSM-5 [25]. Adsorbed oxygen species have to be scavenged either by an activated form of the initial ‘HC’ reductant, such as Cx Hy Oz (alcohol, aldehyde, etc.) or by the initial HC if their total oxidation is simultaneous with NO decomposition–reduction to N2 . These ‘oxygenates’ and/or HC suffer a total oxidation to CO/CO2 and H2 O, regenerating the active site: this is the principle of catalysis. Once the active site is recovered, the reaction continues to turn over. This is the ‘catalytic cycle’. These two stages have to be included in function 3 of the catalyst (Figure 5.1) during the design of any DeNOx catalyst: they give a key to the global NO reduction and N2 formation.

2.5. Global approach of function 2 The catalyst has to get two more functions to assist the third one. It has to produce by itself, the active reductants, i.e. Cx Hy Oz (CH3 OH/HCHO in the case of Figure 5.1) by activating the initial HC of the feed. This mild oxidation of HC occurs owing to the presence of NO2 , actually well recognized as a good oxidant of HC at relatively low temperatures. As dioxygen generally provokes the HC total oxidation to CO/CO2 at quite higher temperatures, there is a new target for designing a good DeNOx catalyst: the lower the temperature of mild oxidation, when compared to the temperature of HC total oxidation, the most efficient will be the catalyst. In some cases, the total oxidation of HC is occurring simultaneously with both the Cx Hy Oz oxidation and the decomposition of NO to N2 .

2.6. Global approach of function 1 Figure 5.1 shows that function 1 is therefore the oxidation of NO to NO2 , this last one being subsequently delivered to function 2 to oxidize HC to Cx Hy Oz which will be delivered to function 3 to scavenge the adsorbed oxygen species left by the (NO)2 – adsorbed dinitrosyl species – decomposition. In order to experimentally demonstrate the model, this chapter will give evidence for each of the three functions. Once convinced by the model, the target will be to design a catalyst permitting the three functions to run simultaneously. It is finally easy to understand why this simultaneity of functions 2 and 3 can lead to a complete misunderstanding of the DeNOx reaction. As function 2 is producing RNOx compounds, simultaneously with function 3 which is producing N2 , it is rather difficult to discriminate between the elementary steps leading to either RNOx or N2 . The present chapter will demonstrate that, in lean conditions, RNOx leads to [Cx Hy Oz + NO] and N2 is formed in another catalytic cycle.

Three-Function Model Reaction

151

3. MAIN FEATURES OF EXPERIMENTAL CONDITIONS 3.1. Model catalyst preparations Various catalytic systems obey the present general model and will be used in this chapter. References to their synthesis are given in the text. The three-function model introduced in the preceding section has been established on an H-mordenite (HMOR) supported cobalt−palladium catalyst [12]. For the sake of demonstration, model catalysts with a unique function, i.e. F1, F2 or F3, (Figure 5.1), were prepared to separately give evidence of the major role of each active site (Figure 5.1). Let us note that ‘three functions’ does not necessarily mean three different active sites, but in the case of CoPd/HMOR material, three different sites were identified. Exchanged mordenites (CoPd/HMOR, Co/HMOR, Pd/HMOR) were purchased from the ‘Institut Régional des Matériaux Avancés (I.R.M.A.)’, located in Ploemeur (France). They were prepared according to Hamon et al.’s patent [23] by exchanging a NH4 mordenite with the appropriate amount of metallic precursors, respectively, cobalt(II) acetate and Pd(NH3 4 Cl2 . Two kinds of pre-treatment were subsequently applied: • 2 h at 773 K (500 C) in flowing nitrogen or argon (inert gases), with a flow rate of 100 mL min−1 . • 4 h at 573 K (300 C) followed by 4 h at 773 K (500 C) under synthetic air. The flow rate was 100 mL min−1 and the increasing temperature rates 1 K min−1 .

3.1.1. The complete CAT I, three-function catalyst – CoPd/HMOR – (Figure 5.3) It was obtained by a pre-treatment of fresh impregnated HMOR in flowing air, up to 773 K. In these conditions, as detected by TEM, EDS and UV–visible spectroscopy (not shown, [12]), a fraction of Co2+ species, exchanged in the pores of HMOR, migrates on the outside of the zeolite grain, to form Co3 O4 on the external surface of the HMOR grain. The material being impregnated by the palladium precursor, once all the internal exchanged positions have been already occupied by Co2+ , PdO is formed on the external surface of the zeolite grain, as observed by TEM (not shown, [12]).

3.1.2. ‘Cat II’ (Figure 5.3): Co/HMOR pre-treated in flowing air It corresponds to the cobalt initially exchanged into the HMOR porosity. Nevertheless, a fraction of cobalt oxide – Co3 O4 – is produced after calcination, as previously seen in the case of Cat I, on the surface of zeolite grains.

3.1.3. Cat III (Figure 5.3): Co/HMOR pre-treated in flowing argon It was prepared by a thermal pre-treatment of the fresh exchanged material, in flowing argon. As a consequence, it maintains all the Co2+ pre-exchanged species in their exchange position, inside the pore of HMOR. As it will be seen hereafter, the comparison

152

Past and Present in DeNOx Catalysis Co3O4

PdO

Co2+

Cat I: Co-Pd/HMOR Pre-treated in air

Co2+

Cat II: Co/HMOR Pre-treated in air

Co2+

Cat III: Co/HMOR Pre-treated in ar

HMOR

Co3O4 HMOR

HMOR

Co3O4 SiO2

PdO

Cat IV: Co3O4/SiO2 Cat V: PdO/SiO2

SiO2

Figure 5.3. Schemes of the five model catalysts used in the present study [12].

of the catalytic activity of Cat II and Cat III gives a clear evidence of the role of external cobalt oxide Co3 O4 [12].

3.1.4. ‘Cat IV’: Co3 O4 supported over silica (Co3 O4 /SiO2  It was prepared by incipient wetness impregnation using Co(NO3 2 , 6H2 O, 99.9%, as cobalt precursor. Silica Aerosil 380 was purchased from Degussa. The sample was then dried at 573 K (300 C) and subsequently calcined at 773 K (500 C) under synthetic air. The quantity of cobalt introduced was 2 wt.%.

3.1.5. ‘Cat V’: PdO supported over silica (PdO/SiO2 ) It was prepared by incipient wetness impregnation using Pd(NH3 4 Cl2 . The sample was then dried at 573 K (300 C) and subsequently calcined at 773 K (500 C) under synthetic air. The amount of palladium introduced was 0.5 wt.%.

3.2. Model catalyst major role In order to determine the major catalytic activity of the preceding model catalyst, in the three functions of the model, the three reactions were studied separately on each catalyst (Table 5.1). The comparison of the results permits to identify the most active site, for each function, when the complete ‘Cat I’ CoPd/HMOR catalyst is working.

Three-Function Model Reaction

153

Table 5.1. Major role and function of model catalysts [10,12] Model catalyst Co/HMOR (Cat III) PdO/SiO2 (Cat V) Co3 O4 /SiO2 (Cat II, Cat IV)

Feed

Major role

Main function

NO/He1 CH4 /O2 2 NO/O2 3

NO adsorption HC oxidation NO to NO2

F3 F2 F1

1000 ppm NO in helium, flow rate: 50 cm3 min−1 , T = 35 C (308 K) CH4 400 ppm + 5 vol.% O2 in helium. VVH: 62 000 h−1 3 400 ppm NO + 5 vol.% O2 . VVH: 62 000 h−1 1 2

3.2.1. Function 3 Table 5.1 shows that the major NO adsorption is occurring over Co/HMOR (Cat III) when compared to PdO/SiO2 and Co3 O4 /SiO2 . This adsorption on Co cationic sites, located inside the zeolite (Cat III) leads [during temperature-programmed desorption (TPD) studies (Figure 5.5a, [12])] to the desorption of NO, and production of both N2 O and N2 . The formation of N2 (function 3) shows that the reduction of NO to N2 goes through its decomposition, with N2 O being the adsorbed intermediate, able to suffer a desorption at low temperature according to Figure 5.2.

3.2.2. Functions 2 and/or 3 The oxidation of methane (functions 2 and/or 3) is mainly occurring over PdO when compared to the activities of [Co3 O4 + Co2+ /HMOR] (Cat II) and Co2+ /HMOR (Cat III).

3.2.3. Function 1 The major oxidation of NO to NO2 is occurring over Co3 O4 (Cat IV) when compared to the activities of Co2+ /HMOR (Cat III), and PdO/SiO2 (Cat V).

3.3. Catalytic activity measurements This chapter reports the results from transient experiments (mainly, TPD or TPSR) coupled with on-line analysis of reaction mixture at the outlet of a well-stirred reactor. It means that the gas composition detected at the outlet of the reactor is in contact with the catalyst inside the reactor. Catalytic runs in isothermal conditions were also proceeded in order to avoid strong adsorptions of reactants or intermediates. Catalytic runs were performed after a pre-treatment of catalysts in flowing hydrogen, synthetic air or nitrogen, at 773 K. After returning to room temperature (RT) and flowing the following reaction mixtures: • NO (400 ppm), CH4 (400 ppm), 8 vol.% O2 , in the case of DeNOx over CoPd/HMOR [12], • 200 ppm NO/9 vol.% O2 /1000 ppm CH3 OH, total flow rate: 250 cm3 min−1 , VVH: 50 000 h−1 , in the case of DeNOx in the presence of methanol over Co/Al2 O3  [26],

154

Past and Present in DeNOx Catalysis

Micro-GPC (for N2) NO, NO2, NOx

N2O

Microreactor

Feed

CO, CO2 HC (FID)

Plasma reactor

GC-MS

Figure 5.4. Catalytic device for DeNOx reaction coupled with non-thermal plasma.

TPSR were carried out from RT to 770 K, with a heating rate of 10 K min−1 . Before TPD, catalysts were pre-treated as above, then flushed by reactant (generally NO in either N2 or O2 , and flushed by N2 to remove any physisorbed species at RT. Experiments were carried out in a U-type quartz reactor. The sample (0.025–0.2 g) was held between plugs of quartz wool and the temperature was monitored through a WET 4000 or Eurotherm 2408 temperature controllers. Reactant gases were fed from mass flow controllers (Brooks 5850TR). The reactor outflow was continuously analyzed by a set of detectors (Figure 5.4): • A ‘NOXMAT CLD 700 AL’ or a ‘Thermo Environmental Instruments 42CHL NOx Chemiluminescence analyzer’ for NO, NO2 and NOx (NO + NO2 . • An Ultramat 6E IR analyzer to monitor N2 O. • An Ultramat 6E to follow CO/CO2 . • A Fidamat 5E-I to follow the total concentration in HC (and Cx Hy Oz  with a FID detector. • A GC/MS equipped with a GC 6890N, Agilent Technologies, using a ‘Fused Silica – CP7351’ Varian Column, a Chemstation, Agilent Technologies and a Mass Spectral Libraries NIST Rev. D.03.00, Agilent Technologies for analyzing all organic and organic nitroso compounds. • A micro-GPC ‘G2890A’ from Agilent, for quantitatively measuring N2 . Signals from detectors were monitored on-line by Virtual Bench-Lab View computation programs or labVIEW 7 programs. Treatments of data were done using Origin 6.1.

Three-Function Model Reaction

155

The limits of detection were considered to be 1 ppm for NOx and HC, 3 ppm for N2 O and less than 5 ppm for CO/CO2 .

4. EVIDENCE FOR ELEMENTARY STEPS OF FUNCTION 3 ON Co2+ /HMOR (CAT III) Function 3 can be summarized as follows: HCxOy + 2 NO

N2 + H2O + CO2

Function 3

(1) NO dissociation, via an adsorbed dinitrosyl ‘(NO)2 ’ leading to N2 formation and desorption + 2 Oads (ex-NO) (2) Recovering of the active site by scavenging of Oads species by HCx Oy (total oxidation to CO/CO2 

4.1. ‘NO’ dissociation to N2 , during TPD of NO pre-adsorbed at RT on the active site of the third function catalyst: supported Co2+ on HMOR (Cat III, function 3 alone) selected for the sake of demonstration [12] Figure 5.5a corresponds to a catalyst presenting only supported Co2+ over HMOR, the active site of function 3 [10,12]. It shows that the pre-adsorbed NO is able to suffer a decomposition/reduction to N2 at two different temperatures: N2 is clearly detected at about 100 C and N2 + N2 O between 280 and 370 C. The reduction of NO to N2 is already done in these conditions: 2 NOads = N2 + 2 Oads

(2)

‘There is no need for a reductant for this step’. It can be seen that methane is not consumed: it plays the role of an inert gas. This result corresponds to already published data of Li and Armor [25]. As can be seen in Figure 5.5a, the N2 O intermediate can also desorb before dissociating, even at high temperature. The reason is that the temperature is too low to get the scission of the second and remaining N−O bond of N2 O for obtaining N2 : adsorbed N2 O desorbs before reacting to give N2 (Figure 5.2) [13]. Only a fraction of adsorbed N2 O goes to N2 .

156

Past and Present in DeNOx Catalysis (a)

TPD

MS signal intensity/a.u.

2.5

NO N2 N2O CH4

2.0 N2

Zone 1

1.5

Zone 2

1.0

NO

CH4

0.5

N2

NO

0.0

Co/HMOR 3rd function alone

Zone 3

100

200

300

400

500

Temperature/°C (b)

T(TPD) = T(DeNOx)

TPSR

Concentration/ppm

600

NO NO2 CH4

500 400 NO

CoPd/HMOR 3 functions

300 200

Zone 2

Zone 1

CH4

Zone 3

100 NO2

0

100

200

300

400

500

Temperature/°C

Figure 5.5. (a) Temperature programme desorption (TPD) of NO pre-adsorbed alone at RT in the absence of oxygen (function 3 alone: Co/H-MOR ‘Cat III’) (b) Temperature programme surface DeNOx reaction (TPSR) over a complete three-function catalyst CoPd/HMOR ‘Cat I’ (NO (400 ppm), CH4 (400 ppm), 8 vol.% O2 , 93 000 h−1 , Heating rate TPSR:10 K min−1 ) [12].

The elementary steps for this first part of catalytic reaction are: i (1) Dinitrosyl formation on the metal cation 2 NO + ∗ → NO ∗ NO

1

(2) 1st N−O bond scission of dinitrosyl, 1st remaining Oads NO ∗ NO → N2 O ∗ O (3) 2nd N−O bond scission, 2nd remaining Oads N2 O ∗ O → N 2 + O ∗ O Balance  2 NO = N2 + O ∗ O

1 1

where ∗ is the cation cus, and i the stoichiometric number of i step. Let us note that N2 Oads can desorb before dissociation (Figure 5.2) [12,13]. These elementary steps perfectly represent the phenomena observed on Figure 5.5a.

Three-Function Model Reaction

157

It is important to note that the temperature of NO desorption on ‘Cat III’ (F3 alone), at 340 C (613 K), corresponds to: • The temperature of thermal activation of NO on the active site (significance of TPD experiments, Figure 5.5a). • The temperature where NO is able to dissociate on the Co2+ active site (cus). • The temperature of DeNOx reaction (function 3): comparing Figure 5.5a (NO TPD, ‘Cat III’) to Figure 5.5b – TPSR in the presence of a three-function catalyst (CoPd/HMordenite, ‘Cat I’), in a complete flowing feed NO/HC(CH4 /O2 (excess) – the temperature of DeNOx is that of the NO thermal desorption. According to the model, the catalyst will have to produce Cx Hy Oz (CH3 OH, HCHO) (function 2) to proceed to the DeNOx process, as discussed in Section 4.2. • The prediction of the DeNOx temperature range can therefore be done and it should correspond to the temperature at which the HC is activated by a mild oxidation to Cx Hy Oz [109] (see Section 4.2). It will be also verified, in this chapter, on other catalytic systems, that the temperature at which NO dissociates is the temperature of the DeNOx reaction (see Section 7).

4.2. The catalytic cycle of function 3 is able to turn over, when flowing the activated form of HC (Cx Hy Oz : alcohol, aldehyde, etc.) according to the model of Figure 5.2 4.2.1. Case of CoPd/HMOR (three-function catalyst ‘Cat I’) [12] The CoPd/HMOR three-function catalyst is able to produce by itself mild oxygenates of methane, CH3 OH and HCHO, above 100 C (373 K), as seen in Figure 5.6. It can be NO 400 ppm + CH4400 ppm + 5vol% O2 During TPSR 10 K.min–1

3

Formaldehyde

2

4 2

1

0 0 –2

Methanol

MS signal of CH3OH/a.u.

MS signal of HCHO/a.u.

6

–1 100

200

300

400

500

Temperature/°C

Figure 5.6. Evidence for Cx Hy Oz species during transient DeNOx reaction in the presence of Methane over CoPd/HMOR (‘Cat I’).

158

Past and Present in DeNOx Catalysis (b) 200

200

150

150

NOx , NO, NO2 /ppm

NOx , NO, NO2 /ppm

(a)

183°C, 116 ppm

NOx

100

181°C, 76 ppm

NO 182°C, 43 ppm

50

NO2

0

100

200

460°C, 30 ppm 455°C, 24 ppm 475°C, 6 ppm

300

Temperature/°C

400

500

NOx NO NO2

100

50

19 ppm 50 100 150 200 250 300 350 400 450 500 550

Temperature/°C

Figure 5.7. TPD and TPSR over -Al2 O3 alone (a) TPD of NO pre-adsorbed at RT, in the presence of oxygen (function 3 alone). (b) DeNOx activity of -Al2 O3 in the presence of CH3 OH (total flow rate: 250 cm3 min−1 , catalyst weight: 0.2 g, VVH: 50 000 h−1 , 200 ppm NO/9 vol.% O2 /1000 ppm CH3 OH) [26].

observed, above 280 C (553 K), that both consumptions of these oxygenates and DeNOx reactions are occurring.

4.2.2. Case of -Al2 O3 alone: role of Cx Hy Oz on function 3 [26] Following a ‘step-by-step’ methodology, and according to the previous prediction of DeNOx temperature by TPD of NO, Figure 5.7a shows that the reaction can be expected either at 200 C or in the 400–500 C temperature range. Nevertheless, according to the model, the reaction will occur only if the catalyst is able to produce, at the favourable temperature, the appropriate Cx Hy Oz necessary for the function 3 to work. Methanol is not active at 200 C (473 K) (there is no total oxidation by ‘O species’ adsorbed at this temperature), but it appears to be a good ‘reductant’ at high temperature, i.e. methanol is scavenging oxygen species left by previous NO decomposition (Figure 5.7b). The NOx conversion over -Al2 O3 in the presence of CH3 OH (function 3) is shown in Figure 5.7b. The reaction proceeds through successive isotherms, to avoid any preadsorption of reactants or intermediates. A consumption of NOx (NO + NO2  higher than 50% is observed between 350 and 500 C (623 and 773 K), i.e. in the ‘high-temperature window’ predicted by TPD in Figure 5.5a. At 400 C (673 K), the NOx consumption is maximum and is equal to 90%. Let us note that there is no supported metal in the material, the active cus being a cation, probably Al3+ unsaturated site.

4.2.3. Case of 0.5 wt.% Co/-Al2 O3 DeNOx in the presence of CH3 OH [26] When 0.5 wt.% Co is supported over -Al2 O3 (Figure 5.7) – cobalt being mainly dispersed as CoO [26] – the ‘consumption’ of NOx is found to be higher than 50% between

Three-Function Model Reaction

159

100 NOx

80

? = GC/SM

80

?

60

60 N2

40

40

20 0 100 150

Conversion ≈ N2 200

250

300

350

400

450

20

0 500 550

Consumption of NOx (%)

Conversion of NOx to N2 (%)

SUCCESSIVE ISOTHERMS 100

Temperature/°C

Figure 5.8. Formation of N2 during DeNOx reaction (feed: 200 ppm NO, 1000 ppm CH3 OH, 9% O2 and Ar complement) over 0.5 wt.% Co2+ / -Al2 O3 . N2 has been measured by micro GPC. The reaction was conducted by successive isotherms. ‘?’ = acetonitrile [26].

225 and 425 C (498 and 698 K) and 100% (total conversion) between 300 and 350 C (573 and 623 K). For a temperature higher than 400 C (673 K), no inhibition of the reaction by water was observed (not shown here). Water molecules do not compete on active sites (Al3+ and Co2+ ) with NO, methanol or HCHO oxygenates resulting from the interaction between CH3 OH and NO2 (function 2).

4.2.4. Is the consumption of NOx corresponding to the DeNOx reaction, i.e. to the formation of N2 ? [26] In the case of Al2 O3 alone, the higher conversion of NO to N2 was 50% at 500 C (773 K) (not shown here). In the case of 0.5 wt.% Co2+ /Al2 O3 alumina, it can be observed in Figure 5.8 that the plot of formation of N2 does not overlap that of NOx conversion for T < 400 C (673 K). The highest NOx conversion to N2 is 73% at 350 C (623 K). Furthermore, the selected zone between 400 and 500 C (673 and 773 K) shows that the selectivity to N2 is 100%, the two previous plots overlapping. For temperature lower than 400 C (623 K), the difference between the two plots, based on NOx concentration, was found to correspond to the formation of acetonitrile (GC/MS detection).

4.3. Conclusion on function 3 Similar results have been obtained over alumina alone, in the presence of propene [27]. The initial HC of the feed (propene) has to first transform to oxygenates (alcohol, aldehyde, etc.) – simultaneously to the NO decomposition (function 3) – to scavenge adsorbed oxygen species left by NO decomposition and regenerate the active sites of function 3. The mild oxidation of HC to oxygenates is the role of function 2 of the present model.

160

Past and Present in DeNOx Catalysis

5. FUNCTION 2: ACTIVATION, – MILD OXIDATION –, OF HC (INITIAL REDUCTANT) BY NO2 Function 2 can be summarized as follows:

NO2 + HC

Cx Hy Oz + NO

Function 2 via RNOx

The net reaction for a Hy Cx hydrocarbon is: Hy Cx + NO2 = Cx Hy O + NO

(3)

The interaction between HC [the initial and ‘global’ reductant in reaction (1)] and NO2 leads to the formation of an organic nitrogen-containing compound (RNOx , which subsequently suffers decomposition, releasing some ‘Cx Hy O’ oxygenate and NO. This is the crucial point: we shall see that NO is formed, but not N2 in this fundamental step. According to literature, one of the models [6] considers the following global scheme: HC + NOx → RNOx → N2 where dinitrogen comes from the direct ‘decomposition’ of RNOx (see Martens et al. [6,8], Section 1). In contrast, the present model is based on an experimental result, showing that the ‘decomposition of RNOx in the presence of oxygen’ leads to oxygenates Cx Hy Oz , necessary for function 3 to turn over, with the release of NO and not N2 [5,28]. (Let us note that the generalization of the present model to the oxidation of carbon particulates by NO2 leads to the same result: carbon partial oxidation and release of NO [31]). At this level of discussion, some confusion can be done on the origin of N2 formation during what is often considered as the ‘oxidation of RNOx releasing N2 ’. The present explanation has two parts: (1) The RNOx decomposition leads to the formation of oxygenates Cx Hy Oz with release of NO (function 2). (2) Furthermore, any efficient catalyst, in DeNOx reaction, has to more particularly present functions 2 and 3 turning over simultaneously. As soon as Cx Hy Oz species (alcohol, aldehyde, etc.) are formed, by function 2, they initiate the rotation of the catalytic cycle corresponding to function 3. So, it is this last cycle which releases N2 as shown experimentally, simultaneously with RNOx decomposition in another cycle. These two reactions do not occur in the same catalytic cycle.

Three-Function Model Reaction

161

The present chapter intends to demonstrate experimentally, with even more details, all these phenomena. The basis of the demonstration can be based on already published data on the surface reaction between NO2 and adsorbed organic compounds. Yokoyama and Misono have shown that the rates of NO2 reduction over zeolite or silica are proportional to the concentration of adsorbed propene [29], whereas Il’ichev et al. have demonstrated that NO2 reacts with pre-adsorbed ethylene and propylene on H-ZSM-5 and Cu-ZSL-5 to form nitro-compounds [30]. Chen et al. [2–4] have observed the same nitrogencontaining deposits on MFI-supported iron catalysts. The question on the pairing of nitrogen atoms is not considered here. A more evident demonstration has been done on supported precious metals [5,28]. The interaction between NO2 and HC is clearly located in the second cycle (Figure 5.1). Baudin [32] have detected the RNOx and showed the decomposition of RNOx to Cx Hy Oz over an Ir/CeZrO2 catalyst. The dinitrogen is formed in the third cycle and is not linked to any organic-nitro compound. Following Yokoyama and Misono, Il’ichev et al. and Chen et al. (Figure 5.9) show the decomposition of an RNOx compound pre-synthesized over an Ir/CeZrO2 catalyst, where iridium is cationic. A first adsorption step was done at RT by flowing the following gas mixture: NO (150 ppm), C3 H7 OH (550 ppm C), O2 (8 vol.%), complement: N2 . The RNOx results in the interaction between adsorbed propan-2-ol and NO2 formed and adsorbed. RNOx are formed and stored on the catalyst surface. C3 H7 OH has been selected for simulating the

100

200

300

400

500

NO, NO2, N2O (ppm)

800

Production : NO, NO2, N2O FID signal CO, CO2

700 600

800 700

NO

600

NO

500 400

NO2

N2O

200 100

NO

200

300

300 200 100

NO2

0

100

Decomposition step RNOx → Cx Hy Oz + NO

400

N2O

300

500

0

400

500

400

500

T (°C)

Signal FID, CO, CO2 (ppm)

2000 1800 1600

100

200

300

CO2

1200

600 400 200

1400

- C3H7OH (550 ppm C)

1200

- O2 (8%)

1000 800

CO2

Signal FID

TPSR (10 K/min) - NO (150 ppm)

1600

CO2

1000 800

1800

FID signal (Cx Hy Oz )

1400

2000

- N2

600 400

CO

CO

200

0

0

100

200

300

400

500

T (°C)

Figure 5.9. Concentration (ppm) vs. T; pre-adsorbed RNOx decomposition: RNOx = Cx Hy Oz + NO; TPSR (10 C min−1 ), NO (150 ppm), O2 (8 vol.%), complement: N2 . First plot: N-containing species; second plot: FID signal including Cx Hy Oz and CO, CO2 species [32].

162

Past and Present in DeNOx Catalysis

reaction of RNOx decomposition, as it is one of the Cx Hy Oz corresponding to the mild oxidation of propene by NO2 , as shown elsewhere [32]. A second step concerns the TPSR in the presence of flowing NO (150 ppm), O2 (8 vol.%), complement: N2 , in the absence of propanol, to simulate a TPSR reaction in the presence of NO/C3 H7 OH/O2 , where NO is oxidized to NO2ads reacting with C3 H7 OHads to lead to RNOxads . During TPSR, it can be seen that the decomposition of the nitro compound is quite violent and leads to the formation of NO (Figure 5.9, NOx graph). Let us consider the following reaction: RNOx = Cx Hy Oz + NO This reaction is one of the steps in cycle 2 (Figure 5.1), when starting with the initial HC of the feed. The two-step sequence is: HC + NO2 → ‘RNOx ’ ‘RNOx ’ → Cx Hy Oz + NO Net reaction: HC + NO2 = Cx Hy Oz + NO where ‘RNOx ’ is the surface intermediate. This reaction can also occur without catalyst, in the gas phase [5,12]. RNOx is an intermediate of cycle 2 (not 3!), leading to NO and not to N2 . The simultaneity between cycles 2 and 3 can lead to the confusion that N2 comes from the RNOx decomposition, as often assumed in literature for the N−N pairing. For the sake of simplicity, a CeO2 −ZrO2 (70/30) mixed oxide will be now used as material. This mixed oxide has been previously shown to be able to proceed to three-way catalysis, the general concept for N2 formation over a metal cation being the same: NO decomposition and oxygen species scavenging, in stoichiometric conditions, by CO as reductant [10,11].

5.1. Detection of Cx Hy Oz and RNOx compounds by GC/MS during DeNOx reaction over CeZrO2 alone [32] It is now quite well recognized, at international level, that NO2 produces a mild oxidation of HC to Cx Hy Oz , at a temperature lower than that of its total oxidation by O2 . Figure 5.10 shows that CeO2 −ZrO2 (70/30) mixed oxide alone is able to proceed to DeNOx reaction. Its activity is quite low [8–10% at 275 C (548 K)], but it shows that a cation is able to proceed to the reaction. Figure 5.11 more particularly shows that oxygenates are detected by GC/MS, as well as RNOx species. Decomposition of RNOx intermediate leads to oxygenates: Baudin et al. [28,32] have shown the simultaneous formation of ‘NO’ for a reaction over Ir/CeZrO2 . The corresponding reaction: Propene + NO2 = Cx Hy Oz + NO has been also found to occur in the gas phase [28].

(4)

Three-Function Model Reaction

163

NO conversion %

100 80 GC/MS analysis

60 40 20 0 100

200

300

400

500

Temperature/°C

u.a.

Figure 5.10. DeNOx reaction over CeZrO2 alone (500 ppm NO; 2000 ppmC1 of C3 H6 ; 8 vol.% O2 ). Successive isotherms [32].

H

O

GC/MS – T = 225°C

H

+ H

RNOx

O H

O

N

O O

4

6

8

10

12

14

Time/min

Figure 5.11. TPSR DeNOx reaction over CeZrO2 ; (200 ppm NO, 1000 ppm CH3 OH, 9% O2 . GC/MS analysis at 225 C as indicated on Figure 5.7 [32].

Berger has shown that NO2 produced over a three-function CoPd/HMOR (Cat I) also produces methanol and formaldehyde at a temperature as low as 120 C (393 K) during the DeNOx reaction in the presence of methane [11,12] (Figure 5.7).

5.2. Conclusion on function 2 Catalyst needs NO2 to proceed to the mild oxidation of HC to oxygenates such as alcohol or aldehyde, avoiding the total oxidation of HC to CO, CO2 /H2 O. As we shall discuss later, alumina is not a good catalyst for oxidation, but can be a good candidate for such a purpose. So, catalyst also needs to oxidize NO to NO2 : this is function 1, which clearly have to ‘turn over’, according to the model, simultaneously with the two other functions 2 and 3, to get an optimized catalyst.

164

Past and Present in DeNOx Catalysis

6. FUNCTION 1: CATALYTIC OXIDATION OF NO TO NO2 Function 1 can be summarized as follows:

2NO + O2

2 NO2

Function 1

The global reaction is given as follows: 2 NO + O2 = 2 NO2

(5)

In the present case, the system needs an active site able to oxidize NO without total oxidation of HC. Platinum has to be avoided and an oxide can be considered. Let us note that some supports are also able to proceed to NO oxidation (CeZrO2 , even alumina, etc.). This reaction has to occur to a sufficient degree of conversion to get sufficient amount of NO2 , and it has to occur at the same temperature as functions 2 and 3. Park [26] has studied the NO oxidation over Co3 O4 /Al2 O3 . These results will be presented for demonstration. Figure 5.12 shows that the NO oxidation without catalyst is very slow, whereas thermodynamics shows that 100% conversion can be obtained up to about 200 C (473 K). Alumina alone has a very low activity. Higher the amount of cobalt, higher is the density of sites and higher the NO2 production. Let us note that the reaction can be done without platinum.

No conversion to NO2 %

100

Thermodynamics 80 60

2wt %Co

40

1wt %Co 20 0

0.5wt %Co Alumina 0

100

200

300

400

500

Temperature/°C

Figure 5.12. Catalytic oxidation of NO to NO2 over 0.5, 1, 2 wt.% Co/Al2 O3 . Successive isotherms. 200 ppm NO/9 vol.% O2 /Ar; total flow rate: 250 cm3 min−1 ; catalyst weight: 0.2 g; VVH: 50 000 h−1 [26].

Three-Function Model Reaction

165

6.1. Partial conclusion on the three functions A partial conclusion can be drawn at this stage of presentation. A larger number of works have been published in literature and the present model can explain the majority of data. An extensive literature review has been already done by Gorce et al. and will not be considered again in this chapter [5]. On this basis, the three functions can be studied separately. A concept of ‘composite catalyst’ can be considered; using three catalysts, each one bringing one of the three functions of the model. Furthermore, let us remember that one material can also bring one or two functions. The main difficulty is to get the three catalytic cycles – associated to the three functions – turning over simultaneously. The KOCAT Society (no patent reference known by authors), in South Korea, solved the DeNOx problem of big burners, by directly injecting oxygenates on the catalyst at the outlet of the burner. This process involves the third function of our model. This example shows that only one model (the present one) for DeNOx reaction can be used for either mobile or stationary sources. Pathways are the same; what is changing is the nature of the reductant, which has to be activated, through its partial oxidation, at the temperature when N−O bonds (dinitrosyl species) are broken. What does it mean? As it is very difficult to find the best design of material, for the catalyst to simultaneously initiate the three functions by itself, an external device can be developed to substitute functions 1 and 2, providing the catalyst the ‘good’ oxygenated species, for the full range of temperature. There are different ways to develop such an idea. One of them is the non-thermal plasma-assisted DeNOx reaction. It will be time consuming – and there is no room in the present chapter – to recall all results already published on this topics by the Society of Automotive Engineers (SAE) Congresses. One of the earlier works, by Penetrante and co-workers [33], has shown that the coupling of a non-thermal plasma reactor, in front of alumina catalyst, was able to produce a significant DeNOx of the feed. They did not give, in their paper, any interpretation for the observed results, but published the composition of the stabilized gas mixture at the outlet of the plasma reactor, just before the catalytic reactor containing alumina. The gas mixture contained more oxygenated compounds. Hoard and Balmer [34] and Doraï and Kushner [35] have also found Cx Hy Oz (CH2 O, CH2 O2 , NO2 and RNOx (CH3 ONO2 , which are ‘intermediates needed for functions 1 and 2 and, furthermore, for the third function itself. Alumina alone, similar to Co/HMOR [Figure 5.5a, reaction (2)], presents two peaks of NO desorption (see hereafter Figure 5.15a): one is at low temperature, 175–325 C (448–598 K) and the second one is at 400–550 C (673–823 K). According to Section 4.1 and reference [10], it can be predicted, for alumina, that: (1) NO is able to dissociate and produce N2 at low temperature (function 3). (2) DeNOx should occur as soon as active reductants are delivered to alumina at the temperature when NO dissociates, to scavenge Oads left by NO (function 3). By coupling a non-thermal plasma reactor to a catalytic reactor containing alumina alone, Baudin [32] have observed the DeNOx function of alumina at low temperature,

166

Past and Present in DeNOx Catalysis

confirming the present model. In this case, the non-thermal plasma plays the role of both the functions 1 and 2, as will be discussed in next section.

7. NON-THERMAL PLASMA-ASSISTED CATALYTIC NOx REMEDIATION, FOR SUBSTITUTING FUNCTIONS 1 AND 2 OF THE MODEL. ACTIVATION OF THE LOW-TEMPERATURE DeNOx -FUNCTION 3 OF ALUMINA [32–38] 7.1. Effect of a dielectric barrier discharge (DBD) type non-thermal plasma on a synthetic gaseous reaction mixture The synthetic mixture contained: C3 H6 (2000 ppm C1  – NO (500 ppm) – O2 (8 vol.%) – N2 , energy density: 36 J L−1 , voltage: 14 kV, VVH: 54 000 h−1 [32]. The resulting gas mixture at the outlet of the plasma reactor is then flowing through the catalytic reactor. Figures 5.13 and 5.14 show the nature of gas species at the outlet of the plasma reactor, in the absence of catalyst.

7.1.1. Plasma plays the role of function 1 The NO oxidation to NO2 is already occurring at RT. In the experimental conditions of Figure 5.13, total NO oxidation is observed between 180 and 270 C (453 and 543 K). Above 280 C (553 K), the system follows thermodynamics.

7.1.2. Plasma plays the role of function 2: formation of Cx Hy Oz and RNOx compounds Figure 5.14 reports the GC/MS analysis of organic species at the outlet of plasma at 150 C (423 K).

Concentration/ppm

500

NOx

400

NO2 300

GC/MS GC/MS

200 100

NO 0 0

50

100

150

200

250

300

350

Temperature/°C

Figure 5.13. NO oxidation to NO2 in a DBD non-thermal plasma [32].

400

CIT

Three-Function Model Reaction

167

O

O2 O

N2 H

+

O

150°C – Plasma ON

O H3C O N

O

H H

O O

H

H3C

N

O

O

O H3C

NH2

ON

N O

O O N O

OH N O

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

18.00

Time/min

Figure 5.14. Detection of organic products oxygenates and organic nitrogen-containing compounds at the outlet of non-thermal plasma reactor without catalyst. Feed: NO (500 ppm) – C3 H6 (2000 ppm C1  – O2 (8 vol.%) – N2 [32].

It is clearly shown, as already found by Penetrante and co-workers [33], Hoard and Balmer [34] and Doraï and Kushner [35], that the plasma reactor is able to produce oxygenates and RNOx from RT to 400 C (673 K). Those species correspond to function 2 and they are necessary for the DeNOx reaction according to the present model.

7.2. Plasma-assisted DeNOx catalysis: case of alumina alone (Figure 5.15) [32]. 7.2.1. Plasma ‘OFF’: high-temperature DeNOx The feed composition has been given just before, for a VVH = 54 000 h−1 . Figure 5.15b, in ‘plasma OFF’ condition, shows that DeNOx is occurring only at high temperature. The reaction is occurring at the temperature of the 2nd peak of NO activation over alumina [Figure 5.15a, 10% conversion for T > 425 C (998 K)].

7.2.2. Plasma ‘ON’: activation of the low-temperature function 3 of alumina When plasma is ‘ON’, the NOx conversion is rising from 2 to 40%, at 280 C (553 K). Figure 5.16 shows the presence, at the outlet of the catalytic reactor, of unreacted Cx Hy Oz and RNOx species, demonstrating that the plasma substitutes the catalytic function 2 of the DeNOx process. Theses compounds are separately delivered in the full range of temperature, at the very beginning of the temperature-programmed reaction. This is quite different from

168

Past and Present in DeNOx Catalysis

NO, NO2, NOx , N2O/ppm

(a) 140 120 100 80 60

NO

40 20

NO2

N2O 0 100 150 200 250 300 350 400 450 500

(b)

NOx Conversion %

NOx

“T” NO activation over AI2O2

Temperature/°C 80 60 Plasma ON

Plasma OFF

40 20 0 100 150 200 250 300 350 400 450 500

Temperature/°C

Figure 5.15. Non-thermal plasma-assisted DeNOx reaction over alumina alone, according to Baudin [32]. (a) NO, NO2 , NOx and N2 O TPD plots. (b) NOx conversion vs. T without plasma.

the catalytic process, where functions 1 and 2 have to be activated at the temperature where function 3 begins to work. Consequently, the plasma is producing Cx Hy Oz at a sufficiently low temperature, including a selection of Cx Hy Oz (Figure 5.16) able to activate the low-temperature function 3 of alumina (Figure 5.15b). This activation is linked to the 1st peak of NO desorption, as predicted by our methodology (Figure 5.15a).

7.2.3. Sequence ‘plasma ON’ (low temperature) – ‘plasma OFF’ (high temperature) in the presence of a mixture of HCs representative of diesel engine exhaust Figure 5.17 shows that a mixture of HCs and a VVH of 18 000 h−1 corresponding to a volume of alumina three times higher than the precedent case (Figure 5.15b) leads to a NOx conversion between 50 and 60%, in the temperature range 180–425 C (453–698 K), plasma being ‘OFF’ for temperature higher than 320 C (593 K). The reason for such a process is that at higher temperature, HCs play their role. Let us note that DeNOx was followed by N2 quantitative measurements for every 3 minutes. The preceding results suggest an advantageous plasma–catalyst coupling effect on the NOx remediation, in full accordance with the proposed mechanism [38]. The Cx Hy Oz and RNOx compounds, produced by the non-thermal plasma before the catalytic reactor

Three-Function Model Reaction

169

Conversion NOx %

Main gas phase species (GC/MS Analysis) 100

25% 25 %

80

Plasma (propene)/Alumina coupling On-line GC/MS at 215°C (DeNOx : 25%)

60 40 20 0 100 150 200 250 300 350 400 450 500

Temperature/°C

CxHyOz O

90 ⋅ 10

4

O

H3C O N

O O

RNOx

CIT

O

45 ⋅ 104

H3C

N

HC

O N

O

O O O

O O

2

4

N

O

6

8

HO

10

N

12

14

16

O O

18

Time/min

Figure 5.16. Non-thermal plasma-assisted DeNOx reaction over alumina alone, according to Baudin [32]. GC/MS analysis of outlet gases at 215 C (488 K) and 25% NOx conversion (Figure 5.14b). Feed: n -C10, C7H8, C3H6, C3H8, Ar 100

DeNOx = N2 (μ-GC) For T > 200°C

Al2O3

NOx Conversion %

80

60 Plasma OFF

40 Plasma ON

20

0 100

150

200

250

300

350

400

450

500

Temperature/°C

Figure 5.17. Non-thermal plasma-assisted DeNOx reaction over alumina alone, according to Baudin [32]. Reaction in the presence of a mixture of HCs (see feed composition on the top of the figure). VVH = 18 000 h−1 .

with alumina, demonstrate the strategy of the model at both low (plasma ‘ON’) and high (plasma ‘OFF’) temperatures, as well as the equivalence of the role of plasma and that of functions 1 and 2 of the catalyst, when non-plasma assisted.

170

Past and Present in DeNOx Catalysis

8. CONCLUSION AND METHODOLOGY It has been chosen, for presenting the three-function model, to start from the true ‘DeNOx catalytic cycle’ corresponding to function 3, which leads to the N−N bonding and N2 release from the catalyst. Subsequently, it appears that two other functions are necessary to assist function 3. Function 1, oxidation of NO to NO2 , can be studied separately [3926] and its temperature of activation is adjusted to the temperature at which function 3 is working. (It does not mean that the amount of NO2 , at the outlet of the reactor, will be the same in the presence of a reductant, as there is a consumption of NO2 during the interaction HC/NO2 ). Function 3 can be studied separately by direct injection of the Cx Hy Oz oxygenates (alcohol, aldehyde, etc.) corresponding to the mild oxidation process of HC by NO2 . It is rather difficult to study function 2 separately, as the catalyst generally presents, simultaneously, functions 1 and/or function 3. Nevertheless, the mild interaction of HC with NO2 can be approached through the direct NO2 /HC reaction, even in the absence of dioxygen. It has to be compared to the total oxidation of HC in the presence of oxygen, as it is a competitive reaction for the HC consumption. In contrast, it will be very interesting to compare the two catalytic pathways of elementary steps, and reaction intermediates, in both oxidation reactions (mild and total oxidation of reductant).

8.1. Some important features of the model can be summarized as follows • Simultaneity between the three catalytic cycles, as presented hereafter (Figure 5.18): The three cycles have to turn over in the same range of temperature. This catalytic approach of the DeNOx reaction is not new. There is the same process for isomerization of alkanes, where there are also 3 catalytic cycles which have to turn over simultaneously (bifunctional catalysis). The kinetics of isomerization is given by only one cycle, the other two turning over very rapidly and are near equilibrium [13].

NO + O2

NO2

NO + Hx Cy O F2 *RNOx

F1

Hy Cx

•N2 • CO, CO2 • H2O

F3

NO + O2

Figure 5.18. Catalytic assisted DeNOx reaction: each cycle corresponds to one function (F1, F2 and F3).

Three-Function Model Reaction

171

• Regulation of the total oxidation of ‘reductants’ (HC, Cx Hy Oz  by O2 , taking into account the difference of temperature between their mild oxidation by NO2 and their total oxidation by dioxygen. This point requires the choice of a total oxidation function, but not too much active. • If the preceding requirements are fulfilled, then the DeNOx process (function 3) does not need a large amount of reductant, as it is very often claimed: the stoichiometry of ‘2NO + Cx Hy Oz = N2 + xCO/CO2 + y/2 H2 O’ should be considered. Clearly, it is generally impossible to avoid the competition between the ∗ Oads left by NO and the ∗ Oads due to O2 dissociation, for the total Cx Hy Oz oxidation on function 3 (this competition corresponds to a kinetic coupling of at least two catalytic cycles, through Oads [13]). Both of them contribute to the total oxidation of reductants.

8.2. A methodology can be suggested from these conclusions, for designing the best efficient catalyst in DeNOx reaction For a given catalyst, in the framework of the present model, the proposal is as follows: [I] Prediction of the temperature where DeNOx can take place, by TPD of NO preadsorbed with or without oxygen. In the absence of oxygen, check the formation of N2 , N2 O, NO2 and NO during TPD. If N2 O and/or N2 are formed, it means that the reaction already took place. If not, the system needs the reductant to take place. This experiment also means that function 3 can work [10,25]. [II] Study of function 3, by studying the reaction NO/O2 /Cx Hy Oz (corresponding to the HC initially in the feed) [5,10,26,28]. [III] Checking of function 1: NO/O2 to NO2 . Check the domain of temperature compared to that of NO desorption (thermal activation). [IV] Function 2 can be approached in the absence of O2 , by the reaction HC/NO2 . If function 3 exists on the catalyst, then the reaction can partially proceed, when the temperature is rising, to the total oxidation of HC.

8.3. Finally, some targets and suggestions for the future can be done (1) If cycle 1 is easy to study (NO to NO2 ) [39], then pathways of functions 2 and 3 still remain to be established in detail. (2) Kinetics of cycles 2 and 3 have to be done. That of cycle 2 will probably have to take into account the presence of cycle 3. (3) What is the catalytic cycle able to drive the kinetics of the whole process (rate determining cycle)? (4) How to control the kinetics of the process, regulating the kinetics of two of the three cycles, to get only one rate determining cycle? (5) The design of a three-function catalyst, for a given application with specific reductants, will be easier in the framework of the model. (6) Considering the model, several possibilities to solve the problem of DeNOx on stationary or mobile sources can be defined, at different levels.

172

Past and Present in DeNOx Catalysis

The model can also be extended to three-way catalysis as far the active site is a cation of transition metal, the reductant being CO and the feed CO/NO/HC being very near stoichiometry [10,11].

ACKNOWLEDGEMENTS One of the authors (GDM) greatly acknowledges all engineers from industries (PSAPeugeot-Citroën, Renault, Gaz de France, ADEME) who worked with the ‘Laboratoire Réactivité de Surface’, as well as Michel Boudart for many fruitful discussions on DeNOx . The Group of Researchers who worked in the framework of the French-Polish ‘Jumelage’ (granted by CNRS, Polish Academy of Science, Polish Ministry of Scientific Research and Information Technology, French Foreign Office) on ‘Catalytic Materials for Environment’ are also greatly acknowledged for so many works and PhD thesis which contributed to applying and confirming our model.

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

[17] [18] [19]

[20] [21] [22] [23] [24]

Wang, X., Chen, H.-Y. and Sachtler, W.M.H. (2001) J. Catal., 197, 281. Chen, H.-Y., Voskoboinikov, T. and Sachtler, W.M.H. (1999) Catal. Today, 54, 483. Chen, H.-Y., Voskoboinikov, T. and Sachtler, W.M.H. (1999) J. Catal., 186, 91. Chen, H.-Y., Voskoboinikov, T. and Sachtler, W.M.H. (1998) J. Catal., 180, 171. Gorce, O., Baudin, F., Thomas, C. et al. (2004) Appl. Catal., B54, 69. Martens, J.A., Cauvel, A., Francis, A., et al. (1998) Angew. Chem. Int. Ed., 37, 1901. Kharas, K.C.C. (1993) Appl. Catal., B2, 207. Brosius, R. and Martens, J.A. (2004) Topics Catal., 28, 119. Iwamoto, M. and Takeda, H. (1996) Catal. Today, 27, 71. Djéga-Mariadassou, G. (2004) Catal. Today, 90, 27. Djéga-Mariadassou, G., Fajardie, F., Tempère, J.-F. et al. (2000) J. Mol. Cat., A: Chemical, 161, 179. Berger, M. (2000) Thèse de l’Université Pierre et Marie Curie. Djéga-Mariadassou G. and Boudart M. (2003) J. Catal., 216, 89. Ferreira, A.P., Capela, S., Da Costa, P. et al. (2007) Catal. Today, 119, 156. Ferreira, A.P., Henriques, C., Ribeiro, M.F. et al. (2005) Catal. Today, 107–108, 181. (a) Boudart, M. and Djéga-Mariadassou, G. (1982) Cinétique des Réactions en Catalyse Hétérogène, Masson, Paris. (b) Boudart, M. and Djéga-Mariadassou, G. (1984) Kinetics of Heterogeneous Catalytic Reactions, Princeton University Press: Princeton, NJ, USA. Iiuzuka, T. and Lunsford, J.H. (1980) J. Mol. Catal., 8, 391. Hendriksen, D.E., Meyer, C.D. and Eisenberg, R. (1977) Inorg. Chem., 16, 970. Cataluna, R., Arcoya, A., Scoane, X.L. et al. (1995) Catalysis and automotive pollution control III. In Studies in Surface Science and Catalysis (A. Frennet and J.M. Bastin, eds) 96, p. 215. Fajardie, F., Tempère, J.-F., Manoli, J.-M. et al. (1998) J. Catal., 179, 469. Manuel, I., Thomas, C., Bourgeois, C. et al. (2001) Catal. Lett., 77, 193. Manuel, I., Chaubet, J., Thomas, C. et al. (2004) J. Catal., 224, 269. Hamon, C., Le Lamer, O., Morio, N. et al. (1998) PCT WO 98/15339. Germain, J.E. (1969) Catalytic Conversion of Hydrocarbons, Academic Press: New York, p. 259.

Three-Function Model Reaction

173

[25] Li, Y. and Armor, J.N. (1991) Appl. Catal., 76, L1. [26] Park, J.-W. (2005) Thèse de l’Université Pierre et Marie Curie. [27] Lê Van, T., Thiet, N.-Q., Thomas, C. et al. (2004) Proc. of Global Symposium on Recycling, Waste Treatment and Clean Technology, "REWAS", 1, 785. [28] Baudin, F., Da Costa, P., Thomas, C. et al. (2004) Topics Catal., 30–31, 97. [29] Yokoyama, C. and Misono, M. (1996) J. Catal., 160, 95. [30] Il’ichev, A.N., Matyshak, V.A., Korchak, V.N. et al. (2000) Kinet. Catal., 41, 706. [31] Darcy, P. (2005) Thèse de l’Université Pierre et Marie Curie. [32] Baudin, F. (2004) Thèse de l’Université Pierre et Marie Curie. [33] Penetrante, B.M., Brusasco, R.M., Meritt, B.T. et al. (1998) SAE Technical Paper 982508. [34] Hoard, J. and Balmer, M.L. (1998) SAE Technical Paper 982429. [35] Doraï, R. and Kushner, M.J. (1999) SAE Technical Paper 01-3683. [36] Gorce, O., Jurado, H., Thomas, C. et al. (2001) SAE Technical Paper 2001-01-3508. [37] Khacef, A., Cormier, J-M., Pouvesle, J-M. et al. (2002) Hakone 8, P5.4. [38] Baudin, F., Schneider, S., Lendresse, Y. et al. (2004) Patent 2.877.693-O4 11882-A1. [39] Marques, R., Darcy, P., Da Costa, P. et al. (2004) J. Mol. Cat., 221, 127.

This page intentionally left blank

Chapter 6

IDENTIFICATION OF THE REACTION NETWORKS OF THE NOx STORAGE/REDUCTION IN LEAN NOx TRAP SYSTEMS P. Forzatti∗ , L. Castoldi, L. Lietti, I. Nova and E. Tronconi Dipartimento di Chimica, Materiali e Ingegneria Chimica ‘G.Natta’, Centro NEMAS – Nano Engineered MAterials and Surfaces, Politecnico di Milano, Milano, Italy ∗ Corresponding author: Dipartimento di Chimica, Materiali e Ingegneria Chimica ‘G.Natta’, Centro NEMAS – Nano Engineered MAterials and Surfaces, Politecnico di Milano, Milano, Italy. Tel.: +39 02 2399 3238, Fax.: +39 02 2399 3318, E-mail: [email protected]

Abstract The reaction network of the NOx storage and reduction over Pt−Ba/Al2 O3 ‘lean NOx trap’ catalysts is presented in this paper. The NOx storage was investigated at first. The collected results showed that a dual pathway is operating when starting from NO/O2 mixtures: the first route implies the wellknown oxidation of NO to NO2 , and its subsequent adsorption via disproportionation to form nitrates (nitrate route), whereas the second novel route consists of a stepwise oxidation of NO in the presence of oxygen to form nitrite ad-species, which are progressively oxidized to nitrates (nitrite route). Then the reduction of stored NOx with hydrogen is addressed. The bulk of data points out that the reduction of stored nitrates occurs under near isothermal conditions through a Pt-catalysed surface reaction that does not involve the thermal desorption of the stored nitrates as a preliminary step. A specific role of a Pt−Ba interaction was suggested, which plays a role in the NOx storage phase as well. The effect of the operating conditions (temperature, presence of CO2 and water) on both adsorption–reduction phases has been also analysed.

1. INTRODUCTION Nitrogen oxides, or NOx , is the generic term for a group of gases including NO (nitrogen monoxide), NO2 (nitrogen dioxide) and N2 O (nitrous oxide), which are currently considered among the most important air pollutants, together with ozone (O3 ), sulphur Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

176

Past and Present in DeNOx Catalysis

dioxide (SO2 ), particulate matter (PM) and carbon monoxide (CO) [1]. In fact, NOx , being transformed to nitric acid in the presence of water, is one of the main causes of acid rains and contributes to the so-called photochemical smog. Nitrogen oxides have also direct harmful effects on human health by causing the reduction of the breathing functionality and damages to the cardio-circulatory system. Furthermore, NO2 reacts with haemoglobin by forming meta-haemoglobin that causes several pathologies in children. For all these reasons, in the last few decades, the concentration of NOx in the atmosphere has been constantly monitored by the environmental agencies of the most developed countries, and the NOx emissions have been regulated by more and more restrictive legislations [1,2]. Even if part of NOx emissions derives from natural sources such as lightning and microbiological activities, a considerable amount of nitrogen oxides has got a human origin. In particular, NOx is generated during combustion processes, and major sources are thermal power stations (stationary sources) and vehicles (mobile sources): nowadays, they are blamed for a roughly equal contribution to anthropogenic NOx emissions [3]. However, the number of motor vehicles is expected to significantly increase in the future (values as high as 920 million are estimated by 2010 [3]), so that their contribution to global NOx emission is expected to grow. At the same time, the treatment of stack gases from stationary sources is becoming nowadays widely applied with good results in terms of NOx abatement. For such reasons, over the past few years, research has been focused primarily on the purification of automobile exhausts, also due to the very strict legislations established by US, Europe and Japan concerning the emissions from mobile sources [4]. Besides, NOx is also produced by other mobile sources (e.g. ships and locomotives), for which specific legislations are foreseen in the near future. A first major result of this effort was the development of the three-way catalysts (TWC) technology, which allows the simultaneous removal of CO, unburned hydrocarbons (HCs) and NOx by using a noble metal-based catalyst. The application of such a technology, which was introduced in the 1980s, allowed gasoline engines to comply with the emission standards. Leaded fuel was thus replaced by unleaded gasoline, in order to keep the catalyst away from lead poisoning. A requirement for automotive three-way catalysis is that the air-to-fuel (A/F) combustion ratio must be at the stoichiometric point, which is for a gasoline engine about 14.6 on a weight basis. However, higher A/F ratios would result in a decrease in fuel consumption (up to 30%) along with a welcome lower generation of CO2 , a well-known ‘greenhouse’ gas. This condition is achieved in lean-burn spark-ignited gasoline engines and diesel engines. Unfortunately, the TWC cannot reduce NOx in excess of air, and thus the challenge of the research is to develop catalytic systems able to remove NOx from lean burn and diesel engines. The technologies for the reduction of NOx under lean burn conditions, on which most of the research activities are focused, are the NH3 - or urea-selective catalytic reduction (SCR) process and the so-called ‘NOx storage–reduction’ (NSR) technique. The NH3 SCR is a well-established process in the case of stationary sources. The high activity and selectivity, the good resistance to catalyst poisoning and ageing have made the NH3 -SCR process the most promising technology to comply with the NOx emission standards in the near future for mobile sources as well, in particular for heavy trucks. In this case, urea is used as a source of NH3 . In 2005, most of the European truck producers launched urea-SCR-equipped trucks able to meet the Euro 4 and Euro 5 emission standards.

Identification of the Reaction Networks of the NOx Storage/Reduction

177

The ‘NSR’ catalysts, also referred to as NOx adsorbers or lean NOx traps (LNTs), were developed and launched on the market by Toyota in the 1990 [5–6]. The technology consists while treating the exhausts with a double functionalized catalyst, a basic component able to store the nitrogen oxides while operating the engine under lean conditions, and a noble metal, which allows oxidation of NO and the reduction of the stored nitrates by means of HCs when operating the engine under fuel-rich conditions. Accordingly, this technology implies that the engine operates under cyclic conditions, by alternating long periods under lean conditions during which NOx are stored on the catalyst surface, with short periods under rich conditions during which the stored NOx are reduced to N2 and water by the HCs contained in the exhaust, as well as by CO and H2 , which may also be present. The obvious advantage of this technology with respect to the SCR process is that NSR catalysts do not make use of any external reductants (ammonia or urea). Ternary catalysts made by an alumina support on which the active elements (primarily Ba and Pt) are dispersed are a typical catalytic material for NSR applications. The Pt−BaO/Al2 O3 catalytic system is a representative of this class of catalysts. NOx is oxidized and trapped on Ba in the oxygen-rich atmosphere and then reduced to nitrogen in the oxygen-deficient atmosphere on Pt. Among alkali and alkaline earth elements, Ba has been reported as the most effective element to store NOx in the NSR catalyst [5]. Other components, commonly added to the ternary Pt−BaO/Al2 O3 catalyst, include titanium oxide (TiO2 ), which minimizes the adsorption of SOx , as well as rhodium (Rh) and zirconium oxide (ZrO2 ), which are used to promote the formation of hydrogen, hence playing an important role during the reduction of stored NOx and in the catalyst regeneration from sulphates [6]. Nowadays, the NSR process is a very attractive lean-DeNOx technology, but it has been commercialized only in Japan where low sulphur gasoline is available. In fact, the major drawback of the NSR catalyst is its sensitivity to SOx due to the fact that surface sulphates are invariably more stable if compared to nitrates [7]. The durability aspects of the NSR catalysts were addressed by several researchers [8] and there seems to be general agreement that poisoning of the NOx storage function is directly related to the amount of SO2 passed over the catalyst. This is an important aspect since it suggests that in order to employ these catalysts on the US or European markets, appropriate strategies aimed at developing sulphur resistant NOx trap must be worked out. However, the ultra-low sulphur diesel fuel introduced in the US market in October 2007 as per US EPA regulations contains less than 15 ppm sulphur, and also in Europe a decrease in the S-content of fuels is in progress to comply with the new legislation requirements. The potential of NSR catalysts in the removal of NOx from mobile sources has motivated in the last few years extensive investigations from both the academic and the industrial world, and several studies have been published in the open literature dealing with fundamental and practical aspects of LNT catalysts [4–53]. However, the mechanisms that operate the NOx adsorption and the respective subsequent reduction have not been completely clarified so far. It has been shown that under oxidizing conditions, NOx are stored on the surface of a Ba-containing catalyst in various forms (surface nitrites/nitrates), whose precise nature is, however, still a matter of debate [9–29]. Even less clear are the mechanisms, which are responsible for the reduction of stored NOx when the A/F ratio is set to rich and the stored NOx species are reduced over Pt to N2 , ammonia, N2 O or back to NOx [11].

178

Past and Present in DeNOx Catalysis

An extensive investigation on the analysis of LNT systems has been carried out during the last few years in our labs, and the respective results are presented hereafter. In particular, the mechanisms of the NOx storage and of their subsequent reduction have been addressed, along with the effect of the operating conditions on both the storage and the reduction phases.

2. CATALYST PREPARATION, CHARACTERIZATION AND TESTING PROCEDURES 2.1. Preparation and characterization Various model catalytic systems have been considered for the study of LNT systems. These systems are basically constituted by Ba as NOx -storage component, and by Pt as noble metal, both dispersed on alumina as high surface area carrier. In particular, a ternary Pt−Ba/-Al2 O3 (1/20/100 w/w) catalyst was prepared and tested as a reference system, composition and preparation procedures were selected according to the Toyota patents [5,6,9,10]. Briefly, a Pt/-Al2 O3 (1/100 w/w) sample was prepared by incipient wetness impregnation with an aqueous solution of dinitro-diammine platinum (Strem Chemicals, 5% Pt in ammonium hydroxide) of a commercial alumina support material (Versal 250 from La Roche Chemicals calcined at 700 C), followed by drying overnight and calcination at 500 C for 5 h. The Pt−Ba/-Al2 O3 (1/20/100 w/w) reference catalyst was prepared by incipient wetness impregnation of Pt/-Al2 O3 with an aqueous solution of Ba(CH3 COO)2 (Strem Chemicals, 98.5%) followed by drying overnight and calcination at 500 C for 5 h. In order to analyse the effect of the different catalyst component, a Ba/-Al2 O3 (20/100 w/w) sample was also prepared by impregnation of the bare -alumina support with barium acetate. Ternary catalysts with different amounts of Ba were also prepared, in the range 0–30% (w/w); for this purpose, the Pt/-Al2 O3 (1/100 w/w) sample was impregnated with solutions containing various concentrations of Ba(CH3 COO)2 . The reference Pt−Ba/-Al2 O3 (1/20/100 w/w) catalyst shows surface area values in the range 140–160 m2 /g, a pore volume of 0.7–0.8 cc/g and an average pore radius close to 100 Å (measured by N2 adsorption–desorption at 77 K by using a Micromeritics TriStar 3000 instrument). Slight differences in the characterization data are associated to various batches of the ternary catalyst [24,25]. X-ray diffraction (XRD) analysis, performed by using a Brüker D8 Advanced Instrument equipped with graphite monochromator of the diffracted beam, showed the presence of -Al2 O3 (JCPDS 10-425) along with traces of microcrystalline orthorhombic (JCPDS 5-378) and monoclinic (JCPDS 78-2057) BaCO3 phases. Quantitative analysis of the XRD spectra indicated that the amount of crystalline BaCO3 accounts for 5% of the Ba loading in the freshly prepared sample and increased up to 30% of Ba in the sample exposed to air at room temperature (RT) for prolonged time. This indicated that Ba is well dispersed on the surface of the catalyst and that extensive surface restructuring occurs with time. Restructuring of the catalyst surface also occurs during NSR catalysis; this will be discussed in the following sections. The Pt dispersion of the fresh samples was measured by dynamic hydrogen chemisorption by using a temperature-programmed desorption (TPD)/R/O 1100 ThermoFisher

Identification of the Reaction Networks of the NOx Storage/Reduction

179

Instrument. Before the H2 chemisorption, each sample was heated in pure H2 at 300 C for 90 min, subsequently it was heated in He at 290 C for 60 min in order to desorb hydrogen from the sample. The chemisorption measurement was performed at 0 C by several H2 pulses with an Ar purge in between, in order to desorb physisorbed hydrogen. A 1/1 H/Pt ratio was used to estimate the Pt dispersion. Values in the range 50–70% have been obtained on the fresh samples Pt−Ba/-Al2 O3 (1/20/100 w/w) catalyst. The characterization carried out on the samples having different Ba loading showed that upon increasing the Ba loading from 0 to 30% (w/w), the surface area progressively decreases from 186 m2 /g down to 110 m2 /g, respectively, and the pore volume from 1.02 cm3 /g to 0.6 cm3 /g [27]. The Pt dispersion, measured by H2 chemisorption at 0 C, was about 82% in the Ba-free sample and decreases upon addition of Ba, thus reaching the value of 40% at the highest barium loading (Pt−Ba/-Al2 O3 1/43/100 w/w, corresponding to 30%, w/w Ba). The decrease in the Pt dispersion was ascribed to the fast and exothermic decomposition of barium acetate precursor eventually leading to sintering of the Pt crystallites, and/or to the masking of the Pt crystallites upon addition of the Ba component [24,25]. In all the examined cases, XRD measurements showed the presence of microcrystalline -Al2 O3 (JCPDS 10-425). For Ba loadings below 10% (w/w), no other crystalline phases have been detected, thus suggesting that the Ba component is well dispersed on the support or present as amorphous phase. On the other hand, in the samples with higher Ba loadings, the presence of crystalline BaCO3 phases was evident [24,25,27] both in the monoclinic (JCPDS 78-2057) and in the orthorhombic (JCPDS 5-378) form.

2.2. Testing procedures The NSR capability of the catalysts was investigated under transient conditions in a flow microreactor system with samples in the powder form. The dynamics of the NOx adsorption–reduction was investigated by runs at constant temperature [transient response methods (TRM)]. Typically, rectangular step feeds of NO (1000 ppm) in He + 3% (v/v) O2 were alternated with rectangular step feeds of hydrogen (2000 ppm) in He, a purge period in He was performed between the two phases, which allowed to make a clear distinction between the different processes. TRM experiments were performed at different temperatures in the range 250–400 C. The effect of water and CO2 in the feed mixture was also investigated by performing TRM experiments in the presence of carbon dioxide (3000 ppm) and/or water (1%). The adsorption of NOx was also carried out with NO (1000 ppm) in He and/or with NO2 (1000 ppm) in the presence and in the absence of 3% O2 . The thermal stability of NOx adsorbed species and their reactivity in the presence of gaseous reductant molecules was addressed by thermal decomposition in He (TPD) or by heating in flowing H2 /He mixtures [temperature-programmed surface reaction (TPSR)], respectively. In these cases, after NOx adsorption and He purge at the adsorption temperature (300–400 C), the samples were cooled to RT under flowing He. Then the samples were heated at 15 C/min up to 500–600 C in He (TPD) or in He + H2 (2000 ppm) (H2 -TPSR). In each run, 120 mg of catalyst (75–100 m) were used and a total flow rate by 200 cm3 /min STP was maintained in the different phases. The flow microreactor system

180

Past and Present in DeNOx Catalysis

was directly connected to both a mass spectrometer (Balzers QMS 200) and to a gas chromatograph (HP 6890) for product analysis. Further details about the experimental apparatus and procedures can be found elsewhere [24,25,45].

3. NOx STORAGE 3.1. Nature of the barium species involved in the NOx storage process In order to obtain some additional information on the nature and reactivity of the Ba surface species involved in the NOx storage process, the initial storage–reduction cycles performed over the fresh reference catalyst sample (Pt−Ba/Al2 O3 1/20/100 w/w) were analysed [24]. The catalyst was pretreated at 500 C in He/O2 for 1 h; since the characterization data pointed out the presence of barium carbonate on the fresh catalyst, it is expected that BaCO3 and BaO are the Ba surface species primarily present on the catalyst, along with Ba(OH)2 in minor amount. The results collected during the initial NOx storage cycles performed at 350 C onto the fresh catalyst sample are presented in Figure 6.1 in terms of NOx , H2 O and CO2 outlet concentration as a function of time. Upon the first adsorption cycle, the evolution of CO2 (with a delay) points out that NOx was adsorbed at first at BaO and then at BaCO3 (with CO2 release), in line with the greater basic character of the former site. After the storage, regeneration with hydrogen has been performed. Upon the second adsorption cycle, water and CO2 evolution were observed, whereas in the case of the third adsorption cycle only water was detected, along with minor amounts of CO2 . From the third adsorption cycles onwards, the NOx storage occurs with similar features. This change in the nature of the Ba surface species was also accompanied by an increase of the NOx breakthrough, i.e. the NOx storage was enhanced. These data are consistent with the hypothesis that the storage of NOx occurred first at BaO, then at Ba(OH)2 and finally at BaCO3 , in line with the basic character of these different Ba adsorption sites. Catalyst regeneration with H2 restored the adsorption Ba sites; however, Ba(OH)2 and BaO were formed instead of BaCO3 since H2 O (and not CO2 ) was produced during reduction (see Section 4). As a result, the amounts of BaO and Ba(OH)2 species on the catalyst surface progressively increased at the expenses of BaCO3 . On the other hand, if the feed gas contains H2 O and/or CO2 the most abundant barium species involved in the storage process are Ba(OH)2 and/or BaCO3 , respectively. These aspects will be discussed and clarified in the following sections.

3.2. NOx storage mechanism An extensive study of NOx storage over the homemade reference Pt−Ba/Al2 O3 (1/20/100 w/w) system was performed in our laboratories [24–26,28]. The adsorption of NO and of NO2 in the presence and in the absence of oxygen was carried out in a fixed bed flow microreactor at 350 C over the Pt−Ba/Al2 O3 ternary catalyst, over the corresponding binary Ba/Al2 O3 and Pt/Al2 O3 samples and over the bare -Al2 O3

Identification of the Reaction Networks of the NOx Storage/Reduction

181 I

NOinlet

900

NOx 600

CO2

300

H2O 0

Concentration, ppm

II 900

NOx 600

300

H2O

CO2

0

III 900

NOx 600

CO2

300

H2O

0 0

300

600

900

1200

Time, s

Figure 6.1. Subsequent NOx storage runs in NO (1000 ppm) + O2 (3%, v/v), He balance (total flow 200 Ncc/min, catalyst weight 120 mg) at 350 C on a fresh sample of Pt−Ba/Al2 O3 (1/20/100 w/w). NOx , H2 O and CO2 are outlet concentrations, and NO is inlet concentration.

support. The stored NOx species were removed from the catalyst surface by heating in He up to 600 C under temperature programming (TPD). The adsorption/desorption sequence was repeated several times in order to fully condition the catalytic systems; accordingly, BaO was the most Ba abundant species present on the catalyst surface. Fourier transform infrared (FTIR) was used as a complementary technique to investigate the nature of the stored NOx species. The adsorption of NO2 over the Ba/-Al2 O3 (20/100 w/w) sample was investigated at first. In fact, NO2 has been suggested as intermediate in the NO adsorption in the presence of oxygen [21]. As shown in Figure 6.2, the NO2 storage is slow and accompanied by the evolution of NO, which is observed with a small breakthrough time (35 s). As reported elsewhere [28,36,37], the NO2 storage occurred in this case according to the stoichiometry of the following global disproportion reaction: BaO + 3NO2 → BaNO3 2 + NO ↑

(1)

182

Past and Present in DeNOx Catalysis (b) 1320 1420

1.2 10 min 1 min

0.8

1200

NO2,in

Concentration, ppm

1.0

0.6

900

0.4

1560

600

0.2

NO2

NO

1800

1600

1400

1200

0.0 1000

–1

Wavenumber/cm

300

0 0

1000

2000

3000

4000

5000

6000

Time, s

Figure 6.2. Storage run in NO2 (1000 ppm), He balance (total flow 200 Ncc/min, catalyst weight 120 mg) at 350 C over Ba/Al2 O3 (20/100 w/w) catalyst. NO and NO2 are outlet concentrations, and NO2 is inlet concentration. (b) Results of NO2 adsorption using FTIR experiments; spectra are reported after 1 and 10 min of exposure to 5 mbar of NO2 at 350 C. Each spectrum is reported as difference from the spectrum before NO2 admission.

which leads to the formation of nitrates onto the catalyst surface and to the evolution of gaseous NO. The FTIR data reported in Figure 6.2b showed that only nitrate species were formed upon NO2 adsorption, mainly of the ionic type (bands at 1320, 1420–1440 cm−1 , asym NO3 split for the partial removal of the degeneracy; 1035–1020 cm−1 , sym NO3 ) and in minor amounts of bidentate type (1560 cm−1 , asym NO2 mode expected around 1300 cm−1 obscured by the modes of ionic nitrates). Notably, the adsorbed nitrates were related to the Ba component as the surface of the alumina support was almost completely covered by Ba, as pointed out by FTIR data [25], which showed the disappearance of OH groups of the alumina support. NO2 adsorption experiments were also performed over the reference Pt−Ba/-Al2 O3 (1/20/100 w/w) sample and similar results have been obtained (see Figure 6.3). It is worthwhile noting that in this case, the NO outlet concentration can be related to both the NO2 disproportionation and the NO2 decomposition reactions. In fact, O2 formation was also observed in this case, in line with NO2 decomposition to NO and oxygen over Pt sites. FTIR spectra collected over the Pt−Ba/-Al2 O3 reference catalyst (Figure 6.3b) pointed out also in this case that nitrate species are formed upon NO2 adsorption. The uptake of NO in the presence of oxygen (3%, v/v) on the Pt-free sample (Ba/ -Al2 O3 ) proceeded as illustrated in Figure 6.4a. In this case, small quantities of NOx species are adsorbed on the catalyst surface. The FTIR spectra (Figure 6.4b) recorded at different exposure times showed a progressive formation of ionic nitrites (1220 cm−1 ) upon increasing the exposure time up to 10 min, along with small amounts of bridging nitrates. At higher exposure times, the band due to nitrite species decreased in intensity and completely disappeared after 20 min. At the same time, bands characteristic of ionic

Identification of the Reaction Networks of the NOx Storage/Reduction

183

(b) 1.2 1.0

1320

1200

0.8

Concentration, ppm

NO2,in

1420

10 min 1 min

0.6

900 0.4

NO

1560 0.2

600

O2

300

1800

NO2

1600

1400

1200

0.0 1000

Wavenumber/cm–1

0

0

1000

2000

3000

4000

5000

6000

Time, s

Figure 6.3. Storage run in NO2 (1000 ppm), He balance (total flow 200 Ncc/min, catalyst weight 120 mg) at 350 C over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst. NO and NO2 are outlet concentrations, and NO2 is inlet concentration. (b) Results of NO2 adsorption using FTIR experiments; spectra are reported after 1 and 10 min of exposure to 5 mbar of NO2 at 350 C. Each spectrum is reported as difference from the spectrum before NO2 admission.

nitrates (1420, 1320, 1030 cm−1 ) and in minor amounts of bidentate nitrates developed, so that after 20 min of exposure only nitrates were evident in the spectra. Figure 6.4c showed the storage uptake behaviour by using NO/O2 over the reference Pt−Ba/-Al2 O3 sample. Upon admission of NO (at t = 0 s) both the NO and NO2 outlet concentrations show a significant delay. Then the NO concentration increases, followed by that of NO2 . NO2 formation is ascribed to the oxidation of NO by O2 according to the stoichiometry of reaction (2): NO + 1/2O2 → NO2

(2)

The breakthrough time observed in the NOx concentration profile (obtained by addition of the NO and NO2 concentrations) indicates that during the initial part of the pulse, the NO fed to the reactor is completely stored on the catalyst surface. The FTIR spectra gave the results reported in Figure 6.4d. The adsorption of NO initially occurred primarily in the form of nitrites, which have readily transformed into nitrates so that at the end of the NO pulse, at catalyst saturation, nitrates were the prevalent species. Notably, the rate of both nitrite formation and their oxidation to nitrates was higher on Pt−Ba/-Al2 O3 than on Ba/-Al2 O3 , thus pointing out a catalytic role of Pt. Similar results have been obtained by performing FTIR spectra under operating conditions [26]. It was shown that the nitrite band intensity exhibited a broad maximum and then decreased with time on stream, while the nitrate bands presented a monotonous increase during the entire storage phase, so that at the end of the experiment nitrates are the prevalent adsorbed species. Thus, in line with FTIR data, in the adsorption process nitrites are intermediate species, which then evolve leading to the formation

184

Past and Present in DeNOx Catalysis

(a)

(b)

1200

0.6

Concentration, ppm

NOin

1320 20 min

900

NO

1420

0.4 15 min 1220

600

10 min 5 min

300

0.2 3 min 1 min

NO2

0 0

1000

2000

3000

0.0

Wavenumber/cm–1

Time, s (c)

(d)

1200

Concentration, ppm

NOin 900

1320 1420

NOx NO

600

1.0 20 min 10 min 5 min

0.8

3 min

0.6

NO2

300

0.4 1220 1 min 3 min

0.2

0 0

1000

Time, s

2000

3000

1800

1600

1400

1200

0.0 1000

Wavenumber/cm–1

Figure 6.4. Storage run in NO (1000 ppm) + O2 (3%, v/v), He balance (total flow 200 Ncc/min, catalyst weight 120 mg) at 350 C over (a) Ba/Al2 O3 (20/100 w/w) and (c) Pt−Ba/Al2 O3 (1/20/100 w/w) catalysts. NO and NO2 are outlet concentrations and NO is inlet concentration. Results of NO/O2 adsorption using FTIR experiments over (b) Ba/Al2 O3 (20/100 w/w) and (d) Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst; spectra are reported after 1, 3, 5, 10, 15 and 20 min of exposure to NO/O2 mixtures (1:4, pNO = 5 mbar) at 350 C. Each spectrum is reported as difference from the spectrum before NO/O2 admission.

of nitrates. Notably, the maximum in the nitrite band intensity (observed during FTIR under operating conditions) roughly corresponded to the NO breakthrough observed by TRM experiments. Accordingly, upon NO adsorption in the presence of oxygen over Pt−Ba/alumina: (1) nitrites are the major NOx adsorbed species at the initial stage of adsorption before NOx breakthrough (observed after 4 min); (2) nitrites are progressively transformed into nitrates during storage; (3) a parallel route involving the adsorption of NO2 is also present. All the previous data lead to the proposal of the reaction pathway shown in Figure 6.5 for the storage of NOx over Pt−Ba/alumina catalysts. In the presence of oxygen, NO was effectively stored through a stepwise oxidation at a Pt site followed by adsorption at a neighbouring Ba site to form Ba-nitrites at first and Ba-nitrates later on. This is referred to as the ‘nitrite route.’ The nitrite oxidation to nitrates is catalysed by Pt and likely involves NO2 formed by NO oxidation on Pt and/or O2 species. Accordingly, a cooperative effect between Pt−Ba neighbouring couples might be relevant for this route.

Identification of the Reaction Networks of the NOx Storage/Reduction NO–2

NO/O2 Pt + Ba

NO + O2

BaO BaO Al 22O33

Nitrite species

Pt

O2 NO2

NO3– NO

NO2

Pt Pt

185

Ba

BaO BaO Al 22O33 Nitrate species

Figure 6.5. Reaction pathway for NOx adsorption over supported Pt-Ba catalysts

NO is also oxidized to NO2 over Pt in the presence of oxygen. NO2 could be adsorbed onto the Ba sites to form Ba nitrates through the global disproportion reaction (1), which is accompanied by the evolution of NO and results in the formation of nitrates. This is referred to as the ‘nitrate route.’ It is worth of note that the role of the alumina support, which showed a non-negligible NOx adsorption capacity in the presence of NO2 [28], was not considered in the scheme due to its almost complete coverage by the Ba component in the Pt−Ba/-Al2 O3 (1/20/100 w/w) reference sample.

3.3. Effect of Pt−Ba interaction on NOx adsorption mechanism It has been suggested that the presence of Pt−Ba couples (i.e. the existence of a Pt−Ba interaction) might be relevant for the NOx adsorption via the nitrite route. In this respect, it is of interest to investigate the effect of the catalyst Ba loading, since it is expected that this would affect the number of Pt−Ba couples and hence the Pt−Ba interaction [19,36,37]. In order to verify the existence of Pt−Ba proximity CO chemisorption measurements by FTIR have been carried out on Pt−Ba/-Al2 O3 samples having different Ba loading [29]. Upon increasing the Ba loading, the band intensity of CO linearly adsorbed on Pt sites decreases (due to the decrease of the Pt dispersion with Ba loading, see Section 1 in Chapter 2) and shifts towards lower energy (from 2072 to 2049 cm−1 ) according to the increase of the system basicity. Moreover, a shoulder at lower energy is well detected at low Ba content (at 1967 and 1949 cm−1 for Pt−Ba/ -Al2 O3 1/5.3/100 w/w and 1/20/100 w/w, respectively) that could be related to CO linearly adsorbed on Pt sites with a strong interaction with Ba. Hence, the data indicate a strong interaction between Pt and the basic oxygen anions of the Ba phase, thus suggesting that the exposed Pt sites and the Ba component are in close proximity [25]. The effect of the Ba loading on the NOx adsorption capability was investigated by performing TRM experiments with NO/O2 mixtures at 350 C and the results are shown in Figure 6.6 for Pt−Ba/-Al2 O3 catalysts having different Ba loadings. The outlet NO, NO2 and NOx (NO + NO2 ) concentration curves are displayed as a function of time, along with that of the NO inlet concentration (dotted lines).

186

Past and Present in DeNOx Catalysis

Concentration, ppm

1000

Concentration, ppm

1000

Concentration, ppm

1200

1000

NOin

800

NOin

(a)

NOx

(b)

NOx

NO

NO

600

NO2

400

NO2

200 0 1200

NOin NOx

800 600

(c)

NOin

(d) NOx

NO

NO

NO2

NO2

400 200 0 1200

NOin

(e)

NOin

(f) NOx

NOx

800

NO

NO

600 400

NO2

200

NO2

0

0

500

1000

1500

Time, s

2000

2500

3000 0

500

1000

1500

2000

2500

3000

Time, s

Figure 6.6. Storage run in NO (1000 ppm) + O2 (3%, v/v), He balance (total flow 200 Ncc/min, catalyst weight 120 mg) at 350 C over Pt−Ba/Al2 O3 (1/x/100 w/w) catalysts. (a) x = 0; (b) x = 53; (c) x = 115; (d) x = 20; (e) x = 30; (f) x = 43. NO, NO2 and NOx are outlet concentrations, and NO is inlet concentration.

In all cases, NO and NO2 are detected at the reactor outlet and it clearly appears that the breakthrough time in the NOx outlet concentration and the NO2 formation markedly depend on the Ba loading of the catalysts. As regards the NOx breakthrough, when NO/O2 adsorption is carried out on the Pt/-Al2 O3 (1/100 w/w) sample, no dead time is observed, which indicates a negligible storage of NOx species on the surface. As a matter of fact, only minor amounts of NOx have been stored in this case up to catalyst saturation, which however desorb upon switching off the NO feed flow. The presence of Ba modified the response of the catalysts in the adsorption of NO/O2 . In all cases, the NO outlet concentration shows a breakthrough time that increased from about 50 s for the Pt−Ba/-Al2 O3 1/5.3/100 w/w sample up to about 700 s for the Pt−Ba/-Al2 O3 1/30/100 w/w and then decreased to 560 s for the sample with the highest Ba content (i.e. Pt−Ba/-Al2 O3 1/43/100 w/w). The amounts of NOx stored up to saturation increase with Ba loading up to a value near 12 × 10−4 molNO /gcat for the sample with the highest Ba loading (Figure 6.7). The fraction of Ba that participates in the storage process (‘active’ Ba), i.e. the percentage ratio of Ba involved in the storage with respect to the total amount of Ba present in the catalyst sample, can be evaluated. Assuming the formation of either Ba(NO2 2 or Ba(NO3 2 , the amounts of Ba participating in the process is calculated starting from the values of NOx species adsorbed during the storage periods. As shown in Figure 6.7, the

Identification of the Reaction Networks of the NOx Storage/Reduction

187 35 30

12 25 9

20 15

6

10 3 5 0

% Ba involved in the storage

mol NOx /gcat stored (*104)

15

0 0

5

10

15

20

25

30

Ba loading (%)

Figure 6.7. Moles of adsorbed NOx (point) and fraction of Ba involved in the storage (square) as function of Ba loading at catalyst saturation on Pt−Ba(x)/-Al2 O3 catalysts.

fraction of Ba involved up to catalyst saturation is low (about 4%) for Pt−Ba/-Al2 O3 1/5.3/100 w/w sample, and then increased with the catalyst Ba content up to the maximum value of about 33% for the Pt−Ba/-Al2 O3 1/30/100 w/w sample, which accordingly exhibits the best utilization of the Ba component. Notably, the estimation of the Ba coverage for the different samples showed that a value close to one was almost achieved for Ba content of 16–20% (w/w). This suggests that the maximum storage capacity was presented by the systems, which are characterized by the formation of the monolayer. The collected results are in line with the ‘nitrite route’ represented in Figure 6.5. Indeed, upon increasing the Ba loading, the possibility to have Pt−Ba neighbouring species is also increased. Accordingly, in the case of the samples with a low Ba content, it is reasonable to assume that the number of Pt−Ba neighbouring couples is low: in this case the ‘nitrite route’ is less efficient and the catalyst likely adsorbs NOx through the ‘nitrate route,’ with release of NO. As a matter of fact, results showed that the samples with low Ba content were characterized by very short breakthrough times. In the case of higher Ba loadings, a greater number of Pt−Ba neighbouring couples are present on the catalyst surface, as also suggested by FTIR measurements; this would favour the ‘nitrite’ pathway, thus resulting in a better utilization of the Ba component and in significant breakthrough times. Notably, NO adsorption experiments in the presence of oxygen were also performed over a physical mixture of Pt/-Al2 O3 and Ba/-Al2 O3 having the same Pt and Ba loading of the Pt−Ba/-Al2 O3 1/20/100 w/w sample. Comparing the results (here not reported) with the correspondent ternary Pt−Ba/-Al2 O3 1/20/100 w/w sample, it appeared that physical mixture and ternary sample were characterized by a similar breakthrough time, but the physical mixture presented a higher oxidizing capacity, as indicated by a higher NO2 concentration measured at catalyst saturation (steady-state) at the reactor outlet. Hence, the data obtained on the mechanical mixture confirm the storage mechanism proposed in literature where NO is oxidized to NO2 at first; NO2 is adsorbed on the catalytic surface to form nitrates. In this case, no direct interaction is necessary to have a good and efficient adsorption.

188

Past and Present in DeNOx Catalysis

3.4. Kinetic model The role of the Pt−Ba interaction in the mechanism of adsorption of NOx species was also studied by a kinetic model reported in the literature [16]. The model, which consists of 10 elementary reversible steps, is based on the oxidation of NO to NO2 over Pt and on the storage of NO2 over Ba, and it was used to simulate the data collected over both the physical mixture and the ternary Pt−Ba/-Al2 O3 1/20/100 w/w sample. A spillover reaction between Pt and Ba oxide sites has also been included in the model to account for the observed lower thermal stability of Ba-nitrates in the presence of Pt [16]. Essentially, the model assumes that the adsorption of NOx proceeds through the nitrate route and does not consider the nitrite route. The simulations [30] showed that the model satisfactorily reproduced the results collected over the physical mixture upon NO step addition in the presence of oxygen. Both the oxidation of NO to NO2 and the dead time for the breakthrough of NOx are reasonably simulated by the model, as indeed expected according to nitrate route adsorption. In the case of the Pt−Ba/alumina ternary sample, the dead time in the NOx breakthrough upon a step change of NO in the presence of oxygen is reasonably well predicted by the model but the oxidation of NO to NO2 is markedly overestimated. If the Pt dispersion in the model is lowered, the oxidation of NO to NO2 is properly described but the NOx breakthrough is no longer predicted. Therefore, the literature model is not able to describe at the same time the oxidation of NO to NO2 and the NO breakthrough. The model inadequacy was particularly evident for the systems having different Ba loadings, which showed an increase in the breakthrough time associated with similar NO oxidizing capacity. Work is currently in progress in order to gain a better adequacy of the model to our data; in particular the nitrite route has also been included to provide an additional NO adsorption pathway, which is in line with obtained data, and preliminary results obtained in this direction seem to be promising.

3.5. Effect of CO2 on the NOx storage mechanism In order to verify if the NOx storage mechanism proposed for BaO sites is operating also in the presence of CO2 , i.e. under conditions more representative of the real composition of the gas mixture in the engine exhausts, the influence of CO2 on the NOx storage mechanism over the reference Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst and over the binary Ba/Al2 O3 20/100 w/w reference sample was investigated. As discussed in the first section, primarily BaCO3 species are present on the catalyst surface when CO2 is contained in the feed stream. The TRM runs of NO2 adsorption in the presence of CO2 on the ternary Pt−Ba/Al2 O3 catalyst at 350 C (Figure 6.8a) showed that, like in the absence of CO2 , (see Figure 6.3), the NOx adsorption occurred via NO2 disproportionation with evolution of NO. Also in the presence of carbon dioxide, the NOx storage is accompanied by NO2 decomposition on Pt sites, with evolution of NO and O2 .

Identification of the Reaction Networks of the NOx Storage/Reduction

189

(a) 1200

Concentration, ppm

NO2,in 900

NO2

NOx 600

300

CO2

NO

O2

0 0

2000

4000

6000

(b) 1200

Concentration, ppm

NOin 900

NOx

600

CO2

NO2

NO

300

0 0

300

600

900

1200

1500

Time, s

Figure 6.8. (a) Storage run in NO2 (1000 ppm)/CO2 (3000 ppm) balance He (total flow 100 Ncc/min, catalyst weight 60 mg) at 350 C; (b) Storage run in NO (1000 ppm)/CO2 (3000 ppm)/O2 (3% v/v) balance He (total flow 100 Ncc/min, catalyst weight 60 mg) at 350 C over Pt-Ba/Al2 O3 (1/20/100 w/w) catalyst. NO, NO2 , NOx , O2 and CO2 (+3000 ppm) are outlet concentrations, and NO is inlet concentration.

Notably, during the NO2 storage, the release of CO2 was observed (Figure 6.8a), according to the stoichiometry: 3NO2 + BaCO3 → BaNO3 2 + NO ↑ + CO2 ↑

(3)

This indicates that the catalyst surface is fully carbonated and that the displacement of carbonates is occurring upon NO2 adsorption. In agreement with these data, FTIR spectra [52] showed that the presence of CO2 does not significantly modify the adsorption of NO2 . In particular, both in the presence and in the absence of CO2 , nitrates are mainly formed at the catalyst surface, primarily of the ionic type (1410, 1320 and 1020 cm−1 ), along with minor amounts of bridging species (1550 cm−1 ). Accordingly, from both TRM and FTIR data, it is concluded that the adsorption of NO2 /CO2 mixtures strictly parallels that of NO2 in the absence of CO2 , i.e. the occurrence of the ‘nitrate route’ is not greatly affected by the presence of CO2 .

190

Past and Present in DeNOx Catalysis

A slightly different picture is obtained when the storage is carried out with NO/O2 instead of pure NO2 . In fact, TRM data obtained over the reference Pt−Ba/Al2 O3 sample upon NO/O2 admission in the presence of CO2 at 350 C (Figure 6.8b) indicated that the NOx storage capacity was to some extent lower than that determined in the absence of CO2 (see Figure 6.4c). More significant is the change in the dynamics of the NOx breakthrough, which indicates that in the presence of CO2 the adsorption process becomes slower. Notably, also in this case NOx adsorption was accompanied by CO2 evolution thus suggesting that only BaCO3 sites are involved in the adsorption phase. The FTIR data recorded upon NO/O2 adsorption in the presence of CO2 showed that bidentate carbonates were immediately formed along with nitrite species. By increasing the time of contact, carbonates were partially displaced, while nitrites evolved to nitrate species [52]. Notably, the amounts of surface nitrites present at each contact time are lower in the presence of CO2 than in its absence. This indicates that CO2 competes for the surface oxygen sites able to give nitrites at the beginning of the adsorption process: in fact, after several minutes of exposure to the NO/O2 /CO2 mixture, the amount of nitrates stored was comparable to that obtained in the absence of CO2 . In conclusion, while in the presence of CO2 the ‘nitrite route’ is inhibited to some extent due to the competition between NO and CO2 for the surface oxygen sites of the Ba phase, the ‘nitrate route’ is only marginally affected by the CO2 presence, if any.

3.6. Effect of the operating conditions on NOx adsorption The storage of NOx in the presence of 3% oxygen over Pt−Ba/-Al2 O3 (1/20/100 w/w) sample has been investigated (in the absence of CO2 and H2 O) in a wide temperature range, from 150 to 400 C. In all cases H2 was used in the regeneration step, hence BaO and Ba(OH)2 are the Ba sites involved in the NOx storage, as discussed above. The data discussed below refer to a fully conditioned sample, i.e. a sample on which several storage–reduction cycles were carried out until reproducible results were obtained [24]. Figure 6.9a and b show the amounts of NOx stored onto the Pt−Ba/-Al2 O3 (1/20/100 w/w) sample at breakthrough and at saturation as a function of temperature (base case). A maximum in the NOx stored at breakthrough is observed near 350 C, with values of adsorbed NOx close to 35 × 10−4 molNO /gcat . This corresponds to an overall amount of Ba involved in the storage near 14% of the total Ba loading. During the NOx storage, a small amount of water was released indicating that the adsorption of NOx species occurs on BaO and on Ba(OH)2 sites, leading to the formation of nitrites and nitrates species according to the reaction pathway depicted in Figure 6.5. Notably, the relative amounts of Ba(OH)2 involved in the NOx storage process is large (50–60%) at low temperature but diminishes significantly with increased temperatures (down to 10–15% at 400 C), as expected. Considerably higher amounts of NOx are stored onto the Pt−Ba/-Al2 O3 up to catalyst saturation, reaching a maximum at 300 C (5.81 × 10−4 molNO /gcat ) that corresponds to an utilization of the Ba component in the storage closes to 24%. Experiments were also performed in the presence of CO2 during both the storage and the regeneration steps. Accordingly, in this case BaCO3 are the Ba sites involved in the NOx storage. By increasing the temperature the NOx breakthrough increases in the whole investigated temperature range (see Figure 6.9a). Notably, no water was observed

Identification of the Reaction Networks of the NOx Storage/Reduction

191

4

Mol NOx stored/gcat (*104)

(a) Base case

3

2

H2O + CO2

H2O 1

CO2

0

Mol NOx stored/gcat (*104)

(b) 6

Base case

4

H2O H2O + CO2

2

CO2

0 150

200

250

300

350

400

Temperature, °C

Figure 6.9. Effect of temperature on the amounts of NOx stored (a) up to NOx breakthrough and (b) up to catalyst saturation in the base case (point) [NO (1000 ppm)/O2 (3%, v/v) balance He], in the presence of H2 O (square) [NO (1000 ppm)/H2 O (1%, v/v)/O2 (3%, v/v) balance He], in the presence of CO2 (triangle) [NO (1000 ppm)/CO2 (3000 ppm)/O2 (3%, v/v) balance He] and in the presence of H2 O + CO2 (star) [NO (1000 ppm)/CO2 (3000 ppm)/H2 O (1%, v/v)/O2 (3%, v/v) balance He] over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst (total flow 200 Ncc/min, catalyst weight 120 mg).

at the reactor outlet but only CO2 , without any delay, according to the presence of a fully carbonated catalyst surface. The comparison between the amounts of NOx stored at breakthrough in the standard runs and in the presence of CO2 showed a significant reduction of the storage in the presence of 0.3% CO2 (Figure 6.9a). A different picture is apparent if one considers the total amount of NOx stored at catalyst saturation (Figure 6.9b). Indeed, in this case the amounts of stored NOx are similar in both cases, thus indicating that the presence of CO2 affects the NOx breakthrough (i.e. the dynamic of the storage) more than the total storage capacity of the catalytic system. The presence of water in the feed gas during the NOx storage was also studied, in the range 150–400 C. Data indicated that the storage of NOx was always accompanied by a simultaneous release of water (without any dead time), thus suggesting that the adsorption of NOx occurs primarily on Ba(OH)2 sites. Besides, a striking effect of water (with respect to the ‘base case’) was the increase of NOx breakthrough at low temperature and a slight decrease at high temperature (see Figure 6.9a). The same effects can be observed at saturation (Figure 6.9b), so that it seems that water has got a promoting

192

Past and Present in DeNOx Catalysis

action at low temperature and a negative effect at high temperature on the reactions involved in the NOx storage process. The effect of water on the NOx adsorption is still a matter of debate in the literature. In line with our data, Svedberg et al. [53], report an increase in the low-temperature storage. Also Theis et al. [54] noticed an improved NOx storage efficiency in the presence of water, and suggest that H2 O enhances the spillover of NO2 from Pt sites to the NOx storage sites, possibly by the formation of nitric acid, which reacts with the NOx storage sites to form the nitrates. As opposite, Epling et al. [55] reported that water has a negative impact on NOx adsorption and discussed these effects in terms of the relative stability of the O/OH/CO3 /NO3 surface species, also considering how the relative stability of these species changes with temperature. To date, a clear picture able to reconcile this apparent controversy is still missing. Finally, the combined effect of 0.3% CO2 and 1% H2 O on the NOx storage capacity of the catalyst was also considered. The presence of water tends to compensate the negative effect of CO2 on the storage (particularly at breakthrough), and this eventually results in a lower but still significant storage capacity of the catalyst at any temperature (see Figure 6.9).

4. REDUCTION OF THE STORED NOx SPECIES 4.1. Reduction of NOx adsorbed species by hydrogen over Pt−Ba/-Al2 O3 catalysts A typical result obtained during the reduction by H2 at 350 C of NOx stored up to saturation at the same temperature over the Pt−Ba/-Al2 O3 (1/20/100 w/w) catalyst is presented in Figure 6.10. Upon the stepwise addition of 2000 ppm of H2 at t = 0 s the stored NOx was reduced to N2 . Indeed H2 was completely consumed while the N2 outlet concentration increased immediately to the level of 360 ppm and then it kept almost constant. Accordingly, at the beginning the reaction is very fast and selective to nitrogen, and limited by the concentration of H2 . Small amounts of NO were also observed, immediately after the H2 stepwise addition. Subsequently, after about 400 s the nitrogen concentration starts to decrease and, at the same time, evolution of hydrogen in the gas phase and formation of ammonia were observed. Ammonia is by far the most important by-product of the reduction of NOx adsorbed species over LNT systems, as also reported by several authors [11,27,50,51]. The reduction of NOx also produced water, which however did not desorb immediately, showing a delay of about 50 s due to adsorption onto the catalyst and most likely onto Ba sites to form Ba(OH)2 . The stepwise addition of hydrogen to the reactor was accompanied by a small increase of the catalyst temperature (3–5 C), due to the occurrence of the exothermic reduction, so that the run was actually performed in the absence of significant thermal effects. The following main reactions were thus likely involved in the reduction of stored NOx by H2 : BaNO2 2 + 3H2 → BaO + N2 + 3H2 O

(4)

BaNO3 2 + 5H2 → BaO + N2 + 5H2 O

(5)

Identification of the Reaction Networks of the NOx Storage/Reduction

Concentration, ppm

2000

193

H2inlet H2

1600

1200

800

N2

400

NO*10

NH3

0 0

1000

2000

3000

Time, s

Figure 6.10. Reduction of stored NOx with H2 (2000 ppm) balance He (total flow 100 Ncc/min, catalyst weight 60 mg) at 350 C over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst after storage at saturation at the same temperature. N2 , NH3 , NO and H2 are outlet concentrations, and H2 is inlet concentration.

BaNO3 2 + 8H2 → BaO + 2NH3 + 5H2 O

(6)

BaO + H2 O → BaOH2

(7)

It is worth noticing that the N2 concentration of 360 ppm in Figure 6.10 corresponded more likely to that expected from reaction (5) (400 ppm) and was significantly lower than that derived from reaction (4) (660 ppm). This eventually confirms that nitrates are the most abundant species when the storage process has been extended to saturation (Section 2 in Chapter 3). Notably, the treatment with H2 at 350 C leads to almost complete removal of the stored NOx ; indeed, only a minor fraction of nitrates remains on the surface after the regeneration procedure, as shown by dedicated TPD experiments performed after the rich phase [24].

4.2. Mechanistic aspects of the reduction of stored NOx by hydrogen The common idea on the mechanisms governing the reduction of NOx adsorbed species over LNT catalysts is that the regeneration process includes at first the release of NOx from the catalyst surface (i.e. from the alkali- or alkali-earth metal compound), followed by the reduction of the released NOx to N2 or other products [11]. The reduction of the released NOx in a rich environment is thought to occur according to the TWC chemistry and mechanisms; in particular, it was suggested that NO is decomposed on reduced Pt sites [38], or that a direct reaction occurs between released NOx species and the HC reductant molecules on the precious metal sites [39].

194

Past and Present in DeNOx Catalysis

In the case of LNT, different proposals have been advanced to explain the mechanisms governing the NOx release. Recent papers suggested that the NOx release is provoked by the heat generated upon the reducing switch (thermal release) [40], by the decrease of the gas-phase oxygen concentration that destabilizes the stored nitrates [41], by spillover and reduction of NO2 onto reduced Pt sites or by the establishment of a net reducing environment, which decreases the equilibrium stability of nitrates [12,42,43]. In order to gain further insight in the mechanism governing the reduction of adsorbed NOx species and to better elucidate the role of the different catalyst components (noble metal and storage component) in the process, mechanistic aspects of the reduction of stored NOx were extensively investigated in our labs [44,45]. NOx was stored on the catalyst surface under controlled conditions at 350 C (see Section 1 in Chapter 3); then the catalyst regeneration was performed at constant temperature by step addition of H2 (TRM), by thermal decomposition in He (TPD) and by heating in flowing H2 (TPSR). This allowed the analysis of the thermal stability/reactivity of the stored nitrates. The Pt-free sample [i.e. the binary Ba/-Al2 O3 (20/100 w/w) catalyst] was considered at first. In this case, the sample was saturated by feeding NO2 at 350 C (see Figure 6.2), since it has been shown that the adsorption with NO/O2 mixtures did not allow any significant storage of NOx species over the Pt-free sample (Section 2 in Chapter 3). The results of the TPD and TPSR experiments performed over the Ba/-Al2 O3 catalyst are displayed in Figure 6.11. In the case of the TPD experiment, no desorption peaks were observed below 350 C, i.e. below the adsorption temperature. The decomposition of nitrate species present on the catalyst surface is apparent only above 350 C, and the process is not yet completed at temperatures as high as 600 C. In line with several literature indications [25,28,33,35] decomposition of nitrates results in this case in the initial evolution of NO2 , followed by NO and O2 . The data hence indicate that nitrates do not appreciably decompose below the adsorption temperature during the TPD run performed under inert atmosphere. The stability/reactivity of stored nitrates was also investigated with H2 (TPSR experiment), and results are shown in Figure 6.11. Upon heating the catalyst in H2 after NOx storage at 350 C, H2 consumption was apparent only above 350 C, i.e. above the NOx adsorption temperature; H2 consumption was accompanied by the evolution of NO and of minor amounts of NO2 and N2 . The comparison with the TPD data showed that the presence of hydrogen did not affect significantly the temperature threshold for nitrate decomposition, but leads to a different product distribution. As suggested by Cant and Patterson [42], the observed product distribution can be ascribed to the occurrence of the following reactions: NO2 + H2 → NO + H2 O

(8)

NO + H2 → 1/2N2 + H2 O

(9)

Notably, these reactions are catalysed by the Ba component or involve a specific reactivity of adsorbed NOx since no reaction was observed between NO2 and H2 in an empty reactor up to 500 C. Hence, TPD and TPSR data obtained over the binary catalyst showed that the presence of a net reducing environment did not play any significant role in the decomposition/reduction of nitrates stored over Ba. Indeed a partial reduction of stored NOx could

Identification of the Reaction Networks of the NOx Storage/Reduction

195

800

Concentration, ppm

TPD 600

NO 400

NO2 200

O2

Concentration, ppm

0 800

600

2000

H2

TPSR

1500 400 1000

NO 200

500

NO2 0 100

N2 NH3

200

300

400

500

0 600

Temperature °C

Figure 6.11. The TPD in He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) and TPSR in H2 (2000 ppm) balance He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) after NO2 adsorption at 350 C over Ba/Al2 O3 (20/100 w/w) catalyst. NO, NO2 , O2 , N2 , NH3 and H2 are outlet concentrations.

only be obtained after their thermal decomposition from the catalyst, once released into the gas phase, NOx undergo a partial reduction by flowing hydrogen. As a matter of fact, when the reduction step was carried out at the same temperature used for adsorption, no reduction by hydrogen of the stored nitrates was observed. A different picture was obtained in the case of the reference Pt−Ba/-Al2 O3 (1/20/100 w/w) sample (Figure 6.12). In the case of the TPD experiment performed after NOx adsorption at 350 C, the decomposition of stored NOx species was observed only above 350 C. Evolution of NO and O2 was observed in this case, along with minor quantities of NO2 [25,28,33,35]. Complete desorption of NOx was attained already slightly below 600 C. As in the case of the binary Ba/-Al2 O3 sample, the data hence indicate that nitrates formed upon NO/O2 adsorption at 350 C followed by He purge at the same temperature, did not appreciably decompose below the adsorption temperature during the TPD run under inert atmosphere. It is, however, worth of note that over the ternary Pt−Ba/-Al2 O3 sample, the complete decomposition of stored NOx is achieved at lower temperatures if compared to the binary sample. This indicates that Pt promotes the rate of nitrate decomposition [45]. The

196

Past and Present in DeNOx Catalysis

observation that Pt affects the decomposition of stored nitrates is documented by several authors [11,16,25,28,41,42,49]: Coronado et al. [48] suggested that the decomposition of nitrates takes place at the interface between the Ba component and the noble metal, so that this latter favours the nitrate decomposition; Olsson et al. [16] invoked the presence of a spillover mechanism of NO2 from Ba to Pt to explain the decrease of desorption temperature of nitrates in Pt−Ba/-Al2 O3 as compared to Ba/-Al2 O3 . The TPD experiments were also performed after NOx adsorption at different temperatures, namely 300 and 400 C [44]. The temperature threshold for nitrate decomposition was always observed close to the adsorption temperature, thus indicating that the adsorption temperature rules the thermal stability of the NOx adsorbed species. A different picture was apparent upon heating the stored nitrates in H2 instead of He (TPSR run in Figure 6.12). Indeed, in this case, the reduction of the stored NOx was observed at temperatures as low as 140 C. The reaction was very fast and showed a complete consumption of H2 at 170 C with production of N2 and of significant amounts of NH3 as well. Temperature, °C 100 500

200

300

400

500

600

TPD

Concentration, ppm

400

NO 300

200

O2

100

NO2*10 0 500

500

TPSR 400

N2 T

300

300

H2/5 200

200

NH3 100

Temperature °C

Concentration, ppm

400

100

0

0 0

400

800

1200

1600

2000

Time, s

Figure 6.12. The TPD in He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) and TPSR in H2 (2000 ppm) balance He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) after NO/O2 adsorption at 350 C over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst (Batch B). NO, NO2 , O2 , N2 , NH3 and H2 are outlet concentrations.

Identification of the Reaction Networks of the NOx Storage/Reduction

197

The TPSR experiments were also performed after NOx adsorption at different temperatures (300 and 400 C) [44]. H2 consumption was always observed near 140 C, in spite of the fact that different adsorption temperatures were used. The data clearly indicated that the reduction of NOx ad-species is initiated at temperatures well below that of thermal decomposition; besides, the temperature onset for the reaction is not affected by the adsorption temperature. This proves that the reduction of stored NOx does not involve as a preliminary step the thermal release of NOx in the gas phase, but occurs through a Pt-catalysed chemical route, already active at low temperatures. To better address the role of Pt in the reduction mechanism, a physical mixture of the binary Pt/-Al2 O3 and Ba/-Al2 O3 samples was also prepared and tested. This catalytic system is constituted by all the components of the reference ternary system, even if in the physical mixture Pt and Ba are deposited on different alumina particles [44]. Also in this case, NOx was stored at 350 C with NO in the presence of O2 [29]. Then the stability/reactivity of stored nitrates was carried out by TPD experiment and H2 TPSR run, and results are displayed in Figure 6.13.

500

TPD

Concentration, ppm

400

300

NO

200

O2 100

NO2

0 500

H2 2000

Concentration, ppm

400

TPSR 300

1500

200

1000

NH3

100

NO

0 100

N2*5

200

300

400

500

500

0 600

Temperature °C

Figure 6.13. The TPD in He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) and TPSR in H2 (2000 ppm) balance He (heating rate 15 C/min, total flow 100 Ncc/min, catalyst weight 60 mg) after NO/O2 adsorption at 350 C over Pt/-Al2 O3 (1/100 w/w) −Ba/-Al2 O3 (20/100 w/w) physical mixture. NO, NO2 , O2 , N2 , NH3 and H2 are outlet concentrations.

198

Past and Present in DeNOx Catalysis

The TPD data showed that decomposition of adsorbed nitrates occurs at temperatures very close to that of adsorption. The process was not yet completed at temperatures as high as 600 C. This was in line with literature indications [11], claiming that NOx spillover processes from Ba to Pt (not possible on the physical mixture) could affect the nitrate decomposition process. The results of the TPSR experiment performed on the physical mixture showed a consumption of H2 only starting from 350 C with evolution of ammonia, along with minor amounts of nitrogen. Hence, as in the case of the Pt-free sample, also on the physical mixture, the reduction of the stored NOx is monitored at temperatures above those corresponding to their thermal decomposition. Therefore, the rate-determining step for NOx reduction on the Pt/-Al2 O3 −Ba/-Al2 O3 physical mixture is the thermal decomposition of NOx species adsorbed over Ba/-Al2 O3 ; once released in the gas phase, NOx are reduced over Pt/-Al2 O3 . These findings were further confirmed by data collected under isothermal conditions. Upon hydrogen addition at 350 C on the physical mixture previously saturated with NOx species at the same temperature, no reaction products were detected. This indicated that the stored nitrates could not be regenerated by H2 at constant temperature, i.e. without a prior release in the gas phase. Accordingly, the bulk of data collected over the different catalytic systems showed that Pt catalyses the reduction of stored NOx already at low temperature, and that the copresence of Pt and Ba on the same support is required for the occurrence of this route. A possible mechanism, which is consistent with our findings, is depicted in Figure 6.14. Several pathways can be suggested. The Pt-catalysed pathway may involve the activation of H2 on Pt sites, followed by its spillover on the alumina support towards nitrate ad-species (Figure 6.14a). The reduction of stored nitrates is hence operated because H ad-species promote the decomposition of nitrates to gaseous NO and/or NO2 , possibly via the intermediacy of nitrites [41]. NO/NO2 is then reduced on Pt. Alternatively, the reduction of the Pt sites creates a driving force for O atoms migration from the Ba site to Pt, resulting in the reduction of stored nitrates and leading to their destabilization/decomposition. Also, a mechanism involving the surface diffusion of NOx ad-species towards reduced Pt sites cannot be excluded (Figure 6.14b) [11]. In this case, NOx spills over the surface and is decomposed at reduced Pt sites. The role of the reductant in this mechanism is N2 H2O

(a) NO3

H2 Pt

H

NOx

BaO

Pt

Al2O3 N2 H2O

(b) NO3 BaO

H2 Pt Al2O3

NO3 BaO

Figure 6.14. Mechanistic proposals for NOx reduction over supported Pt−Ba catalysts.

Identification of the Reaction Networks of the NOx Storage/Reduction

199

to keep Pt in a reduced state. This mechanism is in line with the observed effect of Pt on the thermal decomposition of stored nitrates; it also implies a high mobility of NOx species adsorbed onto Ba. Finally, a specific route involving NOx ad-species present on Ba sites neighbouring Pt might also be suggested. In this respect, the interaction between Pt and Ba, already pointed out in [29] as a key factor for NOx adsorption, would play an important role also in determining activity and selectivity of nitrate reduction. The above mentioned mechanisms can explain the difference observed between the ternary Pt−Ba/-Al2 O3 sample and the Pt/-Al2 O3 −Ba/-Al2 O3 physical mixture where Pt and Ba sites are dispersed on different support particles. Still it is noted that in this case the reduction of NOx ad-species (i.e. catalyst regeneration) can be anyway accomplished by the thermal decomposition/reduction pathway discussed above and induced by thermal effects associated with lean-rich cycling in oxygen-containing exhausts. These aspects are currently under investigation in our labs.

4.3. Effect of the operating conditions on the reduction of NOx adsorbed species by H2 The effect of different operating conditions (e.g. temperature, presence of CO2 and water) on the reduction process by H2 was studied in a large temperature range (200–400 C) over the ternary Pt−Ba/-Al2 O3 catalyst. In all cases, the catalyst was previously saturated with NO/O2 mixtures at the same temperature used for the reduction step (Sections 1 and 6 in Chapter 3).

4.3.1. Effect of temperature on the NOx adsorption–reduction The results obtained during the reduction phase carried out at 200, 300 and 400 C in the absence of CO2 and water after NO/O2 adsorption at the same temperatures are presented in Figure 6.15. In all cases, the features of the data collected at different temperatures are very similar to those obtained at 350 C (Figure 6.10). Hydrogen is completely consumed at the beginning of the reduction process, indicating that the reaction is fast and limited by the concentration of H2 , already at 200 C. The reduction of the NOx adsorbed species is monitored by the N2 outlet concentration, which sharply increases to about 360 ppm, and then it keeps constant until depletion of the reactive stored NOx . Notably, the adsorption processes at different temperatures led to different amounts of stored NOx (Section 6 in Chapter 3), and this explained the different extent of the reduction reaction. Small amounts of NO and of N2 O were observed at the beginning of the reduction process. However, these amounts were always small thus resulting in high selectivity towards N2 , which was close to 95% in both cases. Notably, as opposite to the case presented in Figure 6.10, in these experiments ammonia was never detected as a by-product. This feature was ascribed to the fact that these data were obtained over a different batch of the reference Pt−Ba/-Al2 O3 catalyst having apparently the same structural and morphological characteristics of the sample whose results have been reported in Figure 6.10, if one excludes a slight difference in the Pt dispersion (65% vs 50%). Accordingly, this indicates that small differences in the catalyst characteristics may lead to non-negligible effects in the catalytic behaviour. This is still an open question, which deserves further

200

Past and Present in DeNOx Catalysis 2000

H2O

H2inlet 400°C

1500

400°C

300°C

H2 300°C

1000

Concentration, ppm

200°C 200°C

500

0 400

NO

N2 300 300°C 200 400°C 400°C 100 200°C 0

200°C 0

300

Time, s

600

900

0

300

600

900

Time, s

Figure 6.15. Effect of temperature on stored NOx reduction in H2 (2000 ppm) balance He (total flow 200 Ncc/min, catalyst weight 120 mg) over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst at 200–300–400 C. H2 , N2 , H2 O and NO are outlet concentrations, and H2 is inlet concentration.

studies. In any case, data not reported herein showed that ammonia production is also affected by the operating parameters including temperature, amounts of adsorbed NOx , etc [56]. Water was also produced in the reduction and was adsorbed on the Ba sites to form Ba(OH)2 . As already reported, this eventually accounted for the observed delay in the outlet H2 O concentration, which slightly increased with temperature. The traces of H2 O and H2 continued to change after the N2 outlet concentration has diminished to zero, in particular at low temperature. This could be ascribed both to desorption of water previously stored as Ba(OH)2 , and to the reduction of poorly reactive catalyst oxygen species.

4.3.2. Effect of CO2 presence on the NOx adsorption–reduction The effect of the presence of CO2 during the reduction of stored NOx at different temperatures is presented in Figure 6.16. Also in this case the reaction is fast and H2 is completely consumed upon its addition. At any temperature, the N2 outlet concentration increased sharply to the maximum value of about 200 ppm at 200 C and of about 360 ppm at higher temperature, and then it kept constant until depletion of the reactive stored NOx . The reaction was limited by the H2 concentration above 250 C (data not reported in the figure), but not at 200 C where H2 spillover had to be invoked together with

Identification of the Reaction Networks of the NOx Storage/Reduction 2000

201 2000

H2inlet

H2O

300°C

200°C

1500

1500 400°C

400°C

1000

1000

500

500

H2

200°C 0

0 600

CO2 400°C

N2 or CO

400

3900

3600

Concentration, ppm

Concentration, ppm

300°C

200°C

300°C

3300 200

300°C

0 0

200

3000

400°C

200°C 400

Time, s

600

800

0

200

400

600

800

2700

Time, s

Figure 6.16. Effect of the CO2 presence on stored NOx reduction in H2 (2000 ppm)/CO2 (3000 ppm) balance He (total flow 200 Ncc/min, catalyst weight 120 mg) over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst at 200–300–400 C. H2 , N2 /CO, H2 O and CO2 are outlet concentrations, and H2 is inlet concentration.

the inhibiting effect of CO2 on the reduction of the stored NOx to account for the complete consumption of H2 and for the relatively low concentration of N2 [below the stoichiometry of reaction (5)] at the same time. The inhibiting effect of CO2 at low temperature is likely related to the formation of CO (by reduction with H2 ), which then poisons the Pt sites [11]. The trace of the mass-to-charge signal 44, which was ascribed to CO2 , was complex. Significant amounts of CO2 were desorbed immediately after the addition of H2 , then CO2 was consumed until depletion of stored NOx . The desorption of CO2 was likely due to the occurrence of the following reaction: BaCO3 + H2 O → BaOH2 + CO2

(10)

Water was produced through the reduction of stored NOx and was detected at the reactor exit with a time delay of about 50 s, that compared well with the characteristic time of CO2 desorption. Likewise, the consumption of CO2 was ascribed to the reverse of reaction (10), which implied readsorption of CO2 on BaO/Ba(OH)2 once NOx had been reduced. After all stored reactive NOx groups were reduced and at sufficiently high temperature, i.e. at temperature >300 C, the mass-to-charge signal 28 increases again and reaches an asymptotic value that was higher at high temperature. GC analyses confirmed that in this case the mass-to-charge signal 28 was associated entirely with CO. The increase

202

Past and Present in DeNOx Catalysis

in the CO outlet concentration was accompanied by the parallel consumption of H2 (the H2 outlet concentration was lower at high temperature), and of CO2 (the CO2 trace does not recover the level corresponding to CO2 feed concentration), and by the parallel production of H2 O (the H2 O trace kept high until H2 shutoff). It was, therefore, concluded that, after all stored reactive NOx groups have been reduced and provided that the temperature was sufficiently high, the catalyst promoted the reverse water gas shift (WGS) reaction: CO2 + H2 ⇔ CO + H2 O

(11)

The WGS reaction was limited by thermodynamics under the experimental conditions employed in this study, since the value of the reaction constant Ksp (based on the asymptotic concentration of reactants and products) was close to the value of the equilibrium constant Keq . Storage–reduction experiments were also performed with synthetic exhaust and by reducing gases containing 3% CO2 at 300 C, as well as at 350 C, and the same effects described above were observed.

4.3.3. Effect of H2 O presence on the NOx adsorption–reduction The effect of H2 O during the reduction of stored NOx at 200, 300 and 400 C after adsorption at the same temperatures is presented in Figure 6.17. The reaction is fast and H2 is fully consumed immediately upon its addition. Furthermore, the reaction is highly selective, since NO and N2 O were not detected, not even in trace amount, during reduction. It is worth of note that at 200 C, immediately upon the H2 step addition, the outlet N2 concentration of about 500 ppm has been detected, which suggests that nitrites are present in significant amount upon NOx storage up to catalyst saturation. Indeed, 500 ppm N2 is a concentration significantly higher than that expected for the reduction of nitrates [400 ppm, from reaction (5)], even neglecting the consumption of H2 due to the reduction of trace amount of gaseous oxygen, and it is lower than that expected for the reduction of nitrites [600 ppm, from reaction (4)], considering in this case the consumption of hydrogen caused by the reduction of trace amounts of gaseous oxygen. This implies that water inhibits the reactions responsible for the formation of nitrates. Still, the promoting effect of water at low temperature on NOx stored up to the NOx breakthrough and up to catalyst saturation might be associated with the participation of surface hydroxyls in the storage process. This aspect deserves further study and will be investigated in the future.

4.3.4. Effect of the Ba loading on the NOx adsorption–reduction The results obtained in the case of catalysts having different Ba loadings (0–30%, w/w) during the reduction by H2 of stored NOx at 350 C is presented in Figure 6.18. As previously discussed (Section 3 in Chapter 3), the Ba-free catalyst showed no significant storage capacity of NO under lean condition and accordingly no N2 was formed upon the H2 switch over this sample. A different picture was apparent for the

Identification of the Reaction Networks of the NOx Storage/Reduction 2000

203

H2inlet

1600 1200

H2

400°C 200°C

300°C

Concentration, ppm

800 400 0

H2O

300°C 12000 200°C 400°C 10000 200°C 300°C

400

N2

200 400°C 0 0

300

600

900

Time, s

Figure 6.17. Effect of the H2 O presence on stored NOx reduction in H2 (2000 ppm)/H2 O (1%, v/v) balance He (total flow 200 Ncc/min, catalyst weight 120 mg) over Pt−Ba/Al2 O3 (1/20/100 w/w) catalyst at 200–300–400 C. H2 , N2 and H2 O are outlet concentrations, and H2 is inlet concentration.

Pt−Ba/-Al2 O3 (1/5.3/100 w/w) catalyst: upon H2 admission (t = 0 s), the NOx adsorbed species were readily reduced as indicated by the complete H2 consumption and the correspondent evolution of N2 . Data reported in Figure 6.18 indicated the almost stoichiometric occurrence of reaction (4) (the N-balance closed within 5%), while very small amounts of NO (not shown in the figure) were formed during the catalyst regeneration and the reduction process appeared to be very selective (Figure 6.19). When the Ba loading increased up to 10% (w/w) (Pt−Ba/-Al2 O3 1/11.5/100 w/w sample, Figure 6.18c), the reduction step was longer than in the previous case, in line with the higher amounts of NOx species that the catalyst was able to store during the previous adsorption phase (see Figure 6.6). Also in this case, low amounts of NO were found at the reactor outlet so that the process presented a selectivity rate close to 100% (Figure 6.19). A slightly different picture was observed for the other systems (Figure 6.18d–f). In line with the greater amounts of NOx stored at higher Ba content, the reduction step is longer. Furthermore, the formation of NH3 was also observed, whose amounts increased when increasing the Ba loading of the catalyst. As a matter of fact, N2 selectivity values as low as 30% have been measured for Ba loadings near 30% (w/w)

204

Past and Present in DeNOx Catalysis

Concentration, ppm

2400 2100

Concentration, ppm

H2 in

(a) H2

(b) H2

1500 1200 900

N2

600 300 0 2400 2100

N2 H2 in

H2 in

(c)

(d)

1800

H2

H2

1500 1200 900 600

N2

N2

300 0 2400

Concentration, ppm

H2 in

1800

2100

(e)

H2 in

NH3

(f)

H2 in

1800 1500

H2

1200

H2

900 600

N2

N2

NH3

300

NH3

0

0

500

1000

1500

0

500

1000

Time, s

1500

2000

2500

300

Time, s

Figure 6.18. Reduction run in H2 (2000 ppm) balance He (total flow 100 Ncc/min, catalyst weight 60 mg) at 350 C over Pt−Ba/Al2 O3 (1/x/100 w/w) catalysts. (a) x = 0; (b) x = 53; (c) x = 115; (d) x = 20; (e) x = 30; (f) x = 43. NO, N2 , NH3 and H2 are outlet concentrations, and H2 is inlet concentration. 100 90

Selectivity (%)

80 70 60 50 40 30 20 10 0

5

10

15

20

25

30

Ba loading (%)

Figure 6.19. Selectivity behaviour during reduction of the adsorbed NOx species (reduction phase) upon NOx adsorption (NOx storage) vs Ba loading.

Identification of the Reaction Networks of the NOx Storage/Reduction

205

(Pt−Ba/-Al2 O3 = 1/43/100 w/w, see Figure 6.19). This behaviour could be ascribed to the decrease in the Pt dispersion values measured on the systems with high Ba content, but other explanations are possible. Indeed, a low selectivity was also associated with high storage capability and the marked decrease was observed for Ba coverage approaching the monolayer formation [27]. Furthermore, specific experiments carried out in our labs showed that the formation of NH3 depended also on the amounts of stored NOx , on temperature and on H2 concentration [56]. Accordingly, the match of data collected on the systems with different Ba loading evidenced that the reduction process is complex and that it can lead to nitrogen and/or ammonia depending on the formulation and properties of the catalysts and on the adopted experimental conditions.

5. CONCLUSIONS This paper illustrates the results of an extensive investigation carried out in our labs during the past few years on the analysis of Pt−Ba/Al2 O3 ‘LNT’ catalysts. The mechanisms of the NOx storage phase and of their subsequent reduction are addressed. Accordingly, the reaction network of the NOx storage and reduction over Pt−Ba/Al2 O3 is presented and discussed taking also into account to the effect of the operating conditions on both phases. The NOx storage was investigated at first. The adsorption of NO and NO2 in the presence and in the absence of oxygen was performed over model Pt−Ba/-Al2 O3 based samples under realistic operating conditions (temperature, NOx content, etc.), by imposing stepwise changes in the inlet NOx concentration while analysing the gas phase composition at the reactor outlet. A dual pathway for NO/O2 adsorption over Pt−Ba/-Al2 O3 systems was proposed. In addition to the acknowledged oxidation of NO to NO2 , and its subsequent adsorption via disproportionation to form nitrates (nitrate route), a novel route has been proposed, which implies a stepwise oxidation of NO in the presence of oxygen to form nitrite ad-species, which are progressively oxidized to nitrates (nitrite route). Then the reduction of stored NOx with hydrogen was addressed. The stability/reactivity of the NOx adsorbed species was analysed under different atmospheres (inert and reducing) both at constant temperature and under temperature programming. The bulk of data pointed out that in the absence of significant thermal effects in the catalyst bed, the reduction of stored nitrates occurs through a Pt-catalysed surface reaction that does not involve the thermal desorption of the stored nitrates as a preliminary step. A specific role of a Pt−Ba interaction was suggested, which plays a role in the NOx storage phase as well.

ACKNOWLEDGEMENTS Federica Prinetto and Giovanna Ghiotti from University of Torino are gratefully acknowledged for FTIR analysis.

206

Past and Present in DeNOx Catalysis

REFERENCES [1] “Transportation and Air Quality” and related documents in http://www.epa.gov./otaq/ [2] Health Assessment Document for Diesel Engine Exhaust, EPA/600/8-90/057F, May 2002. [3] Johnson, T. The Environmental Benchmark for Heavy-Duty Applications. Keynote Lecture at ASME 2002 Spring Technical Conference. [4] Guibet, J.C. and Faure-Birchem, E. (1999) Fuels and Engines: Technology, Energy, Environment, Volume 1. Technip, Paris, pp. 1–385. [5] Miyoshi, N., Matsumoto, S., Katoh, K. et al. (1995) Development of New Concept Three-Way Catalyst for Automotive Lean-Burn Engines, SAE Technical Paper 950809. [6] Matsumoto, S. (2000) Catalytic Reduction of Nitrogen Oxides in Automotive Exhaust Containing Excess Oxygen by NOx Storage-Reduction Catalyst, Cat. Tech., 4, 102. [7] Engstrom, P., Amberntsson, A., Skoglundh, M. et al. (1999) Sulphur Dioxide Interaction with NOx Storage Catalysts, Appl. Catal. B Environ., 22, L241. [8] Gobel, U., Hoehne, J., Lox, E.S. et al. (1999) Durability Aspects of NOx Storage Catalysts for Direct Injection Gasoline Vehicles, SAE Technical Paper 99FL-103. [9] Shinjoh, H., Takahashi, N., Yokota, K. et al. (1998) Effect of Periodic Operation over Pt Catalysts in Simulated Oxidizing Exhaust Gas, Appl. Catal. B Environ., 15, 189. [10] Hodjati, S., Petit, C., Pitchon, V. et al. (2000) Absorption/Desorption of NOx Process on Perovskites: Nature and Stability of the Species formed on BaSnO3 , Appl. Catal. B Environ., 27, 117. [11] Epling, W.S., Campbell, L.E., Yezerets, A. et al. (2004) Overview of the Fundamental Reactions and Degradation Mechanisms of NOx Storage/Reduction Catalysts, Catal. Rev., 46, 163. [12] Fridell, E., Skoglundh, M., Westerberg, B. et al. (1999) NOx Storage in Barium-Containing Catalysts, J. Catal., 183, 196. [13] Fridell, E., Persson, H., Westerberg, B. et al. (2000) The Mechanism for NOx Storage, Catal. Lett., 66, 71. [14] Broqvist, P., Panas, I., Fridell, E. et al. (2002) NOx Storage on BaO(100) Surface from First Principles: A Two Channel Scenario, J. Phys. Chem. B, 106, 137. [15] Westerberg, B. and Fridell, E. (2001) A Transient FTIR Study of Species Formed During NOx Storage in the PtBa/Al2 O3 System, J. Mol. Catal. A Chem., 165, 249. [16] Olsson, L., Persson, H., Fridell, E. et al. (2001) A Kinetic Study of NO Oxidation and NOx Storage on Pt/Al2 O3 and PtBa/Al2 O3 , J. Phys. Chem. B, 105, 6895. [17] Anderson, J.A., Bachiller-Baeza, B. and Fernández-García, M. (2003) Role of Pt in Pt/Ba/Al2 O3 NOx Storage and Reduction Traps, Phys. Chem. Chem. Phys., 5, 4418. [18] Hess, C. and Lunsford J.H. (2002) Mechanism for NO2 Storage in Barium Oxide Supported on Magnesium Oxide Studied by in Situ Raman Spectroscopy, J. Phys. Chem. B, 106, 6358. [19] Hess, C. and Lunsford, J.H. (2003) NO2 Storage and Reduction in Barium Oxide Supported on Magnesium Oxide Studied by in Situ Raman Spectroscopy, J. Phys. Chem. B, 107, 1982. [20] Huang, H.Y., Long, R.Q. and Yang, R.T. (2001) The Promoting Role of Noble Metals on NOx Storage Catalyst and Mechanistic Study of NOx Storage under Lean-Burn Conditions, Energy Fuels, 15, 205. [21] Mahzoul, H., Brilhac, J.F. and Gilot, P. (1999) Experimental and Mechanistic Study of NOx Adsorption over NOx Trap Catalysts, Appl. Catal. B Environ., 20, 47. [22] Schmitz, P.J. and Baird, R.J. (2002) NO and NO2 Adsorption on Barium Oxide: Model Study of the Trapping Stage of NOx Conversion via Lean NOx Traps, J. Phys. Chem. B, 106, 4172. [23] Cant, N.V. and Patterson, M.J. (2002) The Storage of Nitrogen Oxide on Alumina-Supported Barium Oxide, Catal. Today, 73, 271. [24] Lietti, L., Forzatti, P., Nova, I. et al. (2001) NOx Storage Reduction over Pt–Ba/ -Al2 O3 Catalyst, J. Catal., 204, 175.

Identification of the Reaction Networks of the NOx Storage/Reduction

207

[25] Prinetto, F., Ghiotti, G., Nova, I. et al. (2001) FT-IR and TPD Investigation of the NOx Storage Properties of BaO/Al2 O3 and Pt-BaO/Al2 O3 Catalysts, J. Phys. Chem., 105, 12732. [26] Nova, I., Castoldi, L., Prinetto, F. et al. (2004) Top. Catal., 30/31, 181. [27] Castoldi, L., Nova, I., Lietti, L. et al. (2004) NOx Adsorption Study over Pt–Ba/Alumina Catalysts: FT-IR and Reactivity Study, Catal. Today, 96, 43. [28] Nova, I., Castoldi, L., Prinetto, F. et al. (2004) Ox Adsorption Study over Pt–Ba/alumina Catalysts: FT-IR and Pulse Experiments, J. Catal., 222(2), 377. [29] Nova, I., Castoldi, L., Lietti, L. et al. (2005) The Pt-Ba Interaction in Lean NOx Trap Systems, SAE Technical Paper 2005-01-1085. [30] Scotti, A., Nova, I., Tronconi, E. et al. (2004) Kinetic Study of Lean NOx Storage over the Pt-Ba/Al2 O3 System, Ind. Eng. Chem. Res., 43, 4522. [31] Szailer, T., Kwak, J.H., Kim, D.H. et al. (2006) Effects of Ba Loading and Calcination Temperature on BaAl2 O4 Formation for BaO/Al2 O3 NOx Storage and Reduction Catalysts, Catal. Today, 114, 86. [32] Kim, D.H., Chin, Y.H., Kwak, J.H. et al. (2005) Changes in Ba Phases in BaO/Al2 O3 upon Thermal Aging and H2 O Treatment, Catal. Lett., 105, 259. [33] Szanyi, J., Kwak, J.H., Hanson, J. et al. (2005) Changing Morphology of BaO/Al2 O3 During NO2 Uptake and Release, J. Phys. Chem. B, 109, 7339. [34] Kabin, K.S., Khanna, P., Muncrief, R.L. et al. (2006) Monolith and TAP Reactor Studies of NOx Storage on Pt/BaO/Al2 O3 : Elucidating the Mechanistic Pathways and Roles of Pt, Catal. Today, 114, 72. [35] Piacentini, M., Maciejewski, M., Bürgi, T. et al. (2004) Role of Support in Lean DeNOx Catalysis, Top. Catal., 30/31, 71. [36] Fanson, P.T., Horton, M.R., Delgass, W.N. et al. (2003) FTIR Analysis of Storage Behavior and Sulfur Tolerance in Barium-Based NOx Storage and Reduction (NSR) Catalysts, Appl. Catal. B Environ., 46, 393. [37] Laurent, F., Pope, C.J., Mahzoul, H. et al. (2003) Modelling of NOx Adsorption over NOx Adsorbers, Chem. Eng. Sci., 58, 1793. [38] Burch, R., Breen, J. and Meunier, F. (2002) A Review of the Selective Reduction of NOx with Hydrocarbons under Lean-Burn Conditions with Non-Zeolitic Oxide and Platinum Group Metal Catalysts, Appl. Catal. B Environ., 39, 283. [39] Maunula, T., Ahola, J. and Hamada, H. (2000) Reaction Mechanism and Kinetics of NOx Reduction by Propene on CoOx /Alumina Catalysts in Lean Conditions, Appl. Catal. B Environ., 26, 173. [40] Kabin, K.S., Muncrief, R.L. and Harold, M.P. (2004) NOx Storage and Reduction on a Pt/BaO/Alumina Monolithic Storage Catalyst, Catal. Today, 96, 79. [41] Liu, Z. and Anderson, J.A. (2004) Influence of Reductant on the Thermal Stability of Stored NOx in Pt/Ba/Al2 O3 NOx Storage and Reduction Traps, J. Catal., 224, 18. [42] Cant, N.W. and Patterson, M.J. (2003) The Effect of Water and Reductants on the Release of Nitrogen Oxides Stored on BaO/Al2 O3 , Catal. Lett., 85, 153. [43] Poulston, S. and Rajaram, R. (2003) Regeneration of NOx Trap Catalysts, Catal. Today, 81, 603. [44] Nova, I., Castoldi, L., Lietti, L. et al. (2006) A Low Temperature Pathway Operating the Reduction of Stored Nitrates in Pt-Ba/Al2 O3 Lean NOx Trap Systems, SAE Technical Paper, 2006-01-1368. [46] Szailer, T., Kwak, J.H., Kim, D.H. et al. (2006) Reduction of Stored NOx on Pt/Al2 O3 and Pt/BaO/Al2 O3 Catalysts with H2 and CO, J. Catal., 239, 51. [47] James, D., Fourré, E., Ishii, M. et al. (2003) Catalytic Decomposition/Regeneration of Pt/Ba(NO3 2 Catalysts: NOx Storage and Reduction, Appl. Catal. B Environ., 45, 147. [48] Coronado, J.M. and Anderson, J.A. (1999) FTIR Study of the Interaction of NO2 and Propene with Pt/BaCl2 /SiO2 , J. Mol. Catal. A Chem., 138, 83.

208

Past and Present in DeNOx Catalysis

[49] Sharma, M., Harold, M.P. and Balakotaiah, V. (2005) Analysis of Periodic Storage and Reduction of NOx in Catalytic Monoliths, Ind. Eng. Chem. Res., 44, 6264. [50] Lesage, T., Terrier, C., Bazin, P. et al. (2003) Phys. Chem. Chem. Phys., 5, 4435. [51] Abdulhamid, H., Fridell, E. and Skoglundh, M. (2004) Influence of the Type of Reducing Agent (H2 , CO, C3 H6 and C3 H8  on the Reduction of Stored NOx in a Pt/BaO/Al2 O3 Model Catalyst, Top. Catal., 30/31, 161. [52] Frola, F., Prinetto, F., Ghiotti, G. et al. (in press) Catal. Today. [53] Svedberg, P., Jobson, E., Erkfeldt, S. et al. (2004) Influence of the Storage Material on the Storage of NOx at Low Temperatures, Top. Catal., 30/31, 199. [54] Theis, J.R., Jen, H.W., McCabe, R.W. et al. (2006) Reductive Elimination as a Mechanism for Purging a Lean NOx Trap, SAE Technical Paper 2006-01-1067. [55] Epling, W.S., Campbell, G.C. and Parks, J.E. (2003) The Effects of CO2 and H2 O on the NOx Destruction Performance of a Model NOx Storage/Reduction Catalyst, Catal. Lett., 90, 45. [56] Nova, I., Castoldi, L., Lietti, L. et al. (2007) How to Control the Selectivity in the Reduction of NOx with H2 over Pt-Ba/Al2 O3 Lean NOx Trap Catalysts, Top. Catal., 42/43, 21.

PART 2

Novel developments and future trends to ensure continuous restrictive standard regulations

This page intentionally left blank

Chapter 7

CURRENT TASKS AND CHALLENGES FOR EXHAUST AFTER-TREATMENT RESEARCH: AN INDUSTRIAL VIEWPOINT J. M. Trichard∗ Renault SA, Technocentre Renault, 1 Avenue du Golf, 78288 Guyancourt Cedex, France Corresponding author: Renault SA, Technocentre Renault, 1 Avenue du Golf, 78288 Guyancourt Cedex, France. Tel.: +33 1 76 85 74 58, Fax: +33 1 76 85 77 16, E-mail: [email protected]

Abstract The market of Diesel vehicles in Western Europe is, presently, still growing. In Asia, this market is still in an embryonic state (but should result as gigantic), while tomorrow the USA could become ‘The Market’. Diesel technology is an important issue for the carmakers, because it emits noticeably less ‘Greenhouse gas’ than its gasoline counterpart. Compliance with the EuroIII standards (2000) forced the fitting of Diesel oxidation catalysts (DOC) in the exhaust line [for the after-treatment of unburnt hydrocarbons (HC) and carbon monoxide (CO)]. Additionally, the exhaust gas recirculation (EGR) was adapted to reduce the engine-out emissions of nitrogen oxides (NOx ). Under the EuroIV regulation (2005), the Diesel Particle Filter (DPF) emerged and is necessary for the ‘heavily loaded’ applications. Renault plans to make a CSF offer for all the diesel product range. This system enables to treat nearly 100% of the particle matter (PM, soot). When loaded, the DPF requires a regeneration phase to burn the accumulated soot. This regeneration is mandatory to keep not only the engine performance and reliability, but also the DPF reliability itself. The main drawbacks of this regeneration are the dilution of ‘post injected’ fuel in engine lubricant, leading to reducing the lubricant draining interval; the fuel over-consumption and the risk of forced vehicle immobilizations (for example, filter clogging under city driving conditions). The actual issues of EuroV standards aim at optimizing engine’s design to decrease the engine-out NOx emissions in order to avoid the need for ‘expensive’ after-treatments in the exhaust line. Only some ‘heavily loaded’ applications would need such NOx after-treatment. Today, two major technological ways of NOx treatment are identified: the NOx Trap and the selective catalytic reduction with ammonia (SCR-NH3 ). After an adsorption period (during which the system traps the NOx ), the NOx Trap requires a regeneration phase (in fact a rich incursion with HC and CO injections) enabling the reduction of accumulated NOx . As for the DPF, the engine combustion strategies carried on imply an Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

212

Past and Present in DeNOx Catalysis

increase of fuel diluted in engine lubricant, but also supplementary unburnt HC emissions, which could jeopardize the compliance with EuroV standards. This after-treatment system is also poisoned by sulfur (present in fuel and in engine lubricant) and must be recurrently desulfated (these desulfations are more severe than regenerations, rising concerns about the engine reliability, the fuel over-consumption and the emitted HC). In spite of these efforts, a part of sulfur adsorption on the NOx Trap active sites (normally dedicated to NOx storage) becomes irreversible and a slow process of system degeneration has to be addressed. The SCR-NH3 is a continuous process for NOx treatment and shows very efficient treatment efficiency. But this technology needs to put in the vehicle an additional tank for the urea storage. Moreover, this technology is constrained from an architectural point of view because two DOCs are necessary before and after the DeNOx catalyst to hydrolyze urea and form ammonia and to prevent the NH3 release in the exhaust line. In the challenge for preservation of the Environment, the European carmakers favor Diesel, which must comply with the most stringent standards constraints. The real key point for the future is, on the one hand, the improvement of DPF regeneration phase and on the other hand, but as first priority, the development of alternative NOx after-treatment technologies reliable on a long period. End of Abstract Nowadays, the Diesel vehicles market in Western Europe keeps growing. Diesel technology is of interest for car manufacturers because the ‘Greenhouse gas’ emissions are noticeably lower with Diesel than with gasoline (do not forget the ACEA1 commitment for 2008: 140 g/km of CO2 for the new cars sold in European Union). A major issue for the Diesel vehicles is to clean-up HC, CO, NOx and soot particles released by the engine at a minimum cost. The evolution of Diesel standards will be discussed and the constraints and limits of Diesel after-treatment systems put in series or in development will be clarified (in particular, those dealing with NOx conversion and soot particles removal).

1. DIESEL EUROPEAN STANDARDS REVIEW 1.1. Diesel standards from 1996 to 2005 (EuroII to EuroIV) The four pollutants in Diesel submitted to European standards are: • • • •

overall unburnt hydrocarbons (HC), carbon monoxide (CO), nitrogen oxides (NOx ), soot particles (PM).

The European standards levels for HC, CO, NOx and PM emissions have been diminishing since 1996 (Figure 7.1). Compliance with EuroIII standards (applied in 2000) has led to the implementation of Diesel oxidation catalyst (DOC) in the exhaust line to convert the residual HC and CO released by the Diesel engine. Additionally, the exhaust gas recirculation (EGR) system at the engine intake has been mandated to reduce NOx emissions. This device mixes a part of exhaust gases

1

ACEA: European Carmakers Association

Current Tasks and Challenges for Exhaust After-Treatment Research

213

PM (g/km) 0,08

EUROII (1996)

0,05

EUROIII (2000)

0,025

EUROIV (2005) 0,005 1 0,64 0,5 0,5

CO (g/km)

EUROV? 0,05

0,2 0,25 0,5 0,7

NOx (g/km)

0,3 0,56 0,7

HC (g/km)

Figure 7.1. Diesel emissions standards from EuroII to EuroV on a four-axis space corresponding to the pollutants: soot particles, HC, NOx and CO.

with freshly admitted air. By this way, the oxygen content of the intake air is lowered and the combustion temperature is reduced, leading to the reduction of engine-out NOx emissions. But another consequence of the EGR system is an increase in particles formation due to an increase in the engine local richness2 . As the droplets of fuel injected in the combustion chamber are therefore less prone to react with oxygen, a fraction of these droplets is pyrolyzed instead of being oxidized. The pyrolysis process leads, then inexorably, to an increase in soot particles emissions. The EGR system implies a NOx /particle compromise. Two ways can be used to adjust the NOx /particle compromise and allow to maintain the soot particles and NOx emission levels below the EuroIII standards: • by controlling the recirculated gas rates, • by modifying EGR distribution channels. Overall, EGR and combustion/injection systems constitute the key factors to comply with the EuroIV standards (applied in January, 2005). The EuroIV step exhibits EuroIII NOx and soot particles limits divided by 2. Besides, vehicle’s weight is always increasing due to the introduction of new safety systems and equipment. Therefore, pollutants emissions increase and a supplementary effort to reach the normative threshold is to be made. To comply with this target, some evolutions have been introduced, as for example multi-injection or water-cooling of the EGR system. The NOx /particle compromise adjustment remains possible for most of the applications without any after-treatment system like the Diesel particle filter (DPF).

2

richness = air/fuel ratio

214

Past and Present in DeNOx Catalysis

1.2. EuroV and EuroVI Diesel standards hypothesis and its consequences 1.2.1. EuroV and its consequences Assumptions for EuroV standards have been proposed by the European Commission at the end of December 2005 (see Figure 7.1). With a limit of soot particles emissions proposed at 5 mg/km, the DPF has to be installed in the exhaust line and becomes mandatory for EuroV (the date of application for EuroV is unknown, of course, but foreseen in the middle of 2009). To reach the NOx emissions limits of 200 mg/km, the requested EGR rates lead to a consequent increase in particle emissions. From a legislative point of view, the totality of the particle emissions could be eliminated by the DPF. However, this situation would generate an increase in DPF loading rate whose consequences are unreasonable (discussed in the following sections). This adjustment principle reaches its limit in a more stringent legislative context when NOx and particle emissions thresholds diminish. It was thus necessary to modify the EGR system in a way to minimize the engine-out NOx emissions. In a very macroscopic point of view, EGR is used in EuroV context to introduce an inert gas into the combustion chamber. The main purpose of this gas is to absorb a part of the calories released by combustion. Therefore, the combustion temperature decreases and so does the NOx engine-out. However, to prevent excessive increase in particle emissions, the stoichiometry of combustion must be respected and a sufficient quantity of oxygen must be introduced in the combustion chamber in a way to minimize the pyrolysis phenomenon. This is achieved by densification of EGR gases: at a colder temperature, the gas expansion is reduced and allows the introduction of a larger quantity of oxygen. To illustrate the two fashions of EGR management and their consequences in the NOx /particle compromise management refer to Figure 7.2. As a conclusion, with a pertinent densification of EGR gases, it is possible to decrease NOx emissions, while keeping under control the increase in particles emissions at a level compatible with the DPF treatment ability. 0.025

NOx /PM trade-off using ‘’EuroIV’’ EGR management

0.02

PM [g/km]

t

en em ag tion n a a m fic R odi EG m

0.015

EuroIV

0.01

0.005

EuroV NOx /PM trade-off using ‘’EuroV’’ EGR management

0.05

0.1

0.15

0.2

NOx [g/km]

Figure 7.2. Both philosophies of exhaust gas recirculation management.

0.25

Current Tasks and Challenges for Exhaust After-Treatment Research

215

0.025

PM [g/km]

0.02

tt en

em ag ion an at m fic R odi EG m

0.015 0.01 0.005

NOx /PM trade-off trade-off using using NOx/PM ‘’EuroIV’’ EGR EGR management management ‘’EuroIV’’

EuroIV EuroIV

EuroV EuroV

0.05

NOx /PM trade-off using ‘’EuroV’’ EGR management

0.1

0.15 0.2 NOx [g/km]

0.25

HC

Faces drastic increase of HC engine-out emissions

NOx

1 1

Figure 7.3. NOx /HC compromise in EuroV, which substitutes for NOx /particle compromise in EuroIV.

However, even if this strategy seems very attractive, the EGR gases densification also shows several drawbacks (see Figure 7.3). Because of a lower temperature range, the combustion no longer allows the complete oxidation of the fuel carbon species that can be released in the exhaust line in the form of unburnt hydrocarbons. With EuroV constraints, the previous NOx /particle compromise is now substituted by a new one: the NOx /HC compromise. This NOx /HC compromise must be controlled in EuroV and will limit the use of the ‘cooled’ EGR technology, as far as the increase of HC emissions has to remain compatible with the HC conversion capacities of the DOC. Actually, the NOx /HC compromise thus becomes the most constraining factor in the limitation of the NOx reduction potential. Currently, car manufacturers spread the water-cooler technology for EGR equipped with a bypass to limit HC and CO increase during the cold phases. Increasing the EGR densification (at a colder temperature) while staying at an iso O2 rate – iso richness rate can be addressed by several means such as: • increasing the heat exchanger area of the water-cooler, • improving the water-cooler technology (exchanger type), • setting a cooling system for the water circulating in the cooler in a way to minimize the water temperature. If the evolutions in combustion/injection and the EGR system modifications are not sufficient, it can be helpful to put a NOx after-treatment system in the exhaust line

216

Past and Present in DeNOx Catalysis

such as the NOx Trap or the SCR-NH3 system. Without any surprise, these applications correspond to the heaviest vehicles with the highest inertias. In this context, the automated transmissions and/or the four driving wheels versions give the more complex cases. Just a few words to explain that in EuroV is that the clean-up strategy adopted by heavy-duty vehicles differs from that chosen by passenger cars. The heavy-duty vehicles have decided to reduce the engine-out particles emission levels and to treat the NOx released in the exhaust line by adopting the SCR-NH3 system.

1.2.2. EuroVI and its consequences EuroVI standards are still unknown, but if we want to be able to anticipate the changes for the engine or for the after-treatment technologies, every car manufacturer has to imagine and to build his legislative scenario internally. The aim of this paragraph is to examine the risks and the consequences induced by the EuroVI standards application. If EuroV finally leads to a moderate reduction of the NOx emissions limit (≈20%), when regarding to US Tier 2–bin 5 levels3 , we can presume that EuroVI will consist of much more stringent NOx limitation4 . A NOx after-treatment system in combination with a DPF (which is mandatory since EuroV) in the exhaust line should become ordinary with EuroVI. The PM limit values will be unchanged from EuroV. The road transport seems to be the main source of NOx emissions (see Figure 7.4).

NOx (kt/an) Agriculture

12.4

Waste treatments

7.4

Combustion (industrial, kitchen, tertiary)

32.5

Electricity production

9.8

Other mobile sources

7.8 6.9

Road traffic

Air traffic Heavy-duty vehicles >3.5 T

31.3

Light commercial vehicles

13.9

Passenger cars

38.3

Total : 160.3 kt/an

Year 2000

Figure 7.4. NOx emissions estimation around Paris for year 2000 (source: ‘AIRPARIF Actualité’ number 24).

3 4

Tier 2–bin 5’s NOx limit is fixed at 0.07 g/km key scenario = 80 mg/km – 5 years after EuroV application

Current Tasks and Challenges for Exhaust After-Treatment Research

217

In Europe, the air quality objective for NO2 in 2010 is 40 g/m3 (annual mean value): every European state will thus be subjected to penalties if it does not comply with this limitation of 40 g/m3 . Today, some institutes and organizations in charge of air quality analysis raise an alert, because the NO2 level over the last years has remained quite unchanged while the NOx level has significantly decreased. The spotlight is put on the Diesel vehicles because their DOCs in the exhaust line produce NO2 by NO oxidation (nevertheless NO remains the major species emitted by the engine). Consequently, considering the future air quality constraints, a new limit value in NO2 emissions may be introduced for Diesel passenger cars. Another potential risk is the introduction of a limit value in the soot particles number (for example the limit could be fixed at 1011 particles/km). This new limitation could have a consequence on the DPF filtering medium. In such a context, the DPF technology with macro porous ‘open’ medium will not be relevant any more. Finally, in EuroVI context there is a risk of ‘off cycle emissions test’, meaning that we would have to consider engine-operating points that are not reckoned by the NEDC5 normative cycle. This ‘off-cycle’ risk must be integrated in the development and the study of after-treatment solutions in the EuroVI context. To anticipate the ‘off cycle emission test’, the possible working hypotheses are numerous, but the most well-known hypothesis is inspired by US legislation with the consideration of the US aggressive cycle (US06 cycle) or the consideration of ‘Not To Exceed’ thresholds as for the heavy-duty vehicles. In this short section, let us make a quick presentation of the US06 cycle (see Figure 7.5), which is a normative cycle dedicated to vehicles sold in US (more often vehicle with high cylinder capacity comparatively to European vehicles). The US06 is much more dynamic and spasmodic cycle than the NEDC. The sudden variations of engine operating points during the US06 cycle make it an aggressive cycle and maximize the NOx engine-out emissions (the NOx emissions being strongly increased during the engine accelerations and with high thermal conditions).

Exhaust gas flow rate in function of time Exhaust gas flow rate (m3/h)

Direct injection Diesel - 2,2L Vehicle speed in function of time

Vehicle speed (km/h)

140 NEDC US06 cycle US06 cycle

120 100 80 60 40 20 0 0

200

400

600

800

1000

Time (s)

1200

300 250 NEDC US06 cycle

200 150 100 50 0 0

200

400

600

800

1000

1200

Time (s)

Figure 7.5. Vehicle speed and exhaust gas flow rate evolutions for the NEDC and US06 cycle for a 2.2 l Diesel direct injection engine.

5

NEDC: Normative European Driving Cycle

218

Past and Present in DeNOx Catalysis

2. DIESEL PARTICLE FILTER MANDATORY IN EUROV – ITS CONSTRAINTS 2.1. DPF physical constraints and their consequences 2.1.1. Soot loading The first function of the DPF is to trap the soot particles. The DPF, in its current technical definition, is a porous material allowing the flowing of gases but trapping, with an efficiency close to 100%, the soot particles contained in engine emissions. The exhaust gases must circulate through the porous medium because the input channels are closed at their end with a plug (see Figure 7.6). Consequently, the gases pass into the adjacent channels (output channel) which are open at their end (and closed at their entry). The main result of the particles trapping is a fooling effect for the porous medium that leads to a reduction of the material permeability. As shown on Figure 7.7, this fooling

Output channel

plug

Input channel

Figure 7.6. DPF operating mode.

Soot loading Filter Impact of soot loading on exhaust back pressure/performance

Engine

Exhaust gas

Back pressure in CSF (mbar)

1200 1000 800 10g/l 600 400 0g/l 200 0 100

200

300

400

500

600

Gas flow rate m3/h

Figure 7.7. Impact of the soot mass trapped in the DPF on the exhaust back pressure.

700

800

Current Tasks and Challenges for Exhaust After-Treatment Research

219

effect causes an increase in the exhaust back pressure as the soot mass trapped in the DPF increases. Before the driver can notice any loss of power for his engine, the DPF requires a regeneration phase to burn the accumulated soot.

2.1.2. DPF regeneration 2.1.2.1. Regeneration conditions As illustrated on Figure 7.8, the DPF regeneration step is just a soot particles combustion reaction (soot is mainly composed by a carbon matrix), which requires a temperature in the range of 600 C and oxygen presence in the exhaust gases. The sequences of soot particles trapping and DPF regeneration alternate successively. The regeneration frequency is in the range of several hundreds of kilometers and the regeneration duration is in the range of several minutes (depending on the operating conditions, on the soot quantity to burn, on the DPF material, etc.). 2.1.2.2. Impact on DPF reliability The combustion of soot leads to an important exothermicity in the DPF during regeneration. This exothermic effect (proportional to the soot loading in the DPF) can lead to an irreversible degradation of the filter material, if the soot mass trapped in the DPF is higher than a soot mass limit (depending on the porous material properties). Consequently, it is necessary to protect the DPF by specific engine control strategies. Nevertheless, some car manufacturers have decided to install different alert levels such as warning lights on vehicle dashboard to alert the driver and protect the after-treatment system integrity. 2.1.2.3. Impact on engine reliability 600 C is the minimal temperature necessary for DPF regeneration in the case of non additivated soot. In nominal engine operating conditions, the temperature at the DPF inlet remains widely lower than 600 C. In the case of an additivated soot (additives added directly within the fuel), the temperature to be reached to burn the soot is decreased by about 50 C to 100 C: even in that case, the temperature target is always difficult to reach in nominal engine operating conditions.

Soot combustion

Soot loading Filtering material

DPF CO2 + H 2O

regeneration Engine emissions

Exhaust gas w/o PM

Engine emissions

Necessary conditions: T > 600°C and oxygen

Figure 7.8. DPF operating mode with alternation of soot particles trapping sequence and DPF regeneration step.

220

Past and Present in DeNOx Catalysis

Car manufacturers have been using specific engine control strategies (associated to convenient rolling conditions), which allow increasing the exhaust temperatures as well as the HC and CO fractions sent to the DOC located in front of the DPF. These HC and CO fractions, when converted by the oxidation catalyst, cause an increase in the exhaust gases temperature, due to the exothermicity of the oxidation reactions. This additional increase in temperature can be used to complete the heat directly produced by the engine. The DOC presence is useful for two main reasons: to convert the HC and CO engine-out emissions in compliance with the Diesel Euro standards and to assist the DPF regeneration step. The DPF regenerations done by specific engine control strategies lead to a fuel overconsumption. Besides, these specific strategies of DPF regeneration can generate a dilution of post-injected fuel in engine lubricant (through the lubricant film or directly through the piston rings during the washing of the combustion chamber walls by the fuel injected in large quantities). Consequently, this phenomenon is reducing the engine lubricating properties, which influences potentially the engine reliability. This phenomenon is even more marked when the engine operating points correspond to low regime and low effective mean pressure (i.e., to urban rolling). Thus, the way to optimize the frequency of DPF regeneration is to examine the dilution of fuel in engine lubricant, while keeping under control the DPF reliability criteria. In conclusion, the DPF management is relatively complex and needs a good management of the compromise that is to secure the after-treatment system performance and the engine reliability: Regenerate sufficiently to protect the DPF system and not too often to limit the engine lubricant dilution.

2.1.3. DPF ageing The main cause of DPF ageing (and thus of its functional characteristics) is the accumulation of lubricant residues, which is directly proportional to the engine lubricant consumption and to the lubricant composition (ashes rate). These residues are present in the exhaust gases within the soot particles and are irreversibly trapped in the DPF. They mainly consist of calcium sulfate (with other components containing zinc and iron). For the DPF technology using the fuel additivation, additives also participate in the residues formation. The residues deposit causes a reduction in the material permeability during the DPF life. This phenomenon leads to an increase in the exhaust back pressure, and is noticeably more marked when the soot quantity trapped in the DPF increases (see Figure 7.9). The bigger the DPF volume, the larger the available surface, the greater the amount of residues and soot trapped by the filter and the lower the exhaust back pressure. In summary, the DPF ageing has two important consequences: • an increase in the regeneration frequency, • a decrease in the engine performances during the DPF life.

Current Tasks and Challenges for Exhaust After-Treatment Research

221

1000

Pressure lost in DPF (mbar)

900 800

Soot 10 g/l 100 g Ash

700 600

Soot 10 g/l 0 g Ash

500 400

Soot 0 g/l 100 g Ash

300 200

Soot 0 g/l 0 g Ash

100 0 100

200

300

400

500

600

700

800

Exhaust gas flow rate m3/h

Figure 7.9. Residues (accumulated in the DPF) influence on the exhaust back pressure.

2.2. Driver rolling profiles impact We have just mentioned that one of the negative impacts of the DPF regeneration is the dilution of fuel in engine lubricant. This phenomenon depends on the engine operating conditions and speeds up during the regeneration sequence. Indeed, a particle filter regeneration generates much more dilution effect on an urban rolling than on a highway rolling (with a ratio of about 5). The difficulty to manage on the one side the DPF performance and reliability and on the other side the engine reliability will be very dependent on the drivers rolling profiles. The regeneration step must be punctual and should not be performed during urban rolling. Thus, the understanding and the characterization of the customer behavior is essential and particularly his/her rolling sequence: that is to say the alternation of unfavorable rolling in urban mode and of favorable rolling in extra urban mode. Let us come back to the rolling conditions in extra urban mode and its consequences. They are favorable to the use of DPF for the following main reasons: • The lubricant dilution: this phenomenon is less important in highway rolling conditions. • The oxido-reduction reaction that can occur between the soot and the NO2 contained in exhaust gases (usually called passive or continuous regeneration). This reaction is favored in the highway rolling conditions because the NO2 production is maximized and because the temperature conditions are also favorable (let us remind that the DPF is always located behind a DOC volume that oxidizes NO into NO2 ). This phenomenon plays a role on the DPF loading autonomy and on the DPF regeneration frequency. In highway rolling conditions, the DPF autonomy is globally increased.

222

Past and Present in DeNOx Catalysis

The rolling conditions in urban mode have two negative effects: • The lubricant dilution phenomenon is maximized by these low speed conditions. • The particle emissions are maximized due to a lower NOx production in the exhaust gases and a lower temperature range (limiting the passive regeneration effect). Therefore, the DPF loading autonomy is globally reduced.

2.3. DPF design process The DPF presence in the exhaust line has two consequences on the engine: • An increase in the exhaust back pressure compared to a technical definition without DPF. • A variation in the exhaust back pressure during the DPF and the vehicle life (because the DPF ageing leads to residues accumulation), which has an impact on the maximal engine performances. The back pressure fluctuation in the DPF can be explained physically by greater difficulty for exhaust gases to cross the DPF according to the accumulation of soot and residues (cf. paragraph 2.1.3). The determination of the optimal DPF volume has to take into account the two important phenomena previously presented, namely: • the dilution of fuel in engine lubricant, associated to the engine reliability, • the durability of the after-treatment system associated with the decrease of engine performances.

3. NOx AFTER-TREATMENT In the EuroV context, the NOx emissions limit, proposed by the European Commission at 200 mg/km, will impose for some ‘heavy’ applications the implementation of a NOx after-treatment system in the exhaust line. Nowadays, there are two technical ways identified by car manufacturers for NOx conversion: • the NOx Trap, • the selective catalytic reduction by ammonia or ‘SCR-NH3 ’. Do not forget that there will be a DPF in the exhaust line dedicated to the soot particles treatment and that the NOx after-treatment system thus adds to the DPF.

3.1. The NOx Trap 3.1.1. The NOx Trap operating mode The NOx Trap operating mode is made of two phases: • NOx storage over alkaline earth sites during lean periods which we will call the storage phase. This phase occurs during the nominal engine operation (in

Current Tasks and Challenges for Exhaust After-Treatment Research

223

lean conditions). Contrary to the DPF, this phase does not present an efficiency of 100%. • NOx release and reduction assisted by noble metal. This phase, which we will call the regeneration phase, is necessary to clean-up the catalyst surface before its saturation (the NOx Trap material has a limited number of adsorption sites). During lean periods, the system ensures two functions: • oxidation of NO, HC and CO species assisted by noble metal, • NOx adsorption on the catalyst surface. During the NOx Trap regeneration phase, the engine operates with specific parameters to generate rich conditions (richness >1) in the exhaust line. These engine operating conditions allow the generation of compounds in the exhaust line, which are able to reduce the NOx accumulated during the storage phase on the NOx Trap surface. From a macroscopic point of view, the quantity of reducing elements and the regeneration duration must be adjusted according to the NOx quantity to convert. The simplified diagram shown in Figure 7.10 summarizes the regeneration principle in rich mixture. In fact, it is a bit more complex: because, depending on the material characteristics, the NOx Trap can present oxygen storage ability at the trap surface. During the regeneration phases, the reductants produced by the engine are consumed by the reduction of the stored oxygen as well as by the reduction of the adsorbed NOx . A part of the reductants does not take part effectively in the NOx reduction, making the estimation of total amount of reductants required to regenerate the NOx Trap completely more complicated to assess. The vehicle can cover several kilometers during the storage phase whereas the regeneration phase only lasts for a few seconds. As shown in Figure 7.11, the difference between NOx engine-out emissions and the NOx standards limit value corresponds to the minimal quantity of NOx to be treated. This characteristic constitutes the key factor for the design of the NOx Trap volume.

H2O, CO2 N2

NO

O2

HC CO H2 NO2

NO3– NO2

Pt

Rh Support

Ba

Step 1: NOx Storage (in lean conditions) Step 2: NOxTrap Regeneration (in rich conditions)

Figure 7.10. Schematic diagram of the storage and the regeneration phases of the NOx Trap.

224

Past and Present in DeNOx Catalysis

NEDC

NOxTrap Stored NOxx StoredNO

NOx NOx emission emission (g/km) (g/km)

+

Released NOx (g/km)

Storage process

NOx legislation (g/km)

Mass of NOx stored on the NOxTrap ≥ NOx emission – NOx legislation

Key factorfor factor fortheNOxTrap the NOxTrap is the total amount isthetotal amount of NOxto NOx tobetreated be treated

Figure 7.11. Storage phase of the NOx Trap.

3.1.2. NOx Trap difficulties 3.1.2.1. Fuel over-consumption generated by the rich pulses To regenerate the NOx Trap, the engine must create rich conditions in the exhaust line and the car manufacturers must deal with specific engine control strategies to increase the reductant quantity (H2 , CO and HC). These strategies lead to fuel over-consumption. 3.1.2.2. HC penalties and more specifically CH4 penalties At the beginning of NOx Trap regeneration, the engine generates massive quantities of reductants; thanks to specific engine control strategies. These reductants are constituted by compounds which react with the adsorbed NOx (CO, H2 , HC); but a certain fraction of reductant, formed by methane, goes through the NOx Trap without any conversion. One of the fundamental difficulties of the NOx Trap regeneration process is thus connected to the engine-out methane emissions. In summary, the NOx mass to convert (difference between engine-out NOx emissions and NOx emissions limit imposed by Diesel Euro standards) governs the requested duration for the rich conditions in the exhaust line (NOx regeneration duration), and thus dictates directly the quantity of methane released in the exhaust line. The CH4 emissions can reach significant values and even lead directly to a cleanup unfeasibility in certain cases, since these amounts can be close to the unburnt HC threshold imposed by Diesel Euro standards. However, even if the CH4 values do not reach the unburnt HC limit, these emissions are added in the vehicle exhaust gas and limit the global unburnt HC that the engine can release in lean conditions. Thus, with the NOx Trap, a new constraint is appearing, which will have an impact on the engineout HC emissions during the lean conditions. As NOx emissions levels increase, CH4 emissions in rich mode increase, and it is then necessary to reduce HC emissions in lean mode to stay within the limits of the DOC conversion ability. This context is specific to

Current Tasks and Challenges for Exhaust After-Treatment Research

225

the European legislation. The US legislation for example, having a standard expressed in NMHC6 , would not be subject to this problem. In summary, from a clean-up point of view, the main hard point for NOx Trap concerns a process of richness >1, which minimizes methane emissions. To solve this problem, some car manufacturers associate to the NOx Trap system an additional injector in the exhaust line, to draw-off the necessary reductants directly from the fuel tank and prevent from CH4 release in the exhaust line. 3.1.2.3. Sulfur sensitivity The NOx Trap is a system particularly sensitive to sulfur. Sulfur and more specifically SO2 takes the place of NO2 on the material surface causing a decrease in the number of available sites dedicated to NOx adsorption. This phenomenon is partially reversible and thus some strategies will be implemented for NOx Trap desulfation to try to recover the initial adsorption capacity of the trap. But a part of sulfur poisoning is irreversible, and whatever strategy is tested during the NOx Trap lifetime, we are going to undergo an inevitable and progressive degradation process of the adsorption capacity of this system during its life span. Sulfur is contained in the fuel and in the engine lubricant. Unfortunately, even if on the one hand, a low sulfur content fuel (below 10 ppm) can be supplied by oil companies, on the other hand, the sulfur concentration in the lubricant (mandatory for its lubricating property) is large enough to lead to a noticeable degradation of the system. It is hence necessary to develop specific engine control strategies that will take into account the degradation of the system performance due to ageing and sulfation. These strategies will have several consequences such as (see Figure 7.12):

NOxTrap regeneration Short spike

Leads to HC penalty

… Repeated milage

Filter regeneration

…

Leads to oil dilution

deSOx event

…

Leads to NOx and HC penalties Leads to oil dilution

Figure 7.12. NOx Trap and DPF implementation in the exhaust line and its consequences.

6

NMHC = Non Methane Hydrocarbons

226

Past and Present in DeNOx Catalysis

• NOx and HC penalties, • fuel dilution in the engine lubricant, • impact on the CO2 emissions. Fortunately, desulfation sequences are not frequent and are spaced out by several thousands of kilometers. Desulfation process requires, at the same moment: • richness greater than 1 (as the NOx Trap regeneration does) • high temperature that the engine in nominal operating conditions cannot provide. This process also lasts several minutes, which makes it globally a very penalizing step and relatively complex to implement from an engine control/tuning point of view. 3.1.2.4. Precious metal cost impact The NOx Trap can be assimilated to a three-way catalyst7 complemented by a NO2 storage function. This system contains two types of precious metals: the first one, dedicated to NO, HC and CO oxidation in lean operating mode, will generally consist of platinum or palladium. The second one, dedicated to NOx reduction in rich operating mode, is typically rhodium. Rhodium price spectacularly increases, with a multiplication by three of its value in 2006 (see Figure 7.13)8 .

1100 1400 1300 1000 Palladium, Platine, 1200 900 (Fixing Londres, $IOz) 1100 800 (Fixing Londres, $IOz) 1000 700 900 600 800 500 700 400 600 300 500 200 400 100 300 200 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

USD per troy ounce

6405.00 5405.00

Rhodium ($IOz)

4405.00 3405.00 2405.00 1405.00

19Sep05 15Dec05 16Mar06 14Jun06

11May05

07Apr04 06Jul04 01Oct04 31Dec04

24Nov03

21Oct02 21Jan03 17Apr03 17Jul03

13Jun02

29Mar01

06Aug01 31Oct01 01Feb02

405.00

Figure 7.13. Pt, Pd and Rh price evolution over last 20 years.

7

A three-way catalyst is used normally in gasoline engine context to convert HC, CO and NOx engine-out emissions 8 To fulfill EuroIV standards, the global precious metal content included in the after-treatment is around 4 g in gasoline (three-way catalyst) and around 10 g in Diesel (DOC + DPF)

Current Tasks and Challenges for Exhaust After-Treatment Research

227

This implies that the NOx Trap turns to be a relatively expensive NOx after-treatment system.

3.1.3. Potential of the NOx Trap technology We previously spoke about the risk of ‘off cycle emissions test’ and the obligation to anticipate the ‘off cycle’ for EuroVI. What is the potential of the NOx Trap technology in this context? The important NOx quantities released during highway rolling are going to increase the frequency of NOx Trap regeneration phases. CH4 penalties (depending on the regeneration strategy used) and fuel over-consumption will inevitably increase as well. Consequently, the balance for the NOx Trap use is just average if anything today. We have seen in the previous paragraphs that in EuroIV, it is necessary to manage a NOx /particle compromise. This compromise has been replaced in EuroV by a NOx /HC compromise. The ‘off-cycle’ risk reveals another compromise to be managed: the NOx /CO2 compromise. Overall, the NOx Trap added to the DPF in the exhaust line gives a complex and expensive after-treatment system that has an impact on the engine operation and leads to a fuel over-consumption.

3.2. The selective catalytic reduction with ammonia The second DeNOx technology, the selective catalytic reduction with ammonia (SCR-NH3 ) commercially available in heavy-duty vehicles since 2006, seems to present an interesting potential in terms of efficiency, reliability, HC penalties, etc.

3.2.1. The SCR-NH3 technology for the NOx reduction The SCR-NH3 system consists in reducing engine-out NOx emissions with NH3 as key-reductant. The reactions are as follows: 4NO + 4NH3 + O2 → 4N2 + 6H2 O ‘standard SCR reaction’

(1)

NO + NO2 + 2NH3 → 2N2 + 3H2 O ‘fast SCR reaction’

(2)

6NO2 + 8NH3 → 7N2 + 12H2 O ‘NO2 SCR reaction’

(3)

The study of the reactions kinetics has shown that reaction (2) is the fastest in the temperature range of 170–350 C. For temperatures higher than 350 C, the rates of reactions (1) and (2) are of the same magnitude. However, for a temperature range of 200–500 C, reaction (3) is always the slowest one. Consequently, to optimize NOx conversion at low temperature, NO2 is requested in a ratio NO/NO2 = 1.

228

Past and Present in DeNOx Catalysis

+

ADBLUE

NH3

CO2

Figure 7.14. Adblue (urea aqueous solution of 32.5% wt) hydrolysis yielding to NH3 .

To have a mixture NO + NO2 at the SCR catalyst intake, the presence of a DOC upstream of the system is required (oxidation of NO into NO2 ). The DOC must be sized according to the optimal ratio NO/NO2 = 1. For safety reasons, NH3 can be provided in the vehicle on two precursor forms only, which are: 1. Urea (liquid precursor): its hydrolysis leads to NH3 . This precursor is the most commonly used by car manufacturers. Liquid urea is stored in a tank and distributed upstream of the SCR system; thanks to a proportioning unit. Industrial liquid urea is known as ‘Adblue’. This industrial liquid [chemical formula (NH2 )2CO] is an aqueous solution of 32.5% wt urea solution (see Figure 7.14). Adblue is corrosive and requires stable materials for the components like tank, pipes, injector, etc. Moreover, this solution is not suitable for year-round use in the Northern countries: the freezing point is at −11 C. Urea hydrolysis requires at least 180 C. Another characteristic of this solution is its proneness to crystallization and polymerization. When parts of the exhaust system are constantly welted by Adblue on the same spot, undesired urea crystals or polymers may form if the exhaust line temperature is lower than 300 C. This phenomenon will result in uncontrolled ammonia production when the crystals or polymers melt or sublimate after being heated at significantly higher temperatures (T > 350 C). This may result in ammonia release. Furthermore, the crystals or polymers can also have an impact on the SCR catalyst cells by reducing the catalyst surface and thus reducing the catalyst performances. 2. Ammonium carbamate (solid precursor): leads to NH3 by decomposition. Solid ammonium carbamate is stored in a compartment above the proportioning unit. The sublimation of solid blocks of ammonium carbamate is achieved by a local heater (spray or electrically warmed oil circulation). The sublimation temperature of ammonium carbamate is 60 C. The process being reversible, solid ammonium carbamate may be formed back for temperature lower than 60 C. Solid precursor is interesting for storage: the tank volume required for a given autonomy will indeed be lower than the tank volume needed to store liquid urea. By now, no car manufacturers have chosen this way. The main problem seems to be the energy requested to reach the sublimation point, and the system response time.

Current Tasks and Challenges for Exhaust After-Treatment Research

229

3.2.2. SCR-NH3 difficulties The ammonia necessary for the reaction is the main hindrance to the SCR-NH3 process because pure ammonia is an irritating and toxic gas which cannot be released in the exhaust line. Particular care must be taken to ensure that the maximal NH3 content released in the exhaust does not exceed the threshold of 10 ppm. NH3 release in the exhaust line can be prevented by keeping the overall urea/NOx ratio significantly below stoichiometry or by installing an NH3 clean-up catalyst behind the SCR catalyst. 3.2.2.1. SCR-NH3 architecture The constraints linked to the SCR-NH3 system architecture are very troublesome. The whole system must be located under the vehicle floor, in the exhaust line, since there are no other rooms for this large system to fit. When speaking about SCR system, one should keep in mind that the SCR catalyst is not the single component and that several elements are necessary for its correct operation. The complete system includes (see Figure 7.15):

ADBLUE tank DCU

Volume depends on: - Clean-up standards and associated engine-out NOx emissions levels - Strategies of refueling tank

External ECU: Management of the urea quantity to inject - Receive information from SCR catalyst - Control injector and pump to deliver required Adblue quantity ADBLUE tank

P sensor T sensor Pump DCU

Adblue Injector T sensor Mixing device

SCR

NOx sensor

ADBLUE injector Injection optimization: - Minimize wall wetting → Polymers formation

Urea hydrolysis NH3 formation:

- Urea → NH3 Needs a temperature above 180°C - Mixing device integration → Homogeneous composition at the SCR catalyst inlet

SCR catalyst Kinetics: - deNOx: NOx + NH3 → H2O + N2 - NH3 storage - HC storage

Figure 7.15. Exhaust line with a SCR-NH3 system located under the vehicle floor.

230

Past and Present in DeNOx Catalysis

• • • •

a unit control to monitor the quantity of injected urea, a tank (additional to the fuel tank) to store the Adblue solution, an urea injector that must atomize the Adblue, a mixing device to ensure an homogeneous distribution of the NH3 in the exhaust gas at the SCR catalyst inlet, improving the urea conversion into NH3 , an hydrolysis catalyst or an additional exhaust line length sized to maximize the residence time, the SCR catalyst for NOx conversion, sometimes (depending on the SCR material used), a last oxidation clean-up catalyst to prevent NH3 release in the exhaust line (limit value fixed at 10 ppm) a NOx sensor to control the NH3 quantities injected and to allow diagnosis (on board diagnosis/OBD function) of the SCR-NH3 system (for the heavy-duty vehicles, the NOx sensor will be mandatory since October 2007).

• • • •

The DOC upstream the SCR-NH3 system is necessary to maximize the NO oxidation into NO2 and to optimize the SCR system operation at low temperature. 3.2.2.2. Urea tank volume and urea availability The urea distribution network in Europe, around year 2006, would be limited to one distribution point for a 500 km radius area (heavy-duty vehicles compatibility – data to be checked and actualized), and would not allow a ‘co-fueling’ strategy whose interests are the simultaneous fuel/urea filling up at the service-station and the minimization of the urea tank volume. Today, it is difficult to anticipate the consequences of EuroV. The car manufacturers should focus on a strategy based on a greater autonomy (target = lubricant draining interval). The evaluation of the urea tank volume depends on the European standards contents. To reach an autonomy corresponding to the lubricant draining interval, the required volume for the urea tank is in the range of 20–30 l. This obviously raises a problem regarding an on-board equipment, since the available space in the vehicle is very limited. Filling of the urea tank by the driver must be checked for the quality and the quantity of solution put in the tank. Consequently, some specific monitoring strategies must be developed. 3.2.2.3. Activation of the SCR catalyst and the urea hydrolysis catalyst The optimal temperature range for SCR is 250–400 C. The SCR and urea hydrolysis catalysts are belatedly activated in the NEDC urban part of the cycle (let us remind that the system is located under vehicle floor for architecture reasons). To improve the SCR system efficiency, specific engine control strategies must be implemented to control the temperature and more especially to ensure a quick heating of the catalysts at the beginning of the NEDC. Of course, this leads to an energy cost and generates fuel over-consumption.

3.3. Comparison of the NOx Trap and the SCR-NH3 technologies The comparison of the NOx Trap and SCR-NH3 systems can be summarized as follows (Table 7.1): The SCR technology fits at the best the ‘off-cycle’ risk in EuroVI context, because this technology does not release CH4 and its continuous operation makes the treatment

Current Tasks and Challenges for Exhaust After-Treatment Research

231

Table 7.1. Comparison table for the NOx Trap and the SCR-NH3 technologies

NOxTrap NOx / HC compromise in EuroV

NEDC

Reliability

Thermal ageing Sulfur poisoning

Impact on requirements

Fuel penalty CH4 penalty Oil dilution Performance (exhaust back pressure)

Integration

Catalysts volume Urea tank

Distribution

10 ppm Sulfur content in fuel/urea distribution

EuroVI perspective

NEDC ‘Off cycle’

SCR-NH3

of large amounts of NOx easier. However, there is one condition: the catalysts must be activated at temperatures higher than 250 C. With the NOx Trap technology, it is necessary to limit the regeneration frequency and to control the NOx /CO2 compromise. Both technologies being catalytic are naturally sensitive to the thermal ageing of the material. But the NOx Trap is particularly sensitive to sulfur. Both technologies generate a quite equivalent fuel over-consumption (see Figures 7.16 and 7.17) but the NOx Trap needs some specific engine control strategies during the regeneration or desulfation phases which can cause dilution of fuel in engine lubricant. Both Figures 7.16 and 7.17 illustrate the difference in operating temperature range for both DeNOx systems and show their efficiency difference. NOx Trap presents a narrow Function of NOx Storage Catalyst Golf class vehicle

NOx conversion approx. 50% NEDC approx. 40% FTP approx. 60% US06 Fuel consumption +2 to 3% regeneration

NOx conversion [%]

100 80 60 40 20 0 100

fresh 200

300

aged 400

500

Temperature [°C]

Figure 7.16. NOx Trap efficiency (fresh and aged materials) on NEDC and US cycles.

232

Past and Present in DeNOx Catalysis Function of SCR Catalyst

Golf class vehicle

NOx conversion approx. 40–50% approx. 50% approx. 70% Fuel consumption approx. +2 to 3%

NEDC FTP US06

NOx conversion [%]

100 80 60 40 20 0 100

200

300

400

500

Temperature [°C]

Urea consumption approx. 0.5 – 3.0% of fuel consumption

Figure 7.17. SCR-NH3 efficiency on NEDC and US cycles.

window between 250 and 400 C and seems to have a maximal conversion rate of 60%. The SCR-NH3 system presents a wide operating range above 250 C with a conversion rate higher than 90%.

4. CONCLUSION The severity of the Diesel European standards for HC, CO, NOx and PM has a first consequence: car manufacturers seem to reach the limits of combustion/injection improvement. Some particle and NOx after-treatment systems will become mandatory in EuroV and EuroVI context. Consequently, this implies a cost and generates a complex implementation. The NOx and particle after-treatment systems have today an impact on the engine operation and have an impact on the CO2 emissions because they lead to fuel over-consumption. The over-consumption generated by these after-treatment systems (DPF and DeNOx system associated with the exhaust line) can simply nullify all the efforts made by the ACEA members over the past few years, regarding CO2 emissions (notice that the European association of the car manufacturers has given its commitment for a mean value of 140 g/km of CO2 for 2008). The introduction of these expensive after-treatment technologies is going to increase the price difference existing today between the Diesel and gasoline versions of a vehicle. Rinaldo Rinolfi, Director of the engine division at the Fiat Research Center, predicted that Diesel could be up to 50% more costly to produce than gasoline engines with similar performance, due to their need for NOx and particle abatement and improved engine induction systems and controls9 . By 2010, in EuroVI context with more stringent NOx limits and a possible outbreak of normative restrictions such as the ‘off-cycle’ risk, it will be necessary to make a technico-economic comparison between the SCR-NH3 and the NOx Trap technologies.

9

‘School of hard NOx ’ by Lindsay Brooke – in Automotive Engineering International of February 2006

Current Tasks and Challenges for Exhaust After-Treatment Research

233

Each technology presents key points and weaknesses, but they both constitute complex and expensive after-treatment systems: one has an impact on the engine operation; other one has an impact on the vehicle architecture. For their development, the efforts must focus on the tuning and the engine control frame. It seems that these two technological solutions for NOx reduction have their limits and it is necessary either to find some systems to assist them and decrease their impact on the customer requirements or to find some brand new innovative DeNOx after-treatment systems. Otherwise, the key for the future could be integrating multiple functions in a single device10 . The trick is to meet EuroVI at affordable cost, while retaining the fuel economy and drivability attributes that are keys to Diesel’s acceptance in the marketplace. Now the question is what is the best way to meet it? A variety of options must be considered: cost, robustness, complexity and application constraints. As 2007 begins, car manufacturers and suppliers are scrambling to make sure that, whatever they commit to in production, it will not be more costly than what the other companies devise. Over the past few years and for a long time, Diesel clean-up has been a very competitive and strategic field.

10

Toyota has commercialized the Diesel Particle NOx Reduction or four-way system (DPNR), which integrates NOx Trap and DPF functions in one block and simultaneously reduces NOx and PM. The physical constraint of such a system is to maintain a significant NOx adsorption performance while keeping under control the increase of back pressure due to the system.

This page intentionally left blank

Chapter 8

THE ROLE OF CERIUM-BASED OXIDES USED AS OXYGEN STORAGE MATERIALS IN DeNOx CATALYSIS X. Courtois, N. Bion, P. Marécot and D. Duprez∗ LACCO, Laboratoire de Catalyse en Chimie Organique, CNRS & University of Poitiers, 40 Av. Recteur Pineau 86022 Poitiers Cedex, France ∗ Corresponding author: LACCO, Laboratoire de Catalyse en Chimie Organique, CNRS & University of Poitiers, 40 Av. Recteur Pineau 86022 Poitiers Cedex, France. Tel.: +33 5 49 45 39 98, Fax.: +33 5 49 45 34 99, E-mail: [email protected]

Abstract Materials with high ‘oxygen storage capacity’ (OSC) are now widely used in automotive converter catalysts. They are mainly composed of Ce-based oxides (CeZrOx, CeZrPrOx, etc.) having both multiple cationic valencies and oxygen vacancies. These properties allow the catalyst to store active O species (O2− , superoxide, etc.) in O2 excess and to release them when the O2 concentration in gas phase decreases or becomes nil. After having briefly examined the main properties of these OSC materials and the methods employed for their characterization, their impact in automotive catalysis will be reviewed, with a special insight in DeNOx catalysis: (1) in three-way catalysis (2) in automotive catalysis under lean conditions (lean-burn spark ignition engine and diesel).

1. INTRODUCTION Use of oxygen storage components in three-way catalysis (TWC) was proposed by Ghandi et al. in 1976 [1] and implemented in the real exhaust catalysts at the beginning of the 1980s. Since the pioneer’s work of Yao and Yu Yao [2] and of Su et al. [3,4], numerous studies were devoted to a better knowledge of oxygen storage capacity (OSC) properties of ceria-based compounds [5,6] and, specially, to Cex Zr1−x O2 mixed oxides [7–10]. OSC compounds are known to enlarge the ‘operating window’ of TWC, i.e. the  range within which CO and HC oxidation as well as NOx reduction can occur at the optimal rate (higher than 90% conversion [11]). Although it can be inferred that OSC materials exhibit the highest impact under transient conditions, numerous investigations were carried out at the stationary state. However, kinetic studies carried out at various concentrations may give an idea of what can occur in transient regime. Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Published by Elsevier B.V.

236

Past and Present in DeNOx Catalysis

2. OSC MEASUREMENTS AND OXYGEN MOBILITY 2.1. OSC measurements Oxygen storage capacity is generally measured in a pulse chromatographic reactor using CO [12–16] or, to a lesser extent H2 [17] as reducers. The procedure is depicted in Figure 8.1. Pulses of CO (or H2 ) are injected over the pre-oxidized sample. According to the nomenclature of Yao and Yu Yao, the CO2 produced upon the first CO pulse allows to calculate the OSC while the total amount of CO2 formed upon several CO pulses (typically 10 pulses) gives the ‘oxygen storage capacity, complete’ (OSCC). In most cases, only OSC is retained for characterizing TW catalysts [18]. Figure 8.1 shows the typical behavior of a ceria-based sample. As a rule: (1) The reduction phase (phase 1) is slower than the re-oxidation one (phase 2). The CO2 formation decreases regularly upon each CO pulse while the re-oxidation is achieved upon the first pulse of O2 . This is a rather general phenomenon in catalysis. Oxides (like rare-earth oxides) reduced more slowly than their suboxides may be re-oxidized. It is interesting to note that the reverse phenomenon can be observed with the metals (Pt, Rh and Pd). Their oxides are reduced at a much lower temperature than the metal can be re-oxidized [19–21] even though the nature of support and the metal particle size may change the redox properties significantly [20,22,23]. (2) Depending on the temperature, there may be a carbon deficit in the mass balance upon individual pulses, i.e. CO consumption may be higher than CO2 formation. A certain amount of carbon may be stored in the catalyst and released upon the following CO pulses or upon the first pulses of O2 . In this case, some CO2 appears in phase 2. A detailed investigation of C and O mass balance during OSC measurements has been made by Holmgren et al. [24]. (3) In most cases, there is a good agreement between the OSC measured upon the first pulse of CO (phase 1), and the OSC deduced from alternate pulses CO and O2 (phase 3), which tends to prove that the catalysts show stable redox properties all along the procedure of measurement [25]. Several reaction steps may occur in the course of the oxygen storage process [24,26]: Step 1: Oxygen adsorption, which may be dissociative (Eqn. 1) or not (Eqn. 2): 1/ O + ? 2 2 s

→ Os

2− O2 + ?2− s → O2

(1) (2)

where ?s represents a free vacant site, which may be negatively charged as in Eqn. 2. Dioxygen species, such as superoxides or peroxides, were evidenced by Fourier transform infrared (FTIR) studies [27–29]. Step 2: Reaction between COg and surface oxygen or dioxygen species to give CO2 (Eqn. 3) or carbonates (Eqn. 4) COg + Os → CO2g + ?

(3)

2− COg + O2− 2s → CO3s

(4)

Cerium-Based Oxides Used as Oxygen Storage Materials

237

Phase 2 O2 saturation

Phase 1 CO saturation

Phase 3 CO/O2 alternately

CO O2 CO2

1 2 3 4 5 6 7 8 9 10

1 2 3 4 5

OSCC

1

2

3

OSC

Figure 8.1. General procedure for OSC measurements over TWC catalysts. CO and CO2 are separated over a small Porapak column inserted between the sample and the TC detector. The first CO pulse is injected over pre-oxidized sample. CO can be replaced by H2 .

Surface carbonites, carbonates, inorganic carboxylates and sometimes formates (specially at low temperature) were identified by Fourier transform infrared (FTIR) [30,31]. Step 3: Oxygen diffusion from the bulk sites to the surface O b + ?s → O s + ?b

(5)

Step 4: Side reaction such as CO dissociation (Eqn. 6), the Boudouard reaction (Eqn. 7) or the water gas shift (WGS) reaction (Eqn. 8, with surface OH species) may also occur: COg + 2?s → Cs + Os

(6)

2COg + ?s → Cs + CO2

(7)

COg + OHs → CO2 + 1/2H2 +?s

(8)

On ceria or ceria−zirconia, OHs species are generally associated to the formation of CeOOH groups with a correlative reduction of the surface [32,33].

2.2. OSC of ceria and Cex Zr1−x O2 oxides Selected OSC values are reported in Table 8.1 for ceria and cerium−zirconium mixed oxides. These results confirm that the isomorphous substitution of Ce4+ by Zr4+ ions clearly improves the catalyst stability. BET (Brunauer, Emmett, Teller) area of ceria treated at 900 C is close to 20 m2 g−1 while it amounts to 35–45 m2 g−1 for most mixed

238

Past and Present in DeNOx Catalysis

Table 8.1. The OSC values of ceria and Cex Zr1−x O2 oxides Catalyst

Tox ( C) BET area (m2 g−1 )

CeO2

900 (6 h)

22

CO

400

71

32

Ce06 Zr04 O2

900 (6 h)

52

CO

400

232

45

9

CeO2 NP1

900 (4 h)

13

CO

400

64

49

10

Ce09 Zr01 O2 NP1

900 (4 h)

18

CO

400

125

69

10

Ce075 Zr025 O2 NP1

900 (4 h)

19

CO

400

112

59

10

Ce09 Zr01 O2 SG

900 (4 h)

43

CO

400

252

59

10

Ce075 Zr025 O2 SG2

900 (4 h)

32

CO

400

217

68

10

Pt/CeO2

500 (1 h)

49

CO

600

250

Pt/Ce075 Zr025 O2

500 (1 h)

72

CO

600

723

10

13

2

O storage Red. ( C)

OSC mol O g−1 m−2

51

Reference 9

13

Pt/CeO2

1000 (1 h)

2

CO

600

150

75

13

Pt/Ce075 Zr025 O2

1000 (1 h)

14

CO

600

522

37

13

CeO34 2

377 (2 h)

5–18

H2

377

70

61

17

Ce09 Zr01 O34 2

377 (2 h)

20–25

H2

377

280

124

17

Ce065 Zr035 O34 2 CeO35 2 Ce09 Zr01 O35 2 Ce065 Zr035 O35 2

377 (2 h)

14–18

H2

377

360

225

17

377 (2 h)

5–18

H2

377

377 (2 h)

20–25

H2

377

377 (2 h)

14–18

H2

377

12 13 75

010

17

058

17

047

17

1

Prepared by nitrate precipitation Prepared by a sol-gel technique (Zr propoxide) 3 Prepared by ball milling of oxides 4 Total OSC (value close to OSCC) 5 Dynamic OSC (pulse of H2 ) 2

oxides prepared by coprecipitation or sol-gel methods. Theoretical OSC values may be calculated according to the following assumptions: • Only Ce4+ is reduced into Ce3+ in the OSC process. Zr4+ virtually cannot be reduced. • One oxygen atom out of four (among those, which are associated with Ce ions) is available in the storage process. The theoretical surface density of oxygen ions was evaluated by Madier et al. for different crystallographic planes of CeO2 and Cex Zr1−x O2 oxides [14]. For ceria, the theoretical O density would be of 13.7, 9.7 and 15.8 at.O nm−2 for (100), (110) and (111) surfaces respectively, which gives a mean surface density of 13.1 at.O nm−2 if one assumes an equidistribution of the three crystallographic planes. This figure leads to a theoretical OSC of 5.4 mol O m−2 . The hypothesis of equidistribution may be not valid in all cases, which can explain some difference in the reported results. Note that the (111) surface is thermodynamically the most stable [34,35]. Due to the decrease of the lattice parameter upon Zr substitution, O surface density increases with the Zr content in the materials (see Ref. [36] for a review on the

Cerium-Based Oxides Used as Oxygen Storage Materials

239

micro- and nanostructures of Cex Zr1−x O2 mixed oxides). The mean oxygen surface density may be approximated by the following equation: SO atO nm−2 = 1307 + 128 1 − x

(9)

which leads to a theoretical OSC of: OSCth mol O m−2 = 543x + 0531x1 − x

(10)

The values of OSCth for the compounds listed in Table 8.1 are given in Table 8.2. If data of Table 8.1 is compared with the theoretical OSC values given in Table 8.2 (for one O layer), it appears that: • The main effect of substituting Ce by Zr is to increase the thermal stability of the materials in significant proportion. • OSC of Cex Zr1−x O2 mixed oxides is higher than that of pure ceria. It may also be higher than the theoretical OSC, which proves that several oxygen layers can be involved in the storage process. Comparison of OSC evolutions with temperature for CeO2 (25 m2 g−1 ) and Ce063 Zr037 O2 oxide (43 m2 g−1 ) was reported by Bedrane et al. [15,37]. Figure 8.2 summarizes the specific behavior of these compounds in the oxygen storage process. Up to 300 C, there is virtually no difference between OSC properties of ceria or of Ce063 Zr037 O2 . By contrast, above 300 C, OSC of the mixed oxide dramatically increases while that of pure ceria remains practically constant. Table 8.2. Theoretical OSC values of Cex Zr1−x O2 mixed oxides corresponding to the reduction of one surface layer x OSCth (mol O m−2 )

1

0.9

0.75

0.65

0.60

5.43

4.93

4.17

3.65

3.39

500

OSC (μmol-CO2/g)

450 400 350 300 250 200 150 100 50 0

200

250

300 350 400 Tem perature (°C)

450

Figure 8.2. Evolution of the OSC for CeO2 () and Ce063 Zr037 O2 ().

500

240

Past and Present in DeNOx Catalysis

Table 8.3. Comparison of OSC of ceria and Ce063 Zr037 O2 at 400 and 500 C Oxide

CeO2 Ce063 Zr037 O2

OSC 400 C (mol)

OSCth (m−2 ) 5.43 3.54

OSC 500 C (mol)

g−1

m−2

n

g−1

m−2

n

49 167

1.96 3.88

0.36 1.10

64 480

256 112

0.47 3.2

The OSC values at 400 and 500 C for the two samples are compared in Table 8.3. At these temperatures, the layer number n involved in the reduction process is significantly higher for the mixed oxide than for ceria. At 500 C, more than three cerium ion layers (i.e. almost two cells) are reduced upon the first pulse of CO.

2.3. Correlation with oxygen mobility Oxygen diffusion (Eqn. 5) being a crucial step in the global oxygen storage process, OSC measurements have been tentatively correlated to oxygen mobility measured by 18 O/16 O isotopic exchange. The principle has been described in detail elsewhere [14,38–40]. It is based on the exchange between gaseous 18 O2 and 16 O species of the support via the metal particles (Figure 8.3). The reaction is carried out in close-loop reactor connected to a mass spectrometer for 18 O2 , 18 O16 O and 16 O2 analyses as a function of time [38]. The gases should be in equilibrium with the metallic surface (fast adsorption/desorption steps: 1 and 1 ). If the bulk diffusion is slow (step 6) and the direct exchange (step 5) does occur at a negligible from the simple relationship rate, coefficients of surface diffusion DS can be calculated √ between the number of exchanged atoms Ne and t given by the model of circular sources developed by Kramer and Andre [41]:  2 Ne = √ Cm18 I0 Ds t

(11)

where I0 is the length of the metal/support interface, i.e. the total perimeter of the metal particles per m2 of catalyst, and Cm18 is the surface concentration (at. m−2 ) of 18 O atoms on the metal. The specific particle perimeter I0 is a function of the metal loading xm % and the dispersion D%: I0 = xm D2 18

16

O O 18O2

16

O2

16

(12) 18

Surface mobility

1

5

O 18O

1′

Metal

O2

2 18

3

O

Support

18

4

O

16

O

6

Figure 8.3. Measurement of surface diffusion by isotopic exchange.

Cerium-Based Oxides Used as Oxygen Storage Materials

241

Table 8.4. Relative oxygen surface mobility for some oxides at 400 C (base 10 for alumina) Oxides CeO2 MgO ZrO2 CeO2 /Al2 O3 Al2 O3 SiO2

BET area (m2 g−1 )

Relative oxygen mobility at 400 C

60 150 40 100 100 200

2300 50 28 18 10 01

The coefficient of surface diffusion for alumina is 2 × 10−18 m2 s−1

where  is a parameter depending on nature of metal and particle shape. For most metal,  is comprised between 2 × 105 and 106 m g−1 [40,42] so that the specific perimeter of particles can be greater than 108 m g−1 . This huge length allows understanding the critical role played by the metal/surface junction in diffusion processes and in catalysis. Equation 11 is generally applied in the first seconds of exchange: Cm18 is then close to the metal surface concentration, all adsorbed O atoms being 18 O atoms. Some results obtained by Martin and Duprez [38] with this isotopic exchange technique are reported in Table 8.4. Ceria shows the highest coefficient of diffusion in the range of 10−15 –10−16 m2 s−1 , which is coherent with the high OSC values obtained with this oxide. When surface diffusion is the only process of exchange, g tends to an equilibrium value ∗ at t → . In most cases, after a rapid step of surface diffusion, it can be observed that g continues slowly decreasing. This phenomenon corresponds to a slow step of bulk diffusion (coefficient Db ). A model of bulk diffusion in spherical grains was developed by Kakioka et al. which led to the following equation [43]: −Ln

g 2 A  =√ Db t ∗

Ng

(13)

where A is the BET area of the oxide used as support and its density. Equations 11 and 13 can be applied for oxides showing a relatively low surface diffusion step. Measurement of Db could be carried out with alumina and zirconia and was found in the range of 10−22 –10−23 m2 s−1 [38]. Interestingly, it was noticed that the bulk O diffusion was an order of magnitude higher in zirconia than in alumina. With ceria, measurement of Db by Eqn. 13 is not possible because of the very fast surface diffusion process, which did not allow discriminating the two regimes of diffusion. The problem is still more complex with ceria−zirconia samples for which both surface and bulk diffusion occur at a very high rate. A spatio-temporal 3D model was then developed in which all the physical steps (adsorption, desorption, surface and bulk diffusion) involved in the exchange process are taken into account simultaneously [44,45]. A program based on this model was developed and used to determine coefficients of surface and bulk diffusion in CeZrOx -supported metal catalysts. Some results are reported in Table 8.5 for three Pt/CeZrOx samples [46]. CZ-O is a conventional mixed oxide prepared by hydrolysis of ZrO(NO3 2 with an aqueous ammonia solution in the presence of a fine ceria powder. CZ-R was obtained by a reducing treatment in CO at 1200 C while CZ-D was prepared by high-energy ball

242

Past and Present in DeNOx Catalysis

Table 8.5. Coefficients of diffusion of Pt/CeZrOx catalysts [46] Catalyst BET area (m2 g−1 ) Metal particle size (nm) Temperature ( C) of exchange Surface diffusion Ds (×10−20 m2 s−1 ) Bulk diffusion Db (×10−23 m2 s−1 )

Pt/CZ-O

Pt/CZ-R

Pt/CZ-D

104 52 323 109 45

3 59 332 1450 36

37 2 334 17 034

milling of CeO2 and ZrO2 powders. This study confirmed that a high temperature of reduction can induce an extremely high mobility in CeZrOx oxides [36]. 18 O/16 O isotopic exchange can detect the presence of binuclear oxygen species (super2− oxide O− 2 , peroxide O2 ). When O2 adsorption leads to the formation of such species, exchange proceeds via a ‘multiple exchange’ mechanism in which two atoms at once are exchanged:     18 O2g + 16 O • •16 O s →16 O2g + 18 O • •18 O s (14) While 16 O18 O is the primary product of exchange in the simple mechanism (exchange atom per atom), 16 O2 is the primary product when exchange occurs via a multiple mechanism [14,40,47]. This type of mechanism could be linked to the presence of superoxides species, which can be detected by FTIR (bands at 1126 cm−1 ) [48,49]. At the beginning of exchange, the ratio P32 /P34 is close to 1 on ceria, which means that both mechanisms (simple and multiple) occur with an equal probability. On ceria−zirconia, this ratio is much higher than unity, which implies that the multiple mechanism is predominant [14]. In parallel, exchange of superoxide species can be followed by FTIR. It was shown that [16 O−16 O]− species exchange directly into [18 O−18 O]− ones without intermediary formation of [18 O−16 O]− (Figure 8.4). Finally, a clear correlation was observed between OSC and intensity of the band at 1126 cm−1 .

16O16O–

(1126 cm–1)

18

O18O–

(1062 cm–1)

0.005

1100 Wavenumber (cm–1)

1050

Figure 8.4. The FTIR spectra of adsorbed superoxide species during exchange of pre-oxidized Ce063 Zr037 O2 with 18 O2 [14].

Cerium-Based Oxides Used as Oxygen Storage Materials

243

3. IMPACT OF OSC MATERIALS IN THREE-WAY CATALYSTS The OSC materials are able to increase the efficiency by enlarging the operating window of TW catalysts [50,51]. NOx reduction in TW catalysts is closely coupled to oxidation reactions since there is a competition between NOx and O2 reaction on the reducers present in the exhaust gases [52]. We will first examine the effect of these OSC materials on the oxidation efficiency of TW catalysts.

3.1. Impact of OSC on oxidation activity – steam effects Ceria-based OSC compounds may have an impact on oxidation reactions especially when the catalysts are working around the stoichiometry (as this is the case under TW conditions). One of the first systematic studies was reported by Yu Yao [53,54]. Most results were obtained in O2 excess (0.5% CO + 0.5% O2 or 0.1% HC + 1% O2 ). Several series of Pt, Pd and Rh/Al2 O3 of various dispersion, as well as metal foils, were investigated in CO, alkane and alkene oxidation. The effect of metal dispersion in CO and the propane oxidation are shown in Figure 8.5. In every case, large particles of metal are more active in oxidation than the smallest ones. CO oxidation is moderately structure-sensitive (less than one order of magnitude between metal foil and much dispersed catalysts). By contrast, propane oxidation (and in general oxidation of small alkanes) are strongly structure-sensitive (two orders of magnitude between large and small particles). Rate equations were also expressed as

CO ox 250°C Rh Pd Pt

Rate per metal site (s–1)

10

1

0.1

C3H8 ox 250°C Rh Pd Pt

0.01

0.001

0.0001

0

20

40

60

80

100

Metal dispersion %

Figure 8.5. Effect of metal dispersion on turnover frequencies in CO and propane oxidation (M/Al2 O3 catalysts). Adapted from the data of Yu Yao [53,54].

244

Past and Present in DeNOx Catalysis

Table 8.6. Kinetic parameters of CO, propane and propene oxidation over M/Al2 O3 and M/CeO2 −Al2 O3 catalysts (M = Pt, Pd or Rh) CO E Pt/Al 104–125 Pt/Ce−Al 84 Pd/Al 108–133 Pd/Ce−Al 50 Rh/Al 92–113 Rh/Ce−Al 104 Pt 2 (250 C) Pd 1 (250 C) Rh 5 (250 C)

C3 H8

C3 H6

m

n

E

m

n

E

m

n

+10 +05 +09 0 +10 0

−09 +03 −09 +10 −08 +02

84–105 96 66–96 63 100 84 0.5 (250 C) 0.2 (350 C) 3 (400 C)

−10 −10 +01 +01 0 +01

+20 +20 +06 +06 +05 +04

67–125 80 63–117 63 67–92 92 3 (300 C) 0.5 (250 C) 2 (300 C)

+20 +15 +15 +07 −08 0

−10 −06 −05 −03 +09 +05

a function of activation energy, E, and kinetic orders with respect to O2 (m) and to CO or HC (n):   E n r = k0 exp − POm2 PCOHC (15) RT A summary of the results is given in Table 8.6. Also reported in the Table are the ratios  between the specific activity of M/CeO2 −Al2 O3 and M/Al2 O3 catalysts. Kinetic orders in CO oxidation on M/Al2 O3 can be explained by the classical Langmuir–Hinshelwood expression for the rate equation, as a function of the rate constant k, the adsorption constants K and the partial pressures P: r =k

KO PO KCO PCO 1 + KO PO + KCO PCO 2

(16)

On alumina, there is a competition between CO and O2 adsorption on the same metal sites. As CO is much more strongly adsorbed than O2 on Pt, Pd and Rh, one has: 1 + KO PO << KCO PCO and Eqn. 16 reduces to: r =k

K O PO  i.e., an order of 1 in O2 and of 1 in CO KCO PCO

(17)

Equations 16 and 17 imply that O2 adsorption is not dissociative, which is coherent with the kinetic data. However, O2 should be dissociated in further steps of the surface reaction. On ceria, new sites for O2 activation are created at the metal/support interface or in the vicinity of metal particles. As CO and O2 do not compete with the same sites, the rate equation becomes: r =k

K O PO KCO PCO × 1 + KO PO 1 + KCO PCO

(18)

with orders between 0 and 1 for CO and O2 , which is again coherent with data of Table 8.6. Numerous studies on CO oxidation have confirmed these conclusions, most

Cerium-Based Oxides Used as Oxygen Storage Materials

245

of them differing by the intimate nature of the site for O2 activation as well as by the chemical state of the metals in the reaction. For Pt/CeO2 −Al2 O3 , Serre et al. [55,56] have described a mixed site Pt−O−Ce, specially created in reducing conditions. The catalyst deactivates after a prolonged exposition in oxidative medium. The presence of specific sites for O2 adsorption on ceria was elegantly demonstrated by Johansson et al. on Pt/CeOx and Pt/SiO2 catalysts prepared by electron beam lithography [57]. On both samples, a kinetic bistability depending on the gas mixing ratio,  = PCO PCO + PO2 , can be observed. The bistable region is shifted considerably to much higher  values on Pt/CeOx than on Pt/SiO2 . This could be simulated in kinetic modeling by introducing an oxygen reactant supply via the CeOx support, which maintains a high CO conversion rate even in CO excess. Similar rate expressions can be derived for propane oxidation. The first step would be the dehydrogenating adsorption of propane on metals giving a Cx Hy fragment (coverage: C ). Contrary to CO, propane is much less strongly adsorbed on the metals than O2 . The order 2 in HC and 1 in O2 on Pt could be explained by a surface reaction between O2ads and two Cx Hy species. On Rh and Pd, O2 shows a moderate inhibitory effect much less significant than on Pt, kinetic orders being all comprised between 0 and 1. Clearly, OSC materials improve CO and to a lesser extent HC oxidation in O2 substoichiometry. OSC materials can also promote reactions of steam with CO (water–gas shift) or HC (steam reforming) [58,59]. Both reactions produce hydrogen (Eqns 19 and 20): CO + H2 O → CO2 + H2

(19) 

 m Cn Hm + nH2 O → nCO + n + H2 2

(20)

The promoter role of ceria in WGS reaction is demonstrated in Figure 8.6. Temperature-programmed conversion of CO was carried out in an oxygen-deficient medium. The oxygen content was adjusted to obtain a maximum conversion of 40% by 100 90

CO conversion %

80 70 60 50 40

PtA

30

RhA PtCeA

20

RhCeA

10

CeA

0 50

100

150

200

250

300

350

400

450

500

Temperature (°C)

Figure 8.6. Temperature-programmed reaction of 0.8% CO in 0.16% O2 + 10% H2 O over 1% Pt or 0.2% Rh catalysts supported on alumina (A) or 12% CeO2 −Al2 O3 (CeA) [58].

246

Past and Present in DeNOx Catalysis

oxidation. The catalysts were 1% Pt and 0.2% Rh deposited over alumina (A) or 12% CeO2 −Al2 O3 (CeA). The CO conversion is not limited by thermodynamic. It is higher than 99% over the entire range of temperature. Two regions can be seen on the reaction profiles: the low-temperature domain corresponding to CO oxidation (limited at a 40% conversion) and the high temperature one where the CO having not been oxidized react with water (WGS domain). For every reaction, Pt is the best metal and ceria acts as a promoter. However, an exceptional increase of conversion can be observed in WGS when the metals are deposited on CeA. Similar experiments were carried out using propane instead of CO as reducer [58]. It was shown that Pt was the best metal for C3 H8 oxidation, while Rh was the most active one in propane steam reforming. Introduction of 12% CeO2 in the alumina support confirms that CeO2 rather inhibits C3 H8 oxidation over Pt while it acts as a promoter on Rh. In both cases, especially with Rh, CeO2 increases the rate of propane steam reforming.

3.2. Impact of OSC on NO conversion The role of OSC materials in NO conversion is more complex. Most metals, specially Rh, are able to decompose NO, but this reaction is rapidly inhibited by O species resulting from this decomposition [60]. On Rh, for instance, Rh−O species are replaced by Rh−NO+ ones in which NO is no longer dissociated [61]. O species may react with adsorbed species of the reducer (CO, HC) to form CO2 . The first role of the OSC materials could be to liberate metal sites by accepting O species. Indirect effects can also occur, with the Ce3+ /Ce4+ redox system being able to regulate the metal state (zero-valent or ionic) during richness oscillations.

3.2.1. Reduction of NO by CO The reaction may lead to N2 (Eqn. 21) or N2 O (Eqn. 22). NO + CO → 1/2N2 + CO2

(21)

2NO + CO → N2 O + CO2

(22)

For environmental reasons, reaction (Eqn. 21) (NO → N2 ) should be promoted, N2 O having a dramatic greenhouse gas effect. The different steps of reaction (Eqn. 23) have been investigated in detail, mainly by FTIR spectroscopy [61–63]. One of the possible intermediate is isocyanate. NCO species could be formed on the metal and migrate on the support, which may explain the large differences observed when Rh is supported on different oxides (alumina, silica, zirconia, ceria-alumina, etc.). However, the main step should be the dissociative adsorption of NO: NOg + 2∗ → N∗ + O∗

(23)

This adsorption reaction has been extensively studied on most noble metals, especially on Rh by NO thermodesorption [64–66]. On Rh/ZrO2 [65], it was shown that N2 left the surface from two separate features: a sharp 1 peak at 170 C due to N2 desorption

Cerium-Based Oxides Used as Oxygen Storage Materials

247

by reaction (Eqn. 24) and a broad peak (2 ) between 180 C and 430 C corresponding to N2 recombination (Eqn. 25): NOg + N∗ → N2g + O∗

(24)

N∗ + N∗ → N2g

(25)

These reactions are very sensitive to particle size of rhodium. On big particles (10 nm), both reactions can occur while, on the smallest ones (3–6 nm), only the recombination feature was observed. One of the first systematic studies of the NO reduction by CO was reported by Taylor and Kobylinski in 1974 [67]. The reaction was carried out in large excess of CO (0.5% NO + 2% CO), which favors the formation of N2 . The light-off temperature T50 (50% conversion of NO) allowed ranking the metals (0.5% M/Al2 O3 ) by increased activity: Ru 205 C > Rh 296 C > Pd 431 C > Pt 471 C Except Ru (not usable in TWC because of the volatility of its oxide [68]), the most active metal is the rhodium. This has been largely confirmed by further studies so that Rh may be considered as a key-component of TWC for NO reduction [69,70]. As far as Pd is concerned, it seems that the active site is composed of Pdn+ −Pd0 pairs, which may explain the higher activity of Pd in NO+CO+O2 mixture (T50 ≈ 200 C) [71]. A detailed kinetic study by Pande and Bell on Rh catalysts has evidenced a significant support effect [72]. The kinetic data were represented by a conventional power law expression:  RNO = kNO PNO PCO

(26)

where RNO is the reaction rate per second and per metal site, P, the pressure in atm and kNO the kinetic constant in atm− + s−1 . The main results are reported in Table 8.7. The different catalysts may be ranked according to their relative activity: TiO2 95 > La2 O3 58 > SiO2 52 > MgO 36 > Al2 O3 1 Between 100 and 200 C, the selectivity to N2 O is very high (70–85%) and decreases with the temperature in accordance with most of the studies on the reduction of NO by CO. The kinetics of this reaction (1.5% NO + 3% CO, 200 C) on a 1% Rh/SiO2 catalyst promoted by Mo, Ce and Nb (1–10 wt-%) was studied by Hecker et al. [73]. A significant particle size effect was observed: the biggest particles of Rh are the most active per Rh site (Figure 8.7). On silica, the promoter effects are mainly due to particle size effects, certain promoters, especially ceria being able to stabilize Rh particles. The presence of Mo tends to slow down the reaction so that Mo appears as an inhibitor of the reaction. Kinetic parameters obtained by Hecker et al. are close to those reported by Pande and Bell: orders slightly negative in NO (−02 to −05, exceptionally −12 for Mo) and orders nil or slightly positive with respect to CO (0 to 0.4); activation energies between 117 and 139 kJ mol−1 .

248

Past and Present in DeNOx Catalysis

Table 8.7. Kinetics of the CO + NO reaction at 210 C over Rh catalysts. Gas composition is 1.3%NO+ 3.9%CO [72] Dispersion Order/NO Order/CO Ea (kJ mol−1 ) kNO 10−3 (%)

Catalyst

BET area (m2 g−1 )

4.6% Rh/SiO2 4.2% Rh/Al2 O3 4.6% Rh/MgO 4.3% Rh/TiO2 4.2% Rh/La2 O3

250

55

−02

+01

140

4.30

175

64

−04

∼0

101

0.82

100

46

−02

+01

108

2.92

50

26

+01

−02

82

7.75

14

10

−02

∼0

126

4.75

Turnover frequency (s–1)

0.020

0.015

Mo series

0.010

Ce series Nb series Unmodified

0.005

0.000 1

2

3

4

5

6

7

Rh particle size (nm)

Figure 8.7. Effect of the dispersion of Rh on the NO reduction by CO at 200 C over promoted Rh/SiO2 catalysts (1.5% NO + 3% CO) [73].

The reduction of NO by CO was investigated in detail by Oh et al. in closer conditions than those encountered in automotive converters: weaker NO and CO concentrations, close to the stoichiometry, more realistic metal content (<1%) [74,75]. The main conclusions of Oh’s works are: • The high sensitivity of the reaction to particle size of Rh is confirmed: at 230 C, in a mixture of 0.5% NO + 1% CO, the turnover frequency increases from 0.017 s−1 for a highly dispersed catalyst to 0.74 s−1 for a catalyst dispersed at 1.7%, the activity per metal site on unsupported Rh catalysts being still much higher. • Ceria remarkably increases the activity of Rh/Al2 O3 : 2–9 wt-% CeO2 added to a 0.014% Rh/Al2 O3 catalyst increases its activity by a factor 5 by 250 C. However, ceria decreases the activation energy (120 kJ mol−1 on Rh/Al2 O3 : instead of 80 on Rh/CeO2 −Al2 O3 ) so that the two catalysts have nearly the same activity around 300–310 C.

Cerium-Based Oxides Used as Oxygen Storage Materials

249

• The reaction CO + NO may be seen as an oxidation of CO by NO, which competes with the oxidation of CO by O2 . On Rh, CO oxidation by dioxygen is almost two orders of magnitude faster than CO oxidation by NO. Nevertheless, if the three reactants are present together, the rate of CO oxidation by O2 dramatically decreases while that of CO by NO increases. As a consequence, both reactions occur practically at the same rate [76]. Most of the kinetic observations can be interpreted as follows: • The key-step is the NO decomposition (Eqn. 23): the global reaction rate depends for a great part of the rate on this step. • At low temperature, N2 is mainly formed by the reaction of NOg with adsorbed N∗ species (Eqn. 24) or by the same reaction with adsorbed NO (Eqn. 27): NO∗ + N∗ → N2 + O∗ + ∗

(27)

while N2 O is formed by a similar reaction (Eqn. 28): NO∗ + N∗ → N2 O∗ + ∗

(28)

At high temperature, NO is almost totally dissociated: N2 O cannot then be formed and N2 stems essentially from recombination of adsorbed N∗ species (Eqn. 25). The selectivity to N2 increases with the temperature and virtually reaches 100% around 300 C (Figure 8.8). The chemical state of the metal can play a decisive role on the reaction mechanism. In TWC, Rh is thought to remain in the zero-valent state, which favors NO dissociation [77,78]. However, the role of the OSC materials is complex, and it is not inert with respect to NO activation. Ranga Rao et al. [79] showed that, when bulk oxygen vacancies are formed in a reduced Ce06 Zr04 O2 solid solution, NO was efficiently decomposed on the support to give N2 O and N2 . Further studies by the same group

200

U.A.

CO CO2

150

NO 100

N2 50

N2O 0 50

100

150

200

250

300

350

400

Temperature (°C)

Figure 8.8. Temperature-programmed reaction of CO (1%) + NO (1%) over a pre-oxidized 0.2% Rh/CeO2 −Al2 O3 catalyst (Pirault and Marécot, unpublished results).

250

Past and Present in DeNOx Catalysis

Table 8.8. Effect of the reducing pretreatment of 0.5% Rh catalysts on the NO + CO reaction rate (1% NO + 1% CO) [80,81] Support

Rh precursor

Reaction rate at 200 C after pretreatment in H2 , for 2 h at 200 C (mole NO g−1 s−1 ×109 )

Reaction rate at 200 C after pretreatment in CO + NO from 200 to 500 C (mole NO g−1 s−1 ×109 )

Al2 03 Ce04 Zr06 O2 Ce06 Zr04 O2 CeO2 Ce06 Zr04 O2

Chloride Chloride Chloride Chloride Nitrate

70 331 336 1120 16301

32 90 46 83 15901

1

Reaction rate at 160 C

using different Cex Zr1−x O2 composition and different metals confirmed this prominent property of supports containing reduced ceria [80,81]. Selected results of these studies are reported in Table 8.8. A significant increase of activity can be observed on Rh catalysts supported on reducible oxides. Activity exaltation is severely annihilated when the catalysts are treated in the reaction mixture. Nevertheless, the presence of chlorine largely upset the results. Cl probably slows down the reduction of the support, particularly in the CO + NO mixture. By surface science techniques, Mullins and Overbury showed that the presence of reduced ceria might create new active sites for NO dissociation [82]. The degree of decomposition is increased and the onset temperature for decomposition is reduced when Rh is supported on reduced ceria (Rh/CeOx ) compared to Rh on oxidized ceria (Rh/CeO2 ) NO dissociation being self-inhibiting. The promotion by reduced ceria could be due to a spillover phenomenon of O species from metal to support. In fact, this is not sufficient to explain all the results of Mullins and Overbury. An exposure of the Rh/CeOx surface to water leads to a re-oxidation accompanied by a hydroxylation of the support while the metal surface is left unchanged. In fact, it seems that preferential orientation of Rh surface on reduced ceria may also explain the specific role of CeOx surface. This is consistent with the fact that NO dissociation occurs at lower temperatures on Rh (110) and on Rh (100) than on Rh (111) [83,84]. However, reduced ceria is able, alone, to dissociate NO. Martinez-Arias et al. [85] have first investigated by electron paramagnetic resonance (EPR) and FTIR spectroscopies NO reaction on ceria pre-outgassed at different temperatures and showed the role of superoxides differentially coordinated in the formation of hyponitrites species further decomposed into N2 O. Later Haneda et al. [86] have demonstrated that reduced ceria and reduced praseodymium oxide dissociate NO even though the presence of a noble metal (Pt) significantly increases the formation of N2 or N2 O. The main results of this study are summarized in Table 8.9. Lanthanum and samarium show virtually no NO dissociation activity even in the presence of Pt. These supports are not reducible and have no OSC property. The intrinsic NO dissociation activity of platinum is very weak, probably in reason of the low metal dispersion. The behavior of terbium oxide is more surprising. Although it is reducible in H2 , it is unable to dissociate NO except in the presence of Pt.

Cerium-Based Oxides Used as Oxygen Storage Materials

251

Table 8.9. NO dissociation over reduced rare-earth oxides and over 1% Pt catalysts deposited on these oxides. Gas Prior to NO dissociation (970 ppm NO), the samples are reduced for 1 h in H2 at 500 C [86] BET area (m2 g−1 )

H2 temperatureprogrammed reduction (25–500 C) mol H2 g−1

N2 (N2 O) formed (mmol g−1 ) at 200 C

at 400 C

RE oxide La2 O3 CeO2 Pr6 O11 Sm2 O3 Tb4 O7

75 55 95 80 22

25 75 773 15 1008

00 00 147 26 196 46 00 00 00 00

Pt/RE oxides Pt/La2 O3 Pt/CeO2 Pt/Pr6 O11 Pt/Sm2 O3 Pt/Tb4 O7

17 271 1823 41 1356

00 00 500 11 225 (226) 00 00 272 81

18 02 523 00 431 (10.5) 55 02 248 (0.0)

The specific role of OSC materials in NO activation and NO dissociation has largely been confirmed by many authors over Pt−Rh [87,88] and Pd catalysts [89,90] or even over bare OSC oxides [91]. By EPR, Lecomte et al. [87] evidence the presence of − O− 2 superoxide species over a Pt Rh/Al2 O3 catalyst modified by ceria. The formation of these species could be closely related to the performance of the Pt−Rh/CeO2 −Al2 O3 catalyst in CO + NO reaction. When the ceria-containing catalyst is pre-reduced before reaction, a typical temperature-programmed reaction profile can be observed (Figure 8.9). While 100

NO conversion (%)

90 80

Pt-Rh/CeO2–

70 60 50 40 30

Pt-

20 10

CeO2–

0 50

100

150

200

250

300

350

400

450

Temperature (°C)

Figure 8.9. Effect of ceria on the NO reduction by CO (gas composition: 0.5% NO + 0.5% CO, 25 000 h−1 ). After Ref. [87].

252

Past and Present in DeNOx Catalysis

prereducing, the catalyst has virtually no effect on Pt−Rh/Al2 O3 , a significant activity peak can be observed at low temperature on the ceria-based catalyst. This peak disappears upon calcination, and a profile like that of Figure 8.9 can then be recorded. It is not systematically recorded in bench tests with complex synthetic mixtures [92]. The behavior of CeO2 −Al2 O3 support is rather general and can be observed with other ceria-containing catalysts such as Rh/CeO2 −ZrO2 catalysts [80,81] or Pd/CeO2 −ZrO2 [93]. It has also been shown that pre-reduced ceria−zirconia supports present a noticeable activity in NO + CO in the absence of metal. This activity totally disappears when the support is calcined [88].

3.2.2. Reduction of NO by H2 and hydrocarbons Compared to CO, these reactions were much less studied over TW catalysts. Kobylinski and Taylor [67] have compared the NO reduction by CO and by H2 . Their main results are summarized in Tables 8.10 (light-off activity) and 8.11 (selectivity). Pt and Pd are by far the most active metals in NO reduction by H2 while Rh and Ru present the highest activity in NO reduction by CO. However, when the two reducers are injected together (last column), CO tends to impose its behavior in NO reduction. This is due to a strong adsorption of CO, which inhibits the reduction by H2 . Although CO seemed to inhibit the reduction by H2 , this reducer still maintains a very significant activity. The CO inhibition is largely compensated by the high temperature of working. Interestingly, Table 8.11 confirms the good selectivity of Rh (and Ru) to N2

Table 8.10. Temperatures (T  C) for 50% NO conversion as a function of the reducing agent. Catalysts: 0.5% metal on alumina. Space velocity: 24 000 h−1 Catalyst

NO + H2

NO + CO

NO + CO + H2

121 106 163 237

471 431 296 205

398 330 275 210

Pt Pd Rh Ru

Table 8.11. Product selectivity (% NO converted in N2 and NH3 ) and reaction selectivity (% NO converted by NO + CO and NO + H2 ). Gas composition 1.5% NO + 4.5% CO + 4.5% H2 Catalyst

Pt Pd Rh Ru

T ( C)

515 515 482 432

NO conv. (%) 94 94 100 100

Selectivity NO → N2

NO → NH3

NO + CO

NO + H2

23 26 67 92

77 74 33 8

8 9 20 29

92 91 80 71

% N2 from NO + CO 40 38 29 31

Cerium-Based Oxides Used as Oxygen Storage Materials

253

while Pt and Pd lead to a significant formation of ammonia. Maunula et al. [94] have confirmed the superiority of Pt over Rh in the reduction of NO by H2 . They also found that N2 O would be an intermediate in the formation of N2 . Relatively, few studies were devoted to the reduction of NO by HC in TW conditions. Platinum seems to be a good reducer of NO by light alkanes, specially the methane [95]: CH4 + 4NO → 2N2 + CO2 + 2H2 O

(29)

However, some contradictory results were obtained in several studies. For instance, in the CH4 −NO reaction, some authors have reported that N2 O was the primary product [95] while others found that ammonia was first produced [96]. The presence of water can play a decisive role since H2 O allows generating H2 by WGS or steam reforming [59]. Olefins generally show a higher activity than alkanes. Propene for instance has been found more reactive than propane. Some exceptions should be quoted, ethylene has been found less reactive than CH4 in NO reduction at stoichiometry [97]. The role of ceria is practically not evoked in these previous studies. More recently, Pérez-Hernández et al. have compared the catalytic behavior of Pt/ZrO2 , Pt/CeO2 and Pt/CeO2 −ZrO2 in NO reduction by CH4 or CO [98]. All the catalysts were found more active and more selective to N2 in NO reduction by CH4 . However, the cerium content plays a decisive role in decreasing the differences of NO light-off activity between CH4 + NO and CO + NO. Rather different results were obtained by Bera et al. [99] who found that Pt/CeO2 is more active in CO+NO (100% conversion achieved at 270 C) than in NO + CH4 (100% conversion at 350 C). However, the superiority of the ceria support with respect to alumina in NO reduction was demonstrated both for Pt and Pd catalysts. As said before, ceria may also modify the catalyst behavior by increasing the H2 content in rich conditions, which may induce a higher capacity of the catalyst to reduce NO. A close parallelism could be established between the OSC of aged catalysts and their light-off activity in CO, HC and NOx conversion [100]. Different evolutions of both the OSC and the catalytic activity were observed depending on the method of ageing (laboratory ageing or engine bench ageing) and on the type of catalyst (PtRh/CeO2 −Al2 O3 or PdRh/CeO2 −Al2 O3 ). In every case, the conversion of NOx is the most sensitive to catalyst ageing. This has been confirmed by Muraki and Zhang who compared CO, HC and NO conversions over aged Rh catalysts supported on ceria and ceria−zirconia [70]. The conversions were significantly higher over ceria−zirconia supported Rh catalysts, much more stable than those supported on pure ceria. Ageing the catalyst, especially in rich/lean oscillation, may also lead to heterogeneity in the Ce local concentration. Finally, possible migration of Ce4+ ions was observed by Fan et al. in Pt/Ce067 Zr033 O2 catalyst in rich/lean oscillations. Cerium ions may then decorate Pt particles and deeply modified their behavior in NO reduction [101].

4. IMPACT OF OSC MATERIALS IN LEAN-BURN AND DIESEL CATALYSIS The impact of oxygen storage in DeNOx catalysis in O2 excess is more complex and strongly depends on the process used for NOx reduction. The impact of OSC materials will be examined on two processes: the selective reduction by HC (HC-SCR) and the NOx -trap process.

254

Past and Present in DeNOx Catalysis

4.1. Impact of OSC materials in HC-SCR In the 1990s, most investigations were devoted to passive DeNOx , i.e. NOx reduction by the HC present in the exhaust gas. Even in lean conditions, there remains Ce3+ cations and oxygen vacancies, which may participate in NO or HC activation. In 2000, Djega-Mariadassou [77] proposed the following model for NO adsorption in Rh/CeO2 where Rh is partially oxidized while there remain reduced sites of ceria on the support (Figure 8.10) However, there are also evidences that Ce3+ cations could participate in NO activation and dissociation [85,102,103]. The model of Figure 8.10 was later modified to take into account this possibility [104]. It seems that the behavior of Pd/CeO2 −ZrO2 can be explained by similar models of NO and HC activation [105]. NO would be adsorbed on reduced sites of ceria−zirconia while the hydrocarbon (propylene in this study) would be partially oxidized into Cx Hy Oz intermediates over PdOx sites. For the reduction of NO in oxygen excess, it seems, however, that conventional catalysts, even doped with OSC materials, cannot adsorb the HC’s to reduce the NOx efficiently. Engineers of Toyota have proposed a system composed of a dual bed in which the conventional Pt catalyst is mixed with different zeolites [106]. A relatively close contact between Pt and the zeolite, which stores the HC’s is required, but the advantage of this technique is that the zeolite as well as the presence of an OSC materials prevent the poisoning of Pt by heavy HC’s. Some experiments were carried out with Pt in the zeolite but the most interesting system would the multi-component catalyst schematized in Figure 8.11. The idea to insert an OSC component such as Ce ions in zeolite has received much attention. Cordoba et al. [107] have shown that H-ZSM5 promoted both by Ce and Pd has excellent properties in lean-DeNOx by dodecane (Figure 8.12). As in the previous N N O

O2−

Rh+δ O

O2−

Figure 8.10. Model of adsorption site for NO in the presence of O2 (adapted from Ref. [77]). The cubes represent oxygen vacancies. Rh+ would be the site for HC (or CO) activation.

NO N2 HC HC HC

HC

Zeolite

Pt

O CeO2-ZrO2

Support

Figure 8.11. Cooperative effect of an OSC component and a zeolite in SCR-HC of NO in O2 excess.

Cerium-Based Oxides Used as Oxygen Storage Materials

255

80

NO conversion (%)

HMOR 60

Pd-HMOR Ce-HMOR PdCe-HMOR

40

20

0 100

200

300

400

500

Temperature (°C)

Figure 8.12. Cooperative effect of Pd and Ce in lean-DeNOx by dodecane (900 ppm NO+100 ppm NO2 + 30 ppm N2 O + 400 ppm C12 H26 + 6% O2 [107].

example (Figure 8.11) the zeolite should help storing and probably transforming the dodecane molecule into highly reducing species. Thanks to their redox properties, Ce ions could adsorb NO and maintain Pd in the better chemical state to reduce NO. Very few studies were devoted to NO reduction by H2 in lean conditions because of the high reactivity of H2 with O2 . Costa et al. [108,109] discovered a new low loaded Pt catalyst able to perform H2 -DeNOx with a good selectivity. It consists of 0.1% Pt supported on La05 Ce05 MnO3 . This catalyst was found very active (74% NO conversion at 140 C in a mixture, 0.25% NO, 1% H2 , 5% O2 , 5% H2 O/He, GHSV, and 80 000 h−1 ) and more selective to N2 (87% at 140 C) than a conventional 0.1% Pt/alumina catalyst [108]. By transient isotopic exchange reaction, Costa and Efstathiou showed that the interface metal/support played a significant role in NO reduction. Two kinds of active NOx species were evidenced: M−NO+ 2 located on the La05 Ce05 MnO3 support and M−O−(NO)−O−M at the interface metal/support, while on a non-selective Pt catalyst, virtually all the active species are on Pt (Pt−NO+ or Pt-nitrate) [109].

4.2. Impact of OSC materials on NOx -trap systems One of the most promising processes is the active DeNOx based on NOx -trap materials. It has been developed for lean-burn gasoline engines. Cerium compounds are thought to intervene in different steps of the whole process: (1) NO oxidation, (2) NOx storage, (3) Nitrate desorption and NOx reduction. Most probably, the main role of OSC materials is to accelerate HC partial oxidation during rich-spikes (giving CO and H2 as NOx reducers). However, this beneficial effect of OSC compounds competes with a detrimental reaction, i.e. the reduction of the OSC materials itself that may delay HC decomposition and thus NOx reduction [110]. Nakatsuji et al. [111], in cooperation with researchers of Isuzu, have studied the effect of OSC materials on simplified catalysts in as much as Rh was directly deposited on oxides known for their OSC properties (Ce, Ce−Zr, Ce−Pr, Ce−Nd−Pr, Ce−Gd−Zr oxides). These catalysts were submitted to periodic lean (55 s)/rich (5 s) excursions as in the conventional NOx -trap system. Compared to non-OSC supports, ceria allows maintaining a high DeNOx activity even in large O2 excess during the lean period (Figure 8.13).

256

Past and Present in DeNOx Catalysis

NO conversion (%)

100 80 1%Rh/ceria @ 250°C

60

1%Rh/alumina @ 300°C

40 20 0 0

1

2

3

4

5

6

7

8

9

10

Oxygen concentration (%)

Figure 8.13. Effect of O2 concentration over Rh-loaded catalysts at the temperature of their maximum activity (lean/rich spans: 55/5 s; 50 000 h−1 ).

Even in absence of the classical Ba component, OSC materials may play a role in the NOx reduction using severely lean/rich conditions. Most of other studies implying OSC materials were devoted to the chemical interaction between the NOx -trap materials (Ba) and the OSC oxides. The order of introduction of the different functions (metal = Pd, OSC and NOx -trap) was studied by Kolli et al. [112]. Although the aim of this work was to study the catalysts in TWC conditions, some interesting results were obtained, which may be generalized to other conditions. It was shown that the best catalysts consisted of Pd−Ba or Ba−Pd deposited over OSC-Al2 O3 support, i.e. the OSC materials should be impregnated first. Although their catalyst was not optimized, Liotta et al. [113] have investigated a Pt-OSC/Ba−Al2 O3 catalyst in NOx trap conditions. Interestingly, they showed that Ba ions migrated through the OSC layer (CeZrOx). This property could allow a better control of the Ba dispersion as well as an increased resistance to SO2 poisoning. However, the presence of Ba is not essential for the NOx -trap when OSC material is used as support. Eberhardt et al. [114] and Philipp et al. [115] have compared the NOx trap behavior of Ba/Al2 O3 and Ba/CeO2 materials. Ba/CeO2 show higher NOx storage efficiency than Ba/Al2 O3 within the temperature range of 200–400 C. No solid/solid reaction between Ba and ceria was observed below 780 C while Ba reacted at lower temperature with Al2 O3 to form inactive Ba aluminate. In these previous studies, however, no noble metal was impregnated on the support, which may bias the results in real catalysts. Corbos et al. have compared the NOx storage capacities of Pt/CeZrOx with a classical Pt/Ba/Al2 O3 and with Pt/Ba/CeZrOx. The values given in Table 8.12 show a better capacity for Pt/CeZrOx at certain temperatures [116]. Table 8.12. NOx storage capacities (mol g−1 ) calculated for the first 100 s. The catalysts were stabilized at 700 C under O2 , H2 O and N2 ; storage mixture: 350 ppm NO, 10% O2 , 10% H2 O, 10% CO2 and N2 . (Ref. [116]) NOx storage capacities mol g−1  Storage temperature ( C) Pt/Ba/Al (129 m2 g−1 ) Pt/CeZr (61 m2 g−1 ) Pt/Ba/CeZr (47 m2 g−1 )

200 131 171 94

300 137 146 118

400 183 150 233

Cerium-Based Oxides Used as Oxygen Storage Materials

257

Similar results were obtained by Lin et al. [117] who investigated the effect of La and Ce on NOx storage properties of Pt/Ba−Al2 O3 . Substituting La for Ce increased the NOx storage capacity from 341 mol g−1 in Pt25 La305 Ba334 Al100 to 1020 mol g−1 for the Pt25 Ce305 Ba334 Al100 catalyst. Finally, unconventional OSC supports have also been investigated. Machida et al., for instance, studied a Pd/MnOx −CeO2 as NOx -storage materials [118]. This catalyst was tested in lean/rich conditions using H2 as reducer with, however, an unusual period (9 min. lean/3 min. rich). The catalyst showed both excellent activity and selectivity (almost 100% to N2 ).

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

Ghandi, H.S., Piken, A.G., Shelef, M. et al. (1976) SAE Paper 760201. Yao, H.C. and Yu Yao, Y.F. (1984) J. Catal., 86, 254. Su, E.C., Montreuil, C.N. and Rothschild, W.G. (1985) Appl. Catal., 17, 75. Su, E.C. and Rothschild, W.G. (1986) J. Catal., 99, 506. Engler, B., Koberstein, E. and Schubert, P. (1989) Appl. Catal., 48, 71. Miki, T., Ogawa, T., Haneda, M. et al. (1990) J. Phys. Chem., 94, 6464. Ozawa, M., Kimura, M. and Isogai, A. (1993) J. Alloys Comp., 193, 73. Fornasiero, P., Di Monte, R., Ranga Rao, G.J. et al. (1995) J. Catal., 151, 168. Cuif, J.P., Blanchard, G., Touret, O. et al. (1997) SAE Paper Nr 970463. Rossignol, S., Gérard, F. and Duprez, D. (1999) J. Mater. Chem., 9, 1615. Lox, E.S.J. and Engler, B.H. (1999) In Environmental Catalysis (G. Ertl et al. eds.), Wiley-VCH, New York, p. 1. Kacimi, S., Barbier, J. Jr., Taha, R. et al. (1993) Catal. Lett., 22, 343. Hori, C.E., Permana, H., Simon Ng, K.Y. et al. (1998) Appl. Catal. B Environ., 16, 105. Madier, Y., Descorme, C., LeGovic, A.M. et al. (1999) J. Phys. Chem. B, 103, 10, 999. Bedrane, S., Descorme, C. and Duprez, D. (2002) Catal. Today, 73, 233. Rohart, E., Larcher, O., Hédouin, C. et al. (2004) SAE Paper 04-01-1274. Trovarelli, A., Zamar, F., Llorca, J. et al. (1997) J. Catal., 169, 490. Duprez, D. and Descorme, C. (2002) In Catalysis by Ceria and Related Materials (A. Trovarelli ed.), Chapter 7. Imperial College Press, London, pp. 243–80. Barbier, J., Bahloul, D. and Marécot, P. (1992) J. Catal., 137, 377. Martin, D. and Duprez, D. (1995) Appl. Catal. A Gen., 131, 297. Pisanu, A.M. and Gigola, C.E. (1996) Appl. Catal. B Environ., 11, L37. Douidah, A., Marécot, P., Szabo, S. et al. (2002) Appl. Catal. A Gen., 225, 21. Yoshida, H., Yazawa, Y. and Hattori, T. (2003) Catal. Today, 87, 19. Holmgren, A., Andersson, B. and Duprez, D. (1999) Appl. Catal. B Environ., 22, 215. Rossignol, S., Descorme, C., Kappenstein, C. et al. (2001) J. Mater. Chem., 11, 2587. Bedrane, S., Descorme, C. and Duprez, D. (2002) Catal. Today, 75, 401. Soria, J., Martinez-Arias, A. and Conesa, J. (1995) J. Chem. Soc. Faraday Trans., 91, 1669. Li, C., Domen, K., Maruya, T. et al. (1989) J. Am. Chem. Soc., 111, 7683. Descorme, C., Madier, Y. and Duprez, D. (2000) J. Catal., 196, 167. Li, C., Sakata, Y., Arai, T. et al. (1989) J. Chem. Soc. Faraday Trans., 1, 85, 929; Li, C., Sakata, Y., Arai, T. et al. (1989) J. Chem. Soc. Faraday Trans., 1 85, 1451. Binet, C., Badri, A., Boutonnet-Kizling, M. et al. (1994) J. Chem. Soc. Faraday Trans., 90, 1023. Daturi, M., Finnochio, E., Binet, C. et al. (1999) J. Phys. Chem. B, 103, 4884. Norman, A. and Perrichon, V. (2003) Phys. Chem. Chem. Phys., 5, 3557.

258 [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73]

[74] [75] [76] [77]

Past and Present in DeNOx Catalysis Sayle, T.X.T., Parker, S.C. and Catlow, C.R.A. (1994) Surf. Sci., 316, 329. Conesa, J.C. (1995) Surf. Sci., 339, 337. Di Monte, R. and Kašpar, J. (2005) J. Mater. Chem., 15, 633. Bedrane, S. (2002) Ph.D dissertation, Poitiers University, France. Martin, D. and Duprez, D. (1996) J. Phys. Chem., 100, 9429. Holmgren, A., Duprez, D. and Andersson, B. (1999) J. Catal., 182, 441. Duprez, D. (2006) In Isotopes in Catalysis (J. Hargreaves et al. eds.), Chapter 7. Imperial College Press, London, 2006. Kramer, R. and Andre, M. (1979) J. Catal., 58, 287. Duprez, D., Pereira, P., Miloudi, A. et al. (1982) J. Catal., 75, 151. Kakioka, H., Ducarme, V. and Teichner, S.J. (1971) J. Chim. Phys., 68, 1715. Galdikas, A., Descorme, C. and Duprez, D. (2004) Solid State Ionics, 166, 147. Galdikas, A., Descorme, C. and Duprez, D. (2004) Appl. Surf. Sci., 236, 342. Galdikas, A., Descorme, C., Duprez, D. et al. (2004) Top. Catal., 30/31, 405. Duprez, D. (1997) Stud. Surf. Sci. Catal., 112, 13. Li, C., Domen, K., Maruya, K. et al. (1989) J. Am. Chem. Soc., 111, 7683. Li, C., Domen, K., Maruya, K. et al. (1990) J. Catal., 123, 436. Nunan, J.G., Silver, R.G. and Bradley, S.A. (1992) In Catalytic Control of Air pollution (R.G. Silver, J.E. Sawyer, J.C. Summers eds.), ACS Symposium Series, Vol. 495, Chapter 7. p. 83. Ozawa, M. (1998) J. Alloys Comp., 275–277, 886. Pˆarvulescu, V.I., Grange, P. and Delmon, B. (1998) Catal. Today, 46, 233. Yu Yao, Y.F. (1980) Ind. Eng. Chem. Prod. Res. Dev., 19, 293. Yu Yao, Y.F. (1984) J. Catal., 87, 152. Serre, C., Garin, F., Belot, G. et al. (1993) J. Catal., 141, 1. Serre, C., Garin, F., Belot, G. et al. (1993) J. Catal., 141, 9. Johansson, S., Österlund, L. and Kasemo, B. (2001) J. Catal., 201, 275. Barbier, J. Jr. and Duprez, D. (1993) Appl. Catal. B Environ., 3, 61. Barbier, J. Jr. and Duprez, D. (1994) Appl. Catal. B Environ., 4, 105. Dubois, L.H., Hansma, P.K. and Somorjai, G. (1980) J. Catal., 65, 318. Dictor, R. (1988) J. Catal., 109, 89. Solymosi, F., Volgyesi, L. and Rasko, J. (1980) Z. Phys. Chem., 120, 79. Solymosi, F., Bánsági, T. and Novák, E. (1988) J. Catal., 112, 183. Cho, B.K., Shanks, B.H. and Bailey, J.E. (1989) J. Catal., 115, 486. Zafiris, G.S. and Gorte, R.J. (1991) J. Catal., 132, 275. Kasp˘ar, J., de Leitenburg, C., Fornasiero, P. et al. (1994) J. Catal., 146, 136. Kobylinski, T.P. and Taylor, B.W. (1974) J. Catal., 33, 376. Taylor, K. (1984) In Catalysis, Science and Technology (J.R. Anderson, M. Boudart eds.), Vol. 5. Springer Verlag, Berlin, p. 119. Leclercq, C., Dathy, C., Mabilon, G. et al. (1991) In Catalysis and Automotive Pollution Control CaPoC 2 (A. Crucq, ed.), Stud. Surf. Sci. Catal., 71, 181. Muraki, H. and Zhang, G. (2000) Catal. Today, 63, 337. Duplan, J.L. and Praliaud, H. (1991) In Catalysis and Automotive Pollution Control CaPoC 2. (A. Crucq ed.), Stud. Surf. Sci. Catal., 71, 667. Pande, N.K. and Bell, A.T. (1986) J. Catal., 98, 7. Hecker, W.C., Wardinsky, N.D., Clemmer, P.G. et al. (1992) In Progress in Catalysis. Proceedings of 12th Canadian Symposium on Catalysis. (K.J. Smith, E.C. Sanford eds.), Stud. Surf. Sci. Catal. 73, 211. Oh, S.H. (1990) J. Catal., 124, 477. Oh, S.H. and Eickel, C.C. (1991) J. Catal., 128, 526. Oh, S.H. and Carpenter, J.E. (1986) J. Catal., 101, 114. Djega-Mariadassou, G., Fajardie, F., Tempère, J.-F. et al. (2000) J. Mol. Catal., 161, 179.

Cerium-Based Oxides Used as Oxygen Storage Materials [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118]

259

Djega-Mariadassou, G. (2004) Catal. Today, 90, 27. Ranga Rao, G., Fornasiero, P., Di Monte, R. et al. (1996) J. Catal., 162 1. Di Monte, R., Fornasiero, P., Graziani, M. et al. (1998) J. Alloys Comp., 275–277, 877. Fornasiero, P., Ranga Rao, G., Kaspar, J. et al. (1998) J. Catal., 175, 269. Mullins, D.R. and Overbury, S.H. (2002) Surf. Sci., 511, L293. Schmatloch, V., Jirka, I. and Kruse, N. (1994) J. Chem. Phys., 100, 8471. Hopstaken, M.J.P. and Niemantsverdriet, J.W. (2000) J. Phys. Chem., 104, 3058. Martinez-Arias, A., Soria, J. and Conesa, J.C. et al. (1995) J. Chem. Soc. Faraday Trans., 91, 1679. Haneda, M., Kintaichi, Y. and Hamada, H. (2002) Phys. Chem. Chem. Phys., 4, 3146. Lecomte, J.J., Granger, P., Leclercq, L. et al. (1999) Colloids Surf. A, 158, 241. Granger, P., Delannoy, L., Lecomte, J.J. et al. (2002) J. Catal., 207, 202. Noronha, F.B., Baldanza, M.A.S., Monteiro, R.S. et al. (2001) Appl. Catal. A, 210, 275. Kolli, T., Rahkamaa-Tolonen, K., Lassi, U. et al. (2005) Catal. Today, 100, 297. Gonzalez-Velasco, J.R., Guttierez-Ortiz, M.A., Marc, J.L. et al. (2001) Appl. Catal. B Environ., 33, 303. Zu, Z., Wan, C.Z., Lui, Y.K. et al. (1996) Catal. Today, 30, 83. Martínez-Arias, A., Fernández-García, M., Iglesias-Juez, A. et al. (2001) Appl. Catal. B Environ., 31, 51. Maunula, T., Ahola, J., Salmi, T. et al. (1997) Appl. Catal. B Environ., 12, 287. Harrison, B., Wyatt, M. and Gough, K.C. (1982) In Catalysis (C. Kemball, D.A. Dowden eds.), Vol. 5. The Royal Society of Chemistry, London, p. 127. Vartui, J.C. and Gonzalez, R.D. (1973) Ind. Eng. Chem. Proc. Res. Dev., 12, 265. Ault, J.W. and Ayen, R.J. (1971) AIChE J., 17, 265. Pérez-Hernández, R., Aguilar, F., Gómez-Cortès, A. et al. (2005) Catal. Today, 107–108, 175. Bera, P., Patil, K.C. and Jayaram, V. (2000) J. Catal., 196, 293. Taha, R., Duprez, D., Mouaddib-Moral, N. et al. (1998) In Catalysis and Pollution Control CaPoC 4. (N. Kruse, A. Frennet, J.M. Bastin eds.), Stud. Surf. Sci. Catal., 116, 549. Fan, J., Wu, X., Ran, R. et al. (2005) Appl. Surf. Sci., 245, 162. Overbury, S.H., Mullins, D.R., Huntley, D.R. et al. (1999) J. Catal., 186, 296. Daturi, M., Bion, N., Saussey, J. et al. (2001) Phys. Chem. Chem. Phys., 3, 252. Thomas, C., Gorce, O., Villain, F. et al. (2006) J. Mol. Cat. A, 249, 71. Thomas, C., Gorce, O., Fontaine, C. et al. (2006) Appl. Catal. B, 63, 201. Tanabe, T., Hatanaka, M., Tsuji, R. et al. (2002) R&D Review of Toyota CRDL, 37, 1. Cordoba, L.F., Flytzani-Stephanopoulos, M. and Montes de Correa, C. (2001) Appl. Catal. B Environ., 33, 25. Costa, C.N., Stathopoulos, V.N., Belessi, V.C. et al. (2001) J. Catal., 197, 350. Costa, C.N. and Efstathiou, A.M. (2004) J. Phys. Chem. B., 108, 2620. Epling, W.S., Campbell, L.E., Yezerets, A. et al. (2004) Catal. Rev., 46, 163. Nakatsuji, T., Ruotoistenmäki, J., Komppa, V. et al. (2002) Appl. Catal. B Environ., 38, 101. Kolli, T., Lassi, U., Rahkamaa-Tolonen, K. et al. (2006) Appl. Catal. A Gen., 298, 65. Liotta, L.F., Macaluso, A., Arena, G.E. et al. (2002) Catal. Today, 75, 439. Eberhardt, M., Riedel, R., Göbel, U. et al. (2004) Top. Catal., 30/31, 135. Philipp, S., Drochner, A., Kunert, J. et al. (2004) Top. Catal., 30/31, 235. Corbos, E.C., Elbouazzaoui, S., Courtois, X. et al. (2007) Top. Catal., 42/43, 9. Lin, H.-Y., Wu, C.-J., Chen, Y.-W. et al. (2006) Ind. Eng. Chem. Res., 45, 134. Machida, M., Kurogi, D. and Kijima, T. (2003) Catal. Today, 84, 201.

This page intentionally left blank

Chapter 9

ASPECTS OF CATALYST DEVELOPMENT FOR MOBILE UREA-SCR SYSTEMS – FROM VANADIA-TITANIA CATALYSTS TO METAL-EXCHANGED ZEOLITES O. Kröcher∗ ∗

Paul Scherrer Institute, Villigen, Switzerland Corresponding author: Paul Scherrer Institute, Villigen, Switzerland. E-mail: [email protected]

Abstract Today, urea-SCR is the most efficient process to reduce the nitrogen oxide emissions of diesel vehicles. However, a variety of difficulties had to be overcome in order to adopt the established ammonia-SCR process for the reduction of NOx from stationary power plants, which is characterized by simple steady-state conditions, sufficiently high temperatures and rather low space velocities to diesel vehicles with its highly dynamic operating conditions and space restrictions, requiring a broadening of the operational window of the process as well as improvements of the low-temperature activity, the high-temperature stability and the volumetric activity of the catalysts. The main challenges on this stony way are addressed, stressing the development of suitable catalysts, which are compatible with these demanding operation conditions.

1. INTRODUCTION The selective catalytic reduction of NOx by ammonia (ammonia-SCR) is an established technology since the 1970s to control the nitrogen oxide emissions of fossil fuel power plants and chemical processes [1–3]. It has been successfully adopted to stationary diesel engines by substitution of the poisoning and irritating ammonia by urea, which is much easier to handle in a small scale [4–9]. Already 15 years ago, the idea was discussed to apply urea-SCR also on mobile diesel engines, which is now an emerging technology for the NOx reduction from heavy-duty diesel vehicles [10–15]. The scope of this chapter is to touch on important challenges on the way from stationary to mobile SCR systems [15] and to exemplarily show how they can be mastered. Special focus lies on the development of suitable catalysts, which is the key factor for the successful implementation of the SCR process not only for heavy-duty but also for light-duty diesel vehicles and passenger cars in the future [16–17]. Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Published by Elsevier B.V.

262

Past and Present in DeNOx Catalysis

1.1. Specific challenges for the development of mobile SCR systems SCR systems at stationary diesel engines profit from the high exhaust gas temperatures of about 350–400 C, caused by the usually constant high load operation conditions of the diesel engine. In this temperature window nearly all known SCR catalysts are very active. Moreover, weight and size of the exhaust gas catalyst are usually not strictly limited, which results in a good NOx reduction efficiency (DeNOx ). However, DeNOx is not the only criterion for an SCR catalyst. Further requirements are excellent selectivities regarding NOx and urea/ammonia as well as low ammonia slip, which is an undesired secondary emission of the SCR process. Therefore, all SCR catalysts exhibit surface acidity, which is necessary to store ammonia on the catalyst surface and, thus, to prevent ammonia slip. At a mobile diesel engine in a truck or a passenger car, the operating conditions are far from being constant. Engine load and speed vary often and abruptly, which directly changes the volumetric flow and temperature of the exhaust gas as well as the NOx emissions. These changes have to be considered by a real-time computer model, which calculates the amount of urea currently required. Besides these parameters the temperature dependency of the DeNOx activity has to be factored in the control strategy. At low exhaust gas temperatures, the catalytic activity usually drops, which is particularly critical if the catalyst volume is restricted as it is typical for a mobile aftertreatment system. Diesel engines are characterized by lower exhaust gas temperatures than gasoline engines and larger exhaust gas volume flows, which further aggravate this problem. In gasoline engines, the three-way catalyst is established, for which space velocities (GHSV) of about ≈100 000 h−1 are usually applied. From this number it is obvious that the typical space velocities for SCR catalysts of stationary power plants in the order of 10 000 h−1 had to be highly increased. This could only be achieved by significant improvements of the volumetric activity of the catalyst.

2. GENERATION OF THE REDUCING AGENT AMMONIA FROM UREA For reasons of safety and toxicity, urea is the preferred selective reducing agent for mobile SCR applications. Under the hydrothermal conditions in the exhaust system, urea decomposes to ammonia which reduces the nitrogen oxides on the surface of the SCR catalyst [18,19]. If urea is used instead of ammonia, the DeNOx chemistry involves isocyanic acid as an important intermediate which will lead to a complication of the SCR chemistry [20]. Urea is usually applied as an aqueous solution. If urea solution is atomized into the hot exhaust gas stream, the first step is the evaporation of water from the droplets thus leading to molten urea: NH2 −CO−NH2 (aqueous) → NH2 −CO−NH2 (liquid) +xH2 O(gas) Pure urea will then heat up and decompose thermally according to: NH2 −CO−NH2 (liquid) → NH3 (gas) + HNCO(gas)

Catalyst Development for Mobile Urea-SCR Systems

263

Equimolar amounts of ammonia and isocyanic acid are thus formed. These two endothermic processes, which can be slightly accelerated by catalysts, partially occur already in the gas phase ahead of the SCR catalyst. Investigations of the selective noncatalytic reduction showed that the mixture of ammonia and isocyanic is stable up to 850 C [21]. The final step is the hydrolysis of HNCO to another NH3 molecule and CO2 : HNCOgas + H2 Ogas → NH3 gas + CO2 gas The slightly exothermic hydrolysis reaction is kinetically hindered, but proceeds rapidly on many single and mixed metal oxides [19,22–25]. This means that in practice HNCO is usually hydrolyzed on the SCR catalyst itself, which has to be big enough in order to fulfill the double function [19].

2.1. Utilization of a hydrolysis catalyst The major problems with the substitution of the reducing agent ammonia for urea are on the one hand the homogeneous mixing of urea and exhaust gas and on the other hand the limited residence time in SCR systems for the different decomposition steps, i.e. the evaporation of water from the droplet, the thermolysis of urea to isocyanic acid and the following hydrolysis to ammonia [18]. Figure 9.1 shows the long distance required for a homogenization of the exhaust gas flow without a mixing unit. In this early experiment, urea solution was sprayed into the hot exhaust gas 85 and 350 cm upstream of the SCR catalyst (Figure 9.2). In case of the short distance, a very uneven distribution of reducing agent was achieved over the cross-section of the catalyst entrance. And in case of the long injection distance, the even distribution of reducing agent could be more ascribed to the mixing effect of the pipe bends than the distance. The mixing of urea spray and exhaust gas could be enhanced by the modification of the nozzle geometry to produce a broader spray cone, by the variation of the injection angle to induce turbulences or by an increase of the injection pressure to reduce the droplet size, which resulted in a homogeneous SCR performance at a distance of only 35 cm between injection nozzle and catalyst entrance.

Deviation (%)

60 40 20 0 –20 –40 –60 0

5

10

15

20

25

Catalyst monolith diameter (cm)

Figure 9.1. Spatial distribution of the reducing agent depending on the position of the injection nozzle. Distance injection nozzle–catalyst entrance: () 85 cm, and () 350 cm.

264

Past and Present in DeNOx Catalysis

Turbocharger Urea injection distance: 3.5 m

Urea injection distance: 85 cm SCR catalyst

45

40

20

Figure 9.2. Scheme of the experimental set-up for the investigation of the urea spray distribution.

Also the thermohydrolysis of the urea solution after the injection into the hot exhaust gas upstream of the SCR catalyst has been investigated at the diesel test rig. Urea solution was atomized about 3 m upstream of the SCR catalyst into the hot exhaust equivalent to a residence time in the pipe section of ≈01 s at 440 C. As expected for the thermolysis reaction, ammonia and isocyanic acid were found at the catalyst entrance at all temperatures (Figure 9.3). The ∼1:1 ratio of both components shows that only the thermolysis but not the hydrolysis is taking place in the gas phase. It can also be seen that the residence time of 0.1 s is not sufficient for the quantitative thermolysis of urea, as appreciable amounts of undecomposed urea were always found. The urea share even raises with lowering the flue gas temperature, although the residence time

100.0 90.0

urea

N-fraction (%)

80.0 70.0 60.0 50.0

HNCO

40.0 30.0 20.0

NH3

10.0 0.0

255 °C

290 °C

330 °C

400 °C

440 °C

Temperature (°C)

Figure 9.3. Ammonia, isocyanic acid and urea at various temperatures at the catalyst entrance. Residence time urea injection–catalyst entrance: 0.09 s at 440 C.

Catalyst Development for Mobile Urea-SCR Systems

265

was extended due to the lower gas volume flows. At the lowest temperature of 255 C, 85% of the total reducing nitrogen entered the catalyst in the form of urea. Therefore, a considerable section of the catalyst will be misused for the purely thermal steps of water evaporation and thermolysis of urea. The hydrolysis of HNCO and the SCR reaction can only begin further downstream. This means that the SCR catalyst is not efficiently used for these two reactions, thus, increasing the slip of HNCO and NH3 . In the case of an automotive SCR system, the time available from urea injection to the catalyst entrance will be even much smaller due to the desired compactness of such a system. Therefore, the sometimes low performance of SCR catalysts is probably less due to slow hydrolysis of HNCO, but more due to the delayed evaporation of water and the thermolysis of urea. Below 200 C, reliable urea thermohydrolysis is very hard to achieve, therefore urea dosage is usually stopped in real-world urea-SCR systems in this temperature regime. Another serious problem connected with the urea injection at low temperatures is the formation of white to yellowish deposits, which are observed when urea solution is injected at very low exhaust gas temperatures or if the urea spray forms a thick film at the walls of the SCR system. The analysis of these deposits [26] showed that they mainly consist of urea and some biuret at low temperatures and of cyanuric acid and some biuret at higher exhaust gas temperatures around 350 C. From laboratory investigations of the urea decomposition, it is known that biuret is easily formed from 150 to 190 C [27], whereas the formation of cyanuric acid is predominant from 200 to 300 C, according to the following reactions [12]: O

O +

H2N

NH2

N

C

H2N

H

Urea

O

O

Isocyanic acid

N H Biuret

NH2

O HN N

3

C

NH

O

H Isocyanic acid

O

N O H Cyanuric acid

Moreover, triuret, ammeline, ammelide, melamine and other products may be formed from isocyanic acid, biuret and combinations of them. If urea is heated up very fast, these reactions are suppressed and the decomposition into ammonia and isocyanic acid is the preferred reaction. Due to the high reactivity of isocyanic acid, its primary formation may subsequently lead to the formation of the aforementioned compounds of higher molecular weight. In order to avoid the formation of by-products, the heating-up must be carried out fast. Only then ammonia and isocyanic acid are obtained as sole products. In any case, local undercooling of the gas duct should be avoided and rapid dilution of the thermolysis products in the exhaust gas has to be ensured in order to avoid locally high concentrations of reactive compounds. The limited space available for the SCR system onboard of diesel vehicles had lead to the idea of using a special hydrolysis catalyst in front of the SCR catalyst [23,28]

266

Past and Present in DeNOx Catalysis

in order to quantitatively decompose urea to ammonia before it enters the SCR catalyst and to relieve the SCR catalyst from the burden of coping with both urea decomposition and SCR reaction. Among the different steps of the urea decomposition, i.e. water evaporation, urea thermolysis and HNCO hydrolysis, the last step is mostly accelerated by action of a catalyst. The hydrolysis proceeds very rapidly on a number of oxides such as silica, alumina, and titania, but also on typical SCR catalysts based on V2 O5 /WO3 −TiO2 and Fe-ZSM5 [19,22–25]. Interestingly enough, the rate of HNCO hydrolysis is very high on a typical SCR catalyst and lies nearly at the level of the gas phase diffusion rate of HNCO (Figure 9.4) [18]. Based on the excellent activity of titania for the hydrolysis of HNCO, this material is preferably used as coating for hydrolysis catalysts. Although higher HNCO conversions were observed on pure TiO2 than on a typical SCR catalyst, only a marginal effect on SCR performance can be expected by adding a special hydrolysis catalyst upstream of the SCR catalyst as HNCO hydrolysis is not the rate determining step of the over-all process. However, a significant positive effect is determined in practical SCR systems when using a hydrolysis catalyst [28–29], which must be traced back to other reasons. In practice hydrolysis catalysts are coated on special structured metal substrates [30], on which the urea solution is sprayed. Different from conventional honeycomb structures, these corrugated structures induce a high degree of turbulence in the flow [30], thus increasing the heat and mass transport required for the fast evaporation of water and the thermolysis of urea. Subsequently, the hydrolysis can proceed on the catalytic coating. One advantage of using titania compared to an SCR catalyst for hydrolysis is the much higher mechanical stability of a titania coating [28]. The permanent strike of urea droplets on the hydrolysis catalyst is extremely abrasive. SCR catalysts optimized with binders, pore forming agents, etc. can hardly withstand these harsh conditions. Furthermore, these

10000

kmass [cm3/g*s]

EA ≈ 8.1 kJ/mol

EA ≈ 12.7 kJ/mol 1000 D.HNCO*10000 hydrolysis of HNCO EA ≈ 85 kJ/mol

SCR reaction

150°C 250 200 500 450 400 350 300 100 0.0012 0.0014 0.0016 0.0018 0.0020 0.0022 0.0024

0.0026

1/ T [1/K]

Figure 9.4. Uncorrected reaction rate constants of the hydrolysis of HNCO and of the SCR reaction over the V2 O5 /WO3 −TiO2 powder catalyst [13].

Catalyst Development for Mobile Urea-SCR Systems

267

special structured substrates are very effective in the mixing and homogenization of the gas flow, which is necessary to achieve uniform concentrations and space velocities in every single channel of the following SCR catalyst [29].

3. VANADIA-BASED SCR CATALYSTS 3.1. Standard-SCR activity The simplest variant of the selective catalytic reduction of NOx with NH3 is the ‘standardSCR’ reaction, in which NH3 and NO comproportionate in a 1:1 stoichiometry to nitrogen. This reaction is efficiently catalyzed with high activity and selectivity between 300 and 400 C by V2 O5 /WO3 −TiO2 catalysts, which are wide-spread in stationary SCR systems [1]. 4NH3 + 4NO + O2 → 4N2 + 6H2 O ‘Standard-SCR’ is also the basic reaction over SCR catalysts in diesel vehicles, as more than 90% of the NOx in diesel exhaust is composed of NO under usual operation conditions. The primary aim of the development of SCR catalysts for mobile applications was to increase the intrinsic activity of the standard V2 O5 /WO3 −TiO2 catalysts for stationary SCR systems. In order to improve the activity of the catalyst, the vanadia concentration can be elevated from around 1% in the catalyst for stationary applications to around 2% or higher (Figure 9.5a). However, above a V2 O5 concentration of 2%, the increase in DeNOx is rather small and is accompanied with a lower temperature stability, which is revealed by the reduced DeNOx for the sample with 3% V2 O5 after ageing at 600 C (Figure 9.5b). On the contrary, the thermal treatment at 600 C even increased the activity (b) 100

100

80

80

DeNOx (%)

DeNOx (%)

(a)

60 40 20 0 150

60 40 20

250

350

450

Temperature (°C)

550

0 150

250

350

450

550

Temperature (°C)

Figure 9.5. DeNOx at 10 ppm ammonia slip in dependency of the vanadia concentration under standard-SCR conditions. Commercially available SCR catalysts based on extruded V2 O5 /WO3 −TiO2 with () 3%, () 1.9% and (•) 1.7% V2 O5 . ‘Extruded V2 O5 /WO3 −TiO2 ’ means that the whole monolith consists of the active mass V2 O5 /WO3 −TiO2 , which is enforced by glass fibers and binders. Cell density: ≈400 cpsi. (a) Thermal treatment at 550 C for 50 h. (b) Thermal treatment at 550 C for 50 h and at 600 C for 30 h. Model gas investigation with 10% O2 , 5% H2 O, 1000 ppm NO, 0–1500 ppm NH3 , balance N2 . Vcat = 75 cm3 . GHSV = 52 000 h−1 .

268

Past and Present in DeNOx Catalysis

of the samples containing 1.7 and 1.9% V2 O5 , as the formation of the active phase is completed by spreading of the V2 O5 on the titania support. Three different phenomena have been identified upon ageing, which are important for the understanding of the catalytic behavior: three-dimensional growth of the supported vanadia, increase of two-dimensional polymeric vanadyl surface species and anatase sintering [31]. The observed improvement of the SCR performance of the catalysts containing up to 2% V2 O5 upon ageing was attributed to an increase of the amount of two-dimensional polymeric vanadyl surface species, whereas the activity decrease of the catalyst containing 3% is caused by an extensive loss of surface area and due to the three-dimensional growth of supported vanadia.

3.2. Selectivity The term ‘selective catalytic reduction’ implies that the reaction of NOx with NH3 or urea is considered to be highly selective. However, in practice, a number of possible side-reactions have to be taken into account. On the one hand, regarding the oxidizing reactants, the reduction of NOx must prevail over O2 , which is present in excess in lean exhaust gases. On the other hand, the reaction shall exclusively produce nitrogen and not higher oxidized nitrogen species such as N2 O, NO and NO2 . Thus, when discussing the SCR reaction ‘reactant-’ as well as ‘product-selectivity’ have to be kept in mind. From Figure 9.6, the excellent selectivity of V2 O5 /WO3 −TiO2 catalysts is clearly discernible, almost independent from the vanadia concentration. Over a broad temperature range up to 400 C, a 1:1 stoichiometry is observed for ammonia and nitrogen oxides in accordance with the SCR reaction equation. At higher temperatures, ammonia is consumed in excess, which is more pronounced for higher vanadia concentrations. This over-consumption is caused by the oxidation of ammonia to nitrogen, N2 O or even

(b) NH3 /NOx Stoichiom. (–)

NH3 /NOx Stoichiom. (–)

(a) 2.0 1.8 1.6 1.4 1.2 1.0 0.8 150

250

350

450

Temperature (°C)

550

2.0 1.8 1.6 1.4 1.2 1.0 0.8 150

250

350

450

550

Temperature (°C)

Figure 9.6. Compliance of the SCR stoichiometry in dependency of the vanadia concentration under standard-SCR conditions. Extruded V2 O5 /WO3 −TiO2 catalysts with () 3%, () 1.9% and (•) 1.7% V2 O5 . (a) Thermal treatment at 550 C for 50 h. (b) Thermal treatment at 550 C for 50 h and at 600 C for 30 h.

Catalyst Development for Mobile Urea-SCR Systems (b) 100

100

80

80

N2O (ppm)

N2O (ppm)

(a)

269

60 40

40 20

20 0 150

60

250

350

450

550

0 150

Temperature (°C)

250

350

450

550

Temperature (°C)

Figure 9.7. N2 O formation at 10 ppm ammonia slip in dependency of the vanadia concentration. Extruded V2 O5 /WO3 −TiO2 catalysts with () 3%, () 1.9% and (•) 1.7% V2 O5 . (a) Thermal treatment at 550 C for 50 h. (b) Thermal treatment at 550 C for 50 h and at 600 C for 30 h.

NO according to the following formal reaction equations, thus limiting the maximum NOx conversion: 4NH3 + 3O2 → 2N2 + 6H2 O 2NH3 + 2O2 → N2 O + 3H2 O 4NH3 + 5O2 → 4NO + 6H2 O The production of N2 O is shown in Figure 9.7. It should be noticed that on V2 O5 /WO3 −TiO2 N2 O is not only formed by ammonia oxidation (refer Chapter 1) but also due to the following reaction [32,33] in analogy to the SCR reaction: 4NH3 + 4NO + 3O2 → 4N2 O + 6H2 O More than 2% of V2 O5 should be avoided, because higher vanadia concentrations result in elevated N2 O emissions, which again increase after ageing at 600 C due to the three-dimensional growth of unselective crystalline V2 O5 [31]. Due to the trade-off between low-temperature activity on the one hand and selectivity as well as stability on the other hand, a compromise has to be found for the vanadia concentration. Most extruded commercial SCR catalysts for mobile applications contain between 1.7 and 1.9% V2 O5 , but there is a trend to further reduce the vanadia concentration in order to suppress the N2 O formation at higher temperatures and to increase the temperature stability of the catalyst.

3.3. Catalyst development The direct way to improve the volumetric activity of monolithic catalysts is to increase its cell density, by which the volumetric surface is tremendously enlarged [15,34]. Figure 9.8 shows that an increase in cell density from 300 to 400 cpsi is more effective on the DeNOx performance than an increase of the vanadia concentration from 1.9 to 3%, which is, moreover, accompanied by the above mentioned problems.

270

Past and Present in DeNOx Catalysis 100

DeNOx (%)

80 60 40 20 0 150

250

350

450

550

Temperature (°C)

Figure 9.8. Influence of cell density on DeNOx at 10 ppm ammonia slip. Extruded V2 O5 / WO3 −TiO2 catalysts with ( ) 3%, and ( ) 1.9% V2 O5 . ( ) 400 cpsi, and ( ) 300 cpsi. Thermal treatment at 550 C for 50 h.

Extruded SCR catalysts based on V2 O5 /WO3 −TiO2 with cell densities of 300 cpsi are commercialized and 400 cpsi prototypes are available. ‘Extruded SCR catalyst’ means that the whole monolith consists of the active mass V2 O5 /WO3 −TiO2 , which is enforced by glass fibers and binders. As this extrusion technology is limited by the mechanical properties of the extruded active mass, much higher cell densities can only be realized by coating ceramic or metallic substrates. Especially, the metallic substrates can be produced with a very low wall thickness down to 30 m, facilitating structures with a cell density of more than 1000 cpsi. Coated SCR catalysts with a cell density of 400 cpsi are standard and research samples with up to 800 cpsi have been tested. These SCR catalysts exhibit a further increased SCR performance compared to its 400 cpsi analogues, but at 600 cpsi the maximum activity should be reached, because gas phase diffusion should not be limiting any more. Moreover, if the cell density is increased at constant wall thickness, the pressure drop rises due to a reduced open frontal area. The effect of a higher cell density is pronounced for temperatures above 350 C. At low temperatures (below ≈300 C), the active mass-to-volume ratio is decisive for the attainable DeNOx , thereby favoring extruded catalysts with their high content of active mass [35]. For coated V2 O5 /WO3 −TiO2 catalysts, a thickness of 20 m appeared to be an optimum compromise between low temperature activity, pressure drop and stability of the coating. Due to the absence of coated SCR catalysts on the market in the 1990s, a vanadia-based SCR catalyst was developed in combination with a coating process at the Paul Scherrer Institute, which posed the state-of-the-art for coated SCR catalysts at that time (Table 9.1). In order to ensure optimal sticking of the catalyst layer colloidal silica was used as binder, resulting in a quaternary V2 O5 /WO3 −SiO2 −TiO2 system. It should be pointed out that the amount of active mass is only rate determining if urea is completely decomposed to ammonia by a hydrolysis catalyst in front of the SCR catalyst. Usually, the urea spray is captured and thermohydrolyzed by the SCR catalyst itself. In this case, not the active mass-to-volume ratio but the capability of the SCR catalyst to capture the urea droplets is decisive for the DeNOx activity [19]. Higher cell densities tremendously enhance the probability to capture the urea droplets near the catalyst entrance.

Catalyst Development for Mobile Urea-SCR Systems

271

Table 9.1. Preparation of V2 O5 /WO3 −SiO2 −TiO2 SCR catalysts coated on metal and cordierite substrates Catalyst NH4 VO3 solution + colloidal silica + 10%WO3 /TiO2 optional: + pore building agent ⇒ Suspension Coating: Slow immersion of substrates in suspension for 10–20 s Blow-off of excess suspension Gentle drying of the module with a hot-air drier (90 C) Weighing of module and repetition of coating procedure until 180–200 mgcat /mlsubstrate are reached (in case of a pore building agent: +10% catalyst weight) Calcination: 50 h at 550 C

In general, for the design of a catalyst with given volume, a compromise between diameter and length has to be found, taking into account the Reynolds number in the channels, pressure drop, cell density, flow distribution at the catalyst entrance and heat management. In case of a short monolith, low Reynolds numbers, a low pressure drop and high cell densities are possible, but a homogeneous flow distribution is difficult to produce and urea droplets might slip through the catalyst. On the contrary, flow homogenization is easier in front of a thin monolith but the pressure drop is considerably higher.

3.4. The fast-SCR reaction The most effective way to improve the SCR performance is to increase the rate of reaction itself by the help of NO2 [28,36,37]. When equimolar amounts of NO and NO2 are used, a very high DeNOx is observed (‘fast-SCR’ reaction), as the very potent oxidizing agent NO2 reoxidizes the catalyst much faster than oxygen [38]: 4NH3 + 2NO + 2NO2 → 4N2 + 6H2 O In Figure 9.9, the SCR performance of a coated V2 O5 /WO3 −TiO2 catalyst is plotted for NO in the feed and for a 1:1 mixture of NO + NO2 . The largest effect is observable at low temperatures, e.g. at 200 C only ≈24% of DeNOx is obtained with pure NO but ≈93% with the NO + NO2 mixture. By plotting ammonia slip versus DeNOx , the two most important properties of an SCR catalyst are combined in one graph. Starting with a constant amount of NOx at

272

Past and Present in DeNOx Catalysis (b)

100

100

80

80

NH3 slip (ppm)

NH3 slip (ppm)

(a)

60 40 20

60 40 20 0

0 0

20

40

60

80

0

100

20

40

60

80

100

DeNOx (%)

DeNOx (%)

Figure 9.9. Performance of a coated V2 O5 /WO3 −TiO2 SCR catalyst developed in our laboratory (Table 9.1) with (a) pure NO and with (b) NO:NO2 = 1  1. (♦) 120 C, () 150 C, (∗ ) 180 C, () 200 C, (•) 250 C, () 300 C, () 350 C, () 400 C, and () 450 C. Cell density: ≈400 cpsi. Vcat = 75 cm3 . Model gas investigation with 10% O2 , 5% H2 O, 1000 ppm NO or 500 ppm NO + 500 ppm NO2 , 0–1500 ppm NH3 , and balance N2 in a laboratory test unit [13]. High Resolution FTIR gas analysis [13]. GHSV = 52 000 h−1 .

a fixed temperature, ammonia is dosed in increasing amounts. First, all ammonia is consumed in the SCR reaction and the curve follows the x-axis. With increasing ammonia concentrations, the maximum conversion is approached which is limited by the activity of the catalyst. Near the activity limit, more and more ammonia is emitted and the curve is bended upwards. The almost rectangular curve shape in Figure 9.9 is a consequence of the high SCR activity of V2 O5 /WO3 −TiO2 combined with its surface acidity. This helps to withdraw the ammonia in the catalyst and to provide sufficiently high ammonia surface concentrations at already low ammonia feed gas concentrations. Thus, excessive ammonia dosage only slightly increases the ammonia concentration at the active sites and DeNOx remains almost constant, even in the case of ammonia overdosage. Usually, ammonia emissions of about 10 ppm in average are regarded as harmless for automotive applications [13]. The influence of the NO2 /NOx feed ratio on the NOx removal efficiency is depicted in Figure 9.10 for different temperatures [37]. Below 300 C, DeNOx always increases linearly with increasing NO2 fractions up to 50%, for which the highest DeNOx was

DeNOx (%)

100 80 60 40 20 0 0

20

40

60

80

100

NO2,in /NOx ,in (%)

Figure 9.10. Influence of NO/NO2 ratio on the DeNOx activity of a coated V2 O5 /WO3 −TiO2 SCR catalyst. () 200 C, () 250 C, () 300 C, (•) 350 C.

Catalyst Development for Mobile Urea-SCR Systems

273

observed. With NO2 in the feed, the fast-SCR reaction proceeds at up to ten times higher rates than the standard-SCR reaction with NO. As soon as all NO2 along with the same amount of NO is consumed according to the 1:1 stoichiometry of the fast-SCR reaction, the remaining NO reacts with ammonia according to the standard-SCR reaction. For NO2 /NOx ratios larger than 50%, a nearly mirrored behavior is observed. Again, NO and NO2 react in the ratio 1:1 in the fast-SCR reaction, but in this case NO2 is left over, which is consumed in the ‘NO2 -SCR’ reaction [37]. 4NH3 + 3NO2 → 31/2N2 + 6H2 O The decline of the DeNOx curve for NO2 fractions above 50% is much stronger than the incline below 50% due to the different reaction rates of standard- and NO2 -SCR. The latter is much slower than the fast-SCR reaction and even slower than the standard-SCR reaction. The promoting effect of NO2 levels off above 350 C, because the rate constants of standard-, fast- and NO2 -SCR reactions are converging at higher temperatures. Another reaction between NH3 and NO2 has to be considered, especially at low temperatures. NH3 and NO2 form ammonium nitrate (NH4 NO3 ), which deposits on the catalyst in solid or liquid form at T ≤ 180 C (m.p. 170 C), thereby reducing the DeNOx performance (Figure 9.11) [32,33,37–41]. 2NH3 + 2NO2 → NH4 NO3 ↓ +N2 + H2 O Additional experiments have shown that the formation of ammonium nitrate is a reversible process due to the observation that the catalyst recovers its original activity when heated at temperatures above 200 C [33,41]. NH4 NH3 solid ↔ NH3 + HNO3

100 T = 200°C T = 190°C

80

DeNOx (%)

T = 180°C

60 T = 170°C

40 T = 160°C

20

T = 150°C

0 5

15

25

35

45

55

Time (min)

Figure 9.11. Poisoning of a coated V2 O5 /WO3 −TiO2 SCR catalyst at various temperatures due to the formation of ammonium nitrate. Cell density ≈400 cpsi. Vcat = 75 cm3 . Model gas: 10% O2 , 5% H2 O, 500 ppm NO + 500 ppm NO2 , 0–1500 ppm NH3 and balance N2 . GHSV = 52 000 h−1 .

274

Past and Present in DeNOx Catalysis NH3 2NO2

N2O4

H2O

HNO2

[NH4NO2]

+

½NO 3/2 NO2

HNO3

V2O5/WO3 —TiO2

N2 + 2H2O

½H2O

NH3 NH4NO3

Figure 9.12. Reaction scheme explaining the NO/NO2 -SCR chemistry over V2 O5 /WO3 − TiO2 [32,41].

The NO in the diesel exhaust gas will react with the formed HNO3 to yield NO2 . 2HNO3 + NO → 3NO2 + H2 O This deposition of ammonium nitrate on the SCR catalyst is in full analogy with the well-known deposition of ammonium sulfates [(NH4 2 SO4 and NH4 HSO4 ] on SCR catalysts. The NO/NO2 -SCR chemistry on a typical V2 O5 /WO3 −TiO2 catalyst was revealed by Koebel et al. and is summarized in the reaction scheme depicted in Figure 9.12 [32,41]. Please note that the sum of the reactions in Figure 9.12 results in the reaction equation of the ‘fast-SCR’ reaction. Recent results of Nova et al. indicate that the reduction of HNO3 by NO might be better formulated as [39,40]: HNO3 + NO → HNO2 + N2 + H2 O Subsequently, the formed HNO2 undergoes the same reactions via ammonium nitrite to nitrogen as shown in the upper part of Figure 9.12. More details about the SCR chemistry can, for example, be found in [32,33,37–41]. In typical diesel exhaust gas, the NO2 /NOx fraction is only 5–10%. This fraction may be increased by passing the gas over a strong oxidation catalyst containing platinum as the active component. However, it is difficult to obtain useful fractions of NO2 at temperatures below 200 C at high space velocities due to the strong temperature dependency of the NO oxidation over platinum [42]. As expected the NO conversion rises exponentially with temperature, but declines at higher temperatures due to the thermodynamic limit of the reaction (Figure 9.13). The oxidation catalyst should not be oversized. Otherwise, too much NO2 is produced at intermediate temperatures and this will lead to a decrease of the reaction rate of the SCR reaction due to the low rate of the NO2 -SCR reaction. A way out of the problem would be a kind of controlled oxidation catalyst producing ≈50% NO2 over the entire temperature range [43]. Despite these limitations, the application of an oxidation catalyst has a strong positive effect on the SCR performance (Figure 9.14) [14].

Catalyst Development for Mobile Urea-SCR Systems

275

NO2 / NOx (%)

100 80 60 40 20 0 100

200

300

400

500

600

Temperature (°C)

Figure 9.13. Conversion of various concentrations of NO to NO2 in a humid feed on Pt/SiO2 as a function of temperature [42]. Feed: 10% O2 , 5% H2 O, and 100–1500 ppm NO in N2 . () 100 ppm, () 500 ppm, () 1000 ppm, and (•) 1500 ppm NO. Sample weight = 0.8 g. V∗ = 150 LN /h. At intermediate temperatures a conversion maximum is found, because the increasing oxidation activity of the catalyst is limited by the (---) thermodynamic equilibrium between NO and NO2 .

DeNOx (%)

100 80 60 40 20 0

0

20 40 185 225 260 300 335

60

80 405

100 465

Load (kWel)/T Cat (°C)

Figure 9.14. Influence of an oxidation catalyst on the DeNOx activity, measured at the Diesel test rig ‘HARDI’ of the Paul Scherrer Institute (PSI) [12,14]. 6.64 l four-cylinder turbocharged diesel engine with intercooler (Liebherr D924 TI-E A2). Steady-state experiments at a fixed engine speed of 1500 rpm and different engine loads. Extruded V2 O5 /WO3 −TiO2 SCR catalyst from Frauenthal (≈3% V2 O5 , Vcat = 96 l, 300 cpsi). Oxidation catalyst V09 from OMG (90 g ft−3 Pt, Vcat = 19 l, 400 cpsi). () SCR catalyst + oxidation catalyst. () SCR catalyst.

3.5. Ammonia slip Good SCR systems combine high NOx conversions with low ammonia slip, which is an undesired secondary emission. Therefore, SCR catalysts have to be designed such that ammonia is converted to a high degree in the catalyst module and that the residual unconverted ammonia is withheld on the catalyst surface. For this reason, all highly active SCR catalysts have acidic surfaces, which bond ammonia and provide high local ammonia concentrations at the active SCR sites. However, the storage capacity of the acidic SCR catalysts for ammonia depends on the catalyst temperature, as depicted in Figure 9.15 for an extruded and a coated catalyst [44,45]. Measured with the same catalyst volume, the extruded SCR catalyst adsorbs much more ammonia than the coated

3.5

2.57

K50 (coated catalyst)

2.5

2

0

200

250

300

350

400

0.25

0.19

0.5

0.28

0.29

0.60

1

0.34

1.5

1.24

1.37

NH3 desorbed (mg NH3)

3

D41 (extruded catalyst)

2.28

4

3.35

Past and Present in DeNOx Catalysis 11.31

276

450

T (°C)

Figure 9.15. Comparison of the total ammonia adsorption of coated and extruded V2 O5 /WO3 −TiO2 catalysts. Catalyst volume = 7 cm3 . Model gas for loading: 10% O2 , 5% H2 O, NH3 = 1000 ppm, and balance N2 . GHSV = 52 000 h−1 . Model gas for temperature-programmed desorption (TPD) experiment: 10% O2 , 5% H2 O, NO = 1000 ppm, NH3 = 1000 ppm, and balance N2 . NH3 desorbed is calculated as the sum of thermally desorbed NH3 , directly measured at the catalyst outlet, and chemically desorbed NH3 , measured by the reduction of the NO concentration due to the SCR reaction.

catalysts due to its higher content of active mass. As the exhaust gas temperatures vary from ≈100 C at idle up to ≈500 (600) C at full load, the consideration of the ammonia storage capacity is decisive for the quality of the urea dosing strategy. In Koebel and Elsener [35], it was shown that at low temperatures, the entire catalyst is used in the SCR reaction; whereas at high temperatures, only the utmost part of the catalyst layer is involved in the SCR reaction. Consequently, extruded catalysts are less effectively used under these conditions than coated SCR catalysts. In addition to the storage capacity of the SCR catalyst, the dynamics of the ammonia storage and release has to be considered in a urea dosing strategy. This is exemplarily shown in Figure 9.16 for a fast load increase [13,45]. If the urea dosing is simply increased according to the load jump, a large ammonia peak is observed. However, the catalyst temperature increases much slower and parallel to it also increases the activity. The second curve shows how the ammonia emissions are reduced in this case. If the ammonia emissions downstream of the catalyst are measured, a feedback control can be realized to reduce the urea dosage by the amount of ammonia released, which further improves the result. With highly developed urea dosing strategies, over 90% DeNOx is possible at both low average ammonia slip and nearly no peak emissions of ammonia [14]. More information on how to develop a suitable urea dosing strategy and a control model can be found in references [46–50]. Even in the case of an ideal dosing strategy, ammonia slip cannot be excluded if the SCR catalyst is cold, undersized, deactivated or if load jumps are large and very fast. As a consequence, the addition of an oxidation catalyst downstream of the SCR catalyst has

Catalyst Development for Mobile Urea-SCR Systems

277

1000

Slip of NH3 + HNCO (ppm)

Dos = f(load) Dos = f(load, temperature) Dos = f(load, temperature) + closed feed back control 100

10 –5

0

5

10

15

20

25

30

Time from load increase (min)

Figure 9.16. Effect of varying dosing strategies on the ammonia and HNCO slip during a load increase from 25 to 75 kW, which corresponds to a temperature increase from 230 to 400 C.

been proposed in order to further reduce the ammonia emissions. Hug et al. [10] were the first who described such a two-catalyst SCR system for off-road vehicles. A much more sophisticated aftertreatment system was proposed by Jacob, using an ammonia oxidation catalyst in the so-called ‘VHRO’ system (Figure 9.17) [13,28,29]. Ammonia oxidation catalysts (sometimes called slip catalyst) are conventional oxidation catalysts based on precious metals. The most active types are based on Pt. Their activity is strongly dependent on the temperature and, thus, relatively large catalyst volumes are required for the ammonia oxidation below 250 C. At rising temperatures, their oxidation power increases and this leads to the formation of N2 O and NO. Especially undesired is their strong tendency to form N2 O at intermediate temperatures (250–300 C) [2] if the gas coming from the SCR catalyst also contains unreacted NO, which allows for the reaction: 4NH3 + 4NO + 3O2 → 4N2 O + 6H2 O

V

V-KAT H-KAT H-KAT O-KAT

2 NO2 +

H

R

O

O2

2 NO2

CO(NH2)2 HNCO + H2O O2 + 4 NH3 4 NO + 2 NO + 2 NO2 + 4 NH3

NH3 + HNCO NH3 + CO2

4 NH3 + 3 O2

2 N2 + 6 H2O

Figure 9.17. Scheme of the MAN VHRO system.

4 N2 + 6 H2O 4 N2 + 6 H2O

278

Past and Present in DeNOx Catalysis

4. ZEOLITE-BASED SCR CATALYSTS The forthcoming Euro5 and Euro6 emission limits for diesel vehicles necessitate the combination of urea-SCR systems for the reduction of NOx with diesel particulate filters (DPF) for the abatement of soot emissions. There are two possibilities for the regeneration of a DPF: continuous or thermal. The continuous regeneration utilizes NO2 for the oxidation of the soot particles and leads to the problem of the control of the NO/NO2 ratio entering the DPF and the SCR system depending on the operating conditions of the motor. For a thermal regeneration, the exhaust gas temperatures are increased up to 700 C to oxidize the soot with the residual oxygen in the exhaust gas. Such high temperatures were not yet considered for SCR systems. At present, V2 O5 / WO3 −TiO2 are used as SCR catalysts, which can resist at most 650 C for a short time before they deactivate due to the release of vanadia species and the phase transformation of TiO2 (anatase) into TiO2 (rutile) accompanied by a loss of BET surface. On the basis of the present standard of knowledge, vanadia emissions cannot be strictly excluded for SCR catalysts with high V2 O5 content near or above monolayer surface coverage (>2%), which pose a problem as health risks are discussed for vanadia [51]. Therefore, alternative SCR catalysts are required that contain only harmless substances and that also resist harsh operating conditions. Metal-exchanged zeolites are considered as potential alternative catalysts and especially Fe-ZSM5 has received much attention because it seems to fulfill these requirements [52–54]. Cu-ZSM5 was also tested as potential SCR catalyst, but its hydrothermal stability is insufficient for a high-temperature application.

4.1. Comparison of metal-exchanged zeolites and vanadia catalysts under standard-SCR conditions Figure 9.18 compares the catalytic performance of Fe-ZSM5 and Cu-ZSM5 on cordierite monoliths (Umicore AG) with V2 O5 / WO3 −TiO2 prepared and coated according to Table 9.1. It is discernible, that Cu-ZSM5 is much higher active for SCR than both Fe-ZSM5 and V2 O5 / WO3 −TiO2 at T ≤ 300 C and that it is also more active than 100

DeNOx (%)

80 60 40 20 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.18. DeNOx at 10 ppm NH3 slip versus temperature for 1000 ppm NO in the feed. () Cu-ZSM5 (Umicore AG), (•) V2 O5 /WO3 −TiO2 (PSI, see Table 9.1), and () Fe-ZSM5 (Umicore AG). Feed gas: 10% O2 , 5% H2 O, 1000 ppm of NO, variable amounts of NH3 and N2 . GSHV = 52 000 h−1 .

Catalyst Development for Mobile Urea-SCR Systems

279

V2 O5 / WO3 −TiO2 at the high temperature end. However, Fe-ZSM5 outperforms both V2 O5 / WO3 −TiO2 and Cu-ZSM5 at T > 550 C. Fe-ZSM5 seems to be especially suitable for high-temperature applications, whereas V2 O5 / WO3 −TiO2 is a good catalyst for a broad intermediate temperature range. On base of the SCR activity measurements alone, Cu-ZSM5 appears to be the best choice for the low and intermediate temperature region up to 550 C. However, besides the SCR activity, the selectivity is an important parameter for the assessment of the catalysts. In case of the SCR reaction, the selectivity with respect to both the product nitrogen and the reactant ammonia has to be considered. The product selectivity is important, as side products such as N2 O can be formed and the reactant selectivity is important, as ammonia can be converted to nitrogen not only in the SCR reaction but also by the selective catalytic oxidation with oxygen [54]. The ammonia selectivity for the three different catalysts is shown in Figure 9.19. The excellent selectivity of V2 O5 /WO3 −TiO2 as well as Fe-ZSM5 is obvious. Whereas V2 O5 /WO3 −TiO2 strictly obeys the 1:1 stoichiometry for NH3 and NOx within the measuring accuracy (1–2%), for Fe-ZSM5 a few percent overconsumption of ammonia is noticeable from 200 to 600 C, which steeply increases above this temperature window. However, this effect is low compared to V2 O5 /WO3 −TiO2 , which completely loses its ‘reactant-selectivity’ at very high temperatures. The vanadia-based catalyst is not any more acting as an SCR catalyst, but mainly as an oxidation catalyst at T ≥ 550 C. Over

(b) NH3,reacted /NOx , reacted (–)

NH3,reacted /NOx , reacted (–)

(a) 3.5 3.0 2.5 2.0 1.5 1.0 0.5

0.0

2.0

1.0

3.5 3.0 2.5 2.0 1.5 1.0 0.5

3.0

0.0

2.0

1.0

α (–)

3.0

α (–)

NH3,reacted /NOx , reacted (–)

(c) 3.5 3.0 2.5 2.0 1.5 1.0 0.5

0.0

2.0

1.0

3.0

α (–)

Figure 9.19. Compliance of the SCR stoichiometry for 1000 ppm NO in the feed. (a) V2 O5 / WO3 −TiO2 (b) Fe-ZSM5. (c) Cu-ZSM5. () 150 C, () 200 C, (•) 250 C, () 300 C, () 350 C, () 400 C, () 450 C, (♦) 500 C, (∗ ) 550 C, (—) 600 C, (---) 650 C, and (–––) 700 C.

280

Past and Present in DeNOx Catalysis 350

N2O (ppm)

300 250 200 150 100 50 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.20. N2 O formation at 10 ppm ammonia slip for 1000 ppm NO in the feed. () Cu-ZSM5, (•) V2 O5 /WO3 −TiO2 , and () Fe-ZSM5.

Cu-ZSM5, already above 250 C, a considerable amount of ammonia is not going into in the SCR reaction, which suggests that this catalyst is used more for low-temperature applications. Regarding the product selectivity, the N2 O formation over the different catalysts is most important (Figure 9.20). Negligible amounts of N2 O (∼3 ppm) were produced over Fe-ZSM5 up to 450 C and about 25 ppm of N2 O over Cu-ZSM5 up to 500 C, respectively. At higher temperatures, metal-exchanged zeolites showed no N2 O formation at all, unlike V2 O5 /WO3 −TiO2 , which starts to produce N2 O in this temperature regime. The decrease in the N2 O formation at T > 550 C over the vanadia system is due to the ammonia oxidation to nitrogen, which is getting the predominant reaction in this temperature region. Ammonia in principle can be oxidized to N2 , N2 O and NO according to the reactions given in Section 3.2 (and to NO2 as NO itself can be further oxidized to NO2 , depending on the oxidation potential of the catalyst) [54]. Besides the oxidation of ammonia to N2 O, which is measured, and the oxidation to N2 , which is calculated from the nitrogen balance, NO may be formed, which cannot be distinguished from the NO dosed. Finally, also the formation of NO2 has to be considered as a secondary product out of the NO oxidation. In fact, Cu-ZSM5 exhibits such a strong oxidation potential that considerable amounts of NO2 are produced at intermediate and high temperatures.

4.2. Ammonia oxidation For a better understanding of the side-reactions, the number of possible reaction pathways was reduced by dosing ammonia over the catalysts without NO in the feed. Due to the lack of NO and, thus, the absence of the SCR reaction, NH3 is oxidized to N2 over the catalysts at lower temperatures than in the presence of NO (Figure 9.21). Cu-ZSM5 proves its strong oxidizing properties for NH3 to N2 by reaching 100% conversion already at 400 C. Also V2 O5 /WO3 −TiO2 shows pronounced ammonia oxidation, but at an increased selectivity for NO and N2 O at T ≥ 450 C compared to Cu-ZSM5. Over Fe-ZSM5 ammonia is fully oxidized not below 700 C and also much less NO and N2 O are formed. Comparing Fe-ZSM5 and Cu-ZSM5, it can be stated that the oxidation

Catalyst Development for Mobile Urea-SCR Systems (a)

(b) 100

100

80

75

SNO (%)

XNH3 (%)

281

60 40

50 25

20

0

0 150

250

350

450

550

650

150

750

250

350

450

550

650

750

Temperature (°C)

Temperature (°C)

(c) 100

SN2O (%)

75 50 25 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.21. Oxidation of ammonia over SCR catalysts. Feed gas: 10% O2 , 5% H2 O, and 1000 ppm NH3 in N2 . GSHV = 52 000 h−1 . () Cu-ZSM5, (•) V2 O5 /WO3 −TiO2 , and () Fe-ZSM5.

potential of the exchanged redox element is correlated with the low-temperature SCR activity (Figure 9.17) but also with the tendency to oxidize ammonia to N2 , N2 O and NO (Figures 9.18–9.20).

4.3. Inhibition of SCR by ammonia and ammonia storage An important phenomenon during SCR over Fe-ZSM5 is the inhibition of the reaction by the reactant ammonia. As expected, DeNOx increases linearly with the amount of ammonia dosed and levels off above a certain ammonia dosage according to the limited activity of the catalyst at a given temperature (Figure 9.22) [54]. However, for low and intermediate temperatures, DeNOx is even going down again if ammonia is overdosed. This behavior is due to a competitive adsorption of ammonia on active Fe sites or due to the reduction of Fe3+ to Fe2+ [54,55]. This ammonia inhibition effect was not observed for Cu-ZSM5 and for V2 O5 /WO3 −TiO2 with a vanadia concentration above 1.5%. However, inhibition of the SCR reaction by ammonia is a known phenomenon for V2 O5 /WO3 −TiO2 catalysts with low vanadia concentrations below and around 1%, which are used for SCR systems in stationary power plants. When the ammonia storage capacities of Fe-ZSM5 and V2 O5 /WO3 −TiO2 are compared (not shown here), the much higher amount of ammonia desorbed from Fe-ZSM5

282

Past and Present in DeNOx Catalysis 100

DeNOx (%)

80 60 40 20 0

1.0

0.5

0.0

1.5

α (–)

Figure 9.22. Inhibition of the standard-SCR reaction over Fe-ZSM5 by ammonia. () 200 C, (•) 250 C, () 300 C, () 350 C, () 400 C, () 450 C, (♦) 500 C, (∗ ) 550 C, and (—) 600 C.

is noticeable, which is caused by the extremely high surface area of the zeolite. This might result in problems if Fe-ZSM5 is used as an SCR catalyst for a diesel passenger car due to the typically very fast load changes of the engines in combination with an average low raw emission level of NOx . In case of a rapid load increase, the rising catalyst temperature results in desorption of large amounts of ammonia, which cannot be compensated anymore by the reduction of the urea dosage. Even if the urea dosage is reduced to 0, ammonia emissions might be observed.

4.4. Stability of SCR catalysts Stability is a very important issue in the development of SCR catalysts, as a certain DeNOx level has to be guaranteed over the lifetime of the diesel vehicle in order to observe the emission limits. Due to the highly transient operation of the engine, the catalysts are subjected to a broad variety of ageing conditions, including also exceptionally high temperatures when a possible DPF is regenerated. Figure 9.23 compares the SCR

100

DeNOx (%)

80 60 40 20 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.23. Comparison of the SCR activity of Fe-ZSM5 after different ageing procedures. () Fresh, (•) dry ageing: 10% O2 in N2 , 650 C, 50 h, () wet ageing: 10% H2 O + 10% O2 in N2 , 650 C, 50 h, () SO2 treatment: 100 ppm SO2 + 5% H2 O + 10% O2 in N2 , 500 C, 50 h.

Catalyst Development for Mobile Urea-SCR Systems

283

100

DeNOx (%)

80 60 40 20 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.24. DeNOx at 10 ppm NH3 slip for the () fresh, (•) dry aged, and () wet aged Cu-ZSM5 monolith catalyst under standard-SCR conditions.

activity of Fe-ZSM5 in the fresh state with the activity after thermal and hydrothermal ageing as well as after a treatment with SO2 [54]. Whereas a thermal treatment at 650 C for 50 h alone did not affect DeNOx at all, the combination of heat and water lowered the SCR activity moderately. The N2 O formation remained unchanged. 27 Al NMR spectroscopy revealed that water favors the dealumination of the zeolite framework at high temperatures, thereby reducing the Brønsted acidity, which is connected with the SCR activity [54]. Fe-ZSM5 was totally stable towards SO2 and showed even a small increase in DeNOx , which may be explained by an increase in surface acidity. Despite the already good stability of Fe-ZSM5 in the SCR process, efforts are made to develop metal-exchanged zeolites, which can withstand even harsher conditions as they might be possible during an erratic regeneration of a DPF. It is apparent from Figure 9.24 that Cu-ZSM5 is much more sensitive towards ageing than Fe-ZSM5. Thermal as well as hydrothermal treatment at 650 C for 50 h considerably reduced the DeNOx activity. Around 25 ppm of N2 O was produced at all temperatures. Surprisingly, the wet aged catalyst performed better than the dry aged catalyst, which contrasts with the behavior of Fe-ZSM5, where the dry aged catalyst performed better than the wet aged catalyst.

4.5. Fast-SCR The catalytic activity was also investigated with a feed containing 50% NO and 50% NO2 (fast-SCR conditions) (Figure 9.25) [56]. The DeNOx activity of both Fe-ZSM5 and V2 O5 /WO3 −TiO2 was tremendously enhanced to >90% at T > 200 C by this measure, whereby Fe-ZSM5 excels V2 O5 /WO3 −TiO2 in both the low and high temperature regions. When comparing the SCR activity with (Figure 9.25) and without NO2 (Figure 9.18), it is clearly discernible that Fe-ZSM5 profits even more by the addition of NO2 than V2 O5 /WO3 −TiO2 . For Cu-ZSM5, showing already high activity without NO2 , a relatively small increase in DeNOx was observed under fast-SCR conditions.

284

Past and Present in DeNOx Catalysis 100

DeNOx (%)

75 50 25 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.25. DeNOx of Fe-ZSM5 and V2 O5 /WO3 −TiO2 under fast-SCR conditions. () Cu-ZSM5, (•) V2 O5 /WO3 −TiO2 , and () Fe-ZSM5. Feed gas: 10% O2 , 5% H2 O, 500 ppm NO + 500 ppm NO2 , variable amount of NH3 and N2 . GSHV = 52 000 h−1 .

4.6. N2 O formation The hydrothermal ageing of Fe-ZSM5, which represents practical operating conditions, resulted in the production of N2 O at intermediate temperatures in the presence of NO2 in the gas feed (Figure 9.26) [57]. For Cu-ZSM5, the addition of NO2 to the feed increased the N2 O production significantly, the maximum value of 111 ppm found at 250 C. For a better understanding of the N2 O formation over Fe-ZSM5 in the presence of NO2 , supplementing experiments with N2 O in the feed were carried out [56]. In the absence of ammonia, N2 O started to react to nitrogen above 450 C (Figure 9.27) according to the simple decomposition reaction: 2N2 O → N2 + O2 When ammonia was dosed, the N2 O conversion was drastically enhanced, lowering both the onset and full conversion temperature by 100–350 C and 600 C, respectively. On the one hand, ammonia promotes the N2 O decomposition reaction and on the other hand, it selectively reduces N2 O to nitrogen in the N2 O-SCR reaction (Figure 9.27) [56]: 3N2 O + 2NH3 → 4N2 + 3H2 O

120

N2O (ppm)

100 80 60 40 20 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.26. N2 O formation under fast-SCR conditions. () fresh, and () wet aged Fe-ZSM5, () fresh Cu-ZSM5, (•) V2 O5 /WO3 −TiO2 .

Catalyst Development for Mobile Urea-SCR Systems

285

100

XN2O (%)

80 60 40 20 0 150

250

350

450

550

650

750

Temperature (°C)

Figure 9.27. N2 O decomposition and N2 O-SCR reaction over Fe-ZSM5. Feed gas: 10% O2 , 5% H2 O, 500 ppm N2 O, () 0 and () 500 ppm NH3 in N2 . GHSV = 52 000 h−1 .

5. CONCLUSIONS The main challenges on the way from stationary SCR systems to systems for mobile diesel engines have been addressed. • In this context, first the utilization of urea as reducing agent has to be mentioned, which has the advantages of being non-toxic and safe to handle, but the high freezing point and the heat required for decomposition are intrinsic problems. Therefore, the efficiency of the urea SCR process is mainly limited by the minimum temperature required for urea dosing and decomposition. At least the risk of byproduct formation can be prevented by a proper design of the urea injection and the catalytic system. However, it should be mentioned that the utilization of ammonia or ammonia solution as reducing agents would help to circumvent these principle limitations and allow us to reach higher average DeNOx values [15]. The decisive parameter for the achievement of a certain DeNOx level with urea SCR is always the exhaust gas temperature, which is determined by the engine technology and the operating conditions. Due to low exhaust gas temperatures under low load conditions the urea injection has to be stopped, which prevents the achievement of high average DeNOx values. Besides the problems with the urea thermohydrolysis, low exhaust gas temperatures also decrease the activity of the SCR catalysts. • The problem of low exhaust gas temperatures is further aggravated by exhaust gas recirculation (EGR), a new engine technology, which further reduces the exhaust gas temperature. Consequently, the engine technology has to be developed so that the positive effect on the engine’s NOx raw emissions overcompensate any potential negative effect on the SCR process by far. One solution for this problem might be the combination of HCCI in the partial load/speed regime of the engine, which almost eliminates the raw NOx emissions, and the application of urea SCR as aftertreatment technology for the residual operational regime [58]. • The utilization of a hydrolysis catalyst is advantageous for capturing urea droplets, improved thermohydrolysis of urea and mixing/homogenization of the flow, but it is not a pre-requisite for a working system.

286

Past and Present in DeNOx Catalysis

• The main reactions, which have to be considered on SCR catalysts, are the standardSCR, fast-SCR, and the NO2 -SCR reactions, beside the ammonia oxidation and the formation of N2 O. The fast-SCR reaction is promoted by NO2 in the feed that can be generated from NO in a pre-oxidation catalyst. However, the right dimensioning of the oxidation catalyst is critical in order to prevent the production of an excess of hazardous NO2 . This problem is further aggravated if a continuous regenerating DPF is installed in front of the SCR system, as part of the NO2 produced by the oxidation catalyst is always consumed in the filter for soot oxidation. • State-of-the-art SCR catalysts tackle space velocities of 50 000 h−1 (or higher) at >95% DeNOx . The cell densities of the monoliths should be around 400 cpsi, but improvements are expected up to 600 cpsi. Vanadia-based SCR catalysts with a V2 O5 concentration of 1.5–3% are established and are approved systems. The optimum vanadia concentration depends on the application, requiring different compromises between low temperature activity and N2 O selectivity as well as high temperature stability. Extruded SCR catalysts are advantageous at low temperatures, where only catalyst mass is rate determining. However, coated SCR catalyst can be produced with higher cell densities, giving more freedom of design and enabling better urea droplet capturing as well as higher activities. • Metal-exchanged zeolites are an interesting alternative to the established vanadiabased SCR catalysts. The main advantages of Fe-ZSM5 are the very good temperature resistance above 550 C and the low tendency for ammonia oxidation and N2 O formation at these high temperatures. Drawbacks are the moderate low temperature activity, if only NO is in the feed and the tendency to produce N2 O at intermediate temperatures in the presence of NO2 . On the contrary, vanadia-based systems are excellent at low and intermediate temperatures, but start to produce significant amounts of N2 O at 550 C and ammonia oxidations gets predominant in this temperature region. The main disadvantage of this catalyst type is the limited stability at temperatures above 600 C. However, by lowering the vanadia concentration and by adding rare earth elements, the stability can be improved [59]. Cu-ZSM5 is extremely active at low temperatures, but oxidizes much ammonia, preventing an application in the high temperature region. However, this catalyst is promising for reduction of NOx in all kind of low temperature applications.

REFERENCES [1] Bosch, H. and Janssen, F. (1988) Catalytic Reduction of Nitrogen Oxides, Catal. Today, 2, 369. [2] Köser, H. (Ed.) (1992) SCR-deNOx-Katalysatoren Qualitätssicherung, Beurteilung und neue Entwicklungen, Vulkan-Verlag, Essen, ISBN 3-8027-8507-X. [3] Forzatti, P. (2001) Present Status and Perspectives in de-NOx SCR Catalysis, Appl. Catal. A, 222, 221. [4] Koebel, M., Elsener, M. and Eichler, H.P. (1990) Mit Harnstoff gegen Stickoxide, Tech. Rundsch., 82, 74. [5] Koebel, M. and Elsener, M. (1991) Stickoxidminimierung bei Dieselmotoren, Schweiz. Ing. Archit., 109, 187.

Catalyst Development for Mobile Urea-SCR Systems

287

[6] Koebel, M., Elsener, M. and Eicher, H.P. (1991) Stickoxidminderung bei Stationären Dieselmotoren Mittels SCR und Harnstoff als Reduktionsmittel, Brennstoff Wärme Kraft, Special 3 E24. [7] Koebel, M. (1992) Stickoxidminderung in Abgasen, Schweiz. Ing. Archit., 110, 693. [8] Koebel, M. and Elsener, M. (1992) Entstickung von Abgasen nach dem SNCR-Verfahren: Ammoniak oder Harnstoff als Reduktionsmittel?, Chem. Ing. Tech., 64, 934. [9] Koebel, M. (1993) Entstickung von Dieselabgasen mit Harnstoff-SCR, VDI Ber., 1019, 195. [10] Hug, H.T., Mayer, A. and Hartenstein, A. (1993) Off-Highway Exhaust Gas AfterTreatment: Combining Urea-SCR, Oxidation Catalysis and Traps, SAE Technical Paper Series 1993–0363. [11] Gabrielsson, P.L.T. (2004) Urea-SCR in Automotive Applications, Top. Catal., 28, 177. [12] Koebel, M., Elsener, M. and Marti, T. (1996) NOx -Reduction in Diesel Exhaust Gas with Urea and Selective Catalytic Reduction, Combust. Sci. Technol., 121, 85. [13] Koebel, M., Elsener, M. and Madia, G. (2001) Recent Advances in the Development of Urea-SCR for Automotive Applications, SAE Technical Paper Series 2001-01-3625. [14] Koebel, M., Elsener, M., Kröcher, O., et al. (2004) NOx Reduction in the Exhaust of Mobile Heavy-Duty Diesel Engines by Urea-SCR, Topics Catal., 30/31, 43. [15] Koebel, M., Elsener, M. and Kleemann, M. (2000) Urea-SCR: A Promising Technique to Reduce NOx Emissions from Automotive Diesel Engines, Catal. Today, 59, 335. [16] König A., Herding G., Hupfeld, B., et al. (2001) Current Tasks and Challenges for Exhaust Aftertreatment Research. A Viewpoint from the Automotive Industry, Topics Catal., 16/17, 23. [17] Tennison, P., Lambert, C. and Levin, M. (2004) NOx Control Development with Urea SCR on a Diesel Passenger Car, SAE Technical Paper Series 2004-01-1291. [18] Koebel, M. and Strutz, E.O. (2003) Thermal and Hydrolytic Decomposition of Urea for Automotive Selective Catalytic Reduction Systems: Thermochemical and Practical Aspects, Ind. Eng. Chem. Res., 42, 2093. [19] Kleemann, M., Elsener, M. Koebel, M., et al. (2000) Hydrolysis of Isocyanic Acid on SCR Catalysts, Ind. Eng. Chem. Res., 39, 4120. [20] Maurer, B., Jacob, E. and Weisweiler, W. (1999) Modelgasuntersuchen mit NH3 und Harnstoff als Reduktionsmittel für die katalytische NOx -Reduktion, MTZ, 60, 308. [21] Koebel, M. and Elsener, M. (1992) Entstickung von Abgasen nach dem SNCR-Verfahren: Ammoniak oder Harnstoff als Reduktionsmittel?, Chem. Ing. Tech., 64, 934. [22] Piazzesi, G., Devadas, M., Kröcher, O., et al. (2006) Isocyanic Acid Hydrolysis over FeZSM5 in Urea-SCR, Catal. Commun., 7, 600. [23] Jacob, E. (1990) Verfahren und Vorrichtung zur selektiven katalytischen NOx -Reduktion in sauerstoffhaltigen Abgasen, German Patent DE 4038054. [24] Piazzesi, G., Kröcher, O., Elsener, M., et al. (2006) Adsorption and Hydrolysis of Isocyanic Acid on TiO2 , Appl. Catal. B, 65, 55. [25] Piazzesi, G. (2006) The Catalytic Hydrolysis of Isocyanic Acid (HNCO) in the Urea-SCR Process, Ph.D. Thesis No. 16693, ETH Zurich. [26] Koebel, M. and Elsener, M. (1995) Determination of Urea and its Thermal Decomposition Products by High-Performance Liquid Chromatography, J. Chromatogr., 689, 164. [27] Schaber, P.M., Colson, J., Higgins, S., et al. (1999) Study of the Urea Thermal Decomposition (Pyrolysis) Reaction and Importance to Cyanuric Acid Production, American Laboratory, August 13. [28] Jacob, E., Emmerling, G., Döring, A., et al. (1998) NOx -Verminderung für Nutzfahrzeugmotoren mit Harnstoff-SCR-Kompaktsystemen (GDKAT), 19. Int. Wiener Motorensymp. (H.P. Lenz, ed.), Fortschritt-Berichte VDI Reihe 12, No. 348, Vol. 2, 366. [29] Jacob, E., Müller, R., R. Scheeder, R. et al. (2006) Hochleistungs-SCR-Katalysatorsystem: Garant für niedrigste NOx -Emission, 27. Internationales Wiener Motorensymposium (H. P. Lenz, ed.) VDI-Fortschritt-Berichte VDI Reihe 12, Nr. 622, 240.

288

Past and Present in DeNOx Catalysis

[30] Müller-Haas, K. and Rice, M. (2005) Innovative Metallic Substrates for Exhaust Emission Challenges for Gasoline and Diesel Engines, SAE Technical Paper Series 2005-01-3851. [31] Madia, G., Elsener, M., Koebel, M., et al. (2002) Thermal Stability of Vanadia-TungstaTitania Catalysts in the SCR Process, Appl. Catal. B, 39, 181. [32] Koebel, M., Elsener, M. and Madia, G. (2001) Reaction Pathways in the Selective Catalytic Reduction Process with NO and NO2 at Low Temperatures, Ind. Eng. Chem. Res., 40, 52. [33] Madia, G., Koebel, M., Elsener, M., et al. (2002) Side Reactions in the Selective Catalytic Reduction of NOx with Various NO2 Fractions, Ind. Eng. Chem. Res., 41, 4008. [34] Williams, J.L. (2001) Monolith Structures, Materials, Properties and Uses, Catal. Today, 69, 3. [35] Koebel, M. and Elsener, M. (1998) Selective Catalytic Reduction of NO over Commercial DeNOx-Catalysts: Comparison of the Measured and Calculated Performance, Ind. Eng. Chem. Res., 37, 327. [36] Kato, A., Matsuda, S., Kamo, T., et al. (1981) Reaction between NOx and NH3 on Iron Oxide-Titanium Oxide Catalyst, J. Phys. Chem., 85, 4099. [37] Madia, G., Koebel, M., Elsener, M., et al. (2002) The Effect of an Oxidation Precatalyst on the NOx Reduction by Ammonia SCR, Ind. Eng. Chem. Res., 41, 3512. [38] Koebel, M., Madia, G., Raimondi, F., et al. (2002) Enhanced Reoxidation of Vanadia by NO2 in the fast SCR Reaction, J. Catal., 209, 159. [39] Nova, I., Ciardellia, C., Tronconi, E., et al. (2006) NH3 –NO/NO2 Chemistry over V-based Catalysts and its Role in the Mechanism of the Fast SCR Reaction, Catal. Today, 114, 3. [40] Ciardelli, C., Nova, I., Tronconi, E., et al. (2004) A ‘Nitrate Route’ for the Low Temperature “Fast SCR” Reaction over a V2 O5 –WO3 /TiO2 Commercial Catalyst, Chem. Commun. 2718. [41] Koebel, M., Elsener, M. and Madia, G. (2002) Selective Catalytic Reduction of NO and NO2 at Low Temperatures, Catal. Today, 73, 239. [42] Despres, J., Elsener, M., Koebel, M., et al. (2004) Catalytic Oxidation of Nitrogen Monoxide over Pt/SiO2 , Appl. Catal. B, 50, 73. [43] Bunge, R. and Bürgler, B. European Patent EP 1495796. [44] Kleemann, M. (1999) Beschichtung von Cordierit-Wabenkörpern für die Selektive Katalytische Reduktion von Stickoxiden, Ph.D. Thesis No. 13401, ETH Zurich. [45] Kleemann, M., Elsener, M., Koebel, M., et al. (2000) Investigation of the Ammonia Adsorption on Monolithic SCR Catalysts by Transient Response Analysis, Appl. Catal. B, 27, 231. [46] Schär, C.M., Onder, C.H., Geering, H.P., et al. (2003) Control of a Urea SCR Catalytic Converter System for a Mobile Heavy Duty Diesel Engine, SAE Technical Paper Series 2003-01-0776. [47] Schär, C.M., Onder, C.H. and Geering, H.P. (2006) Control of an SCR Catalytic Converter System for a Mobile Heavy-Duty Application, IEEE Trans. Contr. Sys. Technol., 14, 641. [48] Schär, C. (2003) Control of a Selective Catalytic Reduction Process, Ph.D. Thesis No. 15221, ETH Zurich. [49] Elsener, M., Geering, H.P., Jaussi, F., et al. (2003) Aufbau und Vermessung eines deNOx Systems auf der Basis von Harnstoff-SCR, MTZ, 11, 966. [50] Schär, C.M., Onder, C.H., Geering, H.P., et al. (2004) Control-Oriented Model of an SCR Catalytic Converter System, SAE Technical Paper Series 2004-01-0153. [51] Costigan, M., Cary, R. and Dobson, S. (2001) Vanadium Pentoxide and Other Inorganic Vanadium Compounds, Concise International Chemical Assessment Document 29, World Health Organization, Geneva, Switzerland. [52] Sun, Q., Gao, Z., Chen, Y., et al. (2001) Reduction of NOx with Ammonia over Fe/MFI: Reaction Mechanism Based on Isotopic Labeling, J. Catal., 201, 89. [53] Devadas, M., Kröcher, O. and Wokaun, A. (2005) Catalytic Investigation of Fe-ZSM5 in the Selective Catalytic Reduction of NOx with NH3 , React. Kin. Catal. Lett., 86, 347. [54] Kröcher, O., Devadas, M., Elsener, M., et al. (2006) Investigation of the Selective Catalytic Reduction of NO by NH3 on FE-ZSM5 Monolith Catalysts, Appl. Catal. B, 66, 208.

Catalyst Development for Mobile Urea-SCR Systems

289

[55] Devadas, M., Kröcher, O., Elsener, M., et al. (2007) Characterization and Catalytic Investigation of FE-ZSM5 for Urea-SCR, Catal. Today, 119, 137. [56] Devadas, M., Kröcher, O., Elsener, M. et al. (2006) Influence of NO2 on the Selective Catalytic Reduction of NO with Ammonia over FE-ZSM5, Appl. Catal. B, 67, 187. [57] Devadas, M. (2006) Selective catalytic reduction (SCR) of nitrogen oxides with ammonia over Fe-ZSM5, Ph.D. Thesis No. 16524, ETH Zurich. [58] Boulouchos, K., Kröcher, O. and Lutz, T. (2005) NOx -Reduction f¨ur PW-Dieselmotoren, auto&technik, 4, 28. [59] Casanova, M., Rocchini, E., Trovarelli, A., et al. (2006) High-Temperature Stability of V2 O5 /TiO2 −WO3 −SiO2 Catalysts Modified with Rare-Earths, J. Alloys Comp., 408–412 1108.

This page intentionally left blank

Chapter 10

THE FORMATION OF N2 O DURING NOX CONVERSION: FUNDAMENTAL APPROACH AND PRACTICAL DEVELOPMENTS P. Granger∗ , J. P. Dacquin, F. Dhainaut and C. Dujardin Unité de Catalyse et de Chimie du Solide, Université des Sciences et Technologies de Lille, UMR CNRS 8181, Bâtiment C3, 59655 – Villeneuve d’Ascq, France ∗ Corresponding author: Unité de Catalyse et de Chimie du Solide, Université des Sciences et Technologies de Lille, UMR CNRS 8181, Bâtiment C3, 59655 – Villeneuve d’Ascq, France. Tel.: 33 (0)3 20 43 49 38, Fax.: 33 (0)3 20 43 65 61, E-mail: [email protected]

Abstract Presently, a growing interest is focused on unregulated emissions of nitrous oxide (N2 O) from stationary and mobile sources in order to anticipate future restrictive legislations, since N2 O exhibits a significant higher global warming power than that of CO2 . The adoption of end-of-pipe technologies is appropriate. However, the simultaneous conversion of NOx and N2 O over catalytic processes is still challenging both for industrial plants and automotive exhaust gases particularly at low temperature. Subsequent selectivity enhancements towards the formation of N2 probably need better insights into the mechanisms involved in the formation and the subsequent conversion of N2 O during the overall reduction of NOx , particularly in O2 excess. Up to now practical solutions for mobile sources imply the use of noble metals. Conventional three-way catalysts (TWC) running under stoichiometric conditions are wide-spread through the world even if their efficiency is still restricted during the cold start engine with a substantial formation of undesired N-containing products such as N2 O. Nowadays, the use of noble metals in TWC, particularly Rh, becomes more and more questionable with the continuous development of lean-burn engines because of their poor efficiency to convert NOx into nitrogen in those running conditions. In such a circumstance, there is a particular interest in developing non-noble metal-based catalysts and also the use of additives and alternative reducing agents. By way of illustration, hydrogen could be an interesting issue for both stationary and mobile sources for the reduction of NOx emissions at low temperature under lean conditions. This paper will discuss on such an opportunity and the correlative development of novel catalysts.

1. BACKGROUND AND SCIENTIFIC CONTEXT Nowadays most of the research programmes are focused on the development of low atmospheric pollutant emission systems integrating low energy consumption in order to Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

292

Past and Present in DeNOx Catalysis

minimise CO2 emissions and other gases, which exhibit greenhouse gas behaviour. Diesel and lean-burn gasoline engines represent promising alternatives within this context. However, under lean conditions, usual after-treatment technologies are not effective and future NOx standard emissions within the Euro VI are still challenging. Otherwise, unregulated emissions of nitrous oxide (N2 O) from industrial plants and during the cold start engine of light vehicles have actually a growing interest. Probably, the regulation of N2 O from industrial sources is the most feasible. Presently, numerous patents have already claimed efficient catalytic systems for solving this environmental problem [1–3] involving thermally stable and selective catalysts for the decomposition of N2 O into N2 and O2 at high temperature (T  900 C). On the other hand, the simultaneous decomposition of NO and N2 O in the exhaust gas should proceed at relatively low temperature. In that case, the concentration of N2 O in the tail gas is also an outstanding parameter [4]. The actual end-of-pipe technologies running at low temperature for stationary sources need the use of reducing agents because of strong inhibiting effects due to the accumulation of oxygen atoms on the active sites and to a low sulphur tolerance. Ammonia is commonly used for the selective conversion of NOx into nitrogen in O2 excess over conventional vanadia–titania or supported copperbased catalysts in the temperature range 200–400 C [5,6]. Interestingly, oxygen has a beneficial effect on the selective conversion of NO to nitrogen. Nevertheless, the simultaneous conversion of N2 O and NO cannot be achieved in the usual running conditions and necessitates higher operating temperatures [7–10]. In fact, the temperature seems to be a key parameter both for stationary and mobile sources. The catalyst composition should also be considered and the different characteristics of both sources in terms of space velocity, gas composition may a priori exclude an unique solution. Farrauto and Heck [11] reported different approaches and future needs for improving the catalytic performances of automotive exhaust catalysts at low temperature. As indicated by both authors, a close-coupled catalyst may represent a significant technological advancement in meeting the cold start hydrocarbon emission problem associated to the ultra-low emission vehicle. However, the high residual temperature may induce a significant loss of activity due to thermal sintering. Consequently, the use of thermally stable materials and electrically heated systems seem to be more adequate options. The use of noble metals is also a major point particularly for automotive exhaust systems, which are characterised by high space hourly gas velocity. Nowadays, most of the car manufacturers are interested to replace noble metals by cheaper active phases or at least to significantly lower their content without altering their tolerance to poisoning effects. Recent developments using gold and silver as active components specifically for low temperature applications could be promising, but their sulphur tolerance is questionable in those temperature conditions [12–14]. Apparently, such an aspect may exclude an extensive development of those silver-based catalysts for NOx abatement processes for stationary sources. This review reports results obtained during the past decades regarding the specific point related to the formation and the subsequent transformation of nitrous oxide (N2 O) during the overall reduction of NO and recent advanced developments concerning the selective catalytic removal of NOx into nitrogen, at low temperature, mainly under lean conditions. A particular attention has been paid towards the use of hydrogen, which may concern both stationary and mobile sources. In the particular case of automotive exhaust systems, H2 can be produced on board in significant amount by reforming and then

Formation of N2 O During NOx Conversion

293

used in cold start/exhaust conditions [15]. The production of hydrogen could also be enhanced via the reforming of additives such as alcohol. Recent investigations claimed that hydrogen significantly enhances the conversion of NOx and unburned hydrocarbons, which makes suitable this option, particularly, in the case of methane usually activated at high temperature. The effective role of H2 in the conversion of NOx and unburned hydrocarbons and the molecular description is actually debatable. Indeed, H2 may directly react with NO at low temperature but may also exhibit a beneficial effect on the conversion of hydrocarbon with controversial statements relative to the nature of the intermediates involved in the reaction and correlatively to the nature of the active sites particularly those involving reaction pathways leading to the formation of nitrogen [16,17]. Regarding the direct reduction of NO by hydrogen previous investigations over noble metals recognised H2 as non-selective [18]. In addition, nitrous oxide (N2 O) is usually the main N-containing product. Both aspects may a priori rule out any successful practical developments. Nevertheless, recent investigations underlined promising results on supported platinum and palladium-based catalysts over reducible supports [19,20]. Significant improvements are reported on Pd-based catalysts supported on perovskite structures. Those materials preserved high activities and selectivities after ageing in severe conditions. It was found that peculiar catalytic performances may be obtained after successive thermal activation steps under controlled atmospheres to obtain optimal and stable interactions between oxidic palladium species and the perovskite support. Various attempts seem to indicate that these materials could represent an interesting issue for mobile and stationary sources, but such an opportunity will be related to further successful developments in the synthesis of those materials to obtain higher specific surface areas and low contents of impurities due to the segregation of monometallic oxides. The in situ generation of those structures under reactive conditions could be an alternative solution and highlights an important factor related to surface reconstructions and the build-up of those structures under reactive conditions. The stabilisation of isolated active species assisted by the reconstruction of the support, highly active and selective towards the conversion of NO into N2 (without any formation of N2 O), could be a matter of outstanding importance.

2. GENERAL ASPECTS CONCERNING THE FORMATION AND THE SUBSEQUENT TRANSFORMATION OF N2 O DURING THE OVERALL TRANSFORMATION OF NO 2.1. Conventional three-way catalysts 2.1.1. The overall conversion of NO Among the different noble metals (Pt, Rh and Pd) in three-way catalysts (TWC), Rh is usually recognised as essential for the conversion of NO into nitrogen. The formation of nitrous oxide as side product typically occurs below the light-off temperature, N2 O being in those conditions the main N-containing product (see Figure 10.1). Under stoichiometric conditions the intermediate formation of ammonia may also occur. Generally such a formation disappears in the presence of a large excess of oxygen.

294

Past and Present in DeNOx Catalysis (b)

(a)

600

Concentration /ppm

Conversion /%

100 80 CO NO C3H8

60 40

C3H6

20 0

500 400 N2O

300

NH3

200 100 0

0

100

200

300

Temperature /°C

400

500

0

100

200

300

400

500

Temperature /°C

Figure 10.1. Light-off curves for the reaction of the standard mixtures (R = 0.98, 5000 ppm H2 O) over Rh/Al2 O3 . (a) Conversions of CO, NO, C3 H6 and C3 H8 ; (b) production of N2 O and NH3 (reproduced with permission from Ref. [21]). Reactions involved over three-way catalysts are oxidation, (1) CO + 1/2O2 → CO2 ; (2) Cx Hy + (x + y/4) O2 → xCO2 + y/2H2 O, and reduction, (3) NO + H2 → 1/2N2 + H2 O; (4) NO + CO → 1/2N2 + CO2 ; (5) (2x + y/2)NO + Cx Hy → (x + y/4)N2 + xCO2 + y/2H2 O.

Cant et al. [21] focused their attention on the concentration of N2 O in the automotive exhaust gas, which are rather low (14 ppm) but quite dependent on the air-to-fuel ratio. Typically 60–80% of NO is converted into N2 O below the light-off temperature on Rh and then the selectivity drops at relatively high temperature 370 C [21,22] when the partial pressures of NO tends below 10 Torr [22–25]. Previous surface science studies showed that Rh is essential for NOx removal over TWC because NO dissociates more readily on metallic Rh than on Pt and Pd sites [26–28]. Nevertheless, the efficiency of Rh to selectively transform NO into N2 is restricted below the light-off temperature with a predominant formation of N2 O. Future practical developments are closely related to a better understanding of the formation and on the transformation of N2 O over noble metals during the cold start engine. Such an aspect is still challenging over TWC and particularly under lean conditions since the extent in NOx conversion is usually significantly lowered. Different reactions pathways on Rh may explain the intermediate formation of ammonia. NH3 can be obtained via successive reaction steps between adsorbed NHx and dissociated hydrogen species [29]. Alternately, the formation of ammonia may occur via the hydrolysis of isocyanic acid (HNCO) [30]. Isocyanate species are formed by reaction between Nads and COads on metallic particles. Those species can diffuse onto the support leading to spectator species or alternately react with Hads yielding ultimately HNCO. Previous infrared spectroscopic investigations pointed out that isocyanate species predominantly form over rhodium-based catalysts [31].

2.1.2. Kinetic aspects of the CO + NO reaction and related N2 O formation/transformation In spite of numerous investigations dealing with the transformation of NO in the past three decades, only few of them carefully examined the formation of N2 O and its

Formation of N2 O During NOx Conversion

295

subsequent transformation. Such a question is still opened particularly on noble metals and the reaction pathways for the formation of N2 and N2 O are sometimes controversial with the involvement of various intermediates such as NCO [29], NO dimer (N2 O2 , N2 O4 , dinitrosyl or nitrosyl species on Rh under stoichiometric and lean conditions [30–34] according to the following elementary steps: Rh−NO− + Rh0 N → 2Rh0 + N2 O

(6)

Rh(NO)2 + Rh0 N → 2Rh−NO+ + Rh−NO− + N2 O

(7)

Rh−NCO + NO → 2Rh+ + N2 O + CO

(8)

The nature of intermediates seems to be closely related to the oxidation state of noble metal species according to those operating conditions with the segregation of reduced or oxidic noble metal species. Presently, there is a general consensus on the mechanism schemes in three-way conditions, whereas under lean conditions the wide variety of catalysts previously investigated leads to more controversies regarding the nature of the reaction pathways leading to the formation of N2 and N2 O. By way of illustration, Burch and Coleman [35] recently proposed from transient kinetic experiments that the formation of N2 O would proceed via the following step on Pt-based catalysts. NOads + NOads → N2 O + Oads +∗

(9)

Such a reaction pathway has been earlier discussed on Rh according to the usual observation of gem-dinitrosyl species on isolated RhI species [36]. On the other hand, such a proposal could be more controversial on metallic Pt sites [37,38]. A particular attention on the mechanisms for the formation of N2 O over noble metals has been paid in our laboratory [37–40]. It was previously found that an enhancement in the initial selectivity towards the production of N2 (Table 10.1) during the CO + NO reaction can be related to an increase in the relative rate of step (13) over supported Pt-based catalysts [33]. Unexpectedly, Rh exhibits a poor selectivity towards the formation of N2 at low conversion and low temperature, which has been mainly related to a stronger NO adsorption on Rh than on Pt and Pd.

∗

NO+∗ ⇔ NOads

(10)

CO+∗ ⇔ COads

(11)

NOads +∗ ⇔ Nads + Oads

(12)

Nads + Nads → N2 + 2∗

(13)

NOads + Nads → N2 + Oads +∗

(14)

NOads + Nads → N2 O + 2∗

(15)

COads + Oads → CO2 + 2∗

(16)

stands for a vacant adsorption site.

296

Past and Present in DeNOx Catalysis Table 10.1. Selectivity towards the formation of nitrogen on supported Pt-based catalysts at 300 C (initial partial pressure of NO and CO equal to 5 × 10−3 atm [37]) Catalyst

NO conversion (%)

SN2 (%)

k13 /k15 2

k14 /k15

Pt/Cr3 C2 Pt/Si3 N4 Pt/Al2 O3

0.9 1.2 2.2

56.1 32.4 27

2012 91 0.25

0 0 0.30

Typically, Rh sites are usually saturated by chemisorbed NO molecules below the light-off temperature with low N coverages. Consequently, the formation of nitrogen would not occur via the associative desorption of two adjacent chemisorbed nitrogen atoms according to step (13) but predominantly via steps (14) [38]. Surface science investigations seem to be in line with the involvement of step (14) rather than step (13) for the formation of N2 at least on Rh. For instance. Zaera and Gopinath [41] who investigated the NO + CO reaction on Rh (111), using a molecular beam technique, showed that N−NO intermediate would be involved in the formation of nitrogen. Infrared spectroscopic investigations suggested that different nitrosyl species would be involved in the formation of N2 and N2 O on rhodium-based catalysts. As illustrated in Figure 10.2, a detrimental effect on the formation of N2 is observed correlatively to the development of positively charged NO species. In fact, it was concluded that those latter species could originate the formation of N2 O according to step (15), whereas neutral nitrosyl species would be preferentially involved in the formation of nitrogen according to step (14). Additives usually enhance the oxygen storage capacity of TWC. Subsequent additions of ceria, lanthana or ceria−zirconia mixed oxides promote the activity in NO conversion [43]. Clearly, ceria and more recently ceria−zirconia mixed oxides [44,45] considerably improve the conversion of NO but only exhibit a weak effect on the selectivity at low conversion and low temperature typically during the cold start engine. Kinetic investigations showed that at low temperature the beneficial effect of ceria on the activity of noble metals may be explained by changes in the mechanism schemes with steps involving ceria, but those corresponding to the formation of N2 and N2 O would not be affected since only noble metal sites would be involved according to steps (13–15). On the other hand, an extensive reduction of ceria has been evidenced during the CO + NO reaction near stoichiometric conditions in the absence of oxygen with subsequent modifications of the adsorptive properties of noble metals in contact with reduced ceria species [40]. Holles et al. [46] assumed that ceria and lanthana did not alter the conventional mechanism suggested over noble metals but would act predominantly on the rates of the individual elementary steps. Basically, the dissociation of chemisorbed NO molecules over noble metals is currently suggested as rate determining step with sites usually covered by NO. In the presence of ceria, the positive order with respect to the partial pressure of NO suggests that the dissociation of NO would not be rate determining when ceria interacts with noble metals. As developed by Holles et al. [46] and Oh [47] an increase in the rate of NO dissociation should lead to significant changes in adsorbate coverage with significant amounts of chemisorbed N atoms. Consequently, an improvement of the selectivity towards the formation of nitrogen was

Formation of N2 O During NOx Conversion

297 (b)

(a)

Intensity (a.u.)

Rh(NO)0 1607 Rh(NO )δ+ 1687 0.2 K.M.

2237 f

Rh(NO)δ+ Rh(NO) Rh(CO)2δ+

e d c 0

b

1820

10

1625

a 2300

2100

1900

Wavenumber

1700

20 30 Time (min)

40

50

1500

(cm–1)

(Peak area)/(Second peak area)

(c) 1 0.8

N2O

0.6 0.4 0.2

N2

0 0

5

10

Number of pulses

Figure 10.2. Operando spectroscopic investigation of the NO transformation on Rh/Al2 O3 . (a) IR spectra recorded after successive NO pulses at 300 C on a prereduced catalyst. (b) and (c) Changes in the intensity of IR bands related to the adsorbates during successive NO pulses on a CO preadsorbed surface at 300 C and correlative changes in the formation of N2 and N2 O, respectively (reproduced with permission from Ref. [42]).

associated with an enhancement of the recombination of adjacent chemisorbed atoms. An alternative explanation can be proposed, which accounts for a different configuration for adsorbed NO molecules on sites located at the metal/support interface with nitrogen bonded to noble metals and the terminal O atoms of NO molecules interacting with anionic vacancies from the support. The subsequent weakening facilitates the N−O bond breaking. A recent comparative investigation of the NO + CO reaction shows a significant rate enhancement in the formation of N2 on Ce098 Pd002 O2− prepared via a combustion synthesis method in comparison with conventional Pd-based catalysts supported on alumina

298

Past and Present in DeNOx Catalysis

and silica. In this case, a constant selectivity of 80% in the conversion and temperature of the study is observed [48]. Such a selectivity enhancement in comparison with metallic palladium particles dispersed on alumina has been explained by the presence of isolated Pd2+ in Ce4+ position. Those above-mentioned investigations suggest that the selectivity under lean conditions will be partly governed by the degree of dispersion of the active phase i.e. the noble metals, their oxidation state and their chemical environment. Hence those considerations underline the importance of the support for obtaining isolated species stabilised in unusual oxidation state under severe conditions at relatively high temperature. Now regarding the controversial statements related to the role of N2 O as primary product or intermediate during the overall CO + NO reaction, strong debatable points were initially introduced by Zhdanov [49] and Cho [50]. Interestingly, such a discussion was still developed in a recently reviewed article [51]. Contrary to previous arguments, no variation in the selectivity towards the production of N2 O in the course of the NO + CO, reaction on Rh/SiO2 does not rule out a two step process involving the intermediate formation of N2 O. As a matter of fact, it was found that the selectivity over noble metal is predominantly governed by the competition between NO and N2 O adsorption [52]. On rhodium, the strong adsorption of NO completely inhibits the readsorption of N2 O, which only takes place when NO is quasi-completely converted. In contrast the subsequent N2 O + CO reaction occurs more readily on Pt-based catalysts.

2.2. The reduction of NO by hydrogen over noble metals Previous kinetic investigations dealing with the NO + H2 reaction over supported noble metal-based catalysts showed different kinetic features according to the nature of the support [29,53–58]. Initially, this reaction has been described in the absence of oxygen on Rh deposited on silica and alumina by the following mechanism [29]. NO +∗ ⇔ NOads H2 + 2∗ ⇔ 2Hads

(17)

NOads + Hads → Nads + OHads

(18)

Nads + Nads → N2 + 2∗ NOads + Nads → N2 + Oads +∗ NOads + Nads → N2 O + 2∗ Nads + Hads → NHads +∗

(19)

NHads + Hads → NH2ads +∗

(20)

NH2ads + Hads → NH3 + 2∗

(21)

Oads + Hads → OHads +∗

(22)

OHads + Hads → H2 O + 2∗

(23)

Formation of N2 O During NOx Conversion

299

Further investigations revealed more complex kinetic features relative to the formation of N2 and N2 O. By way of illustration, Shestov et al. [58] suggested different reaction pathways on Pt/SiO2 for the production of nitrogen involving two equivalent species or physisorbed NO species, which would interact with NO-derived species. Such arguments are in qualitative agreement with the mechanism earlier suggested by Hecker and Bell [29]. However, the findings of Shestov opened the discussion relative to the involvement of chemisorbed NHx species in the production of N2 . Indeed, recent investigations proposed a direct interaction between NHads and NOads leading to the formation of nitrogen, which could explain the effective role of H2 in the conversion of NO into nitrogen [53]. We found that, at low temperature, in the range 70–150 C, noble metal sites are essentially covered by strongly NO species. In such conditions, the desorption of two adjacent chemisorbed N atoms would not occur significantly. This tendency is also supported by a microkinetic approach [59] where calculations leads to a much higher activation barrier for step (13) than that of step (14). Also, different reaction pathways could be involved for the formation of N2 O according to step (15) or (9) [57] in a kinetic regime where active sites are essentially covered by NO. Kinetic data on Rh and on Pd-based catalysts in this kinetic regime showed that the selectivity towards the production of NO and N2 O is insensitive to the reaction conditions particularly to the partial pressure of NO, which indicates that the formation of N2 and NO proceeds via a same intermediate. Consequently, the formation of N2 O and N2 during the NO + H2 reaction, in the presence or in the absence of oxygen, would likely occur via steps (14) and (15) on Pd/Al2 O3 [53,60] or alternately a bimolecular reaction between two adjacent nitrosyl species for the production of N2 should be envisaged for explaining the experimental evidences. Let us mention that this latter suggestion is still in agreement with the previous conclusions drawn by Shestov et al. [58]. On reducible supports typically when Pd is deposited on LaCoO3 then reduced in H2 at 450 C for obtaining Pd0 /CoOx /La2 O3 , an alternative mechanism would likely occur, which accounts for steps involving the creation of active sites at the metal/support interface. These active sites would be composed of metallic Pd in interaction with anionic vacancies from the support potentially active for the dissociation of NO according to step (26) [54]. NO+∗ ⇔ NOads H2 + 2∗ ⇔ 2Hads ‘O’ + Hads → ‘OH’+∗

(24)

‘OH’ + Hads → ‘V ’ +∗ +H2 O

(25)

‘V ’ + NOads → ‘O’ + Nads

(26)

Nads + Nads → N2 + 2∗ NOads + Nads → N2 + Oads +∗ NOads + Nads → N2 O + 2∗ Nads + Hads → NHads +∗

300

Past and Present in DeNOx Catalysis

NHads + Hads → NH2ads +∗ NH2ads + Hads → NH3 + 2∗ Oads + Hads → OHads +∗ OHads + Hads → H2 O + 2∗ ‘O’, ‘OH’ and ‘V ’ stand for reactive oxygen, hydroxyl species and anionic vacancies from the support, respectively. Generally, the kinetic behaviour of noble metal is affected in the presence of oxygen particularly when hydrogen is used, because it is recognised as non-selective. In addition, a detrimental effect of oxygen on the rate of NOx transformation usually takes place due to a strong inhibiting effect. As a matter of fact, oxygen may also affect the nature of the rate determining step. In the absence of oxygen the beneficial effect of hydrogen on the conversion of NO, at relatively low temperature in comparison with CO, is explained by step (18) where Hads assists the dissociation of chemisorbed NO molecule. In the presence of oxygen, Hads reacts more readily with Oads on Pd/Al2 O3 . Consequently, the residual H coverage is negligible and the dissociation of NO would occur conventionally on a nearest-neighbour vacant site according to step (12) as earlier illustrated for the NO + CO reaction [60]. Surprisingly, different kinetic features are obtained on Pd0 /CoOx /La2 O3 under net oxidising conditions with a positive apparent reaction order with respect to the partial pressure of oxygen [61] and a selectivity enhancement towards the formation of nitrogen. Those changes in catalytic performances cannot be explained by the abovementioned mechanism. As a matter of fact, surface oxidation and reconstruction, which could take after O2 exposure as well as the involvement of NO2 as intermediates could be considered for explaining the positive effect of oxygen on the rate of NO conversion into N2 [34]. O2 + 2∗ ⇔ 2Oads

(27)

NOads + Oads → NO2ads

(28)

NO2ads + NO2ads → N2 O4ads

(29)

N2 O4ads → N2 + 4Oads

(30)

2.3. Selective reduction by hydrocarbons Actually, there is no general consensus on the reaction pathways leading to the formation of nitrogen and nitrous oxide during the reduction of NO by hydrocarbon. It seems obvious that the reaction conditions (gas composition and temperature) and the nature of hydrocarbons may strongly influence the reaction pathways involved in the formation of N2 and N2 O. The kinetics for the reduction of NO by saturated and unsaturated hydrocarbons has been extensively investigated by Burch et al. [62–64]. These authors identified two different kinetic regimes for the reduction of NO by C8 H18 depending on the relative

Formation of N2 O During NOx Conversion

301

concentration [C8 H18 ]/[NO] [62]. At low [C8 H18 ]/[O2 ] values, the competition between both reactants is largely in favour of oxygen. Consequently, on a surface saturated by adsorbed O atoms, the formation of NO2 prevails. On the other hand the reverse tendency is observed at high [C8 H18 ]/[O2 ] ratio with a surface predominantly covered by C8 H18 -derived species. As mentioned before, the oxidation state of Pt is of great importance and may strongly affect the kinetic regime. Burch et al. [62] suggested that the local reduction of Pt surface at high C8 H18 -derived species coverage would favour the dissociation of NO according to the sequence (b) in Figure 10.3. On the contrary, this elementary step would be inhibited at high oxygen coverage where the formation of NO2 would predominantly occur then reacting by spill-over from Pt onto alumina with C8 H18 -derived species according to the sequence (a). Similar kinetic features were previously reported with C3 H8 as reducing agent [63]. However, the balance between both regimes also depends on the nature of the reducing agent. It is believed that the C−H bond breaking during the dissociative adsorption of hydrocarbons is the rate determining step. It was found that unsaturated hydrocarbons,

(a) + OH* – O* – * FAST

C8H18 (g)

+ * + o*

C8H17* + OH*

+24.5 O* –25.5* FAST +

NO(g)

–

* *

H2O(g)

8 CO2 (g) + 8.5 H2O(g)

NO* + O*

– O* +

NO2(g)

–

* *

NO2*

C8H18(g)

Pt metal AI2O3 support N2(g), N2O(g),

CxHy

CO2(g), H2O(g)

(b) C8H18(g)

CHx* + 2*

O2(g) NO(g)

8CO2(g) + 9H2O(g)

+

*

–

*

2O* NO*

+

*

N* + O* + N* – 2*

– 2*

N2(g) N2O(g)

Figure 10.3. Proposed mechanism for the C8 H18 −NO−O2 reaction in the low [C8 H18 ]/[O2 ] (a) and in the high [C8 H18 ]/[O2 ] concentration region (b). Reactions above the dotted line occur on the Pt surface, while reactions below occur on the alumina support (reproduced with permission from Ref. [62]).

302

Past and Present in DeNOx Catalysis

such as C3 H6 , more strongly adsorb than their saturated homologous and irreversibly at low temperature on Pt/SiO2 irrespective of the gas composition [64]. Such statements seem in relative agreement with Denton et al. [65] who concluded that NO acts as intermediate during the selective NOx reduction with C3 H6 , the primary role of oxygen would not to oxidise NO into NO2 . As a matter of fact different view points can be compared regarding the nature of the intermediates involved in the reduction of NOx by hydrocarbons in large O2 excess. Most of the recent investigations have been performed in O2 excess over a wide variety of hydrocarbons and catalyst compositions, which make difficult subsequent relevant comparisons. Again, the role of the state of the metal component on the lightoff performances has been underlined using a stoichiometric mixture C3 H6 + CO + NO + O2 particularly on Pd-based catalysts [66]. It was found that a PdO-like phase was unstable and transformed into Pd0 except when palladium is deposited on ceria−zirconia mixed oxides. In such a case, Pd clusters with an average oxidation state +I are stabilised. As a matter of fact, those palladium clusters would preserve a certain metallic-like character associated to the evidence of Pd−Pd chemical bonds; the average oxidation state +I would be likely induced by the presence of reactive oxygen in the bulk [66]. Such a stabilisation provides a source of oxygen for hydrocarbon oxidation. In a recent paper, Joubert et al. [67] compared different families of reducing agents on the same Pt/Al2 O3 catalysts in the presence of 5% O2 and 5% H2 O between 150 and 600 C with a gas hourly specific velocity of approximately 14 000 h−1 . These authors selected different criteria to investigate their efficiency such as the maximum NO conversion, the width of the operating window, the temperature at half and full conversion. As far as the selectivity for the production of N2 is concerned, they observed significant changes according to the structure of the molecules. Taking into account the above-mentioned criteria and the selectivity behaviour, cyclohexadiene, butane-1,3-diol, propan-1-ol and propanoic acid have been considered as the most efficient towards the selective reduction of NO to nitrogen among the selected hydrocarbon in Table 10.2. However, it is noticeable that their catalytic performances become significant mainly above 200 C. The same authors attempted to identify on the same catalyst the different routes, which may originate the formation of N2 and N2 O under lean conditions with propene as reducing agent. This investigation is interesting because it can offer complementary information regarding the intermediate obtained from the direct reaction between NO2 and derived-hydrocarbon species earlier proposed in the mechanism scheme of Burch et al. [62–64]. In fact, different organic compounds, which could be obtained from various reactions between nitro and nitroso compounds and intermediates characterised by different organic functions were examined. It was concluded that two different classes of intermediates may exist and could explain the selectivity changes towards the formation of N2 according to the following scheme represented in Figure 10.4. However, those authors did not indicate the nature of the active sites, if those reactions take place on metallic Pt or oxidic Pt sites and if the support is involved. In summary, a complex chemistry could be envisaged for explaining the formation of N2 and N2 O under lean conditions. In the particular case of N2 O, nitrosyl, dinitrosyl species as well as nitrates and nitrites species could be considered as intermediates particularly under lean conditions.

Reducer

CH4 C2 H4 C2 H2 C2 H6 C3 H6 C3 H4 C3 H8 C6 H14

R

XNOx max (%) T ( C)

Width at half conversion XNOx max 2 ( C)

T50 ( C)

T100 ( C)

Efficiency (temperature integrated conversion) Xnox

Xn2

Xn2 o

Xno2

2.2 1.9 1.9 1.9 1.9 1.3 2.5 3.3

29 42 42 22 55 31 13 40

(225) (225) (250) (200) (225) (250) (200) (250)

105 75 90 115 90 90 55 75

550 200 225 410 210 225 365 210

650 225 250 550 225 250 500 250

80 105 92 67 153 75 18 84

75 42 56 64 73 47 17 39

5 63 36 3 80 28 1 45

294 216 206 328 198 157 263 203

3.4

37 (300)

130

225

250

no

90

20

168

3.8

65 (200)

125

165

200

221

90

131

203

5.2

70 (200)

160

165

200

292

219

73

181

5.2

70 (250)

125

220

250

209

119

90

183

Formation of N2 O During NOx Conversion

Table 10.2. Summary of the performances of hydrocarbons in NOx reduction over the 1% Pt/Al2 O3 catalyst

C6H12

C6H6

C6H8

C6H10 Reproduced with permission from reference [67]

303

304

Past and Present in DeNOx Catalysis NO + propene → Nitro and nitroso compounds Nitro

Ketones + N2O Nef

C NOH

Beckmann

Oximes Nitroso

–H2

Amides H2O

Isocyanates Acids + amines Amines

Amines

diazo

R – N2

+ alcohols N2 + nitrito + nitrato

Figure 10.4. Reaction scheme proposed for the reduction of NO by propene over Pt/Al2 O3 (200–250 C) (reproduced with permission from [68]).

3. NEW DEVELOPMENTS OF NON-NOBLE METAL-BASED CATALYSTS 3.1. Gold as alternative in DeNOx catalysis under lean burned conditions Promising results are reported at the lab-scale associated with the development of gold catalysts dispersed on reducible supports such as TiO2 . The most relevant illustration is provided by the removal of CO at relatively low temperatures. Actually the functioning mode of those catalysts is strongly debated on the effective role of the support for inducing geometrical and electronic modifications, which could originate the lowtemperature activity in the conversion of CO. Such developments also concerned NOx . For instance, Ilieva et al. [69] report a selective conversion of NO into N2 at 200 C on Au dispersed on CeO2 and CeO2 −Al2 O3 . Unfortunately, these experiments were performed in the absence of significant amounts of oxygen and SO2 . The selectivity enhancement observed has not been related to a better dissociation of NO but mainly to the formation of specific sites related to coordinated unsaturated gold surface atoms at the metal/support interface promoted by an increase in Au dispersion.

3.2. Potentialities of silver-based catalysts in automotive exhaust systems Undoubtedly, the most advanced results concerning the development of non-noble metalbased catalysts are related to the use of silver [70–78]. Recent investigations showed a significant rate enhancement of the simultaneous conversion of NOx and hydrocarbon under lean conditions in the presence of hydrogen. As illustrated in Figure 10.5, such

Formation of N2 O During NOx Conversion

305

100

Conversion of NO to N2 (%)

90 80

a) NO + C8H18 + H2 + O2

70

b) NO + C8H18 + O2

60 50 40 30

c) NO + H2 + O2

20 10 0 150

200

250

300

350

400

450

500

550

600

Temperature (°C)

Figure 10.5. Comparisons of the conversion obtained on Ag/alumina in combination with Pt oxidation catalyst in the presence of hydrogen. 500 ppm NO, 375 ppm C8 H18 , 1 vol.% H2 , 6 vol.% O2 , 10 vol.% CO2 , 350 ppm CO, 12 vol.% H2 O in He. GHSV = 60 000 h−1 (reproduced with permission from Ref. [72]).

a rate enhancement is not related to the direct NO + H2 reaction, which preferentially occurs over noble metals at low temperature. Also, this beneficial effect does not seem to be associated with the exothermicity of this reaction inducing local changes in temperature, which may enhance the desorption of strongly adsorbed species at low temperature. Another parameter is related to the selectivity, since no N2 O formation is detectable contrarily to noble metals where N2 O is the main N-containing product at low temperature. Actually, there are strong debates for explaining the effective role of hydrogen. Two different viewpoints are compared, which account for the solid, particularly the oxidation state of silver stabilised under reactive conditions, and the intermediates involved in the selective reduction of NO into N2 . However, both aspects are probably intimately linked. Metallic Ag0 species yield N2 O and enhance the combustion of hydrocarbons. In contrast, the positive effect of hydrogen was attributed exclusively to the formation of small charged metallic Agn + clusters [12,73]. Sazama et al. [12] evidenced a complex surface chemistry over those catalysts under reaction conditions related to the formation of monodentate and bidentate nitrates exhibiting different intrinsic reactivity towards hydrogen (see Figure 10.6). Surface reactions involving −CN bonded to Ag+ into isocyanate species (NCO) interacting with alumina for explaining the production of nitrogen is unclear. Nevertheless, this finding highlights the role of the support as previously shown for the reduction of NOx by hydrocarbons under lean conditions over supported noble metal-based catalysts. In this particular case, the mechanism proposals involved steps on the support. Such a point could be an important issue in developing the activity of supported silver-based catalysts particularly their resistance to the deactivation. The formation of those species would involve interactions between nitrates and nitrites with oxygenates produced from the partial oxidation of heavy or light hydrocarbons. The formation of acetate occurs, as well as formates and acrylates, particularly in the presence of NOx . The addition of hydrogen may affect the individual reaction steps assigned to

306

Past and Present in DeNOx Catalysis H2 off

H2 on

0.12

–1

2150 cm

0.08 0.010 1245 cm

–1

0.04

0.005

Absorbance

Absorbance

0.015

–1

1295 cm

0.000

0.00

–1

2230 cm

0

20

40

60

80

100

120

Time (min)

Figure 10.6. In situ Fourier transform infrared spectra of decane SCR-NO in the presence and absence of hydrogen on Ag/Al2 O3 at 200 C. Evolution of intensities of the bands characteristic for adsorbed species (monodentate nitrates 1245 cm−1 , bidentate nitrates 1295 cm−1 , −CN 2150 cm−1 and −NCO 2230 cm−1 . 1000 ppm NO, 6 vol.% O2 , 750 ppm decane, 0 or 1000 ppm H2 (reproduced with permission from Ref. [12]).

the formation of NO2 and acetates or other intermediates from the transformation of hydrocarbons. Kinetic features could be related to the fact that the formation of acetates is rate determining. By way of illustration, Richter et al. [75] assumed that molecular oxygen dissociates over small Ag0 particles formed by the reduction of appropriately sized Ag2 O clusters and promotes the oxidation of hydrocarbons to acetates and other oxygenates. According to Sazama et al. [12] the presence of Agn + clusters would explain the rate enhancement in the presence of H2 . The number of those clusters would be closely related to the extent of NO conversion. In fact, their stabilisation could be associated to the reducing effect of adsorbed CHx O-containing reaction intermediates. The hydrogen induced effect on the stabilisation of those silver clusters has been questioned in recent papers [12,76,77], which brought better knowledges on the specific role of H2 . Wichterlova et al. [76] show that H2 contrarily of CO promotes the activity, but both reducing agents favour the development of silver clusters. Consistently, they proposed that hydrogen itself may participate to the reaction. Presently, the ability of H2 to promote more reactive intermediate species towards the formation of N2 is questioned. At low temperature (200–300 C) it is suggested that H2 could directly involved the transformation of nitrites and nitrates. Their subsequent transformation into cyanides and isocyanates is a hypothetical route, which has been recently proposed by Bion et al. [77] with the transformation of Ag+ −CN to Al3+ NCO. Alternative mechanisms have been recently proposed [78,79] based on a kinetic investigation of NO reduction by n-octane under isothermal (200 C) and steady-state conditions in the presence of H2 . The authors built up a mathematical model based on supposed reaction pathways, which account for molecular adsorption of NO and CO and dissociative ones for H2 and O2 . The elementary steps, which have been considered for modelling their results are reported in Table 10.3. Interesting kinetic information can be provided by the examination of this mechanism scheme in particular the fast bimolecular

Formation of N2 O During NOx Conversion

307

Table 10.3. Mechanism proposed for depicting the H2 assisted hydrocarbon selective catalytic reduction of NO over Ag/Al2 O3 (reproduced with permission from Ref. [78])

1 2 3 4 5 6 7 8

CO + ∗ ≡ CO∗ NO + ∗ ≡ NO∗ H2 + 2∗ ≡ 2H∗ O2 + 2∗ ≡ 2O∗ C8 H18 + O∗ +∗ → C8 H17∗ + OH∗ C8 H17∗ + NO∗ →∗ C8 H17 NO∗ +∗ C8 H17∗ + O∗ → C8 H17 O∗ +∗ 2C8 H17 NO∗ → N2 + 2C8 H17 O∗

N1

N2

N3

N4

−16 2 0 16 2 2 0 1

−8 0 0 85 1 0 1 0

0 2 2 0 0 0 0 0

1 −8 0 2 0 0 05 75 0 1 0 1 0 0 0 0

2 18 0

1 9 0

0 0 0

0 0 2

1 9 0

0 −1 0

0 0 0

1 9 0

0 0 0 0 0 0 0

0 0 0 0 0 0 0

1 2 0 0 0 0 0

0 0 1 0 0 0 0

0 0 0 1 0 0 0

0 2 0 0 0 0 0

0 0 0 0 1 0 −1

0 0 0 1 0 1 1

fast

9 10 11

C8 H17 O∗ + 24O∗ −→ 8CO + 17OH∗ 2OH∗ ⇔ H2 O + O∗ NO∗ + H∗ → NOH∗ + ∗

12 13 14 15 16 17 18

NOH∗ + NOH −→ N2 + 2OH∗ OH∗ + H∗ → H2 O + 2∗ CO∗ + O∗ → CO2 + 2∗ C8 H17 NO∗ + NO∗ → N2 + C8 H17 O∗ + O∗ NO∗ + O∗ → NO2∗ +∗ C8 H17∗ + NO2∗ → C8 H17 NO∗ + O∗ NO2 +∗ ≡ NO2∗

fast

N5

N6

N7

N8

0 0 1 05 0 0 0 0

0 −8 1 1 0 0 05 7 0 1 0 0 0 0 0 0

2NO + 16O2 + 2C8 H18 → 18H2 O + 16CO + N2 [→NO + 8O2 + C8 Hl8 → 9H2 O + 8CO + (l/2)N2 ] 8.5O2 + C8 H18 → 8CO + 9H2 O 3 2NO + 2H2 → N2 + 2H2 O 4 CO + (1/2)O2 → CO2 5 2NO + 7.5O2 + C8 Hl8 → 9H2 O + 8CO + N2 6 H2 +(1/2)O2 → H2 O 7 NO + (1/2)O2 → NO2 8 NO2 + NO + 7O2 + C8 H18 → 9H2 O + 8CO + N2 1 2

reaction between two chemisorbed NOH species, which leads to the formation of N2 . Such an hypothesis gives arguments for discussion, which can be compared with those earlier reported for explaining the beneficial effect of hydrogen on the reduction of NO by H2 with the involvement of Hads in the dissociation step of NO according to step (18) (see Sect. 2.2).

3.3. Stability and SO2 tolerance of Ag-based catalysts Interestingly, the detrimental effect of SO2 on the catalytic performances of Ag/Al2 O3 could be lowered in the presence of hydrogen, which emphasise the fact that the use of H2 is relevant for catalytic after-treatment processes. In fact, Shimidzu et al. [80] show that 50 ppm SO2 significantly block active sites due to the formation of sulphates. In the presence of hydrogen, sulphate species are reduced and desorbed as SO2 or migrate to alumina surface to form Al−SO4 2− , which are relatively stable even in the presence of hydrogen at 350 C. Such a latter finding is relevant and may partly explain why no

308

Past and Present in DeNOx Catalysis

consensus is observed in the effective role of SO2 , which may act as an inhibitor [81] and sometimes as a promoter [82,83]. In fact, the related performances of silver-based catalysts seem to be in close connection with the nature of the reducing agent. By way of illustration [82], the extent of interaction between hydrocarbons and active sites probably determines the type of mechanism and the related role of SO2 . Various authors explained their observations in the light of previous findings brought by Burch and Watling [84] who previously found on Pt/Al2 O3 that reaction paths for the formation of nitrogen from the reduction of NOx may involve the support material according to the extent of interaction between hydrocarbons and Pt sites. The promotional effect observed by Angelidis et al. [82] seems to be directly associated with C3 H6 as reductant with no significant effect of the support. On the other hand a poisoning effect occurs in the presence of C3 H8 . This result was explained by an extensive sulphur accumulation on alumina, which prevents surface reactions between NO2 and C3 H8 -derived species at the metal support interface and/or on alumina. These results underline the low sensitivity of silver catalysts. Obviously, even in a favourable normative context regarding the constant decrease in sulphur content in refinery oil, the continuous and irreversible accumulation of sulphates over automotive exhaust catalysts will exercise site blocking. Such considerations open the discussion on further developments of silver-based catalysts for mobile and particularly for stationary sources.

3.4. Alternative to the conventional vanadia/titania-based catalysts for the selective reduction of NO by ammonia Presently the catalytic selective NOx reduction by ammonia is efficient and widespread through the world for stationary sources. The remarkable beneficial effect of O2 for the complete reduction of NO into nitrogen is usually observed between 200 and 400 C. However, such a technology is not applicable for mobile sources due to the toxicity of ammonia and vanadium, which composes the active phase in vanadia−titania-based catalysts. Main drawbacks related to storing and handling of ammonia as well as changes in the load composition with subsequent ammonia slip considerably affect the reliability of such a process. On the other hand, the use of urea for heavy-duty vehicles is of interest with the in situ formation of ammonia. The replacement of vanadia-based catalysts in the reduction of NOx with ammonia is of interest due to the toxicity of vanadium. Tentative investigations on the use of noble metals in the NO + NH3 reaction have been nicely reviewed by Bosch and Janssen [85]. More recently, Seker et al. [86] did not completely succeed on Pt/Al2 O3 with a significant formation of N2 O according to the temperature and the water composition. Moreover, 25 ppm SO2 has a detrimental effect on the selectivity with selectivity towards the oxidation of NH3 into NO enhanced above 300 C. Supported copper-based catalysts have shown to exhibit excellent activity for NOx abatement. Recently Suarez et al. and Blanco et al. [87,88] reported high performances of CuO/NiO−Al2 O3 monolithic catalysts with NO/NO2 = 1 at low temperature. Different oxidic copper species have been previously identified in those catalytic systems with Cu2+ , copper aluminate and CuO species [89]. Subsequent additions of Ni2+ in octahedral sites of subsurface layers induce a redistribution of Cu2+ with a surface copper enrichment. Such redistribution

Formation of N2 O During NOx Conversion

309

may induce some significant changes on the catalytic performances particularly on the selectivity. Suarez et al. [87] mentioned that numerous investigations dealt with the reduction of NO but very few consider N2 O formation as a product of the selective catalytic reduction (SCR) reaction even though it is a strong greenhouse gas and contributes to the depletion of the stratospheric ozone layer. Previous investigations shown that [NOx ]/[NO] and [NH3 ]/[NOx ] ratios affect the selectivity towards the formation of nitrogen, optimal values of, respectively, 0.5 and 0.8 being obtained at 200 C on a typical acid nitric SCR unit. Suarez et al. investigated the role of the support on the selectivity behaviour. They explained the more accentuated formation of N2 O over Cu/TiO2 due to the involvement of the following reaction 3NO = NO2 + N2 O. Mizumoto et al. [90] suggested that the formation of N2 O involved reduced copper sites. Such suggestion can be qualitatively compared with previous statements on supported Ag catalysts where similarly the segregation of Ag0 clusters inhibits the formation of N2 . Yamagushi et al. [91] also argued that the formation of N2 O prevails on metallic Pt sites on Pt/Al2 O3 via a bimolecular reaction between NHxads and NOads . The study of Suarez pointed out the role of the support in the formation of N2 O in spite of complex features regarding this specific aspect. These authors concluded that the formation of the undesired oxidation of ammonia does not originate the formation of nitrous oxide, which would occur more probably via a bimolecular reaction between NO3 − ads and NHxads on Cu2+ .

4. POTENTIAL APPLICATION OF PEROVSKITES IN THREE-WAY CONDITIONS Perovskite materials ABO3 may be attractive due to lower costs and flexibility of their composition. Those materials can tolerate significant substitution and non-stoichiometry. However, their use for low temperature catalytic applications may be questionable due to usual low specific surface areas and sometimes to the partial segregation of impurities at the surface, which may affect their performances. Recent developments in the synthesis of those materials leads to significant improvements for controlling their textural and structural properties. Three-dimensionally ordered macroporous perovskite solids can be obtained by a colloidal crystal templating method, which allows specific surface area of about 61 m2 g−1 for LaFeO3 after calcination at 500 C [92]. Such a temperature is relatively low compared to the usual temperature conditions for obtaining the typical structure of perovskite. Reactive grinding methods may lead after calcination at 200 C to the perovskite structures with substantial higher specific surface area [93]. Combustion method can be profitably used to obtain large specific surface areas and seems to be particularly appropriate for the dispersion of LaMnO3 on conventional alumina support [94]. However, such a solution could not be obvious according to the nature of the precursors of the perovskite, which may react with alumina. In addition, conventional -alumina does not exhibit the optimal pore structure, which would lead to highly dispersed perovskite structures (Figure 10.7). Uenishi et al. [95] investigated the redox behaviour of palladium at start up in the perovskite-type structure LaFePdOx . An interesting behaviour is reported due to their self regenerative function, which provides high catalytic performances under cycling

310

Past and Present in DeNOx Catalysis

A

B O : A-site;

: B-site;

: Oxide

Figure 10.7. Structure of the perovskite ABO3 .

conditions with successive reductive and oxidative atmospheres. These authors characterised isolated Pd species inside the perovskite structure with an unusual oxidation state +III and/or +IV. Under reductive conditions, small Pd0 particles segregate at the surface and may be reversibly transformed into well-dispersed Pd(+III) species during exposure under lean conditions. Such behaviour promotes the catalytic performances in terms of activity and selectivity and also the durability due to the fact that thermal sintering reactions, which usually take place in running conditions over conventional alumina support are inhibited in the case of perovskite supports (Figure 10.8).

HC CO NOx

50% conversion temperature (°C)

250 260 270 280 290

Conversion (%)

100

240

HC CO NOx

95

90

85

300 310

Before aging

After aging

LaFeCoPdO3

Before aging

After aging

LaFePdO3

80

Before aging

After aging

LaFeCoPdO3

Before aging

After aging

LaFePdO3

Figure 10.8. Comparative performances under simulated conditions (reproduced with permission from Ref. [96]).

Formation of N2 O During NOx Conversion

311

5. HYDROGEN AS POTENTIAL REDUCING AGENT FOR THE SELECTIVE CONVERSION OF NO INTO NITROGEN FOR STATIONARY AND MOBILE SOURCES 5.1. Flexibility in the use of H2 Taking into account previous normative and environmental aspects, the use of hydrogen could be an interesting alternative for lowering the atmospheric pollutant emissions from mobile and stationary sources. Actually, different strategies involving the use of H2 develop and can be implemented. Reformed exhaust gas recirculation producing hydrogen-enriched gas fed back to the engine can have several beneficial effects such as lowering NOx emissions from gasoline engines and promoting auto ignition of various hydrocarbon fuels [97]. A reformer can be setup in an exhaust gas recirculation loop and can produce in the stream 5–20% H2 at relatively low inlet temperature (350–400 C) [98]. The simultaneous formation of CO seems to enhance the beneficial effect of H2 contrarily to the strong inhibiting effect usually reported in solid polymer fuel cells. However, the reforming of exhaust gas over precious metals could be questionable due to their low sulphur tolerance and the achievement of low sulphur gasoline should not lead to significant improvement due to the cumulative nature of poisoning effects. Hydrogen may also concern after-treatment systems with recent extensive developments. Actually, there is a strong debate on the most efficient system, which can be implemented on car passengers. Up to now, the NOx storage and reduction (NSR) catalysts for lean-burn gasoline engines appears as the most suitable technology in spite of significant thermodynamic limitations. The usual temperature for obtaining optimal performances is around 400 C [99]. The improvement of the fuel economy without altering the efficiency of those systems may be actually considered as one of the main target. Taking of this, the finding of Takahashi et al. [100] may offer significant practical developments. They found that H2 was the best reducing agent below 400 C according to the sequence H2 > CO > C3 H6 , and NOx storage sites can be fully restored by supplying an adequate amount of H2 . The H2 generation occurs more probably through the watergas-shift reaction than the steam reforming, which could explain the higher efficiency of CO compared to that of C3 H6 . NSR technology suffers from significant drawbacks mainly related to thermodynamic limitations. The use of H2 may attenuate those effects. An investigation of NOx reduction mechanism on Pt−Ba/-Al2 O3 lean NOx trap clearly shows that the preliminary decomposition of the adsorbed NOx species is not a determining pathway for achieving the subsequent reduction step of NOx into N2 in the presence of H2 [101]. These authors observed completely different catalytic features with a direct reduction of stored nitrates with H2 taking place at temperatures as low as 140 C. The conventional selective reduction of NOx for car passengers still competes but the efficient SCR with ammonia on V2 O5 /TiO2 for stationary sources is not available for mobile sources due to the toxicity of vanadium and its lower intrinsic activity than that of noble metals, which may imply higher amount of active phase for compensation. As illustrated in Figure 10.9, such a solution does not seem relevant because a subsequent increase in vanadium enhances the formation of undesirable nitrous oxide at low temperature. Presently, various attempts for the replacement of vanadium did not succeed regarding the complete conversion of NO into N2 at low temperature. Suarez et al. [87] who investigated the reduction of NOx with NH3 on CuO-supported monolithic catalysts

312

Past and Present in DeNOx Catalysis 150

80

120

N2O outlet, ppm

NO conversion, vol %

100

210°C 250°C

60

90

280°C 310°C

40

60

20

30 0

0 0

4

2

6

8

10

wt.% V2O5

Figure 10.9. NO conversion and N2 O formation as a function of the vanadia loading at different temperatures. Feed composition: [NO] = [NH3 ] = 1000 ppm, [O2 ] = 3 vol.%, [N2 ] = balance. Operating conditions: GHSV (NTP) = 20 000 h−1 , LV = 066 Nms−1 , P = 0.12 MPa [5] (reproduced with permission from Ref. [102]).

(a)

(b)

150 100

60

50 0

50 0

2

4

6

8

[CuO], wt.%

10

12

N2O in outlet, ppm

200 70

NH3 in outlet, ppm

NOx conversion, %

50

250

80

40 30 20 10 0 0

2

4

6

8

10

12

[CuO], wt.%

Figure 10.10. (a) NOx conversion and NH3 slip, and (b) N2 O in the outlet as a function of the CuO content (filled dark triangle, open triangle) CuO/−Al2 O3 and (filled dark circle, open circle) CuO/TiO2 monolithic catalysts. Feed compositions: [NOx ] = 1000 ppm, [NO]/[NOx ] = 0.54, [NH3 /NOx ] = 0.77, [O2 ] = 4 vol.%, [N2 ] = balance. Operating conditions: GHSV (NTP) = 10 200 h−1 , L = 098 Nms−1 , T = 200 C, P = 120 kPa (reproduced with permission from Ref. [87]).

observed significant N2 O formation at 200 C depending on the nature of the support, the formation being much more favoured on TiO2 than on alumina (see Figure 10.10). They suggested that the formation of N2 O would involve a reaction between adsorbed nitrates and NHx species. Now, regarding the SCR with hydrocarbons in O2 excess, numerous investigations have shown a low activity below 200 C. However, it was found that H2 can promote the reduction of NO below 200 C on molybdenum and sodium-modified Pt/SiO2 and Pt/Al2 O3 catalysts [103]. Such a promotional effect also observed on silver-based catalysts originates extensive investigations in this field and offers new perspectives in the developments of non-noble metal-based catalysts. However, further developments of that variety of catalysts seem to be questionable due to their low sulphur tolerance.

Formation of N2 O During NOx Conversion

313

5.2. Critical aspects in the use of noble metals for hydrogen applications Up to now the main drawback in the use of H2 is its poor selectivity with a strong attenuation of the conversion of NO with an increase in temperature as illustrated in Table 10.4. In addition, N2 O is the main N-containing product at the maximum NO conversion [104]. Recent developments showed promising results over platinum and palladium-based catalysts after deposition on reducible materials [104–106] (see Figures 10.11 and 10.12). Further improvements might be expected on Pd-based catalysts supported on those

Table 10.4. Comparison of the catalytic performances of various supported noble metal-based catalysts (reproduced with permission from Ref. [104]) T = 100 C

Catalyst

Pd/TiO2 Pd/Al2 O3 Pd/MgO Pd/SiO2 Pt/TiO2 Pt/Al2 O3 Pt/MgO

T = 300 C

N2 1

N2 O2

NO3

N2 1

N2 O2

NO3

217 27 78 45 105 66 54

258 0 54 23 401 556 382

475 27 132 68 506 622 436

274 21 81 102 06 28 14

175 0 37 54 0 0 0

449 21 118 156 06 28 14

0.1 vol.% NO, 0.3 vol.% H2 , 5 vol.% O2 , 10% H2 O, space velocity = 20 000 h−1 ml gcata Percentage conversion of NO into N2 2 Percentage conversion of NO into N2 O 3 Overall conversion

1

(b)

(a)

0.30

0.36

Pt /La0.5Ce0.5MnO3 Pt /Al2O3

0.27 0.18 0.09 0.00

80 120 160 200 240 280 320 360 400

Temperature (°C)

RN2O (μ moles/g.s)

RN2 (μ moles/g.s)

0.45

0.24

Pt /La0.5Ce0.5MnO3 Pt /Al2O3

0.18 0.12 0.06 0.00

80 120 160 200 240 280 320 360 400

Temperature (°C)

Figure 10.11. Effect of 5% H2 O in the feed stream on (a) integral rate of N2 and N2 O formation on 0.1% Pt/La05 Ce05 MnO3 (filled dark circle) and Pt/Al2 O3 (filled dark triangle). Reaction conditions: NO = 0.25%, H2 = 10%, O2 = 5%, H2 O = 5%, W = 015 g, GHSV = 80 000 h−1 (reproduced with permission from Ref. [105]).

314

Past and Present in DeNOx Catalysis 100

100

Pt /La0.7Sr0.2Ce0.1FeO3

60

60

XNO SN 2

40

40

20 0

80

SN2 (%)

XNO (%)

80

20

100

150

200

250

300

350

400

0

Temperature (°C)

Figure 10.12. Effect of 5% H2 O in the feed stream on NO conversion (filled dark circle) and N2 selectivity (filled dark triangle) as a function of temperature on 0.1% Pt/La07 Sr02 Ce01 FeO3 . Reaction conditions: NO = 0.25%, H2 = 10%, O2 = 5%, H2 O = 5%, W = 015 g, GHSV = 80 000 h−1 (reproduced with permission from Ref. [106]).

typical reducible materials, such as perovskite structures, because a better selectivity towards the formation of nitrogen on Pd than on Pt has been earlier reported [107]. Presently, the effective role of reducible materials is strongly debated due to the fact that the reaction mechanisms earlier proposed involve steps both on the support and on the metal. Alternately, the nature of the metal-support may strongly modify the adsorptive properties of noble metals further altering the relative rates of elementary steps taking place over noble metal particles.

5.3. Additives to noble metals Among the different parameters, which can influence the reaction, the acidic/basic properties of the support as well as the oxidation state of noble metals under lean conditions may drastically alter the selectivity. Previous investigations suggested that the formation of N2 O on Rh-based catalysts would involve oxidic rhodium species, on the other hand the formation of nitrogen would take place on metallic sites. Earlier investigations found that the addition of Na and Mo could be a promising issue with a subsequent stabilisation of the metallic character of noble metals [108]. However, the control of Na deposition seems to be an important parameter in determining the catalytic performances. By way of illustration, Machida and Watanabe [109] found that an increase in Na loading improves the activity in NO conversion but has a detrimental effect on the selectivity towards the production of nitrogen. Subsequent Na incorporation enhanced the formation of nitrogen according to the mode of incorporation of Na and Pt via conventional successive or coimpregnation. The former route would lead to preferential interactions between Na species surrounding Pt sites. A similar effect is reported by Burch and Coleman [103] with small addition of Na (0.27%) on Pt/SiO2 and Pt/Al2 O3 . As a matter of fact, the subsequent formation of Lewis acid sites would provide a positive effect on the conversion of NO into N2 up to a certain limit. At high Na loadings a deactivation is sometimes reported. The build-up of nitrites and nitrates species may

Formation of N2 O During NOx Conversion

315

originate a significant loss of conversion. The observation of Machida and Watanabe are not currently observed particularly at high Na loading, since Pt/ZSM5 retains a significant residual conversion of NO. Presently such peculiar catalytic properties are not well understood, the Brønsted acidity of the zeolithe could be involved. Alternative suggestions have been proposed by Gonçalves et al. [110] who investigated the synergy effect Pt and K in the catalytic reduction of NO and N2 O. This synergy effect could be related to non-competitive adsorptions of NO and N2 O, respectively, on active sites mainly composed of Pt and K. Consistently, the usual inhibiting effect on the subsequent reduction of N2 O would not occur significantly. Nevertheless, such an explanation could partly explain the results obtained by both authors. Transient kinetic experiments suggest that CO2 desorption over K and the dissociation of NO and/or N2 O would be rate controlling steps. A strong interaction between both metals may induce a charge transfer from K to Pt. Alkalis would be able to lower the work function of noble metals. A subsequent strengthening of the Pt−NO and Pt−N2 O bond would take place and enhance the N−O bond breaking in both adsorbed molecules. Such an electronic effect would explain the selectivity enhancement towards the formation of N2 . According to elementary steps (13–15), a subsequent introduction of K would increase the N coverage at the expense of NO on metallic sites. Consequently, the associative desorption of nitrogen according to step (13) would occur more readily.

5.4. Sulphur and water tolerance of perovskite-based catalysts for stationary sources Presently the effective role of sulphur additive is not well explained because sometimes activation or deactivation phenomena are observed. Such a versatile behaviour is wellillustrated over noble metal-based catalysts particularly when they are dispersed on perovskite supports [111]. The catalytic performances of a prereduced Pt/LaCoO3 in H2 at 450 C are illustrated in Figure 10.13a. After preactivation in H2 subsequent bulk and surface characterisation highlighted an extensive reduction of the perovskite (b) 100

100

100

100

80

80

80

80

60

60

60

60

40

40

40

40

20

20

20

20

0

0

0 0

200

400

Temperature (°C)

Conversion (%)

Conversion (%)

(a)

0 0

200

400

Temperature (°C)

Figure 10.13. Comparative temperature-programmed catalytic performances of prereduced Pd/CoOx /La2 O3 (a) and after ageing (b) at 500 C overnight in the reaction mixture (0.15 vol.% NO, 0.5 vol.% H2 , 3 vol.% O2 , 5 vol.% H2 O and 600 ppm SO2 ).

316

Past and Present in DeNOx Catalysis

leading to the preferential formation of well-dispersed Pt0 and CoOx species on La2 O3 (Pt/CoOx /La2 O3 . Two ranges of conversion of NO in the range 100–150 and up to 200 C are observable. The low temperature conversion curve of NO shows that the competition between the NO + H2 and H2 + O2 reactions is largely in favour of the latter one. Remarkable results have been obtained after ageing. Of course, the lowtemperature conversion of NO disappears due to an extensive oxidation of Pd. However, the competition becomes largely in favour of the NO + H2 reaction above 200 C with a selectivity enhancement towards the formation of nitrogen. Clearly, it has been found that the presence of SO2 originates a strong deactivation at low temperature when metallic noble metal particles catalyse the reduction of NO. Under oxidative conditions the improvement on the catalytic performance has been mainly related to significant structural changes associated to the development of amorphous solids and a partial reconstruction of the perovskite as exemplified in Figure 10.14. In those conditions, the formation of oxidic Pt species may also occur. Such tendencies on the catalytic performances have been observed irrespective of the nature of the support and of the noble metals. Nevertheless, they are much more pronounced when Pd is initially deposited on LaCoO3 with a much higher conversion level obtained after ageing. It is usually difficult to discuss unambiguously on the role of the formation of sulphate, which may explain the deactivation. Their formation can equally occur on the support and on the noble metals. The poisoning effect of SO2 has been reported by Qi et al. on Pd/TiO2 /Al2 O3 [112]. However, in the presence of water, the stabilisation of hydroxyl groups could inhibit the adsorption of SO2 [113]. Burch also suggested a possible redispersion of palladium oxide promoted by the formation of hydroxyl species [114]. Such tentative interpretations could correctly explain the tendencies that we observed irrespective to the nature of the supports, which indicate an improvement in the conversion of NO into N2 at high temperature. Nevertheless, the accentuation of those tendencies particularly on prereduced perovskite-based catalysts could be in connection with structural modifications associated with the reconstruction of the rhombohedral structure of (a)

(b)

Intensity, a.u.

Intensity, a.u.

LaCoO3

25

30

35

40

45

2θ (°)

50

55

60

65

14 19 24 29 34 39 44 49 54 59 64 69 74

2θ (°)

Figure 10.14. Comparison between theoretical and experimental XRD patterns recorded on prereduced Pt/CoOx /La2 O3 (a) and after ageing in the reaction conditions (b) Theoretical XRD patterns of La2 O3 is reported in Figure 10.14a and of La(OH)3 on aged catalyst in Figure 10.14b.

Formation of N2 O During NOx Conversion

317

LaCoO3 . In such a situation, tentative explanations on possible changes in the chemical environment of oxidic palladium species could be proposed with a subsequent stabilisation of oxidic Pd species in the perovskite matrix. The formation of well-dispersed Pdn+ entities could further inhibit the formation of sulphate generally observed on PdO at high temperature [113,115].

5.5. Self-reconstruction and stabilisation of highly dispersed Pd species under reactive conditions Surface reconstruction has been earlier observed and reported in the literature [116]. Sequential reductive and oxidative thermal treatment usually leads to bulk transition from CoOx + La2 O3 to LaCoO3 , respectively. On the other hand, the restoration of the perovskite structure is not observed under severe conditions at higher temperature. In those temperature conditions, the sintering of Co crystallites leads to irreversible redox cycle with the preferential formation of Co3 O4 under lean conditions. In situ X-ray diffraction (XRD) experiments of the NO + H2 + O2 performed on prereduced Pd/LaCoO3 into Pd0 /CoOx /La2 O3 showed significant bulk reconstructions above 500 C associated to the formation of LaCoO3 [111]. Parallel to this observation no bulk Pd species were observed in the temperature range of this study, which could suggest the involvement of thermal sintering. Subsequent X-ray photoelectron spectroscopy (XPS) experiments evidenced that those structural changes start at the surface at much lower temperature (300 C). Data reported in Table 10.5 indicate a complete reoxidation of Co into Co (+III) species, which highlight the reconstruction of the perovskite. Parallel to those observations, Figure 10.15 reveals significant changes of the binding energy level of Pd with an irreversible oxidation leading to Pd(+III) species (B.E. of 337.5 eV) assigned to oxidic palladium species stabilised inside the perovskite structure [94,118,119]. Irreversible changes under lean conditions would lead to a redispersion of the Pd phase with a strong interaction with

Table 10.5. X-ray photoelectron spectroscopy analysis of Pd/LaCoO3 after exposure under controlled atmosphere (reproduced with permission from Ref. [117]) Reaction step

Binding energy (eV) Co2p

Fresh After reduction in H2 at 500 C 30 min at 25 C 30 min at 200 C 30 min at 300 C 30 min at 500 C

N1s

779.7 777.8/780 780.1 779.8 780

780.3

403.4 403.2/407 403.3/407

Surface composition (%) Pd3d5/2

Co(III)

336.2 335.1

100

335.3 335.2 335.2 336.2 337.5 336.2 337.5

51 100 100

Co(II)

Co0

39

61

100 49

318

Past and Present in DeNOx Catalysis

(a)

(b) 407.2

337.5 336.2 335.1

346

341

336

Binding energy/eV

403.4

h g h g f e d c

e

b

d

a

c b 395

331

f

415

410

405

400

Binding energy/eV

Figure 10.15. X-ray photoelectron spectroscopy measurements on prereduced Pd/LaCoO3 into Pd0 /CoOx /La2 O3 after successive exposures under reaction conditions 0.15 vol.% NO, 0.5 vol.% H2 and 3 vol.% O2 . Photopeaks Pd 3d and N 1s recorded on the calcined sample in air at 400 C (a); after reduction in H2 at 500 C (b); after exposure under reaction conditions at 25 C (c); at 200 C for 30 min (d); at 300 C for 30 min (e); at 300 C for 2 h (f); at 500 C for 30 min (g) and at 500 C for 2 h (h) (reproduced with permission from Ref. [117]).

the perovskite structure. Interestingly, it was found that successive redox cycle typically encountered in three-way conditions may lead during these surface reconstructions to the disappearance of impurities related to the segregation of monometallic oxides at the surface, which usually take place during conventional preparation procedures involving thermal treatment in severe conditions [95,96,118]. Those investigations underline the potentialities of perovskite-based catalysts in DeNOx catalysis due to the fact that the structure may preserve a high dispersion of noble metals.

6. CONCLUSION The formation of N2 O may represent in the future a challenging aspect particularly under lean conditions for mobile sources. Basically, the minimisation of N2 O particularly during the cold start engine implies a selective catalyst at low conversion and/or a subsequent reduction of N2 O, which can occur readily at low temperature. Both aspects can be equally considered in three-way conditions where a complete NOx conversion can be achieved near stoichiometric conditions. On the other, they seem to be much more debatable under lean conditions because of lower NO conversion levels accompanied with a slow successive reduction of N2 O. In such a case, novel selective catalysts have to be envisaged parallel to the use of more efficient reducing agents such as hydrogen. In this sense, the development of non-noble metal catalysts such as silver is promising but most of the results have omitted the presence of significant amounts SO2 , which usually strongly deactivate the catalysts. Based on these considerations the development of perovskite-based catalysts could be an alternative due to their ability to stabilised well-dispersed active sites more resistant to deactivation and to their tolerance to poisoning effects. Consequently, those materials

Formation of N2 O During NOx Conversion

319

could be profitably used for reducing the amount of noble metal currently used for such applications and for enhancing their intrinsic activity compared to conventional alumina support where only weak interactions between noble metals and the support take place.

REFERENCES [1] Van Deyck, F., Weiss, F.J., Verbeck, R. et al. (1985) Removal of nitrous oxide from gases containing hydrogen, nitric oxide and nitrous oxide, US Patent 4507271. [2] Nirisen, Ø., Schöffel, K., Waller, D. et al. (2002) Catalyst for decomposing nitrous oxide and method for performing processes comprising formation of nitrous oxide, World Patent 0202230 A1. [3] Neveu, B. and Hamon, C. (1999) Method for reducing nitrous oxide in gases and corresponding catalyst, World Patent 9964139. [4] Ramirez, J.P. (2007) Prospect of N2 O emission regulations in the European fertilizer industry, Appl. Catal. B 70, 31. [5] Kieger, S., Delahay, G. and Coq, B. (2000) Influence of co-cations in the selective catalytic reduction of NO by NH3 over copper exchanged faujasite zeolites, Appl. Catal. B 25, 1. [6] Forzatti, P. (2000) Environmental catalysis for stationary applications, Catal. Today 62, 51. [7] Coq, B., Mauvezin, M., Delahay, G. et al. (2000) The simultaneous catalytic reduction of NO and N2 O by NH3 using an Fe-zeolite-beta catalyst, Appl. Catal. B 27, 193. [8] Guzman-Vargas, A., Delahay, G. and Coq, B. (2003) Catalytic decomposition of N2 O and catalytic reduction of N2 O and N2 O + NO by NH3 in the presence of O2 over Fe-zeolite, Appl. Catal. B 42, 369. [9] Centi, G., Cerrato, G., D’Angelo, S. et al. (1996) Catalytic behavior and nature of active sites in copper-on-zirconia catalysts for the decomposition of N2 O, Catal. Today 27, 265. [10] Monterra, C., Gianello, E., Cerrato, G. et al. (1998) Role of surface hydration state on the nature and reactivity of copper ions in Cu-ZrO2 catalysts: N2 O decomposition, J. Catal. 179, 111. [11] Farrauto, R.J. and Heck, R.M. (2000) Environmental catalysis into the 21st century, Catal. Today 55, 179. ˇ [12] Sazama, P., Capek, L., Drobná, H. et al. (2005) Enhancement of decane-SCR-NOx over Ag/alumina by hydrogen. Reaction kinetics and in situ FTIR and UV-vis study, J. Catal. 232, 302. [13] Wichterlová, B., Sazama, P., Breen, J.P. et al. (2005) An in situ UV-vis and FTIR spectroscopy study of the effect of H2 and CO during the selective catalytic reduction of nitrogen oxides over a silver alumina catalyst, J. Catal. 235, 195. [14] Ilieva, L., Pantaleo, G., Ivanov, I. et al. (2006) Gold catalysts supported on CeO2 and CeO2 -Al2 O3 for NOx reduction by CO, Appl. Catal. B 65, 101. [15] Deluga, G.A., Salge, J.R., Schmidt, L.D. et al. (2004) Renewable hydrogen from ethanol by autothermal reforming, Science 303, 993. [16] Bogdanchikova, N., Meunier, F.C., Avalos-Borja, M. et al. (2002) On the nature of the silver phases of Ag/Al2 O3 catalysts for reactions involving nitric oxide, Appl. Catal. B 36, 287. [17] Bion, N., Saussey, J., Haneda, M. et al. (2003) Study by in situ FTIR spectroscopy of the SCR of NOx by ethanol on Ag/Al2 O3 – Evidence of the role of isocyanate species, J. Catal. 217, 47. [18] Ueda, A., Nakano, T., Azuma, M. et al. (1998) Two conversion maxima at 373 and 573K in the reduction of nitrogen monoxide with hydrogen over Pd/TiO2 catalyst, Catal. Today 45, 135.

320

Past and Present in DeNOx Catalysis

[19] Costa, C.N., Savva, P.G., Andronikou, C. et al. (2002) An investigation of the NO/H2 /O2 (lean De-NOx  reaction on a highly active and selective Pt/La07 Sr02 Ce01 FeO3 catalyst at low temperature, J. Catal. 209, 456. [20] Engelmann-Pirez, M., Granger, P. and Leclercq, G. (2005) Investigation of the catalytic performances of supported noble metal based catalysts in the NO + H2 reaction under lean conditions, Catal. Today, 107, 315. [21] Cant, N.W., Angove, D.E. and Chambers, D.C. (1998) Nitrous oxide formation during the reaction of simulated exhaust streams over rhodium, platinum and palladium catalysts, Appl. Catal. B 17, 63. [22] Belton, D.N. and Schmieg, S.J. (1993) Oxidation of CO by NO over Rh(111), J. Catal. 144, 9. [23] Peden, C.H.F., Belton, D.N. and Schmieg, S.J. (1995) Structure sensitive selectivity of the NO-CO reaction over Rh(110) and Rh(111), J. Catal. 155, 204. [24] Permana, H., Ng, K.Y.S., Peden, C.H.F. et al. (1995) Effect of NO pressure on the reaction of NO and CO on Rh(111), J. Phys. Chem. 99, 16344. [25] Permana, H., Ng, K.Y.S., Peden, C.H.F. et al. (1996) Adsorbed species and reaction rates for NO-CO over Rh(111), J. Catal. 164, 194. [26] Root, T.W., Schmidt, L.D. and Fischer, G.B. (1983) Adsorption and reaction of nitric oxide and oxygen on Rh(111), Surf. Sci. 134, 30. [27] Gorte, R.J., Schmidt, L.D. and Gland, J.-L. (1990) Binding states and decomposition of NO on single crystal planes of Pt, Surf. Sci. 109, 367. [28] Wickam, D.T., Banse, B.A. and Koel, B.E. (1991) Adsorption of nitrogen dioxide and nitric oxide on Pd(III), Surf. Sci. 243, 83. [29] Hecker, W.C. and Bell, A.T. (1985) Reduction of NO by H2 over silica-supported rhodium: Infrared and kinetic studies, J. Catal. 92, 247. [30] Hanoaka, T., Kim, W.Y., Kishida, M. et al. (1997) Enhancement of CO hydrogenation activity of Rh/SiO2 with low rhodium content, Chem. Lett. 645. [31] Dictor, R. (1988) An infrared study of the behaviour of CO, NO and CO + NO over Rh/Al2 O3 catalysts, J. Catal. 109, 89. [32] Chafik, T., Kondarides, D.I. and Verykios, X.E. (2000) Catalytic reduction of NO by CO over rhodium catalysts: 1. Adsorption and displacement characteristics investigated by in situ FTIR and transient-MS techniques, J. Catal. 190, 446. [33] Kondarides, D.I., Chafik, T. and Verykios, X.E. (2000) Catalytic reduction of NO by CO over rhodium catalysts: 3. The role of surface isocyanate species, J. Catal. 193, 303. [34] Nieuwenhuys, B.E. (2000) The surface science approach toward understanding automotive exhaust conversion catalysis at the atomic level, Adv. Catal. 44, 35. [35] Burch, R. and Coleman, M.D. (2002) An investigation of promoter effects in the reduction of NO by H2 under lean-burn conditions, J. Catal. 208, 435. [36] Shelef, M. and Graham, G.W. (1994) Why rhodium in automotive three-way catalyst? Catal. Rev. Sci. Eng. 36, 433. [37] Granger, P., Dathy, C., Lecomte, J.J. et al. (1998) Kinetics of the NO and CO reaction over platinum catalysts: 1. Influence of the support, J. Catal. 173, 304. [38] Granger, P., Lecomte, J.J., Dathy, C. et al. (1998) Kinetics of the CO + NO reaction over rhodium and platinum-rhodium on alumina: II. Effect of Rh incorporation to Pt, J. Catal. 175, 194. [39] Granger, P., Delannoy, L., Leclercq, L. et al. (1998) On the effect of deactivation on the kinetics of CO oxidation by NO over Pt-Rh catalysts, J. Catal. 177, 147. [40] Granger, P., Delannoy, L., Lecomte, J.J. et al. (2002) Kinetics of the CO + NO reaction over bimetallic platinum-rhodium on alumina: Effect of ceria incorporation into noble metals, J. Catal. 207, 202. [41] Zaera, F. and Gopinath, C.S. (2000) Evidence for a N2 O intermediate in the catalytic reduction of NO to N2 on rhodium surfaces, Chem. Phys. Lett. 332, 209.

Formation of N2 O During NOx Conversion

321

[42] Dujardin, C., Mamede, A.-S., Payen, E. et al. (2004) Influence of the oxidation state of rhodium in three-way catalysts on their catalytic performances: An in situ FTIR and MS study, Top. Catal. 30–31, 347. [43] Daganello, F. and Martorana, A. (2002) Phase analysis and oxygen storage capacity of ceria-lanthana-based TWC promoters prepared by sol-gel routes, J. Sol. State Chem. 163, 527. [44] Martinez-Arias, A., Fernandez-Garcia, M., Iglesia-Juez, A. et al. (2002) Influence of thermal sintering on the activity for CO-O2 and CO-O2 -NO stoichiometric reactions over Pd/(Ce,Zr)Ox /Al2 O3 catalysts, J. Soria, Appl. Catal. B 38, 151. [45] Fornasiero, P., Kaspar, J. and Graziani, M. (1997) Redox behavior of high surface area Rh-loaded Ce05 Zr05 O2 mixed oxide, J. Catal. 167, 576. [46] Holles, J.H., Switzer, M.A. and Davis, R.J. (2000) Influence of ceria and lanthana promoters on the kinetics of NO and N2 O reduction by CO over alumina-supported palladium and rhodium, J. Catal. 190, 247. [47] Oh, S.H. (1990) Effects of cerium addition on the CO-NO reaction kinetics over aluminasupported rhodium catalysts, J. Catal. 124, 477. [48] Roy, S., Marimuthu, A., Hedge, M.S. et al. (2007) High rates of NO and N2 O reduction by CO, CO and hydrocarbon oxidation by O2 over nano crystalline Ce098 Pd002 O2− : Catalytic and kinetic studies, Appl. Catal. B 71, 23. [49] Zhdanov, V.P. (1996) Does the CO-N2 O subreaction play an important role in the NO-CO reaction on Rh? J. Catal. 162, 147. [50] Cho, B.K. (1996) Emphasizing the mechanistic importance of intermediate N2 O + CO reaction in the overall NO + CO reaction system, J. Catal. 162, 149. [51] Granger, P., Dujardin, C., Paul, J.F. et al. (2005) An overview of kinetic and spectroscopic investigations of three-way catalysts: Mechanistic aspects of the CO+NO and the CO+N2 O reactions, J. Mol. Catal. 228, 241. [52] Granger, P., Malfoy, P. and Leclercq, G. (2004) Kinetics of the CO + N2 O reaction over noble metals: II. Rh/Al2 O3 and Pt-Rh/Al2 O3 , J. Catal. 223, 142. [53] Dhainaut, F., Pietrzyk, S. and Granger, P. (2007) Kinetics of the NO + H2 reaction over supported noble metal based catalysts: Support effect on their adsorption properties, Appl. Catal. B 70, 100. [54] Dhainaut, F., Pietrzyk, S. and Granger, P. (2007) Kinetic investigation of the NO reduction by H2 over noble metal based catalysts, Catal. Today 119, 94. [55] Frank, B., Emig, G. and Renken, A. (1998) Kinetics and mechanism of the reduction of nitric oxides by H2 under lean-burn conditions on a Pt-Mo-Co/-Al2 O3 catalyst, Appl. Catal. B 19, 45. [56] Burch, R., Shestov, A.A. and Sullivan, J.A. (1999) A steady-state isotopic transient kinetic analysis of the NO/O2 /H2 reaction over Pt/SiO2 catalysts, J. Catal. 188, 69. [57] Burch, R., Shestov, A.A. and Sullivan, J.A. (1999) A transient kinetic study of the mechanism of the NO + H2 reaction over Pt/SiO2 catalysts, J. Catal. 186, 353. [58] Shestov, A.A., Burch, R. and Sullivan, J.A. (1999) A transient kinetic study of the mechanism of the NO + H2 reaction over Pt/SiO2 catalysts: 2. Characteristic features of SSITKA profiles, 186, 362. [59] Shustorovich, E. (1990) The bond-order conservation approach to chemisorption and heterogeneous catalysis: applications and implications, Adv. Catal. 37, 101. [60] Dhainaut, F., Pietrzik, S. and Granger, P. (2007) NO + H2 reaction on Pd/Al2 O3 under lean conditions: kinetic study, Top. Catal. 42–43, 135. [61] Dhainaut, F. (2006) Réduction des oxides d’azote par l’hydrogène sur des catalyseurs à base de métaux nobles supportés, PhD thesis, University of Lille, France. [62] Burch, R., Fornasiero, P. and Southward, B.W.L. (1999) An investigation into the reactivity, deactivation and in situ regeneration of Pt-based catalysts for the selective reduction of NOx under lean burn conditions, J. Catal. 182, 234.

322

Past and Present in DeNOx Catalysis

[63] Burch, R. and Watling, T.C. (1997) Kinetics and mechanism of the reduction of NO by C3 H8 over Pt/Al2 O3 under lean-burn conditions, J. Catal. 169, 45. [64] Burch, R. and Sullivan, J.A. (1999) A transient kinetic study of the mechanism of the NO/C3 H6 /O2 reaction over Pt-SiO2 catalysts, J. Catal. 182, 489. [65] Denton, P., Giroir-Fendler, A., Schuurman, Y. et al. (2001) A redox pathway for selective NOx reduction: stationary and transient experiments performed on a supported Pt catalyst, Appl. Catal. A 220, 141. [66] Fernandez-Garcia, M., Iglesia-Juez, A., Martinez-Arias, A. et al. (2004) Role of the state of the metal component on the light-off performance of Pd-based three-way catalysts, J. Catal. 221, 594. [67] Joubert, E., Courtois, X., Marécot, P. et al. (2006) NO reduction by hydrocarbons and oxygenated compounds in O2 excess over Pt/Al2 O3 catalyst. A comparative study of the efficiency of different reducers (hydrocarbons and oxygenated compounds), Appl. Catal. B 64, 103. [68] Joubert, E., Courtois, X., Marécot, P. et al. (2006) The chemistry of DeNOx reactions over Pt/Al2 O3 : The oxime route to N2 or N2 O, J. Catal. 243, 252. [69] Ilieva, L., Pantalev, G., Ivanov, I. et al. (2006) Gold catalysts supported on CeO2 and CeO2 -Al2 O3 for NOx reduction by CO, Appl. Catal. B 65, 101. [70] Krutzch, B., Goerigk, Ch., Kurze, S. et al. (1999) Process and apparatus for reducing nitrogen oxides in engine emissions, US Patent 5,921,076. [71] Satokawa, S. (2000) Enhancing the NO/C3 H8 /O2 Reaction by Using H2 over Ag/Al2 O3 Catalysts under Lean-Exhaust Conditions, Chem. Lett. 29, 294. [72] Eranen, K., Klingstedt, F., Arve, K. et al. (2004) On the mechanism of the selective catalytic reduction of NO with higher hydrocarbons over a silver/alumina catalyst, J. Catal. 227, 328. [73] Shibata, J., Shimidzu, K., Takada, Y. et al. (2004) Structure of active Ag clusters in Ag zeolites for SCR of NO by propane in the presence of hydrogen, J. Catal. 227, 367. [74] Burch, R., Breen, J.P. and Meunier, F.C. (2002) A review of the selective reduction of NOx with hydrocarbons under lean-burn conditions with non-zeolitic oxide and platinum group metal catalysts, Appl. Catal. B 39, 283. [75] Richter, M., Bentrup, U., Eckelt, R. et al. (2004) The effect of hydrogen on the selective catalytic reduction of NO in excess oxygen over Ag/Al2 O3 , Appl. Catal. B 51, 261. [76] Wichterlova, B., Sazama, P., Breen, J.P. et al. (2005) An in situ UV-vis and FTIR spectroscopy study of the effect of H2 and CO during the selective catalytic reduction of nitrogen oxides over a silver alumina catalyst, J. Catal. 235, 195. [77] Bion, N., Saussey, J., Haneda, M. et al. (2003) Study by in situ FTIR spectroscopy of the SCR of NOx by ethanol on Ag/Al2 O3 – Evidence of the role of isocyanate species J. Catal. 217, 47. [78] Backman, H., Arve, K., Klingstedt, F. et al. (2006) Kinetic considerations of H2 assisted hydrocarbon selective catalytic reduction of NO over Ag/Al2 O3 : Kinetic modelling, Appl. Catal. A 304, 86. [79] Eranen, K., Lindfors, L.E., Klingdted, F. et al. (2003) Continuous reduction of NO with octane over a silver/alumina catalyst in oxygen-rich exhaust gases: combined heterogeneous and surface-mediated homogeneous reactions, J. Catal. 219, 25. [80] Shimidzu, K., Higashimata, T., Tsuzuki, M. et al. (2006) Effect of hydrogen addition on SO2 tolerance of silver-alumina for SCR of NO with propane, J. Catal. 239, 117. [81] Houel, V., James, D., Millington, P. et al. (2005) A comparison of the activity and deactivation of Ag/Al2 O3 and Cu/ZSM-5 for HC-SCR under simulated diesel exhaust emission conditions, J. Catal. 230, 150. [82] Angelidis, T.N., Christoforou, S., Bongiovanni, A. et al. (2002) On the promotion by SO2 of the SCR process over Ag/Al2 O3 : influence of SO2 concentration with C3 H6 versus C3 H8 as reductant, Appl. Catal. B 39, 197.

Formation of N2 O During NOx Conversion

323

[83] Haneda, M., Pusparatu, Kintaichi, Y. et al. (2005) Promotional effect of SO2 on the activity of Ir/SiO2 for NO reduction with CO under oxygen-rich conditions, J. Catal. 229, 197. [84] Burch, R. and Watling, T.C. (1998) The effect of sulphur on the reduction of NO by C3 H6 and C3 H8 over Pt/Al2 O3 under lean-burn conditions, Appl. Catal. B 17, 131. [85] Bosch, H. and Janssen, F. (1988) Formation and control of nitrogen oxides, Catal. Today 2, 369. [86] Seker, E., Yasyerli, N., Gulari, E. et al. (2002) NOx reduction by urea under lean conditions over single step sol-gel Pt/Al2 O3 catalyst, Appl. Catal. B 37, 27. [87] Suarez, S., Martin, J.A., Yates, M. et al. (2005) N2 O formation in the selective catalytic reduction of NOx with NH3 at low temperature on CuO-supported monolithic catalysts, J. Catal. 229, 227. [88] Blanco, J., Avila, P., Suarez, S. et al., (2004) CuO/NiO monolithic catalysts for NOx removal from nitric acid plant flue gas, Chem. Eng. J. 97, 1. [89] Wolberg, A. and Roth, J.F. (1969) Copper oxide supported on alumina: III. X-ray K-absorption edge studies of the Cu2+ species, J. Catal. 15, 250. [90] Mizumoto, M., Yamazoe, M., Seiyama, T. et al. (1978) Catalytic reduction of NO with ammonia over Cu(II) NaY, J. Catal. 55, 119. [91] Yamagushi, M., Matsushite, K. and Takami, K. (1976) Remove nitrogen oxides (NOx  from nitric acid tail gas, Hyd. Proc. 55(8), 101. [92] Sadakane, M., Asanuma, T., Kubo, J. et al. (2005) Facile procedure to prepare threedimensionally ordered macroporous (3DOM) perovskite-type mixed metal oxides by colloidal crystal templating method, Chem. Mater. 17, 3546. [93] Kaliaguine, S., Van Neste, A., Szabo, V. et al. (2001) Perovskite-type oxides synthesized by reactive grinding: Part I. Preparation and characterization, Appl. Catal. A 209, 345. [94] Civera, A., Pavese, M., Saracco, G. et al. (2003) Combustion synthesis of perovskite-type catalysts for natural gas combustion, Catal. Today, 83, 199. [95] Uenishi, M., Tanigushi, M., Tanaka, H. et al. (2005) Redox behavior of palladium at start-up in the perovskite-type LaFePdOx automotive catalysts showing a self-regenerative function, Appl. Catal. B 57, 267. [96] Tan, I., Tanaka, H., Uenishi, M. et al. (2003) Research on the Co-free intelligent catalyst, SAE Paper 01-0812. [97] Peucheret, S., Wyszynsky, M.L., Lehrle, R.S. et al. (2005) Use of catalytic reforming to aid natural gas HCCI combustion in engines: experimental and modelling results of open-loop fuel reforming, Int. J. Hydrogen, Energy 30, 1583. [98] Peucheret, S., Feaviour, M. and Golunski, S. (2006) Exhaust-gas reforming using precious metal catalysts, Appl. Catal. B 65, 201. [99] Takeuchi, M. and Matsumoto, S. (2004) NOx storage-reduction catalysts for gasoline engines, Top. Catal. 28, 151. [100] Takahashi, N., Suda, K.A., Hashisuka, I. et al. (2007) Sulfur durability of NOx storage and reduction catalyst with supports of TiO2 , ZrO2 and ZrO2 -TiO2 mixed oxides, Appl. Catal. B 72, 187. [101] Nova, I., Lietti, L., Castoldi, L. et al. (2006) New insights in the NOx reduction mechanism with H2 over Pt-Ba/-Al2 O3 lean NOx trap catalysts under near-isothermal conditions, J. Catal. 239, 244. [102] Martín, J.A., Yates, M., Ávila, P. et al. (2007) Nitrous oxide formation in low temperature selective catalytic reduction of nitrogen oxides with V2 O5 /TiO2 catalysts, Appl. Catal. B 70, 330. [103] Burch, R. and Coleman, M.D. (2002) An investigation of promoter effects in the reduction of NO by H2 under lean-burn conditions, J. Catal. 208, 435. [104] Ueda, A., Nakano, T., Azuma, M. et al. (1998) Two conversion maxima at 373 and 573K in the reduction of nitrogen monoxide with hydrogen over Pd/TiO2 catalyst, Catal. Today 45, 135.

324

Past and Present in DeNOx Catalysis

[105] Costa, C.N., Stathopoulos, V.N., Belessi, V.C. et al. (2001) An investigation of the NO/H2 /O2 (lean-deNOx  reaction on a highly active and selective Pt/La07 Sr02 Ce01 MnO3 catalyst at low temperatures, J. Catal. 197, 350. [106] Costa, C.N., Savva, P.G., Andronikou, C. et al. (2002) An investigation of the NO/H2 /O2 (lean-deNOx  reaction on a highly active and selective Pt/La05 Ce05 MnO3 catalyst, J. Catal. 209, 456. [107] Engelmann-Pirez, M., Granger, P. and Leclercq, G. (2005) Investigation of the catalytic performances of supported noble metal based catalysts in the NO + H2 reaction under lean conditions, Catal. Today, 107, 315. [108] Tanoka, T., Yokata, K., Doi, H. et al. (1997) Selective catalytic reduction of NO over PtMo catalysts with alkaline or alkaline earth metal under lean static conditions, Chem. Lett. 409. [109] Machida, M. and Watanabe, T. (2004) Effect of Na-addition on catalytic activity of PtZSM-5 for low-temperature NO-H2 -O2 reactions, Appl. Catal. B 52, 281. [110] Gonçalves, F. and Figueiredo, J.L. (2006) Synergistic effect between Pt and K in the catalytic reduction of NO and N2 O, Appl. Catal. B 62, 181. [111] Pirez, M. (2004) Réduction catalytique selective des oxides d’azote (NOx  provenant d’effluents gazeux industriels par l’hydrogène ou le methane, Ph’D Thesis, University of Lille, France. [112] Qi, G., Yang, R.T. and Thompson, L.T. (2004) Catalytic reduction of nitric with hydrogen and carbon monoxide in the presence of excess oxygen by Pd supported on pillared clays, Appl. Catal. A 259, 261. [113] Summers, J.C. (1979) Reaction of sulfur oxides with alumina and platinum/alumina, Env. Sci. Technol. 13, 321. [114] Burch, R. and Urbano, F.J. (1995) Methane combustion over palladium catalysts: The effect of carbon dioxide and water on the activity, Appl. Catal. A 123, 173. [115] Mowery, D.L., Graboski, M.S., Ohno, T.R. et al. (1999) Deactivation of PdO-Al2 O3 oxidation catalyst in lean-burn natural gas engine exhaust: aged catalyst characterization and studies of poisoning by H2 O and SO2 , Appl. Catal. B 21, 157. [116] Peˇna, M.A. and Fierro, J.L. (2001) Chemical structures and performance of perovskite oxides, Chem. Rev. 101, 1981. [117] Twagisrashema, I., Engelmann-Pirez, M., Frère, M. et al. (2007) An in situ study of the NO + H2 +O2 reaction on Pd/LaCoO3 based catalysts, Catal. Today 119, 100. [118] Nishibata, Y., Mizuki, J., Akao, T. et al. (2002) Self-regeneration of a Pd-perovskite catalyst for automotive emissions control, Nature 418, 164.

Chapter 11

DESIGN OF EXPERIMENTS COMBINED WITH HIGH-THROUGHPUT EXPERIMENTATION FOR THE OPTIMIZATION OF DeNOx CATALYSTS R. Vijay and J. Lauterbach∗ Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, DE 19716, USA ∗

Corresponding author: Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, DE 19716, USA. E-mail: [email protected]

Abstract Design of experiments in combination with high-throughput experimentation (HTE) is a powerful toolbox for the systematic study of vast parameter spaces encountered in the design and optimization of heterogeneous catalysts. We will present the general approach as applied to NOx storage and reduction (NSR) catalysts using response surface analysis. Empirical models were developed to predict the catalyst performance as a function of cycle time, lean fraction of cycle time, and catalyst composition. These models provide useful insight about the factors controlling the NOx storage and NOx conversion of NSR catalysts. Using these empirical models, new catalyst formulations that maximize NOx conversion and selectivity to N2 were found. In addition, high-throughput experimentation allows for simultaneous synthesis and screening of large arrays of different materials which further accelerates the discovery and optimization process. We have also tested a variety of new materials for NSR applications, composed of different transition metals added to standard Pt/Ba-based NSR catalysts, and have discovered that a noble metal free 5Co/15Ba catalyst stores NOx as efficiently as a standard 1Pt/15Ba NSR catalyst. Using the response surface strategy we have verified that the addition of Co to NSR catalsyts improves the performance at higher lean fractions, allowing a substantial improvement in the fuel efficiency. These studies clearly establish the utility of HTE when combined with design of experiments for the efficient analysis of such vast multidimensional systems and for the discovery of new materials, which is the guiding factor for any major technological advances.

1. INTRODUCTION High-throughput experimentation (HTE) has attracted much (and not always positive    ) attention in the scientific and industrial communities in the past few years [1–3]. Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Published by Elsevier B.V.

326

Past and Present in DeNOx Catalysis

The drive for HTE in heterogeneous catalysis is motivated by the large parameter space that can potentially affect the performance of a catalyst. When one, in addition to the compositional space, also considers the variety of preparation methods and reaction conditions that a given catalyst could be subjected to, the total number of possible experimental combinations grows even larger and makes HTE very attractive. To increase the throughput at which catalytic systems can be studied, HTE approaches are increasingly used for materials development [1,4]. High-throughput catalytic techniques have been used to study selective catalytic reduction of nitric oxide (NO) using hydrocarbons [5–8] and one study has also looked at seven different catalysts for NOx storage activity under a single reaction condition [8]. The genesis for combinatorial experimentation in materials science has been reviewed recently by Schubert et al. [9], in which they discuss ‘Combinatorial’ experiments that have been performed as early as the late 1800s. In 1970, Hanak outlined the basic methodology still in use today in HTE [10] and Figure 11.1 summarizes this idea. The first reference in the open scientific literature in heterogeneous catalysis that refers to the use of parallel reactors to study heterogeneous catalysts appeared in 1980 [11]. In 1986, Seddon et al. published a detailed report about the implementation and verification of six parallel reactors for testing heterogeneous catalysts [12]. A first review of the field of high-throughput heterogeneous catalysis was published by Senkan in 2001 [1] and a recent one was published by our group [13]. Likewise, the patent literature was recently been reviewed by Dar [14]. Therefore, the majority of the work cited in these reviews will not be repeated here. We will be mainly focusing on the work done on HTE in our group over past several years and the way we have applied these tools to the study of the catalytic systems for the abatement of nitrogen oxides (DeNOx  for automotive exhaust applications using nitrogen storage and reduction (NSR) catalysts. Even with HTE, it is experimentally expensive to study the entire range of combinations and concentrations affecting NSR catalysts in a reasonable amount of time. Therefore, we are employing a combination of both statistical design of experiments (DOE) [15] and HTE to study NSR catalysts. Statistical experimental design methods have been applied to zeolite synthesis [16], catalyst optimization and discovery for a single reaction condition using HTE [17,18], reaction condition optimization for single catalysts discovered with HTE [19], and recently discussed in the literature, without the

Synthesis of large libraries of potentially useful catalytic formulations for the specific system

Catalyst characterization using physical and chemical analytical techniques

Predicting trends and designing new catalyst formulations based on the knowledge extracted

Figure 11.1. Experimental strategy for HTE techniques.

Testing the performance (or properties) of catalyst using HTE rapid screening techniques

Data acquisition and processing. Interpretation using statistical analysis tools and visual aids like response surface plots etc.

Design of Experiments for Optimization of DeNOx Catalysts

327

use of experimental data, as a tool in HTE [2]. A recent paper presents the application of DOE to HTE in studying both the catalyst composition and the reaction conditions [3]. This paper, however, does not provide enough details about the procedure used to generate the design and it is uncertain to what extent interactions were studied and how the optimization was conducted. Here, we will provide full details of the experimental procedure and demonstrate the application of HTE to the study of NSR catalysts in combination with experimental design. For simplicity, we have divided the chapter into different sections each concentrating on different aspect.

2. HIGH-THROUGHPUT EXPERIMENTAL SETUP AND DATA ANALYSIS Extensive work has been performed in the past decade toward the development of novel analytical tools for high-throughput studies of heterogeneous catalysts [1–3]. Many initial analytical tools for HTE were not able to follow the selectivity of catalytic reactions. The majority of high-throughput analytical techniques consisted of the adaptation of traditional analytical tools to HTE through automation or parallelization. The mass spectrometer has proven to be one of the most popular analytical tools due to its analysis speed, established technology, and versatility. Various experimental setups have been developed to transport the reaction products corresponding to different catalysts, to the mass spectrometer. These techniques ranged from the straight forward use of switching valve [20–23], to the more complex use of x, y, and possibly z motion control devices [24–26]. Other analytical tools followed similar methods of adaptation of existing technologies to the analysis of multiple catalysts. These techniques included the use of gas chromatography [27–36], sensors [37–40], and other analytical techniques [8,23,41–45]. To progress toward high-throughput catalytic science, chemically sensitive and quantitative high-throughput analytical tools were developed to enable researchers to study and compare the selectivity as well as the activity of multiple catalysts. Of these techniques, those based on optical methods have the advantage of being non-destructive, amenable to in situ studies over a wide range of temperature and pressure, and adaptable to parallel analysis [41,46–48]. One such novel technique implemented by our research group uses a focal plane array (FPA) detector in place of the standard single element IR detector in traditional IR spectroscopy. The replacement of the single element detector with the FPA allowed the collection of spatially resolved IR spectra [49]. This instrument was the first chemically sensitive, quantitative, and parallel analytical technique applied to HTE [20,23,50,51]. The optical setup consists of a Bruker Equinox 55 spectrometer, refractive optical elements, a gas phase array (GPA) sampling accessory [52] and a 64 × 64 pixel mercury cadmium telluride (MCT) focal plane array detector arranged as indicated in Figure 11.2. The IR rays pass straight through the GPA and falls on the FPA at the other end, the gases flow through the GPA with the help of 16 1/8 stainless steel tubing inlets and outlets connected at each end of GPA, and on the left is the image of the GPA as seen through the FPA. The optical elements consist of two plano-convex barium fluoride lenses, two zinc selenide meniscus lenses, a front surface mirror and a 3750 nm longwave-pass filter with a long wavelength cutoff at ∼8230 nm. The FPA detector is sensitive between 4000 and 900 cm−1 and is operated at a frame rate of 315 Hz. The spectral images presented here were collected between 2686 and 1342 cm−1

328

Past and Present in DeNOx Catalysis

FPA

Figure 11.2. Arrangement of GPA and FPA for the FTIR imaging based HTE system.

at 8 cm−1 resolution. With these parameters, an imaging data set consisting of 4096 infrared spectra can be collected in less than 2 s. The GPA consists of an array of 16 stainless steel tubes, capped by barium fluoride windows on each end. Each tube has a separate inlet and outlet to maintain separation of the products being analyzed. The GPA allows the simultaneous spectral analysis of up to 16 different gas-phase samples and is connected to 16 mg-scale reactors for the parallel study of heterogeneous catalysts, as described in [53]. On the reactor side, the gas mixture enters from the top and is separated into 16 individual streams, going to each one of the reactors. Figure 11.3 shows the side view of the high-throughput reactor setup. The reactors are vertical flow-through reactors with a

Figure 11.3. Side view of the top of the 16-way Reactor setup.

Design of Experiments for Optimization of DeNOx Catalysts

329

catalyst powder supported on a stainless steel frit. Each reactor contains a thermocouple in the catalyst bed, and the temperatures of all 16 reactors are continuously displayed and recorded by software written in-house using LabView. Multiple mass flow controllers permit a wide range of feed gas concentrations, compositions, and flowrates to be explored. The reactor operates at ambient pressure with a temperature range between room temperature and ∼1073 K and space velocities between 3500 ml/(h gcat) and 200 000 ml/(h gcat). In addition, a capillary and a needle valve are installed before each reactor to minimize any flowrate differences among the different reactors. Further details regarding the setup can be found in [53–56]. The product streams are then connected to the inlet side of GPA tubes and analyzed using the above mentioned FTIR setup. After collection, the data are processed and analyzed using software written in-house [57]. More recently, this analytical setup has been applied to the analysis of multiple transient reactions in parallel [13,58]. One of the important uses of this rapid, chemically sensitive parallel technique is the possibility to observe transient reaction products, as demonstrated in Figure 11.4. Shown in Figure 11.4 are the IR spectra of the effluent from a single catalyst, a NOx storage and reduction (NSR) catalyst [59], as a function of time. The absorbance intensity of the anti-symmetric stretching band of nitrous oxide (N2 O) at 2227 cm−1 varies considerably during a change in the reaction conditions from fuel rich to fuel lean. The intensity of N2 O band increases initially and then decays to a small steady-state value. Without the temporal analysis, the identification of N2 O would not have been possible and a key reaction intermediate would have gone undetected. The IR spectra are finally analyzed to determine the effluent concentration from each reactor channel. The quantification of species concentration is performed using either univariate or multivariate calibration methods. For non-overlapping peaks, like CO, CO2 , and N2 O, we can use univariate calibration. This is simply performed by baseline correction, the peak areas and/or peak heights and then converting these values

0.6

N2 O 0.4

↑ A

NO2 0.2

CO

NO

0.0 75

←

2200 50

t/s

2000 25

1800 0

1600

~ ν/cm–1 →

Figure 11.4. Transient chemically sensitive analysis of the effluent from a single catalyst [58].

330

Past and Present in DeNOx Catalysis

6000

Absorbance (a.u.)

1.2

4000

1.0 0.8

0.8 0.6 0.4 0.2 0.0 2600

1.2 0.4 0.2

3000

1.0

0.6

0.0 2600

2400

2200

2000

1800

1600

1400

–1

Wavelength (cm )

2000

Absorbance (a.u.)

Absorbance (a.u.)

5000

Concentration (ppm)

1.2

CO

2400 2200

2000 1800

1600

1400

–1

1.0

Wavelength (cm )

0.8 0.6 0.4 0.2 0.0 2600

1000

2400

2200

2000

1800

1600

1400

Wavelength (cm–1)

NO2

0 0

40

80

120

160

200

240

280

Time (sec)

Figure 11.5. Concentration profile for NO2 and CO during a switch from fuel-lean to fuel-rich conditions for a typical NSR catalyst. NO2 decreases and CO increases with time [61].

to concentrations based on previously performed calibrations. The concentrations calculated from different pixels corresponding to a given reactor are then averaged in order to increase the signal to noise ratio and improve the detection limit. To analyze convoluted IR spectra, the PLSplus/IQ add-on (i.e. multivariate calibration) to the Thermo Electron Corporation (Waltham, MA, USA) software GRAMS/AI version 7.0 was used. With the GRAMS software, the multivariate calibration were produced and implemented. Details on both these calibration techniques can be found in [54]. Figure 11.5 shows the concentration profiles for NO2 and CO in the effluent of one reactor when switching from simulated fuel-lean to fuel-rich exhaust conditions, as measured by FTIR imaging [53,54,59]. Concentrations are calculated directly from the IR spectra using chemometric methods. The NO2 band (around 1600 cm−1  decreases with time, while the CO band (around 2143 cm−1  becomes much stronger. The effluent composition can also be directly correlated to the intensity of those specific IR bands in the IR images [59]. To summarize, FTIR imaging is a versatile technique capable of performing steady-state and transient measurements even for complex IR spectra. We have written a Labview interface for monitoring, recording, and controling the actual flow rates for each mass flow controller and the temperature of each one of the catalyst beds. A representative screenshot of this program is shown in Figure 11.6. Each of the 16 thermocouple readouts is shown in the table at the top of the screen, as well as plotted in real-time. The mass flow controller setpoints and real-time gas flow readings are shown at the bottom of the screen, with each fill bar corresponding to a different controller. This program is also interfaced with the mass flow controllers, the temperature controller, the bed thermocouples, and the spectral imaging collection software, allowing for completely unattended collection of HTE data.

Design of Experiments for Optimization of DeNOx Catalysts

331

Figure 11.6. LabVIEW interface to monitor and control the HT reactor system.

3. STATISTICAL DESIGN OF EXPERIMENTS The basic premise of all experimental research is that experimental data contain information from which knowledge can be extracted. However, if the data are not acquired carefully, one may end up with a vast amount of data that do not contain the desired information. Additionally, any information not contained in data cannot be extracted even by the most sophisticated analysis methods. The traditional approach to investigate the effect of different variables in a multidimensional system has been to change a single experimental variable at a time while holding all other experimental variables constant. The most attractive advantage to this methodology, termed ‘one at a time’, is the ability to perform a simple analysis of individual variables of interest. The obvious disadvantages consist of the inherent experimental inefficiencies in testing for the effect of one variable at a time, and the inability to experimentally observe interactions between variables of interest. Ensuring that one is able to acquire informative data in the face of unavoidable random variability and extract the contained information efficiently is the role of statistical DOEs. For high-throughput studies, where one is able to acquire large quantities of data in fairly short periods of time, the judicious application of experimental design is especially important. Statistical experimental design procedures have been used successfully in many science and engineering applications [60,61]. The fundamental goal of these procedures is adequacy and efficiency. Adequacy is ensuring that the acquired data contain the information required to achieve the objectives of the experimental study, and

332

Past and Present in DeNOx Catalysis

efficiency is obtaining the desired information with the minimum number of experiments. Statistical DOE is an adaptable approach based on the current understanding of the system under study and the final desired level of understanding. Given the possibility of different objectives at different phases of experimental studies, there are various experimental designs for each phase. Many times, the experimental process begins with exploratory research referred to as screening designs and then evolves to more refined experimental designs referred to as response surface designs or D-optimal designs. These designs will be explained in the following sections.

3.1. Screening designs (or factorial designs) Factorial designs are used in the intial phases of experimentation to determine the statistically significant factors and interactions, and to quantify the effect of the factors on the response of interest. In screening design, the experimental parameters are normally varied at two different levels. In addition, the experimental parameters are normalized to the same scale with +1 for the high and −1 for the low value of the factor. The model coefficients and terms estimated based on these normalized variables then gives a better indication of the relative importance of different factors and interaction between them. The normalization is done using Eqn (1) where N(X) is the normalized variable and XH and XL are the respective low and high values of X. NX =

2 ∗ X − XH − X L XH − X L

(1)

A linear model is used to relate a performance criterion (Y) to the parameters of interest (X1 , X2 , X3 ,    ) and takes the form of Eqn (2), where C is a constant and i are the coefficients fitted to the experimental data. Y = C + 1 X1 + 2 X2 + 3 X3 + · · · + 12 X1 X2 + 13 X1 X3 + · · · + 123 X1 X2 X3 + · · · (2) Here X1 , X2 , X3     are the normalized variables for the parameters X1 , X2 , X3     [same as NX is for X]. The factorial designs can be classified into two different varieties, namely full factorial designs and fractional factorial designs (FFD). The full factorial design enables the determination of all important parameters and all the interactions between them. Once the range of the experimental conditions have been decided upon, the full factorial experimental design is developed by considering all possible combinations of the two levels of different factors. This is shown graphically for a three variable system in Figure 11.7, where the full factorial design points are indicated by the white circles. These combinations of different factorial levels are then tested to check for their effect on the response (like Y above). The collection of these experimental points assures that the interactions between the experimental variables are considered, in addition to the main effects, since all parameters are changed simultaneously. A significant disadvantage to this type of experimentation is the number of experiments that are required to estimate all model coefficients when the number of experimental parameters (k) is high. The minimum total number of experiments (N) necessary, to estimate all model coefficients, scales according to N = 2k

Design of Experiments for Optimization of DeNOx Catalysts

333

+1

x3

–1

+1 x2 –1

x1

+1

–1

Figure 11.7. Experimental combinations in statistical experimental design for three variables.

Table 11.1. Linear model parameters for seven experimental factors Model coefficient name

Interaction order

Constant Main effect 2-way interaction 3-way interaction 4-way interaction 5-way interaction 6-way interaction 7-way interaction

0 1 2 3 4 5 6 7

Total model parameters

Number of model coefficients 1 7 21 35 35 21 7 1 128

As an example, an experimental program interested in investigating the effect of three parameters would only need to complete eight experiments to estimate the eight coefficients in Eqn (2). However, to complete the same analysis for seven factors, 128 experiments are necessary. The number of model parameters for seven variables can be broken down according to the different variable interactions, as done in Table 11.1. The number of interaction parameters (P) can be calculated with the total number of factors (N) and the interaction level (I) using the correlation: P=

N! N − I!∗ I!

Of these 128 model parameters, the majority of the higher order interactions would most likely be statistically insignificant. To decrease the number of experiments required, FFDs can be used. In these designs, the ability to predict the highest order interactions is sacrificed in order to reduce the number of experiments. This has the effect of reducing

334

Past and Present in DeNOx Catalysis

the required experiments while at the same time confounding the estimated coefficients in the linear model of Eqn (2). The extent of the confounding of the parameters is referred to as the resolution of an experimental design. For example, an experimental design with a resolution of V indicates that the estimates of the main effects (i ) are combined with the 4-way interactions (ijkl ) and are estimated as a single value. Similarly, the estimates of 2-way interactions (ij ) are combined with the 3-way interactions (ijk ). In general, experimental designs, where the resolution is V or higher, are normally considered excellent FFDs because there is minimal confounding between low order interactions and the majority of the higher order interactions would most likely be statistically insignificant. FFDs are run in half (1/2), quarter (1/4), eighth (1/8),    fractions, where the fraction refers to the experimental fraction of the complete full factorial design experiments that are performed. The nomenclature for these designs can be represented as N = 2Rk−p where N is the total number of experiments, k is the number of experimental factors, R is the resolution and p is related to FFD by FFD =

1 2p

As an example, a seven factor quarter fraction factorial design would be written as seen in equation below 7−2 2IV

A summary of the effect of decreasing the total number of experimental runs on the resolution of the design is seen in Table 11.2 below. It can be seen that for more than five factors, running half fraction factorial designs reduce the minimum resolution to V. This implies that a significant savings in experimental efforts can be achieved, with a minimal loss in the information obtained, by reducing the experiments in half using a half fraction factorial design when a large number of variables are investigated.

Table 11.2. Resolution level obtained for different number of factors and different numbers of experimental runs Experimental runs 4 8 16 32 64 128

Factors 2

3

Full

III Full

4

5

6

7

IV Full

III V Full

III IV VI Full

III IV IV VII Full

Design of Experiments for Optimization of DeNOx Catalysts

335

3.2. Response surface design Screening designs are mainly used in the intial exploratory phase to identify the most important variables governing the system performance. Once all the important parameters have been identified and it is anticipated that the linear model in Eqn (2) is inadequate to model the experimental data, then second-order polynomials are commonly used to extend the linear model. These models take the form of Eqn (3), where i are the coefficients for the squared terms in the model and 3-way and higher-order interactions are excluded. Y = C + 1 X1 + 2 X2 + · · · + 12 X1 X2 + · · · + 1 X12 + 2 X22 + · · ·

(3)

This model is capable of estimating both linear and non-linear effects observed experimentally. Hence, it can also be used for optimization of the desired response with respect to the variables of the system. Two popular response surface designs are central composite designs and Box–Behnken designs. Box–Behnken designs were not employed in the experimental research described here and will therefore not be discussed further, but more information on Box–Behnken designs can be obtained from reference [15]. Central composite designs typically combine a full factorial design with additional points, as seen in Figure 11.7. Here the full factorial design points are designated as white circles, while the additional experimental points are designated by ‘star’ points, seen as cross-filled circles, and center points, seen as solid-shaded circles. The distance from the center point to the star points, commonly referred to as , is normally chosen to be the same as the distance from the center point to the full factorial points to ensure that the experimental points are symmetric about the center point. This design is known as full central composite design. However, in the given figure the star points are on the centre of each one of the six faces. This design is known as face-centered central composite design. In this study, we will be using both kinds of designs, either to decide on catalyst composition or to decide on reaction conditions to be tested. In order to estimate the reproducibility of the system, the center point experimental condition is commonly repeated multiple times. When the number of experimental factors is high, it is possible to reduce the total number of experiments in the response surface design similar to the FFD. This is done by combining a FFD, in place of a full factorial design, while maintaining the star points and the center point of the response surface design. Finally, the application of statistical DOEs can also be adapted to the needs of HTE. As explained above, the reactor used for the experiments performed consisted of 16 reactor wells. This enabled the simultaneous testing of up to 16 catalysts under similar conditions. This feature of the high-throughput setup can be adapted into the experimental design by conducting one experimental design, for the catalyst compositions across the different reactors, nested inside another design that controlled the operating conditions for all catalysts. Figure 11.8 graphically represents this idea. In this chapter, we will discuss the application of both the screening as well as the nested response surface designs discussed above, using HTE, for the study of NSR catalysts with respect to catalyst compostition and reaction condition. In practice, the reactors are loaded with different catalysts corresponding to a design over the catalyst composition and

336

Past and Present in DeNOx Catalysis +1 +1 C3

C3

–1

–1

–1 C1 +1 +1

–1

C2

+1

–1 +1

+1 –1

C2

–1

C2

+1

–1

–1

C1

+1

C3

C3

–1

C

+1

+1

–1

C1

+1 –1

C2

+1

OC3

+1 +1 C3

C3

–1

–1 +1 –1C C1 3

2 1C

+1

C3 –1

–1 –

–1

C1

C1

+1 –1

+1

–1

C1

+1 –1

C2

+1

C2

+1

+1

–1

–1

C2 +1 –1

–1

+1

OC

2

–1

–1

OC1

+1

Figure 11.8. Nested experimental design.

then the experimental operating conditions to test the catalysts are designed using a separate experimental design. The combination of these two designs allows a comprehensive search of the parameter space affecting a catalyst performance, i.e. the catalyst composition and the operating conditions or pre-treatment options. Complete details of the statistical analysis which is done, for both the factorial and response surface designs, to identify the statistical significance of different factors and interactions can be found [15]. In summary, the main goal of DOEs is to strategically collect the required information by running minimum number of experiments. A major disadvantage of this kind of approach is that it lacks a fundamental perspective and would not be able to provide a comprehensive understanding of the system by itself. In addition, it is impossible to make any definitive conclusions about the response outside the range of experimental conditions studied. However, using this approach, one can very quickly refine a complex multidimensional system to a small number of potentially interesting areas. The empirical models developed, and the trends identified provide a strong foundation for obtaining the detailed mechanistic insight of the system and developing comprehensive mechanistic models. In addition, a particularly unique characteristic of this approach is the ability to identify the interactions between different variables and their relative importance. This narrows down the parameter space associated with many multivariable systems and further provides physical insights into the variables governing the response.

Design of Experiments for Optimization of DeNOx Catalysts

337

4. CATALYST SYNTHESIS AND CHARACTERIZATION All catalysts were synthesized by the incipient wetness technique. The results shown here will mainly focus on Pt, Ba, Co, Fe, and Mn containing catalysts supported on

-Al2 O3 (Catalox1 Sba-200, 200 m2 /g). Pt and Ba were chosen as typical NSR components, Fe was added because of reports that it improved the resistance of Pt/Ba NSR catalysts to sulfur [62] and Co and Mn were chosen to check for any similar promotional effect. Chloroplatinic acid hexahydrate, barium nitrate, iron (III) nitrate nonahydrate, cobalt nitrate, and magnesium nitrate (Strem Chemicals) were used as metal precursors. Because of the low solubility of Ba(NO3 2 in water, it was necessary to utilize multiple impregnation steps to achieve the desired weight loadings. Finally, the necessary amount of all precursors was dissolved in distilled water to obtain the weight loading for a given catalyst. This solution was then added to the dried support until incipient wetness was obtained. The impregnated supports were dried overnight in a vacuum oven at a temperature of 393 K and then crushed before the next impregnation step. This process was repeated until the entire precursor solution had been added to the support. After completion of the final impregnation step, the powders were crushed and calcined in a tube furnace. The calcination procedure consisted of heating to 473 K over 2 h, holding the temperature at 473 K for 1 h, further heating to 823 K over 3 h, holding at 823 K for 2 h, and then cooling to 298 K over 4 h. Before the first run, the catalysts were reduced in the high-throughput reactor for 1 h in 10% v/v H2 at 773 K. On subsequent days, the catalysts were reduced for 30 min under identical conditions. The naming convention for each catalyst is based on the nominal weight loading. Thus, a catalyst with a nominal weight loading of 1% w/w Pt and 15% w/w Ba is referred to as 1Pt/15Ba. For all results reported, 0.15 g of catalyst was loaded into each reactor. Atomic absorption spectroscopy (AAS) was used to verify the weight loadings of a selected number of catalysts. The details explaining AAS can also be found in textbooks [63] and will not be repeated here. The actual weight loadings of those selected catalysts will be reported wherever necessary. The gases used in the experiments were obtained from BOC Gases and Keen Compressed Gas. The NO was obtained as premixed 1% NO in He. Ethylene, carbon monoxide, oxygen, hydrogen, and helium, were obtained at purity levels ≥9999%.

5. INITIAL SCREENING STUDY Screening designs, as described above, were used to decide on the relative importance of different experimental variables/factors to be tested in the initial exploratory phases. The system we chose to study using these designs was that of Pt-, Ba-, and Fe-based NSR catalysts. For the screening design, we studied the significance of the catalyst temperature, NOx concentration, O2 concentration, reductant concentration, reductant type, and space velocity. Although the statistical analysis, using full factorial design, for catalyst composition was also performed, but we will not discuss it here as it is already well known from previous studies that Pt, Ba, and Fe are all important components of the NSR catalyst [62,64]. Here, we will only report the results from a single catalyst 1Pt/5Fe/15Ba and details about the effect of catalyst composition will be discussed under response surface design. A quarter fraction factorial design was used for studying the

338

Past and Present in DeNOx Catalysis Table 11.3. Reaction investigated

condition

Setting NO conc. (ppm) O2 conc. (%) C2 H4 conc. (%) CO conc. (%) Reductant type Space velocity (mL/h/gcat) Temperature (K)

screening

parameters

Low

High

3500 4 0.58 3.5 C2 H4 30 000 548

8650 8 0.92 5.5 CO 42 500 648

Table 11.4. Actual reaction conditions tested on all screening catalysts Reaction condition ID

NO conc. ppm

O2 conc. %

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

3500 8650 3500 3500 8650 3500 8650 8650 8650 8650 3500 3500 8650 3500 3500 8650

8 4 4 8 8 4 8 4 8 4 8 4 4 8 4 8

CO conc. %

C2 H4 conc. %

Space velocity mL/h/g∗ cat

092

30 000 30 000 42 500 42 500 42 500 30 000 30 000 42 500 30 000 42 500 42 500 30 000 30 000 30 000 42 500 42 500

55 55 35 092 058 35 058 058 35 058 35 092 55 092 55

Temperature Stoichiometric K ratio Lean Rich 548 548 548 548 548 548 548 548 648 648 648 648 648 648 648 648

297 161 152 467 307 239 482 253 482 253 467 239 161 297 152 307

006 016 006 01 016 01 025 025 025 025 01 01 016 006 006 016

effect of reaction conditions, reducing the number of experimental testing points from 26 (64) to 26−2 (16). The high and low settings for the fuel lean inlet reaction conditions used as input to the FFD are shown in Table 11.3. The actual reaction conditions tested are listed in Table 11.4. In this design, either ethylene or carbon monoxide was used as reducing agents. The concentration of the two reducing agents was chosen such that the stoichiometric ratio (SR) between the oxidizing and reducing agents remained constant. The catalysts were studied for saturation NOx storage (SNS) capacity. SNS is defined as the area between the steady-state NOx concentration in the fuel lean state and the actual effluent concentration of NOx from the time at which O2 was added to the feed gases until the time NOx concentration reaches steady state. The integrated area for SNS was converted to moles of NOx by using the flow rates collected for each one of the reactors.

Design of Experiments for Optimization of DeNOx Catalysts

339

Fuel rich and fuel lean conditions are based on the molar ratio of oxidizing to reducing molecules as defined by the SR: SR =

NO + 2∗ O2 CO + 6∗ C2 H4

SR greater than 1 refers to fuel lean conditions, while SR less than 1 refers to fuel rich conditions. To convert from fuel rich to fuel lean experimentally, a portion of the helium in the reactant gases was replaced with an equal volume of O2 using a 4-way switching valve. The pressures of the switching valve outlets were balanced such that only the reactant concentration is changed while keeping the flow rates and the pressure constant. Before collecting data, at least two lean/rich cycles of 15-min lean and 5-min rich were completed for the given reaction condition. These cycle times were chosen so as the effluent from all reactors reached steady state. After the initial lean/rich cycles were completed, IR spectra were collected continuously during the switch from fuel rich to fuel lean and then back again to fuel rich. The collection time in the fuel lean and fuel rich phases was maintained at 15 and 5 min, respectively. The catalyst was tested for SNS at all the different reaction conditions and the qualitative discussion of the results can be found in [75]. Quantitative analysis of the data required the application of statistical methods to separate the effects of the six factors and their interactions from the inherent noise in the data. Table 11.5 presents the coefficient for all the normalized parameters which were statistically significant. It includes the estimated coefficients for the linear model, similar to Eqn (2), of how SNS is affected by the reaction conditions. Table 11.5. Statistical analysis of the Group 2 data for the 1Pt/15Ba/5Fe catalyst in the screening study Normalized variables

SNS (∗ 10−6 mol NOx ) Coefficient

Constant

352

Parameters NO O2 Reductant concentration Reductant type (from C2 H4 to CO) Space velocity Temperature

59 – −59 122 – 163

Parameter interactions NO ∗ reductant concentration NO ∗ reductant type Reductant type ∗ Temp. NO ∗ Temp. Reductant type ∗ Reductant conc. NO ∗ Reductant conc. ∗ Temp. NO ∗ Reductant type ∗ Red conc.

– 47 −31 36 −44 – 35

340

Past and Present in DeNOx Catalysis

It was found that increasing the inlet NO concentration from 3500 to 8650 ppm, while holding everything else constant, increases the SNS by 118 ± 28 mmol NOx , on average. The same trend holds for the remaining parameters. The normalization was accomplished by assigning the low value listed in Table 11.3 as −1 and the high value as +1, as discussed earlier. Only parameters significant to 95% confidence are shown. It should be noted that the reductant concentration in the table refers to either the C2 H4 or the CO concentration. The positive correlation between SNS and inlet NO concentration was caused by the accompanying increase in the NOx gas and/or catalyst surface concentration near the Ba storage sites. This increase in NOx concentration will shift the Ba equilibrium phase to a higher ratio of Ba(NO3 2 to BaCO3 [65,66]. The results in Table 11.5 also indicate that the O2 concentration is not significant between 4% v/v and 8% v/v, confirming previous studies that show the O2 concentration only affects the SNS below 3% [67] or 4% [68]. In contradiction to the conclusions drawn by Fridell et al. [68] (who did not measure NOx storage until steady state), our results indicate that the reductant concentration and the reductant type have an influence on the SNS. The data in Fridell et al. showed a minimal decrease in NOx storage with increasing propylene concentration. However, since only four data points and no statistical analysis or confidence intervals were reported, it is not possible to state with certainty whether the propylene concentration affected the NOx storage. Our results demonstrate that increasing the reducing agent concentration does decrease the SNS. Our results also indicate that SNS changes with the reducing agent, confirming results reported in reference [69], which show a difference between H2 and CO, but also contradicting results in reference [68], which reported only minimal changes in NOx storage for different reducing agents (i.e. C3 H6 , C3 H8 , CO, and H2 . It would be expected that the reducing agent would affect the SNS, since it is known that the reducing agent strongly affects NO reduction under lean conditions over noble metal catalysts [70–72]. It has also been shown that the reducing agent affects the thermal stability of NOx species stored on Ba containing catalysts [73,74]. These effects can be explained by considering the thermodynamic driving force for the formation of Ba(NO3 2 . Increasing the reductant concentration and changing the reductant type, from CO to C2 H4 , leads to an increase in NOx reduction as seen in Table 11.5. Increased reduction lowers the NOx gas and/or catalyst surface concentration near Ba storage sites. This in turn shifts the Ba equilibrium phase to a lower ratio of Ba(NO3 2 as compared to BaCO3 and Ba(OH)2 [65]. Similar results were obtained for all catalysts studied (when we performed a full factorial design over catalyst composition, which will not be discussed here). The results discussed in the preceding paragraphs could have been obtained from traditional methods of changing one variable at a time; however, a significantly larger number of experiments would have been required. Additionally, the interaction information shown in Table 11.5 could only have been obtained by varying more than one variable at a time. The results for the 1Pt/15Ba/5Fe catalyst and the other screening catalysts indicate that interactions in reaction conditions are statistically significant. Previous reports have often neglected the possibility of such interactions. This study demonstrates the importance of a systematic analysis in the study of catalytic systems to understand not only the effects of experimental parameters in isolation, but to also understand the significance of the interactions between parameters [75]. Since a FFD was employed in this screening study, the interactions are combined with each other, and the most likely interactions are listed in Table 11.5. The results for the screening design must

Design of Experiments for Optimization of DeNOx Catalysts

341

be understood in the context of this ambiguity. Since the interactions are uncertain at this stage of testing, they will be discussed later, when we will discuss the response surface design, in more detail. However, it is clear from these screening studies that space velocity is the only reaction variable which was least important and hence it will be kept constant for the response surface studies.

6. RESPONSE SURFACE STUDY From the screening design, it was concluded that all the metal loadings, the reaction conditions and interactions between them significantly affected the catalyst response [75]. Hence, to effectively capture these effects and to develop a more robust model to predict catalyst performance as a function of both reaction conditions and catalyst composition, we performed a nested response surface design study of the above Pt-, Ba- and Fe-based NSR system. The catalyst compositions were designed using a face centered central composite response surface design. The high and low settings for the metal loadings were 0 and 1% w/w Pt, 0 and 15% w/w Ba, and 0 and 5% w/w Fe, respectively. The nominal weight loadings are indicated in Figure 11.15. The center point catalyst, indicated by shading in Figure 11.9 with a nominal weight loading of 0.5Pt/7.5Ba/2.5Fe, was synthesized four times to provide an indication of the reproducibility of the synthesis procedure. The reproducibility of the synthesis procedure was verified through AAS [76]. The four catalysts synthesized at the same nominal weight loading are referred to as 0.5Pt/7.5Ba/2.5Fe_1,2,3, or 4 to distinguish them from each other. The reaction conditions were investigated according to a half fraction central composite design. The reaction conditions and their corresponding experimental levels investigated for the response surface study are shown in Table 11.6. To ensure that changing the reductant type from CO to C2 H4 would not change the SR [75], an equivalent CO concentration was used in the experimental design. The equivalent CO concentration was defined using

5

Fe (% w/w)

0

w)

0

Pt (% w/w)

% a(

w/

15

1 0B

Figure 11.9. Nominal weight loadings of the synthesized catalysts are indicated by the circles in this diagram of a three dimensional box. The solid filled circle indicates the nominal weight loading of a catalyst synthesized four times to verify reproducibility.

342

Past and Present in DeNOx Catalysis

Table 11.6. Experimental levels of reaction conditions which were investigated for response surface study of NOx storage and reduction catalysts Level

Lowest Low Mid High Highest

NO conc. (ppm)

O2 conc. (%)

Equivalent CO conc. (%)

Reductant type

Space velocity (mL/h/g)

Temp. (K)

– 920 1420 1930 2430

2 4 6 8 10

– 1 18 26 –

– C2 H4 50/50 Mix CO –

– – 42 000 – –

548 598 648 698 748

Eqn (4) due to the capacity for 1 mol of C2 H4 to reduce six times more moles of NO as compared to CO. Equiv CO = CO + 6∗ C2 H4

(4)

A complete list of the reaction conditions tested for this response surface design can be found in [76]. The center point reaction condition was repeated six times. This was done to measure the variability of the reaction system. Also, the space velocity is kept constant, as it was the least important factor predicted by screening design, for all the reaction conditions. The purpose of this nested response surface design was to develop an empirical model in the form of Eqn (5) to relate the five reaction condition variables and the three catalyst composition variables to the observed catalytic performance. Y = Cmj  + 1 mj X1 + 2 mj X2 + · · · + 12 mj X1 X2 + · · · + 1 mj X12 + · · · (5) In this equation, Y is the catalyst performance, the variables Xi and mj are normalized variables representing the reaction conditions and catalyst’s metal weight loadings, respectively. The model coefficients C, i , and i , are functions of the catalyst composition, as shown in Eqns (6) and (7), where mj refers to the nominal weight loading of Pt, Ba, or Fe. The equation for i takes the same form as Eqns (6) and (7). Cmj  = Cg + 1 m1 + 2 m2 + · · · + 12 m1 m2 + · · · + 1 m21 + 2 m22 + · · ·

(6)

i mj  = Ci + 1 m1 + 2 m2 + · · · + 12 m1 m2 + · · · + 1 m21 + 2 m22 + · · ·

(7)

To fit the data to the general model of Eqn (5), the experimental data for each catalyst were first treated separately by modeling the data with respect to only reaction conditions. These empirical models are referred to as catalyst specific models. These models gave the catalyst composition dependent parameters i.e. C’s, ’s, and ’s for each catalyst. After developing models for each catalyst, the models for C(xj ), (xj ), and (xj ) were then fitted for each parameter in Eqn (5). The accuracy of this final general model was verified by comparing the model predictions with the experimental results for some additional data points which were not included in the estimation of the model coefficients earlier. One of the most critical limitations of these models is that, like any other empirical model, they can only be used to predict the catalyst performance within the parameter space studied and any extrapolation using these models can give significant errors.

Design of Experiments for Optimization of DeNOx Catalysts

343

The catalytic performance for this response surface study was determined by the initial NOx storage/reduction. The initial NOx storage/reduction is a combination of NSR capabilities during the initial fuel lean phase, i.e. right after the switch from fuel rich to fuel lean conditions. The initial NOx storage/reduction is calculated by setting a cutoff NOx concentration above which further NOx storage or reduction is inconsequential to the usefulness of the catalyst. The area between the inlet NOx concentration and the outlet NOx concentration until reaching the cutoff NOx limit was integrated and converted to moles of NOx in the same manner in the screening design. The cutoff NOx concentration for this study was chosen to be 300 ppm. Further details about this criterion can be found in [76]. The NSR catalysts were tested for 30 min fuel lean and 15 min fuel rich cycles. The cycle times were chosen so that the reaction products from all catalysts achieved steady-state concentrations after switching from lean to rich conditions. Because catalysts were exposed to multiple lean and rich cycles before testing, the performance measured under these conditions was independent of cycling conditions. This procedure permitted the independent study of variables affecting only the NSR process under fuel lean conditions i.e. gas composition, catalyst composition and temperature. All the catalysts were tested twice for all the different reaction conditions to get an idea of repeatability. The results from these testing showed that the largest difference between the intial NOx storage/reduction values, based on the repeated runs of different catalysts, was less than 7.2 mol NOx . In addition, the repeats of the center reaction condition for 0.5Pt/7.5Ba/2.5Fe_4 catalyst showed a standard deviation of 1.2 mol NOx . In general, the initial NOx storage/reduction data collected with the HTE setup were very reproducible. The first step in the data analysis was the development of catalysts specific models similar to Eqn (2), but with a constant C, , and , for all catalysts. All the catalysts were analyzed for all the reaction conditions to develop empirical models for initial NOx storage/reduction as a function of the reaction conditions. However, due to space constraints we will only discuss the model for the 0.5Pt/7.5Ba/2.5Fe_4 catalyst and a more comprehensive discussion can be found in [76]. The fitted model coefficients for the 0.5Pt/7.5Ba/2.5Fe_4 catalyst are shown in Table 11.7. Here the coefficients are for the normalized reaction conditions for the experimental data averaged over the repeated runs for all the reaction conditions. Only coefficients significant to 95% confidence (as indicated by a p value of 0.05 or less) were included. The details regarding the statistical analysis can be found in [15,76]. To demonstrate the model adequacy in fitting the initial NOx storage/reduction data, the model predictions are plotted versus the experimental average, for 0.5Pt/7.5Ba/2.5Fe_4 in Figure 11.10a).The single error bar is estimated from the standard deviation for center point condition repeated six times. Also included are the results for five additional experimental conditions, Figure 11.10b), which were not included in the development of the model parameters, or in the original experimental design but were run to validate the empirical model. The experimental conditions for these five additional experimental runs are listed in Table 11.8. From Figure 11.10, it can be seen that the model not only matches the original data used to develop the model coefficients, but it also correctly predicts the additional experimental data within the experimental error. Therefore, it was concluded that the polynomial model adequately fits the initial NOx storage/reduction data as a function of the five reaction condition parameters over the investigated experimental range.

344

Past and Present in DeNOx Catalysis Table 11.7. Statistical analysis of the reaction condition study for 0.5Pt/7.5Ba/2.5Fe_4 Normalized variables

Init. NOx Stor./Red. (∗ 10−6 mol NOx ) Coefficient

Constant 95% Confidence interval (±)

154 14

Main effects NO O2 Reductant concentration CO fraction (remainder as C2 H4  Temperature

071 096 008 132 −538

Non-linear terms NO ∗ NO O2 ∗ O2 Red. conc. ∗ Red. conc. CO frac. ∗ CO frac. Temp. ∗ Temp. Interactions NO ∗ O2 NO ∗ Red. conc. NO ∗ CO frac. NO ∗ Temp. O2 ∗ Red. conc. O2 ∗ CO frac. O2 ∗ Temp. Red. conc. ∗ CO frac. Red. conc. ∗ Temp. CO frac. ∗ Temp.

– – – 469 – 045 105 −035 09 – −042 048 – – –

The analysis shown in Table 11.7 provides insight into what affected the initial NOx storage/reduction for a Pt/Ba/Fe based NSR catalyst. The most significant terms in predicting the initial NOx storage/reduction were temperature and reducing agent composition (i.e. CO fraction). In this study, it was found that the initial NOx storage/reduction increased as the temperature decreased (see Table 11.7) over the temperature range investigated. It was also found that when the reducing agent was either CO or C2 H4 , the initial NOx storage/reduction was higher for 0.5Pt/7.5Ba/2.5Fe_4 than when the two reducing agents were mixed. This observation is significant for automotive applications because automotive exhaust contains mixtures of hydrocarbons and CO, and a large percentage of published studies relating to NSR catalysts were done with a single reducing agent (see for example, [78–80] for recent examples). More detailed analysis of these results can be found in [76] and will not be discussed here. The single catalyst models for initial NOx storage/reduction were then extended to include the catalyst composition, as shown by Eqns (5), (6) and (7). This general model included the nominal weight loadings of Pt, Ba, and Fe, in addition to the five reaction

Design of Experiments for Optimization of DeNOx Catalysts (a) Initial NOx storage/reduction (10–6 moles NOx )

40

345

Experimental Values Model Prediction

30

20

10

0

Initial NOx storage/reduction (10–6 moles NOx )

(b) Experimental Values Model Predictions 20

15

10

5

0

Figure 11.10. Single catalyst model predictions versus experimental initial NOx storage/reduction results for catalyst 0.5Pt/7.5Ba/2.5Fe_4 for a) Training data set b) Validation data set for batch 2.

conditions. It was found that the general model predictions were within the experimental accuracy for different reactors. The most significant parameters in the general model mirror the results shown in Table 11.7, where the constant and the coefficients for the CO fraction, square of the CO fraction, temperature, and the interaction between the NO concentration and the reducing agent concentration were the largest. The dependence of these normalized parameters on the normalized metal composition of the catalyst is shown in Table 11.9. For these five parameters, the Pt weight loading is the most influential in determining the numerical value of the coefficient. The constant, in Eqn (5) for the general model, is related to the average NOx storage/reduction for all catalysts under all reaction conditions. This constant was dependent on all three metal loadings.

346

Past and Present in DeNOx Catalysis

Table 11.8. Additional reaction conditions collected for validation of the model performance Reaction condition ID

NO conc. (ppm)

O2 conc. (%)

Eq CO conc. (%)

CO frac.

CO conc. (%)

C2 H4 conc.

Space velocity (mL/h/g cat)

Temp. (K)

Add01 Add02 Add03 Add04 Add05

1170 1680 1170 1680 1420

5 7 5 7 6

22 22 14 14 18

05 1 0 05 05

11 22 0 07 09

018 0 023 012 015

42 500 42 500 42 500 42 500 42 500

698 598 598 698 648

Table 11.9. The general model dependence of the largest normalized parameter coefficients on normalized Pt, Ba, and Fe weight loadings Constant

Init. NOx stor./red. (∗ 10−6 mol NOx )

Constant Pt Ba Fe Pt 2 Pt ∗ Ba

168 8 9 32 −57 71

CO fraction Constant Pt Ba Fe Pt2 Fe2 Pt ∗ Ba

24 −2 −29 19 – −24 −37

Temperature Constant Pt Pt2

−53 −14 44

CO fraction2 Constant Pt Pt2

48 17 −32

NO ∗ Reductant concentration Constant Pt Pt2

26 01 −24

Design of Experiments for Optimization of DeNOx Catalysts

347

The least significant effect was the weight loading of the Fe, which has previously been shown to only slightly increase NOx storage for Pt/Ba catalysts [75]. The Pt and the Ba strongly influenced the initial NOx storage/reduction, as seen in Figure 11.11 where the change in the model constant is shown as a function of the Pt and Ba loading.

(a)

Initial NOx storage/reduction (10–6 moles NOx )

40 35 30 25 20 15 10 5 0 0 2

1.0

% 4 6 Ba riu 8 m 10 loa 12 din g 14

0.8 0.6 0.4 0.2 0.0

ing

oad

tl %P

(b)

40

30 25 20

1.0

15

0.8 0.6

10

0.4 5 0

% Pt loading

Initial NOx storage/reduction (10–6 moles NOx )

35

0.2 0

2

4

6

8

10

12

14

0.0

% Barium loading

Figure 11.11. The normalized parameter coefficients and their dependence on the normalized Pt, Ba, and Fe weight loadings, following the form of Eqns (1) and (2) for NOx concentration and N2 selectivity dependence of the constant in the catalyst specific model, on the weight loadings of Pt and Ba : (a) and (b) Two different views.

348

Past and Present in DeNOx Catalysis

From this figure, it can be seen that the maximum initial NOx storage/reduction occurred at the highest weight loading of Pt and Ba. However, for each Ba concentration there is an optimum Pt loading necessary to maximize the initial NOx storage/reduction. A similar result has recently been observed by Castoldi et al., for initial NOx storage/reduction with a cutoff NOx concentration between 25–50 ppm NOx [78]. Castoldi et al. observed that with increasing Ba loading, the Pt dispersion decreased. Consequently, increasing the Ba content beyond 25 w/w% resulted in a decrease in initial NOx storage/reduction as the Pt dispersion decreased. The decrease in NOx storage/reduction for these two cases is most likely due to a decrease in the contact area between Pt and Ba. It has been proposed that NO2 spillover from Pt to neighboring Ba storage sites is a vital step in the NOx storage process [64,69,70,81]. Although it has been shown that NOx can adsorb directly to Ba storage sites [82], the NOx concentration for a 15% w/w Ba catalyst never dropped below 300 ppm in these studies and therefore no initial NOx storage/reduction was observed. Therefore, rapid adsorption of NOx directly to Ba storage sites is most likely a slow process and insignificant when compared to the NOx storage of Ba sites in close proximity to the Pt sites during the initial NOx storage/reduction measured for these studies. In this general model, the Pt weight loading completely determined the model dependence on the temperature, the square of the CO fraction, and the interaction between the NO and reducing agent concentration. In a more complex fashion, the reducing agent composition (i.e. the CO fraction) coefficient was dependent on the weight loadings of all three metals. The multiple metal complexities in the reducing agent composition is complicated due to the nature of the dependence of initial NOx storage/reduction on both the fuel lean phase and fuel rich phase. In addition, using CO as a reducing agent resulted in more complete NOx conversion in the fuel rich phase than C2 H4 . CO more effectively reduced NOx in the fuel rich state, due to a higher oxidation activity [83], and thereby dominated the initial NOx storage/reduction for Pt only catalysts even though C2 H4 more effectively reduced NOx in the fuel lean state [75]. In contrast to the Ba-free catalysts, the dominant factors affecting the initial NOx storage/reduction for Ba-containing catalysts were the conditions of the lean phase. Therefore, in spite of the lower capacity to reduce NOx under fuel rich conditions, C2 H4 demonstrated an increased initial NOx storage/reduction over Ba-containing catalysts. The dependence of the initial NOx storage/reduction on the catalyst composition and the reducing agent was captured by the general initial NOx storage/reduction model. The general model predicted an increase of ∼10 mol NOx for initial NOx storage/reduction in switching the reducing agent from C2 H4 to CO for the 1Pt catalyst. In contrast the general model predicted a decrease of ∼14 mol NOx for initial NOx storage/reduction for the same change in reaction conditions for the 1Pt/15Ba catalyst. Due to the multidimensional nature and size of the high-throughput dataset collected, the development of the empirical NOx storage correlations was very useful in the identification of trends in the data. Further details of the trend predictions and model analysis are discussed in [76]. For all the above studies, the cycle time used for rich and lean phase, 15 and 30 min, respectively, was long and far from being realistic. In addition to such long cycle studies, to get an estimate of the catalyst performance under realistic conditions, we have also done second type of studies known as short cycle studies. In these studies, the total cycle time was kept relatively short (∼s) to simulate the actual exhaust conditions, and

Design of Experiments for Optimization of DeNOx Catalysts

349

the catalysts were screened for average NOx conversion capacity. The total cycle time was defined as the time between the beginning of one lean phase and the beginning of the next lean phase, and the lean fraction was defined as the fraction of the total cycle time spent in the fuel lean phase. A graphical representation of the total cycle time and lean fraction can be found in [84]. We have only studied the effect of catalyst composition and cyclic operating conditions, i.e. total cycle time and fraction of time spent in lean phase. The reaction conditions were held constant at 0.15% v/v NO, 6% v/v O2 , 0.9% v/v CO, and 0.15% v/v C2 H4 in He for the fuel lean phase at a space velocity of 42 000 mL/h/g catalyst. The fuel rich phase was simulated by replacing the oxygen with an equal volume of helium while maintaining all other flow rates constant. All reactions were performed at 648 K and atmospheric pressure. The temperature for these studies was fixed at 648 K as it has been shown to be the temperature for optimum performance of Pt/Ba based NSR catalysts [68]. The total cycle time and the lean fraction, for short cycle studies, varied according to a response surface design for two variables, as indicated by the design points in Figure 11.12. The center cycling condition, of 90 s total cycle time and a lean fraction of 0.675, was repeated at least three times to estimate the variability in the measurements. All the design conditions, including the repeats for the centre cycling condition, were tested randomly and it typically takes 5–6 h to sample all the points. All the catalysts which were used for previous study were also used here. The total cycle time and the lean fraction were fitted to the catalytic response (R) using an empirical model of the form of Eqn (8), where C, , and , were fit to the experimental data. R = C + 1 CT  + 2 LF  + 12 CT LF  + 1 CT 2 + 2 LF 2

(8)

Design points Validation points

140

Cycle time (s)

120

100

80

60

40 0.4

0.5

0.6

0.7

0.8

0.9

1.0

Lean fraction

Figure 11.12. Response surface cycling design for total cycle time and lean fraction of the total cycle time [84].

350

Past and Present in DeNOx Catalysis

Table 11.10. Experimental levels of reaction conditions and catalyst composition which were investigated for response surface study of NOx storage and reduction catalysts Level

Cycle time (s)

Lean fraction

Pt loading (%)

Ba loading (%)

Fe loading (%)

Low High

60 120

05 085

0 1

0 15

0 5

In this equation, R is the catalyst performance, determined by average NOx concentration. CT and LF are normalized variables representing the operating conditions, i.e., cycle time and lean fraction. The operating condition variables were normalized according to Eqn (5) by assigning the low value listed in Table 11.10 as −1 and the high value as +1. Using the empirical model, insight into the operation of the NSR catalysts was gained by considering which factors govern the catalytic performance. All cycling conditions were preceded by running the catalysts under fuel rich conditions for 15 min to ensure that all catalysts began each cycle free of reversibly stored nitrates. After completion of the long fuel rich phase, the reaction gases were cycled between fuel rich and fuel lean conditions according to the times specified by the cycling design shown in Figure 11.12. The catalysts were exposed to multiple lean/rich cycles until the catalysts behavior becomes reproducible, typically two to three cycles. The NOx concentration was then averaged over multiple cycles and used to describe the catalytic performance which was defined using Eqn (9). Here NOx avg is the average NOx effluent concentration and NOx i is the inlet NOx concentration.   NOx avg ∗ NOx Conv = 100 1 − (9) NOx i Figure 11.13 shows the response surface developed using empirical model predictions for 0.5Pt/7.5Ba/2.5Fe_4 catalyst. The red points are the combination of cycle time and lean fraction which were run to validate the model. It was found that the model adequately predicted the experimental data over the entire region of investigation. Figure 11.14 graphically represents the model prediction for NOx conversion as a function of cycling time and lean fraction. A large region of 100% NOx conversion can be seen corresponding to cycle times between 60 and 125 s and lean fractions below 0.65. From Figure 11.14 it would appear that at lean fractions below ∼0.5, the NOx conversion decreased; however, the model data for the regions of long (>120 s) and short (<60 s) total cycle times at lean fractions less than 0.5 are extrapolations of the experimental data. Since the model is an empirical polynomial, one needs to take care in extrapolating beyond the limits of the experimental data. From the model results presented in Figure 11.14, it can be seen that the maximum NOx conversion for a fixed lean fraction occurred at an intermediate cycle time. Similar results have been reported by Han et al. [85] and Kabin et al. [86], and it was concluded that, at very short cycle times, the catalyst responded as if the fuel rich and fuel lean feed gases were mixed. However, an additional explanation for the decrease in NOx

Design of Experiments for Optimization of DeNOx Catalysts

351

100

% NOx conversion

90 80 70 60 50 120 110 100

40

Lean

0.70

fract

0.75

ion

70

0.80

0.85

60

e

80

tim

0.65

cle

0.60

(s )

90

0.55

Cy

30 0.45 0.50

0.90

Figure 11.13. Model prediction and validation for % NOx conversion as a function of cycle time and lean fraction (red points are the validation points) for 0.5Pt/7.5Ba/2.5Fe_4 catalyst.

100 90 80 70 60 50 40 30 20 10 0

120

Total cycle time (s)

110 100 90

NOx conversion (%)

130

80 70 60 50 0.5

0.6

0.7

0.8

0.9

Lean fraction

Figure 11.14. Model results for NOx conversion by 0.5Pt/7.5Ba/2.5Fe_4 as a function of total cycle time and the lean fraction of the total cycle time [84].

conversion at short-cycle time could be the NOx breakthrough that is common for NSR catalysts (see Figure 11.14 at 25 s) upon switching the reactant mixture from fuel lean to fuel rich [68,86–89]. A decrease in the total cycle time increases the frequency of switches from fuel lean to fuel rich in turn increasing the NOx breakthrough fraction of

352

Past and Present in DeNOx Catalysis

the total cycle time which further leads to an increase in NOx concentration or decrease in NOx conversion at short cycle time. Additionally, after the development of similar models for each catalyst, the single catalyst models were then combined into a general model encompassing the metal compositions in addition to the total cycle time and lean fraction, similar to what was done in the previous section for long cycle studies. Details can be found in [84]. Excellent agreement was found between the general model and the experimental data. The general model was further used to interpret the NOx conversion results. The most influential terms in the general NOx conversion model affecting the NOx conversion and their dependencies on the Pt, Ba, and Fe weight loadings, are shown in Table 11.11. From this table, it can be seen that the NOx conversion depends on the loadings of all three metals. Using model predictions, it was found that the highest NOx conversions were generally achieved at Pt loadings above 0.5% w/w and Ba loadings above 7.5% w/w. In addition, there is an optimum amount of Pt, which depends on the amount of Ba present and necessary to maximize the initial NOx storage/reduction. The initial NOx storage/reduction capacity decreases if the Ba content is increased beyond a certain optimum value. This was attributed to decrease in Pt dispersion and also reduced contact area between Pt

Table 11.11. Normalized parameter coefficients and their dependence on the normalized Pt, Ba, and Fe weight loadings, following the form of Eqns (1) and (2) for NOx concentration and N2 selectivity Constant

NOx concentration (ppm)

Constant Pt Ba Fe Pt2 Ba2 Fe2 Pt ∗ Ba Ba ∗ Fe

49 −456 −154 −207 581 164 −148 −286 –

Lean fraction Constant Pt Ba Fe Pt 2 Pt ∗ Ba

366 85 42 80 −159 39

Lean fraction ∗ Lean fraction Constant Pt Ba Fe Pt2 Fe2 Pt ∗ Ba

236 80 43 – −165 – 54

Design of Experiments for Optimization of DeNOx Catalysts

353

and Ba. Also, rapid adsorption of NOx directly to Ba storage sites is most likely a slow process and insignificant compared to the NOx storage on Ba sites in close proximity to the Pt sites during the initial NOx storage/reduction period measured for these studies. The details of single catalyst models and the details of promotional effect of iron, which was also analyzed using these models, are discussed in depth [84]. In addition, the role of catalyst composition, operating variables and reaction condition on N2 O production and selectivity has also been analyzed using the similar DOE strategy.

7. EXPLORING NEW MATERIALS All the above sections dealt with the systematic study of Pt, Ba and Fe based NSR systems using the combination of HTE and DOE strategy. No doubt quickly extracting useful information from wide variety of systems was one of the objectives. However, the other main objective behind the development of HTE techniques was to accelerate the discovery process which in turn is the basis for any major technological advance. Using this as the guiding factor, we also tested a variety of new materials to check for their promotional effect on traditional Pt/Ba based NSR catalysts [90]. Figure 11.15 shows the results for some of these materials tested for initial NOx storage/reduction behavior. The addition of either Fe or Mn shows an increase in the NOx storage by 25–30%. Co, however, has a considerably higher promotional effect, increasing the NOx storage by more than 100%. Due to the significant increase in performance shown by the Co containing NSR catalyst, a detailed study was performed on NSR catalysts having different combinations of Pt, Ba, and Co to verify the promotional effect of Co. Figure 11.16 shows the performance of these catalysts with a variety of different weight loadings of Pt, Ba, and Co. The high promotional effect of Co was also evident from the excellent performance of the 1Pt/5Co/15Ba catalyst, demonstrating NOx storage nearly double that of 1Pt/15Ba catalyst. This promotional effect was mainly attributed to the increased 40

NOx storage (10–6 moles)

35 30 25 20 15 10 5 0

none

Fe

Mn

Co

Promoter

Figure 11.15. Promotional effect of different transition metals on initial NOx storage/reduction activity for 0.5Pt/7.5Ba/2.5 Promoter catalyst [90].

354

Past and Present in DeNOx Catalysis 75 70 65

NOx storage (*10–6 moles)

60 55 50 45 40 35 30 25 20 15 10 5 0

1Pt

5Co

15Ba

1Pt /5Co 1Pt /15Ba 5Co /15Ba 1Pt /15Ba /5Co

Catalyst composition

Figure 11.16. Initial NOx storage/reduction for different catalysts under fuel lean conditions [90].

number of sites available for oxidation of NO to NO2 and the increased contact area for NO2 spillover to neighboring BaNOx storage sites [90]. It was anticipated that Co could replace Pt as the oxidizing metal in NSR catalysts and thereby reduce the total noble metal content in NSR catalysts. A dramatic reduction in cost can be achieved, without sacrificing the performance, by replacing Pt with Co as the active oxidizing metal. To verify this claim, we have further explored Pt/Co/Ba NSR system using DOE strategy for realistic short cycle time studies [91]. It has been found that addition of 5% Co improves the performance of 1%Pt/15%Ba and 1%Rh/15%Ba catalysts at higher lean fractions, allowing a substantial improvement in overall fuel efficiency. Also, addition of 5%Co to 1Pt/15Ba allows for the reduction of noble metal content to 0.25% without affecting the performance of the catalysts. Therefore, the addition of Co to NSR catalysts is expected to have a large impact in terms of improving fuel efficiency as well as reducing the cost of the catalysts. All the above experiments were done on FTIR imaging based HTE setup. With the conventional single reactor setup, it would have taken a long time to do a similar study and would have been impossible to analyze the data without the help of DOE strategies. Not only were we able to successfully analyze a complicated system, but we were also able to discover a potential new material for the specific application (NSR system). In this chapter, we have demonstrated the role of HTE both for speeding up material discovery, and for rapid understanding of the complex catalytic systems. The methodology of the nested experimental design employed in above studies helped to identify simple linear trends, such as temperature dependence, and complex trends, such as the interplay between reducing agent and catalyst composition on the initial NOx storage/reduction. It could further be used to identify the optimal operating regime for a given catalyst or vice versa and to quantify the promotional effect of different promoters [91].

Design of Experiments for Optimization of DeNOx Catalysts

355

We have also applied similar strategy for studying the effect of water on the performance of NSR catalysts and its dependence on the catalyst composition [92]. In addition, we have also explored the effect of SO2 , which is a known poison for these catalysts, for Pt/Ba/Fe based NSR system [75]. The trends and the significant factors identified using response surface models could then be used to guide the mechanistic modeling of these multi factorial systems. Finally, it can be seen that combination of HTE in conjunction with statistical experimental design can be used to gain understanding into the functionality of complex catalytic systems, as well as to design better catalysts for specific applications.

REFERENCES [1] Senkan, S. (2001) Combinatorial heterogeneous catalysis – a new path in an old field. Angew. Chem. Int. Ed., 40, 312. [2] Harmon, L. (2003) Experiment planning for combinatorial materials discovery. J. Mater. Sci., 38, 4479. [3] Maxwell, I.E., van den Brink, P., Downing, R.S. et al. (2003) High-throughput technologies to enhance innovation in catalysis. Top. Catal., 24, 125. [4] Potyrailo, R.A. and Amis, E.J. (2004) High Throughput Analysis: A Tool of Combinatorial Materials Science. Kluwer Academic Plenum Pub. Corp., New York. [5] Krantz, K., Ozturk, S. and Senkan, S. (2000) Application of combinatorial catalysis to the selective reduction of NO by C3H6. Catal. Today, 62, 281. [6] Ozturk, S. and Senkan, S. (2002) Discovery of new fuel-lean NO reduction catalyst leads using combinatorial methodologies. Appl. Catal. B Environ., 38, 243. [7] Richter, A., Langpape, M., Kolf, S. et al. (2002) Combinatorial preparation and highthroughput catalytic tests of multi-component deNO(x) catalysts. Appl. Catal. B Environ., 36, 261. [8] Busch, O.M., Hoffmann, C., Johann, T.R.F. et al. (2002) Application of a new color detection based method for the fast parallel screening of deNO(x) catalysts. J. Am. Chem. Soc., 124, 13527. [9] Hoogenboom, R., Meier, M.A.R. and Schubert, U.S. (2003) Combinatorial methods, automated synthesis and high-throughput screening in polymer research: past and present. Macromol. Rapid Commun., 24, 16. [10] Hanak, J.J. (1970) The “multiple-sample” concept in materials research: synthesis, compositional analysis and testing of entire multicomponent systems. J. Mater. Sci., 5, 964. [11] Thomas, R., Moulijn, J.A., Debeer, V.H.J. and Medema, J. (1980) Structure-metathesisactivity relations of silica supported molybdenum and tungsten-oxide. J. Mol. Catal., 8, 161. [12] Creer, J.G., Jackson, P., Pandy, G. et al. (1986) The design and construction of a multichannel microreactor for catalyst evaluation. Appl. Catal., 22, 85. [13] Hendershot, R.J., Snively, C.M. and Lauterbach, J. (2005) High-throughput catalytic science. Chem. A Eur. J., 11, 806. [14] Dar, Y.L. (2004) High-throughput experimentation: a powerful enabling technology for the chemicals and materials industry. Macromol. Rapid Commun., 25, 34. [15] Lawson, J. and Erjavec, J. (2000) Modern Statistics for Engineering and Quality Improvement. Brooks/Cole-Thomson Learning, Pacific Grove, CA 179–515. [16] Tagliabue, M., Carluccio, L.C., Ghisletti, D. et al. (2003) Multivariate approach to zeolite synthesis. Catal. Today, 81, 405.

356

Past and Present in DeNOx Catalysis

[17] Bem, D.S., Gillespie, R.D., Erlandson, E.J. et al. (2003) Combinatorial experimental design using the optimal-coverage algorithm. In Experimental Design for High Throughput Materials Development (J.N. Cawse, ed.), J. Wiley, Hoboken, NJ, p. 89. [18] Spivack, J.L., Cawse, J.N., Whisenhunt, J. et al. (2003) Combinatorial discovery of metal co-catalysts for the carbonylation of phenol. Appl. Catal. A Gen., 254, 5. [19] Urschey, J., Kuhnle, A. and Maier, W.F. (2003) Combinatorial and conventional development of novel dehydrogenation catalysts. Appl. Catal. A Gen., 252, 91. [20] Snively, C.M., Oskarsdottir, G. and Lauterbach, J. (2001) Parallel analysis of the reaction products from combinatorial catalyst libraries. Angew. Chem. Int. Ed., 40, 3028. [21] Hahndorf, I., Buyevskaya, O.V., Langpape, M. et al. (2002) Experimental equipment for high-throughput synthesis and testing of catalytic materials. Chem. Eng. J., 89, 119. [22] Wang, H., Liu, Z.M. and Shen, J.H. (2003) Quantified MS analysis applied to combinatorial heterogeneous catalyst libraries. J. Comb. Chem., 5, 802. [23] Snively, C.M., Oskarsdottir, G. and Lauterbach, J. (2001) Chemically sensitive parallel analysis of combinatorial catalyst libraries. Catal. Today, 67, 357. [24] Senkan, S., Krantz, K., Ozturk, S. et al. (1999) High-throughput testing of heterogeneous catalyst libraries using array microreactors and mass spectrometry. Angew. Chem. Int. Ed., 38, 2794. [25] Cong, P., Doolen, R., Fan, Q. et al. (1999) High-throughput synthesis and screening of combinatorial heterogeneous catalyst libraries. Angew. Chem. Int. Ed., 38, 484. [26] Orschel, M., Klein, J., Schmidt, H.-W. et al. (1999) Detection of reaction selectivity on catalyst libraries by spatially resolved mass spectrometry. Angew. Chem. Int. Ed., 38, 2791. [27] Desrosiers, P., Guram, A., Hagemeyer, A. et al. (2001) Selective oxidation of alcohols by combinatorial catalysis. Catal. Today, 67, 397. [28] Huybrechts, W., Mijoin, J., Jacobs, P.A. et al. (2003) Development of a fixed-bed continuousflow high-throughput reactor for long-chain n-alkane hydroconversion. Appl. Catal. A Gen., 243, 1. [29] Li, W., Gracia, F.J. and Wolf, E.E. (2003) Selective combinatorial catalysis; challenges and opportunities: the preferential oxidation of carbon monoxide. Catal. Today, 81, 437. [30] Corma, A., Serra, J.M. and Chica, A. (2003) Discovery of new paraffin isomerization catalysts based on SO42-/ZrO2 and WOx/ZrO2 applying combinatorial techniques. Catal. Today, 81, 495. [31] Serra, J.M., Corma, A., Farrusseng, D. et al. (2003) Styrene from toluene by combinatorial catalysis. Catal. Today, 81, 425. [32] Yamada, Y., Ueda, A., Shioyama, H. et al. (2003) High throughput experiments on methane partial oxidation using molecular oxygen over silica doped with various elements. Appl. Catal. A Gen., 254, 45. [33] Serra, J.M., Chica, A. and Corma, A. (2003) Development of a low temperature light paraffin isomerization catalysts with improved resistance to water and sulphur by combinatorial methods. Appl. Catal. A Gen., 239, 35. [34] Hagemeyer, A., Borade, R., Desrosiers, P. et al. (2002) Application of combinatorial catalysis for the direct amination of benzene to aniline. Appl. Catal. A Gen., 227, 43. [35] Hoffmann, C., Schmidt, H.W. and Schuth, F. (2001) A multipurpose parallelized 49-channel reactor for the screening of catalysts: methane oxidation as the example reaction. J. Catal., 198, 348. [36] Reetz, M.T., Kuhling, K.M., Wilensek, S. et al. (2001) A GC-based method for highthroughput screening of enantio selective catalysts. Catal. Today, 67, 389. [37] Yamada, Y., Ueda, A., Zhao, Z. et al. (2001) Rapid evaluation of oxidation catalysis by gas sensor system: total oxidation, oxidative dehydrogenation, and selective oxidation over metal oxide catalysts. Catal. Today, 67, 379.

Design of Experiments for Optimization of DeNOx Catalysts

357

[38] Al-Saeedi, J.N. and Guliants, V.V. (2002) High-throughput experimentation in multicomponent bulk mixed metal oxides: Mo-V-Sb-Nb-O system for selective oxidation of propane to acrylic acid. Appl. Catal. A Gen., 237, 111. [39] Gomez, S., Peters, J.A., van der Waal, J.C. et al. (2003) High-throughput experimentation as a tool in catalyst design for the reductive amination of benzaldehyde. Appl. Catal. A Gen., 254, 77. [40] Desrosiers, P., Guan, S.H., Hagemeyer, A. et al. (2003) Application of combinatorial catalysis for the direct amination of benzene to aniline. Catal. Today, 81, 319. [41] Su, H. and Yeung, E.S. (2000) High-throughput screening of heterogeneous catalysts by laser- induced fluorescence imaging. J. Am. Chem. Soc., 122, 7422. [42] Senkan, S., Ozturk, S., Krantz, K. et al. (2003) Photoionization detection (PID) as a high throughput screening tool in catalysis. Appl. Catal. A Gen., 254, 97. [43] Kubanek, P., Busch, O., Thomson, S. et al. (2004) Imaging reflection IR spectroscopy as a tool to achieve higher integration for high-throughput experimentation in catalysis research. J. Comb. Chem., 6, 420. [44] Johann, T., Brenner, A., Schwickardi, M. et al. (2003) Listening to catalysis – a real time parallel method for high throughput product analysis. Catal. Today, 81, 449. [45] Staiger, C.L., Loy, D.A., Jamison, G.M. et al. (2003) A parallel colorimetric method for the rapid discovery and optimization of heterogeneous hydrodesulfurization catalysts. J. Am. Chem. Soc., 125, 9920. [46] Moates, F.C., Somani, M. Annamalai, J. et al. (1996) Infrared thermographic screening of combinatorial libraries of heterogeneous catalysts. Ind. Eng. Chem. Res., 35, 4801. [47] Gracia, F., Li, W. and Wolf, E.E. (2003) The preferential oxidation of CO: selective combinatorial activity and infrared studies. Catal. Lett., 91, 235. [48] Senkan, S.M. (1998) High-throughput screening of solid-state catalyst libraries. Nature, 394, 350. [49] Colarusso, P., Kidder, L.H., Levin, I.W. et al. (1998) Infrared spectroscopy imaging: from planetary to cellular systems. Appl. Spectrosc., 52, 106A. [50] Snively, C.M., Katzenberger, S., Oskarsdottir, G. et al. (1999) Fourier-transform infrared imaging using a rapid-scan spectrometer. Opt. Lett., 24, 1841. [51] Snively, C.M., Oskarsdottir, G. and Lauterbach, J. (2000) Chemically sensitive high throughput parallel analysis of solid phase supported library members. J. Comb. Chem., 2, 243. [52] Snively, C.M. and Lauterbach, J. (2002) Sampling accessories for the high-throughput analysis of combinatorial libraries using spectral imaging. Spectroscopy, 17, 26. [53] Hendershot, R.J., Lasko, S.S., Fellmann, M.-F. et al. (2003) A novel reactor system for high throughput catalyst testing under realistic conditions. Appl. Catal. A Gen., 254, 107. [54] Hendershot, R.J., Vijay, R., Feist, B.J. et al. (2005) Multivariate and univariate analysis of infrared imaging data for high-throughput studies of NH3 decomposition and NOx storage and reduction catalysts. Meas. Sci. Technol. 16, 302. [55] Snively, C.M. and Lauterbach, J. (1999), FTIR imaging for chemically sensitive, high throughput analysis of combinatorial libraries. In Combinatorial Catalysis and HighThroughput Catalyst Design and Testing (E.G. Derouane, F. Lemos, A. Corma and F.R. Ribeiro, eds.), Kluwer Academic Publishers, Vilamoura, Portugal, pp. 437. [56] Hendershot, R.J. (2004) Statitically Guided High-Throughput Experimentation of NOx storage and Reduction. Doctoral Thesis, Morris Library, University of Delaware. [57] Griffiths, P.R. and de Haseth, J.A. (1986) Fourier Transform Infrared Spectrometry. Wiley, New York. [58] Hendershot, R.J., Fanson, P.T., Snively, C.M. et al. (2003) High-throughput catalytic science: parallel analysis of transients in catalytic reactions. Angew. Chem. Int. Ed., 42, 1152. [59] Takeuchi, I., Lauterbach, J. and Fasolka, M.J. (2005) Combinatorial materials synthesis. Mater. Today, 8, 18.

358

Past and Present in DeNOx Catalysis

[60] Box, G.E.P., Hunter, W.G. and Hunter, J.S. (1978) Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building. Wiley, New York. [61] Hicks, C.R. and Turner, K.V. (1999) Fundamental Concepts in the Design of Experiments, 5th edn., Oxford University Press, New York. [62] Yamazaki, K., Suzuki, T., Takahashi, N. et al. (2001) Effect of the addition of transition metals to Pt/Ba/Al2 O3 catalyst on the NOx storage-reduction cataysis under oxidizing conditions in the presence of SO2 . Appl. Catal. B Environ., 30, 459. [63] Welz, B. and Sperling, M. (1999) Atomic Absorption Spectrometry, 3rd completely revised. edn., Wiley-VCH, Weinheim, New York. [64] Fridell, E., Persson, H., Westerberg, B. et al. (2000) The mechanism for NOx storage. Catal. Lett., 66, 71. [65] Amberntsson, A., Persson, H., Engstrom, P. et al. (2001) NOx release from a noble metal/BaO catalyst: dependence on gas composition. Appl. Catal. B, 31, 27. [66] Balcon, S., Potvin, C., Salin, L. et al. (1999) Influence of CO2 on storage and release of NOx on barium-containing catalyst. Catal. Lett., 60, 39. [67] Mahzoul, H., Brilhac, J.F. and Gilot, P. (1999) Experimental and mechanistic study of NOx adsorption over NOx trap catalysts. Appl. Catal. B, 20, 47. [68] Fridell, E., Skoglundh, M., Westerberg, B. et al. (1999) NOx storage in barium-containing catalysts. J. Catal., 183, 196. [69] James, D., Fourre, E., Ishii, M. et al. (2003) Catalytic decomposition/regeneration of Pt/Ba(NO3)(2) catalysts: NOx storage and reduction. App. Catal. B Environ., 45, 147. [70] Takahashi, N., Shinjoh, H., Iijima, T. et al. (1996) The new concept 3-way catalyst for automotive lean-burn engine: NOx storage and reduction catalyst. Catal. Today, 27, 63. [71] Acke, F. and Skoglundh, M. (1999) Comparison between ammonia and propane as the reducing agent in the selective catalytic reduction of NO under lean conditions over Pt black. Appl. Catal. B Environ., 20, 133. [72] Burch, R., Sullivan, J.A. and Watling, T.C. (1998) Mechanistic considerations for the reduction of NOx over Pt/Al2O3 and Al2O3 catalysts under lean-burn conditions. Catal. Today, 42, 13. [73] Poulston, S. and Rajaram, R.R. (2003) Regeneration of NOx trap catalysts. Catal. Today, 81, 603. [74] Cant, N.W. and Patterson, M.J. (2003) The effect of water and reductants on the release of nitrogen oxides stored on BaO/Al2O3. Catal. Lett., 85, 153. [75] Hendershot, R.J., Rogers, W.B., Snively, C.M. et al. (2004) Development and optimization of nox storage and reduction catalysts using statistically guided high-throughput experimentation. Catal. Today, 98, 375. [76] Hendershot, R.J., Vijay, R., Snively, C.M. et al. (2006) Response surface study of the performance of lean nox storage catalysts as a function of reaction conditions and catalyst composition. Appl. Catal. B Environ., 70, 160. [77] Cawse, J.N. (2003) Experimental Design for High Throughput Materials Development. J. Wiley, Hoboken, NJ. [78] Castoldi, L., Nova, I., Lietti, L. et al. (2004) Study of the effect of Ba loading for catalytic activity of Pt-Ba/Al2 O3 model catalysts. Catal. Today, 96, 43. [79] Amberntsson, A., Skoglundh, M., Jonsson, M. et al. (2002) Investigations of sulphur deactivation of NOx storage catalysts: influence of sulphur carrier and exposure conditions. Catal. Today, 73, 279. [80] Su, Y. and Amiridis, M.D. (2004) In situ FTIR studies of the mechanism of NOx storage and reduction on Pt/Ba/Al2 O3 catalysts. Catal. Today, 96, 31. [81] Gill, L.J., Blakeman, P.G., Twigg, M.V. et al. (2004) The use of NOx adsorber catalysts on diesel engines. Top. Catal., 28, 157.

Design of Experiments for Optimization of DeNOx Catalysts

359

[82] Fanson, P.T., Horton, M.R., Delgass, W.N. et al. (2003) FTIR analysis of storage behavior and sulfur tolerance in barium-based NOx storage and reduction (NSR) catalysts. Appl. Catal. B Environ., 46, 393. [83] Heck, R.M. and Farrauto, R.J. (1995) Catalytic Air Pollution Control: Commercial Technology. Van Nostrand Reinhold, New York. [84] Han, P.-H., Lee, Y.-K., Han, S.-M. et al. (2001) NOx storage and reduction catalysts for automotive lean-burn engines: effect of parameters and storage materials on NOx conversion. Top. Catal., 16/17, 165. [85] Kabin, K.S., Muncrief, R.L. and Harold, M.P. (2004) NOx storage and reduction on a Pt/BaO/alumina monolithic storage catalyst. Catal. Today, 96, 79. [86] Bögner, W., Krämer, M., Krutzsch, B. et al. (1995) Removal of nitrogen oxides from the exhaust of a lean-tune gasoline engine. Appl. Catal. B Environ., 7, 153. [87] Li, Y.J., Roth, S., Dettling, J. et al. (2001) Effects of lean/rich timing and nature of reductant on the performance of a NOx trap catalyst. Top. Catal., 16, 139. [88] Huang, H.Y., Long, R.Q. and Yang, R.T. (2001) A highly sulfur resistant Pt-Rh/TiO2 /Al2 O3 storage catalyst for NOx reduction under lean-rich cycles. Appl. Catal. B Environ., 33, 127. [89] Hendershot, R.J., Vijay, R., Snively, C.M. et al. (2006) High-throughput study of the performance of NOx storage and reduction catalysts as a function of cycling conditions and catalyst composition. Chem. Eng. Sci., 61, 3907. [90] Vijay, R., Hendershot, R.J., Rivera-Jiménez, S.M. et al. (2005) Noble metal free NOx storage catalysts using cobalt discovered via high-throughput experimentation. Catal. Commun., 6, 167. [91] Vijay, R., Snively, C.M. and Lauterbach, J. (2006) Performance of co-containing NOx storage and reduction catalysts as a function of cycling condition. J. Catal., 243, 368. [92] Hendershot, R.J., Vijay, R., Snively, C.M. et al. (2006) High-throughput study of the influence of H2O and CO2 on the performance of Nitrogen storage and reduction (NSR) catalysts. Appl. Surf. Sci., 252, 2588.

This page intentionally left blank

Chapter 12

PLASMA-ASSISTED NOX ABATEMENT PROCESSES: A NEW PROMISING TECHNOLOGY FOR LEAN CONDITIONS M. M˘agureanu1 and V. I. Pârvulescu2∗ 1

2

Department of Plasma Physics and Nuclear Fusion, National Institute for Lasers, Plasma and Radiation Physics, Atomistilor Street 409, R76900 Magurele–Bucharest, Romania

Department of Chemical Technology and Catalysis, University of Bucharest, 4–12 Regina Elisabeta Bvd., Bucharest 030016, Romania ∗

Corresponding author: Department of Chemical Technology and Catalysis, University of Bucharest, 4–12 Regina Elisabeta Bvd., Bucharest 030016, Romania. Tel.: +4021 4103178, Fax.: +4021 3159249, E-mail: [email protected]

Abstract This chapter deals with the removal of nitrogen oxides from gas streams by using non-equilibrium (non-thermal) plasma generated in atmospheric pressure electrical discharges. The advantage of using non-thermal plasma for promoting chemical reactions is its energy selectivity: most of the electrical energy dissipated in the discharge is used to produce high-energy electrons, and not spent on heating the entire gas stream. Typical electron temperatures in these discharges are 2–10 eV, while the background gas remains close to room temperature. The high-energy electrons seldom collide directly with the pollutant molecules, however, they excite, dissociate and ionize the background gas molecules. Thus, chemically active species are generated, which in turn, react with the pollutant molecules and decompose them. Different types of electrical discharges will be discussed with respect to NOx removal: corona discharges, dielectric barrier discharges (DBD), dielectric packed-bed discharges, and microwave discharges. Two issues of utmost importance will be emphasized throughout this chapter, namely: energy efficiency/energy consumption and by-product formation. In order to improve the efficiency for NOx removal by non-thermal plasma, different electrode geometries and electrode shapes have been investigated. The effect of the discharge gap length and of structured electrodes or multipoint electrodes, which enhance the local electric field, will be addressed. The discharges can be operated in d.c., a.c. or in pulsed mode. Comparisons between those operation modes will be discussed, as well as the influence of voltage polarity, amplitude and frequency. In pulsed discharges, the effect of pulse width and pulse repetition rate on NOx removal will be considered. Other experimental parameters found important for NOx decomposition are the gas flow rate, NOx initial concentration, reactor length, etc.

Past and Present in DeNOx Catalysis P. Granger and V.I. Pârvulescu (Editors)

© 2007 Elsevier B.V. All rights reserved.

362

Past and Present in DeNOx Catalysis

The gas mixture containing the nitrogen oxides is very important as well. Experiments and modeling carried out for N2 /NOx mixtures, or with addition of O2 , H2 O, CO2 and hydrocarbons will be discussed. Typical hydrocarbon additives investigated are ethane, propene, propane, 2-propene-1-ol, 2-propanol, etc. As compared to the case without hydrocarbons, NO oxidation occurs much faster when hydrocarbons are present. The reaction paths for NO removal change significantly, in fact the chemical mechanism itself is completely different from that of without hydrocarbon additives. Another additive investigated extensively is ammonia, used especially in corona radical shower systems. The combination of non-thermal plasma and heterogeneous catalysis will be addressed as well. Single stage plasma-catalytic systems, with the catalysts placed in the discharge zone, and two stage systems, with the catalysts placed downstream of the plasma will be discussed. Single-stage reactors have the advantage of reactions on the catalyst surface with short-lived active species generated in the discharge. In two-stage systems, or plasma enhanced selective catalytic reduction (PE-SCR) systems, the oxidation of NO to NO2 in the plasma increases the catalyst performance at low temperature (below 450–500 K).

1. SEVERAL BASIC ASPECTS OF NON-EQUILIBRIUM PLASMA In a non-equilibrium plasma (also called non-thermal plasma), the electrons have much higher kinetic energy than the neutral gas molecules. The advantage of using non-thermal plasma for promoting chemical reactions resides in this energy selectivity, as most of the electrical energy dissipated in the discharge is used to produce high-energy electrons, rather than spent on heating the entire gas stream. The fast electrons mainly initiate the chemical reactions in the plasma. They collide with the background gas molecules and excite, dissociate and ionize them, generating chemically active species, which in turn initiate other chemical reactions. The removal of a wide variety of pollutants by means of non-thermal plasma has been reported: aliphatic hydrocarbons [1–3], aromatics [4–7], chlorinated hydrocarbons [4,8–10], as well as inorganic contaminants such as SO2 , H2 S [11,12] and NOx , which will be discussed in detail in this chapter. Non-equilibrium plasmas can be generated at atmospheric pressure using electron beams, corona discharges, DBD (also known as silent discharges), and microwave discharges. Although electron beam-induced plasmas can decompose nitrogen oxides effectively, high capital costs of electron beam accelerators and auxiliary systems for X-ray shielding cannot be neglected, and thus motivated the study of NOx removal by electrical discharges. In the following, several general characteristics of corona, DBD and microwave discharges will be mentioned.

1.1. Corona discharges Corona discharges are transient discharges generated by strongly inhomogeneous electric fields associated with thin wires, needles or sharp edges of an electrode. Various geometries are used for generating corona discharges, the most common being pin-to-plate, wire-cylinder, and wire-plane, illustrated in Figure 12.1. The discharge can be supplied

Plasma-Assisted NOx Abatement Processes (a)

(b)

363 (c) HV

HV HV

Figure 12.1. Electrode geometries for corona discharges: (A) pin-to-plate; (B) wire-cylinder; (C) wire-plane.

Figure 12.2. CCD photographs of a pulsed corona discharge in air in point-to-plate configuration [14].

by a constant voltage (d.c. corona), an alternating voltage (a.c. corona) or a pulsed voltage. The corona discharge appears as a faint filamentary discharge radiating outward from the corona active electrode. The filaments are called streamers, have diameters of tens to hundreds of micrometers and propagate through the discharge gap with velocities of the order of 105 m/s [13,14]. Photographs of a pulsed corona discharge in air, in point-to-plate configuration [14] are shown in Figure 12.2, illustrating the branching of the streamers, which takes place due to electrostatic repulsion. A large number of papers and books dealing with corona discharges are available. Here, we mention only a few articles which summarize the main characteristics of these discharges and various configurations of the plasma reactors [13–16]. Average electron energies in corona discharges range between 2 and 10 eV. The degree of ionization is small and the electron density is 1013–1014 cm−3 . One of the disadvantages of the corona discharge is the transition to spark, which can occur under certain conditions. A streamer head can propagate until the cathode when the applied voltage is high enough. Then a conducting path is created between the electrodes and a current will flow if the voltage is maintained. When this current is low, the conductivity will be reduced due to recombination and the discharge will extinguish. However, if the current is high enough, it will heat the gas, decrease its density and increase its conductivity. Then, the current will grow to a value much higher

364

Past and Present in DeNOx Catalysis

than the corona current, the voltage will decrease significantly, and a spark will appear. The spark discharge is a dense, thermal plasma and occupies a smaller volume than the streamer corona, therefore, it is not considered for plasma chemical applications. In order to prevent the corona-to-spark transition, one possibility is to operate the discharge in pulsed mode, with short pulses (tens–hundreds of nanoseconds), so that the voltage drops before the spark can occur. Another possibility, which will be discussed in more detail in the following section, is to cover one of the electrodes with a dielectric layer.

1.2. Dielectric barrier discharges The main characteristic of the DBD is that a dielectric layer covers one of the electrodes, or sometimes both electrodes. When the voltage is applied between the electrodes, many microdischarges, or plasma filaments are initiated, which are randomly distributed in space and time. The dielectric limits the charge transported by a single microdischarge and distributes the microdischarges evenly over the entire electrode area. Once the discharge is ignited, the transported charge accumulates on the dielectric and the field due to this charge reduces the electric field in the gap and interrupts the discharge after several nanoseconds. Therefore, spark breakdown can be avoided. DBDs can be operated in a.c. mode or in pulsed mode. The most common discharge configurations are the planar electrode configuration and the coaxial configuration, illustrated in Figure 12.3; however, geometries combining needles or wires and planar or cylindrical electrodes covered by dielectric are also sometimes used. Dielectric packed-bed reactors are used as well for investigations of plasma-chemical reactions in plasma. In this configuration, the discharge gap is filled with dielectric pellets: glass beads, BaTiO3 , and Al2 O3 being some of the most usual. When the high voltage is applied, the pellets are polarized and an intense electric field is formed at the contact points between pellets. Because the pellets are packed close together, the electric field in the microdischarges can be significantly enhanced, thus, potentially increasing the production of chemically active species. The main characteristics of DBD, as well as a detailed theoretical description of their operation, are summarized in Eliasson and Kogelschatz [15,17]. The microdischarges have diameters of some hundreds of micrometers and are usually short-lived, with durations of the order of 100 ns or less. Average electron energies in DBD range between 1 and 10 eV and electron densities are of the order of 1014 cm−3 . (b)

(a) HV

HV 1 1 2

2

Figure 12.3. Electrode geometries for dielectric barrier discharges: (A) planar electrode configuration; (B) coaxial configuration (1 – electrodes; 2 – dielectric).

Plasma-Assisted NOx Abatement Processes

365

1.3. Microwave discharges The frequency range for microwave discharges is 0.3 –10 GHz, and the most common frequency used for applications is 2.45 GHz. One of the advantages of the microwave discharges over other types of electrical discharges is the operation without electrodes or vessel interaction as potential source of impurities. Based on the way the electromagnetic field is imposed on the plasma, the microwave discharges can be divided into two categories [18]. The first one consists of discharges sustained within microwave circuits, in this case, the active zone of the discharge being localized inside the field applicator: waveguides or resonant cavities. The second category of microwave discharges is formed by discharges with a large spatial extent as compared to the wavelength, which are sustained by the electric field of a wave propagating along them and are termed travelingwave discharges. The absorption of energy from the wave by the charged particles in the plasma causes the wave attenuation and the absorbed energy is redistributed to the plasma particles. At these very high frequencies only electrons can follow the electric field oscillations, thus, the plasma is usually far from thermal equilibrium. In the absence of collisions with other particles, the electrons oscillate out of phase with the field, therefore, the energy they acquire over one field period is zero. Microwave energy is transferred to the gas by electrons colliding with the gas neutrals. In this case, the energy of the electrons increases and decreases in small steps, the increase depends on the electric field applied and in general amounting to more than the decrease. A detailed review of theoretical and experimental studies of gas breakdown in high frequency fields is given in MacDonald [19]. It is shown that the Townsend criterion can be applied to describe the microwave breakdown. Because the excess production rate of electrons needs to be only infinitesimally greater than the loss rate; the Townsend criterion is customarily defined by saying that at breakdown, the ionization rate must equal the loss rate.

1.4. The initiation of chemical reactions in a non-equilibrium plasma In non-equilibrium discharges, the high-energy electrons produced in the discharge mainly initiate the chemical reactions in the plasma. They collide with the gas molecules and excite them to higher energy levels, or dissociate them. The rate of these inelastic collisions depends on the electron energy distribution function (EEDF). Under atmospheric pressure conditions, the EEDF depends only on the local value of E/N (E – electric field and N – number density of molecules), and therefore, the rate constants of electronmolecule reactions depend on E/N [16]. The main electron-molecule reactions, which take place in a streamer corona in humid air, together with the parameters of their rate constants, taken from Kulikovsky [16] are shown in Table 12.1, where the rate constant is expressed as:   B K = A exp E/N The largest parts of the species produced by the plasma in air are oxygen atoms and excited nitrogen molecules in the A3  state. The generation of nitrogen atoms requires

366

Past and Present in DeNOx Catalysis

Table 12.1. Electron-molecule reactions and parameters of their rate constants A (cm3 s−1 

Reaction e + O2 → e + O2 a1  e + O2 → e + O + O e + O2 → e + O1 D + O e + N2 → e + N2 A3  e + N2 → e + N2 C3  e + N2 → e + N + N e + H2 O → e + OH + H

10 × 10−9 , E/N ≤ 40 63 × 10−11 , E/N > 40 13 × 10−8 10 × 10−8 10 × 10−8 63 × 10−9 63 × 10−9 20 × 10−11

B (Td) (1 Td = 10−17 V cm2 ) 120, E/N ≤ 40 8.1, E/N > 40 309 338 336 486 949 322

more energy, so their density is about one order of magnitude lower. The chemically active species formed in the discharge initiate other chemical reactions. They are responsible for the decomposition of nitrogen oxides in the plasma, as NOx dissociation by direct electron impact is not important, due to the low concentration of NOx .

2. NOX REMOVAL BY ELECTRICAL DISCHARGES 2.1. NOx removal in DBD Numerous studies of NOx removal in DBD have been carried out [20–36]. The effect of electrode shape, discharge gap length, discharge polarity, gas composition and flow rate, and operation temperature on NOx conversion will be considered. The influence of structured electrodes or multipoint electrodes, which enhance the local electric field [20,21], as well as the effect of discharge polarity [20–22] and gap length [21,23,24], was investigated. A discharge configuration having either a multipoint or a trench electrode and a planar electrode covered by dielectric was compared with the classical DBD geometry with planar electrodes [20,21]. The multipoint electrode had 528–5000 pyramids of 45 tip angle and 1–5 mm height. The trench electrode has knife-edge rails with 5-mm height and 45 tip angle. The electrodes used in this study had the same area. The multipoint electrode configuration had the lowest capacitance, followed by the trench and the planar configurations, in this order. The small capacitance of the reactor has the advantage of lower dielectric loss in the case of multipoint electrode. From spectroscopy measurements, it was found that in the multipoint electrode configuration, electrons with energies higher than 10 eV can be generated [22,25]. NOx removal was investigated in a mixture of N2 , O2 and NO. NO removal reached 100% when using the multipoint and the trench electrode geometries, for input energy densities above 150 J/l [20]. The input energy is defined as the ratio of the power and flow rate. For planar electrodes, NO removal was considerably lower, a maximum of only 40% being achieved for energy densities below 250 J/l. In this case, NOx removal remained below 10% over a wide range of input energy, showing that NO was removed mainly by oxidation to NO2 . However, when using the multipoint electrode configuration, NOx removal increased up to almost 80% with increasing the energy densities up to 300 J/l, and above this value

Plasma-Assisted NOx Abatement Processes

367

showed a marked decrease. It was also found that the distance between adjacent points of the multipoint electrode is an important factor. Thus, when increasing the number of points, NO removal became smaller, reaching similar values as for the planar electrodes geometry when using 5000 points. The authors obtained energy efficiencies for NO removal in the range 6–10 g/kWh when using the multipoint or the trench electrode configurations, and only 1–2 g/kWh for the planar electrodes. Spectroscopic investigations of the discharge revealed that in case of multipoint electrode geometry, the average electron energy depends on polarity, being higher when the multipoint electrode was positive as compared to the negative polarity [20–22]. For planar electrodes, the electron energy was not influenced by polarity. However, no remarkable difference was observed in NO removal between positive and negative polarities in case of the multipoint configuration. The electron energy was also increased by decreasing the length of the discharge gap [23]. By reducing the gap length from 0.5 to 0.05 mm, the electric field increased significantly and spectroscopy measurements showed the increase of electron temperature. However, it was found that NOx removal was lower when using shorter discharge gaps in mixtures of N2 , O2 and NO. Thus, for the 0.5-mm gap length, NO removal increased to 100% at 100 J/l and started to decrease when the input energy density was higher than 300 J/l, indicating NOx formation in the plasma at high energy. About half of the NO was oxidized to NO2 in the discharge. When the short discharge gap (0.05 mm) was used, NO removal was only 50% and all of it was oxidized to NO2 , so that NOx removal was zero. However, in the absence of O2 , no effect of the gap length on NO and NOx removal was observed. It was thus concluded that NO was dominantly removed by oxidation with ozone and they explained the decrease in NO removal when shortening the gap length by the fact that ozone generation in the plasma is less efficient at very high electric fields [23]. The composition of the carrier gas containing the nitrogen oxides is another very important factor in NOx removal. The influence of oxygen content in N2 /O2 /NO mixtures on the conversion of NOx was investigated [26–28]. The addition of water vapor [26,27,29–32], carbon dioxide [26,32] and hydrocarbons [27–35] was studied as well. In the presence of oxygen, especially when the oxygen concentration exceeds 5%, the oxidation of NO in the plasma becomes dominant [28], due to two reasons: 1. The dissociation energy of O2 is smaller than that of N2 . As the average electron kinetic energy is relatively low in atmospheric pressure discharges, the rate for electron-impact dissociation of O2 is much higher compared to that of N2 . The dissociation of O2 produces only oxidative radicals: O, O3 , and OH, which lead to NO oxidation either to NO2 , or to HNO2 and HNO3 . 2. High electron energies are required to optimize the production of N atoms by electron-impact dissociation of N2 . Under conditions optimum for N2 dissociation, a large number of excited nitrogen atoms is produced, which can lead to undesired reactions in the presence of O2 , namely rather than reducing NO, they would react with O2 to produce NO. Chang et al. [26] observed a decrease in NO removal as they increased the oxygen content from 0% to 20% in mixtures of N2 /O2 /NO in a coaxial DBD operated in a.c.

368

Past and Present in DeNOx Catalysis

mode at 60 Hz frequency. The authors explained this result by NO formation in the plasma, which becomes more important as the inlet oxygen concentration increases, so that at 20% O2 the overall removal of NO is nearly zero. NO can be generated in the plasma by the following reactions: N + O2 → NO + O N + O3 → NO + O2 O + NO2 → NO + O2 However, in humid gas (2.6 vol.% H2 O), a significant increase in NO removal was observed when increasing O2 inlet concentration up to 5 vol.% [26,31,36], followed by a slow decrease, as the O2 content was further increased. As the O2 concentration was increased, the rate of generation of OH and HO2 radicals by reactions of atomic oxygen with H2 O molecules increases. NO can be oxidized by reactions with these radicals, forming NO2 , HNO2 , and HNO3 : NO + OH → HNO2 NO + HO2 → HNO3 NO + HO2 → NO2 + OH At high oxygen content, NO formation in the plasma led to the decrease of NO removal, so that the optimum O2 concentration found was 5 vol.%. With increasing the concentration of water vapor, NO and NO2 removal efficiencies increased [26,27,29,30] as a result of generation of more OH and HO2 radicals by electron impact dissociation of H2 O molecules. A positive effect of oxygen on NO removal was also found in the presence of ethylene [27]. NO removal efficiency increased significantly as the oxygen content was increased from 0 to 2.5 vol.%, for applied voltages higher than 16 kV, and showed a slight increase for higher oxygen concentrations. On the other hand, the presence of oxygen inhibited considerably NOx removal efficiency. The authors concluded that the different trends for NO and NOx removal mainly result from the oxidation of NO to NO2 [27]. Higher O2 content facilitates NO oxidation but reduces NO2 removal: NO + O → NO2 NO + O3 → NO2 + O2 It was suggested that, while at low applied voltage, NO formation might still be important, explaining the lower NO removal in the presence of oxygen as compared to the oxygen-free case; at high voltage, NO removal efficiency was higher in the presence of oxygen due to the abundant generation of O radicals as compared to N radicals. However, in this case, the effect of ethylene could be important as well. Ethylene decomposition increased with increasing voltage and oxygen content. The radicals formed from C2 H4 decomposition (alkyl, alkoxy and acyl radicals) also contribute to NO removal.

Plasma-Assisted NOx Abatement Processes

369

The influence of hydrocarbons on the removal efficiency of NOx has been investigated experimentally, especially the effect of ethylene [27,31,33,34,36,37] and propylene addition [28–30,35]. It was found that NO removal increased considerably when ethylene was added to the inlet gas. In N2 /NO mixtures, Lee et al. [27] observed an increase in NO removal efficiency with increasing C2 H4 concentration, and this effect was stronger at higher applied voltage. For example, at 20 kV NO conversion was almost 80% when 2500 ppm C2 H4 was injected in the carrier gas, as compared to only 30% in the absence of C2 H4 . In the absence of oxygen, ethylene is dissociated in the plasma mainly by direct electron impact, and the CH, CH2 , and CH3 radicals formed may react with NO molecules as follows: CH + NO → HCN + O CH + NO → HCO + N CH + NO → NCO + H CH2 + NO → HCN + OH CH2 + NO → HNCO + H CH3 + NO → CH3 NO Some of the intermediate products formed may react further with NO. The authors [27] found that NOx removal was almost the same with NO removal, in the absence of oxygen, which indicated that only chemical reduction occurred in N2 /NO/C2 H4 mixtures. In the presence of oxygen, C2 H4 decomposition in the plasma increased due to reactions with atomic oxygen. Thus, many oxygenated species are formed, which in turn may react with NO. Therefore, NO removal was enhanced in the presence of oxygen, reaching 100% for applied voltages higher than 18 kV, as compared to 70–80% in the N2 /NO/C2 H4 mixtures [27]. A similar behavior, namely the increase of NO removal when adding ethylene to the gas mixture containing N2 , O2 , NO, and H2 O, was also observed in Ravi et al., Niessen et al., and Mok et al. [31,33,36]. Niessen et al. [33] obtained an energy cost for NO removal of 48 eV/NO molecule, for a conversion of 90% in a mixture containing N2 , O2 , H2 O, NO and NO2 . The addition of 2000 ppm C2 H4 to the gas mixture reduced the energy cost to 6 eV/NO molecule. On the other hand, NOx was not removed in the presence of C2 H4 for mixtures containing oxygen, at low input energy [31,33,36]. Only for input energy above 50 J/l NOx removal started to increase, reaching 30–40% at 150 J/l [31]. The major end-products detected were carbon oxides (CO2 and CO) resulted from the total oxidation of C2 H4 , formic acid (CH2 O2 , formaldehyde (CH2 O), NO2 , N2 O, HNO3 and HNO2 [27,36]. The addition of propylene also led to the increase of NO removal efficiency in a pulsed DBD in a mixture containing N2 , O2 , NO and 500 ppm C3 H6 [30,35]. Consequently, the energy cost for NO oxidation decreased from 42 to 25 eV/NO molecule [30]. The authors also observed an increase in NOx removal up to 30%. The major reaction products detected were: carbon oxides, formaldehyde, acetaldehyde, propylene oxide, formic acid, ethyl acetate, methyl nitrate and nitromethane.

370

Past and Present in DeNOx Catalysis

With increasing gas temperature, it was found that NO removal increased significantly [26,27,30]. In the presence of hydrocarbons, the authors explained this improvement by the increased decomposition of C2 H4 or C3 H6 at higher temperature, resulting in formation of chemically active species. NOx conversion was also enhanced, however, to a smaller extent as compared to NO removal. Therefore, the energy consumption for NO and NOx removal decreased with increasing gas temperature. For example, at 533 K, Khacef et al. [30] obtained 43% NOx conversion at an input energy density of 27 J/l, corresponding to an energy cost of 30 eV/NOx molecule.

2.2. NOx removal in corona discharges Corona discharges have been investigated extensively for NOx removal [38–54]. The effect of electrodes configuration, electrical circuit, gas composition and flow rate were studied. When the discharge was operated in pulsed mode, the influence of pulse rise time, duration, and repetition frequency, as well as the effect of the voltage polarity on NOx conversion, were considered by numerous authors. In order to improve NOx removal efficiency, the optimization of the discharge configuration and electrical circuit was attempted [38–42]. Georgescu et al. [42] compared pulsed corona discharges in point-plate, wire-plate and wire-cylinder geometries, and found that the best results with respect to NOx removal were obtained in the wire-cylinder electrode configuration. Masuda [38] studied the influence of voltage polarity and pulse rise time on NOx removal in pulsed corona plasma. He observed better NOx removal efficiency when voltage pulses with shorter rise times were applied to the corona reactor. The effect of pulse width, in the range of 40–120 ns, and pulse repetition frequency was investigated [41]. NO removal efficiency decreased with increasing NO conversion and was higher when using shorter pulses. Wang et al. [39] used a multi-wire-to-plate corona geometry operated also in pulsed mode to remove NOx from diesel exhaust. They investigated the effect of the number of wires and of the distance between adjacent wires on the discharge characteristics and on the NOx removal efficiency. It was found that NOx removal improved when increasing the number of wires and the distance between neighboring wires. For example, NOx conversion was 38% when this distance was 40 mm and increased to 51% for 80 mm and to 56% for 120 mm between wires. It was suggested that lower NOx removal for short wire-to-wire distance could be due to perturbations of the electric fields for adjacent wires. This is in agreement with simulations of streamer corona in wireplate configuration [43,44], which predicted interference effects on the electric field distribution near neighboring wires, consisting in the reduction of the electron density and electric field in the streamer head, as the spacing between wires decreased. Hu et al. [45] investigated the removal of nitrogen oxides in a pulsed corona discharge in coaxial configuration in mixtures containing N2 and NO or/and NO2 or/and N2 O. In N2 /NO mixtures, NO concentration decreased with increasing input energy, while NO2 concentration increased, reached a maximum value at intermediate energy and then started to decrease as the input energy was further increased. N2 O was also formed in the discharge in small amounts, up to 10 ppm. A similar behavior was observed by Masuda [38] in NO/N2 /O2 mixtures, by Zhao et al. [40,46] in NO/N2 mixtures and by Rajanikanth and Ravi in diesel exhaust [47]. In N2 /NO/NO2 /N2 O mixtures, the

Plasma-Assisted NOx Abatement Processes

371

concentrations of all nitrogen oxides showed a slow continuous decrease with increasing input energy [45]. The influence of the flow rate (from several slm up to tens of slm, slm – standard liters per minute) and initial concentration of nitrogen oxides (in the range of 200–600 ppm) on their reduction was investigated [40,45,48]. It was shown that NOx conversion depends only on the energy density, regardless of the gas flow rate. However, lower flow rate was found more advantageous from the point of view of energy consumption: at constant NOx conversion, the efficiency for NOx removal was considerably higher at low flow [48]. When oxygen was present in the gas mixture, oxidation reactions became dominant [46,49,50]. The effect of oxygen content on NO removal in pulsed corona plasma was studied by Zhao et al. [46]. The authors observed that, after a sharp increase with increasing input energy, NO conversion was saturated at 65–75%, due to NO formation in the plasma, which occurred simultaneously with NO removal. The concentration of NO2 generated by NO oxidation increased significantly with increasing oxygen content. Most of the NO removed was converted to NO2 , so that NOx conversion reached a maximum of about 20% only when the O2 concentration was around 2%, while for higher concentrations, negative values for the NOx conversion were obtained, showing that NOx formation in the plasma was dominant. The influence of CO2 on NOx removal by pulsed corona discharges in mixtures containing NO and N2 was investigated as well [51,52]. It was found that small CO2 concentrations, of some hundreds of ppm, did not affect NOx conversion. For CO2 concentrations below 1%, a slight decrease in NOx removal was observed. However, at high CO2 concentrations (1–8%), NOx conversion decreased considerably with increasing the content of CO2 . The authors explained this behavior by the electronegative character of CO2 , and suggested that lower electron concentrations due to electron attachment led to the decrease of the rate constants of electron collisions with N2 with increasing CO2 concentration [51,52]. In the presence of ammonia, NO removal was enhanced considerably, due to reactions with NH2 and NH radicals generated from NH3 decomposition in the plasma [53]. This will be discussed in more detail in Section 2.3. A comparative assessment of pulsed corona, dielectric barrier discharge and dielectric packed-bed discharge with respect to NOx removal was done by Penetrante et al. [54]. By modeling the streamer dynamics, the authors observed that the field inside the streamer channel is space-charge shielded to a value corresponding to the reduced electric field of E/p = 38 kV/cm/atm, which is close to the breakdown threshold for N2 at atmospheric pressure. The results obtained in pulsed corona and DBD were similar and in good agreement with the modeling below 95 J/l input energy density. In this case, NO removal was dominated by reduction with atomic nitrogen, thus, the energy cost for NO removal in N2 /NO mixtures was equal to the energy cost for the production of N radicals of about 240 eV. Above 100 J/l, the experiments showed significant NO production, presumably due to the increase in gas temperature. It was found that the radical production efficiency in the plasma could not be enhanced by varying the voltage pulse parameters, using a dielectric barrier, adding a dielectric packing, or changing the electrode structure. The results obtained in the dielectric packed-bed reactor were consistent with those obtained in corona and DBD reactors for input energy densities below 50 J/l. At higher energy, NO removal was quenched, most likely due to gas heating in the microdischarge channels and the glass beads. The effect of the reactor design on the plasma treatment of NOx was

372

Past and Present in DeNOx Catalysis

studied [24,55]. In these works, various parameters of the plasma reactor were varied systematically in an attempt to understand how this affects the energy efficiency for plasma processing of NOx . These parameters include the packing material, electrode diameter, and voltage frequency. It was shown that the applied voltage is not the relevant parameter when comparing the performance of different plasma reactors. The important control parameter is the input energy density. The authors found that reactor design has little influence on the basic energy consumption of the plasma. Consequently, different reactor designs should yield basically the same plasma chemistry if the experiments are performed under identical gas composition and temperature conditions.

2.3. NOx removal in corona radical shower systems In corona radical shower systems, the high voltage electrode is a pipe with several nozzles and the ground electrodes are planar. A typical configuration for such a system is shown in Figure 12.4a. The discharge was usually operated in d.c. mode. The plasma appears in front of the nozzles and shows a flame-like pattern (Figure 12.4b [56]). It consists of streamers, which propagate from the tip of the nozzles to the planar electrodes (Figure 12.4c [57]). The velocity of the streamer head was estimated to about 2.5 × 105 m/s [57].

(a) HV

1

3

2

2 3

(b)

(c)

Figure 12.4. (A). Schematic drawing of the corona radical shower system: 1 – high voltage electrode (pipe with nozzles), 2 – nozzles, 3 – planar ground electrodes; (B) Photograph of the discharge in a corona radical shower system with 10 nozzles [56]; (C) CCD images of streamers in a corona radical shower system [57] (CCD exposure time: 400 ns).

Plasma-Assisted NOx Abatement Processes

373

Usually, various additives such as hydrocarbons, steam or most often ammonia are injected through the hollow electrode and exit through the nozzles, thus, being decomposed in the plasma. Therefore, radicals are generated, which in turn react with NO and NO2 and contribute to their remediation. In air, in the absence of additives, NO removal takes place by oxidation [56–58]. Wu et al. [56] observed that the concentration of ozone generated in the corona discharge in the presence of NO is significantly smaller as compared to the ozone formed in air without NO under the same conditions. Thus, they concluded that NO oxidation by ozone and atomic oxygen is an important reaction path for the NO conversion. Two-dimensional distributions of ground-state NO were detected by planar laserinduced fluorescence during the process of NOx removal in a corona radical shower system in NO/dry air mixtures [57,58]. The authors observed that the density of NO molecules decreased not only in the plasma region formed by the corona streamers and the downstream region of the reactor, but also in the upstream region of the reactor. They explained this behaviour by oxidation with ozone, which is transported upstream by electrohydrodynamic flow. Wu et al. [56] found that the geometry of the nozzle electrode is very important in the corona radical shower system for NOx treatment. They investigated the influence of the nozzle diameter and number of nozzles on the corona discharge characteristics, ozone and NOx formation in the plasma, as well as on NO removal. At nozzle diameters of 2–3 mm, the discharge had a typical streamer corona character, while decreasing the nozzle diameter below 1 mm leads to the transition to the glow corona regime, which was found to be ineffective for NOx removal. Increasing the number of nozzles enlarged the corona region and increased the discharge current, while it had almost no effect on the onset voltage and sparking voltage. Regarding NO removal, the authors found that NO conversion was enhanced with increasing nozzle radius and number of nozzles. The maximum NO conversion they obtained in humid air reached 82% for a nozzle electrode with 14 nozzles of 3 mm diameter. The energy yield improved as the nozzle diameter and the number of nozzles increased, the maximum obtained being around 20 g/kWh. The composition and the flow rate of the gas fed through the nozzles can influence the electrical characteristics of the corona discharge, as well as NOx removal. Lin et al. [59] reported that the presence of water vapour reduces significantly the discharge current and increases the breakdown voltage as compared to dry gas. In contrast, by flowing a mixture of Ar and NH3 through the nozzles resulted in higher discharge current due to the low ionization potential of argon. With regard to NO removal, the effect of injecting a mixture of CO2 and dry air through the nozzles was negligible [58]. In these experiments, the NO removal rate reached almost 90%, while the energy efficiency was about 10 g/kWh. Water vapour was found to improve NO removal, especially at low input power [59]. By injecting ammonia through the nozzles, NO removal was also increased. Several authors performed experiments aimed at the simultaneous removal of NOx and SO2 in corona radical shower systems with ammonia injection [60,61]. They found that the discharge current and the NOx removal rate increased with increasing NH3 –acid gas molecular ratio due to the presence of more radicals generated by ammonia decomposition in the plasma. The highest NOx removal efficiency was 98% at a molecular ratio of 2 [60].

374

Past and Present in DeNOx Catalysis

When ammonia is injected, NH2 , NH, N, and H radicals are formed in the region close to the nozzles by the following electron impact dissociation reactions [59,60]: NH3 + e → NH2 + H NH3 + e → NH + H + H NH3 + e → N + H + H2 These radicals react to NO, leading to its reduction: NO + NH2 → H2 O + N2 NO + NH → OH + N2 NO + N → N2 + O NO2 + N → N2 + O2 It was found that direct reduction of NO2 is much less significant as compared to NO reduction in the presence of NH3 [60]. However, NO2 and NO are oxidized to HNOx (x = 1 − 3): NO2 + H → HNO2 NO2 + OH + N2 → HNO3 + N2 NO2 + HO2 + N2 → HNO3 + N2 + O NO + H + H → HNO + H NO + OH + N2 → HNO2 + N2 NO + H → HNO + e The formation of aerosol particles was observed [60–63] when ammonia is injected, by reactions of HNOx with undissociated NH3 : HNO3 g + NH3 g → NH4 NO3 s Ammonium nitrate particles formed in the discharge are amorphous and irregularshaped, with dimensions of several microns [60]. Their density increased with increasing applied voltage. It was suggested that these aerosol particles might play an important role in NOx removal by surface reactions with NH3 and NOx . This is supported by experimental data showing that NOx removal efficiency for contaminated electrodes is much higher than for clean electrodes [60,62,63]. Chang et al. conducted pilot scale tests in a corona radical shower system with ammonia injection for simultaneous removal of NOx and SO2 from coal boiler flue gases [61]. The corona radical shower system used had dimensions of 21 × 18 × 2 m3 and consisted of 20 parallel flow channels with five multiple nozzle corona electrodes per

Plasma-Assisted NOx Abatement Processes

375

channel. The corona electrodes had a length of 1.6 m and the distance between adjacent nozzles was 3 cm. The flow rate of the flue gas through the system was 1000–1500 Nm3 /h and NO initial concentration was varied in the range 53–93 ppm. The corona discharge current increased with increasing applied voltage and ammonia-to-acid gas molecular ratio. With regard to NOx removal, it was found that the conversion was improved with increasing input electrical power and ammonia radical injection rate. The maximum NOx removed was 70–80%. The energy efficiency for NOx removal decreased with increasing the specific energy density, even if NOx conversion increased with power. No significant effect of the ammonia-to-acid gas ratio on the removal efficiency was observed. At 75% NOx removal rate, the energy efficiency was approximately 125 g/kWh.

2.4. NOx removal in photo-triggered discharges Photo-triggered discharges have been initially developed for pumping atmospheric pressure lasers [64,65]. In the photo-triggered pumping scheme, the discharge electrodes are connected to an energy storage unit charged up to a voltage V0 , which is chosen so that the initial reduced electric field is higher than the self-sustaining electric field of the discharge. Once the desired voltage has been reached, pre-ionization is produced by means of vacuum ultraviolet (VUV) photons or X-rays and initiates the gas breakdown with a variable time lag. Recently, the photo-triggered discharge was investigated for NO removal [66–68]. In this case, pre-ionization was produced by UV emission from an auxiliary surface discharge located under the cathode of the main discharge. The advantage of this type of electrical discharge is the generation of large volume homogeneous plasma, without filamentary structures, as in corona or dielectric barrier discharges. In these studies concerning gas cleaning, the discharge was operated below atmospheric pressure (up to 500 mbar), however, in principle it can be ignited at atmospheric pressure as well. In N2 /O2 /NO mixtures, it was found that NO removal occurs principally through oxidation by O atoms and O3 molecules. In these conditions, energetic costs lower than 50 eV/NO removal were achieved. Experimental results have been compared to predictions of a fully self-consistent discharge and kinetic modeling [68] in N2 /NO and N2 /H2 O/NO mixtures. Due to the spatial homogeneity of the photo-triggered discharge, a zero-dimensional model is well suited to describe the plasma. The kinetic model predicts that singlet metastable states of the nitrogen molecule (denoted by N2 (a )), which are strongly mixed by collisions and highly populated, are very important species for NO decomposition in the absence of O2 . In this case, NO dissociation by the reaction: N2 a  + NO → N2 + N + O represents an important pathway for NO removal in N2 /NO mixtures, besides NO reduction by N atoms. The drastic decrease in NO removal observed experimentally in the presence of water was attributed to de-excitation of the N2 (a ) states by the H2 O molecules.

376

Past and Present in DeNOx Catalysis

2.5. NOx removal in microwave discharges Microwave discharges have been investigated as well for removing nitrogen oxides from exhaust gases. It was found that continuous wave (CW) discharges are effective in the case of NO/N2 mixtures, when 65–85% overall reductions of NO were obtained, depending on NO input concentration [69]. However, the high gas temperature in these discharges (∼7000 K) favors NOx formation in the presence of oxygen. Ighigeanu et al. [70] reported maximum 6% NOx removal in a microwave discharge at 350–400 W power, in a mixture of air, argon, CO2 , H2 O, NH3 , and NOx . In order to avoid high temperatures, pulsed microwave discharges have been investigated [71,72]. Long microwave pulses (50 s) with low peak power (4 kW) at high pulse repetition rate (2 kHz) also led to gas temperatures above 1000 K, and implicitly to NOx formation in N2 /O2 /NO mixtures [71]. Better results were obtained using shorter microwave pulses (35 s) of high peak power (1.4 MW), at moderate repetition frequency (40 Hz) [72]. In this case, the gas temperature remained low around 500 K. NOx reduction rates up to 95% could be achieved in N2 /NO mixtures. In the presence of O2 , NOx removal decreased significantly, but the reduction could still be observed for O2 concentrations up to 9%. As expected, NOx removal rate increased with increasing energy deposition, however, values larger than 100 J/l led only to higher conversion from NO to NO2 . The authors found energy costs of 25–100 eV/NO molecule, for O2 concentrations lower than 4%. NOx processing by a pulsed microwave discharge in synthetic exhaust gases containing N2 , O2 and NO was described theoretically by Baeva et al. using a zero-dimensional time-dependent kinetic model [73]. The model results agree well with the corresponding experimental data. It was found that N radicals play the major role in the reduction of NO, even in O2 -containing mixtures, although the density of atomic nitrogen is one order of magnitude lower than the density of atomic oxygen. Even if oxidation of NO to NO2 cannot be completely eliminated, an overall NOx reduction can be achieved when using short microwave pulses, if the gas temperature remains below 500 K.

3. REACTION MECHANISMS RESPONSIBLE FOR NOX REMOVAL IN NON-THERMAL PLASMA The high-energy electrons generated in the plasma mainly initiate the chemical reactions by reactions with the background gas molecules (see Table 12.1). Direct electron impact reactions with NOx are usually not important for NOx decomposition, as in real flue gas, as well as in experiments in simulated gas, the concentrations of NOx are very low (some hundreds of ppm), and therefore, the probability of electron collisions is also low. Zhao et al. [74] developed a kinetic model to analyze the removal of nitrogen oxides in a pulsed corona discharge in NO/N2 mixtures. They considered reactions of NO and NO2 with N and O atoms and with excited N2 (A3  molecules. NO + N → N2 + O NO + O → NO2

Plasma-Assisted NOx Abatement Processes

377

NO + N2 A3  → N2 + NO NO2 + N → N2 O + O NO2 + N → N2 + O2 NO2 + N → 2NO NO2 + O → NO + O2 NO2 + N2 A3  → N2 + NO + O The results indicated that the evolution of NO and NO2 is mainly influenced by N atoms in the absence of oxygen. However, it was found that N2 (A3  molecules are important as well, especially for the evolution of N2 O, which is affected by the following reactions: N2 A3  + O2 → N2 O + O N2 A3  + N2 O → 2N2 + O The ionic species do not contribute significantly to the reduction of NO [54,75]. The results of kinetic modeling, in good agreement with experimental data, showed that with increasing the input energy density the concentration of NO decreases, while the concentration of NO2 initially increases, reaches a maximum value, and then starts to decrease towards higher input energy [45,46,54,74]. In the presence of oxygen, the situation is more complicated, as more active species can be formed, which can contribute to NOx decomposition or formation in the plasma. It was found [46] that the excited states of N2 do not contribute significantly to NOx evolution, but mainly influence the dissociation or quenching of O2 . As the concentration of O2 increases, their contribution to O2 dissociation also increases. Only nitrogen atoms in the ground state (N(2 D)) or excited state (N(4 S)) may contribute directly to NO formation by reactions with molecular oxygen: N2 D + O2 → NO + O N4 S + O2 → NO + O The results indicated that N(4 S) atoms are mainly consumed in reactions with NO and NO2 , leading to NOx reduction, while NO formation is mainly due to the reactions of N(2 D) with molecular oxygen. With increasing O2 concentration, the selectivity to NO formation increases as well. Excited electronic states of molecular oxygen are essentially unreactive, and only oxygen atoms in the ground state (O(3 P)) or the first electronic excited state (O(1 D)) are involved in reactions with NOx . The excited O atoms are mostly quenched by the background N2 and O2 molecules, therefore, only ground-state O atoms contribute to the evolution of NOx in the plasma. In the presence of oxygen, NO is mainly oxidized to NO2 , by reactions with O atoms and with ozone [33,46,76–78]. The rate constant of molecular dissociation of O2 by electron collisions is almost two orders of magnitude higher than the dissociation of N2 in

378

Past and Present in DeNOx Catalysis

plasmas with electron energies below 10 eV, due to the lower dissociation energy of O2 (4.8 eV), as compared to N2 (9.2 eV). It was found [46] that when the O2 concentration is very low (less than 2.5 vol.%), NOx conversion is positive, which means that NO reduction exceeds NO oxidation to NO2 . However, when O2 concentration is increased, NOx conversion becomes negative, showing that NOx formation in the plasma exceeds NOx decomposition. In humid air, oxidation of NOx by reactions with OH radicals produced by dissociation of water molecules may occur as well, leading to the formation of HNO2 and HNO3 [77,79]: NO + OH + M → HNO2 NO2 + OH + M → HNO3 Water dissociation may take place either by electron impact, as the dissociation energy is relatively low (4.8 eV), or, to a greater extent by reactions with excited oxygen atoms (O(1 D)) [33,77,79]: H2 O + O1 D → OH + OH The formation of so-called secondary radicals, for example, ozone and HO2 , starts as soon as the discharge pulse is finished. They contribute to NO oxidation to NO2 by the following reactions: NO + O3 → NO2 + O2 NO + HO2 → NO2 + OH Eichwald et al. developed a model to describe the spatio-temporal evolution of the main neutral species involved in a corona discharge used for NO removal, including radial mass diffusion, temperature variation, and chemical kinetics [80]. During streamer propagation (up to 20 ns), ions, radicals, and excited species are formed in the discharge channel. However, modeling results indicate that NO concentration remains unchanged during the discharge phase. For times up to 1 s after the discharge pulse, NO reduction by N atoms plays the most important role in the evolution of NO on the axis of the discharge channel. At the border of this channel, oxidation of NO to NO2 takes place after 10 s, while after 100 s, NO oxidation by O3 becomes dominant. The gas temperature increases with about 30 K on the axis of the discharge channel for times up to 50 ms after the discharge pulse, and then starts to decrease due to thermal diffusion. The gas temperature rise modifies the chemical kinetics, particularly the efficiency of three-body reactions, which are very sensitive to both gas temperature and density variations. When the gas dynamics is considered, NO removal in the discharge channel is limited to about 60% after 1 ms instead of 99% when only the chemical kinetics governs the gas reactivity. This is due to the rapid decrease of the density of primary radicals after 10 s, which limits the removal of the new NO molecules diffusing from the high-density region (the gas volume) towards the low-density region (the discharge channel) after 10 s. Another computational model for the removal of nitrogen oxides in a pulsed dielectric barrier discharge was developed by Gentile and Kushner [75] for gas mixtures containing N2 /O2 /H2 O (85:5:10) and 500 ppm NO. The results show that NO concentration decreases relatively fast in time, whereas the densities of the reaction products (HNO2 ,

Plasma-Assisted NOx Abatement Processes

379

HNO3 , N2 O, and N2 O5  increase slowly with each pulse. An exception is the density of NO2 , which initially increases and then starts to decrease, as NO2 is removed as well by plasma-chemical reactions. The calculations indicate that 100% NO removal can be achieved and the energy cost is 72 eV/molecule. It is suggested that NO removal occurs mainly by reduction, and only approximately 10% is oxidized to HNO3 , while conversion to other nitrogen oxides is very low. The discharge was characterized by three time scales: the current pulse period, with duration of tens of nanoseconds, the post-pulse period, ranging from tens to hundreds of nanoseconds and the interpulse period, ranging from hundreds of nanoseconds to the beginning of the next discharge pulse. During the current pulse, radicals are produced by electron impact. These radicals react with the NO molecules during the post-pulse period and lead to the decrease of NO concentration. Secondary radicals, like O3 and HO2 , are formed on time scales of tens to hundreds of microseconds and contribute to NOx removal by oxidation. The removal of Nx Oy reaction products continues until N and OH radicals are depleted (i.e. during the post-pulse period), whereas NO removal occurs in the interpulse period as well, due to oxidation to NO2 by ozone. The number of discharge pulses is also important, as reaction products formed in early pulses become reactants in later pulses [75,76]. Especially, the behavior of NO2 varies significantly, as for later pulses it becomes a major reactant for N atoms, leading to NO regeneration from NO2 . With respect to energy consumption for NOx removal, the models developed by Gentile and Kushner [75] and Dorai and Kushner [76] predict that energy efficiency generally improves when using more current pulses of lower energy, larger applied voltage, and larger H2 O percentage in the gas stream.

3.1. Reaction mechanism for NO removal in the presence of hydrocarbons When hydrocarbons are present in the gas mixture, NO removal by oxidation to NO2 occurs at much lower input energy and the reaction paths change significantly as compared to the case without hydrocarbons. Numerous works analyze the reaction mechanism of NOx conversion in non-thermal plasma with addition of hydrocarbons, especially ethylene [33,37,77,79,81–83], propylene [35,76,81,83–87], and propane [76,81,85,87].

3.1.1. Ethylene The addition of ethylene leads to an important increase in NO removal and a decrease in the energy costs for conversion. When C2 H4 is present in the gas mixture, it reacts predominantly with OH radicals, forming C2 H4 OH, which rapidly reacts with O2 to form the peroxy radical O2 C2 H4 OH [33,77,79,81]: C2 H4 + OH → C2 H4 OH C2 H4 OH + O2 → O2 C2 H4 OH The O2 C2 H4 OH radical leads to NO oxidation to NO2 by the reaction: O2 C2 H4 OH + NO → NO2 + OC2 H4 OH

380

Past and Present in DeNOx Catalysis

The alkoxy radical OC2 H4 OH formed in the process either dissociates into CH2 O and CH2 OH, or reacts quickly with O2 to form glycolaldehyde (HOCH2 CHO) and HO2 : HO2 is also formed by the reactions of CHO and CH2 OH with O2 , and contributes as well to NO oxidation: CHO + O2 → CO + HO2 CH2 OH + O2 → CH2 O + HO2 HO2 + NO → NO2 + OH Reduction of NO is promoted by N atoms and CH2 radicals formed by ethylene decomposition, but it is much less important as compared to oxidation. The addition of ethylene decreased notably the energy consumption for NO removal. Niessen [33] obtained energy costs of 61 eV/NO removed in the absence of ethylene and only 9.6 eV/NO removed when ethylene is present in the gas mixture. However, the NOx concentration does not change much, as NO is mostly converted to NO2 . The main reaction products obtained in the presence of ethylene are NO2 , glycolaldehyde (OC2 H3 OH), formaldehyde (CH2 O), and oxirane.

3.1.2. Propylene The addition of propylene (C3 H6  to the gas mixture containing NO was also extensively investigated [35,76,81,83,85–89], as it increases significantly the efficiency of NO oxidation to NO2 . Dorai and Kushner [85,87,88] performed a computational study to investigate the effects of C3 H6 on the conversion pathways for NOx in non-thermal plasma. The reaction mechanism is initiated by electron impact on O2 and H2 O, creating O and OH radicals. Atomic oxygen reacts with propylene by the following reactions: O + C3 H6 → C2 H5 CHO O + C3 H6 → C2 H5 + HCO O + C3 H6 → CH3 −CHO−CH2 O + C3 H6 → CH2 CHO + CH3 The removal of NOx , as compared to the case without propylene, results mainly from the reaction with NO forming 2-nitroso ethanal (ONCH2 CHO), especially at higher input energy: CH2 CHO + NO → ONCH2 CHO Hydroxyl radicals react with C3 H6 by the following addition reactions to form hydroxylalkyl radicals: OH + C3 H6 → H3 C−CH−CH2 OH OH + C3 H6 → H3 C−CHOH−CH2

Plasma-Assisted NOx Abatement Processes

381

These radicals react rapidly with O2 to form the peroxy radicals: H3 C−CH−CH2 OH + O2 → H3 C−CHOO−CH2 OH H3 C−CHOH−CH2 + O2 → H3 C−CHOH−CH2 OO The peroxy radicals react with NO and convert it into NO2 : H3 C−CHOO−CH2 OH + NO → H3 C−CHO−CH2 OH + NO2 H3 C−CHOH−CH2 OO + NO → H3 C−CHOH−CH2 O + NO2 The resulting alkoxy radicals either decompose, or react with O2 to form the final reaction products: H3 C−CHO−CH2 OH → H3 C−CHO + CH2 OH H3 C−CHOH−CH2 O → CH3 CHOH + HCHO H3 C−CHO−CH2 OH + O2 → H3 C−COCH2 OH + HO2 H3 C−CHOH−CH2 O + O2 → H3 C−CHOHCHO + HO2 The hydroxy radicals formed by the decomposition of the hydroxylalkoxy radicals react rapidly with O2 to form aldehydes and HO2 , which also contribute to NO oxidation to NO2 : CH2 OH + O2 → HCHO + HO2 CH3 CHOH + O2 → CH3 CHO + HO2 Typical reaction products of NO removal in non-thermal plasma in the presence of propylene, predicted by the model in Martin et al. and Dorai and Kushner [86,88], include formaldehyde (HCHO), acetaldehyde (CH3 CHO) methyl oxirane (C3 H6 O), glycoxal (CHO-CHO), methyl nitrite (CH3 ONO), methyl nitrate (CH3 ONO2 , and 2-nitroso ethanal (ONCH2 CHO). Higher concentration of C3 H6 and especially higher temperature increase NO removal, but have little effect on NOx removal. As most atmospheric-pressure plasmas have a filamentary structure, the energy is dominantly deposited in the confined volume of the microdischarges. Due to the localized temperature rise, as well as non-uniform distribution of the species densities, transport of the species to and from the plasma streamers occurs. Dorai and Kushner investigated the consequences of hydrocarbons addition, in particular propylene, on radial transport dynamics and remediation of NOx in a dielectric barrier discharge [85]. The model predicts that in the presence of hydrocarbons, O and OH radicals are dominantly consumed in the streamer regions in reactions with the hydrocarbons. Therefore, the transport of O and OH radicals to larger radii is reduced, as compared to the case without hydrocarbons and NO removal is also restricted to the streamer regions. This finding is particularly important in systems using high gas flow rates at short residence times, which are more sensitive to gas by-passing.

382

Past and Present in DeNOx Catalysis

Another reaction mechanism for NOx removal in the presence of propylene, described in Dorai and Kushner [87] suggests as first step the following reactions of C3 H6 with O atoms: O + C3 H6 → C2 H5 + HCO O + C3 H6 → CH2 O + C2 H4 O + C3 H6 → CH3 CO + CH3 The C2 H5 radicals produced in the first step react with O2 forming C2 H5 O2 radicals that in turn convert NO into NO2 and generate C2 H5 O radicals: C2 H5 + O2 → C2 H5 O2 C2 H5 O2 + NO → C2 H5 O + NO2 The C2 H5 O radicals then decompose to give formaldehyde and acetaldehyde. The CH3 CO radicals formed in the first step can react directly with O2 and NO2 to give formaldehyde: CH3 CO + NO2 → CH2 O + HCO + NO CH3 CO + O2 → CH2 O + CO + OH The authors suggest that this reaction is one of main sources of OH radicals, as they consider that the electron impact dissociation of water is too slow [87]. OH radicals are more reactive towards propylene than atomic oxygen, therefore, they replaced O atoms as the driver of the chemistry that oxidizes C3 H6 . The reaction of C3 H5 with molecular oxygen, forming peroxy radicals is also included in the model developed by Martin et al. [86], as compared to the work of Dorai and Kushner [84]. Formaldehyde and acetaldehyde formed in the reactions of O and OH with propylene are themselves destroyed by O and OH radicals in processes that mainly regenerate OH. The HCO radicals generated in these steps are the major source of CO and CO2 .

3.1.3. Propane In the presence of propane (C3 H8 , the reaction mechanism is initiated by hydrogen abstraction from C3 H8 by OH radicals, producing alkyl radicals, which then rapidly react with O2 to form peroxy radicals [88]. The peroxy radicals react with NO and oxidize it to NO2 : C3 H8 + OH → H3 C−CH−CH3 + H2 O C3 H8 + OH → H3 C−CH2 −CH2 + H2 O H3 C−CH−CH3 + O2 → H3 C−CHOO−CH3 H3 C−CH2 −CH2 + O2 → H3 C−CH2 −CH2 OO

Plasma-Assisted NOx Abatement Processes

383

H3 C−CHOO−CH3 + NO → H3 C−CHO−CH3 + NO2 H3 C−CH2 −CH2 OO + NO → H3 C−CH2 −CH2 O + NO2 As in the case of ethylene and propylene, these reactions lead to increased NO removal, however, NOx removal remains almost unaffected, because NO is largely converted to NO2 . In addition, it was found that propane is less reactive as compared to propylene [81,88], due to their stable molecular structure with stronger sigma bonds of C−C and C−H.

3.2. The influence of soot on NOx removal The influence of soot on NOx removal in non-thermal plasma has been investigated by Dorai and Kushner and Dorai et al. [88,89] in the presence of propylene and propane. Soot particles were assumed to contain only C and H atoms, and were denoted (Cx Hy . Soot may undergo oxidation when reacting with O and OH radicals generated in the plasma: Cx Hy  + O → Cx−1 Hy −CO Cx Hy  + OH → Cx−1 Hy −C−OH The interaction between NOx and soot takes place through adsorption and reduction processes involving NO2 . The reaction mechanism starts with NO2 adsorption on soot, forming C−NO2 and C−ONO complexes. Spontaneous desorption produces CO, accompanied by a reduction of the soot mass, and NO or H2 . Cx Hy  + NO2 → Cx−1 Hy −C−NO2 Cx Hy  + NO2 → Cx−1 Hy −C−ONO The hydroxyalkyl radicals CH3 CHCH2 OH and CH3 CH(OH)CH2 formed by reactions of OH with C3 H6 can also be adsorbed on the soot. However, the resulting desorption products are not further useful for NOx removal, and therefore, the reactivity of these radicals is decreased. The model predicts that in the presence of soot, the removal of NO decreases due to both heterogeneous generation of NO from NO2 adsorbed on the surface of the soot particles and decrease of reactivity of the hydroxyalkyl radicals. It was found that NOx removal, both in the presence and absence of soot, is about the same, however, the proportion of NO and NO2 depends on the soot. With lower densities (107 cm−3  of soot particles, the final NOx was primarily NO2 , whereas, with higher densities (109 cm−3  of soot particles, the NOx was mainly NO.

4. NOX REMOVAL BY PLASMA-ASSISTED CATALYSIS Catalysts that can effectively decompose NOx to N2 and O2 in oxygen-rich environments, known as lean-NOx catalysts, have been the subject of considerable research. Difficulty

384

Past and Present in DeNOx Catalysis

has been encountered in finding lean-NOx catalysts that have the activity, durability, and temperature window required to effectively remove NOx from the exhaust of lean-burn engines. Prior art lean-NOx catalysts are hydrothermally unstable. A noticeable loss of activity occurs after relatively little use, and even such catalysts only operate over very limited temperature ranges. One alternative is to use catalysts that selectively reduce NOx in the presence of a coreductant, e.g., SCR using ammonia as a coreductant. However, another viable alternative involves using coexisting hydrocarbons in the exhaust of mobile lean-burn gasoline engines as a coreductant and is considered a more practical, cost-effective, and environmentally sound approach. Many such SCR catalysts are known, some containing base metals or precious metals that provide high activity. However, like most heterogeneous catalytic processes, the SCR process is susceptible to chemical and/or thermal deactivation. Many lean-NOx catalysts are too susceptible to high temperatures, water vapor and sulfur poisoning (from SOx . Catalyst deactivation is accelerated by the presence of water vapor in the stream and water vapor suppresses the NO reduction activity even at lower temperatures. In addition, sulfate formation at active catalyst sites and on catalyst support materials cause deactivation. Lean-NOx catalysts promote the conversion of organo-sulfur compounds present in gasoline to SO2 and SO3 during combustion. SO2 is adsorbed onto precious metal sites at temperatures below 300 C and thereby inhibits the catalytic conversions of CO, hydrocarbons and NOx At higher temperatures with an Al2 O3 catalyst carrier, SO2 is converted to SO3 to form a large-volume, low-density material, Al2 (SO4 3 that alters the catalyst surface area and leads to deactivation. Generally, SCR of NOx works effectively at temperatures above 500–550 K, but the performance for NOx removal decreases considerably at lower temperatures [90]. At low temperature, it is known that the efficiency of the SCR of NOx strongly depends on NO2 concentration in the gas stream. About 30–50% of NO2 in the gas stream could greatly enhance the performance of SCR. However, in real conditions, the fraction of NO2 present in the total NOx does not exceed 5%. Hence, an oxidation pre-treatment is necessary to increase the concentration of NO2 . Plasma-assisted SCR is currently being studied by many researchers. It was reported that the presence of plasma greatly enhances the performance of SCR catalyst to achieve high NOx removal efficiencies even at low gas temperatures [36,82,91–96]. Demidiouk et al. [92] compared the effect of gas pre-treatment by plasma or by noble metal catalysis prior to SCR of NOx , in the temperature range 443–543 K. Ammonia was added as reducing agent to the SCR reactor. As oxidation catalyst, a honeycomb type Pt-Al2 O3 catalyst was used. The plasma was generated in a DBD reactor filled with glass beads, operated in a.c. mode at 60 Hz frequency, with 20-W input power. When using the Pt-Al2 O3 catalyst, the oxidation of NO to NO2 increased significantly with increasing temperature, especially at low NO concentration. An opposite effect was observed when using plasma treatment: the oxidation of NO to NO2 decreased slowly with increasing temperature. Therefore, the contribution of plasma oxidation to the SCR process was significant at the lower temperatures used. For example, at 450 K, the NOx removal increased from 25% to 30% with the SCR alone to 50–60% when using plasma oxidation pre-treatment. Oxidation by noble metals was more effective at higher temperature, enhancing NOx removal with almost 20% as compared to SCR alone at 543 K.

Plasma-Assisted NOx Abatement Processes

385

The combination of plasma (d.c. corona, dielectric packed bed or dielectric barrier discharge) and V2 O5 /TiO2 or V2 O5 -WO3 /TiO2 catalysts was investigated [36,37,53,97–101], using ammonia as reducing agent. In Mok et al. [36], the plasma was generated in a corona reactor with point-to-plate geometry. D.C. high voltage (13–25 kV) was applied to the needle electrode, and the time averaged current was in the range of 40–300 A. The catalyst (monolithic V2 O5 /TiO2 with 5 wt.% vanadium) was placed either directly inside the plasma (one-stage configuration), or downstream of the plasma region (two-stage configuration). The experiments were carried out at room temperature. The results obtained in the one-stage and two-stage configurations were similar, suggesting that the plasma did not activate the catalysts, but only changed the gas composition [36]. However, slightly better performance was observed in the onestage configuration was explained by the authors by the continuous removal of a part of the NO2 formed in the plasma by the catalyst, thus, preventing the reverse reaction of NO2 to NO. Mok et al. [36,37,97,99] used either a dielectric packed-bed reactor filled with glass beads, or a DBD reactor, followed by a monolithic V2 O5 /TiO2 catalyst and observed a significant increase in NOx removal in the presence of plasma as compared to the catalytic process in the temperature range of 348–473 K. The authors found that the catalyst alone removed 20–70% of the initial NOx when increasing the reaction temperature from 348 to 473 K; the NOx conversion in the plasma was very low in the range of 10–13%, whereas in the plasma-catalytic system, the NOx removal increased to 80–90%. The energy consumption under these conditions was 40–50 eV/NOx molecule removed. The reaction temperature did not influence much the NOx removal rate, which was almost 80% even at 350 K. However, at low temperature, ammonium nitrate was produced and deposited on the catalyst contributing to its deactivation. Therefore, in order to avoid the formation of NH4 NO3 , the temperature should be kept above 450 K. Hammer et al. [34] observed that most of the NOx removal below 450 K was due to adsorption on the catalyst. Mok et al. [99] used the electrical ignition system of an internal combustion engine as a high voltage pulse generator for the plasma reactor. They employed a dielectric packed-bed reactor filled with glass beads followed by a catalytic reactor with a monolithic V2 O5 /TiO2 catalyst. Ethylene was added to promote the oxidation of NO to NO2 and ammonia was added as reductant. High voltage pulses with peak voltage around 21 kV and duration (full width at half maximum) of about 280 s were generated. The authors found that the performance of this pulsed regime with respect to NOx removal was as good as with a.c. voltage. When a catalyst of V2 O5 and TiO2 deposited on Al2 O3 globule was saturated with ammonia and placed in the plasma region of a corona discharge in point-to-plate configuration, 96% NOx removal was obtained at room temperature [53]. Under these conditions, the energy efficiency was 3.4 g NO/kWh. Without the catalyst, NOx removal in the corona discharge in a gas mixture with ammonia was lower (up to 66%) and the energy efficiency was lower as well (about 1.8 g NO/kWh). However, the lifetime of the catalyst was only 30 h at this low temperature, after which the catalyst started to lose its activity due to NH4 NO3 deposits. SEM imaging showed that the catalyst surface exposed to the corona discharge was almost completely covered with NH4 NO3 solids, which formed relatively large crystals [53]. Experiments in real diesel exhaust gas have been carried out as well [101] using a pulsed DBD and V2 O5 /TiO2 catalysts placed downstream of the plasma reactor. NOx removal reached 90% under no-load conditions, for catalyst temperatures above 373 K

386

Past and Present in DeNOx Catalysis

and input energies above 70 J/l in the plasma. For 50% load conditions, NOx removal was lower due to the high initial concentration of NO. However, about 80% NOx removal was obtained for specific energy densities in the plasma above 90 J/l. The authors also observed that the presence of carbonaceous soot in the gas enhanced the NOx removal. Kang et al. and Wallis et al. [102,103] suggested that plasma electrons with mean energies of 3–4 eV impacting the TiO2 surface can produce electron–hole pairs in the same manner as photon absorption: TiO2 + e32 eV → h+ + e− These electrons can activate oxygen leading to very active electrophile O2 − species. In another attempt, Nakamura et al. [104] have shown that oxygen vacancies are created by plasma excitation of TiO2 and that electrons can be trapped on these vacancies, finally leading to the same electrophile species. Ogata et al. [105] suggested that plasma action on TiO2 directly activates lattice oxygen and OH in the TiO2 . Actually, all these superficial electrophile species, which are extremely active, can very easily transform NO into NO2 . Another important catalytic technology for removal of NOx from lean-burn engine exhausts involves NOx storage reduction catalysis, or the ‘lean-NOx trap’. In the leanNOx trap, the formation of NO2 by NO oxidation is followed by the formation of a nitrate when the NO2 is adsorbed onto the catalyst surface. Thus, the NO2 is stored on the catalyst surface in the nitrate form and subsequently decomposed to N2 . Lean NOx trap catalysts have shown serious deactivation in the presence of SOx because, under oxygenrich conditions, SOx adsorbs more strongly on NO2 adsorption sites than NO2 , and the adsorbed SOx does not desorb altogether even under fuel-rich conditions. The presence of SO3 leads to the formation of sulfuric acid and sulfates that increase the particulates in the exhaust and poison the active sites on the catalyst. Furthermore, catalytic oxidation of NO to NO2 can be operated in a limited temperature range. Oxidation of NO to NO2 by a conventional Pt-based catalyst has a maximum at about 250 C and loses its efficiency below about 100 C and above about 400 C. A two-stage method for NOx reduction in an oxygen-rich engine exhaust was proposed by Merritt et al. [106] and consists of a plasma oxidative stage and a storage reduction stage. The first stage employs a non-thermal plasma treatment of NOx and is intended to convert NO to NO2 in the presence of O2 and hydrocarbons, as propene. The plasma was generated in a corona discharge reactor operated in pulsed regime with high voltage pulses of 30 kV amplitude, 100 ns duration, and pulse repetition rates of 50–5000 Hz. The second stage employs a lean-NOx trap to convert the NO2 formed in the first stage to environmentally benign gases: N2 , CO2 , and H2 O. In the lean-NOx trap, the NO2 is adsorbed on a nitrate-forming material, such as an alkali material, and stored as a nitrate. An engine controller periodically runs a brief fuel-rich condition to provide hydrocarbons for a reaction that decomposes the stored nitrate into benign products such as N2 The lean-NOx trap uses an alkaline material, as for example, a base metal oxide such as CuO, BaO, SrO, K2 O, and MnOx The catalyst also contains catalytic components, which promote the reduction of the stored nitrate in the presence of hydrocarbons to produce N2 . The nitrate reduction component of the catalyst is usually a precious metal, such as Pt, Pd, and/or Rh. Generally, such catalytic components are supported on porous refractory

Plasma-Assisted NOx Abatement Processes

387

oxides such as Al2 O3 , (La2 O-stabilized alumina), zeolite, cordierite, and perovskite. By pre-converting NO to NO2 in the first stage by using the non-thermal plasma, the efficiency of the second stage for NOx reduction is enhanced. Thus, the lean-NOx trap catalyst can be implemented with known NOx adsorbers, and the catalyst may contain less or essentially no precious metals, such as Pt, Pd, and Rh, for reduction of the nitrate to N2 . Therefore, this method for NOx emission reduction is inexpensive and reliable. It improves the activity, durability, and temperature window of lean-NOx trap catalysis, and it also allows the combustion of fuels containing relatively high sulfur contents with a concommitant reduction of NOx , particularly in an oxygen-rich vehicular environment. Tonkyn et al. [94] investigated NOx removal using Na-Y and Ba-Y catalysts placed downstream of a plasma reactor and propene as additive in the discharge to promote NO to NO2 oxidation. They studied a multiple step treatment strategy where two plasmacatalytic reactors were used in series. The plasma reactor has two sets of electrodes, one set connected to high voltage, the other set grounded. The electrodes are metal rods introduced in alumina tubes. The electrode array was designed so that each high voltage electrode is surrounded by four grounded electrodes. The discharge was operated in a.c. regime at 400 Hz, with sinusoidal voltages up to 7 kVrms . The authors obtained NOx conversions up to 70% when using a single stage of plasma-catalytic treatment, while when adding a second plasma reactor followed by a catalytic reactor in series with the first ones, NOx removal improved up to almost 90%. The second stage of plasma-catalytic treatment resulted not only in higher maximum NOx conversion, but also in energy savings, either by reducing the plasma energy required, or by reducing the amount of hydrocarbon required. The catalytic activities of a series of alkali and alkaline earth cation exchanged Y, FAU zeolites were investigated in plasma-assisted NOx reduction using simulated diesel engine exhaust gas [107,108]. The plasma was generated in a DBD reactor between an array of oppositely polarized electrodes. The catalytic reactor was placed downstream of the plasma reactor. The authors found that the catalytic activity of the zeolites showed significant variations with both the nature of the charge compensating cation, and the method of catalyst preparation. The highest NOx conversion level was achieved over Ba-Y, FAU catalyst prepared by a multiple ion exchange method, in which each solution ion exchange step was followed by high temperature calcination. The catalytic activity of Ba-Y increased with increasing Ba2+ /Na+ ratio in the zeolite framework. It is believed that during the high temperature calcination, a redistribution of Ba2+ and Na+ ions take place in the zeolite structure. As a result of this ion migration, additional Na+ ions can be ion-exchanged for Ba2+ , allowing for the preparation of Ba−Y samples with higher Ba2+ /Na+ ratios, than can be achieved by simple multiple solution ion exchanges. The authors found that Ba−Y catalysts can chemisorb much larger amounts of NO2 than Na−Y and the amount of chemisorbed NO2 further increased following the multiple ionexchange/anneal preparation method. Therefore, they suggested that high chemisorption capacity and strong adsorption of NO2 are prerequisites for high activity in plasmaassisted NOx reduction. An enhancement of NOx removal was also obtained when using CuZSM catalysts, placed downstream of the plasma reactor [109]. Dors et al. [49] investigated NOx removal in a hybrid system consisting in a d.c. corona discharge and a molecular sieve placed inside the plasma region. In the corona discharge, all NO converted was oxidized to NO2 , so the removal rate of NOx was zero.

388

Past and Present in DeNOx Catalysis

In the presence of the molecular sieve, NOx removal increased up to 60%. This was explained by the formation of N2 O3 after coadsorption of NO and NO2 on the molecular sieve. If the sieve was saturated with NH3 , NOx removal increased even more. The authors suggested the reduction of NOx due to reactions involving molecular sieve fixed NH4 + ions. Okubo et al. [110] also used a combination of adsorption and non-thermal plasma for NOx reduction. They employed a dielectric packed-bed reactor operated in pulsed mode, and the packing material consisted of molecular sieve pellets of 2 mm, made of zeolite, with 1 nm pore size. First, the gas containing NO was passed through the reactor and adsorption took place. This was followed by the desorption and reduction steps, when either dry air, or nitrogen, or a combination between these was used. When using N2 , the efficiency for NOx reduction was 9–12 g/kWh, however, in the presence of more than 5% oxygen, oxidation took place. Several authors used -alumina in combination with non-thermal plasma for NOx removal [35,36,42,47,93,111,112]. In a pulsed corona discharge in a gas mixture containing humid air, ethanol and NO, with -Al2 O3 placed downstream of the plasma reactor, Georgescu et al. obtained 82% NOx conversion and energy costs of about 19 eV/NOx molecule removed [42]. Tran et al. [111] used a DBD reactor with coaxial geometry, operated in a.c. regime with voltages in the range of 3–16 kV, followed by a catalytic reactor containing In/ -Al2 O3 . The effect of several hydrocarbons (propane, propene, methanol, ethanol, 1-propanol, 2-propanol, acetaldehyde, isooctane, etc.) on NOx removal was tested and it was found that the functionality, the location of the functional group, as well as the molecular size influenced the ability of the hydrocarbon to participate in NOx reduction chemistry. The authors found particularly high conversion at 623 K when primary alcohols were used in combination with In/ -Al2 O3 . Results from tests performed at 473 K showed that indium incorporation slightly activated -Al2 O3 at low temperature, but activity in this low temperature regime is far inferior to materials such as Ba/zeolite Y. Plasma assist benefited significantly at 623 K, but high temperature (773 K) operation was not impacted. A similar behavior was observed for Ag/Al2 O3 catalysts located downstream of the plasma reactor [112]: at 473 K NOx removal did not exceed 60%, whereas at 623 K it increased to 97%. Combinations of Ba−Y zeolite and Ag/ -Al2 O3 enhanced NOx conversion, especially when the zeolite was placed upstream of Ag/ -Al2 O3 . The authors explained the higher activity by significant formation of formaldehyde over Ba-Y zeolite, which can be effectively used as a reducing agent over Ag/Al2 O3 . In this case, NOx conversion ranged from 80 to more than 95% under steady-state operation; whereas under transient operation, a cycle average of 70% reduction was achieved. The drop in efficiency in cycled operation is attributed to NOx desorption during heating transients below 473 K. Alumina and BaTiO3 or TiO2 mixed with alumina were tested [93], directly inside the plasma region of a dielectric packed-bed reactor, at 423 K. The electrical discharge was operated in pulsed mode, with high voltage pulses with rise-time of about 20 ns, duration of 2.5 ms and repetition rate of 300–350 Hz. The results indicated that in the absence of any additive, reduction was minor and the dominant NOx removal path was oxidative adsorption. In contrast, the reductive pathway played a dominant role in the presence of hydrocarbon additives like methanol or ethanol. Mechanical mixing of the

-Al2 O3 with BaTiO3 or TiO2 did not improve the results, in fact NOx conversion was

Plasma-Assisted NOx Abatement Processes

389

slightly lower as compared to -Al2 O3 alone. The lowest energy cost obtained for NOx removal was 28.5 eV/NOx molecule, when ethanol was used as additive. Holzer et al. [113] proposed the combination of BaTiO3 as ferroelectric and of a perovskite (LaCoO3  as a catalytic active material, located inside the discharge zone of NTP reactors as a promising way to improve the performance for the removal of hazardous substances, especially those appearing in low concentrations. Several coaxial barrier-discharge plasma reactors varying in size and barrier material (glass, Al2 O3 , and TiO2  were used for this purpose. In the ferroelectric packed-bed reactors, better energy efficiency and CO2 selectivity were found for the oxidation of the model substances. The change of the electric discharge behavior was caused by a larger number of nonselective and highly reactive plasma species formed within the ferroelectric bed. When combining ferroelectric (BaTiO3  and catalytically active materials (LaCoO3 , only a layered implementation led to synergistic effects, using both highly energetic species formed in the ferroelectric packed-bed and the potential for total oxidation provided by the catalytically active material in the second part of the packed bed. Huang and Matsuda [114] used a pulsed corona discharge reactor in wire-to-cylinder configuration with the ground electrode covered with a porous layer of Ca(OH)2 for NOx removal. The high voltage pulses were generated by charging a capacitor, which is subsequently discharged by means of a rotating spark-gap switch. The voltage pulses had rise times of about 50 ns, durations of 300 ns and repetition rates of 50 Hz. The authors found that NO removal was significantly higher when the Ca(OH)2 layer was present (76–93%, for voltage amplitudes in the range 16–22 kV) as compared to the results with the plasma alone (40–60%, over the same voltage range), due to adsorption of NO on this layer. In addition, NO2 formation by NO oxidation was much reduced in the presence of Ca(OH)2 . The influence of oxygen content (0–2%), gas temperature (293–373 K) and water vapor on NO removal, and NO2 formation were investigated. In contrast, placing a Ca(OH)2 packed-bed downstream of the plasma reactor resulted in about 30% lower NO removal efficiency. NO2 could be removed by the Ca(OH)2 sorbent, however, unreacted NO could not be removed in this two-stage configuration. Several catalysts were tested in combination with plasma [47]: Pd/Al2 O3 , CuO-ZnOAl2 O3 , Fe2 O3 -Cr2 O3 , shaped as pellets and Pt-Rh/CeO2 -Al2 O3 and CuMnAlO4 , coated on a cordierite honeycomb structure for treating the exhaust gas from a diesel engine under no load conditions. The catalysts were placed either in the plasma region of a pulsed corona discharge, or outside of the plasma, in a catalytic reactor with variable temperature. The results showed that when Pd/Al2 O3 , Fe2 O3 , and CuO pellets were placed inside the plasma zone, even at room temperature, some reactions were taking place on the surface of the catalyst in presence of discharge resulting in the removal of NOx . Using Pd/Al2 O3 downstream and upstream of the plasma reactor, the authors found that the efficiency for NOx removal was higher when the catalytic reactor was placed before the plasma reactor. This prevented the deposition of HNO3 formed in the discharge on the catalyst. The results obtained with Fe2 O3 catalyst placed upstream of the plasma reactor were similar with those achieved with Pd/Al2 O3 . Promising results were also obtained using the CuMnAlO4 catalyst in honeycomb structure, which performed equally well with respect to NOx removal as the conventional three-way catalyst Pt-Rh/CeO2 -Al2 O3 .

390

Past and Present in DeNOx Catalysis

5. NOX AND SOOT REMOVAL BY PLASMA-ASSISTED CATALYSIS Peng et al. [115] reported an active perovskite-type catalyst (La08 K02 Cu005 Mn095 O3  prepared via evaporation of the precursor salts and calcination of the resulted mixture at 800 C as being active in NOx removal under plasma-assisted catalysis in the presence of C3 H6 and soot. As for the previous examples, the O radical was produced by an electron impact dissociation of O2 , and the remediation of NOx and HC by plasma was initiated by O. When adding C3 H6 , NO can be rapidly converted to NO2 under the effects of HO2 and RO2 radicals formed by propene decomposition. The role of the catalyst both for the removal of NOx and destruction of soot was evidenced from the conversion data. Chemical analysis of the exhaust indicated that the addition of catalyst promoted the removal of NOx , HC, and soot. Thus, for a C3 H6 concentration in the feed gas of 0.27%, the maximum NOx removal rate increased from 43.5% (with no catalyst) to 72.2% in the presence of the catalyst. However, the role of the catalyst, as in many other examples, has not been explained and the contribution of the perovskite-promotor species in this process is not defined. But the most important challenge in NOx and soot removal by plasma-assisted catalysis refers to the treatment of diesel exhaust [116]. The discovery of new combinations leading to both an advanced total oxidation of unburned organic fuels and fragments, and of soot is extremely important for the near-future. According to the new regulations concerning the use of biomass in the production of fuels, the achievements in the direction NOx and soot removal by plasma-assisted catalysis should be extended to the removal of tars [117].

6. CONCLUSIONS Numerous studies showed that non-equilibrium plasma can become a promising technology for NOx abatement. Chemical reactions in non-thermal plasma are initiated by high-energy electrons, which collide with the background gas molecules, generating chemically active species. These processes occur close to room temperature, without requiring the heating of the entire gas stream. The chemically active species formed react further with the NOx molecules, which leads to NOx removal. When the carrier gas is nitrogen, NO is removed by reduction, with energy costs of about 240 eV/NO removed, similar to the energy required for the formation of nitrogen atoms. In the presence of oxygen, the oxidative pathway becomes dominant, NO being largely converted into NO2 . The energy costs in this case are much lower. Especially when hydrocarbons are present in the gas, the energy consumption is reduced significantly, to about 10 eV/NO removed, due to reactions with HO2 and peroxy radicals, however, the NOx concentration remains almost unaffected. The oxidation of NO to NO2 can be considered an advantage when combining plasma with selective catalytic reduction, as it is well-known that at low temperature the efficiency of NOx reduction by SCR strongly depends on the NO2 concentration in the gas. An oxidation pre-treatment of the gas by non-thermal plasma, prior to the SCR reactor greatly enhances the performance of SCR at temperatures below 500 K. Therefore, the PE-SCR technique is very promising for efficient NOx reduction.

Plasma-Assisted NOx Abatement Processes

391

REFERENCES [1] Futamura, S., Zhang, A., Prieto, G. and Yamamoto, T. (1998) Factors and intermediates governing byproduct distribution for decomposition of butane in nonthermal plasma, IEEE Trans. Ind. Appl. 34, 967–74. [2] Ogata, A., Mizuno, K., Kushiyama, S. and Yamamoto, T. (1998) Methane Decomposition in a Barium Titanate Packed-Bed Nonthermal Plasma Reactor, Plasma Chem. Plasma Process 18, 363–73. [3] Holzer, F., Roland, U. and Kopinke, F.-D. (2002) Combination of non-thermal plasma and heterogeneous catalysis for oxidation of volatile organic compounds Part 1. Accessibility of the intra-particle volume, Appl. Cat. B: Env. 38, 163–81. [4] Rudolph, R., Francke, K.P. and Miessner, H. (2002) Concentration dependence of VOC decomposition by dielectric barrier discharges, Plasma Chem. Plasma Process. 220, 401–12. [5] Magureanu, M., Mandache, N.B., Elloy, P. et al. (2005) Plasma-assisted catalysis for volatile organic compounds abatement, Appl. Cat. B: Env. 61, 12–20. [6] Roland, U., Holzer, F., Kopinke, F. D. et al. (2002) Improved oxidation of air pollutants in a non-thermal plasma, Catal. Today 73, 315–23. [7] Magureanu, M., Mandache, N.B., Gaigneaux, E.M. et al. (2006) Toluene oxidation in a plasma-catalytic system, J. Appl. Phys. 99, 301–08. [8] Futamura, S., Zhang, A.H. and Yamamoto, T. (1999) Mechanisms for formation of inorganic byproducts in plasma chemical processing of hazardous air pollutants, IEEE Trans. Ind. Appl. 35(4), 760–6. [9] Kirkpatrick, M.J., Finney, W.C. and Locke, B.R. (2003) Chlorinated organic compound removal by gas phase pulsed streamer corona electrical discharge with reticulated vitreous carbon electrodes, Plasmas and Polymers 8, 165–77. [10] Hsiao, M.C., Merritt, B.T., Penetrante, B.M. et al. (1995) Plasma-assisted decomposition of methanol and trichloroethylene in atmospheric pressure air streams by electrical discharge processing, J. Appl. Phys. 78, 3451–6. [11] Ma, H., Chen, P., Zhang, M., et al. Study of SO2 removal using non-thermal plasma induced by dielectric barrier discharge (DBD), Plasma Chem. Plasma Process 22, 239–54. [12] Ma, H., Chen, P. and Ruan, R. (2001) H2 S and NH3 removal by silent discharge plasma and ozone combo-system, Plasma Chem. Plasma Process. 21, 611–24. [13] Chang, J.-S., Lawless, P.A. and Yamamoto, T. (1991) Corona discharge processes, IEEE Trans. Pl. Sci. 19, 1152–66. [14] van Veldhuizen, E.M. and Rutgers, W.R. (2002) Pulsed positive corona streamer propagation and branching, J. Phys. D: Appl. Phys. 35, 2169–79. [15] Eliasson, B. and Kogelschatz, U. (1991) Nonequilibrium volume plasma chemical processing, IEEE Trans. Pl. Sci. 19, 1063–77. [16] Kulikovsky, A.A. (1997) Production of chemically active species in the air by a single positive streamer in a nonuniform field, IEEE Trans. Pl. Sci. 25, 439–46. [17] Eliasson, B. and Kogelschatz, U. (1991) Modeling and applications of silent discharge plasmas, IEEE Trans. Pl. Sci. 19, 309–23. [18] Moisan, M. and Pelletier, J. (eds) (1992) Microwave excited plasmas, Elsevier Science Publishers, Amsterdam, pp. 5–25. [19] MacDonald, A.D. (1966) Microwave Breakdown in Gases, John Wiley, New York, pp. 10–29. [20] Takaki, K., Shimizu, M., Mukaigawa, S. et al. (2004) Effect of Electrode Shape in Dielectric Barrier Discharge Plasma Reactor for NOx Removal, IEEE Trans. Pl. Sci. 32, 32–8. [21] Jani, M.A., Toda, K., Takaki, K. et al. (2000) An experimental comparison between electrode shapes for NOx treatment using a dielectric barrier discharge, J. Phys. D: Appl. Phys. 33, 3078–82.

392

Past and Present in DeNOx Catalysis

[22] Jani, M.A., Takaki, K. and Fujiwara, T. (1999) Streamer polarity dependence of NOx removal by dielectric barrier discharge with a multipoint-to-plane geometry, J. Phys. D: Appl. Phys. 32, 2560–7. [23] Toda, K., Takaki, K., Kato, S. et al. (2001) Removal of NO and NOx using a multipoint-type dielectric barrier discharge at a narrow gap, J. Phys. D: Appl. Phys. 34, 2032–6. [24] McLarnon, C.R. and Penetrante, B.M. (1998) Effect of reactor design on the plasma treatment of NOx , SAE Paper No. 982434. [25] Takaki, K., Jani, M.A. and Fujiwara, T. (1999) Removal of nitric oxide in flue gases by multi-point to plane dielectric barrier discharge, IEEE Trans. Plasma Sci. 27, 1137–45. [26] Chang, M.B., Kushner, M.J. and Rood, M.J. (1992) Gas-phase removal of NO from gas streams via dielectric barrier discharges, Environ. Sci. Technol., 26, 777–81. [27] Lee, H.M., Chang, M.B. and Yang, S.C. (2003) Plasma-assisted process for removing NO/NOx from gas streams with C2 H4 as additive, J. Env. Eng. 129, 800–10. [28] Penetrante, B.M., Brusasco, R.M., Merritt, B.T. et al. (1999) Environmental applications of low-temperature plasmas, Pure & Appl. Chem. 71, 1829–35. [29] Daou, F., Vincent, A. and Amouroux, J. (2003) Point and multipoint to plane barrier discharge process for removal of NOx from engine exhaust gases: Understanding of the reactional mechanism by isotopic labeling, Plasma Chem. Plasma Process 23, 309–25. [30] Khacef, A., Cormier, J.M. and Pouvesle, J.M. (2002) NOx remediation in oxygen-rich exhaust gas using atmospheric pressure non-thermal plasma generated by a pulsed nanosecond dielectric barrier discharge, J. Phys. D: Appl. Phys. 35, 1491–8. [31] Ravi, V., Mok, Y.S., Rajanikanth, B.S. et al. (2003) Temperature effect on hydrocarbonenhanced nitric oxide conversion using a dielectric barrier discharge reactor, Fuel Processing Technology 81, 187–99. [32] McLarnon, C.R. and Penetrante, B.M. (1998) Effect of gas composition on the NOx conversion chemistry in a plasma, SAE Paper No. 982433. [33] Niessen, W., Wolf, O., Schruft, R. et al. (1998) The influence of ethene on the conversion of NOx in a dielectric barrier discharge, J. Phys. D: Appl. Phys. 31, 542–50. [34] Hammer, Th., Kappes, Th. and Baldauf, M. (2004) Plasma catalytic hybrid processes: gas discharge initiation and plasma activation of catalytic processes, Catal. Today 89, 5–14. [35] Penetrante, B.M., Brusasco, R.M., Merritt, B.T. et al. (1998) Plasma-assisted catalytic reduction of NOx , SAE Paper No. 982508. [36] Mok, Y.S., Dors, M. and Mizerazcyk, J. (2004) Effect of reaction temperature on NOx removal and formation of ammonium nitrate in nonthermal plasma process combined with selective catalytic reduction, IEEE Trans. Plasma Sci. 32, 799–807. [37] Mok, Y.S., Ravi, V., Kang, H.-C. and Rajanikanth, B.S. (2003) Abatement of nitrogen oxides in a catalytic reactor enhanced by nonthermal plasma discharge, IEEE Trans. Pl. Sci. 31, 157–65. [38] Masuda, S. (1998) Pulse corona induced plasma chemical process: a horizon of new plasma chemical technologies, Pure & Appl. Chem. 60, 727–31. [39] Wang, D., Namihira, T., Fujiya, K. et al. (2004) The reactor design for diesel exhaust control using a magnetic pulse compressor, IEEE Trans. Plasma Sci. 32, 2038–44. [40] Zhao, G.-B. Garikipati, S.V.B.J., Hu, X. et al. (2005) Effect of reactor configuration on nitric oxide conversion in nitrogen plasma, AIChE J. 51, 1813–21. [41] Namihira, T., Tsukamoto, S., Wang, D. (2000) Improvement of NOx removal efficiency using short-width pulsed power, IEEE Trans. Pl. Sci. 28, 434–42. [42] Georgescu, N., Minea, R. and Lungu, C.P. (2005) Depollution experiments with repetitive pulsed corona plasmas, J. Optoel. & Adv. Materials 7, 2501–7. [43] Kim, Y.-H. and Hong, S.H. (2002) Two-dimensional simulation images of pulsed corona discharges in a wire–plate reactor, IEEE Trans. Pl. Sci. 30, 168–9. [44] Naidis, G.V. (1996) On streamer interaction in a pulsed positive corona discharge, J. Phys. D: Appl. Phys. 29, 779–83.

Plasma-Assisted NOx Abatement Processes

393

[45] Hu, X., Zhang, J.-J. Mukhnahallipatna, S. et al. (2003) Transformations and destruction of nitrogen oxides – NO, NO2 and N2 O – in a pulsed corona discharge reactor, Fuel 82, 1675–84. [46] Zhao, G.-B., Garikipati, S.V.B.J., Hu, X. et al. (2005) Effect of oxygen on nonthermal plasma reactions of nitrogen oxides in nitrogen, AIChE J. 51, 1800–12. [47] Rajanikanth, B.S. and Ravi, V. (2002) Pulsed electrical discharges assisted by dielectric pellets/catalysts for diesel engine exhaust treatment, IEEE Trans. Diel. El. Insul. 9, 616–26. [48] Namihira, S., Tsukamoto, D., Wang, H. et al. (2001) Influence of gas flow rate and reactor length on NO removal using pulsed power, T. IEEE Trans. Pl. Sci. 29, 592–8. [49] Dors, M., Mizeraczyk, J. and Nichipor, G.V. (2004) Influence of ammonia on NOx removal in corona discharge-molecular sieve hybrid system, J. Adv. Oxid. Technol. 7, 142–4. [50] McAdams, R. (2001) Prospects for non-thermal atmospheric plasmas for pollution abatement, J. Phys. D: Appl. Phys. 34, 2810–21. [51] Zhao, G.-B., Hu, X., Argyle, M.D. et al. (2005) Effect of CO2 on nonthermal-plasma reactions of nitrogen oxides in N2 . 1. ppm-level concentrations, Ind. Eng. Chem. Res. 44, 3925–34. [52] Zhao, G.-B., Hu, X., Argyle, M.D. et al. (2005) Effect of CO2 on nonthermal-plasma reactions of nitrogen oxides in N2 . 2. Percent-Level Concentrations, Ind. Eng. Chem. Res. 44, 3935–46. [53] Dors, M. and Mizeraczyk, J. (2004) NOx removal from a flue gas in a corona dischargecatalyst hybrid system. Catal. Today 89, 127–33. [54] Penetrante, B.M., Hsiao, M.C., Merritt, B.T. et al. (1995) Comparison of electrical discharge techniques for nonthermal plasma processing of NO in N2 , IEEE Trans. Pl. Sci. 23, 679–87. [55] Penetrante, B. M., Hsiao, M. C., Merritt, B. T. et al. (1996) Pulsed corona and dielectricbarrier discharge processing of NO in N2 , Appl. Phys. Lett. 68, 3719–21. [56] Wu, Z.L., Gao, X., Luo, Z.Y. et al. (2005) NOx treatment by DC corona radical shower with different geometric nozzle electrodes. Energy & Fuels 19, 2279–86. [57] Kanazawa, S., Sumi, T., Shimamoto, S. et al. (2004) Diagnostics of NO oxidation process in a nonthermal plasma reactor: Features of DC streamer-corona discharge and NO LIF profile, IEEE Trans. Pl. Sci. 32, 25–31. [58] Kanazawa, S., Sumi, T., Sato, N. et al. (2005) Wide-range two-dimensional imaging of NO density profiles by LIF technique in a corona radical shower reactor, IEEE Trans. Ind. Appl. 41, 200–5. [59] Lin, H., Gao, X., Luo, Z. et al. (2004) Removal of NOx with radical injection caused by corona discharge, Fuel 83, 1349–55. [60] Park, J.Y., Tomicic, I., Round, G.F. et al. (1999) Simultaneous removal of NOx and SO2 from NO–SO2 –CO2 –N2 –O2 gas mixtures by corona radical shower systems, J. Phys. D: Appl. Phys. 32, 1006–11. [61] Chang, J.-S., Urashima, K., Tong, Y.X. et al. (2003) Simultaneous removal of NOx and SO2 from coal boiler flue gases by DC corona discharge ammonia radical shower systems: Pilot plant tests, J. Electrostatics 57, 313–23. [62] Kanazawa, S., Chang, J.S., Round, G. et al. (1997) Removal of NOx from flue gas by corona discharge activated methane radical showers, J. Electrostatics 40&41, 651–6. [63] Kanazawa, S., Chang, J.S., Round, G. et al. (1998) Reduction of NOx from flue gas by corona discharge activated ammonia radical showers, J. Combust. Sci. Technol. 133, 93–105. [64] Riva, R., Legentil, M., Pasquiers, S. et al. (1995) Experimental and theoretical investigations of a XeCl phototriggered laser, J. Phys. D: Appl. Phys. 28, 856–72. [65] Richeboeuf, L., Pasquiers, S., Legentil, M. et al. (1998) The influence of H2 and C2 H6 molecules on discharge equilibrium and F-atom production in a phototriggered HF laser using SF6 , J. Phys. D: Appl. Phys. 31, 373–89. [66] Rozoy, M., Postel, C. and Puech, V. (1999) NO removal in a photo-triggered discharge reactor, Plasma Sources Sci. Technol. 8, 337–48.

394

Past and Present in DeNOx Catalysis

[67] Fresnet, F., Baravian, G., Pasquiers, S., et al. (2000) J. Phys. D: Appl. Phys. 33, 1315–22. [68] Fresnet, F., Baravian, G., Magne, L. et al. (2002) Influence of water on NO removal by pulsed discharge in N2 /H2 O/NO mixtures, Plasma Sources Sci. Technol. 11, 152–60. [69] Pott, A., Dörk, T., Uhlenbusch, J. et al. (1998) Polarization-sensitive coherent anti-Stokes Raman scattering applied to the detection of NO in a microwave discharge for reduction of NO, J. Phys. D: Appl. Phys. 31, 2485–98. [70] Ighigeanu, D., Martin, D., Zissulescu, E. et al. (2005) SO2 and NOx removal by electron beam and electrical discharge induced non-thermal plasmas, Vacuum 77, 493–500. [71] Baeva, M., Gier, H., Pott, A. et al. (2001) Studies on gas purification by a pulsed microwave discharge at 2.46 GHz in mixtures of N2 /NO/O2 at atmospheric pressure, Plasma Chem. Plasma Process. 21, 225–47. [72] Baeva, M., Gier, H., Pott, A. et al. (2002) Pulsed microwave discharge at atmospheric pressure for NOx decomposition, Plasma Sources Sci. Technol. 11, 1–9. [73] Baeva, M., Pott, A., and Uhlenbusch, J. (2002) Modelling of NOx removal by a pulsed microwave discharge, Plasma Sources Sci. Technol. 11, 135–41. [74] Zhao, G.-B., Hu, X., Yeung, M.-C. et al. (2004) Nonthermal plasma reactions of dilute nitrogen oxide mixtures: NOx in nitrogen, Ind. Eng. Chem. Res. 43, 2315–23. [75] Gentile, A.C. and Kushner, M.J. (1995) Reaction chemistry and optimization of plasma remediation of Nx Oy from gas streams, J. Appl. Phys. 78, 2074–85. [76] Dorai, R. and Kushner, M.J. (2001) Effect of multiple pulses on the plasma chemistry during the remediation of NOx using dielectric barrier discharges, J. Phys. D: Appl. Phys. 34, 574–83. [77] Orlandini, I. and Riedel, U. (2000) Chemical kinetics of NO removal by pulsed corona discharges, J. Phys. D: Appl. Phys. 33, 2467–74. [78] Stefanovic, I., Bibinov, N.K., Deryugin, A.A., et al. (2001) Kinetics of ozone and nitric oxides in dielectric barrier discharges in O2 /NOx and N2 /O2 /NOx mixtures, Plasma Sources Sci. Technol. 10, 406–16. [79] Orlandini, I. and Riedel, U. Numerical simulation of NOx -removal by pulsed corona discharges – the effect of ethene, AIAA Paper 2000-0720. [80] Eichwald, O., Guntoro, N.A., Yousfi, M. et al. (2002) Chemical kinetics with electrical and gas dynamics modelization for NOx removal in an air corona discharge, J. Phys. D: Appl. Phys. 35, 439–50. [81] Shin, H.-H. and Yoon, W.-S. (2003) Hydrocarbon effects on the promotion of non-thermal plasma NO–NO2 conversion, Plasma Chem. Plasma Process. 23, 681–704. [82] Hammer, T. (2002) Non-thermal plasma application to the abatement of noxious emissions in automotive exhaust gases, Plasma Sources Sci. Technol. 11, A196–A201. [83] Filimonova, E.A., Kim, Y.H., Hong, S.H. et al. (2002) Multiparametric investigation on NOx removal from simulated diesel exhaust with hydrocarbons by pulsed corona discharge, J. Phys. D: Appl. Phys. 35, 2795–807. [84] Dorai, R. and Kushner, M.J. (2003) Consequences of unburned hydrocarbons on microstreamer dynamics and chemistry during plasma remediation of NOx using dielectric barrier discharges, J. Phys. D: Appl. Phys. 36, 1075–83. [85] Dorai, R. and Kushner, M.J. (1999) Effect of propene on the remediation of NOx from engine exhausts, SAE Paper 99FL-472. [86] Martin, A.R., Shawcross, J.T. and Whitehead, J.C. (2004) Modelling of non-thermal plasma after treatment of exhaust gas streams, J. Phys. D: Appl. Phys. 37, 42–9. [87] Dorai, R. and Kushner, M.J. (2000) Consequences of propene and propane on plasma remediation of NOx , J. Appl. Phys. 88, 3739–46. [88] Dorai, R. and Kushner, M.J. (2002) Repetitively pulsed plasma remediation of NOx in soot laden exhaust using dielectric barrier discharges, J. Phys. D: Appl. Phys. 35, 2954–68.

Plasma-Assisted NOx Abatement Processes

395

[89] Dorai, R., Hassouni, K. and Kushner, M.J. (2000) Interaction between soot particles and NOx during dielectric barrier discharge plasma remediation of simulated diesel exhaust, J. Appl. Phys. 88, 6060–71. [90] Pârvulescu, V.I., Grange, P. and Delmon, B. (1998) Catalytic removal of NO, Catal. Today, 46, 233–317. [91] Broer, S. and Hammer, T. (2000) Selective catalytic reduction of nitrogen oxides by combining a non-thermal plasma and a V2 O5 -WO3 /TiO2 catalyst, Appl. Catal. B: Env. 28, 101–11. [92] Demidouk, V., Ravi, V., Chae, J.-O. et al. (2005) Pt-Al2 O3 catalyst and discharge plasma pre-treatment techniques for enhancing selective catalytic reduction of nitrogen oxides: A comparative study, React. Kinet. Catal. Lett. 85, 239–44. [93] Kim, H.H., Takashima, K., Katsura, S. et al. (2001) Low-temperature NOx reduction processes using combined systems of pulsed corona discharge and catalysts, J. Phys. D: Appl. Phys. 34, 604–13. [94] Tonkyn, R.G., Barlow, S.E. and Hoard, J.W. (2003) Reduction of NOx in synthetic diesel exhaust via two-step plasma-catalysis treatment, Appl. Cat. B: Env. 40, 207–17. [95] Chae, J.-O. (2003) Non-thermal plasma for diesel exhaust treatment, J. Electrostatics 57, 251–62. [96] Penetrante, B.M., Brusasco, R.M., Merritt, B.T. et al. (1999) Environmental applications of low-temperature plasmas, Pure & Appl. Chem. 71, 1829–35. [97] Mok, Y.S. and Huh, Y.J. (2005) Simultaneous removal of nitrogen oxides and particulate matters from diesel engine exhaust using dielectric barrier discharge and catalysis hybrid system, Plasma Chem. Plasma Process. 25, 625–39. [98] Mok, Y.S., Lee, H.J., Dors, M. et al. (2005) Improvement in selective catalytic reduction of nitrogen oxides by using dielectric barrier discharge, Chem. Eng. J. 110, 79–85. [99] Mok, Y.S., Koh, D.J., Shin, D.N. et al. (2004) Reduction of nitrogen oxides from simulated exhaust gas by using plasma–catalytic process, Fuel Process Technol. 86, 303–17. [100] Lee, Y.H., Chung, J.W., Choi, Y.R. et al. (2004) NOx removal characteristics in plasma plus catalyst hybrid process, Plasma Chem. Plasma Process 24, 137–54. [101] Rajanikanth, B.S., Srinivasan, A.D. and Ravi, V. (2005) Discharge plasma treatment for NOx reduction from diesel engine exhaust: A laboratory investigation, IEEE Trans. Diel. El. Insul. 12, 72–80. [102] Kang, M., Kim, B.J., Cho, S.M. et al. (2002) J. Mol. Catal. A: Chem. 180, 125–32. [103] Wallis, A.E., Whitehead, J. C. and Zhang, K. (2007) Plasma-assisted catalysis for the destruction of CFC-12 in atmospheric pressure gas streams using TiO2 , Catal. Lett. 113, 29–33. [104] Nakamura, I., Negishi, N., Kutsuna, S. et al. (2000) Role of oxygen vacancy in the plasmatreated TiO2 photocatalyst with visible light activity for NO removal, J. Mol. Catal. A: Chem. 161, 205–12. [105] Ogata, A., Kim, H.H., Oh, S.M. et al. (2006) Evidence for direct activation of solid surface by plasma discharge on CFC decomposition, Thin Solid Films, 506, 373–7. [106] Merritt, B.T., Penetrante, B.M., Vogtlin, G.E. et al. (2002) Plasma-assisted catalytic storage reduction system, US Patent 6374595. [107] Kwak, J.H., Szanyi, J. and Peden, C.H.F. (2003) Nonthermal plasma-assisted catalytic NOx reduction over Ba-Y, FAU: The effect of catalyst preparation, J. Catal. 220, 291–8. [108] Kwak, J.H., Szanyi, J. and Peden, C.H.F. (2004) Non-thermal plasma-assisted NOx reduction over alkali and alkaline earth ion exchanged Y, FAU zeolites, Catal. Today 89, 135–41. [109] Oda, T., Kato, T., Takahashi, T. et al. (1998) Nitric oxide decomposition in air by using nonthermal plasma processing with additives and catalyst, IEEE Trans. Ind. Appl. 34, 268–72. [110] Okubo, M., Inoue, M., Kuroki, T. et al. (2005) NOx reduction after treatment system using nitrogen nonthermal plasma desorption, IEEE Trans. Ind. Appl. 41, 891–9.

396

Past and Present in DeNOx Catalysis

[111] Tran, D.N., Aardahl, C.L., Rappe, K.G. et al. (2004) Reduction of NOx by plasma-facilitated catalysis over In-doped -alumina, Appl. Cat. B: Env. 48, 155–64. [112] Rappé, K.G., Hoard, J.W., Aardahl, C.L. et al. (2004) Combination of low and high temperature catalytic materials to obtain broad temperature coverage for plasma-facilitated NOx reduction, Catal. Today 89, 143–50. [113] Holzer, F., Kopinke, F.D. and Roland, U. (2005) Influence of ferroelectric materials and catalysts on the performance of non-thermal plasma (NTP) for the removal of air pollutants, Plasma Chem. Plasma Processing 25, 595–611. [114] Huang, L. and Matsuda, H. (2004) Removal of NO by a pulsed-corona reactor combined with in situ absorption, AIChE J. 50, 2676–81. [115] Peng, X., Lin, H., Huang, Z. et al. (2006) Effect of catalysis on plasma assisted catalytic removal of nitrogen oxides and soot, Chem. Eng. Technol. 29, 1262–6. [116] Narula, C.K., Daw, C.S., Hoard, J.W. et al. (2005) Materials issues related to catalysts for treatment of diesel exhaust, Int. J. Appl. Ceramic Technol. 2, 452–66. [117] Pemen, G.A.J.M., Devi, L., Yan, K., et al. (2007) Plasma-catalytical removal of tars from fuel gas obtained by biomass gasification, J. Adv. Oxidation Technol. 10, 116–20.

INDEX

Ammonia: oxidation, 280–1 slip, 275–7 Catalyst: preparation, 151, 178–80 stability, 5, 12, 17, 237 Cerium-based catalysts, 238, 241, 246, 253 Coated catalyst, 275–6 Computational spectroscopy, 28–9 Corona discharges, 262–364 Cu-ZSM5, 28, 57–8, 61, 278–81, 283–4, 286 DeNOx , 1 catalysis, 267–8 technology, 212, 227 DFT, 27–60, 70, 85–7 Dielectric barrier discharges, 16–17, 364 Diesel: emissions standards, 213 particle filter, 218–22 vehicles, 3, 211–12, 217, 261, 265, 267, 278 Dinitrogen bond formation, 35, 55, 57, 59–60, 133, 145 Dinitrosyl, 44, 45, 47, 49, 50, 52, 53, 54, 55, 56–7, 58, 70, 114, 116, 165, 302 EPR, 41, 49, 51, 53, 98, 112, 114, 129, 250 Extruded catalyst, 270, 276 Fast-SCR, 283–4 Fe-ZSM5, 266, 278, 279, 282, 283, 284, 286 Fourier transform infrared, 118, 181, 236, 237 FTIR imaging, 330, 354

HNCO, 263, 265, 266, 294 hydrolysis, 266 Hydrogen, 19, 98, 102, 105, 108, 125, 132, 151, 175, 177, 178, 179, 180, 192, 193, 194, 195, 198, 199, 202, 205, 245, 291–4, 298, 300, 304, 305, 306, 311, 337, 382 Impact of OSC materials on NOx abatement, 255–7 In situ and Operando infrared and Raman Spectroscopies, 98, 100, 103, 112 IR, 44, 47 Isocyanic acid, 262, 263, 264, 265, 294 Kinetics, 76–9, 81, 82, 83, 87, 88, 89, 90, 91, 98, 132, 170, 171, 227, 247, 300, 378 Lean DeNOx , 176, 177 Legislative trends, 214, 216 LNT systems, 178, 192 Mechanism, 128, 311 Metal-exchanged zeolite, 278–80, 283, 286 Model and Plasma assisted-DeNOx , 166, 167, 361–90 Modeling of supported catalysts, 83–5 Molecular beams, 67, 72, 77 N2 O formation, 284–5 NH3 -SCR, 4, 5, 6, 7, 8–16, 17, 125, 176 for mobile sources, 14–16 for stationary sources, 8–14 Nitrogen dioxide-hydrocarbon interaction, 2, 118

398 Nitrogen oxides (NO), 1, 2, 7, 10, 18, 63, 68, 71, 90 decomposition, 5, 35, 36, 56, 57, 88, 112, 123, 126, 129, 150, 158, 159, 162, 249, 375 dimer, 295 reduction in three-way catalysts, 3, 106, 243, 246, 293, 294 NO/NO2 -SCR, 274 NO-SCR, 126 Noble metals, 98, 99, 292, 293, 294–300 Non-thermal plasma, 16–18 NOx : abatement, 2, 5, 27, 125, 292, 308, 361–90 reduction mechanism, 311 selective catalytic reduction, 1, 7–20 storage, 131, 223, 231 storage-reduction, 345, 347 storage-reduction catalyst, 18–20 trap, 106 NOx /CO2 compromise, 227, 231 NOx /HC compromise, 215, 227 NOx /soot compromise, 213 NOx Trap, 223–5 NOx -trap catalysts, 117, 124, 175, 386 Oxygen storage capacity, 117, 235, 236, 296 Perovskite, 309–10, 314, 315–17, 387, 389 Plasma-assisted heterogeneous catalysis, 362 Post combustion catalysis, 384 Pt–Ba/Al2 O3 , 124, 175, 180, 188, 190 Rare earth oxides, 112, 236, 251

Index SCR: activity, 267–8, 282–3 catalyst, 278–85 mechanism, 128 mobile, 262, 267–8 reactions, 271–5 selectivity, 268–9 stationary, 8–14 Selective catalytic reduction with ammonia, 8–14, 262–7, 275–7, 281–2 of NOx , 1–8 Single crystal surfaces, 68–9, 72, 78, 79–81, 85 Slip catalyst, 277 Spin density, 27–8, 32, 34, 36, 40, 43–4, 52–5 Standard-SCR, 267–8, 273, 278–80 Statistical design of experiments, 67–90 Surface science, 67–90 Temperature programmed desorption, 68–72 Three-function model, 145–72 Transient response method (TRM), 179, 184–5, 188–9 Transition metal cation, 114, 120, 147–8, 172 Transition-metal ions, 27, 28 Ultrahigh vacuum, 68, 79, 98 Urea, 2–3, 7–8, 14–16, 109, 176–7, 228–31, 261–71, 276, 282 decomposition, 109, 265–6 dosing strategy, 276 SCR, 7, 176, 261, 265, 278 V2 O5 /WO3 -TiO2 , 266–76, 278–81, 283–4 Vanadia catalyst, 278–80 catalyst, 278–85

STUDIES IN SURFACE SCIENCE AND CATALYSIS Advisory Editors: B. Delmon, Université Catholique de Louvain, Louvain-la-Neuve, Belgium J.T. Yates, University of Pittsburgh, Pittsburgh, PA, U.S.A.

Volume 1

Volume 2

Volume 3

Volume 4

Volume 5

Volume 6

Volume 7

Volume 8 Volume 9

Volume 10

Volume 11

Volume 12

Volume 13

Preparation of Catalysts I. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the First International Symposium, Brussels, October 14–17, 1975 edited by B. Delmon, P.A. Jacobs and G. Poncelet The Control of the Reactivity of Solids. A Critical Survey of the Factors that Influence the Reactivity of Solids, with Special Emphasis on the Control of the Chemical Processes in Relation to Practical Applications by V.V. Boldyrev, M. Bulens and B. Delmon Preparation of Catalysts ll. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Second International Symposium, Louvain-la-Neuve, September 4–7, 1978 edited by B. Delmon, P. Grange, P. Jacobs and G. Poncelet Growth and Properties of Metal Clusters. Applications to Catalysis and the Photographic Process. Proceedings of the 32nd International Meeting of the Société de Chimie Physique, Villeurbanne, September 24–28, 1979 edited by J. Bourdon Catalysis by Zeolites. Proceedings of an International Symposium, Ecully (Lyon), September 9–11, 1980 edited by B. Imelik, C. Naccache, Y. BenTaarit, J.C. Vedrine, G. Coudurier and H. Praliaud Catalyst Deactivation. Proceedings of an International Symposium, Antwerp, October 13–15, 1980 edited by B. Delmon and G.F. Froment New Horizons in Catalysis. Proceedings of the 7th International Congress on Catalysis, Tokyo, June 30–July 4, 1980. Parts A and B edited by T. Seiyama and K. Tanabe Catalysis by Supported Complexes by Yu.I. Yermakov, B.N. Kuznetsov and V.A. Zakharov Physics of Solid Surfaces. Proceedings of a Symposium, Bechyˇne, September 29–October 3, 1980 edited by M. Lázniˆcka Adsorption at the Gas–Solid and Liquid–Solid Interface. Proceedings of an International Symposium, Aix-en-Provence, September 21–23, 1981 edited by J. Rouquerol and K.S.W. Sing Metal-Support and Metal-Additive Effects in Catalysis. Proceedings of an International Symposium, Ecully (Lyon), September 14–16, 1982 edited by B. Imelik, C. Naccache, G. Coudurier, H. Praliaud, P. Meriaudeau, P. Gallezot, G.A. Martin and J.C. Vedrine Metal Microstructures in Zeolites. Preparation - Properties - Applications. Proceedings of a Workshop, Bremen, September 22–24, 1982  edited by P.A. Jacobs, N.I. Jaeger, P. Jiru and G. Schulz-Ekloff Adsorption on Metal Surfaces. An Integrated Approach edited by J. Bénard

400 Volume 14

Volume 15 Volume 16

Volume 17

Volume 18

Volume 19

Volume 20

Volume 21

Volume 22 Volume 23 Volume 24

Volume 25

Volume 26

Volume 27 Volume 28

Volume 29 Volume 30

Vibrations at Surfaces. Proceedings of the Third International Conference, Asilomar, CA, September 1–4, 1982 edited by C.R. Brundle and H. Morawitz Heterogeneous Catalytic Reactions Involving Molecular Oxygen by G.I. Golodets Preparation of Catalysts III. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Third International Symposium, Louvain-la-Neuve, September 6–9, 1982 edited by G. Poncelet, P. Grange and P.A. Jacobs Spillover of Adsorbed Species. Proceedings of an International Symposium, Lyon-Villeurbanne, September 12–16, 1983 edited by G.M. Pajonk, S.J. Teichner and J.E. Germain Structure and Reactivity of Modified Zeolites. Proceedings of an International Conference, Prague, July 9–13, 1984  edited by P.A. Jacobs, N.I. Jaeger, P. Jiru , V.B. Kazansky and G. Schulz-Ekloff Catalysis on the Energy Scene. Proceedings of the 9th Canadian Symposium on Catalysis, Quebec, P.Q., September 30–October 3, 1984 edited by S. Kaliaguine and A. Mahay Catalysis by Acids and Bases. Proceedings of an International Symposium, Villeurbanne (Lyon), September 25–27, 1984 edited by B. Imelik, C. Naccache, G. Coudurier, Y. Ben Taarit and J.C. Vcdrinc Adsorption and Catalysis on Oxide Surfaces. Proceedings of a Symposium, Uxbridge, June 28–29, 1984 edited by M. Che and G.C. Bond Unsteady Processes in Catalytic Reactors by Yu.Sh. Matros Physics of Solid Surfaces 1984 edited by J. Koukal Zeolites: Synthesis, Structure, Technology and Application. Proceedings of an International Symposium, Portoroˇz-Portorose, September 3–8, 1984 edited by B. Drˇzaj, S. Hoˆcevar and S. Pejovnik Catalytic Polymerization of Olefins. Proceedings of the International Symposium on Future Aspects of Olefin Polymerization, Tokyo, July 4–6, 1985 edited by T. Keii and K. Soga Vibrations at Surfaces I985. Proceedings of the Fourth International Conference, Bowness-on-Windermere, September 15–19, 1985 edited by D. A. King, N.V. Richardson and S. Holloway Catalytic Hydrogenation edited by L. Cerven´y New Developments in Zeolite Science and Technology. Proceedings of the 7th International Zeolite Conference, Tokyo, August 17–22, 1986 edited by Y. Murakami, A. lijima and J.W. Ward Metal Clusters in Catalysis edited by B.C. Gates, L. Guczi and H. Kn¨ozinger Catalysis and Automotive Pollution Control. Proceedings of the First International Symposium, Brussels, September 8–11, 1986 edited by A. Crucq and A. Frennet

401 Volume 31

Volume 32 Volume 33 Volume 34

Volume 35 Volume 36

Volume 37

Volume 38

Volume 39

Volume 40

Volume 41

Volume 42 Volume 43 Volume 44

Volume 45 Volume 46

Volume 47 Volume 48

Preparation of Catalysts IV. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Fourth International Symposium, Louvain-la-Neuve, September 1–4, 1986 edited by B. Delmon, P. Grange, P.A. Jacobs and G. Poncelet Thin Metal Films and Gas Chemisorption edited by P. Wissmann Synthesis of High-silica Aluminosilicate Zeolites edited by P.A. Jacobs and J.A. Martens Catalyst Reactivation 1987. Proceedings of the 4th International Symposium, Antwerp, September 29–October 1, 1987 edited by B. Delmon and G.F. Froment Keynotes in Energy-Related Catalysis edited by S. Kaliaguine Methane Conversion. Proceedings of a Symposium on the Production of Fuels and Chemicals from Natural Gas, Auckland, April 27–30, 1987 edited by D.M. Bibby, C.D. Chang, R.F. Howe and S. Yurchak Innovation in Zeolite Materials Science. Proceedings of an International Symposium, Nieuwpoort, September 13–17, 1987 edited by P.J. Grobet, W.J. Mortier, E.F. Vansant and G. Schulz-Ekloff Catalysis l987. Proceedings of the 10th North American Meeting of the Catalysis Society, San Diego, CA, May 17–22, 1987 edited by J.W. Ward Characterization of Porous Solids. Proceedings of the IUPAC Symposium (COPS I), Bad Soden a. Ts., April 26–29, 1987 edited by K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral Physics of Solid Surfaces I987. Proceedings of the Fourth Symposium on Surface Physics, Bechyne Castle, September 7–11, 1987 edited by J. Koukal Heterogeneous Catalysis and Fine Chemicals. Proceedings of an International Symposium, Poitiers, March 15–17, 1988 edited by M. Guisnet, J. Barrault, C. Bouchoule, D. Duprez, C. Montassier and G. Pérot Laboratory Studies of Heterogeneous Catalytic Processes by E.G. Christoffel, revised and edited by Z. Paál Catalytic Processes under Unsteady-State Conditions by Yu. Sh. Matros Successful Design of Catalysts. Future Requirements and Development. Proceedings of the Worldwide Catalysis Seminars, July, 1988, on the Occasion of the 30th Anniversary of the Catalysis Society of Japan edited by T. Inui Transition Metal Oxides. Surface Chemistry and Catalysis by H.H. Kung Zeolites as Catalysts, Sorbents and Detergent Builders. Applications and Innovations. Proceedings of an International Symposium, W¨urzburg, September 4–8, 1988 edited by H.G. Karge and J. Weitkamp Photochemistry on Solid Surfaces edited by M. Anpo and T. Matsuura Structure and Reactivity of Surfaces. Proceedings of a European Conference, Trieste, September 13–16, 1988 edited by C. Morterra, A. Zecchina and G. Costa

402 Volume 49

Volume 50

Volume 51 Volume 52

Volume 53

Volume 54 Volume 55

Volume 56

Volume 57A

Volume 57B

Volume 58 Volume 59

Volume 60

Volume 61

Volume 62

Volume 63

Zeolites: Facts, Figures, Future. Proceedings of the 8th International Zeolite Conference, Amsterdam, July 10–14, 1989. Parts A and B edited by P.A. Jacobs and R.A. van Santen Hydrotreating Catalysts. Preparation, Characterization and Performance. Proceedings of the Annual International AlChE Meeting, Washington, DC, November 27–December 2, 1988 edited by M.L. Occetli and R.G. Anthony New Solid Acids and Bases. Their Catalytic Properties by K. Tanabe, M. Misono.Y. Ono and H. Hattori Recent Advances in Zeolite Science. Proceedings of the 1989 Meeting of the British Zeolite Association, Cambridge, April 17–19, 1989 edited by J. Klinowsky and P.J. Barrie Catalyst in Petroleum Refining 1989. Proceedings of the First International Conference on Catalysts in Petroleum Refining, Kuwait, March 5–8, 1989 edited by D.L. Trimm, S. Akashah, M. Absi-Halabi and A. Bishara Future Opportunities in Catalytic and Separation Technology edited by M. Misono, Y. Moro-oka and S. Kimura New Developments in Selective Oxidation. Proceedings of an International Symposium, Rimini, Italy, September 18–22, 1989 edited by G. Centi and F. Trifiro Olefin Polymerization Catalysts. Proceedings of the International Symposium on Recent Developments in Olefin Polymerization Catalysts, Tokyo, October 23–25, 1989 edited by T. Keii and K. Soga Spectroscopic Analysis of Heterogeneous Catalysts. Part A: Methods of Surface Analysis edited by J.L.G. Fierro Spectroscopic Analysis of Heterogeneous Catalysts. Part B: Chemisorption of Probe Molecules edited by J.L.G. Fierro Introduction to Zeolite Science and Practice edited by H. van Bekkum, E.M. Flanigen and J.C. Jansen Heterogeneous Catalysis and Fine Chemicals II. Proceedings of the 2nd International Symposium, Poitiers, October 2–6, 1990 edited by M. Guisnet, J. Barrault, C. Bouchoule, D. Duprez, G. Pérot, R. Maurel and C. Montassier Chemistry of Microporous Crystals. Proceedings of the International Symposium on Chemistry of Microporous Crystals, Tokyo, June 26–29, 1990 edited by T. Inui, S. Namba and T. Tatsumi Natural Gas Conversion. Proceedings of the Symposium on Natural Gas Conversion, Oslo, August 12–17, 1990 edited by A. Holmen, K.-J. Jens and S. Kolboe Characterization of Porous Solids II. Proceedings of the IUPAC Symposium (COPS II), Alicante, May 6–9, 1990 edited by F. Rodríguez-Reinoso, J. Rouquerol, K.S.W. Sing and K.K. Unger Preparation of Catalysts V. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Fifth International Symposium, Louvain-la-Neuve, September 3–6, 1990 edited by G. Ponceiet, P.A. Jacobs, P. Grange and B. Delmon

403 Volume 64 Volume 65

Volume 66

Volume 67

Volume 68

Volume 69

Volume 70

Volume 71

Volume 72

Volume 73

Volume 74

Volume 75

Volume 76 Volume 77

Volume 78

New Trends in CO Activation edited by L. Guczi Catalysis and Adsorption by Zeolites. Proceedings of ZEOCAT 90, Leípzig, August 20–23, 1990 ¨ edited by G. Ohlmann, H. Pfeifer and R. Fricke Dioxygen Activation and Homogeneous Catalytic Oxidation. Proceedings of the Fourth International Symposium on Dioxygen Activation and Homogeneous Catalytic Oxidation, Balatonf¨ured, September 10–14, 1990 edited by L.I. Simándi Structure-Activity and Selectivity Relationships in Heterogeneous Catalysis. Proceedings of the ACS Symposium on Structure-Activity Relationships in Heterogeneous Catalysis, Boston, MA, April 22–27, 1990 edited by R.K. Grasselli and A.W. Sleight Catalyst Deactivation 1991. Proceedings of the Fifth International Symposium, Evanston, IL, June 24–26, 1991 edited by C.H. Bartholomew and J.B. Butt Zeolite Chemistry and Catalysis. Proceedings of an International Symposium, Prague, Czechoslovakia, September 8–13, 1991 edited by P.A. Jacobs, N.I. Jaeger, L. Kubelková and B. Wichterlová Poisoning and Promotion in Catalysis based on Surface Science Concepts and Experiments by M. Kiskinova Catalysis and Automotive Pollution Control II. Proceedings of the 2nd International Symposium (CAPoC 2), Brussels, Belgium, September 10–13, 1990 edited by A. Crucq New Developments in Selective Oxidation by Heterogeneous Catalysis. Proceedings of the 3rd European Workshop Meeting on New Developments in Selective Oxidation by Heterogeneous Catalysis, Louvain-la-Neuve, Belgium, April 8–10, 1991 edited by P. Ruiz and B. Delmon Progress in Catalysis. Proceedings of the 12th Canadian Symposium on Catalysis, Banff, Alberta, Canada, May 25–28, 1992 edited by K.J. Smith and E.C. Sanford Angle-Resolved Photoemission. Theory and Current Applications edited by S.D. Kevan Mew Frontiers in Catalysis, Parts A–C. Proceedings of the 10th International Congress on Catalysis, Budapest, Hungary, 19–24 July, 1992 edited by L. Guczi, F. Solymosi and P. Tétényi Fluid Catalytic Cracking: Science and Technology edited by J.S. Magee and M.M. Mitchell, Jr. New Aspects of Spillover Effect in Catalysis. For Development of Highly Active Catalysts. Proceedings of the Third International Conference on Spillover, Kyoto, Japan, August 17–20, 1993 edited by T. Inui, K. Fujimoto, T. Uchijima and M. Masai Heterogeneous Catalysis and Fine Chemicals III. Proceedings of the 3rd International Symposium, Poitiers, April 5–8, 1993 edited by M. Guisnet, J. Barbier, J. Barrault, C. Bouchoule, D. Duprez, G. Pérot and C. Montassier

404 Volume 79

Volume 80

Volume 81

Volume 82

Volume 83

Volume 84

Volume 85 Volume 86 Volume 87

Volume 88

Volume 89

Volume 90

Volume 91

Volume 92

Volume 93

Catalysis: An Integrated Approach to Homogeneous, Heterogeneous and Industrial Catalysis edited by J.A. Moulijn, P.W.N.M. van Leeuwen and R.A. van Santen Fundamentals of Adsorption. Proceedings of the Fourth International Conference on Fundamentals of Adsorption, Kyoto, Japan, May 17–22, 1992 edited by H. Suzuki Natural Gas Conversion II. Proceedings of the Third Natural Gas Conversion Symposium, Sydney, July 4–9, 1993 edited by H.E. Curry-Hyde and R.F. Howe New Developments in Selective Oxidation II. Proceedings of the Second World Congress and Fourth European Workshop Meeting, Benalmádena, Spain, September 20–24, 1993 edited by V. Cortés Corberán and S. Vic Bellón Zeolites and Microporous Crystals. Proceedings of the International Symposium on Zeolites and Microporous Crystals, Nagoya, Japan, August 22–25, 1993 edited by T. Hattori and T. Yashima Zeolites and Related Microporous Materials: State of the Art 1994. Proceedings of the 10th International Zeolite Conference, Garmisch-Partenkirchen, Germany, July 17–22, 1994 edited by J. Weitkamp, H.G. Karge, H. Pfeifer and W. H¨olderich Advanced Zeolite Science and Applications edited by J.C. Jansen, M. St¨ocker, H.G. Karge and J. Weitkamp Oscillating Heterogeneous Catalytic Systems by M.M. Slinko and N.I. Jaeger Characterization of Porous Solids III. Proceedings of the IUPAC Symposium (COPS III), Marseille, France, May 9–12, 1993 edited by J. Rouquerol, F. Rodríguez-Reinoso, K.S.W. Sing and K.K. Unger Catalyst Deactivation 1994. Proceedings of the 6th International Symposium, Ostend, Belgium, October 3–5, 1994 edited by B. Delmon and G.F. Froment Catalyst Design for Tailor-made Polyolefins. Proceedings of the International Symposium on Catalyst Design for Tailor-made Polyolefins, Kanazawa, Japan, March 10–12, 1994 edited by K. Soga and M. Terano Acid-Base Catalysis II. Proceedings of the International Symposium on Acid-Base Catalysis II, Sapporo, Japan, December 2–4, 1993 edited by H. Hattori, M. Misono and Y. Ono Preparation of Catalysts VI. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Sixth International Symposium, Louvain-La-Neuve, September 5–8, 1994 edited by G. Poncelet, J. Martens, B. Delmon, P.A. Jacobs and P. Grange Science and Technology in Catalysis I994. Proceedings of the Second Tokyo Conference on Advanced Catalytic Science and Technology, Tokyo, August 21–26, 1994 edited by Y. Izumi, H. Arai and M. Iwamoto Characterization and Chemical Modification of the Silica Surface by E.F. Vansant, P. Van Der Voort and K.C. Vrancken

405 Volume 94

Volume 95 Volume 96

Volume 97

Volume 98

Volume 99 Volume 100

Volume 101

Volume 102

Volume 103

Volume 104

Volume 105

Volume 106

Volume 107

Volume 108

Catalysis by Microporous Materials. Proceedings of ZEOCAT’95, Szombathely, Hungary, July 9–13, 1995 edited by H.K. Beyer, H.G. Karge, I. Kiricsi and J.B. Nagy Catalysis by Metals and Alloys by V. Ponec and G.C. Bond Catalysis and Automotive Pollution Control III. Proceedings of the Third International Symposium (CAPoC3), Brussels, Belgium, April 20–22, 1994 edited by A. Frennet and J.-M. Bastin Zeolites: A Refined Tool for Designing Catalytic Sites. Proceedings of the International Symposium, Québec, Canada, October 15–20, 1995 edited by L. Bonneviot and S. Kaliaguine Zeolite Science 1994: Recent Progress and Discussions. Supplementary Materials to the 10th International Zeolite Conference, Garmisch-Partenkirchen, Germany, July 17–22, 1994 edited by H.G. Karge and J. Weitkamp Adsorption on New and Modified Inorganic Sorbents edited by A. Dabrowski and V.A. Tertykh Catalysts in Petroleum Refining and Petrochemical Industries 1995. Proceedings of the 2nd International Conference on Catalysts in Petroleum Refining and Petrochemical Industries, Kuwait, April 22–26, 1995 edited by M. Absi-Halabi, J. Beshara, H. Qabazard and A. Stanislaus 11th International Congress on Catalysis - 40th Anniversary. Proceedings of the 11th ICC, Baltimore, MD, USA, June 30–July 5, 1996 edited by J.W. Hightower.W.N. Delgass, E. Iglesia and A.T. Bell Recent Advances and New Horizons in Zeolite Science and Technology edited by H. Chon, S.I. Woo and S.-E. Park Semiconductor Nanoclusters - Physical, Chemical, and Catalytic Aspects edited by P.V. Kamat and D. Meisel Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces edited by W. Rudzinski, W.A. Steele and G. Zgrablich ´ Progress in Zeolite and Microporous Materials Proceedings of the 11th International Zeolite Conference, Seoul, Korea, August 12–17, 1996 edited by H. Chon, S.-K. Ihm and Y.S. Uh Hydrotreatment and Hydrocracking of Oil Fractions Proceedings of the 1st International Symposium/6th European Workshop, Oostende, Belgium, February 17–19, 1997 edited by G.F. Froment, B. Delmon and P. Grange Natural Gas Conversion IV Proceedings of the 4th International Natural Gas Conversion Symposium, Kruger Park, South Africa, November 19–23, 1995 edited by M. de Pontes, R.L Espinoza, C.P. Nicolaides, J.H. Scholtz and M.S. Scurrell Heterogeneous Catalysis and Fine Chemicals IV Proceedings of the 4th International Symposium on Heterogeneous Catalysis and Fine Chemicals, Basel, Switzerland, September 8–12, 1996 edited by H.U. Blaser, A. Baiker and R. Prins

406 Volume 109

Volume 110

Volume 111

Volume 112

Volume 113

Volume 114

Volume 115

Volume 116

Volume 117

Volume 118

Volume 119

Dynamics of Surfaces and Reaction Kinetics in Heterogeneous Catalysis. Proceedings of the International Symposium, Antwerp, Belgium, September 15–17, 1997 edited by G.F. Froment and K.C. Waugh Third World Congress on Oxidation Catalysis. Proceedings of the Third World Congress on Oxidation Catalysis, San Diego, CA, U.S.A., 21–26 September 1997 edited by R.K. Grasselli, S.T. Oyama, A.M. Gaffney and J.E. Lyons Catalyst Deactivation 1997. Proceedings of the 7th International Symposium, Cancun, Mexico, October 5–8, 1997 edited by C.H. Bartholomew and G.A. Fuentes Spillover and Migration of Surface Species on Catalysts. Proceedings of the 4th International Conference on Spillover, Dalian, China, September 15–18, 1997 edited by Can Li and Qin Xin Recent Advances in Basic and Applied Aspects of Industrial Catalysis. Proceedings of the 13th National Symposium and Silver Jubilee Symposium of Catalysis of India, Dehradun, India, April 2–4, 1997 edited by T.S.R. Prasada Rao and G. Murali Dhar Advances in Chemical Conversions for Mitigating Carbon Dioxide. Proceedings of the 4th International Conference on Carbon Dioxide Utilization, Kyoto, Japan, September 7–11, 1997 edited by T. Inui, M. Anpo, K. Izui, S. Yanagida and T. Yamaguchi Methods for Monitoring and Diagnosing the Efficiency of Catalytic Converters. A patent-oriented survey by M. Sideris Catalysis and Automotive Pollution Control IV. Proceedings of the 4th International Symposium (CAPoC4), Brussels, Belgium, April 9–11, 1997 edited by N. Kruse, A. Frennet and J.-M. Bastin Mesoporous Molecular Sieves 1998 Proceedings of the 1st International Symposium, Baltimore, MD, U.S.A., July 10–12, 1998 edited by L. Bonneviot, F. Béland, C. Danumah, S. Giasson and S. Kaliaguine Preparation of Catalysts VII Proceedings of the 7th International Symposium on Scientific Bases for the Preparation of Heterogeneous Catalysts, Louvain-la-Neuve, Belgium, September 1–4, 1998 edited by B. Delmon, PA. Jacobs, R. Maggi, J.A. Martens, P. Grange and G. Poncelet Natural Gas Conversion V Proceedings of the 5th International Gas Conversion Symposium, Giardini-Naxos, Taormina, Italy, September 20–25, 1998 edited by A. Parmaliana, D. Sanfilippo, F. Frusteri, A. Vaccari and F. Arena

407 Volume 120A

Volume 120B

Volume 121

Volume 122

Volume 123

Volume 124 Volume 125

Volume 126

Volume 127

Volume 128

Volume 129

Volume 130

Volume 131

Adsorption and its Applications in Industry and Environmental Protection. Vol I: Applications in Industry edited by A. D¸abrowski Adsorption and its Applications in Industry and Environmental Protection. Vol II: Applications in Environmental Protection edited by A. D¸abrowski Science and Technology in Catalysis 1998 Proceedings of the Third Tokyo Conference in Advanced Catalytic Science and Technology, Tokyo, July 19–24, 1998 edited by H. Hattori and K. Otsuka Reaction Kinetics and the Development of Catalytic Processes Proceedings of the International Symposium, Brugge, Belgium, April 19–21, 1999 edited by G.F. Froment and K.C. Waugh Catalysis: An Integrated Approach Second, Revised and Enlarged Edition edited by R.A. van Santen, P.W.N.M. van Leeuwen, J.A. Moulijn and B.A. Averill Experiments in Catalytic Reaction Engineering by J.M. Berty Porous Materials in Environmentally Friendly Processes Proceedings of the 1st International FEZA Conference, Eger, Hungary, September 1–4, 1999 edited by I. Kiricsi, G. Pál-Borbély, J.B. Nagy and H.G. Karge Catalyst Deactivation 1999 Proceedings of the 8th International Symposium, Brugge, Belgium, October 10–13, 1999 edited by B. Delmon and G.F. Froment Hydrotreatment and Hydrocracking of Oil Fractions Proceedings of the 2nd International Symposium/7th European Workshop, Antwerpen, Belgium, November 14–17, 1999 edited by B. Delmon, G.F. Froment and P. Grange Characterisation of Porous Solids V Proceedings of the 5th International Symposium on the Characterisation of Porous Solids (COPS-V), Heidelberg, Germany, May 30–June 2, 1999 edited by K.K. Unger, G. Kreysa and J.P. Baselt Nanoporous Materials II Proceedings of the 2nd Conference on Access in Nanoporous Materials, Banff, Alberta, Canada, May 25–30, 2000 edited by A. Sayari, M. Jaroniec and T.J. Pinnavaia 12th International Congress on Catalysis Proceedings of the 12th ICC, Granada, Spain, July 9–14, 2000 edited by A. Corma, F.V. Melo, S. Mendioroz and J.L.G. Fierro Catalytic Polymerization of Cycloolefins Ionic, Ziegler-Natta and Ring-Opening Metathesis Polymerization By V. Dragutan and R. Streck

408 Volume 132

Volume 133

Volume 134

Volume 135

Volume 136

Volume 137

Volume 138 Volume 139

Volume 140

Volume 141

Volume 142

Volume 143

Volume 144

Proceedings of the International Conference on Colloid and Surface Science, Tokyo, Japan, November 5–8, 2000 25th Anniversary of the Division of Colloid and Surface Chemistry, The Chemical Society of Japan edited by Y. Iwasawa, N. Oyama and H. Kunieda Reaction Kinetics and the Development and Operation of Catalytic Processes Proceedings of the 3rd International Symposium, Oostende, Belgium, April 22–25, 2001 edited by G.F. Froment and K.C. Waugh Fluid Catalytic Cracking V Materials and Technological Innovations edited by M.L. Occelli and P. O’Connor Zeolites and Mesoporous Materials at the Dawn of the 21st Century. Proceedings of the 13th International Zeolite Conference, Montpellier, France, 8–13 July 2001 edited by A. Galameau, F. di Renso, F. Fajula ans J. Vedrine Natural Gas Conversion VI Proceedings of the 6th Natural Gas Conversion Symposium, June 17–22, 2001, Alaska, USA. edited by J.J. Spivey, E. Iglesia and T.H. Fleisch Introduction to Zeolite Science and Practice. 2nd completely revised and expanded edition edited by H. van Bekkum, E.M. Flanigen, P.A. Jacobs and J.C. Jansen Spillover and Mobility of Species on Solid Surfaces edited by A. Guerrero-Ruiz and I. Rodríquez-Ramos Catalyst Deactivation 2001 Proceedings of the 9th International Symposium, Lexington, KY, USA, October 2001 edited by J.J. Spivey, G.W. Roberts and B.H. Davis Oxide-based Systems at the Crossroads of Chemistry. Second International Workshop, October 8–11, 2000, Como, Italy. Edited by A. Gamba, C. Colella and S. Coluccia Nanoporous Materials III Proceedings of the 3rd International Symposium on Nanoporous Materials, Ottawa, Ontario, Canada, June 12–15, 2002 edited by A. Sayari and M. Jaroniec Impact of Zeolites and Other Porous Materials on the New Technologies at the Beginning of the New Millennium Proceedings of the 2nd International FEZA (Federation of the European Zeolite Associations) Conference, Taormina, Italy, September 1–5, 2002 edited by R. Aiello, G. Giordano and F. Testa Scientific Bases for the Preparation of Heterogeneous Catalysts Proceedings of the 8th International Symposium, Louvain-la-Neuve, Leuven, Belgium, September 9–12, 2002 edited by E. Gaigneaux, D.E. De Vos, P. Grange, P.A. Jacobs, J.A. Martens, P. Ruiz and G. Poncelet Characterization of Porous Solids VI Proceedings of the 6th International Symposium on the Characterization of Porous Solids (COPS-VI), Alicante, Spain, May 8–11, 2002 edited by F. Rodríguez-Reinoso, B. McEnaney, J. Rouquerol and K. Unger

409 Volume 145

Volume 146

Volume 147

Volume 148

Volume 149

Volume 150

Volume 151

Volume 152 Volume 153

Volume 154

Volume 155

Volume 156 Volume 157

Volume 158

Science and Technology in Catalysis 2002 Proceedings of the Fourth Tokyo Conference on Advanced Catalytic Science and Technology, Tokyo, July 14–19, 2002 edited by M. Anpo, M. Onaka and H. Yamashita Nanotechnology in Mesostructured Materials Proceedings of the 3rd International Mesostructured Materials Symposium, Jeju, Korea, July 8–11, 2002 edited by Sang-Eon Park, Ryong Ryoo, Wha-Seung Ahn, Chul Wee Lee and Jong-San Chang Natural Gas Conversion VII Proceedings of the 7th Natural Gas Conversion Symposium, Dalian, China, June 6–10, 2004 edited by X. Bao and Y. Xu Mesoporous Crystals and Related Nano-Structured Materials Proceedings of the Meeting on Mesoporous Crystals and Related Nano-Structured Materials, Stockholm, Sweden, 1–5 June, 2004 edited by O. Terasaki Fluid Catalytic Cracking VI: Preparation and Characterization of Catalysts Proceedings of the 6th International Symposium on Advances in Fluid Cracking Catalysts (FCCs), New York, September 7–11, 2003 edited by M. Occelli Coal and Coal-Related Compounds Structures, Reactivity and Catalytic Reactions edited by T. Kabe, A. Ishihara, E.W. Qian, I.P. Sutrisna and Y. Kabe Petroleum Biotechnology Developments and Perspectives edited by R. Vazquez-Duhalt and R. Quintero-Ramirez Fisher-Tropsch technology edited by A.P. Steynberg and M.E. Dry Carbon Dioxide Utilization for Global Sustainability Proceedings of the 7th International Conference on Carbon Dioxide Utilization (ICCDU VII), October 12–16, 2003 Seoul, Korea edited by S.-E. Park, J.-S. Chang and K.-W. Lee Recent Advances in the Science and Technology of Zeolites and Related Materials Proceedings of the 14th International Zeolite Conference, Cape Town, South Africa, 25–30th April 2004 edited by E. van Steen, L.H. Callanan and M. Claeys Oxide Based Materials New Sources, Novel Phases, New Applications edited by A. Gamba, C. Colella and S. Coluccia Nanoporous Materials IV edited by A. Sayari and M. Jaroniec Zeolites and Ordered Mesoporous Materials Progress and Prospects ˇ edited by J. Cejka and H. van Bekkum Molecular Sieves: From Basic Research to Industrial Applications Proceedings of the 3rd International Zeolite Symposium (3rd FEZA), Prague, Czech Republic, August 23–26, 2005 ˇ ˇ edited by J. Cejka, N. Zilková and P. Nachtigall

410 Volume 159

Volume 160

Volume 161

Volume 162

Volume 163 Volume 164 Volume 165

Volume 166

Volume 167

Volume 168 Volume 169 Volume 170A

Volume 170B

New Developments and Application in Chemical Reaction Engineering Proceedings of the 4th Asia-Pacific Chemical Reaction Engineering Symposium (APCRE’05), Syeongju, Korea, June 12–15, 2005 edited by H.-K. Rhee, I.-S. Nam and J.M. Park Characterization of Porous Solids VII Proceedings of the 7th International Symposium on the Characterization of Porous Solids (COPS-VII), Aix-en-Provence, France, May 26–28, 2005 edited by Ph.L. Llewellyn, F. Rodríquez-Reinoso, J. Rouqerol and N. Seaton Progress in Olefin Polymerization Catalysts and Polyolefin Materials Proceedings of the First Asian Polyolefin Workshop, Nara, Japan, December 7–9, 2005 edited by T. Shiono, K. Nomura and M. Terano Scientific Bases for the Preparation of Heterogeneous Catalysts Proceedings of the 9th International Symposium, Louvain-la-Neuve, Belgium, September 10–14, 2006 edited by E.M. Gaigneaux, M. Devillers, D.E. De Vos, S. Hermans, P.A. Jacobs, J.A. Martens and P. Ruiz Fischer-Tropsch Synthesis, Catalysts and Catalysis edited by B.H. Davis and M.L. Occelli Biocatalysis in Oil Refining. edited by M.M. Ramirez-Corredores and Abhijeet P. Borole Recent Progress in Mesostructured Materials. Proceedings of the 5th International Mesostructured Materials Symposium (IMMS2006), Shanghai, P.R China, August 5–7, 2006 edited by D. Zhao, S. Qiu, Y. Tang and C. Yu Fluid Catalytic cracking VII: Materials, Methods and process Innovations. Studies in Surface Science and Catalysis edited by M.L. Occelli Natural Gas Conversion VIII. Proceedings of the 8th Natural Gas Conversion Symposium, Natal, Brazil, May 27–31, 2007 Edited by F.B. Noronha, M. Schmal and E.F. Sousa-Aguiar Introduction to Zeolite Molecular Sieves ˇ edited by Jiˇrí Cejka and Avelino Corma Catalysts for Upgrading Heavy Petroleum Feeds edited by Edward Furimsky From Zeolites to Porous MOF Materials – The 40th Anniversary of International Zeolite Conference. Part A: Proceedings of the 15th International Zeolite Conference, Beijing, P. R. China, 12–17th August 2007 edited by Ruren Xu, Zi Gao, Jiesheng Chen and Wenfu Yan From Zeolites to Porous MOF Materials – The 40th Anniversary of International Zeolite Conference. Part B: Proceedings of the 15th International Zeolite Conference, Beijing, P. R. China, 12–17th August 2007 edited by Ruren Xu, Zi Gao, Jiesheng Chen and Wenfu Yan