Transparent Conductive ZnO

Springer Series in materials science 104 Springer Series in materials science Editors: R. Hull R. M. Osgood, Jr. ...

Author: Klaus Ellmer | Andreas Klein | Bernd Rech

49 downloads 1569 Views 16MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Springer Series in

materials science

104

Springer Series in

materials science Editors: R. Hull

R. M. Osgood, Jr.

J. Parisi

H. Warlimont

The Springer Series in Materials Science covers the complete spectrum of materials physics, including fundamental principles, physical properties, materials theory and design. Recognizing the increasing importance of materials science in future device technologies, the book titles in this series ref lect the state-of-the-art in understanding and controlling the structure and properties of all important classes of materials. 88 Introduction to Wave Scattering, Localization and Mesoscopic Phenomena By P. Sheng

98 Physics of Negative Refraction and Negative Index Materials Optical and Electronic Aspects and Diversiﬁed Approaches Editors: C.M. Krowne and Y. Zhang

89 Magneto-Science Magnetic Field Effects on Materials: Fundamentals and Applications Editors: M. Yamaguchi and Y. Tanimoto

99 Self-Organized Morphology in Nanostructured Materials Editors: K. Al-Shamery and J. Parisi

90 Internal Friction in Metallic Materials A Reference Book By M.S. Blanter, I.S. Golovin, H. Neuh¨auser, and H.-R. Sinning

100 Self Healing Materials An Alternative Approach to 20 Centuries of Materials Science Editor: S. van der Zwaag

91 Time-dependent Mechanical Properties of Solid Bodies By W. Gr¨afe

101 New Organic Nanostructures for Next Generation Devices Editors: K. Al-Shamery, H.-G. Rubahn, and H. Sitter

92 Solder Joint Technology Materials, Properties, and Reliability By K.-N. Tu 93 Materials for Tomorrow Theory, Experiments and Modelling Editors: S. Gemming, M. Schreiber and J.-B. Suck 94 Magnetic Nanostructures Editors: B. Aktas, L. Tagirov, and F. Mikailov 95 Nanocrystals and Their Mesoscopic Organization By C.N.R. Rao, P.J. Thomas and G.U. Kulkarni 96 GaN Electronics By R. Quay 97 Multifunctional Barriers for Flexible Structure Textile, Leather and Paper Editors: S. Duquesne, C. Magniez, and G. Camino

102 Photonic Crystal Fibers Properties and Applications By F. Poli, A. Cucinotta, and S. Selleri 103 Polarons in Advanced Materials Editor: A.S. Alexandrov 104 Transparent Conductive Zinc Oxide Basics and Applications in Thin Film Solar Cells Editors: K. Ellmer, A. Klein, and B. Rech 105 Dilute III–V Nitride Semiconductors and Novel Dilute Nitride Material Systems Physics and Technology Editor: A. Erol 106 Into The Nano Era Moore’s Law Beyond Planar Silicon CMOS Editor: H. Huff

Volumes 40–87 are listed at the end of the book.

Klaus Ellmer Andreas Klein Bernd Rech Editors

Transparent Conductive Zinc Oxide Basics and Applications in Thin Film Solar Cells

With 270 Figures, 5 in color and 50 Tables

123

Dr. Klaus Ellmer Hahn-Meitner-Institut Berlin GmbH, Abteilung Solare Energetik (SE5) Glienicker Str. 100, 14109 Berlin, Germany E-mail: [email protected]

Dr. Andreas Klein Technische Universit¨at Darmstadt, FB11 Materialwissenschaften Petersenstr. 23, 64287 Darmstadt, Germany E-mail: [email protected]

Professor Dr. Bernd Rech Hahn-Meitner-Institut Berlin GmbH, Abteilung Silizium-Photovltaik Kekulestr. 5, 12489 Berlin, Germany E-mail: [email protected]

Series Editors:

Professor Robert Hull

Professor Jürgen Parisi

University of Virginia Dept. of Materials Science and Engineering Thornton Hall Charlottesville, VA 22903-2442, USA

Universit¨at Oldenburg, Fachbereich Physik Abt. Energie- und Halbleiterforschung Carl-von-Ossietzky-Strasse 9–11 26129 Oldenburg, Germany

Professor R. M. Osgood, Jr.

Professor Hans Warlimont

Microelectronics Science Laboratory Department of Electrical Engineering Columbia University Seeley W. Mudd Building New York, NY 10027, USA

Institut f¨ur Festk¨orperund Werkstofforschung, Helmholtzstrasse 20 01069 Dresden, Germany

ISSN 0933-033X ISBN 978-3-540-73611-0 Springer Berlin Heidelberg New York Library of Congress Control Number: 2007933696 All rights reserved. No part of this book may be reproduced in any form, by photostat, microﬁlm, retrieval system, or any other means, without the written permission of Kodansha Ltd. (except in the case of brief quotation for criticism or review.) This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media. springer.com © Springer-Verlag Berlin Heidelberg 2008 The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Data prepared by SPi using a Springer LATEX macro package Cover concept: eStudio Calamar Steinen Cover production: WMX Design GmbH, Heidelberg Printed on acid-free paper

SPIN: 11935865

57/3180/SPi

543210

Preface

Zinc oxide is a widely applied material in industry. It is produced in hundreds of thousands of tons for paints (chinese white), additive for rubber and plastics, catalysts, pharmaceuticals and cosmetics (sun creams), or as coating material for paper. In electronics industry ZnO is used in nickelor manganese–zinc ferrites, as ingredient of phosphors, in surface-acoustic wave ﬁlters and as a transparent electrode. Zinc oxide is a wide bandgap compound semiconductor (Eg = 3.2 eV) which has been investigated as an electronic material for many decades, starting in the 1930s. It belongs to the class of transparent conducting oxides (TCO). The other important oxides are indium and tin oxide. Recently, fundamental and applied research on zinc oxide experienced a renaissance due to the prospective use of zinc oxide as an optoelectronic material for blue and UV lasers. Moreover, thin ZnO ﬁlms are important components in most thin ﬁlm solar cells. The cost-eﬀective large-scale production of these ﬁlms on the one hand and the development of ZnO ﬁlms with improved properties on the other hand are key challenges in the ﬁeld of production and R&D in photovoltaics. Renewable energies are one or even the only answer to provide the world energy demand on the long term in a sustainable way. Moreover, the world is facing climate changes due to green house gas emissions, and the limitation of these emissions is one of the key challenges of the global society. Solar energy bears the largest potential of all renewable energy sources, however, it is still rather expensive for many large-scale applications today. The direct conversion of sunlight into electricity by photovoltaic (PV) solar modules has grown by 20–40% per annum during the last decade and has emerged to a billion Euro market. Signiﬁcant cost reductions have been achieved and there is a huge potential to bring the costs down further. Currently, PV module production is dominated by crystalline silicon solar cells, which are based on Si wafers with a typical thickness of around 150–300 µm. However, the production of the comparatively thick silicon wafers involves high process temperatures and very pure silicon as an expensive feedstock material, partly limiting the potential for cost reductions. Thin-ﬁlm solar cells require only a few micrometers of ﬁlm thickness to absorb most of the sunlight and thus bear a great potential for signiﬁcantly reducing the cost of photovoltaic energy conversion due to low material consumption, simple production techniques and

VI

Preface

high productivity by depositing on large areas. In addition, low-temperature processing and low material consumption save energy during production leading to short energy pay-back times. Thin ﬁlms of ZnO play an essential role in most thin ﬁlm solar cells being produced today. ZnO ﬁlms serve as transparent and conductive front contact, provide additional optical functions like light scattering and subsequent light trapping or enhance the reﬂection at the back contact. In heterojunction solar cells based on chalcopyrite absorber layers, ZnO is an inherent part of the p/njunction. In general, ZnO ﬁlms with improved optical, electronic or structural properties promise higher conversion eﬃciencies while the development of process technologies providing an optimum ﬁlm quality on large areas at high growth rates are an essential prerequisite to meet the cost targets in production. Examples of long-term scientiﬁc challenges are the development of adapted nano-structured ZnO ﬁlms, like e.g. ZnO nanorods, or p-type ZnO ﬁlms which may both open up new possibilities for designing future thin ﬁlm solar cells. This book is devoted to the properties, preparation and applications of zinc oxide (ZnO) as an transparent electrode material. It focuses on ZnO for thin ﬁlm solar cell applications and hopefully inspires also readers from related ﬁelds. The book is structured into three parts to serve both as an overview as well as a data collection for students, engineers and scientists. The ﬁrst part, Chaps. 1–4, provide an overview of the application and fundamental material properties of ZnO ﬁlms and their surface and interfaces properties. Chaps. 5–7 review thin ﬁlm deposition techniques applied for ZnO preparation on lab scale but also for large area production. Finally, Chaps. 8 and 9 are devoted to applications of ZnO in silicon- and chalcopyrite-based thin ﬁlm solar cells, respectively. One should note that the application of CVD grown ZnO in silicon thin ﬁlm cells is discussed earlier in Chap. 6. The idea to write this book evolved during a research project on zinc oxide in thin ﬁlm solar cells, initiated by the German association “Forschungsverbund Sonnenenergie” and ﬁnanced by the German Ministry of Education and Research which are gratefully acknowledged. Last, but not least, we thank all our colleagues who contributed with questions, discussions and data to this book. Berlin, Darmstadt August 2007

Klaus Ellmer Andreas Klein Bernd Rech

Contributors

Carsten Bundesmann Leibniz-Institut f¨ ur Oberﬂ¨ achenmodiﬁzierung e.V. Permoserstraße 15, Leipzig 04318, Germany Carsten.Bundesmann@ iom-leipzig.de Klaus Ellmer Hahn-Meitner-Institut Berlin GmbH Solar Energy Research Glienicker Str. 100 14109 Berlin, Germany [email protected] Sylvie Fa¨ y University of Neuchatel Institute of Microtechnology (IMT) Rue A.-L.-Breguet 2 2000 Neuchatel, Switzerland [email protected]

Reiner Klenk Hahn-Meitner-Institut Berlin GmbH Solar Energy Research Glienicker Str. 100 14109 Berlin, Germany [email protected] Michael Lorenz Universit¨ at Leipzig, Institut f¨ ur Experimentelle Physik II, Linn´estr. 5, 04103 Leipzig, Germany [email protected] Joachim M¨ uller Forschungszentrum J¨ ulich GmbH Institute of Energy Research and Photovoltaics 52425 J¨ ulich, Germany joachim l [email protected]

J¨ urgen H¨ upkes Forschungszentrum J¨ ulich GmbH Institute of Energy Research and Photovoltaics 52425 J¨ ulich, Germany [email protected]

Bernd Rech Forschungszentrum J¨ ulich GmbH Institute of Energy Research and Photovoltaics 52425 J¨ ulich, Germany [email protected]

Andreas Klein University of Technology Darmstadt Institute of Material Science Petersenstr. 23, 64287 Darmstadt Germany [email protected]

Frank S¨ auberlich University of Technology Darmstadt Institute of Material Science Petersenstr. 23, 64287 Darmstadt Germany [email protected]

VIII

Contributors

R¨ udiger Schmidt-Grund Universit¨ at Leipzig, Institut f¨ ur Experimentelle Physik II, Linn´estr. 5, 04103 Leipzig, Germany [email protected] Mathias Schubert Department of Electrical Engineering, Nebraska Center for Materials and Nanoscience University of Nebraska-Lincoln Lincoln, NE 68588-0511, USA [email protected]

Arvind Shah University of Neuchatel Institute of Microtechnology (IMT) Rue A.-L.-Breguet 2 2000 Neuchatel, Switzerland [email protected] Bernd Szyszka Fraunhofer Institute for Surface Engineering and Thin Films (IST) Bienroder Weg 54 E 38108 Braunschweig, Germany [email protected]

Contents

1 ZnO and Its Applications K. Ellmer and A. Klein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Zinc Oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Properties of ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Material Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 Growth of ZnO Single Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.2 Deposition of ZnO Thin Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.3 Preparation of ZnO Nanostructures . . . . . . . . . . . . . . . . . . . . . . . 1.5 Electronic Structure of ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Intrinsic Defects in ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 Thermodynamic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.2 Self-Diﬀusion in ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7 Applications of ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.1 Transparent Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.2 Varistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.3 Piezoelectric Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.4 Phosphors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.5 Transparent Oxide Thin Film Transistors . . . . . . . . . . . . . . . . . . 1.7.6 Spintronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 2 4 9 9 9 11 12 14 14 19 21 24 25 26 26 26 27 27

2 Electrical Properties K. Ellmer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Electrical Properties of ZnO Single Crystals . . . . . . . . . . . . . . . . . . . . 2.1.1 Dopants in ZnO Single Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2 Electrical Transport in ZnO Single Crystals . . . . . . . . . . . . . . . . 2.2 Electrical Transport in Polycrystalline ZnO . . . . . . . . . . . . . . . . . . . . . 2.2.1 ZnO Varistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Thin ZnO Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Transport Processes in Polycrystalline Films . . . . . . . . . . . . . . . 2.2.4 Experimental Mobility Data of Polycrystalline ZnO . . . . . . . . . 2.3 Outlook: Higher Electron Mobilities in Zinc Oxide . . . . . . . . . . . . . . . 2.4 Transparent Field Eﬀect Transistors with ZnO . . . . . . . . . . . . . . . . . .

35 36 38 41 53 53 56 57 61 67 70

X

Contents

2.5 Search for p-Type Conductivity in ZnO . . . . . . . . . . . . . . . . . . . . . . . . 71 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 3 Optical Properties of ZnO and Related Compounds C. Bundesmann, R. Schmidt-Grund, and M. Schubert . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Basic Concepts and Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Structural Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Vibrational Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Infrared Model Dielectric Function: Phonons and Plasmons . . 3.2.4 Visible-to-Vacuum-Ultraviolet Model Dielectric Function: Band-to-Band Transitions and their Critical-Point Structures . . . . . . . . . . . . . . . . . . . . . . . . 3.2.5 Spectroscopic Ellipsometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Dielectric Constants and Dielectric Functions . . . . . . . . . . . . . . . . . . . 3.4 Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Undoped ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Doped ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Mgx Zn1−x O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.4 Phonon Mode Broadening Parameters . . . . . . . . . . . . . . . . . . . . . 3.5 Plasmons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Below-Band-Gap Index of Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.1 ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6.2 Mgx Zn1−x O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 Band-to-Band Transitions and Excitonic Properties . . . . . . . . . . . . . . 3.7.1 ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7.2 Mgx Zn1−x O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86 88 90 92 92 98 99 100 102 105 105 106 108 108 116 118

4 Surfaces and Interfaces of Sputter-Deposited ZnO Films A. Klein and F. S¨ auberlich . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Semiconductor Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 ZnO in Thin-Film Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.3 Photoelectron Spectroscopy (PES) . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Surface Properties of ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Crystallographic Structure of ZnO Surfaces . . . . . . . . . . . . . . . . 4.2.2 Chemical Surface Composition of Sputtered ZnO Films . . . . . . 4.2.3 Electronic Structure of ZnO Surfaces . . . . . . . . . . . . . . . . . . . . . . 4.3 The CdS/ZnO Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Band Alignment of II–VI Semiconductors . . . . . . . . . . . . . . . . . . 4.3.2 Sputter Deposition of ZnO onto CdS . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Dependence on Preparation Condition . . . . . . . . . . . . . . . . . . . . . 4.3.4 Summary of CdS/ZnO Interface Properties . . . . . . . . . . . . . . . . 4.4 The Cu(In,Ga)Se2 /ZnO Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

125 126 126 127 128 131 131 133 139 149 149 151 156 162 164

79 79 81 81 83 85

Contents

XI

4.4.1 Chemical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Electronic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 The In2 S3 /ZnO Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 Cu(In,Ga)Se2 Solar Cells with In2 S3 Buﬀer Layers . . . . . . . . . . 4.5.2 Chemical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.3 Electronic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

165 169 172 172 173 176 180

5 Magnetron Sputtering of ZnO Films B. Szyszka . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 History of ZnO Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Magnetron Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Glow Discharge Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Processes at the Target Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Magnetron Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Magnetron Sputtering of ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.5 Ceramic Target Magnetron Sputtering . . . . . . . . . . . . . . . . . . . . . 5.3.6 Other Technologies for Sputter Deposition of ZnO . . . . . . . . . . 5.4 Manufacturing Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 In-Line Coating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Rotatable Target Magnetron Sputtering . . . . . . . . . . . . . . . . . . . 5.5 Emerging Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1 Ionized PVD Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2 Hollow Cathode Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.3 Model-Based Process Development . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

187 187 188 189 190 190 192 194 215 217 218 218 225 227 227 227 228 229

6 Zinc Oxide Grown by CVD Process as Transparent Contact for Thin Film Solar Cell Applications S. Fa¨y and A. Shah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 History of ZnO Growth by CVD Process . . . . . . . . . . . . . . . . . . . 6.1.2 Extrinsic Doping of CVD ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Development of Doped CVD ZnO Films . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Inﬂuence of Deposition Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Growth Mechanisms for CVD ZnO . . . . . . . . . . . . . . . . . . . . . . . . 6.2.3 Inﬂuence of Substrate Temperature . . . . . . . . . . . . . . . . . . . . . . . 6.2.4 Inﬂuence of Precursor Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.5 Doping of CVD ZnO Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.6 The Eﬀect of Grain Size on Electrical and Optical Properties of CVD ZnO Layers . . . . . . . . . . . . . . . . 6.2.7 Alternative CVD Methods for Deposition of Thin ZnO Films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

235 235 235 236 238 238 241 252 261 266 277 279

XII

Contents

6.3 CVD ZnO as Transparent Electrode for Thin Film Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Characteristics Required for CVD ZnO Layers Incorporated within Thin Film Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Pulsed Laser Deposition of ZnO-Based Thin Films M. Lorenz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Brief History and Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Fundamental Processes and Plasma Diagnostics . . . . . . . . . . . . . . . . . 7.3 PLD Instrumentation and Parameters for ZnO . . . . . . . . . . . . . . . . . . 7.4 Results on Epitaxial PLD ZnO Thin Films . . . . . . . . . . . . . . . . . . . . . 7.4.1 Structure of Nominally Undoped PLD ZnO Thin Films . . . . . . 7.4.2 Surface Morphology of PLD ZnO Thin Films . . . . . . . . . . . . . . . 7.4.3 Electrical Properties of PLD ZnO Thin Films: Eﬀect of Buﬀer Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4.4 Luminescence of PLD ZnO Thin Films on Sapphire . . . . . . . . . 7.4.5 Chemical Composition of Doped PLD ZnO Films and Doping Eﬀects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5 Demonstrator Devices with PLD ZnO Thin Films . . . . . . . . . . . . . . . 7.5.1 Large-area ZnO Scintillator Films . . . . . . . . . . . . . . . . . . . . . . . . . 7.5.2 Bragg Reﬂector Mirrors and ZnO Quantum Well Structures . . 7.5.3 Schottky Diodes to ZnO Thin Films . . . . . . . . . . . . . . . . . . . . . . . 7.5.4 PLD of ZnO pn-Junctions, First LEDs, and Other Highlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.6 Advanced PLD Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.6.1 Advances in PLD of Thin Films . . . . . . . . . . . . . . . . . . . . . . . . . . 7.6.2 High-Pressure PLD of ZnO-Based Nanostructures . . . . . . . . . . . 7.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells J. H¨ upkes, J. M¨ uller, and B. Rech . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Silicon Thin Film Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.1 Principle of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.2 Amorphous and Microcrystalline Silicon . . . . . . . . . . . . . . . . . . . 8.2.3 Design Aspects of Silicon Thin Film Solar Cells . . . . . . . . . . . . . 8.2.4 Requirements for TCO Contacts . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.5 Solar Module Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.6 Approaches for Light Trapping Optimization . . . . . . . . . . . . . . . 8.3 Sputter Deposition and Etching of ZnO:Al . . . . . . . . . . . . . . . . . . . . . . 8.3.1 Properties of Sputter Deposited ZnO:Al . . . . . . . . . . . . . . . . . . .

280 280 289 298 299 303 303 305 309 313 314 319 322 327 331 336 338 340 341 344 346 346 348 349 350 359 359 361 361 363 365 368 373 375 377 378

Contents

XIII

8.3.2 Etching Behavior of Zinc Oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3.3 Practical Aspects for Etching of Sputtered ZnO:Al . . . . . . . . . . 8.4 High Eﬃciency Silicon Thin Film Solar Cells . . . . . . . . . . . . . . . . . . . . 8.4.1 Optimization of Solar Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.4.2 Highly Transparent ZnO:Al Front Contacts . . . . . . . . . . . . . . . . 8.4.3 Application of Texture Etched ZnO:Al Films in High Eﬃciency Solar Cells and Modules . . . . . . . . . . . . . . . . . 8.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

379 392 393 393 401 403 404 406

9 Chalcopyrite Solar Cells and Modules R. Klenk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1 Heterojunction Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1.1 Why Use an Undoped ZnO Layer? . . . . . . . . . . . . . . . . . . . . . . . . 9.2 Transparent Front Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.1 Monolithic Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.2 Optical Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.1 Chemical Vapour Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.2 Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.1 Module Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Nonconventional and Novel Applications . . . . . . . . . . . . . . . . . . . . . . . 9.5.1 Direct ZnO/Chalcopyrite Junctions . . . . . . . . . . . . . . . . . . . . . . . 9.5.2 Superstrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.3 Transparent Back Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

415 415 418 419 420 423 423 425 425 427 428 431 431 432 432 434

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439

1 ZnO and Its Applications K. Ellmer and A. Klein

1.1 Introduction Zinc oxide has been investigated already in 1912. With the beginning of the semiconductor age after the invention of the transistor [1], systematic investigations of ZnO as a compound semiconductor were performed. In 1960, the good piezoelectric properties of zinc oxide were discovered [2], which led to the ﬁrst electronic application of zinc oxide as a thin layer for surface acoustic wave devices [3]. Currently, research on zinc oxide as a semiconducting material sees a renaissance after intensive research periods in the 1950s and 1970s [4, 5]. The results of these earlier activities were summarized in reviews of Heiland, Mollwo and St¨ ockmann (1959) [6], Hirschwald (1981) [7], and Klingshirn and Haug (1981) [8]. Since about 1990 an enormous increase of the number of publications on ZnO occurred (see Fig. 1.1) and more recent reviews on ZnO have been published [9–11]. The renewed interested in ZnO as an optoelectronic material has been triggered by reports on p-type conductivity, diluted ferromagnetic properties, thin ﬁlm oxide ﬁeld eﬀect transistors, and considerable progress in nanostructure fabrication. All these topics are the subject of a recently published book [11]. A major driving force of research on zinc oxide as a semiconductor material is its prospective use as a wide band gap semiconductor for light emitting devices and for transparent or high temperature electronics [12]. ZnO has an exciton binding energy of 60 meV. This is higher than the eﬀective thermal energy at 300 K (26 meV). Therefore, excitonic gain mechanisms could be expected at room temperature for ZnO-based light emitting devices. However, a prerequisite is to prepare p-type zinc oxide, which is normally an n-type semiconductor. In the last decade, a lot of eﬀorts were undertaken, to prepare p-type ZnO by doping with nitrogen, phosphorous, and arsenic [9, 10, 13, 14]. Maximum hole concentrations of up to 1019 cm−3 and mobilities of a few cm2 V−1 s−1 at room temperature have been reported [15]. The latter are much lower than the electron mobility of ∼200 cm2 V−1 s−1 . A severe problem is that the p-type conductivity is often not persistent, vanishing within days or weeks [13]. The fundamental thermodynamic diﬃculties for achieving p-type conductivity in ZnO are addressed in Sect. 1.6.1 of this chapter.

2

K. Ellmer and A. Klein

Fig. 1.1. Increase of the number of publications about zinc oxide (ZnO) over the last 40 years according to the literature data base SCOPUS

Further information on p-type doping of ZnO is given in other chapters of this book (2 and 7). In this book the chemical, structural, optical, electrical, and interface properties of zinc oxide are summarized with special emphasis on the use of ZnO as transparent conductive electrode in thin ﬁlm solar cells. This application has a number of requirements, which can be fulﬁlled by ZnO: – High transparency in the visible and near infrared spectral region – Possibility to prepare highly-doped ﬁlms with free electron density n > 1020 cm−3 and low resistivity (<10−3 Ω cm) – Good contacts to the active semiconductors (absorber layers) – Possibility to prepare the TCO layers on large areas (>1 m2 ) by deposition methods like magnetron sputtering or metal-organic chemical vapor deposition (MOCVD) – Possibility to prepare ZnO ﬁlms with suitable properties at low substrate temperature (<200◦ C for Cu(In,Ga)(S,Se)2 solar cells) – Possibility for preparation of tailored surfaces with suitable light scattering properties for light trapping, which is particularly important for Si thin ﬁlm solar cells – Low material costs, nontoxicity, and abundance in earth crust

1.2 Zinc Oxide Zinc oxide (ZnO) is an oxidic compound naturally occurring as the rare mineral zincite, which crystallizes in the hexagonal wurtzite structure P63 mc [16]. The mineral zincite was discovered in 1810 by Bruce in Franklin (New Jersey,

1 ZnO and Its Applications

3

Fig. 1.2. (a) An orange zincite crystal from Sterling Mine, Ogdensburg, USA (collection: Rob Lavinsky, www.mineralienatlas.de/lexikon/index.php/Zinkit) and (b) a synthetic zinc oxide crystal (www.gc.maricopa.edu/earthsci/imagearchive/ zincite755.jpg). The mineral in (a) exhibits a size of 30 × 25 × 6 mm3

USA). The zincite ore in New Jersey is an important zinc source of the USA. Other locations where zincite can be found are Sarawezza (Tuscany, Italy), Tsumeb (Namibia), Olkusz (Poland), Spain, Tasmania, and Australia. Zincite is usually colored red or orange by manganese impurities. Photographs of zincite are shown in Fig. 1.2. Zinc oxide crystals exhibit several typical surface orientations. The most important surfaces are the (0001) and (000¯ 1) (basal plane), (10¯ 10) and (11¯ 20) (prism planes) and (11¯21) (pyramidal plane) crystal faces. In principle, the (0001) planes are terminated by Zn atoms only, while the (000¯ 1) surfaces are terminated by oxygen atoms only. However, this simple picture does not hold in reality (see description of the surface structure in Sect. 4.2.1 of this book). Nevertheless, the etching behavior is noticeably diﬀerent for these two surfaces [17] (see also Chap. 8). Today, most of the zinc oxide powder produced worldwide is used in nonelectronic applications for rubber production, chemicals, paints, in agriculture and for ceramics [18]. The pure powder is produced from metallic zinc, which is an abundant material in the earth’s crust. Identiﬁed zinc resources of the world are about 1.9 × 109 tons [18]. The world wide mine production in 2006 was 107 tons [18]. In 2006, an estimated 350,000 tons of zinc was recovered from waste and scrap [18]. Of the total zinc consumed, about 55 % was used in galvanic processes for corrosion protection, 21 % in zinc-based alloys, 16 % in brass and bronze, and 8 % in other uses. Major coproducts of zinc mining and smelting, in order of decreasing tonnage, were lead, sulfuric acid, cadmium, silver, gold, and germanium.

4

K. Ellmer and A. Klein 1st layer c

2nd layer Zn

b O b

a

c

a

Fig. 1.3. Two views of the crystal structure of zinc oxide (ZnO). Left: Perspective view perpendicular to the c-axis. The upper side is the zinc terminated (0001) plane, the bottom plane is oxygen terminated (000¯ 1). Right: View along the c-axis on the zinc terminated (0001) plane

Already in 1914, shortly after the discovery of X-ray diﬀraction, Bragg elucidated the crystal structure of wurtzite ZnO by X-ray diﬀraction, which was published in 1920 [19]. The hexagonal unit cell (a = 0.325 nm, c = 0.52066 nm) of ZnO, which contains 2 molecules, is shown in Fig. 1.3. The zinc atoms are surrounded by oxygen atoms in a nearly tetrahedral conﬁguration. Along the c-axis the Zn–O distance is somewhat smaller (dZn–O [1] = 0.190 nm) than for the other three neighboring oxygen atoms (dZn–O [2] = 0.198 nm). Besides the hexagonal wurtzite phase, a metastable cubic phase with the rocksalt structure is also known. Using synchrotron radiation energydispersive X-ray diﬀraction, Decremps et al. [20] measured the shrinking of the lattice cell of hexagonal ZnO up to hydrostatic pressures of 11 GPa, while Desgrenier extended his measurements even up to 56 GPa [21]. At a pressure of 9.8 GPa (at 300 K) a phase transition to the cubic phase of ZnO that exhibits the rocksalt (NaCl) structure occurs. Upon decreasing the hydrostatic pressure this phase transition is reversible in the pressure range from 2–6 GPa, depending on temperature. This means, that, in contrast to an earlier observation, the high pressure phase is not metastable at normal pressure. The temperature–pressure phase diagram of ZnO is shown in Fig. 1.4 [20]. From the XRD data at diﬀerent pressures and temperatures the equation-ofstate parameters for rocksalt ZnO can be calculated. The bulk modulus B at room temperature is 196±5 GPa, which is higher than that of hexagonal ZnO (B = 142.4 GPa). The corresponding bulk moduli of Ge, Si, and GaAs are 75, 98, and 75.3 GPa [22]. This means ZnO exhibits a high covalent bonding energy.

1.3 Properties of ZnO In Table 1.1 a number of properties of zinc oxide are summarized in comparison to other transparent conducting oxides and to silicon.

1 ZnO and Its Applications

5

Fig. 1.4. Phase diagram of ZnO. The squares mark the wurtzite (B4) – rocksalt (B1), the triangles the B1–B4 transition [20]. The hysteresis of the transition, which extends from 9.8 to 2 GPa at 300 K, depends on the temperature and is absent for T > 1 300 K

Today, the most important TCOs for electrode applications are In2 O3 , SnO2 , and ZnO, which are typically doped using tin (In2 O3 :Sn = ITO), ﬂuorine (SnO2 :F = FTO), and Al (ZnO:Al = AZO), respectively [24, 33, 34]. The transparency is due to their optical band gaps, which is ≥3.3 eV, leading to a transparency for wavelength >360 nm. The direct gap of In2 O3 is Eg,d = 3.6–3.75 eV [35, 36]. An indirect optical gap of Eg,d = 2.6 eV has also been reported for this material [35]. However, literature is not yet conclusive on the existence of an indirect gap (see discussions in [37–39]). CdO, the ﬁrst discovered and applied transparent conductor [40], which also exhibits the highest reported conductivity (see compilation of data in [41]), is less used today because of its toxicity and its low optical band gap (Eg,d = 2.2 eV, Eg,i = 0.55 eV [42]). Photoelectron spectra indicate, however, that the band gap is ∼1 eV [43]. Although the (direct) optical gap is increased for degenerate doping due to the Burstein–Moss eﬀect [44], it remains diﬃcult to prepare noncolored CdO ﬁlms. More recently also TCOs with multiple cations have been investigated by a number of groups [34,45,46]. The search for new TCO materials is partially related to the desired replacement of indium due to its limited availability, but also to achieve better functionality due to modiﬁed interface properties or the requirement of even higher conductivity at the same transparency level. The latter might be achieved by TCOs with higher carrier mobilities (see also Chap. 2). In addition, a number of p-type TCOs have been identiﬁed [47–53]. These typically contain metals with shallow d-levels (mainly Cu 3d). In consequence, these materials can have high carrier concentrations, but typically lack of high mobility [54, 55]. CuInO2 has been shown to exhibit both n- and p-type conductivity [51,52]. The transparency of this material is, however, due to an optically forbidden transition of the fundamental gap [56].

6

K. Ellmer and A. Klein

The electrically active band gap is, therefore, considerably lower than the apparent optical gap. Of the semiconductors in Table 1.1 only ZnO exhibits piezoelectricity, i.e., the generation of electrical charges when subjected to a mechanical deformation. This is caused by the polarity of zinc oxide, i.e., the absence of inversion symmetry in this crystal lattice. Other materials that exhibit piezoelectricity are e.g., α-quartz (SiO2 ), LiNbO3 , ZnS, turmaline, saccharose, or liquid crystals. The high piezoelectric or electromagnetic coupling factor of ZnO, which is deﬁned as the square root of the ratio of electrostatic energy and the mechanical deformation energy stored in the material Table 1.1. Abundance of the metal in the earths’s crust, optical band gap Eg (d: direct; i: indirect) [23, 24], crystal structure and lattice parameters a and c [23, 24], density, thermal conductivity κ, thermal expansion coeﬃcient at room temperature α [25–27], piezoelectric stress e33 , e31 , e15 and strain d33 , d31 , d15 coeﬃcients [28], electromechanical coupling factors k33 , k31 , k15 [29], static ε(0) and optical ε(∞) dielectric constants [23, 30, 31] (see also Sect. 3.3, Table 3.3), melting temperature of the compound Tm and of the metal Tm (metal), temperature Tvp at which the metal has a vapor pressure of 10−3 Pa, heat of formation ∆Hf per formula unit [32] of zinc oxide in comparison to other TCOs and to silicon Parameter

Unit

ZnO

Mineral Zincite Abundance ppm 40 Eg eV 3.4 (d) Lattice Hexagonal structure Wurtzite Space group P 63 mc (number) 186 a, c nm 0.325, 0.5207 Density g cm−3 5.67 κ W m−1 K−1 69 , 60⊥ α 10−6 /K 2.92 , 4.75⊥ −2 1.32, –0.57, –0.48 e33 , e31 , e15 C m d33 , d31 , d15 10−12 C N−1 11.7, –5.43, –11.3 k33 , k31 , k15 0.47, 0.18, 0.2 (0) 8.75 , 7.8⊥ (∞) 3.75 , 3.70⊥ ◦ Tm C 1,975 ◦ Tm (metal) C 420 ◦ C 208 Tvp ∆Hf eV 3.6 a Decomposes into SnO and O2 at 1,500◦ C

In2 O3

SnO2

Si

6.7

Cassiterite 40 3.6 (d) Tetragonal Rutile P 42 /nmm 136 0.474, 0.319 6.99 98 , 55⊥ 3.7 , 4.0⊥

Silicon 2.6 × 105 1.12 (i) Cubic Diamond F d3m 227 0.5431 2.33 150 2.59

8.9 4.6 1,910 157 670 9.6

9.58 , 13.5⊥ 4.17 , 3.78⊥ 1,620a 232 882 6.0

1,410 1,410

0.1 3.6 (d) Cubic Bixbyite Ia3 206 1.012 7.12

–

1 ZnO and Its Applications

7

Fig. 1.5. Binary Zn–O phase diagram [58]. Above 200◦ C only ZnO is stable. L indicates the solubilities of oxygen in Zn at diﬀerent temperatures. Tm (Zn) is the melting point of Zn (419.6◦ C)

(kij2 = d2ij /eij /sij ) [57], led to one of the ﬁrst electronic applications of zinc oxide in surface-acoustic wave devices (see Sect. 1.7). The piezoelectricity of ZnO also induces a charge carrier scattering process (piezoelectric mode scattering, see Chap. 2, Sect. 2.1.2), which is important for the electron mobility in ZnO single crystals at low temperatures (100 K). For highly-doped ZnO ﬁlms, there are indications that the piezoelectricity inﬂuences the carrier transport even at room temperature (see Sect. 2.2.3). The binary oxygen–zinc phase diagram is depicted in Fig. 1.5 [58]. Above 200◦ C only the binary compound ZnO is stable. At low temperatures also zinc peroxide (ZnO2 ) is reported, which can be prepared by chemical synthesis [59]. The melting point of ZnO is 1975◦C. The sublimation of ZnO occurs congruently by decomposition to the gaseous elements according to: ZnO(s) Zn(g) + 0.5 O2 (g) All three oxides listed in Table 1.1 exhibit high melting points of 1,600– 2,000◦C. However, it has to be kept in mind that these oxides decompose into the elements below their melting points if the oxygen partial pressure is too low. This is important for the deposition of such oxides at higher substrate temperatures (see Chaps. 5–7). The melting points of the metals in ZnO, In2 O3 , and SnO2 are quite low. Zinc also has a high vapor pressure at typical substrate temperatures during deposition (<600 K). In contrast indium and tin have lower vapor pressure, and reevaporation of metal during growth at elevated substrate temperature is less important. The formation energies of the oxides, related to one metal atom, are 3.6 eV (ZnO), 4.8 eV (In2 O3 ), and 6 eV (SnO2 ) explaining the increased thermodynamic stability of these oxides going from ZnO to SnO2 . The vapor pressure of ZnO is high already at about 1,400◦C [60], which makes it diﬃcult to grow single crystals from its own melt. The stoichiometric width of ZnO below 600◦C is rather narrow. For temperatures higher than

8

K. Ellmer and A. Klein

600◦ C Hagemark and Toren [61] measured the Zn–Zn1+x O phase boundary by an electro-chemical method and by Hall and conductivity measurements assuming that excess zinc constitutes a shallow donor in ZnO. New measurements of Tomlins et al. [62] suggest that Hagemark and Toren actually measured the phase boundary Zn–ZnO1−x , i.e., the concentration of oxygen vacancies. Recently, Lott et al. [63] measured the excess zinc in the vapor phase directly by optical absorption. Their results are shown in Fig. 1.6. The temperature dependence of the lattice parameters a and c of ZnO were measured by Khan [25] and Reeber [26] and are displayed in Fig. 1.7. For low temperatures (T < 200 K) ZnO exhibits a very small thermal expansion. For temperatures higher than 300 K, the thermal expansion of the a and c axis are linearly dependent on T , which means that the thermal expansion 1200 1100

T [°C]

1000 900 800 700 600 500

0

50

100

150

200

xZn [ppma]

0.3265

0.522

0.3260

0.520

0.3255

0.518

0.3250 (b) 0

200

400

600

0.516

lattice parameter c [nm]

lattice parameter a [nm]

Fig. 1.6. Zinc excess xZn according to Lott et al. [63] near the stoichiometric composition Zn1+x O in the temperature range 600–1,100◦ C

800

temperature [K] Fig. 1.7. Thermal expansion of ZnO (wurtzite) along the a and the c-axis in the temperature ranges from 4 to 296 K [26] (full symbols) and from 300 to 892 K [25] (open symbols), which ﬁt very well to each other

1 ZnO and Its Applications

9

coeﬃcients α/⊥ = (da, dc)/dT are constant. The linear thermal expansion coeﬃcient at room temperature is α = 2.9 × 10−6 K−1 and α⊥ = 4.75 × 10−6 K−1 , which is lower than that of In2 O3 and SnO2 and only slightly higher than that of silicon (see Table 1.1).

1.4 Material Preparation 1.4.1 Growth of ZnO Single Crystals Zinc oxide single crystals were grown already in 1935 by Fritsch by evaporation and condensation of ZnO from pressed and sintered cylinders in air at about 1,450◦C [64]. Later especially three methods were used for crystal growth: – Hydrothermal growth in autoclaves at temperatures up to 450◦ C and pressures up to 2,500 bar [65, 66] – Growth from the gas phase by oxidation of Zn vapor at temperatures between 1,100 and 1,400◦C [57, 67] – Growth from melts of salts with low melting temperature (ZnBr2 ) [68]. Today, only the ﬁrst two methods are used for the growth of single crystals [69]. However, recently also the growth of ZnO from its pressurized melt in a cold crucible was demonstrated [70]. This technique, which uses oxygen pressures up to 100 bar, is scalable to crystal diameters larger than 125 mm. A new, promising method, avoiding the temperature-gradient problems with a cold crucible, uses a hot iridium crucible [71]. To prevent the oxidation (burning) of the iridium crucible the growth is performed in a self-adjusting gas atmosphere of CO2 . Single crystalline ZnO wafers are commercially available from diﬀerent companies as listed in Table 1.2. Photographs of a ZnO single crystal and of wafers are shown in Fig. 1.8. 1.4.2 Deposition of ZnO Thin Films ZnO thin ﬁlms can be prepared by a variety of techniques such as magnetron sputtering, chemical vapor deposition, pulsed-laser deposition, molecular beam epitaxy, spray-pyrolysis, and (electro-)chemical deposition [24,74]. In this book, sputtering (Chap. 5), chemical vapor deposition (Chap. 6), and pulsed-laser deposition (Chap. 7) are described in detail, since these methods lead to the best ZnO ﬁlms concerning high conductivity and transparency. The ﬁrst two methods allow also large area depositions making them the industrially most advanced deposition techniques for ZnO. ZnO ﬁlms easily crystallize, which is diﬀerent for instance compared with ITO ﬁlms that can

10

K. Ellmer and A. Klein Table 1.2. Parameters of commercially available zinc oxide single crystals

Parameter

Unit

ZNT

CI

ML

TDC

Method s-CVT PM HT HT ◦ T C 1,100 2,000 330 300–400 P MPa 50 <100 d mm 51 130 51 51 ρ Ω cm 0.3 0.35 500–1 000 390 cm−3 1 × 1017 9 × 1016 8 × 1013 ND − NA µ cm2 V−1 s−1 209 200 200 −2 cm <50 <300 Ndisl. Ref. [72] [70] [73] [66] Abbreviations and symbols denote ZNT: ZN Technology Inc. (formerly Eagle Pitcher), Brea, USA; CI: Cermet Inc., Atlanta, USA; ML: Mineral Ltd., Alexandrov, Russia; TDC: Tokyo Denpa Co., Tokyo, Japan; s-CVT: seeded chemical vapor transport; PM: pressurized melt; HT: hydrothermal; T : growth temperature; P : growth pressure; d: dimension; ρ: resistivity; ND,A : donor or acceptor concentration; µ: Hall mobility; Ndis : dislocation density

Fig. 1.8. (a) Photographs of a hydrothermally grown zinc oxide single crystal (Mineral Ltd., Alexandrov, Russia) and (b) of zinc oxide wafers prepared from such crystals (Crystec GmbH, Berlin, Germany). The as grown crystal in (a) has a maximum size of 70 mm [73]. The size of the crystals in (b) is 10 × 10 × 0.5 mm3

be grown as amorphous ﬁlms [75]. ZnO ﬁlms exhibit mostly a (0001) texture, i.e., the c-axes of the crystallites are perpendicular to the substrate surface [24, 74]. Plasma-assisted processes (magnetron sputtering, pulsed-laser ablation) induce a pronounced c-axis (0001) texture. While on amorphous substrates or silicon ZnO forms polycrystalline ﬁlms without an in plane alignment, epitaxial ﬁlms, i.e., ﬁlms with an in and out of plane alignment can be grown on oxide single crystalline substrates (sapphire-Al2 O3 , LiNbO3 , MgO etc.) even at room temperature [76]. Structural and electrical properties of polycrystalline and epitaxial ﬁlms are compared in Sect. 2.2.3.

1 ZnO and Its Applications

11

1.4.3 Preparation of ZnO Nanostructures The ﬁrst experiments on single crystal growth of ZnO [64, 77] showed that under equilibrium conditions ZnO grows preferentially along the (0001) axis, leading to needle-like crystals with dimensions in c direction of some millimeter. The ﬁrst reported electrical and optical measurements on ZnO single crystals were performed on such pencil-like crystals with a hexagonal cross section [2] (see Chap. 2). Later, under nonequilibrium conditions a large variety of ZnO nanostructures were synthesized. Yamada and Tobisawa obtained ZnO plates, columns, pyramids, stellar-shaped crystals, spheres, whiskers, and dendrites under the extreme nonequilibrium conditions of a converging shock-wave using an explosive charge [78]. In the last 10 years other techniques were used to grow ZnO nanostructures: pulsed-laser ablation (see also Chap. 7), magnetron sputtering [79], chemical-vapor deposition [80], and chemical solution preparation [81]. ZnO nanostructures exhibit a high crystalline quality as inferred from cathode- and photoluminescence measurements [82]. A review on ZnO nanorods was given recently by Yi et al. [83]. Information on ZnO nanostructures can also be found in [9, 11]. Figure 1.9 shows scanning electron micrographs of an array of ZnO nanowires grown on a magnetron-sputtered ZnO ﬁlm on a ﬂuorine-doped SnO2 covered glass substrate. Because of the strongly c-axis oriented growth of the magnetron-sputtered ZnO ﬁlm, the ZnO nanowires are very wellaligned vertically. The hexagonal cross section is clearly visible. Depending on the preparation conditions, diameter and length of the nanowires can be adjusted [79]. Possible applications of ZnO nanostructures are UV lasers (up to now only with optical excitation) [84], chemical sensors [85], or transparent substrates for thin ﬁlm solar cells, e.g. as an alternative to TiO2 in injection type solar cells or in organic solar cells [79, 86, 87]. In the latter two applications the increased eﬀective surface of arrays of ZnO nanowires leads to

Fig. 1.9. Plain view (left) and cross-sectional view (right) of ZnO nanowires prepared from a chemical solution on a magnetron-sputtered ZnO ﬁlm on ﬂuorinedoped SnO2 on glass, which are used as a substrate for dye-sensitized solar cells [79]

12

K. Ellmer and A. Klein

increased sensitivity or absorption. Recently, a new, interesting application of ZnO nanowires was reported [88]. Combining ZnO nanowires with zig–zag structured electrodes, a device that is able to convert ultrasonic energy into a voltage was demonstrated. In this arrangement the piezoelectric eﬀect of ZnO (see Table 1.1) is used to extract a voltage due to mechanical bending of the nanowires.

1.5 Electronic Structure of ZnO ZnO is a tetrahedrally bonded semiconductor. The electronic structure of tetrahedrally bonded materials with diamond, zincblence, wurtzite, or chalcopyrite structure is very similar [23, 89, 90]. Except for symmetry, the electronic structures of zincblende and wurtzite modiﬁcations are mostly identical. The cation and anion s- and p-orbitals form sp3 -hybrids, which overlap forming bonding and antibonding combinations. The electronic states at the valence band maximum are mainly derived from anion p-states and are, therefore, threefold degenerate. However, the degeneracy is lifted by spinorbit splitting and also by noncentrosymmetric crystal ﬁelds. The latter is pronounced e.g., in the chalcopyrite structure with a I–III–VI2 (e.g., CuInSe2 ) or a II–IV–V2 (e.g., ZnSnP2 ) composition [91] but also occurs in ZnO [23]. At liquid He temperatures the three valence bands are at 0, 4.9, and 48.6 meV binding energy with respect to the valence band maximum, respectively [92]. In the I–III–VI2 chalcopyrites, there is further a considerable hybridization between the anion p-states and the low lying d-states of the group I cation, which leads to a lowering of the optical gap compared with the binary II–VI analogues [93, 94]. In literature this band gap anomaly is often indicated as p–d repulsion as the presence of the metal (Cu, Ag) d-states leads to an upward shift of the valence band maximum [93–95]. In LCAO1 theory the valence band maximum of a tetrahedrally bonded semiconductor is derived from the anion p-levels and its energy given by [89]: 2 2 1/2 Epa + Epc /2 + 1.282 / me d2 , (1.1) EVB = Epa + Epc /2 − where Epa and Epc are the p orbital energies of the anion and the cation, me is the free electron mass and d the interatomic spacing, respectively. According to (1.1), a lower valence band maximum has to be expected for anions with a larger binding energy of the p levels. The binding energy2 of the p orbitals increases monotonically from Te (9.54 eV) over Se (10.68 eV) and S (11.6 eV) to O (16.72 eV) [89]. In fact the valence band oﬀsets between II–VI semiconductors, which are displayed in Fig. 1.10, nicely follow the trend calculated by Wei and Zunger using density functional theory [95]. They correspond well with experimental determinations (see Chap. 4). 1 2

Linear combination of atomic orbitals [89]. Hartree–Fock atomic term values are used in LCAO theory [89].

1 ZnO and Its Applications

13

Fig. 1.10. Band alignment between II–VI compounds according to density functional theory calculations by Wei and Zunger [95]. The energy of the valence band maximum of ZnS is arbitrarily set to 0 eV. A comparison to experimental results is presented in Fig. 4.18 in Sect. 4.3.1 (page 150)

To understand the full details of the band alignment, the contribution of the cation d-states also needs to be considered [95]. Because of the low lying O 2p levels, the interaction between the Zn 3d and the anion p states in ZnX (X = O, S, Se, Te) compounds is strongest for ZnO [43, 96–98]. This also explains the anomaly of the band gaps for the ZnX compounds (see Fig. 1.10): While the band gap generally increases with lower anion mass, the gap of ZnO (3.4 eV) is smaller than the gap of ZnS (3.7 eV). However, instead of shifting the valence band maximum upward in energy, as in the case of the chalcopyrites [93,94] and other II–VI compounds [94,95], the band alignments given in Fig. 1.10 rather suggest a lowering of the ZnO conduction band minimum as the valence band maximum of ZnO roughly follows the trend with decreasing anion mass. As the mixing between the anion p and the cation d levels aﬀects the energetic positions of the energy bands, it is particularly important for the function of quantum well structures. The band alignment between the barrier and the well material is important for establishing the energy levels and, therefore, the optical transitions. ZnO-based quantum well structures can be prepared by alloying of ZnO with MgO or CdO [99–104]. Zn1−x Mgx O retains the wurtzite structure for Mg admixtures up to x ≈ 0.5 [99,102–104] resulting in a variation of the optical gap with x from 3.4–4.5 eV [104]. More details on the optical properties of ZnO and (Zn,Mg)O are presented in Chap. 3. As there are no d levels in Mg, alloying of ZnO with MgO therefore aﬀects the energy positions of the band edges according to the diﬀerent s and p atomic levels (1.1) and by modifying the interaction between the anion p and the cation d states. Figure 1.11 shows an experimental determination of the band alignment at the ZnO/(Zn,Mg)O interface using optical spectroscopy of quantum well structures [100]. The data indicate that the larger band gap of (Zn,Mg)O is

14

K. Ellmer and A. Klein

Fig. 1.11. Energy-level alignment between ZnO and (Zn,Mg)O as determined by optical spectroscopy [100]. The energy-level alignment agrees with a recent indirect determination using photoelectron spectroscopy [105]. The diﬀerence of the band gaps is almost fully accomplished by a conduction band oﬀset

almost completely accomplished by an upward shift of the conduction band energy. This result is in good agreement with a recent indirect determination of the band alignment using photoelectron spectroscopy [105] (for more information on band alignments see Chap. 4).

1.6 Intrinsic Defects in ZnO 1.6.1 Thermodynamic Properties Intrinsic point defects are deviations from the ideal structure caused by displacement or removal of lattice atoms [106,107]. Possible intrinsic defects are vacancies, interstitials, and antisites. In ZnO these are denoted as VZn and VO , Zni and Oi , and as ZnO and OZn , respectively. There are also combinations of defects like neutral Schottky (cation and anion vacancy) and Frenkel (cation vacancy and cation interstitial) pairs, which are abundant in ionic compounds like alkali-metal halides [106,107]. As a rule of thumb, the energy to create a defect depends on the diﬀerence in charge between the defect and the lattice site occupied by the defect, e.g., in ZnO a vacancy or an interstitial can carry a charge of ±2 while an antisite can have a charge of ±4. This makes vacancies and interstitials more likely in polar compounds and antisite defects less important [108–110]. On the contrary, antisite defects are more important in more covalently bonded compounds like the III–V semiconductors (see e.g., [111] and references therein). The formation of one mole of a defect requires an energy ∆Hd . The distribution of the created defects on the available sites N increases the entropy

1 ZnO and Its Applications

15

S of the system. The change in free energy of a crystal by defect formation ∆Gd = ∆Hd − T S has a minimum at a defect concentration Nd = N exp(−∆Hd /kBT ),

(1.2)

which deﬁnes the number of defects in thermodynamic equilibrium. An important aspect of intrinsic defects is the dependence of their formation energy on the chemical potential and the Fermi level position in the gap [112]. The chemical potential corresponds to the concentration of a species, which is for gases related to its partial pressure. An increasing partial pressure of oxygen, e.g., leads to a lowering of ∆Hd for defect reactions that add oxygen or remove Zn (Oi , VZn ) and vice versa. For charged defects, the defect formation enthalpy directly depends on the Fermi energy, which is the chemical potential of the electrons (and holes).3 This dependency is schematically sketched in Fig. 1.12. A donor type defect e.g., can be neutral or positively charged depending on its occupation, which is given by the Fermi level position with respect to the defect energy ED . Donors are occupied (neutral) when the Fermi level is above the defect level and unoccupied (charged) when the Fermi level is below the defect level. The electron from the donor is transferred to its reservoir whose energy is represented by the Fermi energy. This leads to an energy gain δ = q(ED − EF ), where q is the charge of the donor. The energy gained by lowering the Fermi level toward the valence band maximum generally decreases the formation enthalpy of the donors. As intrinsic defects in ZnO carry a charge of 2, the maximum energy gain from a variation of the Fermi

Fig. 1.12. Dependence of the defect formation enthalpy ∆Hd on the Fermi energy. An occupied donor (EF > ED ) is electrically neutral and ∆Hd does not depend on EF . The unoccupied donor (EF < ED ) is positively charged and the electron(s) released from the donor is(are) transferred to the electron reservoir (EF ) leading to an energy gain δ. The defect formation enthalpy hence decreases with decreasing Fermi energy. The resulting variation of defect formation enthalpy with Fermi energy for donor and acceptor defects is shown on the right. The crossing of the neutral and the charged defect state identiﬁes the energy position of the defect states (ED and EA ) 3

The treatment is for homogeneous materials (bulk) where the electrostatic potential is constant.

16

K. Ellmer and A. Klein

(a)

(b)

(c)

Fig. 1.13. Top: Variation of defect formation enthalpies with Fermi level under zinc- (left) and oxygen-rich (right) conditions as obtained from GGA+U calculations. The gray shaded area indicates the diﬀerence between the calculated and the experimental band gap. The numbers in the plot indicate the defect charge state; parallel lines imply equal charge states; Bottom: Transition levels in the band gap calculated within GGA (a), GGA+U (b) and using an extrapolation formula described in [115]. The dark gray shaded areas indicate error bars. Copyright (2006) by the American Physical Society

energy is two times the band gap. This is more than 6 eV and of the order of a typical defect formation enthalpy. In recent years, there has been considerable eﬀort to derive defect formation enthalpies of intrinsic defects in ZnO [108–110, 113–117]. An example is shown in Fig. 1.13 [115]. Horizontal curves belong to neutral defects, curves with positive or negative slopes to charged donors or acceptors, respectively. The donor with the lowest formation enthalpy is the oxygen vacancy VO , the acceptor with the lowest formation enthalpy the zinc vacancy VZn .

1 ZnO and Its Applications

17

This corresponds to a Frenkel type defect behavior. Negative defect formation enthalpies corresponds to unstable crystals (spontaneous defect formation) and indicate the limit for physically reasonable Fermi level positions. Figure 1.13 includes the oxygen dumbbell interstitial or split interstitial defects (Oi,db and Oi,rot−db ). This particular defect was ﬁrst mentioned by Lee et al. [113] and later treated extensively by Erhart and coworkers [114, 115]. The oxygen dumbbell interstitial corresponds to two oxygen atoms in the oxidation state −1, which occupy an oxygen lattice site. It can, therefore, also be considered as a peroxide-like defect. The atomic arrangement of the defects is shown in Fig. 1.14 together with the ideal lattice structure and the octahedral interstitial oxygen defect [114]. The oxygen dumbbell interstitial is an amphoteric defect, i.e., it behaves as a donor and as an acceptor. However, the transition energy for the donor is close to the valence band maximum, providing a very deep donor level. The transition energy for the acceptor is far above the valence band maximum, hence also providing a deep defect state. Although the dumbbell interstitial, therefore, does not contribute to the

Fig. 1.14. Dumbbell oxygen interstitial defects in ZnO in comparison to the ideal lattice structure and the octahedral oxygen interstitial [114, 115]. The structure of the rotated dumbbell interstitial depends on the charge state. Copyright (2005) by the American Physical Society

18

K. Ellmer and A. Klein

electrical conductivity of ZnO, it is essential for describing oxygen diﬀusion in ZnO (see Sect. 1.6.2). The defect transition energies are given in the lower part of Fig. 1.13. These are relevant for establishing the equilibrium Fermi level position and might help to explain the origin of the usually observed residual n-type conductivity of ZnO. Unfortunately, the band gaps calculated by density functional theory are typically much too low. Using the standard local density approximation (LDA) or the generalized gradient approximation (GGA), the band gap of ZnO is determined as 0.7–0.9 eV (see e.g., [115]). A larger, but still too small band gap of ∼1.8 eV is derived when a +U correction, which accounts for the self-interaction terms, is applied (see [115] and references therein). The literature on intrinsic defects in ZnO diﬀers considerably with respect to the defect transition energies [108–110,113–117]. A recent comparison of the diﬀerent approaches is given in [115], where also a sophisticated interpolation scheme is proposed to overcome the band gap problem. The resulting defect transition energies are given in Fig. 1.13 (bottom right). The lowering of the defect formation enthalpy with the movement of the Fermi energy leads to an increase of the defect concentration. Generally, donor (acceptor) doping leads to a rise (lowering) of the Fermi energy and therefore to a lowering of the formation enthalpy of intrinsic acceptors (donors). This mechanism leads to a limitation of the Fermi level movement and is called selfcompensation [118–120]. It is well known for many semiconductors and leads to a limitation of the doping. In the extreme case of insulators, doping does not lead to creation of electrons or holes but only introduces compensating intrinsic defects [106, 107]. Doping limits in ZnO are important in two aspects: The possibility for p-type doping and the possibility for degenerate n-type doping. P-type conductivity requires a Fermi energy close to the valence band maximum, where the formation enthalpy of compensating donors is very low. According to defect calculations, the most important intrinsic defect for the compensation of p-type conductivity is the oxygen vacancy. On the other hand, the formation of zinc vacancies will limit n-type doping in ZnO. That this limitation occurs for Fermi level positions well above the conduction band minimum is a prerequisite for a transparent conducting electrode material, as otherwise no high electron concentrations are possible. However, it is indeed observed experimentally that the doping eﬃciency (number of free carriers per dopant atom) decreases with the increase of the dopant concentration [121] (see Chaps. 6 and 8). The electron concentration in donor-doped TCOs becomes compensated with increasing oxygen partial pressure. The nature of the compensating defect thereby depends on the material. As mentioned earlier, compensation of n-type doping in ZnO occurs by introduction of zinc vacancies. In contrast, compensation in In2 O3 is accomplished by oxygen interstitials [117]. Their importance in Sn-doped In2 O3 has been already pointed out by Frank

1 ZnO and Its Applications In1.98Sn0.02O3

19

Zn0.99Al0.01O

Conc. [cm-3]

∆µO [eV] ~p-1/8

AlZn

SnIn

n(Ta) VO

VZn n(Ta)

Oi

~p-1/4

Ta =1073K VIn

VO

log [[p(O 2 )/atm]

Fig. 1.15. Electron concentration (dashed line) of Sn-doped indium oxide and Al-doped ZnO in dependence on oxygen partial pressure for a dopant concentration of 1 % [117]. With increasing oxygen partial pressure the donors become compensated by oxygen interstitials (In2 O3 ) or by zinc vacancies (ZnO). Reprinted with permission from [117]. Copyright (2007) by the American Physical Society

and K¨ ostlin in 1982 [122]. The electron concentration for Al-doped ZnO and Sn-doped In2 O3 in dependence on oxygen pressure has been calculated by Lany and Zunger [117]. The result is presented in Fig. 1.15. From Fig. 1.15, it is also evident that, in contrast to frequent assertions, oxygen vacancies are irrelevant for establishing the free carrier concentration in doped TCOs. The defect structure described by Fig. 1.13 does not provide a straightforward explanation for the residual n-type conductivity [116,117,123]. First, the oxygen vacancy is a deep donor and cannot introduce enough carriers because of the large ionization energy. On the other hand, the zinc interstitial has a low ionization energy but a rather high formation enthalpy and is, therefore, not present in required amounts. This diﬃculty is explicitly treated by Lany and Zunger, who suggest a metastable oxygen vacancy defect induced by optical excitation as origin of the residual n-type conductivity [117, 123]. Alternatively, van de Walle has suggested that hydrogen, which forms a shallow donor in ZnO, causes the n-type conductivity [124] as also conﬁrmed experimentally [10, 125–128] (See also Chap. 2). 1.6.2 Self-Diﬀusion in ZnO Self-diﬀusion in materials occurs by repeated occupation of defects. Depending on the defects involved one can distinguish between (1) vacancy, (2) interstial, and (3) interstitialcy mechanisms [107]. As an example, diﬀerent diﬀusion paths for oxygen interstitials are illustrated in Fig. 1.16 [129]. For a detailed description of diﬀusion paths for oxygen vacancies, zinc vacancies and zinc interstitials the reader is also referred to literature [129, 130]. When moving in a crystal, an atom has to surmount an energy barrier, which is called the migration enthalpy ∆Hm . The mobility of a diﬀusing species is, therefore, thermally activated and diﬀusion is described by the

20

K. Ellmer and A. Klein

X

A1 X

B2

Oi,db Oi,rot-db

B1

A2

[0110] C

[2110]

[1210]

[1100]

B2

A1 X

(a)

X

B1

A2

[0001] C

(b)

(c)

[1100]

Fig. 1.16. Diﬀusion paths accessible to oxygen interstitials on the wurtzite lattice via jumps to ﬁrst or second nearest neighbor sites [129]. Panels (a) and (b) show in-plane and out-of-plane diﬀusion paths to ﬁrst nearest oxygen neighbors, panel (c) illustrates out-of-plane diﬀusion via second nearest oxygen neighbors. The size of the spheres scales with the position of the atom along the [0001] axis. Copyright (2006) by the American Physical Society

diﬀusion coeﬃcient D given by a prefactor D0 and an activation energy ∆Hm according to D = D0 exp (−∆Hm /kB T) (1.3) However, the diﬀusivity also depends on the concentration of defects, which is also thermally activated according to (1.2). Hence, the self-diﬀusion coefﬁcient D is described as exp (−∆Hm /kB T) (1.4) D ∝ exp (−∆Hd /kB T) where the sum covers all possible diﬀusion paths. Since the defect formation enthalpy depends on the chemical potential and on the Fermi level position, the diﬀusivity will too. Furthermore, the dominant diﬀusion mechanism can change with chemical potential (oxygen partial pressure). This has been explicitly described for oxygen diﬀusion in ZnO by Erhart and Albe [129] where at low oxygen partial pressure a vacancy mechanism is prevalent while an interstitialcy mechanism dominates at higher oxygen partial pressures. In doped materials, the defect concentration can be constant over a range of temperatures and partial pressures [106,107]. Only in this case, the activation energy of diﬀusion is simply given by ∆Hm . Diﬀusion of oxygen [131–136] and zinc [62, 131, 137–145] in ZnO (selfdiﬀusion) has been investigated experimentally by a number of authors starting in the early 1950s (see also the review of experimental data in [129] for

1 ZnO and Its Applications

21

oxygen and in [130] for zinc diﬀusion). However, no consistent picture has emerged from these studies. Erhart and Albe have calculated oxygen diﬀusion in ZnO using density functional theory [129, 130]. The advantage of this approach is that both defect concentrations and migration enthalpies can be determined. The calculation shows that the migration barriers for oxygen diffusion crucially depend on the charge state of the defects and, therefore, also on the Fermi energy position in the gap, which is determined by extrinsic doping and intrinsic defects. Lowest migration barriers for oxygen are obtained by diﬀusion employing the dumbbell and the rotated dumbbell interstitials, which highlights the importance of these kinds of defects. Figure 1.17 shows the calculated diﬀusivity of oxygen in ZnO in dependence on the chemical potential and Fermi energy position in the gap at T = 1 300 K (top). At the bottom, the calculated temperature dependence of the diﬀusivity for cases where either the vacancy mechanism or the interstitialcy mechanism dominate is shown in comparison to compiled experimental data. Except for the older data points, which are probably subject to an experimental artifact (see discussion in [129] and [134]), all experimental values are well reproduced by the calculation. Erhart and Albe also calculated zinc diﬀusion in ZnO [130]. The results are displayed in Fig. 1.18 together with a comparison to experimental data. Depending on chemical potential and Fermi level position either zinc vacancy or zinc interstitial diﬀusion can dominate. In the case of n-type material, where the Fermi level is close to the conduction band, zinc diﬀusion is mostly accomplished via the vacancy mechanism. The diﬀusivities of Zn are generally larger than those of oxygen. Following the calculations, the diﬀusivities in ZnO can be ordered according to Zni > Oi > VZn > VO [130]. In particular, the migration barriers are very small (0.2–0.4 eV) for zinc interstitials [130]. The small migration barriers lead to an onset of migration already at temperatures of ∼100 K, which explains the remarkably high radiation hardness of ZnO caused by the annealing of defects at rather low temperatures [146–148].

1.7 Applications of ZnO Zinc oxide is a very old technological material. Already in the Bronze Age it was produced as a byproduct of copper ore smelting and used for healing of wounds. Early in history it was also used for the production of brass (Cu–Zn alloy). This was the major application of ZnO for many centuries before metallic zinc replaced the oxide [149]. With the start of the industrial age in the middle of the nineteenth century, ZnO was used in white paints (chinese white), in rubber for the activation of the vulcanization process and in porcelain enamels. In the following a number of existing and emerging electronic applications of ZnO are brieﬂy described.

K. Ellmer and A. Klein

Self-diffusion coefficient (cm2/s)

22

l

Re

ve

l. c

he

mi

m

ca

er

F el.

lp

ote n

R

tia

Self-diffusion coefficient (cm2/s)

i le

l

interstitialcy mechanism (I) vacancy mechanism (II)

Inverse temperature (10-15/K) Fig. 1.17. Oxygen diﬀusion in ZnO [129]. Top: Dependence of diﬀusivity on chemical potential and Fermi level at a temperature of 1 300 K illustrating the competition between vacancy and interstitialcy mechanisms. The dark grey areas indicate the experimental data range around 1 300 K. Bottom: Comparison between calculation and experiment. Experimental data from Moore and Williams [131], Hofmann and Lauder [132], Robin et al. [133], Tomlins et al. [134], Haneda et al. [135], and Sabioni et al. [136]. Solid and dashed lines correspond to regions I (interstitialcy mechanism dominant) and II (vacancy mechanism dominant) in the top graph, respectively. Copyright (2006) by the American Physical Society

23

Self- diffusivity (cm2/s)

1 ZnO and Its Applications

Re

l

l. c

ve

he

mi

i le

ca

rm

lp

ote

. el

nti

Fe

R

Self- diffusivity (cm2/s)

al

vacancy (I) interstitial (cy) (II)

Inverse temperature (10-5/K)

Fig. 1.18. Zinc diﬀusion in ZnO [130]. Top: Dependence of diﬀusivity on chemical potential and Fermi level at a temperature of 1 300 K illustrating the competition between vacancy and interstitial mechanisms. The shaded grey areas indicate the ranges selected for comparison with experimental data. Bottom: Comparison between calculation and experiment. Experimental data from Lindner [137], Secco and Moore [138,139], Moore and Williams [131], Wuensch and Tuller [143], Tomlins et al. [62], and Nogueira et al. [144, 145]. Solid and dashed lines correspond to regions I (vacancy mechanism) and II (interstitial(cy) mechanism) in the top graph, respectively. Reprinted with permission from [130]. Copyright (2006), American Institute of Physics

24

K. Ellmer and A. Klein

1.7.1 Transparent Electrodes Thin ﬁlm solar cells need a transparent window electrode for light transmission and extraction of the photocurrent. Currently used conﬁgurations for thin ﬁlm solar cells are schematically shown in Fig. 1.19. Highly-doped ZnO ﬁlms are used particularly in amorphous silicon [150] and Cu(In,Ga)(S,Se)2 [151, 152] cells. High doping levels with carrier concentrations up to 1.5 × 1021 cm−3 and resistivities as low as 2 × 10−4 Ω cm are achieved by addition of trivalent dopants like boron, aluminium, or gallium. For amorphous silicon cells, the degenerately n-doped transparent electrode forms a tunnel junction to a highly p- or n-doped material. In Cu(In,Ga)(S,Se)2 cells ZnO is a part of the electric p/n junction. To obtain high eﬃciencies, a bilayer structure of a thin (∼50 nm) nominally undoped ZnO and a highly n-doped layer is typically used. The main advantage of zinc oxide is that it is much cheaper than indium oxide, a prerequisite for large area technologies like thin ﬁlm solar cells. The role of ZnO in amorphous silicon and Cu(In,Ga)(S,Se)2 thin ﬁlm solar cells is discussed in detail in Chaps. 4, 6, 8, and 9. In display technology mostly ITO (Sn-doped In2 O3 ) is used today. However, because of the limited availability of In, there is a strong interest in replacing ITO by other materials. The advantages of ITO compared with other transparent conducting oxides are the still higher conductivities, the possibility to prepare very ﬂat ﬁlms and the good etching behavior, which enables highly reproducible structure formation. Today, most ﬂat panel

a-Si:H

organic

dye

CuInSe2

CdTe

ZnO

metal

Ag

CdS

CdTe

metal

electrolyte

a-Si:H

CIGS

CdS

organic

TiO2

ZnO,SnO /SnO ZnO 2

Mo

O/SnO2

ITO

SnO2

glass

gl glass

glass

glass

glass

(SnO2,ZnO)/a-Si

ZnO/CdS

SnO2/CdS

ITO/organic

TiO2/dye SnO2/TiO2 SnO2/Pt

n/p

n

n

p

n

SnO2 Pt dye

Fig. 1.19. Transparent conducting oxide electrodes in diﬀerent types of thin ﬁlm solar cells. TCO contacts are given at the bottom of each structure. The bottom most row indicates the doping type of the semiconductor, which is in contact to the TCO

1 ZnO and Its Applications

25

Fig. 1.20. Schematic energy band diagram of a two-layer organic light emitting diode (OLED), in which tin-doped indium oxide (ITO) is used to inject holes into the highest occupied molecular orbital (HOMO) and a low work function metal to inject electrons into the lowest unoccupied molecular orbital (LUMO)

displays are using LCD4 technology [153]. In these applications, the interface between the transparent electrode and the polymer layer inserted for orientation of the liquid crystals does not have an active electronic function in the device; i.e., it is simply used to transmit light and to apply an electric ﬁeld for reorientation of the crystals. The situation is diﬀerent in organic light emitting devices (OLED), where ITO is almost exclusively used as anode material [154]. A schematic layout together with an energy band diagram of an OLED is displayed in Fig. 1.20. In OLEDs, ITO is used to inject holes into the organic conductor. Typically an oxidizing surface treatment is performed prior to the deposition of the organic material [155–157]. These lead to an increase of the work function, which is believed to reduce the hole injection barrier. ZnO has also been tested as electrode material in OLEDs [158, 159]. 1.7.2 Varistors Varistors are voltage-dependent resistors, which are extensively used for overvoltage protection [160]. Their size varies from a few millimeters on printed electronic circuit boards for low voltage operation to more than 1 m for high voltage operation in electrical power grids. The latter are made by a stack of individual resistors of up to 10 cm diameter. ZnO varistors were ﬁrst developed by Matsuoka in Japan at the beginning of the 1970s [161]. They 4

Liquid crystal display.

26

K. Ellmer and A. Klein

are made from sintered polycrystalline ceramics using diﬀerent additives as Bi2 O3 , Sb2 O3 , or other metal oxides. The material is poorly doped and the additives segregate to the grain boundaries during sintering leading to a large barrier for electron transport [162,163]. More details on the function and the properties of ZnO varistors are described in Chap. 2. 1.7.3 Piezoelectric Devices There are several applications of ZnO that are due to its excellent piezoelectric properties [28, 164]. Examples are surface-acoustic wave (SAW) devices and piezoelectric sensors [28, 165–167]. Typically, SAW devices are used as band pass ﬁlters in the tele-communications industry, primarily in mobile phones and base stations. Emerging ﬁeld for SAW devices are sensors in automotive applications (torque and pressure sensors), medical applications (chemical sensors), and other industrial applications (vapor, humidity, temperature, and mass sensors). Advantages of acoustic wave sensors are low costs, ruggedness, and a high sensitivity. Some sensors can even be interrogated wirelessly, i.e., such sensors do not require a power source. In an SAW device, a mechanical deformation is induced by electrical contact ﬁngers in a nearly isolating, highly (0001)-textured ZnO ﬁlm (see insert in Fig. 1.21). The insulating wave travels along the ZnO ﬁlm surface with the velocity of sound in ZnO and is detected at the end of the device by another metal–ﬁnger contact. High-frequency electrical signals (10 MHz– 10 GHz) can be transformed to SAWs with typical wave velocities of about 3 km s−1 . Because of the much lower acoustic velocity, compared with the velocity of light, such SAW devices can also be used as acoustic delay lines with a characteristic frequency dependence suitable for high-frequency ﬁlters. A typical frequency curve of an SAW device is shown in Fig. 1.21. 1.7.4 Phosphors In displays ZnO powders are used as green phosphors [169]. Recently, magnetron sputtered ﬁlms of ZnO-based compounds, for instance Zn2 SiO4 :Mn or ZnGa2 O4 :Mn were used as green phosphors in thin-ﬁlm electroluminescence displays [170, 171]. Even white cathodoluminescence was observed for self-assembled ZnO micropatterns [172]. 1.7.5 Transparent Oxide Thin Film Transistors Recently, ﬁeld-eﬀect transistors based on zinc oxide were reported [173, 174], opening the opportunity to design microelectronic devices that are transparent and/or work at high temperatures [175]. More details on thin ﬁlm transistors employing transparent conducting oxides are given in Chap. 2.

1 ZnO and Its Applications

27

Fig. 1.21. Frequency spectrum of a 10 µm wavelength SAW device on a 1.5 µm thick ZnO ﬁlm on an r-sapphire substrate. The ZnO ﬁlm was deposited by MOCVD using diethyl-zinc and oxygen. The inset shows the geometry of the device with the interdigitated contact ﬁngers. Reprinted with permission from [168]

1.7.6 Spintronics Another prospective application of zinc oxide is the alloying with magnetic atoms like manganese, cobalt, or nickel to prepare diluted magnetic semiconducting alloys that are interesting as materials for spintronics, promising the possibility to use the spin of the electrons for electronic devices [176]. Acknowledgement. The authors are grateful to Paul Erhart and Karsten Albe for extensive discussions of the defect properties of ZnO and for the supply of original versions of their ﬁgures.

References 1. 2. 3. 4.

J. Bardeen, W.H. Brattain, Phys. Rev. 74, 230 (1948) A.R. Hutson, Phys. Rev. Lett. 4, 505 (1960) F.S. Hickernell, Proc. IEEE 64, 631 (1976) C. Klingshirn, H. Priller, M. Decker, J. Br¨ uckner, H. Kalt, R. Hauschild, J. Zeller, A. Waag, A. Bakin, H. Wehmann, K. Thonke, R. Sauer, R. Kling, F. Reuss, C. Kirchner, Adv. Solid State Phys. 45, 275 (2005) 5. C. Klingshirn, M. Grundmann, A. Hoﬀmann, B. Meyer, A. Waag, Phys. J. 5, 33 (2006)

28

K. Ellmer and A. Klein

6. G. Heiland, E. Mollwo, F. St¨ ockmann, Solid State Phys. 8, 191 (1959) 7. W. Hirschwald, in Current Topics in Material Science, vol. 7, ed. by E. Kaldis (North-Holland, Amsterdam, 1981), pp. 144–482 8. C. Klingshirn, H. Haug, Phys. Rep. 70, 315 (1981) ¨ ur, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, 9. U. Ozg¨ V. Avrutin, S.J. Cho, H. Morkoc, J. Appl. Phys. 98, 041301 (2005) 10. N.H. Nickel, E. Terukov (eds.), Zinc Oxide-A Material for Micro- and Optoelectronic Applications (Springer, Dordrecht, 2005) 11. C. Jagadish, S.J. Pearton (eds.), Zinc Oxide: Bulk, Thin Films and Nanostructures (Elsevier, Oxford, 2006) 12. K. Nomura, H. Ohta, K. Ueda, T. Kamiya, M. Hirano, H. Hosono, Science 300, 1269 (2003) 13. D.C. Look, Semicond. Sci. Technol. 20, S55 (2005) 14. B.K. Meyer, J. Sann, A. Zeuner, Adv. Solid State Phys. 45, 289 (2005) 15. B. Claﬂin, D.C. Look, S.J. Park, G. Cantwell, J. Cryst. Growth 287, 16 (2006) 16. R.G.W. Wyckoﬀ, Crystal Structures (Interscience, New York, 1963) 17. A.N. Mariano, R.E. Hanneman, J. Appl. Phys. 34, 384 (1963) 18. http://minerals.usgs.gov/minerals/pubs/commodity/zinc/ 19. W.L. Bragg, Phil. Mag. 39, 647 (1920) 20. F. Decremps, J. Zhang, R.C. Lieberman, Europhys. Lett. 51, 268 (2000) 21. S. Desgreniers, Phys. Rev. B 58, 14102 (1998) 22. A. Dargys, J. Kundrotas, Handbook of Physical Properties of Ge, Si, GaAs and InP (Science Encycl. Publ., Vilnius, 1994) 23. O. Madelung (ed.), Semiconductors Basic Data (2nd ed.) (Springer Verlag, Berlin, 1996) 24. H.L. Hartnagel, A.L. Dawar, A.K. Jain, C. Jagadish, Semiconducting Transparent Thin Films (Institute of Physics Publishing, Bristol, 1995) 25. A.A. Khan, Acta Cryst. A 24, 403 (1968) 26. R.R. Reeber, J. Appl. Phys. 41, 5063 (1970) 27. S. Adachi, Properties of Group IV, III-V and II-VI Semiconductors (John Wiley & Sons, West Sussex, England, 2005) 28. Y. Lu, N.W. Emanetoglu, Y. Cheng, in Zinc Oxide. Bulk, Thin ﬁlms and Nanostructures, ed. by C. Jagadish, S.J. Pearton (Elsevier, Oxford, 2006), pp. 443–489 29. R. Helbig (ed.), Landolt B¨ ornstein, New Series III/17d; II-VI and I-VII Compounds (Springer, Berlin Heidelberg Newyork, 1984) 30. I. Hamberg, C.G. Granqvist, J. Appl. Phys. 60, R123 (1986) 31. H. Fujiwara, M. Kondo, Phys. Rev. B 71, 075109 (2005) 32. M. Chase, JANAF Thermochemical Tables, 4th edn. J. Phys. Chem. Ref. Data, Monograph 9 (NIST, 1998) 33. D.S. Ginley, C. Bright, MRS Bull. 25(Aug), 15 (2000) 34. T. Minami, Semicond. Sci. Technol. 20, S35 (2005) 35. R.L. Weiher, R.P. Ley, J. Appl. Phys. 37, 299 (1966) 36. I. Hamberg, C.G. Granqvist, K.F. Berggren, B.E. Sernelius, L. Engstr¨om, Phys. Rev. B 30, 3240 (1984) 37. Y. Gassenbauer, R. Schafranek, A. Klein, S. Zafeiratos, M. H¨avecker, A. Knop-Gericke, R. Schl¨ ogl, Phys. Rev. B 73, 245312 (2006) 38. S.P. Harvey, T.O. Mason, Y. Gassenbauer, R. Schafranek, A. Klein, J. Phys. D: Appl. Phys. 39, 3959 (2006)

1 ZnO and Its Applications

29

39. P. Erhart, A. Klein, R.G. Egdell, K. Albe, Phys. Rev. B 75, 153205 (2007) 40. K. B¨ adeker, Ann. Phys. (Leipzig) 22, 749 (1907) 41. T.O. Mason, D.R. Kammler, B.J. Ingram, G.B. Gonzalez, D.L. Young, T.J. b. Coutts, Thin Solid Films 445, 186 (2003) 42. F.P. Koﬀyberg, Phys. Rev. B 13, 4470 (1976) 43. C. McGuinness, C.B. Stagarescu, P.J. Ryan, J.E. Downes, D. Fu, K.E. Smith, R.G. Egdell, Phys. Rev. B 68, 165104 (2003) 44. Z. Zhao, D.L. Morel, C.S. Ferekides, Thin Solid Films 413, 203 (2002) 45. A.J. Freeman, K.R. Poeppelmeier, T.O. Mason, R.P.H. Chang, T.J. Marks, MRS Bull. 25(Aug), 45 (2000) 46. T. Minami, MRS Bull. 25(Aug), 38 (2000) 47. H. Kawazoe, M. Yasukawa, H. Hyodo, M. Kurita, H. Yanagi, H. Hosono, Nature 389, 939 (1997) 48. A. Kudo, H. Yanagi, H. Hosono, H. Kawazoe, Appl. Phys. Lett. 73, 220 (1998) 49. H. Kawazoe, H. Yanagi, K. Ueda, H. Hosono, MRS Bull. 25(Aug), 28 (2000) 50. K. Ueda, S. Inoue, S. Hirose, H. Kawazoe, H. Hosono, Appl. Phys. Lett. 77, 2701 (2000) 51. H. Yanagi, T. Hase, S. Ibuki, K. Ueda, H. Hosono, Appl. Phys. Lett. 78, 1583 (2001) 52. H. Yanagi, K. Ueda, H. Ohta, M. Orita, M. Hirano, H. Hosono, Solid State Commun. 121, 15 (2001) 53. H. Yanagi, J. Tate, S. Park, C.H. Park, D.A. Keszler, Appl. Phys. Lett. 82, 2814 (2003) 54. B.J. Ingram, T.O. Mason, R. Asahi, K.T. Park, A.J. Freeman, Phys. Rev. B 64, 155114 (2001) 55. B.J. Ingram, B.J. Harder, N.W. Hrabe, T.O. Mason, Chem. Mater. 16, 5623 (2004) 56. X. Nie, S.H. Wei, S.B. Zhang, Phys. Rev. Lett. 88, 066405 (2002) 57. R. Helbig, J. Cryst. Growth 15, 25 (1972) 58. T.B. Massalski, H. Okamoto, P.R. Subramanian, L. Kacprzak (eds.), Binary Alloy Phase Diagrams, Vol. 3 (ASM International, Materials Park, Ohio, 1990) 59. N.G. Vannerberg, Ark. Kemi 14, 119 (1959) 60. D.F. Anthrop, A.W. Searcy, J. Phys. Chem. 68, 2335 (1964) 61. K.I. Hagemark, P.E. Toren, J. Electrochem. Soc. 122, 992 (1975) 62. G.W. Tomlins, J.L. Routbort, T.O. Mason, J. Appl. Phys. 87, 117 (2000) 63. K. Lott, S. Shinkarenko, T. Kirsanova, L. T¨ urn, E. Gorohova, A. Grebennik, A. Vishnjakov, Phys. Stat. Sol. (c) 1, 2237 (2004) 64. O. Fritsch, Ann. Phys. [5], 22, 375 (1935) 65. D. Rykl, J. Bauer, Kristall und Technik 3, 375 (1968) 66. K. Maeda, M. Sato, I. Niikura, T. Fukuda, Semicond. Sci. Technol. 20, S49 (2005) 67. G. Bogner, E. Mollwo, J. Phys. Chem. Solids 6, 136 (1958) 68. M. Hirose, Y. Furuya, I. Kubo, Jpn. J. Appl. Phys. 9, 726 (1970) 69. D.F. Bliss, in Encyclopedia of Materials: Science and Technology, vol. 10, ed. by K.H.J. Buschow, R. Cahn, M. Flemings, B. Ilschner, E. Kramer, S. Mahajan (Elsevier, Oxford, 2001), pp. 9888–9891 70. J. Nause, B. Nemeth, Semicond. Sci. Technol. 20, S45 (2005) 71. D. Schulz, S. Ganschow, D. Klimm, M. Neubert, M. Roßberg, M. Schmidbauer, R. Fornari, J. Cryst. Growth 296, 27 (2006)

30

K. Ellmer and A. Klein

72. D.C. Look, D.C. Reynolds, J.R. Sizelove, R.L. Jones, C.W. Litton, G. Cantwell, W.C. Harsch, Solid State Commun. 105, 399 (1998) 73. E.V. Kortunova, P.P. Chvanski, N.G. Nikolaeva, J. Phys. IV France 126, 39 (2005) 74. K. Ellmer, J. Phys. D: Appl. Phys. 33, R17 (2000) 75. J.R. Bellingham, W.A. Phillips, C.J. Adkins, J. Phys.: Cond. Mat. 2, 6207 (1990) 76. P. Kuppusami, G. Vollweiler, D. Rafaja, K. Ellmer, Appl. Phys. A 81, 183 (2005) 77. E. Scharowsky, Z. Phys. 135, 318 (1953) 78. K. Yamada, S. Tobisawa, J. Appl. Phys. 66, 5309 (1989) 79. J. Tornow, K. Schwarzburg, J. Phys. Chem. C 111, 8692 (2007) 80. M.H. Huang, Y. Wu, H. Feick, N. Tran, E. Weber, P. Yang, Adv. Mater. 13, 113 (2001) 81. R.B. Peterson, C.L. Fields, B.A. Gregg, Langmuir 20, 5114 (2004) 82. L. Wischmeier, T. Voss, S. Brner, W. Schade, Appl. Phys. A 84, 111 (2006) 83. G.C. Yi, V. Wang, W.I. Park, Semicond. Sci. Technol. 20, S22 (2005) 84. M. Huang, S. Mao, H. Feick, H. Yan, Y. Wu, H. Kind, E. Weber, R. Russo, P. Yang, Science 292, 1897 (2001) 85. Y.W. Heo, F. Ren, D.P. Norton, in Zinc Oxide: Bulk, Thin Films and Nanostructures, ed. by C. Jagadish, S. Pearton (Elsevier, Oxford, 2006), pp. 491–523 86. T. Soga (ed.), Nanostructured Materials for Solar Energy Conversion (Elsevier, Oxford, 2006) 87. M.S. White, D.C. Olson, S.E. Shaheen, N. Kopidakis, D.S. Ginley, Appl. Phys. Lett. 89, 143517 (2006) 88. X. Wang, J. Song, J. Liu, Z.L. Wang, Science 316, 102 (2007) 89. W.A. Harrison, Electronic Structure and the Properties of Solids (Dover Publications, New York, 1989) 90. P.Y. Yu, M. Cardona, Fundamentals of Semiconductors. Physics and Materials Properties, 3rd edn. (Springer, Berlin Heidelberg Newyork, 2001) 91. J.L. Shay, J.H. Wernick, Ternary Chalcopyrite Semiconductors (Pergamon, Oxford, 1975) 92. B.K. Meyer, H. Alves, D.M. Hofmann, W. Kriegseis, D. Forster, F. Bertram, J. Christen, A. Hoﬀmann, M. Straßburg, M. Dworzak, U. Haboeck, A.V. Rodina, Phys. Stat. Sol. (a) 241, 231 (2004) 93. J.E. Jaﬀe, A. Zunger, Phys. Rev. B 29, 1882 (1984) 94. S.H. Wei, A. Zunger, Phys. Rev. B 37, 8958 (1988) 95. S.H. Wei, A. Zunger, Appl. Phys. Lett. 72, 2011 (1998) 96. D. Vogel, P. Kr¨ uger, J. Pollmann, Phys. Rev. B 52, R14316 (1995) 97. D. Vogel, P. Kr¨ uger, J. Pollmann, Phys. Rev. B 54, 5495 (1996) 98. C.L. Dong, C. Persson, L. Vayssieres, A. Augustsson, T. Schmitt, M. Mattesini, R. Ahuja, C.L. Chang, J.H. Guo, Phys. Rev. B 70, 195325 (2004) 99. A. Ohtomo, M. Kawasaki, T. Koida, K. Masubuchi, H. Koinuma, Y. Sakurai, Y. Yoshida, T. Yasuda, Y. Segawa, Appl. Phys. Lett. 72, 2466 (1998) 100. A. Ohtomo, M. Kawasaki, I. Ohkubo, H. Koinuma, T. Yasuda, Y. Segawa, Appl. Phys. Lett. 75, 980 (1999) 101. T. Makino, Y. Segawa, M. Kawasaki, A. Ohtomo, R. Shiroki, K. Tamura, T. Yasuda, H. Koinuma, Appl. Phys. Lett. 78, 1237 (2001)

1 ZnO and Its Applications

31

102. S. Choopun, R.D. Vispute, W. Yang, R.P. Sharma, T. Venkatesan, Appl. Phys. Lett. 80, 1529 (2002) 103. R. Schmidt, B. Rheinl¨ ander, M. Schubert, D. Spemann, T. Butz, J. Lenzner, E.M. Kaidashev, M. Lorenz, A. Rahm, H.C. Semmelhack, M. Grundmann, Appl. Phys. Lett. 82, 2260 (2003) 104. R. Schmidt-Grund, M. Schubert, B. Rheinl¨ander, D. Fritsch, H. Schmidt, E.M. Kaidashev, C.M. Herzinger, M. Grundmann, Thin Solid Films 455–456, 500 (2004) 105. G.V. Rao, F. S¨ auberlich, A. Klein, Appl. Phys. Lett. 87, 032101 (2005) 106. F.A. Kr¨ oger, The Chemistry of Imperfect Crystals (North-Holland, Amsterdam, 1964) 107. J. Maier, Physical Chemistry of Ionic Materials (Wiley-VCH, Weinheim, 2004) 108. A.F. Kohan, G. Ceder, D. Morgan, C.G. van de Walle, Phys. Rev. B 61, 15019 (2000) 109. S.B. Zhang, S.H. Wei, A. Zunger, Phys. Rev. B 63, 075205 (2001) 110. F. Oba, S.R. Nishitani, S. Isotani, H. Adachi, I. Tanaka, J. Appl. Phys. 90, 824 (2001) 111. H. Overhof, J.M. Spaeth, Phys. Rev. B 72, 115205 (2005) 112. S.B. Zhang, J.E. Northrup, Phys. Rev. Lett. 67, 2339 (1991) 113. E.C. Lee, Y.S. Kim, Y.G. Jin, K.J. Chang, Phys. Rev. B 64, 085120 (2001) 114. P. Erhart, A. Klein, K. Albe, Phys. Rev. B 72, 085213 (2005) 115. P. Erhart, K. Albe, A. Klein, Phys. Rev. B 73, 205203 (2006) 116. A. Janotti, C.G. van de Walle, J. Cryst. Growth 287, 58 (2006) 117. S. Lany, A. Zunger, Phys. Rev. Lett. 98, 045501 (2007) 118. W. Walukiewicz, Physica B 302–303, 123 (2001) 119. S.B. Zhang, S.H. Wei, A. Zunger, J. Appl. Phys. 83, 3192 (1998) 120. J.A. van Vechten, J.D. Zook, R.D. Horning, B. Goldenberg, Jpn. J. Appl. Phys. 31, 3662 (1992) 121. C. Agashe, O. Kluth, J. H¨ upkes, U. Zastrow, B. Rech, M. Wuttig, J. Appl. Phys. 95, 1911 (2004) 122. G. Frank, H. K¨ ostlin, Appl. Phys. A 27, 197 (1982) 123. S. Lany, A. Zunger, Phys. Rev. B 72, 035215 (2005) 124. C.G. van de Walle, Phys. Rev. Lett. 85, 1012 (2000) 125. S.F.J. Cox, E.A. Davis, S.P. Cottrell, P.J.C. King, J.S. Lord, J.M. Gil, H.V. Alberto, R.C. Vilo, J.P. Duarte, N.A. de Campos, A. Weidinger, R.L. Lichti, S.J.C. Irvine, Phys. Rev. Lett. 86, 2601 (2001) 126. H. Moormann, D. Kohl, G. Heiland, Surf. Sci. 100, 302 (1980) 127. L.Y. Chen, W.H. Chen, J.J. Wang, F.C.N. Hong, Y.K. Su, Appl. Phys. Lett. 85, 5628 (2004) 128. G.A. Shi, M. Saboktakin, M. Stavola, S.J. Pearton, Appl. Phys. Lett. 85, 5601 (2004) 129. P. Erhart, K. Albe, Phys. Rev. B 73, 115207 (2006) 130. P. Erhart, K. Albe, Appl. Phys. Lett. 88, 201918 (2006) 131. W.J. Moore, E.L. Williams, Discuss. Faraday Soc. 28, 86 (1959) 132. J.W. Hoﬀman, I. Lauder, Trans. Faraday Soc. 66, 2346 (1970) 133. R. Robin, A.R. Cooper, A.H. Heuer, J. Appl. Phys. 44, 3770 (1973) 134. G.W. Tomlins, J.L. Routbort, T.O. Mason, J. Am. Ceram. Soc. 81, 869 (1998) 135. H. Haneda, I. Sakaguchi, A. Watanabe, T. Ishigaki, J. Tanaka, J. Electroceram. 4, 41 (1999)

32 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151.

152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171.

K. Ellmer and A. Klein A.C.S. Sabioni, M.J.F. Ramos, W.B. Ferraz, Mater. Res. 6, 173 (2003) R. Lindner, Acta Chem. Scand. 6, 457 (1952) E.A. Secco, W.J. Moore, J. Chem. Phys. 23, 1170 (1955) E.A. Secco, W.J. Moore, J. Chem. Phys. 26, 942 (1957) D.G. Thomas, J. Phys. Chem. Solids 3, 229 (1957) J.P. Roberts, C. Wheeler, Phil. Mag. 2, 708 (1957) J.P. Roberts, C. Wheeler, Trans. Faraday Soc. 56, 570 (1960) B.J. Wuensch, H.L. Tuller, J. Phys. Chem. Solids 55, 975 (1994) M.A. Nogueira, W.B. Ferraz, A.C.S. Sabioni, Mater. Res. 6, 167 (2003) M.A.N. Nogueira, A.C.S. Sabioni, W.B. Ferraz, Defect Diﬀus. Forum 237–240, 163 (2005) C. Coskun, D.C. Look, G.C. Farlow, J.R. Sizelove, Semicond. Sci. Technol. 19, 752 (2004) Y.V. Gorelkinskii, G.D. Watkins, Phys. Rev. B 69, 115212 (2004) K. Lorenz, E. Alves, E. Wendler, O. Bilani, W. Wesch, M. Hayes, Appl. Phys. Lett. 87, 191904 (2005) H.E. Brown, Zinc Oxide. Properties and Applications (Int. Lead Zinc Research Org., New York, 1976) B. Rech, H. Wagner, Appl. Phys. A 69, 155 (1999) K. Ramanathan, M.A. Contreras, C.L. Perkins, S. Asher, F.S. Hasoon, J. Keane, D. Young, M. Romero, W. Metzger, R. Nouﬁ, J. Ward, A. Duda, Progr. Photovolt.: Res. Appl. 11, 225 (2003) R. Klenk, J. Klaer, R. Scheer, M.C. Lux-Steiner, I. Luck, N. Meyer, U. R¨ uhle, Thin Solid Films 480–481, 509 (2005) M. Katayama, Thin Solid Films 341, 140 (1999) L.S. Hung, C.H. Chen, Mater. Sci. Eng. R 39, 143 (2002) T. Kugler, W.R. Salaneck, H. Rost, A.B. Holmes, Chem. Phys. Lett. 310, 391 (1999) K. Sugiyama, H. Ishii, Y. Ouchi, K. Seki, J. Appl. Phys. 87, 295 (2000) Y. Gassenbauer, A. Klein, J. Phys. Chem. B 110, 4793 (2006) X. Jiang, F.L. Wong, M.K. Fung, S.T. Lee, Appl. Phys. Lett. 83, 1875 (2003) S.H.K. Park, J.I. Lee, C.S. Hwang, H.Y. Chu, Jpn. J. Appl. Phys. 44, L242 (2005) L.M. Levinson, H.R. Philipp, Ceram. Bull. 65, 639 (1986) M. Matsuoka, Jpn. J. Appl. Phys. 10, 736 (1971) M. Rossinelli, G. Blatter, F. Greuter, in Electrical Ceramics, vol. 36, ed. by B.C.H. Steele (Inst. of Ceramics, Shelton, 1985), pp. 1–17 G. Blatter, F. Greuter, Phys. Rev. B 34, 8555 (1986) A. Ballato, in IEEE Ultrasonics Symposium Proceedings (1996), pp. 575–583 B. Drafts, Sensors 17(October), 68 (2000) J.W. Grate, S.J. Martin, R.W. White, Analyt. Chem. 65, 940 (1993) J.W. Grate, S.J. Martin, R.W. White, Analyt. Chem. 65, 987 (1993) C. Gorla, N.W. Emanetoglu, S. Liang, W.E. Mayo, H.S. Y. Yu, M. Wraback, J. Appl. Phys. 85, 2595 (1999) K. Vanheusdalen, C.H. Seager, W.L. Warren, D.R. Tallant, J.A. Voigt, Appl. Phys. Lett. 68, 403 (1996) A.H. Kitai, Thin Solid Films 445, 367 (2003) T. Miyata, Y. Mochizuki, T. Minami, IEICE Trans. Electron. E88-C, 2065 (2005)

1 ZnO and Its Applications

33

172. N. Saito, H. Haneda, T. Sekiguchi, N. Ohashi, I. Sakaguchi, K. Koumoto, Adv. Mater. 14, 418 (2002) 173. K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, H. Hosono, Nature 432, 488 (2004) 174. H.Q. Chiang, J.F. Wager, R.L. Hoﬀman, J. Jeong, D.A. Keszler, Appl. Phys. Lett. 86, 013503 (2005) 175. G. Thomas, Nature 389, 907 (1997) 176. I. Satoh, T. Kobayashi, K. Katayma, T. Okada, T. Itoh, Appl. Phys. A 79, 1445 (2004)

2 Electrical Properties K. Ellmer

Electrical investigations of zinc oxide date back to 1912, when Somerville measured the resistivity of cylindrical ZnO rods up to temperatures of 1,200◦ C [1]. The reported data at the beginning of the twentieth century scattered significantly, depending on the preparation of the samples and the used electrical contacts. Today we know that this was caused by the diﬃcult preparation of compact zinc oxide samples and by the missing understanding of semiconductors at that time. At the beginning of the 1930s, Wagner investigated the dependence of the electrical properties of oxides on their stoichiometry. These investigations were devoted to the proof of the theory of defect equilibria by Schottky and Wagner for ionic crystals. Wagner and Schottky [2] as well as Hauﬀe and Block [3] could show that the oxygen content of the crystals, varied by annealing at diﬀerent oxygen partial pressures, strongly determines the electrical carrier concentration and hence the conductivity due to a reaction balance between oxygen vacancies (2.1) or interstitial zinc atoms (2.2) and electrons in the conduction band according to: O2 (gas) + 2Zn(l) + 2O··(v) + 4e ⇐⇒ 2ZnO O2 (gas) + 2Zn·· (i) + 4e ⇐⇒ 2ZnO

(2.1) (2.2)

where l, v, and i mean lattice place, vacancy, and interstitial position, which gives, by applying the law of mass action: [Zn·· (i)]2 [e]4 pO2 = const.

(2.3)

Assuming that only the doubly charged zinc interstitials (or oxygen vacancies) contribute to the excess electrons leads to the following oxygen partial pressure dependence of the conductivity: −1/6

σ = [e]const. = pO2 const.”

(2.4)

Von Baumbach and Wagner [4] argued that the zinc interstitial is more probable because of the smaller ionic radius of the Zn++ ion (74 pm) compared with the oxygen ion (138 pm). What could not be decided for decades was the question whether the oxygen vacancy or the zinc interstitial constitutes the donor [5], see Sect. 2.1.1.1. Wagner and also Hauﬀe found a signiﬁcant dependence of the conductivity on the oxygen partial pressure; however, the exponent in (2.4) was not

K. Ellmer

conductivity [S/cm]

36

10

ZnO:Ga

0.1

ZnO:Al

1

ZnO:Cr

0.01 ZnO

0.001 0.1

1

10

100

1000

oxygen partial pressure [mbar] Fig. 2.1. Dependence of the conductivity on the oxygen partial pressure for undoped ZnO and ZnO-doped with Cr, Al, and Ga according to Hauﬀe and Block [3]. Parameters: ZnO (T = 665◦ C), ZnO + 0.5 mol%Cr2 O3 (T = 665◦ C), ZnO + 1 mol%Al2 O3 (T = 800◦ C), ZnO + 1 mol%Ga2 O3 (T = 800◦ C). A ﬁt according to (2.4) yields the following exponents n: ZnO (n = −1/4), ZnO:Cr (n = −1/5.6), ZnO:Al (n = −1/10.5), ZnO:Ga (n = −1/13.9)

n = −1/6 but n = −1/4 to −1/5.6 (see Fig. 2.1), i.e., lower than expected, and was ascribed to the porosity of the samples, which exhibited only 60–70 % of the bulk ZnO density. Adding other metal oxides (Al2 O3 , Cr2 O3 , Ga2 O3 ) to zinc oxide increased the conductivity by orders of magnitude and changed the exponent n from n = −1/4 to n = −1/10 to n = −1/14 (see Fig. 2.1). In conclusion, it was found that the conductivity of ZnO depends signiﬁcantly on the stoichiometry, adjusted by the oxygen or zinc partial pressure during growth or annealing. Also, the reducing eﬀect of a hydrogen containing atmosphere was discovered [4, 6]. Both facts play an important role for the growth of ZnO thin ﬁlms, discussed in Sect. 2.2.

2.1 Electrical Properties of ZnO Single Crystals The electrical parameters, especially the conductivity, were investigated early in history of ZnO research. An overwiew over electronic properties of ZnO up to the end of the 1950s was given by Heiland et al. in 1959 [7]. Most of the early investigations up to about 1955 were performed on sintered polycrystalline ZnO samples [4,8], which suﬀered from the general problem of conduction in porous, inhomogeneous materials with a lower density compared with

2 Electrical Properties

37

the corresponding bulk material. Fritsch, the pioneer of zinc oxide research, reported already in 1935 comprehensive measurements of conductivity, Hall eﬀect and thermovoltage on polycrystalline zinc oxide plates, on single crystalline ZnO needles and also on evaporated zinc oxide ﬁlms [9]. One of the aims of his work was to verify the theory of electronic conduction developed at that time. His samples, which were sintered in air at temperatures between 1,400 and 1,500◦C from ZnO powder, had very high densities of 98–99 % of the bulk density. Both, polycrystalline as well as ZnO single crystalline needles exhibited resistivities of about 0.2–0.7 Ω cm at room temperature. Annealing at 900◦ C in oxygen increased the resistivities by about two orders of magnitude. The annealed crystals exhibited resistivities up to 2.5 × 105 Ω cm. The temperature dependence of the conductivity was measured by Fritsch between room temperature and the temperature of liquid air (−190◦C). Some of his results are reproduced in Fig. 2.2. The conductivity curves can be divided into three groups belonging to: 1. Polycrystalline samples in as prepared state 2. Polycrystalline plates, annealed in oxygen 3. Single crystalline needles, annealed in oxygen All samples show the typical behavior of a semiconductor, i.e., decreasing conductivity with decreasing temperature. The diﬀerent sample groups

10 10 10 σ [S/cm]

10 10 10 10 10 10 10

1 0

1

-1 -2 -3 -4

2

-5 -6 -7

3

-8

2

4

6

8

10

12

1000/T [K] Fig. 2.2. Temperature-dependent conductivity of diﬀerent ZnO samples, measured by Fritsch in 1935 [9]. All samples were not intentionally doped. (1) and (2) are as grown and annealed polycrystalline ZnO samples, respectively. The lowest conductivities and the largest activation energies exhibit the annealed single crystals (3)

38

K. Ellmer

mentioned earlier exhibit diﬀerent activation energies in the range of 7–17 meV (group 1), 60–80 meV (group 2), and 200–400 meV (group 3). On the larger sintered polycrystalline samples, Fritsch also performed Hall measurements yielding mobility values at room temperature between 7 and 30 cm2 V−1 s−1 . The Hall voltages were negative, i.e., the samples exhibited n-type conductivity. From these measurements carrier densities of about 1 × 1018 cm−3 for as prepared ZnO plates and 2 × 1016 cm−3 for oxygen annealed samples can be inferred, the latter value being quite comparable to ZnO single crystals commercially available today. The Seebeck coeﬃcient of the polycrystalline samples was also negative, in accordance with the Hall measurements. These results of Fritsch are remarkable, taking into consideration the experimental possibilities and the knowledge about semiconductors at that time. Further progress could be achieved only in the 1950s, when better crystals could be grown and the foundations of semiconductor physics were widely known [10]. In the following section, only measurements on single crystals are taken into consideration. Synthetic single crystals were grown already in the 1950s by chemical vapor transport [11–13] as well as by hydrothermal growth [14]. Typical crystal dimensions were some millimeter in all three dimensions. Often needle-like crystals grown in c-direction were used [9, 11], see also Chap. 1. 2.1.1 Dopants in ZnO Single Crystals 2.1.1.1 Intrinsic Dopants The ﬁrst crystals, which were electrically investigated, were not intentionally doped [9, 15–17]. As already explained in the introduction, the crystals were well conducting because of deviations from stoichiometry. This means, intrinsic defects in zinc oxide – either zinc interstitials or oxygen vacancies – constitute intrinsic dopants in ZnO. Annealing the crystals in an oxygen atmosphere produced crystals of higher resistivity. On the contrary annealing in vacuum, under reducing conditions (hydrogen atmosphere) or in zinc vapor produced better conducting specimens. Recently, Look et al. performed a study on very pure ZnO single crystals in which intrinsic defects were generated by high-energy (MeV) electron irradiation [18]. Since the defect production was higher when the crystals were irradiated in (0001) direction (Zn-face up) compared with electron bombardment along the (000¯1) direction (O-face up), it was concluded that the zinc interstitials are the intrinsic shallow donors. Tomlins et al. [19], on the other hand, investigated the self-diﬀusion of zinc in ZnO single crystals as well as electrical transport properties. They found that the zinc diﬀusion is most likely controlled by a vacancy mechanism. Taking their electrical measurements into consideration they argue that oxygen vacancies are the intrinsic defects leading to n-type conductivity of zinc oxide. Recently, Erhart et al. [20–22] performed comprehensive density-functional calculations of defect formation energies and diﬀusion constants in ZnO (see also description of intrinsic defect structure

2 Electrical Properties

39

in Chap. 1). Erhart et al. assign the zinc vacancy and the oxygen interstitials as acceptors, while only zinc interstitials and oxygen vacancies have energetic shallow donor positions in the band gap. Depending on the growth atmosphere diﬀerent defects are most likely: Under zinc-rich conditions the oxygen vacancy dominates, while the zinc vacancy and the oxygen interstitials are the most likely defects for an oxygen-rich atmosphere. Taking the lower energy of formation of the zinc interstitial, it should be the dominant donor in zinc oxide. 2.1.1.2 Extrinsic Dopants Zinc oxide is easily n-type doped, while p-type doping is still a problem (see Sect. 2.5 and 1.6.1). Already in the 1950s, the role of Al, Cr, Ga, and In as n-type dopants in ZnO was found. Hauﬀe and coworkers [23] investigated extensively the relation between the electrical parameters and the addition of metal oxides to zinc oxide [3]. The addition of small amounts of oxides of a diﬀerent valency causes changes of the conductivity by orders of magnitude, for instance when the group III-oxides B2 O3 , Al2 O3 , In2 O3 , or Ga2 O3 are added. It is assumed that the group III-dopant atoms are built in onto zinc lattice sites, spending the additional electron not required for the bonding to the conduction band according to the equation (M means metal) [24, 25]: M2 O3 ⇐⇒ 2MZn + 2e + 2O0 + 1/2O2

(2.5)

leading to the following oxygen partial pressure dependence: −1/8

[e] = [MZn ] ∼ pO2

.

(2.6)

Also other oxides of metals with three valence electrons (for instance Cr2 O3 , Y2 O3 , Ce2 O3 ), but also TiO2 , ZrO2 , HfO2 can be used to increase the conductivity of ZnO [3, 26]. The preparation of donor-doped zinc oxide crystals at high oxygen partial pressures or annealing in oxygen atmosphere leads to highly resistive crystals. One explanation for this eﬀect is the formation of zinc vacancies, caused by their low enthalpy of formation under oxygen-rich conditions, which are acceptors thus compensating the donors (see also Chap. 1). Another possibility is the oxidation of the dopant metal (for instance gallium) to the corresponding oxide, which is supported by nuclear magnetic resonance (NMR) measurements of the group of Sleight 2− as double-electron traps, which can be viewed as a who suggested Ga3+ 2 Oi precursor of gallium oxide [27]. 2.1.1.3 Hydrogen in Zinc Oxide Hydrogen is a very important dopant in zinc oxide, since it is present during almost all growth processes, introducing a background doping in single crystals and in ﬁlms. The role of hydrogen as a shallow donor was already

40

K. Ellmer

found at the beginning of the zinc oxide research. Mollwo [6], Thomas and Lander [13, 28], and Hutson [17] reported in the 1950s the doping eﬀect of hydrogen. Unfortunately, this knowledge was forgotten (or ignored) for some decades. Only in the year 2000 hydrogen as a donor in ZnO was “rediscovered” by van de Walle by a density functional theory study [29], which showed that under all circumstances H acts as a donor in ZnO. Shortly after, it was unambiguously proved that hydrogen constitutes a shallow donor in zinc oxide [30]. The role of hydrogen in zinc oxide is peculiar compared with other semiconductors since it is a donor, independently on the position of the Fermi level, diﬀerent from the behavior of H in silicon or other semiconductors. Recently, Robertson et al. [31, 32] calculated the band structures of diﬀerent semiconductors, especially of oxides, and the position of the hydrogen defect level with respect to the valence and conduction bands, respectively. His results are comprised in Fig. 2.3 and show that only ZnO, SnO2 , and SrTiO3 exhibit an H defect level in the conduction bands. The concentration of hydrogen in ZnO varies between 5×1016 cm−3 in single crystals grown by chemical transport and 1 × 1020 cm−3 in magnetron sputtered ZnO:Al ﬁlms [33]. The n-type dopants constitute eﬀective-mass-like (hydrogenic) donors with ionization energies around ED = 13.595 eV · m∗ /ε2 ∼ 66 meV, which is conﬁrmed by temperature-dependent Hall and conductivity measurements [34–38]. Donor data from literature are summarized in Table 2.1. vacuum

energy [eV]

0 -2

EC

-4

H0

-6 -8

EV

LaAlO3

ZrSiO4

La2O3

HfO2

ZrO2

SrZrO3

SrTiO3

SnO2

ZnO

MgO

Al2O3

SiO2

-12

Si

-10

Fig. 2.3. Positions of the defect level of hydrogen (•) in diﬀerent oxides and in silicon relative to the valence and conduction band edges as calculated by Robertson et al. [32]

2 Electrical Properties

41

Table 2.1. Donors in zinc oxide according to Meyer et al. [39] and others Element

Structure

Energy ECB -E (meV)

Type

Source

Al B Cr Ga H In Li O Zn

AlZn BZn CrZn GaZn Hi InZn Lii VO ? Zni

65/53

Donor Donor Donor Donor Donor Donor Donor Donor Donor

[38, 39] [23] [23] [39] [30, 38] [39] [40] [38] [17, 38]

54.5 37/46 63.2 300 46

2.1.1.4 Compensation and p-Type Doping The addition of Li2 O or Na2 O during crystal growth or annealing in Li or Na containing atmospheres leads to highly resistive crystals, but never to a crossover to p-type ZnO [41]. This explains why hydrothermally grown ZnO crystals exhibit very high resistivities because the solvents contain sodium, potassium, and/or lithium salts leading to a built in of these compensating atoms. For acoustical applications of ZnO lithium was used to increase the resistivity of zinc oxide up to 1×1012 Ω cm [42]. Also silver can be used to compensate donors in zinc oxide [43]. Other candidates for acceptors are the group V elements nitrogen, phosphorous, and arsenic. The defect chemistry of these expected p-type dopants is very complex (see also Sect. 2.5). For instance, nitrogen on oxygen lattice sites should act as an acceptor, while N2 on the same site forms a donor. The latter property of nitrogen in ZnO was ﬁrst predicted by density functional calculations [44] and recently experimentally proved by extended X-ray absorption ﬁne structure spectroscopy (EXAFS) [45]. Meyer et al. summarized recently the results on doping of single crystals and epitaxial ﬁlms [39, 46]. The group V-elements are built in onto diﬀerent lattice sites, either on oxygen sites, where they are expected to form acceptors, or on zinc sites. Sometimes also interstitial lattice sites are assumed, for instance for lithium, explaining why lithium acts both as a donor and as an acceptor, connected with a very low electrical activation for compensation [40]. Acceptor data from literature are summarized in Table 2.2. 2.1.2 Electrical Transport in ZnO Single Crystals First reliable conductivity data were given by Fritsch already in 1935, shown in Fig. 2.2. Signiﬁcant progress was achieved when the semiconductor era started [11]. Hahn [15] and Harrison [16] were the ﬁrst who reported

42

K. Ellmer Table 2.2. Acceptors in zinc oxide according to Meyer et al. [39] and others Element

Structure

Energy E-EVB (meV)

Type

Source

Ag As Cu Li N Na P

AgZn /Agi AsZn CuZn LiZn NO NaZn PO

200

Amphoteric Donor Deep acceptor Deep acceptor Acceptor Deep acceptor Acceptor

[47, 48] [23] [48, 49] [39, 41, 50] [51] [50] [52]

380/190 800 110 600

temperature-dependent conductivity and Hall measurements on single crystals that were grown accidentally in an industrial process for ZnO production. Naturally, these samples were of poor crystalline and electrical quality. In 1957, Hutson from Bell Labs performed the ﬁrst comprehensive study of the electronic transport along the c-direction in single crystalline ZnO needles, grown by the vapor transport method and intentionally doped by hydrogen, interstitial zinc, and lithium [34]. Hutson measured the temperaturedependent conductivity and the Hall eﬀect between 55 and 1 000 K and calculated electron concentrations ND and mobilities µ. From the analysis of the ND (T ) curves, he could derive a hydrogen-type donor center with an ionization energy of 51 meV for ND < 5 × 1016 cm−3 and an eﬀective electron mass of 0.27 × me , where me is the free electron mass. The room temperature mobilities of his crystals with low carrier concentrations (<1017 cm−3 ) were in the range of 180 cm2 V−1 s−1 . Figure 2.4 displays the temperature dependence of the Hall mobility of three samples of Hutson with diﬀerent carrier concentrations. For comparison the temperature-dependent mobilities of undoped ZnO crystals measured by Wagner and Helbig in 1974 [35], Li and Hagemark in 1975 [53] and by Look et al. in 1998 [37] are included. It can be seen that the mobilities of crystals with low carrier densities (<1017 cm−3 ) have not changed at all from 1957 to 1998, pointing to the high purity and good crystalline quality of the crystals grown at Bell Labs. This reﬂects the fact that ZnO was one of the ﬁrst very pure single crystals prepared (after Ge, Si, and InSb) as mentioned by Rode [54]. Already in the 1950s, Hutson compared his experimental mobility data with theoretical models for the carrier transport in semiconductors, which was continued by others [35, 37, 53]. As an example, the theoretical ﬁt of Wagner and Helbig (1974) is discussed in detail in the following section. The increase of the mobility with decreasing temperature is due to a reduced carrier scattering by polar-optical and acoustical phonons. The maxima of the mobilities around 50–100 K are caused by the onset of ionized impurity scattering. The highest measured mobility at about 80 K is

2 Electrical Properties 10

(a)

2

10

6 4

N [cm-3]

µH [cm2/Vs]

1000

2

100 6 4

10 10

43

18

(b) 17

16

15

2

10

10 2

1

4

6 8

2

4

10

6 8

100

2

4

6 8

1000

14

5

T [K]

10

15

20

1000/T [K-1]

Fig. 2.4. (a) Temperature-dependent mobilities of ZnO single crystals with diﬀerent carrier concentrations at 300 K, reported by Hutson [17] (open triangle, 7.6 × 1016 cm−3 ; open triangle down, 1.6 × 1017 cm−3 ; ﬁlled triangle down, 1 × 1018 cm−3 ). For comparison the mobilities given by Wagner and Helbig [35] (open circle, 2.8 × 1016 cm−3 ); Li and Hagemark [53] (+, 3.1 × 1016 cm−3 ) and Look et al. [37] (open square, 8 × 1016 cm−3 ) are included. The continuous line is a theoretical ﬁt given by Wagner and Helbig in 1974. (b) Carrier concentration as a function of the reciprocal temperature of the samples from (a)

2 000 cm2 V−1 s−1 , pointing to a good crystal quality. The decrease of the mobility with increasing carrier concentration (compare the experimental curves of Hutson in Fig. 2.4) is only due to ionized impurity scattering since the lattice scattering due to optical and acoustical phonons and piezoelectric scattering is independent on the carrier density. In the 1970s, Wagner and Helbig [35] and Hagemark and Chacka [36] performed comprehensive studies of transport properties of zinc oxide single crystals (also vapor phase grown) treated at controlled zinc vapor pressures, to change the carrier concentration of the nominally pure and undoped ZnO crystals. These experimentators, as well as Rode [54] in an excellent, comprehensive theoretical study, took the following scattering processes into account. 2.1.2.1 Optical Mode Scattering This scattering process is due to the interaction of electrons with the electric ﬁeld induced by the lattice vibration polarization (polar longitudinal-optical phonons) occurring in polar semiconductors with partially ionic bonding. According to Devlin [55], the optical Hall mobility can be calculated by

e ω0 µHopt = rHopt φ − 1 (2.7) exp 2αω0 m∗ kT where the polaron coupling constant α is given by

∗ 1 m EH 1 α= − ε∞ εs me ω0

(2.8)

44

K. Ellmer

ε∞ and εs are the high frequency and the static dielectric constants and EH is the ﬁrst ionization energy of the hydrogen atom (13.595 eV). m∗ and me are the eﬀective and the vacuum electron masses, while ω0 is the energy of the longitudinal optical phonon (73.1 meV). rHopt is the Hall coeﬃcient factor for optical mode scattering. rHopt φ is a slowly varying function of the temperature. Usually it is assumed that rHopt φ is equal to 1 [35, 36, 55]. 2.1.2.2 Acoustical Mode Scattering Acoustic phonons lead to lattice deformation scattering because of a local energetic shift of the band edges. According to Bardeen and Shockley [56] the acoustical lattice mode Hall mobility is: √ 8π4 cl e

µHac = rHac (2.9) 3E12 m∗5 (kT )3 where cl is the averaged longitudinal elastic constant. E1 is the deformation potential (energy shift of the conduction band per unit dilation). The Hall coeﬃcient for acoustic phonon scattering is given by rHac = 3π/8 = 1.178 [57]. The deformation potential E1 is not very well known for ZnO. In literature E1 values scatter from 1.4 [58] to 31.4 eV [35]. For the calculation of the overall carrier scattering in ZnO this is not very important since other scattering mechanisms, especially polar-optical and piezoelectric scattering, are dominant [54]. 2.1.2.3 Piezoelectric Mode Scattering This scattering mode occurs only in piezoelectric materials, i.e., in crystals without inversion symmetry, and is caused by the electric ﬁeld associated with acoustical phonons. Zinc oxide exhibits very high electro-mechanical coupling coeﬃcients P (see later), exceeding that of quartz [42, 59], which is one of the mostly used piezoelectric materials. Zook [60] has calculated the piezoelectrically limited mobility on the basis of the elastic and piezoelectric constants as (see also Rode [54]): √ 16 2πh2 εε0 √ µHpie = rHpie , (2.10) 2 3eP⊥, m∗3 kT with the Hall coeﬃcient for piezoelectric mode scattering rHpie = 45π/128 = 1.1045. The averaged piezoelectric electro-mechanical coupling coeﬃcients P⊥, for electrical transport perpendicular or parallel to the c-axis are given by [54, 60]: P⊥2 =

4(21e215 + 6e15 ex + e2x ) (21e233 − 24e33 ex + 8e2x ) + 105εs ε0 εl 105εsε0 εl

(2.11)

2 Electrical Properties

P2 =

2(21e215 + 18e15 ex + 5e2x ) (63e233 − 36e33 ex + 8e2x ) + 105εsε0 εl 105εsε0 εl

45

(2.12)

where cl,t are the longitudinal and transversal elastic constants, eij are the piezoelectric coeﬃcients and ex = e33 − e31 − 2e15 . The longitudinal and transversal elastic constants cl,t can be calculated from the elastic constants cij by cl = (8c11 + 4c13 + 3c33 + 8c44 )/15 ct = (2c11 − 4c13 + 2c33 + 7c44 )/15

(2.13) (2.14)

For most semiconductors, the piezoelectric electro-mechanical coupling coefﬁcients P are of the order of 10−3 [61], while ZnO exhibits P values of about 0.2–0.4 (see Table 2.3). The piezoelectric mode scattering leads to an anisotropic mobility in ZnO calculated by Zook [60] and by Rode [54] for II–VI semiconductors. For the mobility ratio µ⊥ /µ for transport perpendicular and parallel to the c-axis Zook and Rode calculated µ⊥ /µ = 2.8 and 2.94, respectively. Since this scattering process is only signiﬁcant at low temperatures around 100 K, the mobility at room temperature is isotropic (see Wagner and Helbig [35]). Because of the diﬀerent temperature dependencies of acoustic, polar optical, and piezoelectrical scattering, the mobility at room temperature is limited by polar optical scattering. These three lattice scattering processes depend only on the temperature and are thus intrinsic, only determined by the material parameters mentioned above and summarized in Table 2.3. The material parameters for optical and piezoelectric scattering can be obtained by independent measurements (optical, mechanical, and piezoelectric) and are quite accurately known. On the other hand, the deformation potential is much less accurate (see Table 2.3) and is often extracted from ﬁtting measured mobility data to (2.9). If this is done quite large E1 values are obtained, which diﬀer signiﬁcantly from the deformation potentials of other semiconductors, which are typically in the range of 5–10 eV (see [54, 62]). 2.1.2.4 Ionized Impurity Scattering This process describes the scattering of free carriers by the screened Coulomb potential of charged impurities (dopants) or defects theoretically treated already in 1946 by Conwell [74, 75], later by Shockley [10] and Brooks and Herring [76, 77]. In 1969, Fistul gave an overview on heavily-doped semiconductors [78]. A comprehensive review of the diﬀerent theories and a comparison to the experimental data of elemental and compound semiconductors was performed by Chattopadhyay and Queisser in 1980 [79]. For nondegenerate semiconductors the ionized impurity mobility µHii is given by [79]: √ 128 2π(εε0 )2 (kT )3/2 (2.15) µHii = rHii √ ∗ )2 m∗ Ni Z 2 e3 ln 24mneεε20(kT 2

46

K. Ellmer

Table 2.3. Material parameters of zinc oxide for the calculation of the latticescattering-limited mobility (2.7–2.14) Parameter [unit]

Value

Source

Band gap Eg (300 K) [eV] Pressure coeﬃcient dEg /dp [meV GPa−1 ] Eﬀective electron mass m∗ /me (polaron mass) Eﬀective conduction band density of states (300) NC [cm−3 ] High frequency dielectric constant ε∞ Static dielectric constant εS Energy of the longitudinal optical phonon ω0 [meV] Nonparabolicity parameter β [eV−1 ] Deformation potential E1 [eV] Elastic constant c11 [GPa] Elastic constant c13 [GPa] Elastic constant c33 [GPa] Elastic constant c44 [GPa] Piezoelectric constant e33 [Cm−2 ] Piezoelectric constant e31 [Cm−2 ] Piezoelectric constant e15 [Cm−2 ] Longitudinal elastic constant cl [GPa] Transversal elastic constant ct [GPa] Piezoelectric coeﬃcients P⊥ ; P

3.4 (dir) 25.6

[5] [63]

0.318; 0.28; 0.24

[54, 62, 64]

3.7 × 1018

[5]

8.34 3.74 73.1

[65] [65] [66]

0.29; 0.66; 1.04 31.4; 3.8 ;18.9; 1.4 209.7; 206; 190; 207 105.1; 118; 90; 106.1 210.9; 211; 196; 209.5 42.5; 44.3; 39; 44.8 −0.62; −0.51 0.96; 1.22 −0.37; −0,45 204.7; 207.2; 185.3; 204.5 47.9; 44.8; 45.7; 48.1 0.21; 0.36

[67–69] [5, 35, 54, 58] [70–73] [70–73] [70–73] [70–73] [73] [73] [73] [70–73] [70–73] [54]

For ZnO material properties see also Sect. 3.3, Table 3.3 and Sect. 3.4.1, Table 3.4

where n and Ni are the carrier and the impurity concentration, respectively. For Ni it can be written: Ni = n + 2NA , with NA the acceptor density. rHii = 315π/512 = 1.933 is the Hall coeﬃcient for ionized impurity scattering [57]. Often, instead of the scattering function ln(x) in (2.15), the scattering function ln(1 + x) − x/(1 + x) is used [76]. However, the derivation of (2.15) is only valid for x 1; hence the function ln(x) is suﬃcient [79]. This means, that (2.15) is valid up to carrier concentrations of about 5×1018 cm−3 [57]. For high charge carrier concentrations (degenerate semiconductors) or low temperatures the expressions ﬁrst derived by Shockley [10] for scattering at the truncated Coulomb potential and by Dingle [80] (see also [81]) for scattering at the screened Coulomb potential have to be used:

2 Electrical Properties

µsh ii

3(εr ε0 )2 h3 = Z 2 m ∗2 e 3

µDi ii = with

ln 1 +

32/3 π 1/3 εr ε0 h2 n1/3 2m∗ e2

2 −1

3(εr ε0 )2 h3 n 1 2 ∗2 2 Z m e Ni Fii (ξd )

Fii = ln (1 + ξd ) −

ξd 1 + ξd

47

(2.16) (2.17)

and ξd = (3π 2 )1/3

εr ε0 h2 n1/3 m∗2 e2

(2.18)

Again, both expressions diﬀer only in the screening terms for an uncompensated fully ionized semiconductor (n = Ni ), where n is the electron density. The signiﬁcant diﬀerences between theory and experiment in the region of degeneracy have soon led to improvements of the theory of ionized impurity scattering by taking into account the nonparabolicity of the conduction band [82]. Zawadzki [81] calculated an analytical expression for the screening function Fiinp for the case of a nonparabolic band structure, which was further simpliﬁed by Pisarkiewicz et al. [83]:

ξnp ξ 5ξnp 4ξnp 1− ln (1 + ξ) − − 2ξ 1 − (2.19) Fiinp = 1 + ξ 8 1+ξ 16 m∗ with the additional parameter ξnp = 1 − 0∗ (2.20) m which describes the nonparabolicity of the conduction band, with m∗0 , the eﬀective mass at the conduction band edge. For a parabolic band (m∗ = m∗0 ) ξnp = 0, and (2.19) reduces to the expression (2.18). The dependence of the eﬀective mass on the electron energy in the conduction band can be approximated by [81] m∗ = m∗0 [1 + 2β(E − EC )]

(2.21)

Here, β is the nonparabolicity parameter (see Table 2.3), E − EC is the electron energy relative to the conduction band edge EC [83]. Though experimental data on the eﬀective mass of ZnO vs. electron density are rare in literature, this approximation seems to be applicable to zinc oxide. Brehme et al. [84] reported for Al-doped ZnO an eﬀective mass of m∗ = 0.5me for n > 1020 cm−3 , which corresponds to a nonparabolicity parameter of β = 0.29 eV−1 , corresponding to the rule β ≈ 1/Eg (Eg - band gap energy) [81]. Using the approximation for the relation between the reduced Fermi energy η = (EF − EC )/kT and the Fermi–Dirac integral F1/2 (η) = n/NC , given by Nilsson [85] (see also [86]), an analytical equation for m∗ in dependence on the electron density n can be written as: √ (3 π(n/4NC )2/3 ) m∗ ln (n/NC ) √ (2.22) = 1+2βkT 2+ m∗0 1 + [0.24 + 1.08(3 π(n/4NC )2/3 ])2 1 − (n/NC ) where NC is the eﬀective density of states in the conduction band.

48

K. Ellmer

2.1.2.5 Neutral Impurity Scattering Neutral shallow-impurity scattering is often discussed in papers about transport in TCO ﬁlms at room temperature [87, 88]. The mobility due to neutral impurity scattering was ﬁrst derived by Erginsoy [89] who scaled the electron scattering at hydrogen atoms to that in a semiconductor by using its dielectric constant and carrier eﬀective mass, which leads to : µHn = rHn

m∗ e 3 A(T )4πεε0 3 Nn

(2.23)

Here, A(T) is the scattering cross-section factor and Nn is the density of neutral scattering centers. Erginsoy [89] used a temperature-independent value A = 20. Itoh et al. used the correct e–H scattering cross-section to describe low-temperature electron and hole mobilities in germanium [90]. The Hall factor for neutral impurity scattering is rHn = 1 (see Seeger [57]). The concentration of neutral impurities is given by Nn = ND − NA − n(T ), where ND and NA are the donor and acceptor concentrations, respectively. Since the shallow donors in TCO materials (for instance the group III elements in ZnO) exhibit ionization energies around about 50 meV [39], the concentrations of neutral donors at room temperature is very low, taking into account the further reduction of the ionization energy for degenerately doped semiconductors. Therefore, in the following this scattering process is not further taken into consideration. Another important process is scattering at crystallographic defects like dislocations or stacking faults. Since the concentration of such defects is suﬃciently low in single crystals, this scattering process is discussed in the section on polycrystalline ZnO ﬁlms (Sect. 2.2.3). Figure 2.5 shows temperature-dependent mobilities of two samples reported by Wagner and Helbig [35]. These samples were measured with the current ﬂowing parallel and perpendicular to the c-axis. The data are compared with theoretical estimates based on (2.7–2.15), which show a good agreement down to temperatures of 60–100 K. At still lower temperatures the deviation from the theoretical curve is signiﬁcant. Wagner and Helbig as well as Hagemark and Chacka [36] attributed this to freezing out of carriers and transport in an impurity band. Also, scattering by neutral impurities is a possible scattering process at such low temperatures [91]. The good agreement for T > 100 K was achieved by varying the deformation potential value E1 to get a good ﬁt to the experimental data (E1 = 3.8 eV, see Table 2.3). 2.1.2.6 Anisotropy of the Electrical Transport Hutson was the ﬁrst who reported the measurement of the electrical conductivity along and perpendicular to the c-axis of ZnO single crystals [34]. Within his measurement accuracy of 10 %, he found isotropy of the electrical

2 Electrical Properties 10

4

µ Hpie

6 4

µH [cm2/Vs]

µ Hopt

µ Hopt

µ Hac

µ theo

µ Hii

2

10

µ Hac

49

µ Hpie

µ theo

3 6 4

µ Hii

2

10

2

J||c , B⊥c

6 4

J⊥c , B|| c

2

10

(a)

1 2

1

4

6 8

2

4

10

6 8

100

T [K]

2

4

6 8

1000 1

(b) 2

4

6 8

2

4

10

6 8

2

100

4

6 8

1000

T [K]

Fig. 2.5. Temperature-dependent mobility of ZnO single crystals measured with the current ﬂowing parallel (a) or perpendicular (b) to the c-axis of the crystals. The theoretical mobilities for the diﬀerent scattering processes (optical, acoustical, and piezoelectric as well as ionized impurity scattering) as calculated by Wagner and Helbig [34, 35] are shown as diﬀerently dashed lines. The calculated combined mobility curves (solid lines) ﬁt the experimental data quite well for temperatures above about 20–50 K. The carrier concentrations at 300 K were about ND − NA = 2.25 × 1016 cm−3 and NA = 2.75 × 1016 cm−3 (compensation ratio NA /ND = 0.55)

transport in ZnO at room temperature. Wagner and Helbig [35] measured the conductivity and the Hall mobility as a function of temperature parallel and perpendicular to the c-axis (Fig. 2.5). At room temperature, the mobilities µ⊥c and µc were equal, thus conﬁrming Hutsons result. However, at low temperatures (40–100 K), the mobility perpendicular to the c-axis was about two times higher than µc . This result is in the same range as the calculations for piezoelectric scattering by Zook [60] and Rode [54]. Venger et al. [92] calculated optical mobilities parallel and perpendicular to the caxis from IR reﬂection data at room temperature. In contrast to Hutson and Wagner/Helbig, their µc values were about 10 % higher than the mobility perpendicular to the c-axis. 2.1.2.7 Mobility as a Function of the Carrier Concentration The mobility and resistivity data of single crystalline zinc oxide samples (measured at room temperature) from diﬀerent authors, which were reported from 1957 to 2005, are displayed in Fig. 2.6 as a function of the carrier concentration (part of these data were taken from [67]). Undoped ZnO crystals exhibit carrier concentrations as low as 1015 cm−3 , while indium-doped crystals reach carrier concentrations up to 7 × 1019 cm−3 . The mobility data show a large scattering between carrier concentrations of 1017 to 5 × 1018 cm−3 . This is caused by the fact that zinc oxide is a compound semiconductor that is not as well developed as other semiconducting compounds. For instance, only

50

K. Ellmer 2

10 Si:B

(a)

Si:P

resistivity [Ωcm]

2

mobility µ [cm /Vs]

250 200 150 100 50 0 15 10

(b)

1

10

0

10

-1

10

-2

10

-3

16

10

17

18

10 10 -3 Nd [cm ]

19

10

20

10

10

15

10

16

10

17

10

18

10

19

10

20

10

-3

Nd [cm ]

Fig. 2.6. (a) Hall mobility and (b) resistivity of undoped (open symbols) and indium-doped (ﬁlled symbols) ZnO single crystals at room temperature as a function of the carrier concentrations from diﬀerent sources: ( ) Hutson (1957) [17], ( ) Rupprecht (1958) [94], ( ) Thomas and Lander (1958) [24], ( ) Baer (1967) [95], ( ) Hausmann and Teuerle (1973) [96], ( ) Wagner and Helbig (1974) [35], ( ) Utsch and Hausmann (1975) [97], ( ) Hagemark and Chacka (1975) [36], ( ) Look et al. [37], ( ) von Wenckstern et al. (2005) [38]. The full lines are semiempirical ﬁts according to (2.24). For comparison the dotted and dashed lines for the mobilities of boronand phosphorous-doped silicon are displayed in (a)

recently, it was possible to grow zinc oxide from its own melt, a technique used for other semiconductors (Ge, Si, GaAs, GaP, etc.) since decades [93]. The maximum reported room temperature mobilities of ZnO are in the range of µn = 200 − 225 cm2 V−1 s−1 . This low value, compared with other semiconductors (µn [Si] = 1415 cm2 V−1 s−1 , µn [GaAs] = 8 500 cm2 V−1 s−1 ) is due to the strong polar optical scattering in ZnO. For comparison electron mobilities of other II–VI semiconductors a well as other transparent conductive oxides (In2 O3 , SnO2 ) are given in Table 2.4 [53] and displayed as a function of the band gap energy in Fig. 2.7. Taking the maximum mobilities in ZnO crystals as a criterion the crystal quality has not changed in the last 50 years. For carrier concentrations above 1017 cm−3 the mobility decreases continuously down to about 50 cm2 V−1 s−1 at ND ≈1020 cm−3 , which is caused by scattering at ionized impurities, used as donors (indium, zinc interstitials). Data for ZnO single crystals doped with other dopants (B, Al, Ga) are not available. The overall shape of the µ(ND ) in Fig. 2.6 curve is qualitatively similar to that of p- and n-type silicon and can be ﬁtted by the same empirical dependence, which was given two decades ago by Masetti et al. [103] for silicon: µmax − µmin µ1 − (2.24) µMa = µmin + 1 + (n/nref 1 )α1 1 + (n/nref 2 )α2

2 Electrical Properties

51

Table 2.4. Band gap energies, lattice mobilities at room temperature and eﬀective masses of II–VI semiconductors and other TCOs (single crystals)

10

Source [55] [55] [55] [55] [98, 99] [55] [55] [55] [100] [101]

5

0.4 4 2

10

0.3

4 4

0.2

2

10

3

0.1

4

effective mass m*/m0

lattice electron mobility [cm2/Vs]

SemiEg µn−lattice m∗ /me 2 −1 −1 conductor (eV) (cm V s ) ZnO 3.43 200 0.32 ZnS 3.8 140 0.27 ZnSe 2.78 530 0.17 ZnTe 2.35 340 0.12 CdO 2.3 320 0.15/0.11 CdS 2.52 350 0.2 CdSe 1.77 650 0.13 CdTe 1.54 1,050 0.096 In2 O3 3.75 210∗ 0.35 3.6 255 ⊥c:0.3, c: 0.23 SnO2 ∗ Extrapolated from higher carrier concentrations

2

10

2

0.0 0.0

1.0

2.0

3.0

4.0

band gap energy [eV] Fig. 2.7. Lattice mobilities (ﬁlled triangle down) and eﬀective electron masses (open circle) of II–VI semiconductors as a function of the band gap energy Eg [102]

The ﬁt parameters can be connected to the physical quantities according to: – µmax - lattice mobility at low carrier concentrations – µmin - ionized impurity mobility at high carrier concentrations – µmax − µ1 - clustering mobility at very high carrier concentrations At carrier densities above 1020 cm−3 the ionized impurities form clusters with a higher scattering power, which is proportional to the square of the cluster charge Z, further reducing the mobility, thus the parameter clustering mobility was introduced by Klassen [103, 104]. It is interesting to note that though the lattice mobilities of silicon and ZnO are very diﬀerent, the

52

K. Ellmer

Table 2.5. Fit parameters according to (2.24) for zinc oxide [67] in comparison with that of phosporous- and boron-doped silicon [103] Fit parameter µmax [cm2 V−1 s−1 ] (lattice mobility) µmin [cm2 V−1 s−1 ] (ionized mobility) µmin − µ1 (cm2 V−1 s−1 ) (clustering mobility) nref1 (cm−3 ) α1 nref2 (cm−3 ) α2

Si (P) 1414 68.5 12.4 9.2 × 1016 0.711 3.41 × 1020 1.98

Si (B) 470.5 44.9 15.9 2.23 × 1017 0.719 6.1 × 2020 2.0

ZnO 200 50 10 1.5 × 1018 1 6 × 2020 2

ionized impurity mobilities and the clustering mobilities are comparable (see Fig. 2.6 and Table 2.5), which is due to the universal mechanism of scattering at ionized impurities, and cannot be surmounted in homogenously-doped crystals [67]. Carrier concentrations above 1020 cm−3 are not yet reported for ZnO single crystals. Only in thin ﬁlms carrier concentrations of up to 1.5 × 1021 cm−3 were achieved, which will be dealt with in Sect. 2.2.3. 2.1.2.8 Epitaxial ZnO In the last 15 years epitaxial zinc oxide ﬁlms were deposited onto diﬀerent single crystalline substrates, especially sapphire (Al2 O3 ) with c- and a-orientation, lithium niobate (LiNbO3 ), periclase (MgO), and scandium– aluminium–magnesium spinel (SCAM, ScAlMgO4 ) [103–105]. The latter material is especially suited for heteroepitaxial growth, because of its very low lattice mismatch relative to ZnO ( a/a ≈ 0.09 %) [64]. Figure 2.8a shows the temperature-dependent mobility of an undoped epitaxial ZnO ﬁlm on a SCAM substrate, reported by Makino et al. [64]. The mobility values of undoped ZnO ﬁlms are comparable or even higher than for bulk single crystalline ZnO with comparable carrier concentration (N < 1017 cm−3 ). This holds for room temperature, where polar-optical scattering dominates, as well as for low temperatures around T ≈ 100 K, where piezoelectric scattering is signiﬁcant. The authors argue that the crystalline quality of these epitaxial ﬁlms is better than that of ZnO single crystals, explaining the higher mobility. Since polar-optical and piezoelectric scattering processes depend only on material parameters and not on the dopant concentration (see (2.7) and (2.10)) that means on parameters, which were determined independently on carrier transport measurements, it is unlikely that the mobility of the epitaxial ﬁlms should be higher than that of good quality single crystals, shown in Figs. 2.4–2.6. This becomes even more obvious from Fig. 2.8(b), where the carrier concentration-dependent mobilities of ZnO:Ga ﬁlms also grown on SCAM by Makino et al. are shown. The experimental points scatter significantly but are quite well described by the semiempirical curve, (2.24) (see Fig. 2.6), which was ﬁtted to the data of ZnO single crystals reported from 1957 to 2005. In contrast, the theoretical curve calculated by the authors is

2 Electrical Properties

10 10 10 10

6

500

µ ii

5

µ ac 4

µ piezo µ po

3

2

Hall mobility [cm2/Vs]

Hall mobility [cm2/Vs]

10

53

400 300 200 100 0

100

150

200

250

temperature [K]

300

10

16

10

17

10

18

10

19

10

20

10

21

carrier density [cm-3]

Fig. 2.8. (a) Hall mobility as a function of the temperature for an undoped epitaxial ZnO layer and (b) Hall mobility of Ga-doped ZnO layers as a function of the carrier concentration. The ZnO ﬁlms were grown epitaxially on lattice-matched ScAlMgO4 (SCAM) by Makino et al. [64]. In (a) the calculated mobilities for acoustical, polaroptical, piezoelectric, and ionized impurity scattering are shown, together with the total theoretical mobility. In (b) the solid curve is the ﬁt curve (2.24) from Fig. 2.6, while the dashed line is the theoretical curve, calculated by Makino et al. [64]. The dotted line was calculated for transport across depletion regions at grain barriers (see Sect. 2.2.3), also present in epitaxial ﬁlms [106]

higher by more than a factor of 2, which is caused by an unplausibly low value of the eﬀective mass of ZnO of only m∗/me = 0.24 instead of 0.32 (see Table 2.3). In the carrier-concentration region from 5×1018 to 8×1019 cm−3 , the mobility curve exhibits a region, where the mobilities are almost zero. This dependence is ﬁtted by the dotted curve, which will be explained in Sect. 2.2.3, where the transport across electrical grain barriers is addressed explaining such very low mobilities.

2.2 Electrical Transport in Polycrystalline ZnO 2.2.1 ZnO Varistors A varistor is an electronic device with a highly nonlinear current–voltage curve characterized by a very strong current increase, when a certain voltage is reached, the breakdown voltage of the varistor. The ZnO varistor was developed by Matsuoka in Japan at the beginning of the 1970s [107]. It consists of a polycrystalline matrix of ZnO (grain size some 10 µm) with additives of other metal oxides (Bi2 O3 , Co2 O3 , Cr2 O3 , MgO, MnO, NiO, Pr2 O3 , Sb2 O3 , SiO2 , TiO2 , etc.) with contents above ∼0.1 mol%. The mixed powder is pressed into a plate-like shape, sintered at temperatures between 1,000 and 1,400◦ C and slowly (∼100 K h−1 ) cooled [108]. Its current–voltage (IV) characteristic is comparable to a back-to-back Zener diode but the varistor can handle much higher currents and energies and is a superior transient and overvoltage

54

K. Ellmer

(a)

electrodes t

(b) Rg

ZnO grain

d intergranular regions

Rp

Cp

Fig. 2.9. (a) Schematic microstructure of a varistor, depicting the two electrodes, the grain structure and the intergranular regions, responsible for the highly nonlinear IV characteristics. Typical values for the grain size d and the thickness t of the intergranular regions are d ≈ 20 µm and t ≈ 10–20 nm. (b) Simple equivalent circuit of a varistor: Rg – grain resistance, Rp , Cp – parallel resistance and capacity of the intergranular regions [109]

protection device [109]. The response time of a varistor is in the range of ∼500 ps making it well suited to switch oﬀ fast voltage transients. Figure 2.9 shows schematically the microstructure and the equivalent circuit of a varistor. The current transport is governed by the intergranular regions, which consist of depletion regions between ZnO grains, constituting electronic barriers for the current transport. Increasing the voltage at the varistor leads eventually to the breakdown of the barriers accompanied by a high current ﬂow. Taking a typical thickness of a varistor (D = 1.6 mm) and the mean grain size (d = 20 µm), the breakdown voltage per grain is about 2–3 V when the switch-on voltage of the varistor is 120 V [110]. Near to the breakdown voltage of the varistor, the current changes by up to 11 orders of magnitude while the ﬁeld varies only by a factor of 3 pointing to the extreme nonlinearity of this device as depicted in Fig. 2.10(a). This behavior is described by the empirical relation between the current J and the ﬁeld F : p J1 F1 = (2.25) J2 F2 where the exponent p characterizes the nonlinearity of the varistor. Typical values are p = 25 − 50. The equivalent circuit of a varistor (see Fig. 2.9b) consists of a series resistor Rg , representing the resistance of the ZnO grains and a capacitance Cp in parallel with a voltage-dependent resistor Rp , both representing the intergranular depletion layer between the ZnO grains. The intergranular regions are very thin (smaller than 20 nm) and are enriched with the additives, for instance bismuth oxide [111,112]. Normally the series resistance of the grains (≈ 1 Ω cm) can be ignored, except for very high currents. The conduction mechanism is dominated by ﬁeld-assisted tunneling through

2 Electrical Properties

(a)

55

(b)

Fig. 2.10. (a) IV characteristics of a ZnO varistor [116]. (b) Energy band diagram and local charge distribution of a double Schottky barrier between two grains of ZnO with an applied voltage V after Greuter and Blatter [112]. The density of states at the interface is also shown. The transport occurs by thermionic emission over the barrier or by ﬁeld-assisted tunneling through the barrier. ΦB , barrier height; E1 , energy level of the interface states; E0 , energy of the shallow donors

barriers between the ZnO grains at high electric ﬁelds, which is depicted schematically in Fig. 2.10. Without or for low applied ﬁelds the electrons can surmount the barrier only by thermionic emission over the barrier yielding a thermally activated current at low voltages (see the IV characteristics in Fig. 2.10). From the temperature-dependent IV curves a barrier height of about 0.5–0.9 eV can be calculated [112,113]. Increasing the voltage and hence the ﬁeld leads to a decrease and a narrowing of the intergranular barriers and Fowler–Nordheim tunneling sets in accompanied by a steep increase of the current ﬂowing over the barriers. In this region, the IV characteristics exhibits a very low temperature coeﬃcient around dV /(V dT ) ≈ −2 × 10−4 K−1 , an indication of a tunneling process [114]. For a simple rectangular barrier and neglecting image force barrier lowering the current density is given by [115]: 3/2 −7.22 × 10−7 φb F (2.26) j = 33.9 exp φb F where j is the current density (mA cm−2 ), F is the applied ﬁeld (V cm−1 ) and ΦB the barrier height in electronvolt. It has to be mentioned that the barriers are essentially inﬂuenced by the additives that are precipitated at the interfaces between the grains [111]. It is interesting to note that the conductivity of a varistor increases with the pressure (a factor of 5 at 400 bar), which can be explained by the high piezoelectric eﬀect of zinc oxide [113],

56

K. Ellmer

lowering the barrier height by piezoelectrically induced charges. To describe a real varistor, which consists of a serial-parallel combination of many grains and intergranular regions, one has to take into account that the measured current is the summation over all current paths with diﬀerent electric ﬁelds and barrier heights. Although in a varistor the grain barriers are essential for its electronic function, in transparent, conductive electrodes, for instance in thin ﬁlm solar cells, the grain barriers are perturbing the electrical transport. How to avoid detrimental eﬀects of grain barriers is described in the following section. 2.2.2 Thin ZnO Films Polycrystalline zinc oxide ﬁlms were already prepared by Fritsch in 1935 by evaporation [9]. He found for undoped ZnO ﬁlms the typical behavior of a semiconductor (see Sect. 2.1). However, most investigations of thin ﬁlms were done since about 1960, ﬁrst for using the piezoelectric properties of ZnO in surface acoustic wave devices [117, 118]. For this application the ﬁlms had to be grown with a pronounced c-axis texture and with a very high resistivity (in the range of >108 Ω cm). In the following we will concentrate on doped zinc oxide ﬁlms for their application as transparent electrodes, which appeared around 1980. Electrical properties of polycrystalline ZnO ﬁlms were recently reviewed by Hartnagel et al. [87], Minami [26], and Ellmer [67]. With respect to the application of ZnO as transparent electrode, it is the aim to achieve a resistivity as low as possible with the constraint of a high transparency, both by increasing the carrier concentration and the mobility. Polycrystalline ZnO ﬁlms usually grow with the c-axes of the crystallites oriented approximately perpendicular to the ﬁlm plane [118, 119]. Therefore, the electrical transport in polycrystalline ﬁlms, which is measured laterally, occurs in almost all cases perpendicular to the c-axes of the crystallites. 2.2.2.1 Preparation of ZnO Films Thin ﬁlms of zinc oxide can be prepared by a variety of deposition methods: – – – – – – –

Reactive and nonreactive magnetron sputtering [120, 121] Metalorganic chemical vapor deposition (MOCVD) [122, 123] Pulsed-laser ablation (PLD) [124, 125] Evaporation [126, 127] Spray pyrolysis [128, 129] Sol–gel preparation [130, 131] Electrochemical deposition [132]

In the following we will focus on the ﬁrst three deposition methods, since these deliver the best ﬁlms with respect to low resistivity and high transparency. Especially, magnetron sputtering is a technique, which is already

2 Electrical Properties

57

used for large area coating deposition, both in research and industry (see also Chap. 5). Today, magnetron sputtering is one of the dominant deposition technique for the preparation of ZnO window layers for thin ﬁlm solar cells. This is due to the fact that magnetron sputtering allows large-area depositions at low substrate temperatures with the best properties reported. MOCVD is also used for industrial purposes. However, this method requires higher substrate temperatures (typically higher than 150◦ C), which cannot be applied in all cases for thin ﬁlm solar cells. The deposition of zinc oxide onto amorphous silicon solar cells is often performed by MOCVD, potentially also for a-Si:H solar cells (see Chap. 6). Pulsed-laser deposition is a very versatile technique (see Chap. 7), which can be used for material screening and for the determination of the intrinsic material data. It has the draw backs of a very low deposition rate and that it is diﬃcult to be scaled up to large coating areas, which prevents its industrial application up to now. Low resistivities for deposition at room temperature, the main goal for the application of TCO ﬁlms as transparent electrodes, can be achieved in two ways: – Creation of intrinsic donors by lattice defects (for instance oxygen vacancies or metal atoms on interstitial lattice sites) or – Introduction of extrinsic dopants (for instance metals with oxidation number three on substitutional metal lattice sites or halogens with oxidation number minus one on oxygen lattice sites in MO oxides) Carefully adjusting the oxygen partial pressure and the metal deposition rate is one way to introduce intrinsic donors. After the deposition, a reduction process of the deposited ﬁlms by annealing in vacuum or in a hydrogen containing atmosphere leads to the same eﬀect. However, intrinsically-doped ZnO ﬁlms are not well-suited for application because of their relatively high resistivity of about 10−2 –10−3 Ω cm. In addition they are unstable in oxidizing atmospheres leading to a strong increase of the resistivity by reoxidation of the oxygen deﬁcient ﬁlms. 2.2.3 Transport Processes in Polycrystalline Films 2.2.3.1 Dislocation Scattering This process is obviously a natural scattering process in polycrystalline materials, since polycrystalline ﬁlms exhibit a high concentration of crystallographic defects, especially dislocations [133, 134]. However, this process is rarely used to explain experimental data of carrier transport in polycrystalline semiconductors and especially transparent conducting oxides [88], which is mainly due to the fact that in most works on transport properties of polycrystalline ﬁlms the density of defects was not determined. P¨ od¨ or [135] investigated bended n-type Ge crystals with a dislocation density around 107 cm−2

58

K. Ellmer

and could his results describe by taking into account scattering by charged dislocations (see also Seeger [57]) leading to a mobility given by : √ √ 30 2π(εr ε0 )3/2 a2 nkT √ (2.27) µHdisl = rHdisl e3 f 2 m∗ Ndisl where a is the distance between acceptor centers along the dislocation line that catch electrons from the crystallite volume, f is the occupation number (0≤f≤1) of these acceptors and Ndisl is the density of dislocations. P¨ od¨ or sets for the Hall factor rHdisl = 3π/8 = 1.1781 [135]. In the 1990s this scattering process was discussed for epitaxial GaN/InGaN/AlGaN ﬁlms, also exhibiting the hexagonal-wurtzite structure like ZnO, which are already used for commercial light emitting diodes and lasers [136]. These devices work eﬃciently despite threading defect densities up to 1011 cm−2 [137]. Look and Sizelove [136] concluded from temperature-dependent carrier transport measurements that dislocations with a negatively charged core act as scattering centers. Recently, the line charge density of dislocations was measured directly by electron holography in epitaxial GaN [137] and ZnO ﬁlms [138] yielding line charge densities (that corresponds to f /a in (2.27)) of about 107 cm−1 . Recently, heteroepitaxial ZnO ﬁlms grown on GaN ﬁlms on sapphire substrates were analyzed by transmission electron microscopy [139]. These ﬁlms are characterized by a high stacking fault density in the range of 1018 cm−3 . According to Gerthsen et al. [139] these stacking faults are mainly generated by the precipitation of Zn interstitials accompanied by the formation of oxygen vacancies in the vicinity of a stacking fault. Since oxygen vacancies exhibit a charge of Z = 2 its scattering power is higher than that of singly ionized dopands like the group III elements in ZnO (B+ , Al+ , Ga+ , Z = 1) leading to a reduced mobility (see (2.15–2.17)). For polycrystalline ﬁlms an even higher stacking fault and dislocation density is plausible. Sagalowicz and Fox [134] analyzed undoped polycrystalline ZnO ﬁlms by TEM and found dislocation densities of around 1012 cm−2 , corresponding to a mean distance between dislocation of about 10 nm. Hence, the scattering of carriers at dislocations and oxygen vacancies would be even higher in such materials. Recent density-functional calculations [20] have shown that “point defect formation enthalpies in zinc oxide are very small” supporting the considerations given earlier (see Chap. 1). It is also noteworthy that ZnO exhibits a high radiation hardness, which is likely due to a rapid annealing of Frenkel pairs [22, 140], again caused by low defect formation energies, which lead to high diﬀusion coeﬃcients. Zinc oxide is a polar semiconductor, i.e., electrical charges are induced by stress along the c-axis. With this respect ZnO is similar to GaN [141]. This means intrinsic stress, because of lattice mismatch and/or of growth defects, could induce charges at grain barriers or extended defects leading to additional scattering.

2 Electrical Properties

59

2.2.3.2 Grain Barrier Limited Transport Polycrystalline ﬁlms exhibit, depending on their mean grain size, a vast amount of grain barriers, which constitute crystallographically disturbed regions, leading to electronic defects in the band gap of semiconductors. These defects are charged by carriers from the interior of the grains. Depending on the type of the carriers (electrons or holes) and the type of the defects (electron trap or hole trap) charge balance causes depletion or accumulation zones around the barrier. In TCO ﬁlms, which are typically n-type, a depletion zone is generated on both sides of a grain barrier accompanied by an energetic barrier of height ΦB for the electrons. This is due to the electron trap character of the defects. The carrier transport in polycrystalline silicon was ﬁrst described comprehensively by Seto [142]. He assumed a δ-shaped density of electron trap states in the band gap, which are completely ﬁlled. A schematic band diagram according to Setos model is shown in Fig. 2.11. An improved model was presented by Baccarani et al. [143], who considered a continuous energy distribution of trap states in the band gap. Additionally, these authors treated the possibility that the traps are only partially ﬁlled. Qualitatively, both models lead to the same conclusions. In the following, we restrict our discussion to Setos model. The carrier transport across the grain barriers is described by the classical thermionic emission (see for instance Sze [144]) depicted by path TE in Fig. 2.11. For very high carrier concentrations in the grains the depletion width is very narrow thus enabling quantum-mechanical tunneling of the barriers by the electrons, shown as path “TE” in Fig. 2.11. Both models yield an eﬀective mobility µeﬀ dominated by thermionic emission across the grain barriers with an energetic height ΦB : µeﬀ = µ0 exp (−

ΦB ) kT

(2.28)

TE

Φb

e

T

e E CB EF

Nt

E VB L Fig. 2.11. Linear row of grains of identical length L, doping N , and with grain barriers of height ΦB caused by a continuous distribution of electron trap states of density Nt [142]. Two diﬀerent transport paths for electrons are indicated: TE, thermionic emission across the barrier; T, tunneling through the barrier

60

K. Ellmer

where ΦB is the energetic barrier height at the grain boundary, T the sample temperature, and k is the Boltzmann constant, respectively. The prefactor µ0 in (2.28) can be viewed as the mobility inside a grain given by [142]: eL µ0 = √ 2πm∗ kT

(2.29)

Depending on the doping concentration in the grains, two expressions for the barrier height can be derived: e2 Nt2 8εε0 N e2 L 2 N ΦB = 8εε0

ΦB =

for LN > Nt

(2.30)

for LN < Nt

(2.31)

where e is the elementary charge, Nt is the charge carrier trap density at the boundary, εε0 is the static dielectric constant, N is the carrier density in the bulk of the grain, and L the grain size. For LN > Nt the traps are only partially ﬁlled and hence the crystallites are completely depleted, while for LN < Nt only part of the grain is depleted and the traps are ﬁlled completely. The maximum barrier height ΦB,max occurs for a doping concentration of N (ΦB,max ) = Nt /L, accompanied by a minimum of the eﬀective mobility according to (2.28). Figure 2.12 displays the barrier height as a function of the carrier concentration for a ﬁxed grain size, calculated from (2.30–2.31) for a trap density of 5×1012 cm−2 and a grain size of 100 nm. For three situations (1–3), marked in Fig. 2.12, the band structure is schematically shown in Fig. 2.13. In the original models of Seto and Baccarani et al. [142, 143] only thermionic emission was taken into account. For very high carrier concentrations (N > 1020 cm−3 ) additional tunneling through the barriers takes place (see Sect. 2.2.1), which

2

barrier height [eV]

1.4 1.2 1.0 0.8 0.6

1

0.4 Nt /L

0.2 0.0 0.0

0.2

0.4

0.6

3

0.8

1.0

electron concentration [1018cm-3]

Fig. 2.12. Barrier height at the grain barriers as a function of the electron concentration in the grains for a ﬁxed grain size (L = 100 nm) and a trap density Nt = 5 × 1012 cm−2 (calculated using (2.30–2.31))

2 Electrical Properties

1

2

ΦB

3 ΦB

E CB EF

Nt

61

E CB

ΦB

EF N t

Nt

E VB E VB L

L

L

Fig. 2.13. Schematic band diagrams in the grains for diﬀerent doping concentrations N in grains of identical size L (after Kamins [153]). The situations 1–3 correspond to carrier concentrations 1–3 in Fig. 2.12. The barrier height increases with increasing N up to a maximum at Nmax = Nt /L. Further increasing N decreases the barrier height

increases the current ﬂow between the grains. If the thermionic model discussed earlier is applied for such high carrier concentrations lower barrier heights are calculated as expected from (2.30), which has to be kept in mind. Concerning the prefactor µ0 , which is viewed as the intragrain mobility [145], in many cases too low values are extracted from the experiments [142, 143, 146]. Here, additional carrier scattering processes take place in the grains, for instance point defect and dislocation scattering. In the Seto model, it is assumed that every grain barrier exhibits the same barrier height, obviously an idealization of a real polycrystalline ﬁlm. In 1992, Werner [147] made an important extension of the Seto model by applying a Gaussian distribution of the barrier height around a mean barrier height ΦB with a standard deviation of ∆Φ due to variations among the diﬀerent grains and to potential ﬂuctuations within one grain. This model leads to the following expression of the grain barrier limited mobility [147]: 2 ΦB − ∆Φ eL 2kT µeﬀ = √ (2.32) exp − kT 2πm∗ kT which is often better suited to describe the temperature-dependent mobilities of polycrystalline ﬁlms, especially at larger grain barrier heights [146]. Recently, Lipperheide and Wille [148–150] worked out a theory for the combined ballistic and diﬀusive transport across grain barriers, which was applied to the carrier transport in polycrystalline silicon ﬁlms [151, 152]. However, their numerical approach yields only a gradual improvement when compared with the analytical model of Seto. 2.2.4 Experimental Mobility Data of Polycrystalline ZnO Figure 2.14 shows a compilation of mobility and resistivity data of doped and undoped ZnO ﬁlms prepared in the last 25 years as a function of the carrier

K. Ellmer 2

–2

(a)

10

mobility [cm2/Vs]

100 6 4

1 2

2

10

F

6 4

2 –3

10

4

F 1 2

2 –4

10

4

B

2 –5

2

10

(b)

4

B resistivity [½cm]

62

10 19

20

21

10 10 charge carrrier concentration [cm–3]

10

19

20

21

10 10 charge carrrier concentration [cm–3]

Fig. 2.14. (a) Mobilities and (b) resistivities of doped and undoped ZnO ﬁlms vs. the carrier concentration (part of the data were taken from [67,157]). The ﬁlms were deposited by magnetron sputtering (ﬁlled square, open square, open triangle down), MOCVD (open diamond ) or pulsed-laser deposition (open circle). The dashed lines (B) are a theoretical estimation of the mobility and the resistivity of Bellingham et al. [158]. The full lines (F) are the calculated data from the semiempirical ﬁt according to (2.24), while the dotted lines (1, 2) represent theoretical mobility calculations according to (2.15–2.22)

concentration [67]. Included are mostly ﬁlms deposited on glass substrates. It can be seen that the lowest resistivities are in the range of 1.5–3×10−4 Ω cm, about a factor of 2–3 higher than the lowest resistivities of tin-doped indium oxid (ITO, for which resistivities as low as 1.2×10−4 Ω cm are reported [154]), which is commercially used as transparent electrode in ﬂat panel displays [155]. The highest mobilities are in the range of 50–60 cm2 V−1 s−1 , which are quite well ﬁtted by the semiempirical curve (marked with (F) in Fig. 2.14) derived from the single-crystalline ZnO data (see Fig. 2.6). The dopants used are group III (B, Al, Ga, In) and, occasionally, group VII (F, Cl) elements. The diﬀerent dopants lead to nearly the same resistivities. However, aluminium is the dopant mostly used, followed by gallium and boron, the latter used for ZnO ﬁlms deposited by MOCVD [156] (see Chap. 6). All deposition methods yield resistivities well below 10−3 Ω cm. However, the plasmaassisted processes – magnetron sputtering and pulsed-laser deposition – lead to the lowest resistivities for doped ZnO ﬁlms. This is due to the fact that these methods rely not only on thermal activation of the growth process but beneﬁt from the additional energy input from energetic particles (ions, sputtered atoms, energetic neutrals). These methods can be used to prepare ﬁlms of better crystalline and electronic quality and/or to deposit ﬁlms at lower substrate temperatures (see Chaps. 5 and 7). It is interesting to note that though zinc oxide ﬁlms have been prepared now for three decades with large technical eﬀort, the lowest resistivities (and highest mobilities) have not changed since about 20 years (see [26, 67]). This is a hint to limitations concerning the introduction of dopants and the electron transport. The same is true for the other transparent conductive oxides (TCO) as indium

2 Electrical Properties

63

oxide, mostly tin-doped (ITO), and tin oxide (see the review of Minami [26]). The only diﬀerence is the level of resistivities achieved, which is somewhat lower for ITO and higher for SnO2 compared with ZnO. These values are comparable with the data of degenerately doped silicon [103] and gallium arsenide [79], though the intrinsic (lattice) mobilities of these semiconductors are much higher (see also Fig. 2.6). This points to the universality of the ionized-impurity scattering process, which limits the mobility of degeneratelydoped semiconductors. The highest reported carrier concentrations of the doped ZnO ﬁlms are about 1.5 × 1021 cm−3 , which is close to the solubility limit of the diﬀerent group III dopants in zinc oxide [159]. There also seems to be a lower limit of the carrier concentration in polycrystalline ZnO of about 2×1019 cm−3 . However, this is only a measurement eﬀect by the Hall method, which yields too low Hall voltages for highly resistive polycrystalline ﬁlms, where the depletion layers at grain barriers limit the electron transport. Hence, no mobility values can be determined for N < 1 × 1019 cm−3 . Included in Fig. 2.14 is an older theoretical estimation of mobilities and resistivities of Bellingham et al. [158] (curve B), which is obviously too optimistic compared with the experimental data. Though the scattering of the data points is high, the experimental mobility values decrease tentatively, especially for carrier concentrations N > 5×1020 cm−3 , an eﬀect observed also in other polycrystalline semiconductors, like silicon [103] or tin dioxide [160]. This dependence can only be explained by taking into account the eﬀect of impurity clustering (see [104, 161, 162]) and the nonparabolicity of the conduction band of ZnO, which was shown recently [67]. For that purpose theoretical mobilities µHii for a nonparabolic conduction band have been calculated from (2.17–2.20), which are shown in Fig. 2.14 too. The inclusion of the nonparabolicity of the conduction band reduces µHii , especially for n > 5 × 1020 cm−3 (curve 1). However, the improvement toward the experimental data is not suﬃcient for a satisfying agreement. The best overall agreement has been achieved by setting the charge of the impurities to Z = 2, which corresponds to doping by oxygen vacancies (curve 2). Considering that oxygen vacancy doping and extrinsic doping lead to the same mobilities, this could be an indication that the true doping mechanism is due to oxygen vacancies in both cases. In this line of argumentation extrinsic doping would generate oxygen vacancies because of the formation of the corresponding extrinsic dopant oxide (for instance B2 O3 , Al2 O3 , Ga2 O3 ). Moreover, this is supported by the fact that also the oxide forming elements Hf, Ti, Y, Zr can be used to dope zinc oxide [26]. The comparison of experimental data and theoretical calculations shows that still in 2006, as already pointed out by Chattopadhyay and Queisser [79], “transport in heavily-doped semiconductors is not well understood.” For polycrystalline ﬁlms this limited understanding is nicely illustrated by our own mobility data for Al-doped ZnO ﬁlms (see Fig. 2.15) deposited both on glass and sapphire substrates as a function of the carrier concentration

64

K. Ellmer

Hall mobility [cm2/Vs]

80

F

60

40

20

0 10

18

19

20

10 10 charge carrier density N [cm-3]

10

21

Fig. 2.15. Hall mobilities of undoped and doped zinc oxide thin ﬁlms as a function of the carrier concentration. Our own data are shown for ﬁlms deposited onto ﬂoat glass (ﬁlled square) as well as sapphire substrates (open triangle, triangle right, triangle left) [106]. For comparison mobility data of other groups were added, which have been measured for ﬁlms deposited by magnetron sputtering and by pulsedlaser deposition (PLD): Minami (open square, #, ﬁlled circle - PLD) [69], Brehme et al. (ﬁlled triangle down, open triangle down) [84], Kon et al. (boxtimes) [163], Agura (bowtie - PLD) [164], Suzuki (open circle - PLD) [124], and Lorenz et al. (otimes, oplus - PLD) [165]. The mobility values of ZnO single crystals (ﬁlled diamond ), already shown in Fig. 2.6 and the ﬁt curve (F) according to (2.24), are also displayed. The thin ﬁlm mobility data have been ﬁtted by the combined ionized impurity and grain barrier model (2.17–2.28), yielding the grain boundary trap densities summarized in Table 2.6

together with experimental data of other groups: Minami et al. [26], Kon et al. [166], and Brehme et al. [84]. These data have been selected from experiments where the carrier concentration was varied systematically. For carrier concentrations above 1 − 3 × 1020 cm−3 the mobility data of the single crystalline ﬁlms are almost comparable to the literature data reported for ZnO ﬁlms grown on glass substrates, though the former ﬁlms exhibit a signiﬁcantly better crystalline quality compared with ﬁlms grown on glass or silicon [106]. This means that in this carrier concentration region, the mobility is dominated by ionized impurity scattering as already pointed out by Bellingham et al. [158], Minami [26], and Ellmer [67] (see also Sect. 2.1.2). If the carrier concentration is reduced below about 3 × 1020 cm−3 the mobility in the epitaxial ﬁlms decreases steeply. This is in qualitative agreement with data of Minami [167] for undoped ZnO ﬁlms and with data of Brehme et al. [84] and

2 Electrical Properties

65

Kon et al. [163] for ZnO:Al ﬁlms. The data of the latter two groups exhibit lower mobilities at all, probably caused by additional scattering centers due to crystallographic defects (see later). Comparable trends were also reported recently by Agashe et al. [168]; these data were not included in order not to overload Fig. 2.15. In this carrier concentration region (N < 1−3×1020 cm−3 ) the mobility values of the polycrystalline and heteroepitaxial ﬁlms are significantly lower than that of single crystalline ZnO (compare with the semiempirical ﬁt curve in Fig. 2.15). This is caused by the limitation of the carrier transport by electrical grain barriers (see Sect. 2.2.3.2). Using (2.28), the experimental data for N < 3 × 1020 cm−3 have been ﬁtted (thin lines) for the data of Minami [167] for nominally undoped polycrystalline ZnO ﬁlms and for our epitaxial ZnO:Al ﬁlms [106]. The following trap densities Nt at the grain boundaries had to be assumed to ﬁt the experimental data for ZnO and ZnO:Al ﬁlms: Nt(ZnO) = 7 × 1012 cm−2 ; Nt(ZnO:Al) = 1.3 × 1013 cm−2 . Also the data of the other authors have been ﬁtted to determine the grain boundary trap densities, which are summarized in Table 2.6, which also includes the deposition method and the discharge mode (DC or RF) as well as the discharge voltages (if reported). It is interesting to note that the trap density changes signiﬁcantly (by about a factor of 6) depending on the deposition conditions. The highest Nt values occur for DC or pulsed-DC plasma excitation. RF magnetron sputtering reduces Nt already by a factor of 2.5. The lowest trap densities are achieved by magnetron sputtering onto substrates mounted perpendicularly relative to the sputtering target (Minami et al. [169]), which reduces direct particle bombardment of the growing ﬁlms, and for pulsedlaser deposition [124, 165]. This dependence points to the decisive role of the energy of the species contributing to the ﬁlm growth, which will be discussed in the following. The compilation of Table 2.6 shows that grain barriers or other structural inhomogeneities signiﬁcantly aﬀect the carrier transport in polycrystalline as well as in heteroepitaxially grown semiconducting oxide ﬁlms, independently of the material. This is supported by recent reports on gallium nitride ﬁlms grown by MOCVD on sapphire substrates [176] and for GaAs layers [177], where internal potential barriers (i.e., grain barriers) inﬂuence the carrier transport. Also for polycrystalline silicon high trap densities are reported, which are in the range of 3×1012 cm−2 [142,143]. Annealing our ZnO:Al ﬁlms in vacuum at about 500◦C caused a surprising result, exemplarily depicted by the arrow in Fig. 2.15. Although the carrier concentration does not change at all, the mobility is increased by about 30–50 %, approaching the mobility values of the general ﬁt curve (F), i.e., the values of ZnO single crystals. Even more important is the fact that this signiﬁcant electrical improvement is not accompanied by a change of the structural parameters as measured by X-ray diﬀraction: strain, grain size, and orientation [106]. This means that no recrystallization has occurred. Instead, it is plausible that the point defect and/or dislocation density has been reduced. Since the highest defect density

66

K. Ellmer

Table 2.6. Grain barrier trap densities Nt of zinc oxide and other polycrystalline TCO ﬁlmsa Film:dopant

Growth

Method

Excitation Nt (Udis [V]) (cm−2 )

Source

ZnO:Al ZnO:Al ZnO:Al ZnO ZnO ZnO ZnO ZnO ZnO In2 O3 :Sn

Epitaxial On glass On glass On glass Epitaxial Epitaxial On SiO2 On silicon On glass On glass

MS RMS RMS MS∗ PLD PLD RMS RMS RMS RMS/MS

[106] [84] [163] [26] [165] [64] [170] [171] [172] [173]

In2 O3 :Sn

On glass

MS

SnO2 :F,Cl CdIn2 O4

On glass On glass

SP RMS

RF (200) 1.3×1013 DC (425) 3×1013 MF (340) 3×1013 RF 7×1012 – 5×1012 – 1.5×1013 DC 2.2×1012 RF 5×1012 RF (100) 1.5–3×1012 DC (400) 1.5×1012 RF (425) RF-diode 2.5×1013 (2000) – 4×1012 DC (2000) 1.5×1013

[174] [175] [83]

MS, magnetron sputtering; RMS, reactive magnetron sputtering; RF, MF, radio frequency or mid frequency plasma excitation; p-DC, pulsed-DC excitation ∗ Substrate arranged perpendicular relative to the target (reduced ion bombardment of the ﬁlm) a The deposition method and the discharge voltages are also given

exists at the grain boundaries [178], the annealing of the above-mentioned defects most probably takes place at grain boundaries, thus reducing the grain barrier height leading to a higher mobility. Considering this eﬀect one can also explain that other mobility data from literature (see the data from [84,163] in Fig. 2.15) have a comparable dependence on N as mentioned earlier, although their absolute values are lower by a factor of 2–4. Probably, these samples, deposited under diﬀerent deposition conditions, contain a higher amount of point defects and/or dislocations at grain boundaries, leading to higher grain barriers and lower mobilities. Mobility data of ﬁlms prepared by pulsed-laser deposition, recently published by Lorenz et al. [165], are included in Fig. 2.15. Interestingly, these ﬁlms exhibit mobilities that are comparable to mobility values reported for zinc oxide single crystals, pointing to a higher structural quality and hence a lower grain barrier defect density of these PLD ﬁlms. However, a tendency to lower mobilities (compared with single crystal values) can also be observed for N < 1019 cm−3 , which can be explained by a trap density of Nt(PLD) = 5 × 1012 cm−2 .

2 Electrical Properties

67

2.3 Outlook: Higher Electron Mobilities in Zinc Oxide As mentioned in the preceding section the mobility of degenerately-doped zinc oxide (as well as of other TCO materials and semiconductors) is limited by ionized impurity scattering in homogeneously-doped materials. Since about 30 years it is well known that the mobility can be increased by the so-called modulation doping method, introduced by Dingle et al. [179] for GaAs/Ga1−xAlx As superlattice structures (for a review see [180]). The modulation doping principle separates the doping and the transport functions in a semiconductor material by stacking lowly- and heavily-doped ﬁlms with slightly diﬀerent band gaps, schematically depicted in Fig. 2.16. Although the heavily-doped ﬁlms provide the charge carriers, their transport occurs in the lowly-doped layers of the stack that are not subjected to the ionized impurity scattering limitation of the mobility. To our knowledge, the modulation doping principle was applied only to lowly-doped semiconductors with carrier concentrations of about 1017 cm−3 and it is still open if this principle works also for very high carrier concentrations above 1020 cm−3 . Rauf [181] was the ﬁrst who reported very high mobilities in highly-doped ITO ﬁlms, which he claimed to be due to the modulation doping eﬀect. In order that the modulation doping principle works, the thickness of the multilayers has to be in the range of nanometers, i.e.,

comparable to the Debye length in the doped semiconductor ﬁlms LD = (εε0 kT /e2 2N ) to achieve a suﬃcient carrier transfer from the doping to the transport layer (see [182]). Cohen and Barnett [183] have performed simulations of the potential distributions of ZnO/ZnMgO/ZnMgO:Al multilayers and of the electron density distributions that are generated due to charge transfer from the doping layer (ZnMgO:Al) to the transport layer (ZnO). From these calculations the authors conclude that modulation-doped structures with this layer system are possible with mobilities as high as 145 cm2 V−1 s−1 at an average carrier density of 3.8 × 1018 cm−3 . The modulation-doping approach can be used for enhancing TCO properties, for instance for the design of transparent elec-

Fig. 2.16. Carrier transport in a homogenously doped ZnO:Al ﬁlm (left) and in a modulation-doped ﬁlm, consisting of alternating doped Zn1−x Mgx O:Al and undoped ZnO ﬁlms (right). The light grey areas represent the ionized impurities in the ﬁlms, which act as scattering centers. The zigzag lines symbolize the scattering paths of electrons, while the thin straight lines are the electron path without scattering

68

K. Ellmer

tronics (see Sect. 2.4). However, the resistivity of modulation-doped layers would be limited to 1.5 × 10−3 Ω cm since the transferred maximum electron sheet density is only about 1013 cm−2 . To prove this approach, which was suggested by the author in 2001 [67], multilayers of ZnO/ZnO:Al and ZnO/Zn1−xMgx O:Al were prepared on a- and c-plane sapphire substrates with a ﬁxed total thickness of about 550 nm and single layer thicknesses (ZnO and ZnO:Al or Zn1−x Mgx O:Al) down to 3 nm [106]. Sapphire substrates were selected to grow ZnO ﬁlms of high crystalline quality, i.e., ﬁlms that exhibit not only a pronounced c-axis texture but also an in plane orientation. The variation of the structural and electrical parameters is shown in Fig. 2.17 as a function of the single layer thickness. Figure 2.17a shows the full width at half maximum (FWHM2θ ) of the (0002)-diﬀraction peak of ZnO, the half width of the rocking curve the (0002) peak (FWHMω ) together with the resistivity. The electrical parameters carrier concentration and Hall mobility are displayed in Fig. 2.17b. It can be concluded that the structural quality is quite comparable to single ﬁlms, especially for multilayer ﬁlms grown on a-plane sapphire, both with respect to the grain size in c-direction (FWHM2θ ) as well as to the orientation of the grains (FWHMω ). This means that the chemical variation throughout the multilayers does not inﬂuence the crystal growth. However, the resistivity of these multilayers scatters signiﬁcantly 1

10

–1

10

#

–3

10 5

#

4 3 2 1 0 1.0 0.8 0.6 0.4 0.2 0.0

N [cm–3]

(a)

10 10

21

(b)

20

19

18

10 20 µ [cm2/Vs]

FWHM (002)2θ [˚]FWHMω [˚] ρ [Ωcm]

10

#

15 10 5 0

0

200 400 600 ZnO, ZnO:Al thickness [nm]

0

200 400 600 ZnO, ZnO:Al thickness [nm]

Fig. 2.17. (a) Structural parameters (FWHM2θ and FWHMω ) and resistivity of modulation doped ZnO/ZnO:Al ﬁlms deposited on c- (ﬁlled circle) and a-plane (ﬁlled triangle) sapphire and (b) charge carrier density and Hall mobility in dependence of the single layer thickness. The total thickness was kept constant at about 550 nm, while the single layer thickness (ZnO and ZnO:Al) was varied down to 3 nm. Films marked by (#) were deposited at a low oxygen partial pressure. ZnO/Zn1−x Mgx O:Al ﬁlms are marked by (asterisk, open circle). Deposition parameters: p = 0.8 Pa, P = 50 Wrf , Tsub = 250◦ C, xO2 = 0.2 %

2 Electrical Properties

69

and increases by up to 3 orders of magnitude compared with single ﬁlms. Resistivities comparable to single layer ﬁlms could be achieved only for low oxygen partial pressures, including a low background pressure using a high pumping speed (see the values marked by a double cross). This behavior is attributed to a high oxidation sensitivity of thin ﬁlms leading to a passivation of the donors in the ZnO:Al or Zn1−x Mgx O:Al ﬁlms. From Fig. 2.17b it can be seen that especially the carrier concentration decreases signiﬁcantly when the single layer thickness decreases, while the Hall mobility was hardly aﬀected. This eﬀect we ascribe again to the high oxidation sensitivity of very thin ﬁlms. The variation of the work function in the vertical direction of multilayers was investigated by Kelvin probe force microscopy (KPFM) on a fracture cross-section of a ﬁlm deposited on a silicon substrate, shown in Fig. 2.18. This method is described in detail in [184]. A variation of the work function of about 60–100 meV is clearly seen from the bright-dark contrast as well

Fig. 2.18. Kelvin probe force microscopy (KPFM) picture of the cross-section of a modulation doped Zn1−x Mgx O:Al/ZnO ﬁlm on a silicon substrate. The contact potential is given relative to pyrolytic graphite (φ = 4.07 eV). The local variation of the contact potential is shown on the left side, while the chemical composition, determined by SIMS is displayed on the right side. Deposition parameters: p = 0.2 Pa, P = 75 Wrf , Tsub = 300◦ C, single layer thickness dZnO = dZn1−x Mgx O:Al = 160 nm

70

K. Ellmer

as from the local variation of the contact potential for a vertical scan. This corresponds closely to the expected shift of the conduction band edge for a magnesium content of about x = 0.05 [185]. For comparison, the chemical modulation measured by SIMS analysis is shown together with the work function scan. These measurements were performed for layers with a thickness of about 160 nm, since the lateral resolution of the KPFM method is only about 20 nm. The question whether this work function modulation does exist also for much thinner ﬁlms (some nanometer) still has to be investigated. Some multilayers were prepared with Zn1−x Mgx O:Al ﬁlms (marked by ∗ and ◦ in Fig. 2.17). Astonishingly, these layers showed higher resistivities compared with ZnO/ZnO:Al multilayers, which we attribute to the particular deposition conditions when sputtering from a zinc magnesium oxide target. Since Zn1−x Mgx O has a lower electron aﬃnity compared with zinc oxide [186], the formation of negative oxygen ions is favored on a target containing magnesium oxide [187].

2.4 Transparent Field Eﬀect Transistors with ZnO Zinc oxide grows with a good crystalline quality even for room temperature deposition, e.g., when deposited by magnetron sputtering [157]. In this case rather high mobilities of 10–50 cm2 V−1 s−1 are maintained for carrier concentrations below 1019 cm−3 . This is much higher than the mobility in amorphous silicon with typical mobilities of ≈ 1 cm2 V−1 s−1 [188]. Therefore, ZnO was investigated also as a material for thin ﬁlm ﬁeld eﬀect transistors (TFT), which oﬀers the combination of transparent, high mobility TFTs for the next generation of invisible and ﬂexible devices [189]. One perspective could be switching transistors for addressing active matrix displays for organic light emitting diode-based displays [190]. Especially, the low deposition temperatures of ZnO and other TCO ﬁlms are advantageous for the preparation of ﬂexible, lightweight electronics for displays on plastic substrates, for instance polyethylene terephthalate (e.g., Mylar). Compared with organic semiconductors, ZnO exhibits better electronic properties (much higher mobility) and a superior stability at ambient conditions [172]. Such a TFT structure with an RF magnetron sputtered ZnO ﬁlm is shown in Fig. 2.19a [191]. The gate electrode is indium-tin oxide, while the gate insulator consists of Al2 O3 /TiO2 multilayers (thickness 220 nm). Both, the channel layer of undoped ZnO (resistivity ≈ 108 Ω cm) and the source/drain contact layers (ZnO:) were deposited by RF magnetron sputtering at room temperature. Figure 2.19b shows typical current–voltage characteristics of such a ZnO-TFT from which a threshold voltage of 19 V and a saturation mobility of ∼27 cm2 V−1 s−1 can be derived. The on/oﬀ resistances are about 45 kΩ and 20 MΩ, respectively. This example demonstrates that a ZnO-TFT works in principle, even for ZnO ﬁlms deposited at room temperature, opening a new ﬁeld of applications of ZnO as a semiconductor.

2 Electrical Properties

71

(a)

Fig. 2.19. (a) Scheme of a transparent ﬁeld eﬀect transistor based on ZnO [191]. The gate electrode consists of tin-doped indium oxide (ITO) and the gate dielectric is a multilayer of Al2 O3 /TiO2 (ATO). (b) Output characteristics (drain-source current as a function of the drain-source voltage) for diﬀerent gate voltages. The saturation current is about 530 µA at a gate bias of 40 V. From this output characteristics a threshold voltage of 19 V and a ﬁeld-eﬀect mobility of 27 cm2 V−1 s−1 were calculated [192]

Recently, Hossain et al. [178] performed a comprehensive device simulation study to investigate the inﬂuence of the grain barriers (number of barriers per gate length and trap density per barrier) on the properties of ZnO-based thin ﬁlm transistors. By applying the grain boundary trap model (see Sect. 2.2.3) they estimated ﬁeld-eﬀect mobilities of about 10 cm2 V−1 s−1 for grain boundary trap densities Nt in the range of 1 − 3 × 1012 cm2 , which is consistent with reported trap densities in ZnO thin ﬁlm transistors [172]. These trap densities are lower or comparable to trap densities estimated from electrical transport measurements (see Table 2.6). Another important result of the simulation was the quantitative estimation of the inﬂuence of the number of grain barriers over the channel length of the TFT on the eﬀective mobility µFE in the channel. It was found that µFE decreases exponentially as the number of grain barriers increases. In conclusion, transparent ﬁeld eﬀect transistors based on zinc oxide are very promising for future applications.

2.5 Search for p-Type Conductivity in ZnO First attempts to produce p-type ZnO were reported in the 1950s by Lander from Bell Labs [40]. He introduced alkali atoms, especially lithium, into intrinsic ZnO crystals. Depending on the concentration and the diﬀusion temperature, lithium acts both as a donor (interstitial Li+ ) as well as an acceptor. The lithium becomes an acceptor by displacing a zinc ion from a lattice site according to the equation: + − − ZnZn + Li+ i + e ⇐⇒ Zni + Lii ◦

(2.33)

By a lithium treatment at 300 C the intrinsic ZnO crystals with a resistivity of about 10 Ω cm reached resistivities up to 109 Ω cm. However, p-type

72

K. Ellmer

conductivity was not observed. Lander comes to the conclusion that an equi− librium of Li+ i and Li over a broad range of zinc pressures leads to a nearly perfect compensation. P-type conductivity could be achieved only at impractically high oxygen partial pressures. A comparable behavior was observed for Cu and Ag doping of ZnO. While a strong compensation can be achieved, p-conductivity was not observed [49, 193]. A new search for p-type ZnO set in at the end of the 1990s after the tremendous success of gallium nitride as a wide band gap semiconductor. However, only occasionally persistent p-type conduction was reported in zinc oxide. Although in many papers p-type ZnO was claimed, often the results were doubtful and the p-type character vanished after hours, days, or weeks [52]. While the built-in of acceptors in zinc oxide is proven unambiguously by optical methods (especially photoluminescence measurements [39, 194]) electrical analysis often shows only n-type conductivity because of a severe compensation of the acceptors. In view of this unclear situation, only a short overview is given, based on recent work of Look et al. [195]. Look et al. prepared phosporous-doped ZnO ﬁlms on sapphire, which exhibited an acceptor concentration of about 2 × 1017 cm−3 and a mobility of 2–4 cm2 V−1 s−1 . Depending on the illumination conditions (UV light), the conductivity could be reversibly changed between p- and n-type, which was tentatively attributed to the generation of donor states at the surface. In another paper of Looks group, ZnO:N was investigated as a p-type ZnO [195]. Though a mobility of about 1 cm2 V−1 s−1 could be measured, illumination changed that material to a high-mobility n-type ZnO. This n-type behavior persisted for several days. In contrast to earlier expectations the nitrogen acceptor produces layers of lower mobility. Phosphorous-doping seems to be more attractive with mobilities of about 3 cm2 V−1 s−1 . The authors conclude that a fundamental understanding of the p-type ZnO begins to emerge but additional work is required. A recent density-functional theory study found plausible reasons for the compensation mechanism of N acceptors in ZnO [196]. At low N doping levels the nitrogen acceptors are compensated by oxygen vacancies. Increasing the built-in of nitrogen (for instance by using a nitrogen plasma) the N acceptors are still mostly compensated by N2 molecules at oxygen sites (compare with Sect. 2.1.1.1) and by N–acceptor–N2 complexes. Acknowledgement. The assistance of G¨ otz Vollweiler, Rainald Mientus and Franziska Liersch in thin ﬁlm preparation and electrical measurements is gratefully acknowledged.

References 1. A.A. Somerville, Phys. Rev. I 34, 311 (1912) 2. C. Wagner, W. Schottky, Z. Phys. Chem. B 11, 163 (1930)

2 Electrical Properties

73

3. K. Hauﬀe, J. Block, Z. Phys. Chem. 196, 438 (1950) 4. H.H. von Baumbach, C. Wagner, Z. Phys. Chem. 22, 199 (1933) 5. E. Mollwo, in Landoldt-B¨ ornstein, NS III 17b, ed. by O. Madelung, M. Schulz, H. Weiss (Springer-Verlag, Berlin, 1982) 6. E. Mollwo, Z. Physik 138, 478 (1954) 7. G. Heiland, E. Mollwo, F. St¨ ockmann, Solid State Phys. 8, 191 (1959) 8. W. Meyer, H. Neldel, Z. techn. Phys. 18, 588 (1937) 9. O. Fritsch, Ann. Phys. 5, 375 (1935) 10. W. Shockley, Electrons and Holes in Semiconductors, (van Nostrand, Toronto, 1950) 11. E. Scharowsky, Z. Phys. 135, 318 (1953) 12. G. Bogner, E. Mollwo, J. Phys. Chem. Solids 6, 136 (1958) 13. D.G. Thomas, J.J. Lander, J. Chem. Phys. 25, 1136 (1956) 14. R.A. Laudise, E.D. Kolb, A.J. Caporaso, J. Am. Cer. Soc. 47, 9 (1964) 15. E.E. Hahn, J. Appl. Phys. 22, 855 (1951) 16. S.E. Harrison, Phys. Rev. 93, 52 (1954) 17. A.R. Hutson, Phys. Rev. 108, 222 (1957) 18. D.C. Look, J.W. Hemsky, J.R. Sizelove, Phys. Rev. Lett. 82, 2552 (1999) 19. G.W. Tomlins, J.L. Routbort, T.O. Mason, J. Appl. Phys. 87, 117 (2000) 20. P. Erhart, K. Albe, Phys. Rev. B 73, 115207 (2006) 21. P. Erhart, K. Albe, Appl. Phys. Lett. 88, 201918 (2006) 22. P. Erhart, K. Albe, A. Klein, Phys. Rev. B 73, 205203 (2006) 23. K. Hauﬀe, C. Seyferth, Reaktionen in und an festen Stoﬀen, (Springer-Verlag, Berlin, 1966) 24. D.G. Thomas, J. Phys. Chem. Solids 9, 31 (1958) 25. F.A. Kr¨ oger, The Chemistry of Imperfect Crystals. Imperfection Chemistry of Crystalline Solids, vol. 2, 2nd edn. (North-Holland, Amsterdam, 1973) 26. T. Minami, MRS Bull. 25(Aug), 38 (2000) 27. N. Roberts, R.P. Wang, A.W. Sleight, W.W. Warren, Phys. Rev. B 57, 5734 (1998) 28. D.G. Thomas, J.J. Lander, in Halbleiter und Phosphore, ed. by M. Sch¨ on, H. Welker (Vieweg, Braunschweig, 1958), p. 534 29. C.G. van de Walle, Phys. Rev. Lett. 85, 1012 (2000) 30. D.M. Hofmann, A. Hofstaetter, F. Leiter, H. Zhou, F. Henecker, B.K. Meyer, S.O. Orlinskij, J. Schmidt, P.G. Baranov, Phys. Rev. Lett. 88, 045504 (2002) 31. P.W. Peacock, J. Robertson, Appl. Phys. Lett. 83, 2025 (2003) 32. J. Robertson, K. Xiong, S.J. Clark, Thin Solid Films 496, 1 (2006) 33. N. Nickel, in Zinc Oxide - A Material for Micro- and Optoelectronic Applications, ed. by N. Nickel, E. Terukov (Springer, Dordrecht, 2005), pp. 145–155 34. A.R. Hutson, J. Phys. Chem. Solids 8, 467 (1959) 35. P. Wagner, R. Helbig, J. Phys. Chem. Solids 35, 327 (1974) 36. K.I. Hagemark, L.C. Chacka, J. Solid State Chem. 15, 261 (1975) 37. D.C. Look, D.C. Reynolds, J.R. Sizelove, R.L. Jones, C.W. Litton, G. Cantwell, W.C. Harsch, Solid State Commun. 105, 399 (1998) 38. H. von Wenckstern, S. Weinhold, G. Biehne, R. Pickenhain, H. Schmidt, H. Hochmuth, M. Grundmann, Adv. Solid State Phys. 45, 263 (2005) 39. B.K. Meyer, J. Sann, D.M. Hofmann, C. Neumann, A. Zeuner, Semicond. Sci. Technol. 20, S62 (2005) 40. J.J. Lander, J. Phys. Chem. Solids 15, 324 (1960)

74

K. Ellmer

41. O.F. Schirmer, J. Phys. Chem. Solids 29, 1407 (1968) 42. A.R. Hutson, Phys. Rev. Lett. 4, 505 (1960) 43. R. Wang, J.H. Xin, Y. Yang, H. Liu, L. Xu, J. Hu, Appl. Surf. Sci. 227, 312 (2004) 44. Y. Yan, S.B. Zhang, S.T. Pantelides, Phys. Rev. Lett. 86, 5723 (2001) 45. P. Fons, H. Tampo, A.V. Kolobov, M. Ohkubo, S. Niki, J. Tominaga, R. Carboni, F. Boscherini, S. Friedrich, Phys. Rev. Lett. 96, 045504 (2006) 46. B.K. Meyer, H. Alves, D.M. Hofmann, W. Kriegseis, D. Forster, F. Bertram, J. Christen, A. Hoﬀmann, M. Straßburg, M. Dworzak, U. Haboeck, A.V. Rodina, Phys. Stat. Sol. (b) 241, 231 (2004) 47. J. Fan, R. Freer, J. Appl. Phys. 77, 4795 (1995) 48. A.N. Gruzintsev, V.T. Volkov, E.E. Yakimov, Semicond. 37, 259 (2003) 49. R.E. Dietz, H. Kamimura, M.D. Sturge, A. Yariv, Phys. Rev. 132, 1559 (1963) 50. D. Zwingel, J. Lumin. 5, 385 (1972) 51. D. Pﬁsterer, J. Sann, D.M. Hofmann, M. Plana, A. Neumann, M. Lerch, B.K. Meyer, Phys. Stat. Sol. (a) 243, R1 (2005) 52. B. Claﬂin, D.C. Look, S.J. Park, G. Cantwell, J. Cryst. Growth 287, 16 (2006) 53. P.W. Li, K.I. Hagemark, J. Solid State Chem. 12, 371 (1975) 54. D.L. Rode, in Semiconductors and Semimetals, ed. by R.K. Willardson, A.C. Beer (Academic, New York, 1975), pp. 1–89 55. S.S. Devlin, in Physics and Chemistry of II-VI Compounds, ed. by M. Aven, J.S. Prener (North-Holland, Amsterdam, 1967), pp. 549–609 56. J. Bardeen, W. Shockley, Phys. Rev. 80, 72 (1950) 57. K. Seeger, Semiconductor Physics (Springer-Verlag, Berlin, 1991) 58. C. Solbrig, Z. Phys. 211, 429 (1968) 59. A. Ballato, IEEE Ultrasonics Symp., 575 (1996) 60. J.D. Zook, Phys. Rev. 136, A869 (1964) 61. K.W. B¨ oer, Survey of Semiconductor Physics. Electrons and Other Particles in Bulk Semiconductors (Van Nostrand Reinhold, New York, 1990) 62. C. Klingshirn, H. Priller, M. Decker, J. Br¨ uckner, H. Kalt, R. Hauschild, J. Zeller, A. Waag, A. Bakin, H. Wehmann, K. Thonke, R. Sauer, R. Kling, F. Reuss, C. Kirchner, Adv. Solid State Phys. 45, 275 (2005) 63. A. Mang, K. Reimann, S. R¨ ubenacke, Solid State Commun. 94, 251 (1995) 64. T. Makino, Y. Segawa, A. Tsukazaki, A. Ohtomo, M. Kawasaki, Appl. Phys. Lett. 87, 022101 (2005) 65. N. Ashkenov, B.N. Mbenkum, C. Bundesmann, V. Riede, M. Lorenz, D. Spemann, E.M. Kaidashev, A. Kasic, M. Schubert, M. Grundmann, G. Wagner, H. Neumann, V. Darakchieva, H. Arwin, B. Monemar, J. Appl. Phys. 93, 126 (2003) ¨ ur, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, 66. U. Ozg¨ V. Avrutin, S.J. Cho, H. Morkoc, J. Appl. Phys. 98, 041301 (2005) 67. K. Ellmer, J. Phys. D: Appl. Phys. 34, 3097 (2001) 68. J.D. Albrecht, P.P. Ruden, S. Limpijumnong, W.R.L. Lambrecht, K.F. Brennan, J. Appl. Phys. 86, 6864 (1999) 69. T. Minami, H. Sato, K. Ohashi, T. Tomofuji, S. Takata, J. Cryst. Growth 117, 370 (1992) 70. T.B. Bateman, J. Appl. Phys. 33, 3309 (1962) 71. G. Carlotti, D. Fioretto, G. Socino, E. Verona, J. Phys.: Condens. Mat. 7, 9147 (1995)

2 Electrical Properties

75

72. T. Azuhata, M. Takesata, T. Yagi, A. Shikanai, S.F. Chichibu, K. Torii, A. Nakamura, T. Sota, G. Cantwell, D.B. Eason, C.W. Litton, J. Appl. Phys. 94, 968 (2003) 73. I.B. Kobiakov, Solid State Commun. 35, 305 (1980) 74. E. Conwell, V.F. Weisskopf, Phys. Rev. 77, 388 (1950) 75. E. Conwell, V.F. Weisskopf, Phys. Rev. 69, 258 (1946) 76. H. Brooks, Adv. Electron. Electron Phys. 7, 85 (1955) 77. H. Brooks, Phys. Rev. 83, 879 (1951) 78. V.I. Fistul, Heavily Doped Semiconductors (Plenum, New York, 1969) 79. D. Chattopadhyay, H.J. Queisser, Rev. Mod. Phys. 53, 745 (1981) 80. R.B. Dingle, Phil. Mag. 46, 831 (1955) 81. W. Zawadzki, in Handbook on Semiconductors, ed. by T.S. Moss (NorthHolland, Amsterdam, 1982), pp. 713–803 82. R.T. Bate, R.D. Baxter, F.J. Reid, A.C. Beer, J. Phys. Chem. Solids 26, 1205 (1965) 83. T. Pisarkiewicz, K. Zakrzewska, E. Leja, Thin Solid Films 174, 217 (1989) 84. S. Brehme, F. Fenske, W. Fuhs, E. Neubauer, M. Poschenrieder, B. Selle, I. Sieber, Thin Solid Films 342, 167 (1999) 85. N.G. Nilsson, Phys. Stat. Sol. (a) 19, K75 (1973) 86. J.S. Blakemore, Solid-State Electr. 25, 1067 (1982) 87. D. Mergel, Z. Qiao, J. Appl. Phys. 95, 5608 (2004) 88. H.L. Hartnagel, A.L. Dawar, A.K. Jain, C. Jagadish, Semiconducting Transparent Thin Films (Institute of Physics Publishing, Bristol, 1995) 89. C. Erginsoy, Phys. Rev. 79, 1013 (1950) 90. K.M. Itoh, W. Walukiewicz, H.D. Fuchs, J.W. Beeman, E.E. Haller, J.W. Farmer, V.I. Ozhogin, Phys. Rev. B 50, 16995 (1994) 91. G.B. Stringfellow, H.T. Hall, R.A. Burmeister, J. Appl. Phys. 46, 3006 (1975) 92. E.F. Venger, A.V. Melnichuk, L.Y. Melnichuk, Y.A. Pasechnik, Phys. Stat. Sol. (b) 188, 823 (1995) 93. J. Nause, B. Nemeth, Semicond. Sci. Technol. 20, S45 (2005) 94. H. Rupprecht, J. Phys. Chem. Solids 6, 144 (1958) 95. W.S. Baer, Phys. Rev. 154, 785 (1967) 96. A. Hausmann, W. Teuerle, Z. Phys. 259, 189 (1973) 97. B. Utsch, A. Hausmann, Z. Phys. B 21, 27 (1975) 98. F.P. Koﬀyberg, Can. J. Phys. 49, 435 (1971) 99. H. Finkenrath, H. K¨ ohler, M. Lochmann, Z. Angew. Phys. 21, 512 (1966) 100. R.L. Weiher, J. Appl. Phys. 33, 2834 (1962) 101. Z.M. Jarzebski, J.P. Marton, J. Electrochem. Soc. 123, 299C (1976) 102. J.C. Bean, in High-Speed Semiconductor Devices, ed. by S.M. Sze (Wiley, New York, 1990), pp. 13–55 103. G. Masetti, M. Severi, S. Solmi, IEEE Trans. Electron Dev. ED30, 764 (1983) 104. D.B.M. Klaassen, Solid-State Electr. 35, 953 (1992) 105. P. Fons, K. Iwata, A. Yamada, K. Matsubara, S. Niki, K. Nakahara, T. Tanabe, H. Takasu, Appl. Phys. Lett. 77, 1801 (2000) 106. K. Ellmer, G. Vollweiler, Thin Solid Films 496, 104 (2006) 107. M. Matsuoka, Jpn. J. Appl. Phys. 10, 736 (1971) 108. G.D. Mahan, J. Appl. Phys. 54, 3825 (1983) 109. L.M. Levinson, H.R. Philipp, Ceram. Bull. 65, 639 (1986) 110. T.K. Gupta, J. Am. Cer. Soc. 73, 1817 (1990)

76 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125.

126. 127. 128. 129.

130. 131. 132. 133. 134. 135. 136. 137. 138.

139. 140. 141. 142. 143. 144.

K. Ellmer D.R. Clarke, J. Appl. Phys. 50, 6829 (1979) F. Greuter, G. Blatter, Semicond. Sci. Technol. 5, 111 (1990) P.R. Emtage, J. Appl. Phys. 48, 4372 (1977) L.M. Levinson, H.R. Philipp, J. Appl. Phys. 46, 1332 (1975) J.G. Simmons, J. Appl. Phys. 34, 1793 (1963) H.R. Philipp, L.M. Levinson, J. Appl. Phys. 48, 1621 (1977) F.S. Hickernell, J. Appl. Phys. 44, 1061 (1973) N.F. Foster, G.A. Rozgonyi, Appl. Phys. Lett. 8, 221 (1966) N.F. Foster, J. Vac. Sci. Technol. 6, 111 (1969) J.B. Webb, D.F. Williams, M. Buchanan, Appl. Phys. Lett. 39, 640 (1981) G.L. Harding, B. Window, E.C. Horrigan, Sol. Energy Mater. 22, 69 (1991) J. Hu, R.G. Gordon, Solar Cells 30, 437 (1991) K. Saito, Y. Watanabe, K. Takahashi, T. Matsuzawa, B. Sang, M. Konagai, Sol. Energy Mat. Sol. Cells 49, 187 (1997) A. Suzuki, T. Matsushita, N. Wada, Y. Sakamoto, M. Okuda, Jpn. J. Appl. Phys. 35, L56 (1996) E.M. Kaidashev, M. Lorenz, H. vonWenckstern, A. Rahm, H.C. Semmelhack, K.H. Han, G. Benndorf, C. Bundesmann, H. Hochmuth, M. Grundmann, Appl. Phys. Lett. 82, 3901 (2003) J.H. Morgan, D.E. Brodie, Can. J. Phys. 60, 1387 (1982) J. Ma, F. Ji, D. Zhang, H. Ma, S. Li, Thin Solid Films 357, 98 (1999) S. Major, A. Banerjee, K.L. Chopra, Thin Solid Films 108, 333 (1983) D.R. Acosta, O. Lovera, A. Maldonado, R. Asomoza, H. Gomez, A. Palafox, M. d. l. Olvera, J. Palacios-Gomez, Mater. Res. Soc. Symp. Proc. 355, 587 (1994) T. Okamura, Y. Seki, S. Nakagari, H. Okushi, Jpn. J. Appl. Phys. 31, 3218 (1992) T. Nagase, T. Ooie, J. Sakakibara, Thin Solid Films 357, 151 (1999) T. Yoshida, M. Iwaya, H. Ando, T. Oekermann, K. Nonomura, D. Schlettwein, D. W¨ ohrle, H. Minoura, Chem. Commun., 400–401 (2004) I. Sieber, N. Wanderka, I. Urban, I. D¨ orfel, E. Schierhorn, F. Fenske, W. Fuhs, Thin Solid Films 330, 108 (1998) L. Sagalowicz, G.R. Fox, J. Mater. Res. 14, 1876 (1999) B. P¨ od¨ or, Phys. Stat. Sol. 16, K167 (1966) D.C. Look, J.R. Sizelove, Phys. Rev. Lett. 82, 1237 (1999) D. Cherns, C.G. Jiao, Phys. Rev. Lett. 87, 205504 (2001) E. M¨ uller, D. Litvinov, D. Gerthsen, C. Kirchner, A. Waag, N. Oleynik, A. Dadgar, A. Krost, in Zinc Oxide - A Material for Micro- and Optoelectronic Applications, ed. by N. Nickel, E. Terukov (Springer, Dordrecht, 2005), pp. 99–111 D. Gerthsen, D. Litvinov, T. Gruber, C. Kirchner, A. Waag, Appl. Phys. Lett. 81, 3972 (2002) C. Coskun, D.C. Look, G.C. Farlow, J.R. Sizelove, Semicond. Sci. Technol. 19, 752 (2004) J.J. Harris, K.J. Lee, J.B. Webb, H. Tang, I. Harrison, L.B. Flannery, T.S. Cheng, C.T. Foxon, Semicond. Sci. Technol. 15, 413 (2000) J.Y. Seto, J. Appl. Phys. 46, 5247 (1975) G. Baccarani, B. Ricco, G. Spadini, J. Appl. Phys. 49, 5565 (1978) S.M. Sze, Physics of Semiconductor Devices (John Wiley & Sons, New York, 1981)

2 Electrical Properties

77

145. J. Orton, M. Powell, Rep. Prog. Phys. 43, 1263 (1980) 146. J. Oertel, K. Ellmer, W. Bohne, J. R¨ohrich, H. Tributsch, J. Cryst. Growth 198/199, 1205 (1999) 147. J. Werner, in Polycrystalline Semiconductors III-Physics and Technology, ed. by H. Strunk, J. Werner, B. Fortin, O. Bonaud (Scitec Publ., Switzerland, Zug, Switzerland, 1993), p. 534 148. R. Lipperheide, T. Weis, U. Wille, Sol. Energy Mat. Sol. Cells 65, 157 (2001) 149. R. Lipperheide, T. Weis, U. Wille, J. Phys.: Condens. Mater. 13, 3347 (2001) 150. R. Lipperheide, U. Wille, Phys. Rev. B 68, 115315 (2003) 151. T. Weis, S. Brehme, P. Kanschat, W. Fuhs, R. Lipperheide, U. Wille, J. Non-Cryst. Solids 299–302, 380 (2002) 152. T. Weis, R. Lipperheide, U. Wille, S. Brehme, J. Appl. Phys. 92, 1411 (2002) 153. T. Kamins, Polycrystalline Silicon for Integrated Circuit Applications (Kluwer, Boston, 1988) 154. C.G. Granqvist, A. Hult˚ aker, Thin Solid Films 411, 1 (2002) 155. K. Blankenbach, in Adv. Solid State Phys., vol. 40, ed. by B. Kramer (Vieweg, Braunschweig, 2000), pp. 657–668 156. J. Hu, R.G. Gordon, J. Electrochem. Soc. 139, 2014 (1992) 157. K. Ellmer, J. Phys. D: Appl. Phys. 33, R17 (2000) 158. J.R. Bellingham, W.A. Phillips, C.J. Adkins, J. Mater. Sci. Lett. 11, 263 (1992) 159. R. Wang, L.L.H. King, A.W. Sleight, J. Mater. Res. 11, 1659 (1996) 160. J. Bruneaux, H. Cachet, M. Froment, A. Messad, Electrochim. Acta 39, 1251 (1993) 161. D.B.M. Klaassen, Solid-State Electr. 35, 961 (1992) 162. P. Ebert, Z. Zhang, F. Kluge, M. Simon, Z. Zhang, K. Urban, Phys. Rev. Lett. 83, 757 (1999) 163. M. Kon, P.K. Song, Y. Shigesato, P. Frach, S. Ohno, K. Suzuki, Jpn. J. Appl. Phys. 42, 263 (2003) 164. H. Agura, A. Suzuki, T. Matsushita, T. Aoki, M. Okudaa, Thin Solid Films 445, 263 (2003) 165. M. Lorenz, E.M. Kaidashev, H. vonWenckstern, V. Riede, C. Bundesmann, D. Spemann, G. Benndorf, H. Hochmuth, A. Rahm, H.C. Semmelhack, M. Grundmann, Solid-State Electr. 47, 2205 (2003) 166. M. Kon, P.K. Song, Y. Shigesato, P. Frach, A. Mizukami, K. Suzuki, Jpn. J. Appl. Phys. 41, 814 (2002) 167. T. Minami, S. Suzuki, T. Miyata, Mater. Res. Soc. Symp. Proc. 666, F1.3.1 (2001) 168. C. Agashe, O. Kluth, J. H¨ upkes, U. Zastrow, B. Rech, M. Wuttig, J. Appl. Phys. 95, 1911 (2004) 169. H. Nanto, T. Minami, S. Shooji, S. Takata, J. Appl. Phys. 55, 1029 (1984) 170. K.S. Weißenrieder, J. M¨ uller, Thin Solid Films 300, 30 (1997) 171. F.R. Blom, F.C.M. van de Pol, G. Bauhuis, T.J.A. Popma, Thin Solid Films 204, 365 (1991) 172. P.F. Carcia, R.S. McLean, M.H. Reilly, G. Nunes, Appl. Phys. Lett. 82, 1117 (2003) 173. K. Ellmer, R. Mientus, Thin Solid Films Available online since June 14th (2007), DOI: 10.1016/j.tsf2007.05.084, will be published in 2007. 174. J. Szczyrbowski, K. Schmalzbauer, H. Hoﬀmann, Thin Solid Films 137, 169 (1986)

78

K. Ellmer

175. A. Messad, J. Bruneaux, H. Cachet, M. Froment, J. Mater. Sci. 29, 5095 (1994) 176. H. Witte, A. Krtschil, E. Schrenk, K. Fluegge, A. Dadgar, A. Krost, J. Appl. Phys. 97, 043710 (2005) 177. V. Kazukauskas, G. K¨ uhnel, W. Siegel, Appl. Phys. Lett. 70, 1751 (1997) 178. F.M. Hossain, J. Nishi, S. Takagi, A. Ohtomo, T. Fukumura, H. Fuijioka, H. Ohno, H. Koinuma, M. Kawasaki, J. Appl. Phys. 94, 7768 (2003) 179. R. Dingle, H.L. St¨ ormer, A.C. Gossard, W. Wiegmann, Appl. Phys. Lett. 33, 665 (1978) 180. J.J. Harris, J.A. Pals, R. Woltjer, Rep. Prog. Phys. 52, 1217 (1989) 181. I.A. Rauf, Mater. Lett. 18, 123 (1993) 182. A.A. Grinberg, S. Luryi, J. Appl. Phys. 61, 1181 (1987) 183. D.J. Cohen, S.A. Barnett, J. Appl. Phys. 98, 053705 (2005) 184. T. Glatzel, D. FuertesMarron, T. Schedel-Niedrig, S. Sadewasser, M.C. Lux-Steiner, Appl. Phys. Lett. 81, 2017 (2002) 185. A. Ohtomo, M. Kawasaki, T. Koida, K. Masubuchi, H. Koinuma, Y. Sakurai, Y. Yoshida, T. Yasuda, Y. Segawa, Appl. Phys. Lett. 72, 2466 (1998) 186. G.V. Rao, F. S¨ auberlich, A. Klein, Appl. Phys. Lett. 87, 032101 (2005) 187. J.J. Cuomo, R.J. Gambino, J.M.E. Harper, J.D. Kuptsis, J.C. Webber, J. Vac. Sci. Technol. 15, 281 (1978) 188. R. K¨ onenkamp, Photoelectric Properties and Applications of Low-Mobility Semiconductors (Springer, Berlin Heidelberg Newyork, 2000) 189. G. Thomas, Nature 389, 907 (1997) 190. H. Ohta, K. Nomura, H. Hiramatsu, K. Ueda, T. Kamiya, M. Hirano, H. Hosono, Solid-State Electr. 47, 2261 (2003) 191. E.M.C. Fortunato, P.M.C. Barquinha, A.C.M.B. Pimentel, A.M.F. Goncalves, A.J.S. Marques, L.M.N. Pereira, R.F.P. Martins, Adv. Mater. 17, 590 (2005) 192. E.M.C. Fortunato, P.M.C. Barquinha, A.C.M.B.G. Pimentel, A.M.F. Goncalves, A.J.S. Marques, R.F.P. Martins, L.M.N. Pereira, Appl. Phys. Lett. 85, 2541 (2004) 193. J. Fan, R. Freer, J. Am. Cer. Soc. 77, 2263 (1994) 194. B.K. Meyer, J. Sann, A. Zeuner, Adv. Solid State Phys. 45, 289 (2005) 195. D.C. Look, Semicond. Sci. Technol. 20, S55 (2005) 196. E.C. Lee, Y.S. Kim, Y.G. Jin, K.J. Chang, Phys. Rev. B 64, 085120 (2001)

3 Optical Properties of ZnO and Related Compounds C. Bundesmann, R. Schmidt-Grund, and M. Schubert

In this chapter some of the presently known optical properties of zinc oxide are reviewed. In particular, the anisotropic dielectric functions (DFs) of ZnO and related compounds from the far-infrared (FIR) to the vacuum-ultraviolet (VUV) spectral range are studied. Thereupon, many fundamental physical parameters can be derived, such as the optical phonon-mode frequencies and their broadening values, the free-charge-carrier parameters, the static and “high-frequency” dielectric constants, the dispersion of the indices of refraction within the band-gap region, the fundamental and above-band-gap bandto-band transition energies and their excitonic contributions.

3.1 Introduction ZnO has regained much interest because of a large variety of properties, which make ZnO superior to currently used materials. For instance, ZnObased alloys might become alternative materials for GaN-based alloys, which are currently used in ultraviolet (UV) optoelectronic devices. The fundamental band gap energy of ZnO is similar to that of GaN, but the ground-state exciton binding energy of ZnO is more than two times larger than that of GaN (Table 3.1). Hence, stimulated excitonic emission even above room temperature (RT) is possible in ZnO, but not in GaN. ZnO-based laser devices with low threshold currents operating above RT are quite likely [1]. In fact, optically pumped lasing at RT was already demonstrated by several groups [1–5]. Upon alloying with MgO or CdO, the fundamental band gap of ZnO can be shifted to higher or lower energies, respectively (Table 3.1) [11–16]. Furthermore, the electrical n-type conductivity of ZnO can be controlled over many orders of magnitude by doping with Al or Ga [17–20]. On the other hand, reproducible p-type conductivity in ZnO is still a challenge. Doping with group-I elements (Li, Na, K, etc.), which are supposed to substitute the Zn-atoms, or doping with group-V elements (N, P, As, Sb, etc.), which are supposed to substitute the O-atoms, are promising pathways toward p-type conductivity [18, 21, 22]. Upon alloying with Mn or other transition metals, ZnO can reveal ferromagnetic properties with a Curie temperature above RT [23–28]. Essential for the performance of the above addressed materials is the knowledge of fundamental properties.

80

C. Bundesmann et al.

1 Table 3.1. Band-gap energy Eg and the (ground state) exciton binding energy Exb of ZnO, GaN, MgO, and CdO

ZnO Eg 1 Exb

(eV) (meV)

3.37 60

[6] [10]

GaN 3.30 18–28

MgO [7] [7]

7.6 80

[8] [8]

CdO 2.3 −

[9]

By evaluation of the phonon-mode frequencies, information about strain [29] or about the incorporation of doping or alloying atoms can be derived. Besides the phonon-mode frequency, the phonon-mode broadening parameter provides information about crystal quality [30], because scattering due to a lower crystal quality or due to alloying makes the phonon-mode broadening parameter larger. Plasmons – the collective free-charge-carrier modes – interact with the longitudinal phonons, when the screened plasma frequency ωp is in the frequency range of the optical phonons. Thereupon, coupled phonon–plasmon modes are formed. In principle, both Raman scattering and infrared (IR) spectroscopic techniques are able to detect these modes. Unfortunately, the Raman scattering intensity of the uncoupled longitudinal phonon modes of ZnO is weak [31], such that the coupled modes cannot be detected. In contrast to that, IR spectroscopic techniques, for instance infrared spectroscopic ellipsometry (IRSE), can clearly reveal the plasmon-mode contributions to the DFs, and the free-charge-carrier properties can be quantiﬁed. For highlyconductive ZnO the plasma frequency moves from the mid-IR into the near-IR or even into the visible spectral region. Therefore, spectroscopic techniques in the near-IR and visible spectral region are more appropriate for studying the free-charge-carrier parameters in highly-conductive ZnO thin ﬁlm samples (see paragraph optical properties in Sects. 6.2.5.2, p. 273 and 8.3.1, Fig. 8.13). The DFs of ZnO reveal a complex behavior in the vicinity of the fundamental band gap energy, as well as for photon energies above the band gap energy, where multiple electronic transitions occur, dispersed over wide regions of the Brillouin zone. The unambiguous assignment of type, symmetry, and location of a given transition within the Brillouin zone can only be done in conjunction with theoretical band structure analysis, except for the lowest Γ -point transitions. Here, experimentally observed DF spectra are discussed, as well as their analysis in terms of critical point (CP) structures and associated band-to-band transitions without attempting to connect those to individual symmetry points. The fundamental band gap energy Eg is the lowest energy, at which absorption sets in. It can be determined, for instance, by transmission measurements. In ZnO, a semiconductor with dipole-allowed direct band-to-band transitions, excitons couple strongly to the radiation ﬁeld, and their absorption lines are superimposed onto the CP structures because of the band-to-band transitions. Homogeneous and inhomogeneous broadening eﬀects smear out the absorption features, giving rise to a wide range of reported Eg values, in addition to intrinsic origins for energy shifts,

3 Optical Properties of ZnO and Related Compounds

81

such as impurities, strain, composition variations, etc. Spectroscopic ellipsometry (SE) can be used to separate excitonic contributions from band-to-band transitions and for their quantitative assessment. Further attention is paid to the temperature-dependencies of phonon modes, indices of refraction, and fundamental-band-gap energies. Wurtzite-structure ZnO thin ﬁlms grown by a variety of deposition techniques, as well as commercially available single crystal bulk samples are discussed. Furthermore, data for ZnO thin ﬁlms intermixed with numerous elements are reviewed. Most of the results are obtained by SE, which is a precise and reliable tool for measurements of the DFs. The SE results are supplemented by Raman scattering and electrical Hall-eﬀect measurement data, as well as data reported in the literature by similar or alternative techniques (reﬂection, transmission, and luminescence excitation spectroscopy). Raman scattering was often applied for studying the phonon modes of ZnO bulk samples [31–38]. It has become a fast and reliable tool to study ZnO thin ﬁlms [29, 38–43], and ZnO nano- and/or microstructures [44–46]. Raman scattering studies were also reported for ZnO samples doped with Li [43, 47, 48], N [43, 49–51], Al [48, 52–54], P [43, 55], Mn [43, 56–58], Fe [43, 48], Co [43], Ni [43], Cu [43], Ga [48, 51], As [59], Ce [60], or Sb [48, 61], and (Mg,Cd)x Zn1−x O [43, 62] samples. IR reﬂection [63–65] and transmission measurements [66–68] were reported mainly for ZnO bulk materials. IR optical studies of doped ZnO and ZnO-based thin ﬁlms are, in general, restricted to transmission and reﬂection measurements in the near-IR (NIR) spectral region, and to highly conductive Al-doped ZnO thin ﬁlms. Some experiments were performed in the mid-IR (MIR) spectral region [52, 69, 70], where the optical phonon modes can be studied. Recently, IRSE was applied to study undoped and doped ZnO ﬁlms, and ZnO-based alloy ﬁlms [30, 38, 43, 62, 71–74]. The number of spectroscopic studies of ZnO and related materials in the spectral region around the fundamental band gap is too large to be listed here. Undoped and doped ZnO as well as ZnO-based alloy samples were studied by photoluminescence (PL), transmission and reﬂection measurements (see review articles [6,75]). Also, SE measurements were reported for ZnO [76–85] and related materials, for instance, metal-doped ZnO [70,86–89], Mgx Zn1−x O [15, 16, 82, 90, 91], Cox Zn1−x O [92], Mnx Zn1−x O [93], or Fex Zn1−x O [94].

3.2 Basic Concepts and Properties 3.2.1 Structural Properties Undoped and doped ZnO, and most of the ZnO-based alloys crystallize under normal conditions in the wurtzite structure, but ZnO-based alloys can reveal a rocksalt structure for a high content of alloying atoms. One example is Mg-rich Mgx Zn1−x O. Thus, a phase transition with change of coordination

82

C. Bundesmann et al.

(b)

(a)

a2

c c-plane (0001)

a1

a3

c

a-plane (1120)

a

r-plane (1102)

Fig. 3.1. (a) Primitive cell (heavy lines) of the wurtzite-structure lattice placed within a hexagonal prism. a and c are the lattice constants. (b) Schematic drawing of surfaces cut from a hexagonal single crystal with diﬀerent crystallographic orientations (surface planes)

must occur, which aﬀects many physical properties. The phase transition for PLD-grown Mgx Zn1−x O thin ﬁlms was observed in the composition range between 53% and 68% Mg content, whereas scattered reports exist on more diﬀuse transition regions for other Mgx Zn1−x O samples [11, 14–16, 82, 90, 91, 95–100]. 3.2.1.1 Wurtzite Crystal Structure (Hexagonal) The wurtzite crystal structure belongs to the hexagonal system with space 4 group C6v (P 63 mc) in the Schoenﬂies (short standard) notation. Two atom species occupy the positions of a closest packed hexagonal lattice each. These two sublattices are shifted along the c-axis against each other (Fig. 3.1 a). The wurtzite-structure lattice is fourfold coordinated. That is, each atom has four nearest neighbor atoms. Figure 3.1b shows the cuts of diﬀerent orientations of a crystal with hexagonal structure.1

1

A sample is called “c-plane oriented” or shortly “c-plane,” when the investigated surface is the hexagonal (0001)-plane. Analogously, for “a-plane” or “r-plane” samples the investigated surface is the hexagonal (1120) or (1102) plane, respectively. Wurtzite-structure ZnO and ZnO-based thin ﬁlms grown on c-plane and a-plane sapphire adopt c-plane orientation, while those grown on r-plane sapphire adopt a-plane orientation [71, 101–104]. (See Sect. 4.2.1 for detailed discussion of ZnO surfaces and see Sect. 1.3, Table 1.1 for further structural parameters).

3 Optical Properties of ZnO and Related Compounds

83

3.2.1.2 Rocksalt Crystal Structure (Cubic) The rocksalt crystal structure belongs to the cubic system with space group Oh1 (Pm3m). It consists of two face-centered cubic (fcc) sublattices, which are occupied by one atom species each. The two sublattices are shifted along one half of the diagonal of the primitive unit cell against each other. The rocksalt lattice is sixfold coordinated. 3.2.2 Vibrational Properties 3.2.2.1 Wurtzite-Structure Phonons The wurtzite-structure optical phonons at the Γ -point of the Brillouin zone belong to the following irreducible representation [32] Γ opt = 1A1 + 2B1 + 1E1 + 2E2 .

(3.1)

Hereby, the branches with E1 - and E2 -symmetry are twofold degenerated. Both A1 - and E1 -modes are polar, and split into transverse optical (TO) and longitudinal optical (LO) phonons with diﬀerent frequencies ωTO and ωLO , respectively, because of the macroscopic electric ﬁelds associated with the LO phonons. The short-range interatomic forces cause anisotropy, and A1 - and E1 -modes possess, therefore, diﬀerent frequencies. The electrostatic forces dominate the anisotropy in the short-range forces in ZnO, such that the TO-LO splitting is larger than the A1 -E1 splitting. For the lattice vibrations with A1 - and E1 -symmetry, the atoms move parallel and perpendicular to the c-axis, respectively (Fig. 3.2). Both A1 - and E1 -modes are Raman and IR active. The two nonpolar E2 (1) (2) modes E2 and E2 are Raman active only. The B1 -modes are IR and Raman inactive (silent modes). Phonon dispersion curves of wurtzite-structure and rocksalt-structure ZnO throughout the Brillouin Zone were reported in [106–108]. For crystals with wurtzite crystal structure, pure longitudinal or

[0001]

B1(1)

B1(2)

E2(1)

E2(2)

A1

E1

[2110]

Fig. 3.2. Displacement patterns of the optical phonons of a lattice with wurtzite crystal structure. Reprinted with permission from [105]

84

C. Bundesmann et al. z

y

y x

z(xx)z’ z(xy)z’

z x

y x

x(yy)x’ x(yz)x’

z

x(zz)x’ x(zy)x’

Fig. 3.3. Raman scattering conﬁgurations for a thin ﬁlm on a thick substrate in the backscattering regime. The laboratory coordinate system is chosen such that the z-axis is parallel to the sample normal. The direction and polarization of the incident laser beam are indicated by the vertical arrow and the horizontal double arrow, respectively. By aligning a polarizer within the light path of the scattered light parallel or perpendicular with respect to the polarization direction of the incident laser beam, the two scattering conﬁgurations indicated below the drawings can be realized. The notation of the scattering conﬁgurations follows the “Porto notation” [32]: The letters before and after the parenthesis show the direction of the incident and scattered light, respectively, while the letters inside the parenthesis indicate the corresponding polarization

pure transverse phonons of well-deﬁned symmetry can be observed only if the phonon propagation is along or perpendicular to the c-axis. Group theory in combination with polarization and propagation considerations allows to identify the symmetry of the Raman active optical modes by applying diﬀerent scattering conﬁgurations [33]. In Fig. 3.3 possible scattering conﬁgurations for a thin ﬁlm on a substrate are drawn schematically. Table 3.2 summarizes the optical phonons of crystals with wurtzite structure and the scattering conﬁgurations, in which the optical phonons are predicted to produce backward signal in ﬁrst-order Raman scattering. 3.2.2.2 Rocksalt-Structure Phonons The Γ -point optical phonons of a crystal with rocksalt structure belong to the irreducible representation Γ opt = F1u .

(3.2)

The F1u -mode is polar and splits into TO and LO modes. The F1u -mode is IR active and Raman inactive [109].

3 Optical Properties of ZnO and Related Compounds

85

Table 3.2. Raman selection rules of the optical phonon modes of crystals with wurtzite structurea [33] Scattering conﬁguration

Allowed Raman modes

Corresponding Raman tensor elements

x(zz)x A1 (TO) αzz x(zy)x , x(yz)x E1 (TO) αzy = αyz x(yy)x E2 , A1 (TO) αyy = ±αxx z(xx)z E2 , A1 (LO) αxx = ±αyy z(xy)z E2 αxy = αyx a It is assumed that the optical c-axis is parallel the z-direction of the laboratory system

3.2.3 Infrared Model Dielectric Function: Phonons and Plasmons In the IR spectral region, DFs εi (ω) are sensitive to phonon and plasmon contributions. Hence, IR model dielectric functions (MDFs) are written as (ω) contributions [73] the sum of lattice εLi (ω) and free-charge-carrier εFCC i (ω). εi (ω) = εLi (ω) + εFCC i

(3.3)

The subscript i refers to the two polarization states parallel (i = ||) and perpendicular (i = ⊥) to the c-axis, which have to be distinguished for optically uniaxial samples, for instance wurtzite-structure ZnO or sapphire. Cubic crystals, for instance rocksalt-structure Mgx Zn1−x O, are optically isotropic and have only one DF, because the dielectric tensor is reduced to a scalar. 3.2.3.1 Lattice Contributions (Phonons) A common way to describe εL i (ω) with l lattice modes is the factorized form with Lorentzian broadening εLi (ω) = ε∞,i

l 2 ωLO,ij − ω 2 − iγLO,ij ω . ω2 − ω 2 − iγTO,ij ω j=1 TO,ij

(3.4)

The polar lattice modes split into TO- (ωTO,ij ) and LO-modes (ωLO,ij ), with broadening parameters γTO,ij and γLO,ij , respectively [73]. The parameters ε∞,i denote the high-frequency limits in this model approach, which are related to the static dielectric constants ε0,i by the Lydanne–Sachs–Teller relation [110] (Sect. 3.3) ε0,i = ε∞,i

l 2 ωLO,ij . ω2 j=1 TO,ij

(3.5)

86

C. Bundesmann et al.

3.2.3.2 Free-Charge-Carrier Contributions (Plasmons) Free-charge-carrier contributions εFCC (ω) are typically described by the clasi sical Drude approximation εFCC (ω) = − i

2 ωp,i , ω(ω + iγp,i )

(3.6)

where ωp,i and γp,i are the screened plasma frequencies and their broadening parameters, respectively. ωp,i can be related to the free-charge-carrier concentration N and the free-charge-carrier eﬀective mass m∗i 2 ωp,i =

N e2 , ε∞,i 0 m∗i

(3.7)

where 0 and e denote the vacuum permittivity and the elementary charge, respectively. Assuming a constant-carrier-scattering regime, the plasma broadening parameters γp,i are equal to the inverse of the energy-averaged carrier-momentum relaxation time τm i , and can be rewritten with the opti∗ cal carrier mobility µopt i , e, and mi [111] γp,i ≡

1 e = ∗ opt . τm i mi µi

(3.8)

3.2.4 Visible-to-Vacuum-Ultraviolet Model Dielectric Function: Band-to-Band Transitions and their Critical-Point Structures The DF in the visible to deep-ultraviolet (DUV) spectral region is dominated by critical point (CP) structures, which are related to electronic band-to-band transitions. For ZnO, the CP structures due to band-to-band transitions are superimposed by excitonic polarizabilities rendering the fundamental absorption edge. Within the electron-band density-of-states function, Van Hove singularities in one, two, and three dimensions occur at CPs of the type M0 –M3 [112]. Close to energies of such singularities, band-to-band transitions occur, which give rise to the CP structures in ε. The lowest bandto-band transitions of ZnO occur at the Γ -point of the Brillouin zone, and the associated CP structures are typically of the 3DM0 type (Sect. 3.7). At higher energies CP-structures occur, which are often described as 3DM1 -, or, equivalently, as 2DM0 - or M2 -type singularities. Diﬀerent MDF approaches exist for the description of the photon energy dependence of characteristic CP structures. In the spectral region below the electronic band-to-band transitions, the Cauchy approximation (transparency region, (3.24)) or the damped harmonic oscillator function (both transparency as well as absorption region) are often utilized as MDF approaches.

3 Optical Properties of ZnO and Related Compounds

87

General parametric MDFs, which describe CPs and the line shape of the DF of semiconductors, are more complex. The complete MDF is the sum of multiple terms, including contributions due to 3DM0 CP structures (ε3DM0 ), 2DM0 CP structures (ε2DM0 ), adjacent broadened CP structures (εL ), and excitonic contributions (εdex and εcex ). For ZnO, typically, Adachi’s composite MDF is applied [76,113]. In this approach, the contributions due to 3DM0 CP structures read 1/2 1/2 − (1 − χα 2 − (1 + χα 0) 0) 3DM0 α α −3/2 , (3.9) ε (E) = A0 (E0 ) 2 (χα 0) α with α χα 0 = (E + iΓ0 )/E0 .

Aα 0

(3.10)

E0α

and are the amplitude and the transition energy of the CP structures, respectively and Γ0 is the broadening parameter. E is the photon energy. For wurtzite-structure and rocksalt-structure ZnO α covers α = A, B, C and α = D, E, respectively. The contributions due to 2DM0 CPs are described by −2 2 ε2DM0 (E) = −Aβ χβ ln(1 − χβ ) (3.11) β

with χβ = (E + iΓ β )/E β .

(3.12)

Aβ , E β , and Γ β are amplitude, transition energy, and broadening parameter, respectively. Lorentzian-broadened harmonic oscillators may be used as a good approximation for the description of adjacent (spectrally unresolvable) broadened CP structures εL (E) =

Aβ Γ β E β

β

(E β ) − E 2 − iΓ β E

2

,

(3.13)

with averaged parameters Aβ , E β , and Γ β . The discrete contributions (n-series) due to free excitons is approximated by [76, 113] εdex (E) =

∞ α

An,α xb . α − E n ) − E − iΓ (E xb 0 xb n=1

(3.14)

n An,α xb and Γxb are amplitude and broadening parameter, respectively. Exb is the free exciton binding energy of the nth excited state n Exb =

R03D . n2

(3.15)

R03D is the 3D exciton Rydberg energy. At RT it is suﬃcient to consider n = 1 and, in some cases, n = 2. At low temperatures n = 1, 2, 3 might be of

88

C. Bundesmann et al.

interest. The exciton continuum contributions to the DF can be described by εcex (E) =

α

α Aα xc Exc

4R03D (E + iΓxc )

2

ln

(E0α )2 2

(E0α ) − (E + iΓxc )

2

,

(3.16)

with the continuum-exciton strength and broadening parameters Aα xc and α α Γxc , respectively. The ground-state continuum-exciton energy Exc is typically approximated by E0α . 3.2.5 Spectroscopic Ellipsometry The main experimental technique applied in this chapter is SE. Several textbooks were written on SE [73, 114–118]. Therefore, only some basic concepts are described. SE examines the relative phase change of a polarized light beam upon reﬂection (or transmission) at a sample surface. In Fig. 3.4 the setup of an ellipsometry experiment is shown. Upon model analysis of the experimental data, the DFs and thicknesses of the sample constituents can be extracted. Two diﬀerent experimental approaches have to be distinguished, standard and generalized ellipsometry. 3.2.5.1 Standard Spectroscopic Ellipsometry When neither s-polarized light (light polarized perpendicular to the plane of incidence) is converted into p-polarized light (light polarized parallel to the plane of incidence) nor vice versa, standard SE is applied. This is the case for isotropic samples and for uniaxial samples in the special case, where the optical axis is parallel to the sample normal, for example (0001) ZnO [119]. E

Ap

As plan

e of

incid

ence

Φa

Bp

E

Bs

Fig. 3.4. The geometry of an ellipsometry experiment. The linearly polarized incidence light beam becomes elliptically polarized after reﬂection at the sample surface. The plane of incidence is shown hatched. Φa is the angle of incidence

3 Optical Properties of ZnO and Related Compounds

89

Standard SE determines the complex ratio ρ of the reﬂection coeﬃcients for p-polarized and s-polarized light

Bs Bp = tan Ψ exp(i∆). (3.17) ρ= Ap As Ai and Bi denote the intensities of the incident and reﬂected light beam, respectively. Ψ and ∆ are the ellipsometric parameters, where tan Ψ is the magnitude and ∆ the phase of ρ [116]. 3.2.5.2 Generalized Spectroscopic Ellipsometry In the general case, when s-polarized light is converted into p-polarized light and/or vice versa, the standard SE approach is not adequate, because the oﬀ-diagonal elements of the reﬂection matrix r in the Jones matrix formalism are nonzero [114]. Generalized SE must be applied, for instance, to wurtzitestructure ZnO thin ﬁlms, for which the c-axis is not parallel to the sample normal, i.e., (1120) ZnO thin ﬁlms on (1102) sapphire [43, 71]. Choosing a Cartesian coordinate system relative to the incident (Ai ) and reﬂected plane waves (Bi ), as shown in Fig. 3.4, the change of polarization upon reﬂection can be described by [117, 120]

Bp Ap rpp rsp Ap =r = . (3.18) Bs As rps rss As The generalized ellipsometric parameters Ψij and ∆ij are deﬁned by rpp ≡ Rpp = tan Ψpp exp(i∆pp ) , rss rps ≡ Rps = tan Ψps exp(i∆ps ) , rpp rsp ≡ Rsp = tan Ψsp exp(i∆sp ) . rss

(3.19) (3.20) (3.21)

3.2.5.3 Data Analysis Only in the special case of isotropic bulk samples, the experimental data can be transformed directly into the sample’s DF. In all other cases a model analysis is required, where the layered sample structure has to be considered appropriately, i.e., the DFs, thickness, and, in case of anisotropic materials, the crystal orientation of each single layer. Unknown parameters are varied until experimental and model data match as close as possible. Upon the use of parameterized MDFs, physically relevant parameters of the samples can be obtained. Typical MDFs for ZnO are described in Sects. 3.2.3 and 3.2.4.

90

C. Bundesmann et al.

3.3 Dielectric Constants and Dielectric Functions Figure 3.5 summarizes typical DF spectra ε = ε1 + iε2 of ZnO. The DF spectra were obtained by SE in the Mid-IR (MIR), Near-IR (NIR), visible (VIS), ultraviolet (UV), and vacuum-ultraviolet (VUV) spectral region from both single-crystal bulk and thin ﬁlm samples. DF spectra of ZnO in the NIR– VIS–UV spectral region were also reported in [76, 78, 80–82, 121, 122], and for Mgx Zn1−x O in [82, 90]. Phonon-mode contributions (Sect. 3.4), plasmonmode (free-charge-carrier) contributions (Sect. 3.5), refractive indices (Sect. 3.6), critical-point and electronic band-to-band transition parameters as well as their excitonic contributions (Sect. 3.7) can be obtained from the DF spectra, including their anisotropy. While the DFs of ZnO in the IR exhibit appreciable anisotropy, in the below-band-gap spectral region and above the band gap, ε⊥ and ε show only small diﬀerences. ZnO is uniaxial positive, i.e., Re{ε||} > Re{ε⊥ }, at frequencies below and above the reststrahlen band [123], and throughout the entire band-gap region [77, 124]. The static dielectric constants ε0,i can be determined from the analysis of IR spectra with suﬃcient spectral data coverage below and above the reststrahlen regions. For photon energies far above the phonon resonances but still suﬃciently below the electronic band-to-band transitions, the DFs converge to the “high-frequency” dielectric constants ε∞,i , which is related to ε0,i by the Lyddane–Sachs–Teller relation (3.5). ε∞,i measure the sum of 300 400 500 600 700

ω [cm–1] 20000

40000

60000 4

ZnO

150

E1(TO)

2

50

1 0 8

0 A1(LO)

E ||c

E1(LO)

50 ε1

ε2

A1(TO)

E1 E⊥c

7 E2 E3E4 E5 E6E7 6 5

0

4

E0 α-Exb 1,α

-50

3

E0α 0.04

0.06

0.08

1

2

3

ε1

ε2

3 100

4

5

6

7

8

9

2

Photon energy [eV]

Fig. 3.5. Real (ε1 ) and imaginary part (ε2 ) of the DFs ε = ε1 + iε2 of ZnO at RT for polarization perpendicular (solid lines) and parallel (dashed lines) to the c-axis determined by SE [15, 38, 71, 91]. Optical phonon modes (Table 3.4; Sect. 3.4) and transition energies (Tables 3.9, 3.10; Sect. 3.7) are marked. Note the diﬀerent scales for the IR-to-NIR and the VIS-to-VUV spectral regions

3 Optical Properties of ZnO and Related Compounds

91

all linear electronic polarizabilities for all photon energies within and above the fundamental band-to-band transition energy until the shortest end of the electromagnetic spectrum. Table 3.3 summarizes “high-frequency” and static dielectric constants of ZnO. The dielectric constants for the ZnO thin ﬁlms are smaller than those of the ZnO bulk samples, which can be explained by the lower ﬁlm density relative to that of the bulk samples, or because of slightly diﬀerent compositions. Figure 3.6 summarizes “high-frequency” and static dielectric constants of PLD-grown Mgx Zn1−x O thin ﬁlms [43, 62, 72, 74]. Besides the natural Table 3.3. “High-frequency” ε∞,i and static ε0,i dielectric constants of ZnO bulk samples (b) and ZnO thin ﬁlms (f) Ref.

Method

Bond (1965) Yoshikawa (1997) Teng (2000) Ashkenov (2003) Ashkenov (2003) Bundesmann (2004) Bundesmann (2006) Bundesmann (2006) Bundesmann (2006)

[124] [76] [95] [38] [38] [71] [43] [43] [43]

ε∞,⊥ ε∞,|| ε0,⊥ ε0,||

NIR-VIS minimum deviation NIR-VIS SE NIR-VIS prism coupling IRSE IRSE IRSE IRSE IRSE IRSE

(b) (b) (f)b (b) (f)b (f)c (f)d (f)e (f)f

3.70 3.75 3.68 3.72 3.60 3.66 3.70 3.78 3.61 3.76 3.53 3.60 3.67† 3.38† 2.99†

7.78 8.74 7.74 8.67 7.44 8.45 7.78 8.81 7.46 8.69 7.29 8.37 7.96† 6.82† 6.03†

a

ε∞,i follows from ε0,i with the Lydanne–Sachs–Teller-relation (3.5) and the phonon-mode frequencies in Table 3.4 b (0001) ZnO ﬁlm (PLD) on (0001) sapphire c (1120) ZnO ﬁlm (PLD) on (1102) sapphire d (0001) ZnO ﬁlm (PLD) on (001) silicon e ZnO ﬁlm (magnetron sputtering) on metallized foil f ZnO ﬁlm (magnetron sputtering) on metallized glass † Isotropically averaged 4.0

(a)

MgxZn1-xO

12

3.5

wurtzitestructure

ε0

ε∞

10 8

3.0 rocksaltstructure

wurtzitestructure 2.5

(b)

MgxZn1-xO

0.0

0.2

0.4

0.6 x

0.8

rocksaltstructure

6 1.0

0.0

0.2

0.4

0.6

0.8

1.0

x

Fig. 3.6. “High-frequency” dielectric constants (a) and static dielectric constants (b) of wurtzite-structure (E||c: up-triangles, E⊥c: down-triangles) and rocksaltstructure (circles) Mgx Zn1−x O thin ﬁlms [43, 62, 72, 74]. The shaded area indicates the composition range of the phase transition. Reprinted with permission from [74]

92

C. Bundesmann et al.

disappearance of the anisotropy both “high-frequency” and static dielectric constants change abruptly upon the phase transition from wurtzite to rocksalt structure because of the coordination number change (4–6) and the associated change of bond polarizability (increase in splitting between TO and LO phonon-mode frequencies, Fig. 3.13, Sect. 3.4.3), and critical-point characteristics of the two polytypes (Sect. 3.7.2).

3.4 Phonons 3.4.1 Undoped ZnO Table 3.4 summarizes the phonon-mode frequencies of ZnO bulk samples and ZnO thin ﬁlms, as obtained by Raman scattering spectroscopy, IR-reﬂection measurements, and IRSE. Figure 3.7 presents typical Raman spectra of a ZnO bulk sample and a ZnO thin ﬁlm on sapphire. According to the theoretical considerations in Sect. 3.2.2, the scattering intensities at ω ∼ 379 cm−1 , ∼410 cm−1 , ∼437 cm−1 , and ∼572 cm−1 can be assigned to the A1 (TO)-, E1 (TO)-, (2) E2 -, and A1 (LO)-mode, respectively. The scattering cross-section of the A1 (LO)-mode is markedly smaller than that of the A1 (TO)-mode, which was explained by the destructive interference of the Fr¨ ohlich interaction and the deformation potential contributions to the LO-scattering in ZnO [31]. The intensities at ω∼332, ∼538, and ∼665 cm−1 were assigned to multiple-phonon scattering processes [32, 35], whereas the spectral feature at ω ∼ 589 cm−1 was assigned to the E1 (LO)-mode. The phonon modes of the ZnO thin ﬁlm occur at similar frequencies as those of the ZnO bulk sample, but the Raman spectra contain additional spectral features because of the sapphire substrate [125, 126]. In Fig. 3.8, typical IRSE spectra of a ZnO bulk sample and a ZnO ﬁlm on sapphire are plotted. In the Ψ -spectrum of the ZnO bulk sample a plateau with Ψ ∼ 45◦ can be seen, which corresponds to the bands of total reﬂection (reststrahlen bands), which occurs between the E1 (TO)- and E1 (LO)-mode frequencies [123]. The small dip within the plateau is caused by the loss in p-reﬂectivity, and localizes the A1 (LO)- and E1 (LO)-mode frequencies. The derivative-like structure in the Ψ -spectrum of the bulk ZnO sample at ω ∼ 650 cm−1 is caused by the anisotropy Re{ε|| } > Re{ε⊥ } (Sect. 3.3) [38]. In the Ψ -spectrum of the ZnO thin ﬁlm, a similar plateau as in the Ψ spectrum of the ZnO bulk sample is present. However, the phonon modes of the sapphire substrate introduce additional features, for example at ω ∼ 510, ∼630, and ∼900 cm−1 [38,123]. The spectral feature at ω ∼ 610 cm−1 is called the Berreman resonance, which is related to the excitation of surface polaritons of transverse magnetic character at the boundary of two media [73]. In the spectral region of the Berreman resonance, IRSE provides high sensitivity to the A1 (LO)-mode parameters. For (0001)-oriented surfaces of crystals with wurtzite structure, linear-polarization-dependent spectroscopic

3 Optical Properties of ZnO and Related Compounds

93

Table 3.4. Frequencies of long-wavelength optical phonon modes of ZnO bulk samples (b) and ZnO thin ﬁlms (f)a Ref.

Method

Collins (1959) [63] IR-reﬂ. Heltemes (1967) [64] IR-reﬂ. Venger (1995) [65] IR-reﬂ. Damen (1966) [32] Raman Arguello (1969) [33] Raman Scott (1970) [34] Raman Callender (1973) [31] Raman Calleja (1977) [35] Raman Bairamov (1983) [36] Raman Decremps (2001) [37] Raman Ashkenov (2003) [38] Raman Lu (2000) [39] Raman Zeng (2002) [40] Raman Ashkenov (2003) [38] Raman Ye (2003) [41] Raman Bundesmann (2006) [43] Raman Ashkenov (2003) [38] IRSE Ashkenov (2003) [38] IRSE Bundesmann (2004) [71] IRSE Heitsch (2005) [30] IRSE Bundesmann (2006) [43] IRSE Bundesmann (2006) [43] IRSE Decremps (2001) [37] Theory

(1)

(2)

E2 A1 (TO) E1 (TO) E2 (b) – (b) – (b) – (b) 101 (b) 101 (b) – (b) – (b) 98 (b) 98 (b) 99 (b) – (f)b,c – – (f)c (f)b – (f)d – – (f)c (b) – – (f)b (f)e – – (f)c (f)f – (f)g – 92

– 414 377 406 380 412 380 407 380 413 – – 381 407 378.5 410 378 409.5 382 414 378 409 – – – – 378 410 – – 376 410 – 408.2 – 409.1 376.5 410.7 – 406.0 406.2† 409.9† 397 426

A1 (LO) E1 (LO)

– – 591 – 575 589 – 570 591 437 – 583 444 579 591 – 574 583 441 – 583 438 – 590 437.5 576 588 439 574 580 437 572 588 438 – 579 437 – 579 437 – 588 436 563 – 437 – – – 577.1 592.1 – 574.5 588.3 – 574.1 590.4 – 571.3 – – 577.0† – 581.9† 449 559 577

All values are given in units of cm−1 (0001) ZnO ﬁlm (PLD) on (0001) sapphire c (0001) ZnO ﬁlm (PLD) on (001) silicon d (0001) ZnO ﬁlm (MOCVD) on (111) silicon e (1120) ZnO ﬁlm (PLD) on (1102) sapphire f ZnO ﬁlm (magnetron sputtering) on metallized foil g ZnO ﬁlm (magnetron sputtering) on metallized glass † Isotropically averaged a

b

techniques, as for instance IRSE, are not sensitive to the A1 (TO)-mode, and are only slightly sensitive to the E1 (LO)-mode. This limitation can be overcome by application of generalized IRSE to optically uniaxial thin ﬁlms, which are not (0001)-oriented [43, 71, 73]. Figure 3.9 shows the generalized IRSE data (Ψij only) of a (1120) ZnO thin ﬁlm on (1102) sapphire, which allows to extract simultaneously the complete DF spectra for ε|| and ε⊥ (Fig. 3.5), the crystal orientation for both the ZnO thin ﬁlm and the sapphire substrate, and the thin ﬁlm thickness. Typical IRSE spectra of a PLD-grown ZnO thin ﬁlm on silicon and two magnetron-sputtered ZnO thin ﬁlms on metallized foil or glass are shown in Fig. 3.10. The corresponding

94

C. Bundesmann et al. (a) MP

ZnO

A1(TO) A1(LO) E1(TO) E2(2) MP E1(LO) MP

(b)

Raman intensity [arb.u.]

Raman intensity [arb.u.]

x(yz)x'

x(yy)x'

600

700

800

MP

S

S

x(yy)x'

z(xy)z'

500

E1(LO)

S S

z(xy)z'

400

MP

S

z(xx)z'

300

E2(2)

x(yz)x'

z(xx)z'

200

E1(TO)

x(zz)x'

x(zz)x'

ZnO / α-Al2O3

A1(TO) MP

200

300

400

ω[cm–1]

500

600

700

800

ω[cm–1]

180

(0001)ZnO/(0001) α-Al2O3

(b)

45 30 Φa = 70˚

15

E1(LO)

E1(TO)

15

Ψ [˚]

(a) 30

A1(LO)

Ψ [˚]

60

A1(LO) E1(LO)

75

(0001) ZnO

45

E1(TO)

Fig. 3.7. Polarized micro-Raman spectra of a (0001) ZnO bulk sample (a) and a (0001) ZnO thin ﬁlm (d ∼ 1, 970 nm) on (0001) sapphire (b). The vertical dotted and dashed lines mark ZnO and sapphire (S) phonon modes, respectively. MP denotes modes due to multi-phonon scattering processes in ZnO. Excitation with Ar+ -laser line λ = 514.5 nm and laser power P ≤ 40 mW. Reprinted with permission from [38]

Φa = 70˚

75

A2u

60

Eu

Ψ [˚]

∆ [˚]

120 60

α-Al2O3

45 30 15

0 400

600

800

ω [cm−1]

1000

1200

400

600

800

1000

1200

ω [cm−1]

Fig. 3.8. (a) Experimental (dotted lines) and best-ﬁt model (solid lines) IRSE spectra of a (0001) ZnO bulk sample. The ZnO phonon modes, as obtained by IRSE, are marked by vertical arrows. (b) Experimental (dotted lines) and best-ﬁt model (solid lines) IRSE spectra (Ψ only) of a (0001) ZnO thin ﬁlm on (0001) sapphire (upper panel, thickness d ∼ 1 970 nm) and of a bare (0001) sapphire substrate (lower panel ). The ZnO phonon modes, as obtained by IRSE, are marked by arrows. The IR-active modes of sapphire are indicated by solid (TO) and dotted (LO) vertical markers. Reprinted with permission from [38]

A1(LO) E1(LO)

A1(TO) E1(TO)

3 Optical Properties of ZnO and Related Compounds

95

(1120) ZnO/(1102) α-Al2O3

Ψpp [˚]

(a) (b) 30

(c)

0

Ψps [˚]

(d) (e) 15 (f)

Ψsp [˚]

0

(g) (h) 10 (i) 0 400

600

800 ω [cm–1]

1000

1200

Fig. 3.9. Experimental (dotted lines) and best-model (solid lines) generalized IRSE spectra Ψij of a (1120) ZnO thin ﬁlm (d = 1455 ± 5 nm) on (1102) sapphire [43,71]. Spectra are shifted for clarity. Vertical dashed lines indicate the ZnO phonon mode frequencies. Spectra in (a,d,g), (b,e,h), and (c,f,i) belong to diﬀerent sample azimuth angles, respectively. The best-model values of the Euler angle θZnO , θSapphire , which describe the c-axis inclination with respect to the sample normal are 89.0◦ ± 1.0◦ and 54.9◦ ± 0.8◦ , respectively. Reprinted with permission from [71]

phonon-mode frequencies are listed in Table 3.4. Because of the submicrometersized randomly-oriented polycrystalline structure of the magnetronsputtered ZnO thin ﬁlms, the optical response is isotropic, and the MDF is that for an isotropic ﬁlm providing phonon-mode parameters intermediate to those of A1 - and E1 -modes of a single-crystalline ZnO thin ﬁlm. (2) Temperature-dependent Raman data were reported for the E2 -mode of ﬂux-grown ZnO platelets in the temperature range from T ∼ 15 K to (1) (2) T ∼ 1 050 K [127], and for the E2 -mode, the E2 -mode, and the MP-mode at ω ∼ 332 cm−1 of a ZnO bulk sample in the temperature range from T ∼ 300 K to T ∼ 700 K [43] In Fig. 3.11 the unpolarized Raman spectra and the temperature-dependence of the phonon-mode frequencies from [43] are

96

C. Bundesmann et al. 45 ZnO (0001) / Si (001)

Φa = 70˚

A1(LO)

15

E1(TO)

Ψ [˚]

30

(a)

0 45

32 ZnO / Mo / Glass

ZnO / Mo /PI

30

TO

Φa = 70˚

15 LO 0

400

600

(b) 800

1000

1200

ω [cm–1]

Ψ [˚]

Ψ [˚]

30

LO

28

Φa = 70˚

26 TO 24

(c) 22

400

600

800

1000

1200

ω [cm–1]

Fig. 3.10. Experimental (dotted lines) and best-ﬁt model (solid lines) IRSE spectra of a PLD-grown (0001) ZnO thin ﬁlm on (001) silicon (panel (a), ﬁlm thickness d ∼ 670 nm), and magnetron-sputtered ZnO thin ﬁlms on metallized polyimide foil (panel (b), d ∼ 500 nm) and on metallized glass (panel (c), d ∼ 30 nm) [43]. ZnO phonon-mode frequencies, as obtained by best-model analysis, are marked by vertical arrows

plotted. The temperature-dependence can be modeled by the empirical Bose– Einstein equation [128, 129] ω(T ) = ω(0) −

A . exp[Bω(0)/kB T ] − 1

(3.22)

A and B are model parameters representing the high-temperature linear slope (∂ω/∂T |T →∞) and eﬀective phonon-mode temperature (Bω(0)/kB ), respectively. ω(0) is the phonon-mode frequency at T = 0 K. Table 3.5 summarizes the best-model parameters reported in [43]. In [127], a linear temperature(2) dependence with ∂ω[E2 ]/∂T = −1.85 × 10−2 cm−1 K−1 was reported for temperatures above RT. The pressure-dependence of ZnO phonon-mode frequencies measured by Raman scattering was reported in [37, 130]. From that the Gr¨ uneisen parameters γj ∂ ln ωj (3.23) γj = − ∂lnV of the vibrational modes can be obtained. Table 3.6 summarizes the Gr¨ uneisen parameters reported in [37].

3 Optical Properties of ZnO and Related Compounds E2(1)

E2(2)

ZnO bulk

MP

(b)

436

MP

434

T = 690 K

Raman intensity [arb.u.]

ZnO bulk

438

(a)

97

E2

432

ω [cm–1]

T = 591 K

T = 490 K

(2)

332 330 328

MP

326 104

T = 389 K

102

E2(1)

100 98

T = 293 K

100 200 300 400 500 600 700

300

400

ω [cm–1]

500

600

700

T [K]

Fig. 3.11. (a) Temperature-dependent, unpolarized Raman spectra of a (0001) ZnO bulk sample [43]. Spectra are shifted for clarity. (b) Phonon-mode frequencies vs. temperature as determined from the Raman data in Fig. 3.11a. The solid lines are model approximations according to (3.22). Excitation with Nd:YAGlaser line λ = 532 nm and laser power P ∼ 60 mW Table 3.5. Best-model parameters of the temperature-dependent phonon mode shift in Fig. 3.11b, calculated by (3.22) (Bose–Einstein model) [43]

Mode (1)

E2 MP (2) E2

ω(0) (cm−1 )

A (cm−1 )

Bω(0)/kB (K)

∂ω/∂T |T →∞ (cm−1 K−1 )

101.6 ± 0.1 334.1 ± 0.4 439.2 ± 0.3

6.7 ± 1.7 29.3 ± 6.6 48 ± 9

1 000 ± 150 1 020 ± 140 1 290 ± 120

−0.007 ± 0.002 −0.03 ± 0.01 −0.037 ± 0.01

Table 3.6. Gr¨ uneisen parameters of the zone-center phonon modes of ZnO [37] Mode j

Experiment γj

Theory γj

(1)

–1.6 2.1 1.8 2.0 1.4

–1.67 1.70 1.80 1.84 1.30

E2 A1 (TO) E1 (TO) (2) E2 E1 (LO)

98

C. Bundesmann et al.

3.4.2 Doped ZnO Figure 3.12 shows typical Raman spectra of several doped ZnO thin ﬁlms. Additional modes (AM), occurring at ω ∼ 275, ∼ 510, ∼ 582, ∼ 643, and ∼ 856 cm−1 (the ﬁrst four of them are shown and marked by vertical solid lines in Fig. 3.12), were ﬁrst assigned to N-incorporation [49–51], because the intensity of these modes was reported to increase with increasing N-content [50]. However, the AMs appear also in Raman spectra of ZnO samples doped with other elements (Fig. 3.12a), [48,52,53]). Therefore, it was suggested that the AMs are related to defect-induced modes [48]. Theoretical considerations conﬁrmed this assignment [131]. It was discussed that the AMs could be related to modes of ZnO, which are Raman-inactive within a perfect crystal. Upon doping-induced defect formation, the translational crystal symmetry can be broken, and Raman-inactive modes may become Raman-active. The Raman spectra of the ZnO thin ﬁlms with transition metals in Fig. 3.12b show a diﬀerent behavior than those in Fig. 3.12a [43, 48]. Raman spectra of Fe0.08 Zn0.92 O contain the above described AMs, but with diﬀerent intensity ratios. For MnZnO, CoZnO, and NiZnO a broad band between ω ∼ 500 cm−1

Raman intensity [arb.u.]

Fe0.08Zn0.92O

(b)

*

x(yy)x'

x(yy)x'

ZnO:Sb

* ZnO:Ga ZnO:P

ZnO:Al

Raman intensity [arb.u.]

(a)

Mn0.03Zn0.97O

* Co0.17Zn0.83O

Ni0.02Zn0.98O

Cu0.01Zn0.99O

ZnO:N

* 300

400

*

500

ω [cm–1]

ZnO:Li 600

700

ZnO

300

400

500

600

700

ω [cm–1]

Fig. 3.12. Micro-Raman spectra in the x(yy)x scattering conﬁguration of several PLD-grown, element-doped wurtzite-structure (0001) ZnO thin ﬁlms on (0001) sapphire [43,48]. Defect-induced modes are marked by solid vertical lines. The asterisks indicate modes, which seem to occur for speciﬁc dopant species only. Excitation with Ar+ -laser line λ = 514.5 nm and laser power P ≤ 40 mW

3 Optical Properties of ZnO and Related Compounds

99

and ω ∼ 600 cm−1 occurred [43, 58], which was also assigned to defects. Additionally, some AMs appear to be characteristic for their dopant species (Fig. 3.12). 3.4.3 Mgx Zn1−x O The phonon mode frequencies of wurtzite- and rocksalt-structure Mgx Zn1−x O thin ﬁlms vs. x, as obtained by combination of Raman scattering and IRSE, are plotted for 0 ≤ x ≤ 1 in Fig. 3.13. For the wurtzite-structure Mgx Zn1−x O (x ≤ 0.53) thin ﬁlms, an one-mode behavior with a further weak mode between the TO- and LO-mode for the phonons with E1 - and A1 -symmetry was found. The A1 (TO)-, A1 (LO)-, and the upper branch of the E1 (LO)-modes of the wurtzite-structure thin ﬁlms show an almost linear behavior, whereas the lower branch of the E1 (LO)modes and the two E1 (TO)-branches exhibit a nonlinear behavior. In [132] the modiﬁed random element isodisplacement (MREI) model was suggested

700

MgxZn1-xO

A1

E||c

F1u

ω [cm–1]

600 wurtzitestructure

500

rocksaltstructure

400

(a)

300 700

E1 E⊥c

F1u

ω [cm–1]

600 500 400

(b)

300 0.0

0.2

0.4

0.6

0.8

1.0

x

Fig. 3.13. Phonon-mode frequencies of wurtzite-structure PLD-grown Mgx Zn1−x O thin ﬁlms with A1 -symmetry (panel a, triangles) and E1 -symmetry (panel b, triangles), and of rocksalt-structure PLD-grown Mgx Zn1−x O thin ﬁlms (circles in both panels) vs. x [43, 62, 72, 74]. Open and solid symbols represent TO- and LOmodes, respectively. The dashed lines are linear approximations of the rocksaltstructure phonon modes from [74], the solid lines represent MREI calculations for the wurtzite-structure phonon modes redrawn from [132]. The shaded area marks the composition range, where the phase transition occurs. Reprinted with permission from [74]

100

C. Bundesmann et al.

to describe the phonon-mode behavior vs. x. A good agreement for the E1 (TO)-, A1 (TO)-, and A1 (LO)-branches was obtained. The AM of the upper TO-branch with E1 -symmetry was assigned to the mixed mode of the Mgx Zn1−x O alloy, which originates from the local mode ωloc,ZnO:Mg of Mg in ZnO. The extrapolation to x = 0 yields an experimental value of ωloc,ZnO:Mg = 509 cm−1 . This value agrees well with the calculated local mode ωloc,ZnO:Mg = 518 cm−1 [43, 74]. For rocksalt-structure Mgx Zn1−x O thin ﬁlms (x ≥ 0.68), an one-mode behavior was found by IRSE, where both the TO- and LO-modes shift linearly with x. The phonon-mode frequencies of the MgO thin ﬁlm agree well with values of MgO single crystals (ωTO = 401 cm−1 , ωLO = 719 cm−1 [133]). The shift of the TO- and LO-modes can be modeled by a linear compositiondependence ωTO;LO (x) = mTO;LO x+nTO;LO with best-ﬁt coeﬃcients mTO = 97 ± 4 cm−1 , nTO = 300 ± 3 cm−1 , mLO = 157 ± 10 cm−1 , and nLO = 571 ± 9 cm−1 [43]. The extrapolation to x = 0 yield a value of ωTO (0) ∼ 300 cm−1 and ωLO (0) ∼ 570 cm−1 , which should address ωTO and ωLO , respectively, of rocksalt-structure ZnO. No experimental data have been reported for rocksalt-structure ZnO at normal ambient conditions yet. In [134], ab initio calculations for phonon properties of rocksalt-structure ZnO were presented, which used experimental data of rocksalt-structure ZnO studied at high pressures (∼8 GPa) as input parameters. According to these calculations, ωTO and ωLO of rocksalt-structure ZnO were predicted to be 235 and 528 cm−1 , respectively. The values are smaller than those obtained from the IRSE analysis described earlier, but both extrapolations follow the same trend in predicting phonon-mode frequencies, which are smaller than those of wurtzite-structure ZnO. 3.4.4 Phonon Mode Broadening Parameters The phonon mode broadening parameter of the ZnO bulk sample2 in Fig. 3.8a is γ⊥ = 13 ± 1 cm−1 . Similar values were reported for the (0001) ZnO thin ﬁlm on (0001) sapphire in Fig. 3.8b: γ⊥ = 10 ± 1 cm−1 , and the (1120) ZnO thin ﬁlm on a-plane sapphire in Fig. 3.9: γ|| = 15.1 ± 0.2 cm−1 , γ⊥ = 10.7 ± 0.5 cm−1 .3 Accordingly, the crystal quality of these ZnO thin ﬁlms are comparable with that of the ZnO bulk sample, which was conﬁrmed by X-ray diﬀraction [43]. Typical phonon mode broadening parameters for a set of ZnO thin ﬁlms grown on silicon by PLD with varying oxygen partial pressure and/or substrate heater power are shown in Fig. 3.14. Heitsch et al. [30] observed that the 2

3

For (0001)-oriented ZnO ﬁlms, IRSE data are not sensitive to the A1 (TO) mode. Accordingly, the broadening parameter γ|| cannot be determined. For setting up the MDF parameters, it is then often assumed γ|| = γ⊥ , which has no inﬂuence on the validity of γ⊥ [73]. For oﬀ-axis oriented ZnO samples, both γ|| and γ⊥ can be determined [73]

3 Optical Properties of ZnO and Related Compounds

γ⊥ [cm–1]

18

101

ZnO E⊥c

16 14 12 1E-3

0.01 p(O2) [mbar]

0.1

Fig. 3.14. Phonon mode broadening parameters, as determined by IRSE, vs. oxygen partial pressure for a set of PLD-grown ZnO thin ﬁlms on (111) silicon. Triangles and squares represent data of thin ﬁlms grown with substrate heater power of P = 400 W and P = 600 W, respectively. Reprinted with permission from [30] 28 rocksalt-structure MgxZn1-xO

26 γ [cm–1]

24 22 20 18 16 14

0.7

0.8

0.9

1.0

x

Fig. 3.15. Phonon mode broadening parameters of PLD-grown rocksalt-structure Mgx Zn1−x O thin ﬁlms on sapphire. Reprinted with permission from [74]

phonon-mode frequencies vary only slightly, but the phonon mode broadening parameters showed a systematic variation with varying p(O2 ). This variation is assigned to a varying crystal quality, which was conﬁrmed by TEM and PL measurements [30]. The crystal quality is suggested to be best, when the oxygen partial pressure is chosen between p(O2 ) = 0.01 mbar and p(O2 ) = 0.03 mbar. Too much or too few oxygen seems to introduce more defects. The broadening parameters are also inﬂuenced by the growth technique, the substrate material, and the ﬁlm thickness. For example, as can be seen from the IRSE spectra of the ZnO thin ﬁlms in Fig. 3.10a (PLD, silicon substrate, d ∼ 670 nm), b (magnetron sputtering, metallized foil, d ∼ 500 nm), c (magnetron sputtering, metallized glass, d ∼ 30 nm),4 where γ⊥ = 14 ± 1 cm−1 , γ = 23 ± 1 cm−1 , and γ = 32 ± 1 cm−1 was found, respectively, indicating a decreasing crystal quality from (a) to (c). In addition to defects and impurity incorporation, alloy-induced disorder further increases the phonon mode broadening parameters, as discussed 4

Magnetron-sputtered ﬁlms contain randomly oriented crystallites, and the optical response is isotropic with one lattice mode broadening parameter only.

102

C. Bundesmann et al.

in [43,72,74] for rocksalt-structure Mgx Zn1−x O thin ﬁlms (Fig. 3.15). Except for the data point with x = 0.78, a systematic increment of γ with increasing Zn-content is obvious.5

3.5 Plasmons So far, all observed free-charge-carrier contributions to the optical properties of ZnO and related materials are only due to n-type conductivity, because reproducible and suﬃciently high p-type conductivity is still a challenge. Therefore, all data presented here address n-type conductive samples. Figures 3.16 and 3.17 depict typical IRSE spectra of PLD-grown ZnO thin ﬁlms doped with diﬀerent Ga- and Al-contents, respectively, reﬂecting the inﬂuence of the free charge carriers on the IR response of the doped ZnO thin ﬁlms [43]. ZnO:Ga / α-Al2O3 +19

N=3.9*10 d=787nm

+18

Ψ [°]

N=6.3*10 d=954nm

–3

cm

–3

cm

+18 –3 N=1.8*10 cm d=1147nm

60 45 30 15 0

N=8.3*10+17cm–3 d=1305nm

400

600

800 1000 ω [cm–1]

1200

Fig. 3.16. Experimental (dotted lines) and best-model calculated (solid lines) IRSE spectra of PLD-grown Ga-doped ZnO thin ﬁlms on sapphire with diﬀerent free-charge-carrier concentration and thickness parameters as indicated next to the respective graphs [43]. Spectra are shifted for clarity

5

The Mg0.78 Zn0.22 O thin ﬁlm was grown at a higher oxygen partial pressure (p(O2 ) = 0.16 mbar) than all other ﬁlms (p(O2 ) = 0.01 . . . 0.05 mbar). Therefore, γ is not only increased by the alloying-eﬀect, but also by a lower crystal quality (see Fig. 3.14).

3 Optical Properties of ZnO and Related Compounds ZnO:Al / α-Al2O3

42

Ψ [˚]

103

39

36 400

600

800

1000

1200

ω [cm−1]

Fig. 3.17. Experimental (dotted lines) and best-model (solid lines) IRSE spectra of an highly Al-doped ZnO thin ﬁlm (d ∼ 1 400 nm) grown by PLD on sapphire [43]. The best-model free-charge-carrier parameters are N = (5.7±0.1)×1019 cm−3 , µopt || 2 −1 −1 = 7 cm2 Vs−1 , and µopt s ⊥ = 106 cm V

The experimental IRSE data were analyzed assuming an isotropicallyaveraged eﬀective electron mass parameter of m∗ = 0.28 me [135].6 Thereupon, the free-charge-carrier concentration N and the optical mobility (i = ||, ⊥) were obtained [43]. The results of the IRSEparameters µopt i analysis of two sets of Ga-doped ZnO thin ﬁlms are summarized in Fig. 3.18. For comparison, the results of electrical Hall-eﬀect measurements performed on the same samples are also included in Fig. 3.18. Both sample sets reveal increasing free-charge-carrier concentration and free-charge-carrier mobility parameters with decreasing oxygen pressure and a maximum freecharge-carrier concentration of about N ∼ 4 × 1019cm−3 . N and µ|| obtained by IRSE and Hall-eﬀect agree reasonably well. Note that the Hall-eﬀect measures for c-plane orientation µ⊥ only. For (0001) ZnO thin ﬁlms, IRSE analysis revealed an anisotropy of the mobility parameters, which was found to opt fulﬁll always the relation µopt || < µ⊥ . Figure 3.19 depicts IRSE spectra of two magnetron-sputtered polycrystalline Al-doped ZnO thin ﬁlms with high and extremely-high N values. Because of the large electron density of N = 1 × 1021 cm−3 , the strong plasmon contribution to the DF in the IR range completely screens the polar lattice mode excitation in this sample. In highly-doped samples, or samples with codoping, it is often found that free charge carriers are not homogeneously distributed across the layer depth, but rather adopt a certain concentration proﬁle. Figures 3.20a,b depict IRSE spectra of a Cu-doped ZnO thin ﬁlm, together with two diﬀerent best-model calculated data sets. In Fig. 3.20a, two layers were included into the model calculation, allowing for independent free-charge-carrier parameters within the two sublayers, whereas the phonon-mode parameters within the two layers share a common set. Figure 3.20b shows best-model calculations assuming 6

It is implied that the inverse conduction band eﬀective mass tensor does not depend on the wavevector k, and that the inverse eﬀective mass tensor renders the unit matrix times a k-independent scalar parameter (m )−1 .

104

C. Bundesmann et al. 60

(a)

(c)

50 40

1E19

30 20 1E18

µ [cm2 / Vs]

N [cm-3]

10

(b) 1E19

60

(d)

50 40 30 20

1E18

1E-4

10 1E-3

0.01

0 1E-4

0.1

1E-3

0.01

0.1

p(O2) [mbar]

p(O2) [mbar]

Fig. 3.18. Free-charge-carrier concentration (a,b) and mobility parameters (c,d) of Ga-doped ZnO thin ﬁlms on sapphire vs. oxygen pressure during PLD-growth [43]. Triangles and circles correspond to the results determined by IRSE and Hall-eﬀect measurements, respectively. Panels (a,c) and (b,d) contain the results of the ﬁlms grown with 0.1 and 0.5 mass percent Ga2 O3 powder within the PLD target, respectively. Up- and down-triangles in panels (c) and (d) represent the anisotropic optical and µopt mobility parameter µopt ⊥ , respectively ||

Ψ [˚]

44

ZnO:Al/Mo/PI

42

40

38

Φa = 55˚

N=1.0(0.2)*1021cm–3 µopt=1.0(0.3)cm2/Vs d=200nm

Ψ [˚]

44 N=1.0(0.2)*1019cm–3 µopt=2.03(0.02)cm2/Vs d=210nm

42 40 38 400

600

800 1000 1200 1400

ω [cm−1]

Fig. 3.19. Experimental (dotted lines) and best-model (solid lines) IRSE spectra of two polycrystalline Al-doped ZnO thin ﬁlms grown by magnetron sputtering on metallized polyimide foil [43]. The best-model free-charge-carrier concentration, optical mobility, and thickness parameters are indicated

3 Optical Properties of ZnO and Related Compounds (a)

Ψ [˚]

60

ZnO:Cu#2 ZnO:Cu#1 α-Al2O3

75

ZnO:Cu / α-Al2O3

Φa = 70˚

45

ZnO:Cu /α-Al2O3

(b) ZnO:Cu

60

Ψ [˚]

75

105

α-Al2O3

Φa = 70˚

45 30

30

15

15 400

600

800

1000

ω [cm–1]

1200

400

600

800

1000

1200

ω [cm–1]

Fig. 3.20. Experimental (dotted lines) and best-model (solid lines) IRSE spectra of a PLD-grown Cu-doped ZnO thin ﬁlm (d ∼ 1450 nm) on sapphire [43]. Panel (a) contains the best-model calculation, which is obtained by dividing the ZnO layer into two sublayers with diﬀerent free-charge-carrier parameters, as sketched in the inset. The best-model free-charge-carrier parameters in sublayer 1 (d ∼ 900 nm) are N = (8.15 ± 0.01) × 1018 cm−3 , µopt = (32.5 ± 0.3) cm2 V s−1 , and ⊥ opt 2 −1 −1 µ|| = (29.9 ± 0.4) cm V s . The free-charge-carrier concentration in sublayer 2 (d ∼ 550 nm) is below the IRSE detection limit of N ∼ 5 × 1017 cm−3 . Panel (b) contains the best-model data, which are obtained by modeling the ZnO thin ﬁlm as one homogeneous layer

a single layer for this sample. As can be seen, the best-model calculation improved considerably upon inclusion of a second layer, where it turns out that a large free-charge-carrier depleted zone (depletion layer) has formed near the top of the deposited thin ﬁlm [43].

3.6 Below-Band-Gap Index of Refraction 3.6.1 ZnO Figure 3.21a,c shows the index-of-refraction spectra n|| and n⊥ of ZnO below the fundamental band gap. The dispersion of the spectra can be described by the Cauchy formula [117] 2 (k) (k) Bi xk Ci xk (k) k Ai x + , (3.24) + ni (x, λ) = λ2 λ4 k=0

where λ denotes the vacuum wavelength and i indicates the two polarization directions (i = ||, ⊥). The parameter x describes the composition dependencies of the Cauchy parameters for Mgx Zn1−x O compounds (Sect. 3.6.2). The use of the Cauchy formulae requires a high-frequency (short-wavelength) limitation, which is approximated here by 90% of the respective fundamental band-to-band transition energy (band gap energy). The parameter Ai is √ equal to ε∞,i , ignoring ﬁnite absorption due to defect- or impurity-related

106

C. Bundesmann et al. Photon energy [eV] 2.0

2.5

3.5

x = 0 (ZnO) (a)

MgxZn1-xO

0.10 0.15 0.17

2.0

n⊥

3.0

0.19 0.23

1.9

1

2

3

4

5

6

x =0 .68 0.72 0.82 0.88 0.89 0.96 1.00

7

(b)

2.0 1.9

0.29

1.8

1.8

0.37

1.7

1.7 0.05

5.3e V

(c)

x = 0 (ZnO)

(d)

0.00

n||-n⊥

2.1

2.0

0.10 0.15 0.17 0.19

–0.05

3.4e V 0.23 0.29

–0.10

1.9 1.8

1.0e V

0.37

–0.15 1.0

1.5

2.0

2.5

2.1

3.0

Photon energy [eV]

3.5

n(⊥)

2.1

1.5

n

1.0

1.7 0.0

0.2

0.4

0.6

0.8

1.0

x

Fig. 3.21. Index-of-refraction spectra of PLD-grown ZnO and Mgx Zn1−x O thin ﬁlms. (a) and (b): n⊥ for wurtzite-structure and n for rocksalt-structure Mgx Zn1−x O, respectively. (c) Birefringence n –n⊥ for wurtzite-structure Mgx Zn1−x O. For some of the curves error bars are indicated. (d): n⊥ (x < 0.45) and n (x > 0.6) for selected photon energies according to (3.24) and Tables 3.7 and 3.8 (solid lines), and experimental data points (symbols) from individual thin ﬁlm samples. The dashed lines in panels (a) and (c) represent values from an a-plane ZnO single crystal bulk sample [121], and in panel (b) from an MgO single crystal bulk sample [136]. Reprinted with permission from [15, 16]

below-band-gap transitions. The Cauchy-parameters of ZnO are summarized in Table 3.7. The birefringence, which is deﬁned by ∆n = n – n⊥ , is positive with reported values ranging from ∆n = 0.001 to ∆n = 0.05 [15, 76, 95, 121, 124, 137]. Figure 3.22 plots the temperature derivative dn/dT for n⊥ within the temperature range from T = 40 K to T = 800 K with an almost linear increase of dn⊥ /dT with photon energy. 3.6.2 Mgx Zn1−x O Typical index-of-refraction spectra of Mgx Zn1−x O thin ﬁlms with wurtzite and rocksalt structure are plotted in Figs. 3.21a–d. Data for 0 ≤ x ≤ 1 originate from thin ﬁlm measurements, those for x = 0 and x = 1 are

3 Optical Properties of ZnO and Related Compounds

107

Table 3.7. Cauchy parameters (3.24) for wurtzite-structure ZnO and Mgx Zn1−x O (0)

Ai Mollwo (1954)

[137]a

Bond (1965)

[124]a

Hu (1997) [138]b Yoshikawa (1997) [76]a Jellison (1998)

[121]a

Sun (1999) Teng (2000)

[79]b [95]c

Schmidt (2003)

[15]b

Schmidt (2003)

[15]d

a b c

2.001 1.991 1.925 1.911 1.9281 1.962 1.961 1.947 1.939 1.909 1.915 1.899 1.966 1.916 1.844 1.916

(0)

−0.26 −0.25 −0.78 −0.57

(1)

Bi Bi (10−2 µm2 ) −2.08 −2.52 3.05 2.91 −0.0011 −1.19 −2.10 1.18 0.63 1.51 2.92 2.85 1.81 1.76 1.81 1.76

−2.9 −3.2 −4.5 −4.5

(0)

(1)

Ci Ci (10−3 µm4 ) 9.8 10.5 1.61 1.74 5.97 9.4 11.5 4.7 5.4 4.7 1.7 1.6 3.6 3.9 3.6 3.9

−3.0 −3.1 −4.9 −4.9

ZnO single crystal bulk ZnO thin ﬁlm Mgx Zn1−x O thin ﬁlms with 0 ≤ x ≤ 0.36 Mgx Zn1−x O thin ﬁlms with 0.1 ≤ x ≤ 0.37

dn⊥ /dT [1/104K]

d

n n⊥ n n⊥ n⊥ n n⊥ n n⊥ n n n⊥ n n⊥ n n⊥

(1)

Ai

1.0

ZnO E⊥c

0.9

0.8 1.5

2.0

2.5

Photon energy [eV] Fig. 3.22. Temperature dependence dn⊥ /dT of the index of refraction of ZnO below the band gap for selected photon energies, as determined in the temperature range from T = 60 K to T = 800 K. dn⊥ /dT was found to be nearly constant in the studied temperature range

supplemented by data from single-crystal bulk samples. The corresponding Cauchy parameters are listed in Tables 3.7 and 3.8. In general, with increasing x, the indices of refraction decrease. Figure 3.21c shows the birefringence spectra. For x ≥ 0.1, ∆n is negative [15], in contrast to ZnO (Sect. 3.6.1), and increases with increasing Mg content x. Figure 3.21d shows the composition

108

C. Bundesmann et al. Table 3.8. Same as Table 3.7 for rocksalt-structure Mgx Zn1−x O (0)

Ai

(1)

Ai

(2)

Ai

(0)

Bi

(1) Bi −2

(10 Chen (2005) [139]a 2.016 −0.485 0.160 0.7 Schmidt-Grund (2006) [16]b 2.146 −0.508 0.083 1.6 a b

(2)

Bi µm2 )

(0)

Ci

(1)

(2)

Ci Ci (10 µm4 ) −3

1.7 −1.8 1.9 −2.5 0.7 −2.5 1.3 1.38 −1.48 0.36

Mgx Zn1−x O thin ﬁlms with 0.57 ≤ x ≤ 1 Mgx Zn1−x O thin ﬁlms with 0.68 ≤ x ≤ 1

dependence of the index of refraction, exemplarily, for three diﬀerent photon energies [15, 16]. In analogy to the discontinuous composition dependence of the static and “high-frequency” dielectric constants (Sect. 3.3), the phononmode frequencies (3.4.3), and the fundamental band-to-band transition energies E0α (Sect. 3.7.2), the index of refraction reveals a strong discontinuity and increment upon phase transition. Note that second-order Cauchy coefﬁcients are required to describe the composition dependence of the index of refraction for the rocksalt-structure compounds. Sum rule considerations [140] imply that the discontinuity in the refractive index must be accompanied by a substantial increase of the oscillator strengths of higher energy transitions above the spectral range, from which data are available (E < 10 eV). Otherwise the discontinuity of the lowest band-to-band transition energy (Sect. 3.7) would cause a decrease in the refractive index if the oscillator strengths within the rocksalt-structure part of the alloy would remain comparable to those within the wurtzitestructure part.

3.7 Band-to-Band Transitions and Excitonic Properties 3.7.1 ZnO 3.7.1.1 Band-to-Band Transitions The DF spectra of wurtzite-structure ZnO within the VIS-to-VUV spectral region contain CP structures, which can be assigned to band-gap-related electronic band-to-band transitions E0α with α = A, B, C and to above-band-gap band-to-band transitions E β with β = 1, . . . , 7. The E0α -related structures can be described by lineshape functions of the 3DM0 -type (3.9 and 3.10), the CP structures with β = 3, 4 by lineshape functions of the 2DM0 -type (3.11), and the CP structures with β = 1, 2, 5, 6, 7 can be described by Lorentziandamped harmonic oscillator functions (3.13). The CP structures E0α are supplemented by discrete (3.14) and continuum (3.16) excitonic contributions. Tables 3.9 and 3.10 summarize typical parameters of the CPs E0α and E β , respectively, of ZnO [15].

3 Optical Properties of ZnO and Related Compounds

109

Table 3.9. Typical energy, amplitude, and broadening parameters of near-bandgap CP structures E0α (α = A, B, and C) and their exciton polarizabilities of ZnO obtained from SE analysis of multiple sample sets at RT ε⊥ E0A AA 0 AA xc A1,A xb E0B AB 0 AB xc A1,B xb E0C AC 0 AC xc A1,C xb 1 Exb Γ0 ∼ Γxc ∼ Γxb

(eV) (eV3/2 ) (eV2 ) (eV) (eV) (eV3/2 ) (eV2 ) (eV) (eV) (eV3/2 ) (eV2 ) (eV) (eV) (eV)

3.366 6.5 ± 1.1 0.33 ± 0.02 0.023 ± 0.003 3.380 1.5 ± 1.1 0.097 ± 0.016 0.044 ± 0.009 3.434 0.21 ± 0.4 0.017 ± 0.015 0.026 ± 0.008 0.060 0.020

± 0.004 1.5 0.03 0.015 ± 0.004 0.0 0.00 0.000 ± 0.018 1.6 0.57 0.078 ± 0.002 ± 0.009

ε ± 1.3 ± 0.03 ± 0.010 ± 1.0 ± 0.10 ± 0.010 ± 0.8 ± 0.07 ± 0.009

Table 3.10. Same as Table 3.9 for CP structures E β (β = 1 . . . 7) E1 A1 Γ1 E2 A2 Γ2 E3 A3 Γ3 E4 A4 Γ4 E5 A5 Γ5 E6 A6 Γ6 E7 A7 Γ7

(eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV) (eV)

ε⊥ 5.494 ± 0.063 0.50 ± 0.08 3.6 ± 0.6 6.494 ± 0.02 0.16 ± 0.03 1.0 ± 0.1 7.135 ± 0.021 0.34 ± 0.04 0.29 ± 0.02 7.537 ± 0.040 0.18 ± 0.06 0.32 ± 0.03 – – – 8.707 ± 0.005 0.48 ± 0.05 1.03 ± 0.08 9.044 ± 0.008 0.13 ± 0.04 0.49 ± 0.08

5.268 0.22 3.0 6.419 0.06 0.53 7.086 0.09 0.11 7.341 0.07 0.10 8.167 0.55 1.95 8.881 0.85 1.05

ε ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± – – –

0.09 0.13 0.8 0.17 0.01 0.09 0.013 0.02 0.02 0.015 0.01 0.03 0.044 0.04 0.15 0.005 0.04 0.03

110

C. Bundesmann et al.

3.7.1.2 Γ -Point Fundamental Band-to-Band Transition The fundamental absorption edge in ZnO corresponds to the direct transition from the highest valence band to the lowest conduction band at the Γ -point of the Brillouin zone [141]. RT-data for the energy of the lowest Γ -point band-to-band transition are summarized in Table 3.11. 3.7.1.3 Γ -Point Valence-Band Ordering The lowest conduction band of ZnO exhibits Γ7 -symmetry. The highest valence band of ZnO is split into two bands with Γ7- and one band with Γ9 -symmetry due to the spin-orbit and crystal-ﬁeld interactions [76,154–158]. The corresponding three lowest Γ -point band-to-band transition energies are denoted here by E0α . So far, the order of the highest valence bands of unstrained ZnO, Γ7 -Γ9 -Γ7 (Γ797 ) or Γ9 -Γ7 -Γ7 (Γ977 ), is still a subject of debate [157, 158]. The RT properties of the E0α -band-to-band transitions of ZnO thin ﬁlms deposited on sapphire substrate were reported as follows: Although for ε⊥ all amplitude parameters B C B B AA 0 , A0 , and A0 are ﬁnite, for ε the E0 transition is suppressed (A0 = 0) in agreement with theoretical predictions for the case of the valence-band order Γ797 , where transitions between the conduction band and the valence band with Γ9 -symmetry are forbidden to couple to photons polarized perpendicular to the optical c-axis [76]. 3.7.1.4 Crystal-Field Splitting and Spin–Orbit Coupling Parameters The quasi-cubic model approximation can be invoked to calculate the crystalﬁeld splitting (∆cf ) and spin-orbit coupling (∆so ) parameters from E0α [76, 159, 160]

1 2 ∆cf = − ∆12 − ∆13 + (∆12 + ∆13 ) − 6∆12 ∆13 , (3.25) 2

1 2 ∆12 − ∆13 − (∆12 + ∆13 ) − 6∆12 ∆13 , ∆so = − (3.26) 2 α

where ∆ij = E0αi − E0 j are the valence-band splitting energies. Reported ∆cf - and ∆so -values for ZnO are summarized in Table 3.12. 3.7.1.5 Excitonic Properties Discrete and continuum free exciton contributions can be identiﬁed for each of the band-to-band transitions E0α . Sharp resonance features are superimposed to each of the E0α CP structures because of discrete exciton lines, where the ground and ﬁrst excited state (n = 1, 2) can be seen at RT, and the second

3 Optical Properties of ZnO and Related Compounds

111

Table 3.11. RT-data for E0α (lowest Γ -point band-to-band transitions) and Eg (fundamental absorption edge determined from transmission measurements), and 1 of ZnO single crystal bulk samples (b) and ZnO thin ﬁlms (f) Exb Eg Ref.

Sample

Yoshikawa (1997) [76] (b) Jellison (1998) [121] (b) Washington (1998) [78] (f) Muth (1999) [142] (f) Kang (2000) [90] (f) Rebien (2002) [122] (f) Zhao (2002) [96] (f) Djurisic (2003) [143] (b) Djurisic (2003) [143] (b) Ozaki (2003) [144] (b) Schmidt (2003) [15] (f) Kang (2004) [82] (b) Kang (2004) [82] (f) Srikant (1997) [145] (f) Ohtomo (1998) [11] (f) Studenikin (1998) [146] (f) Paraguay (1999) [147] (f) Meng (2000) [148] (f) Minemoto (2000) [99] (f) Postava (2000) [80] (f) Santana (2000) [149] (f) Park (2002) [87] (f) Tokumoto (2002) [150] (f) Takeuchi (2003) [98] (f) Chen (2003) [100] (f) Shan (2003) [151] (f) Misra (2004) [152] (f) Shan (2004) [97] (f) Zhao (2005) [153] (f) 1,α a Exb with α = A, B, and C b ε⊥ : 3.471 eV, ε : 3.432 eV c ε⊥ : 3.548 eV d ε⊥ : 3.453 eV e ε⊥ : 3.484 eV, ε : 3.484 eV 1,α f Exb with α = A, B, and C

(eV)

E0A (ε⊥ ) (ε ) (eV)

E0B

E0C

1 Exb (ε⊥ ) (ε ) (eV) (eV) (meV)

a b 3.450 3.450 59 59 3.372 3.405 56 50 3.372 60 c 3.40 3.45 3.55 3.34 60 3.37 3.4 60 3.406 3.445 72 73 e 3.407 3.4889 d 90 78 f 3.380 3.394 3.438 3.366 3.380 3.434 60 3.373 3.407 55.3 51.3 3.378 3.413 61.1 72.8

3.24 . . . 3.32 3.29 3.195 . . . 3.370 3.28 3.21 3.24 3.35 . . . 3.44 3.28 3.25 3.28 3.28 3.273 3.25 . . . 3.28 3.28 3.274 3.25 . . . 3.27 is 68, 65, and 63 meV, respectively

45. . . 71

is 63.1, 50.4, and 48.9 meV, respectively

exited state (n = 3) evolves at low temperatures (3.15, Figs. 3.5, 3.23 [15]). n,A n,B n,C n := Exb = Exb = Exb for transitions E0α . It is often assumed that Exb 1 Table 3.11 summarizes data reported for Exb by various authors. An MDF parameter set for the exciton contributions to the DF is listed in Table 3.9.

112

C. Bundesmann et al.

Table 3.12. Crystal-ﬁeld splitting (∆cf ) and spin-orbit coupling (∆so ) energies for ZnO single crystal bulk samples (b) and ZnO thin ﬁlms (f) depending on the assumed valence-band orderinga Ref.

Valence-band ∆cf ordering (meV)

Sample Temp.

Liang (1968) [161] Langer (1970) [162] Reynolds (1999) [163] Mang (1995) [164] Srikant (1997) [145] Ozaki (2003) [144] Ozaki (2003) [144] Schmidt-Grund (unpubl.) Schmidt-Grund (unpubl.)

(b) (b) (b) (b) (f) (b) (b) (f) (f)

RT RT RT RT RT RT RT RT → 5K RT → 5K

Γ797 Γ977 Γ977 Γ797 Γ797 Γ977 Γ797 Γ977 → Γ797 † Γ797 → Γ977 †

∆so (meV)

40.8 −4.7 43 19 43 16 39.4 −3.5 42 −5 45 27 49 −19 50 → −82 19 → 16 51 → −82 −15 → −18

a

Regardless of the assumed valence-band ordering, ∆cf is positive at RT and negative at T = 5 K. This implies that the valence-band ordering must change († ) due to temperature-induced biaxial-strain variation

The amplitude parameter selection rules for the discrete and continuum exciton contributions were found concordant with those of the E0α transitions. 3.7.1.6 Temperature Inﬂuence on the Γ -Point Transitions The Γ -point transition energies shift to lower energies with increasing temperature. The energy shift is accompanied by a continuous lineshape broadening, and is caused mainly by the increase in electron–phonon interaction. The inﬂuence of the electron–phonon interaction on the band-to-band transition energies E0α at low temperatures is often addressed by the Varshni equation [165] αV T 2 . (3.27) E0α (T ) = E0α (0) − T + θD E0α (0) is the transition energy at T = 0 K, αV is a model parameter, and θD is the Debye temperature. At higher temperatures the semiempirical Bose–Einstein model function can be applied [129, 166] E0α (T ) = E0α (0) −

αBE θBE , exp(θBE /T ) − 1

(3.28)

which assumes an eﬀective phonon-mode temperature θBE . E0α (0) is again the transition energy at T = 0 K, αBE is a model parameter. Reported parameters for the Varshni and the Bose–Einstein model for ZnO bulk samples are given in Table 3.13. It was pointed out that contributions due to acoustical and optical phonons must be considered for ZnO [169–172]. This is done by the twooscillator model [169, 173]

3 Optical Properties of ZnO and Related Compounds

14

T=5K

Im {ε⊥}

10

e dex,

2

n=

ZnO E⊥c

e dex, A

12 1

23

113

e dex,

B

e cex,

C

A,B,C

1 e 3DM0,

5

A,B,C

T=5K 0 3.3

3.4

3.5

RT T=830K 0 3.0

3.2 3.4 Photon energy [eV]

3.6

Fig. 3.23. Im{ε⊥ } of a PLD-grown ZnO thin ﬁlm at T = 5 K, RT, and T = 830 K in the photon energy range of the lowest Γ -point band-to-band transitions determined by SE. The inset depicts the contributions of the discrete (εdex,A,B,C ) and the continuum (εcex,A,B,C ) exciton polarizabilities and the lowest Γ -point band-to-band transitions (ε3DM0 ,A,B,C ) at T = 5 K. The ground, ﬁrst, and second excited exciton states (n = 1, 2, 3) are labeled (3.15). Note the axis brake in the inset. Reprinted with permission from [85] Table 3.13. Parameters of the temperature dependence of E0A according to the Varshni and Bose–Einstein model for ZnO single crystal bulk samples Varshni-model Bose–Einstein-model E0A (0 K) αV θD E0A (0K) αBE θBE (eV) (10−4 eV K−1 ) (K) (eV) (10−4 eV K−1 ) (K)

Ref. Boemare (2001) [167] Ozaki (2003) [144] Wang (2003) [168]

3.4407 3.441 3.440

6.7 6.5 8.2

E0α (T ) = E0α (0) − α2O

672 660 700 2 i=1

3.436

2.5

203

3.440

3.75

240

Wi θi , exp(θi /T ) − 1

(3.29)

which describes the electron–phonon coupling in ZnO for both low and high temperatures. θi are the associated phonon temperatures, and Wi represent

114

C. Bundesmann et al.

Table 3.14. Parameters of the temperature dependence of the transition energy E0A and the discrete free exciton FWHM W2O according to the 2-oscillator model for a ZnO thin ﬁlma E0A /W2O E0A W2O a

E0A (0)/W2O (0) α2O (eV) (10−4 eV K−1 ) 3.4365 (0.0004) 0.0124 (0.0010)

4.1 (0.1) 2.8 (0.5)

W1

θ1 (K)

θ2 (K)

θ2O (K)

δ

θD (K)

0.30 145 529 414 0.42 621 (0.05) (25) (30) (28) (0.10) (42) 0.27 129 680 531 0.46 797 (0.18) (50) (50) (50) (0.12) (75)

Note that W2 = 1 − W1 ; Numbers in parenthesis are error bars

their relative weight. E0α (0) is the transition energy at T = 0 K. α2O is the slope of the energy shift at temperatures far above the highest phonon temperature.7 Figure 3.25b depicts E0A (T )-data obtained from a (0001) ZnO thin ﬁlm on (0001) sapphire together with the best-model calculation according to (3.29) with k=2. The best-model parameters are given in Table 3.14. Here, the E0A (0)-data are not corrected for an energy shift, which originates in the thermal-expansion-mismatch induced biaxial strain8 between the ZnO thin ﬁlm and the substrate. The phonon energy of the ﬁrst branch kB θ1 ∼ 13 meV (ω ∼ 101 cm−1 ) falls into the range of the acoustical phonon branch of ZnO [106, 174, 175]. The energy of the second phonon branch kB θ2 ∼ 46 meV (ω ∼ 368 cm−1 ) is in good agreement with the phonon-bandedge, i.e., close to the energy of the lowest occupied states of the highenergy ZnO optical phonon branch. The latter extends from about 47 meV (ω[A1 (TO)] = 380 cm−1 ) to about 73 meV (ω[E1 (LO)] = 588 cm−1 ) [38]. Calculations of the phonon density of states showed a strong maximum at the lower part of the upper phonon branch [174]. According to W1 = 0.30 and W2 = 0.70 optical phonons contribute much stronger to the temperature dependence of E0A than acoustical phonons. The Debye-temperature can be estimated by θD ∼ (3/2)θ2O [176]. Tables 3.14 summarizes the parameters for the temperature dependence of E0A for a ZnO thin ﬁlm. The electron–phonon interactions also inﬂuence the linewidth of the interband and exciton transitions. This temperature-dependent homogeneous 7

8

An eﬀective phonon temperature can be deﬁned, e.g., for k=2, by the weighted mean-value of the discrete phonon temperatures, θ2O ≡ W1 θ1 + W2 θ2 . The

material-related parameter δ = (θ2 − θ2O )(θ2O − θ1 )/θ2O can be used to distinguish between weak (δ < 0.33), intermediate (0.33 ≤ δ ≤ 0.577), and strong coupling regimes (δ > 0.57) [173]. For ZnO the coupling regime is intermediate (Table 3.14). Of more subtle inﬂuence is the biaxial strain in thin ﬁlm samples, which is introduced upon diﬀerent thermal expansion coeﬃcients for the ﬁlm and the substrate material. Furthermore, the biaxial strain can depend on the growth history. On this matter the amount of available information is not exhaustive.

3 Optical Properties of ZnO and Related Compounds 250

ZnO

200 W [meV]

115

Experiment Model

150 100 50 0

0

200

400 T [K]

600

800

Fig. 3.24. Experimental data (symbols) of the FWHM W of the discrete free exciton together with model calculations (solid line) according to the two-oscillator model (W2O (T ), 3.30) vs. temperature of a ZnO thin ﬁlm

broadening is superimposed by the (often assumed) temperature-independent lifetime and inhomogeneous broadening due to exciton-, disorder-, and defectinduced exciton scattering. Within the 2-oscillator model, the temperature dependence of the full width at half maximum value (FWHM) W can be described, similarly to the temperature dependence of the fundamental bandto-band transition energy [177], by W2O (T ) = W2O (0) − α2O

2 i=1

Wi θi exp(θi /T ) − 1

(3.30)

with W2O (T = 0K) = W2O (0) and all other parameters as deﬁned in (3.29). Figure 3.24a depicts the temperature dependence of W for the E0A exciton line and the best-model calculation according to (3.30). The experimental data were obtained from MDF analysis of temperature-dependent SE spectra taken from a (0001) ZnO thin ﬁlm on (0001) sapphire. The corresponding parameters α2O , Wi , and θi are summarized in Table 3.14. The best-model free-exciton FWHM at T = 0 K is W2O (0) = 12.4 ± 1 meV. This value is larger than the FWHM value of ∼ 2 . . . 5 meV for the donor bound exciton measured by low-temperature cathodoluminescence and photoluminescence excitation spectroscopy [15, 19]. Figure 3.25a shows the temperature dependence of the E0α transition energy parameters. At high temperatures the slope is linear. Table 3.15 summarizes (∂E0α /∂T )|T →∞ and (∂Eg /∂T )|T →∞ data for ZnO. Superimposed on the electron–phonon coupling-induced band shift are again eﬀects due to thermal-induced biaxial strain, which eﬀects the crystal-ﬁeld splitting parameter. The spin-orbit coupling parameter should depend only slightly on temperature.

116

C. Bundesmann et al.

ZnO 3.50 3.435

3.45 3.430

Energy [eV]

3.40 0

50

100

3.35

3.30

PL (E0A)

3.25

SE (E0A) 3.20 (b)

(a) 3.15 0

200

400

600

800

0

200

400

600

800

T [K]

Fig. 3.25. (a): Γ -point band-to-band transition energies E0A (squares), E0B (triangles), and E0C (circles) of a PLD-grown ZnO thin ﬁlm as a function of temperature obtained by SE. (b) Data for E0A from SE (solid symbols) and low-temperature PL (open symbols), and model calculations according to (3.29) (solid line). The inset enlarges the temperature range T ≤ 120 K. Table 3.15. High-temperature slope (∂E0α /∂T )|T →∞ in units of 10−4 eV K−1 of ZnO single crystal bulk samples (b) and ZnO thin ﬁlms (f) Ref. Watanabe (1964) Ozaki (2003) Ozaki (2003) Schmidt-Grund (unpubl.)

[178] [144] [144]

Sample

E0A

E0B

E0C

(b) (b) (b) (f)

−3.4 −2.9 −4.3

−3.4 −2.9 −4.3

−3.4 −2.9 4.1

Eg −8

3.7.2 Mgx Zn1−x O 3.7.2.1 Γ -Point Band-to-Band Transitions Mgx Zn1−x O crystallizes in the wurtzite or in the rocksalt structure, depending on the Mg mole fraction x. The alloys remain direct-gap materials over the whole composition range. The wurtzite-structure part reﬂects a valenceband structure, which is similar to ZnO. For the rocksalt-structure part the

3 Optical Properties of ZnO and Related Compounds

117

8.0 MgxZn1-xO

rocksalt-structure

Energy [eV]

7.0

6.0

5.0 wurtzite-structure 4.0

3.0

0.0

0.2

0.4

0.6

0.8

1.0

x

Fig. 3.26. Energies of the fundamental absorption edge Eg (open symbols) and the lowest Γ -point band-to-band transition E0A or E0D (ﬁlled symbols) of Mgx Zn1−x O, as determined by transmission measurements (T) and SE, respectively. Filled squares: Refs. [15, 16] (SE), ﬁlled circles: [95] (T), ﬁlled diamonds: [90] (SE), ﬁlled up-triangles: [96] (T), ﬁlled down-triangles: [82] (SE), open circles: [11] (T), open diamonds: [97] (T), open up-triangles: [14] (T), open down-triangles: [98] (T), open left-triangles: [99] (T), open right-triangles: [100] (T). The lines depict numerical approximations according to (3.31) with the parameters in Table 3.16 for E0A (wurtzite-structure Mgx Zn1−x O) and E0D (rocksalt-structure Mgx Zn1−x O)

crystal-ﬁeld splitting parameter vanishes, and the valence band is split into two bands only, with E0D = E0 and E E = E0 + ∆so , and their CP structures can be well adopted by those of the 3DM0 -type (3.9). Figure 3.26 summarizes data of the band gap energies of Mgx Zn1−x O vs. x. With increasing Mg-content the Γ -point band-to-band transitions shift to higher photon energies with a small bowing, which can be described by E0α (x) = E0α (x = 0) + pα x + q α x2 .

(3.31)

The best-model parameters pα and q α from [15, 16] are given in Table 3.16. The data in Table 3.16 may be used to estimate the band gap energy for unstrained wurtzite-structure MgO of E0A = 6.9 eV and for rocksalt-structure ZnO of E0D = 7.6 eV, with stronger bowing for the rocksalt-structure than for the wurtzite-structure occurrence of the alloys. Theoretical band-structure calculations for ZnO revealed the high-pressure rocksalt-structure phase as

118

C. Bundesmann et al.

Table 3.16. Best-model parameters for the Mg-content dependence of the Γ -point band-to-band transition energies E0α (3.31) at RT [15, 16] α A B C D E

x 0 . . . 0.51 0 . . . 0.51 0 . . . 0.23 0.68 . . . 1 0.68 . . . 1

E0α (x = 0) (eV) 3.369 ± 0.002 3.379 ± 0.003 3.427 ± 0.007 7.6 ± 0.5 8.0 ± 0.8

pα (eV) 1.93 ± 0.06 2.0 ± 0.1 2.2 ± 0.3 −7 ± 1 −7.5 ± 2

q α (eV) 1.57 ± 0.25 1.9 ± 0.4 1.4 ± 0.9 7.0 ± 1 7.2 ± 1.5

an indirect gap material (Eind ∼ 5.4 eV), for which the Γ -point band gap energy E0D = 6.54 eV is predicted to be much larger than E0A = 3.34 eV of the wurtzite-structure ZnO [121, 179, 180], which is in qualitative agreement with the extrapolation in Fig. 3.26. Acknowledgement. The authors thank Prof. M. Grundmann, Prof. W. Grill, and Prof. B. Rheinl¨ ander (all Universit¨ at Leipzig) for continuing support and enlightening discussions. We acknowledge technical support by A. Rahm (XRD), D. Speman (RBS), U. Teschner (Raman), and H. v. Wenckstern (Hall, all Universit¨at Leipzig), D. Faltermeier, Dr. B. Gompf, and Prof. M. Dressel (low temperature ellipsometry, all Universit¨ at Stuttgart), and Dr. C. M. Herzinger (VUV-ellipsometry, J. A. Woollam Co., Inc., Lincoln, Nebraska, USA). We thank H. Hochmuth, Dr. E. M. Kaidashev, G. Ramm, and Dr. M. Lorenz for the PLD-preparation (see Chap. 7) of most of the thin ﬁlm samples studied in this chapter, and the Solarion GmbH Leipzig for providing magnetron sputtered thin ﬁlm samples. Finally, we thank all current and former members of the Workgroup Ellipsometry at the Universit¨at Leipzig for diverse support: N. Ashkenov, A. Carstens, T. Chavdarov, Dr. T. Hofmann, Dr. A. Kasic, Dr. G. Leibiger, B. N. Mbenkum, M. Saenger, and C. Sturm. Financial support was provided by the German Federal Ministry of Education and Research BMBF (FK 03WKI09) and the German Research Foundation DFG (Grant No. SCHUH 1338/3-1, SCHUH 1338/4-1, SCHUH 1338/4-2, Gr 1011/10-1, and Gr 1011/14-1).

References 1. D.M. Bagnall, Y.F. Chen, Z. Zhu, T. Yao, S. Koyama, M.Y. Shen, T. Goto, Appl. Phys. Lett. 70, 2230 (1997) 2. D.C. Reynolds, D.C. Look, B. Jogai, Solid State Commun. 99, 873 (1996) 3. Y. Segawa, A. Ohtomo, M. Kawasaki, H. Koinuma, Z. Tang, P. Yu, G. Wong, Phys. Status Solidi B 202, 669 (1997) 4. P. Zu, Z.K. Tang, G.K.L. Wong, M. Kawasaki, A. Ohtomo, H. Koinuma, Y. Segawa, Solid State Commun. 103, 459 (1997) 5. Z.K. Tang, G.K.L. Wong, P. Yu, M. Kawasaki, A. Ohtomo, H. Koinuma, Y. Segawa, Appl. Phys. Lett. 72, 3270 (1998)

3 Optical Properties of ZnO and Related Compounds

119

6. W. Hirschwald, P. Bonasewicz, L. Ernst, M. Grade, D. Hofmann, S. Krebs, R. Littbarski, G. Neumann, M. Grunze, D. Kolb, H.J. Schulz, Current Topics in Materials Science, vol. 7 (North-Holland, Amsterdam, 1981) 7. I. Vurgaftman, J.R. Meyer, J. Appl. Phys. 94, 3675 (2003) 8. D.M. Roessler, W.C. Walker, Phys. Rev. 159, 733 (1967) 9. F. Koﬀyberg, Phys. Rev. B 13, 4470 (1976) 10. D. Look, D. Reynolds, J. Sizelove, R. Jones, C. Litton, G. Cantwe, W. Harsch, Solid State Commun. 105, 399 (1998) 11. A. Ohtomo, M. Kawasaki, T. Koida, K. Masubuchi, H. Koinuma, Y. Sakurai, Y. Yoshida, T. Yasuda, Y. Segawa, Appl. Phys. Lett. 72, 2466 (1998) 12. A.K. Sharma, J. Narayan, J.F. Muth, C.W. Teng, C. Jin, A. Kvit, R.M. Kolbas, O.W. Holland, Appl. Phys. Lett. 75, 3327 (1999) 13. T. Makino, Y. Segawa, M. Kawasaki, A. Ohtomo, R. Siroki, K. Tamura, T. Yasuda, H. Koinuma, Appl. Phys. Lett. 78, 1237 (2001) 14. S. Choopun, R.D. Vispute, W. Yang, R.P. Sharma, T. Venkatesan, H. Shen, Appl. Phys. Lett. 80, 1529 (2002) 15. R. Schmidt, B. Rheinl¨ ander, M. Schubert, D. Spemann, T. Butz, J. Lenzner, E.M. Kaidashev, M. Lorenz, A. Rahm, H.C. Semmelhack, M. Grundmann, Appl. Phys. Lett. 82, 2260 (2003) 16. R. Schmidt-Grund, A. Carstens, B. Rheinl¨ ander, D. Spemann, H. Hochmut, M. Lorenz, M. Grundmann, C.M. Herzinger, M. Schubert, J. Appl. Phys. 99, 123701 (2006) 17. T. Minami, MRS Bull. 25(Aug), 38 (2000) 18. S.J. Pearton, D.P. Norton, K. Ip, Y.W. Heo, T. Steiner, Superlattice Microst. 34, 3 (2003) 19. E.M. Kaidashev, M. Lorenz, H. von Wenckstern, G. Benndorf, A. Rahm, H.C. Semmelhack, K.H. Han, H. Hochmuth, C. Bundesmann, V. Riede, M. Grundmann, Appl. Phys. Lett. 82, 3901 (2003) 20. M. Lorenz, E.M. Kaidashev, H.v. Wenckstern, V. Riede, C. Bundesmann, D. Spemann, G. Benndorf, H. Hochmuth, A. Rahm, H.C. Semmelhack, M. Grundmann, Solid State Electron. 43, 2205 (2003) 21. E.C. Lee, K.J. Chang, Phys. Rev. B 70, 115210 (2004) 22. D.C. Look, B. Claﬂin, Y.I. Alivov, S.J. Park, Phys. Status Solidi A 201, 2203 (2004) 23. T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287, 1019 (2000) 24. K. Ueda, H. Tabata, T. Kawai, Appl. Phys. Lett. 79, 988 (2001) 25. K. Sato, H. Katayama-Yoshida, Jpn. J. Appl. Phys. 39, L555 (2000) 26. Z. Jin, M. Murakami, T. Fukumura, Y. Matsumoto, A. Ohtomo, M. Kawasaki, H. Koinuma, J. Cryst. Growth 214–215, 55 (2000) 27. H. Katayama-Yoshida, K. Sato, T. Yamamoto, Jpn. Soc. Appl. Phys. 6, 20 (2002) 28. M. Diaconu, H. Schmidt, H. Hochmuth, M. Lorenz, G. Benndorf, J. Lenzner, D. Spemann, A. Setzer, K.W. Nielsen, P. Esquinazi, M. Grundmann, Thin Solid Films 486, 117 (2004) 29. T. Gruber, G.M. Prinz, C. Kirchner, R. Kling, F. Reuss, W. Limmer, A. Waag, J. Appl. Phys. 96, 289 (2004) 30. S. Heitsch, C. Bundesmann, G. Wagner, G. Zimmermann, A. Rahm, H. Hochmuth, G. Benndorf, H. Schmidt, M. Schubert, M. Lorenz, M. Grundmann, Thin Solid Films 496, 234 (2006)

120 31. 32. 33. 34. 35. 36. 37. 38.

39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50.

51. 52. 53. 54. 55. 56. 57.

C. Bundesmann et al. R. Callender, S. Sussman, M. Selders, R. Chang, Phys. Rev. B 7, 3788 (1973) T.C. Damen, S.P.S. Porto, B. Tell, Phys. Rev. 142, 570 (1966) C.A. Arguello, D.L. Rousseau, S.P.S. Porto, Phys. Rev. 181, 1351 (1969) J.F. Scott, Phys. Rev. B 2, 1209 (1970) J.M. Calleja, M. Cardona, Phys. Rev. B 16, 3753 (1977) B.H. Bairamov, A. Heinrich, G. Irmer, V.V. Toporov, E. Ziegler, Phys. Status Solidi B 119, 227 (1983) F. Decremps, J. Pellicer-Porres, A.M. Saitta, J.C. Chervin, A. Polian, Phys. Rev. B 65, 092101 (2002) N. Ashkenov, B.N. Mbenkum, C. Bundesmann, V. Riede, M. Lorenz, E.M. Kaidashev, A. Kasic, M. Schubert, M. Grundmann, G. Wagner, H. Neumann, J. Appl. Phys. 93, 126 (2003) Y.F. Lu, H.Q. Ni, Z.M. Ren, W.J. Wang, T.C. Chong, T.S. Low, B.A. Cheng, J.P. Wang, Y.X. Jie, J. Laser Appl. 12, 54 (2000) J.N. Zeng, J.K. Low, Z.M. Ren, T. Liew, Y.F. Lu, Appl. Surf. Sci. 197–198, 362 (2002) J. Ye, S. Gu, S. Zhu, T. Chen, W. Liu, F. Qin, L. Hu, R. Zhang, Y. Shi, Y. Zheng, J. Vac. Sci. Technol. A 21, 979 (2003) J.D. Ye, S.L. Gu, S.M. Zhu, F. Qin, S.M. Liu, W. Liu, X. Zhou, L.Q. Hu, R. Zhang, Y. Shi, Y.D. Zheng, J. Appl. Phys. 96, 5308 (2004) C. Bundesmann, Ph.D. thesis, Universit¨at Leipzig, 2005 (Shaker, Aachen, 2006) C.X. Xu, X.Y. Sun, Jpn. J. Appl. Phys. 42, 4949 (2003) Y.J. Xing, Z.H. Xi, Z.Q. Xue, X.D. Zhang, J.H. Song, R.M. Wang, J. Xu, Y. Song, S.L. Zhang, D.P. Yu, Appl. Phys. Lett. 83, 1689 (2003) Y.K. Tseng, H.C. Hsu, W.F. Hsieh, K.S. Liu, I.C. Chen, J. Mater. Res. 18, 2837 (2003) H.Q. Ni, Y.F. Lu, Z.Y. Liu, H. Qiu, W.J. Wang, Z.M. Ren, S.K. Chow, Y.X. Jie, Appl. Phys. Lett. 79, 812 (2001) C. Bundesmann, N. Ashkenov, M. Schubert, D. Spemann, T. Butz, M. Lorenz, E.M. Kaidashev, M. Grundmann, Appl. Phys. Lett. 83, 1974 (2003) X. Wang, S. Yanga, J. Wanga, M. Lia, X. Jianga, G. Dua, X. Liub, R.P.H. Chang, J. Cryst. Growth 226, 123 (2001) A. Kaschner, U. Haboeck, M. Strassburg, M. Strassburg, G. Kaczmarczyk, A. Hoﬀmann, C. Thomsen, A. Zeuner, H.R. Alves, D.M. Hofmann, B.K. Meyer, Appl. Phys. Lett. 80, 1909 (2002) F. Reuss, C. Kirchner, T. Gruber, R. Kling, S. Maschek, W. Limmer, A. Waag, P. Ziemann, J. Appl. Phys. 95, 3385 (2004) M. Tzolov, N. Tzenov, D. Dimova-Malinovska, M. Kalitzova, C. Pizzuto, G. Vitali, G. Zollo, I. Ivanov, Thin Solid Films 379, 28 (2000) M. Tzolov, N. Tzenov, D. Dimova-Malinovska, M. Kalitzova, C. Pizzuto, G. Vitali, G. Zollo, I. Ivanov, Thin Solid Films 396, 274 (2001) Z. Zhaochun, H. Baibiao, Y. Yongqin, C. Deliang, Mater. Sci. Eng. B 86, 109 (2001) Z.Q. Chen, A. Kawasuso, Y. Xu, H. Naramoto, X.L. Yuan, T. Sekiguchi, R. Suzuki, T. Ohdaira, J. Appl. Phys. 97, 013528 (2005) Y.Q. Chang, D.B. Wang, X.H. Luo, X.Y. Xu, X.H. Chen, L. Li, C.P. Chen, R.M. Wang, J. Xu, D.P. Yu, Appl. Phys. Lett. 83, 4020 (2003) Y.Q. Chang, Y. Chen, D.P. Yu, Z.L. Fang, G.H. Li, F.H. Yang, Mater. Sci. Forum 475–479, 3525 (2005)

3 Optical Properties of ZnO and Related Compounds 58. 59. 60. 61. 62.

63. 64. 65. 66. 67. 68. 69. 70. 71.

72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85.

86. 87. 88.

121

J.B. Wang, H.M. Zong, Z.F. Li, W. Lu, J. Appl. Phys. 97, 086105 (2005) T.S. Jeong, M.S. Han, C.J. Youna, Y.S. Park, J. Appl. Phys. 96, 175 (2004) B. Cheng, Y. Xiao, G. Wu, L. Zhang, Appl. Phys. Lett. 84, 416 (2004) D. Zeng, C. Xie, B. Zhu, R. Jiang, X. Chen, W. Song, J. Wang, J. Shi, J. Cryst. Growth 266, 511 (2004) C. Bundesmann, M. Schubert, D. Spemann, T. Butz, M. Lorenz, E.M. Kaidashev, M. Grundmann, N. Ashkenov, H. Neumann, G. Wagner, Appl. Phys. Lett. 81, 2376 (2002) R.J. Collins, D.A. Kleinman, J. Phys. Chem. Solids 11, 190 (1959) E.C. Heltemes, H.L. Swinney, J. Appl. Phys. 38, 2387 (1967) E.F. Venger, A.V. Melnichuk, L.L. Melnichuk, Y.A. Pasechnik, Phys. Status Solidi B 188, 823 (1995) D.G. Thomas, J. Phys. Chem. Solids 10, 47 (1959) G. Heiland, H. L¨ uth, Solid State Commun. 5, 199 (1967) S. Perkowitz, R.K. Murty-Gutta, A.K. Garrison, Solid State Commun. 9, 2251 (1971) Z.C. Jin, I. Hamberg, C.G. Granquist, J. Appl. Phys. 64, 5117 (1988) A. Pﬂug, V. Sittinger, F. Ruske, B. Szyszka, G. Dittmar, Thin Solid Films 455–456, 201 (2004) C. Bundesmann, N. Ashkenov, M. Schubert, A. Rahm, H. v. Wenckstern, E.M. Kaidashev, M. Lorenz, M. Grundmann, Thin Solid Films 455–456, 161 (2004) C. Bundesmann, M. Schubert, D. Spemann, A. Rahm, H. Hochmuth, M. Lorenz, M. Grundmann, Appl. Phys. Lett. 85, 905 (2004) M. Schubert, Infrared Ellipsometry on Semiconductor Layer Structures: Phonons, Plasmons and Polaritons (Springer, New York, 2004) C. Bundesmann, A. Rahm, M. Lorenz, M. Grundmann, M. Schubert, J. Appl. Phys. 99, 113504 (2006) ¨ Ozg¨ ¨ ur, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Doˇgan, U. V. Avrutin, S.J. Cho, H. Morko¸c, J. Appl. Phys. 98, 041301 (2005) H. Yoshikawa, S. Adachi, Jpn. J. Appl. Phys. 36, 6237 (1997) G.E. Jellison, Thin Solid Films 313–314, 33 (1998) P.L. Washington, H.C. Ong, J.Y. Dai, R.P.H. Chang, Appl. Phys. Lett. 72, 3261 (1998) X.W. Sun, H.S. Kwok, J. Appl. Phys. 86, 408 (1999) K. Postava, H. Sueki, M. Aoyama, T. Yamaguchi, C. Ino, Y. Igasaki, M. Horie, J. Appl. Phys. 87, 7820 (2000) R. Matz, H. L¨ uth, Appl. Phys. 18, 123 (1979) T.D. Kang, H. Lee, W.I. Park, G.C. Yib, Thin Solid Films 455–456, 609 (2004) G.E. Jellison, C.M. Rouleau, Appl. Opt. 44, 3153 (2005) Y.C. Liu, J.H. Hsieh, S.K. Tung, Thin Solid Films 510, 32 (2006) R. Schmidt-Grund, N. Ashkenov, M. M. Schubert, W. Czakai, D. Faltermeier, G. Benndorf, H. Hochmuth, M. Lorenz, and M. Grundmann, AIP Conf. Proc. 893, 271 (2007). G.J. Exarhos, A. Rose, C.F.W. Jr, Thin Solid Films 308–309, 56 (1997) Y.R. Park, K.J. Kim, Solid State Commun. 123, 147 (2002) H.E. Rhaleb, A.E. Naciri, R. Dounia, L. Johann, A. Hakam, M. Addou, Thin Solid Films 455–456, 384 (2004)

122

C. Bundesmann et al.

89. A. Mendoza-Galvan, C. Trejo-Cruz, J. Lee, D. Bhattacharyya, J. Metson, P.J Evans, U. Pal, J. Appl. Phys. 99, 014306 (2006) 90. J.H. Kang, Y.R. Park, K.J. Kim, Solid State Commun. 115, 127 (2000) 91. R. Schmidt-Grund, M. Schubert, B. Rheinl¨ ander, D. Fritsch, H. Schmidt, E. Kaidashev, M. Lorenz, C. Herzinger, M. Grundmann, Thin Solid Films 455–456, 500 (2004) 92. K.J. Kim, Y.R. Park, J. Appl. Phys. 91, 1420 (2002) 93. K.J. Kim, Y.R. Park, J. Appl. Phys. 94, 867 (2003) 94. K.J. Kim, Y.R. Park, J. Appl. Phys. 96, 4150 (2004) ¨ Ozg¨ ¨ ur, M.J. Bergmann, H.O. Everitt, A.K. Sharma, 95. C.W. Teng, J.F. Muth, U. C. Jin, J. Narayan, Appl. Phys. Lett. 76, 979 (2000) 96. D. Zhao, Y. Liu, D. Shen, Y. Lu, J. Zhang, X. Fan, J. Cryst. Growth 234, 427 (2002) 97. F.K. Shan, B.I. Kim, G.X. Liu, Z.F. Liu, J.Y. Sohn, W.J. Lee, B.C. Shin, Y.S. Yu, J. Appl. Phys. 95, 4772 (2004) 98. I. Takeuchi, W. Yang, K.S. Chang, M.A. Aronova, T. Venkatesan, R.D. Vispute, L.A. Bendersky, J. Appl. Phys. 94, 7336 (2003) 99. T. Minemoto, T. Negami, S. Nishiwaki, H. Takakura, Y. Hamakawa, Thin Solid Films 372, 173 (2000) 100. J. Chen, W.Z. Shen, N.B. Chen, D.J. Qiu, H.Z. Wu, J. Phys. Condens. Matter 15, L475 (2003) 101. M. Wraback, H. Shen, S. Liang, C.R. Gorla, Y. Lu, Appl. Phys. Lett. 74, 507 (1999) 102. C.R. Gorla, N.W. Emanetoglu, S. Liang, W.E. Mayo, Y. Lu, M. Wraback, H. Shen, J. Appl. Phys. 85, 2595 (1999) 103. S. Muthukumar, J. Zhong, Y. Chen, Y. Lu, T. Siegrist, Appl. Phys. Lett. 82, 742 (2003) 104. B.P. Zhang, Y. Segawa, K. Wakatsuki, Y. Kashiwaba, K. Haga, Appl. Phys. Lett. 79, 3953 (2001) 105. E. Ruiz, S. Alvarez, P. Alemany, Phys. Rev. B 49, 7115 (1994) 106. K. Thoma, B. Dorner, G. Duesing, W. Wegener, Solid State Commun. 15, 1111 (1974) 107. A.W. Hewat, Solid State Commun. 8, 187 (1970) 108. A. Seko, F. Oba, A. Kuwabara, I. Tanaka, Phys. Rev. B 72, 024107 (2005) 109. A. Anderson (ed.), The Raman Eﬀect, Vol. 2: Applications (Marcel Dekker, New York, 1973) 110. R.H. Lyddane, R.G. Sachs, E. Teller, Phys. Rev. 59, 673 (1941) 111. C.M. Wolfe, N. Holonyak, G.E. Stillmann, Physical Properties of Semicoductors (Prentice Hall, New Jersey, 1989) 112. P. Yu, M. Cardona, Fundamentals of Semiconductors (Springer, Berlin, 1999) 113. S. Adachi, GaAs and Related Materials (World Scientiﬁc, New Jersey, 1994) 114. R.M.A. Azzam, N.M. Bashara, Ellipsometry and Polarized Light (NorthHolland, Amsterdam, 1977) 115. A. R¨ ossler, Infrared Spectroscopic Ellipsometry (Akademie-Verlag, Berlin, 1990) 116. H.G. Tompkins, W.A. McGahan, Spectroscopic Ellipsometry and Reﬂectometry: A User’s Guide (Wiley, New York, 1999) 117. H.G. Tompkins, E.A. Irene (eds.), Handbook of Ellipsometry (William Andrew Publishing, Highland Mills, 2004)

3 Optical Properties of ZnO and Related Compounds

123

118. H. Fujiwara, Spectroscopic Ellipsometry: Principles and Applications (John Wiley & Sons Ltd, Chichester, 2007) 119. M. Schubert, Phys. Rev. B 53, 4265 (1996) 120. M. Schubert, T. Hofmann, C.M. Herzinger, J. Opt. Soc. Am. A 20, 347 (2003) 121. G.E. Jellison, L.A. Boatner, Phys. Rev. B 58, 3586 (1998) 122. M. Rebien, W. Henrion, M. B¨ ar, C.H. Fischer, Appl. Phys. Lett. 80, 3518 (2002) 123. M. Schubert, T.E. Tiwald, C.M. Herzinger, Phys. Rev. B 61, 8187 (2000) 124. W.L. Bond, J. Appl. Phys. 36, 1674 (1965) 125. S.P.S. Porto, R.S. Kishnan, J. Chem. Phys. 47, 1009 (1967) 126. G.H. Watson, W.B. Daniels, C.S. Wang, J. Appl. Phys. 52, 956 (1981) 127. D.G. Mead, G.R. Wilkinson, J. Raman Spectrosc. 6, 123 (1977) 128. J.B. Cui, K. Amtmann, J. Ristein, L. Ley, J. App. Phys. 83, 7929 (1998) 129. L. Vi˜ na, S. Logothetidis, M. Cardona, Phys. Rev. B 30, 1979 (1984) 130. S.S. Mitra, O. Brafman, W.B. Daniels, R.K. Crawford, Phys. Rev. 186, 942 (1969) 131. F.J. Manj´ on, B. Mar´ı, J. Serrano, A.H. Romero, J. App. Phys. 97, 053516 (2005) 132. J. Chen, W.Z. Shen, Appl. Phys. Lett. 83, 2154 (2003) 133. J.R. Jasperse, A. Kahan, J.N. Plendl, S.S. Mitra, Phys. Rev. 146, 516 (1966) 134. J. Serrano, A.H. Romero, F.J. Manj´on, R. Lauck, M. Cardona, A. Rubio, Phys. Rev. B 69, 094306 (2004) 135. E. Mollwo, R. Till, Zeitschrift f¨ ur Physik 216, 315 (1968) 136. R. Synowicki, T.E. Tiwald, Thin Solid Films 455–456, 248 (2004) 137. E. Mollwo, Z. Angew. Phys. 6, 257 (1954) 138. W.S. Hu, Z.G. Liu, J. Sun, S.N. Zhu, Q.Q. Xu, D. Feng, Z.M. Ji, J. Phys. Chem. Solids 58, 853 (1997) 139. N.B. Chen, H.Z. Wu, T.N. Xu, J. Appl. Phys. 97, 023515 (2005) 140. E.D. Palik (ed.), Handbook of Optical Constants of Solids, vol. 3 (Academic Press, New York, 1998) 141. C.F. Klingshirn, Semiconductor Optics (Springer, Berlin, 1997) 142. J.F. Muth, R.M. Kolbas, A.K. Sharma, S. Oktyabrsky, J. Narayan, J. Appl. Phys. 85, 7884 (1999) 143. A. Djurisic, Y. Chan, E. Li, Appl. Phys. A 76, 37 (2003) 144. S. Ozaki, T. Mishima, S. Adachi, Jpn. J. Appl. Phys. 42, 5465 (2003) 145. V. Srikant, D.R. Clarke, J. Appl. Phys. 81, 6357 (1997) 146. S.A. Studenikin, N. Golego, M. Cocivera, J. Appl. Phys. 83, 2104 (1998) 147. F.D. Paraguay, W.L. Estrada, D.R.N. Acosta, E. Andrade, M. Miki-Yoshida, Thin Solid Films 350, 192 (1999) 148. X.Q. Meng, W. Zhen, J.P. Guo, X.J. Fan, Appl. Phys. A 70, 421 (2000) 149. G. Santana, A. Morales-Acevedo, O. Vigil, L. Vaillant, F. Cruz, G. ContrerasPuente, Thin Solid Films 373, 235 (2000) 150. M.S. Tokumoto, A. Smith, C.V. Santilli, S.H. Pulcinelli, A.F. Craievich, E. Elkaim, A. Traverse, V. Briois, Thin Solid Films 416, 284 (2002) 151. F.K. Shan, Y.S. Yu, Thin Solid Films 435, 174 (2003) 152. P. Misra, P. Sahoo, P. Tripathi, V. Kulkarni, R. Nandedkar, L. Kukreja, Appl. Phys. A 78, 37 (2004) 153. J.L. Zhao, X.M. Li, J.M. Bian, W.D. Yu, X.D. Gao, J. Cryst. Growth 276, 507 (2005)

124

C. Bundesmann et al.

154. B. Segall, Phys. Rev. 163, 1674 (1967) 155. S.L. Chuang, C.S. Chang, Phys. Rev. B 54, 4005 (1970) 156. T. Kawashima, H. Yoshikawa, S. Adachi, S. Fuke, K. Ohtsuka, J. Appl. Phys. 82, 3528 (1997) 157. D.C. Reynolds, D.C. Look, B. Jogai, C.W. Litton, T.C. Collins, M.T. Harris, M.J. Callahan, J.S. Bailey, J. Appl. Phys. 86, 5598 (1999) 158. B. Gil, Phys. Rev. B 64, R201310 (2001) 159. J.J. Hopeﬁeld, J. Phys. Chem. Solids 15, 97 (1960) 160. L.C. Lew Yan Voon, M. Willatzen, M. Cardona, N.E. Christensen, Phys. Rev. B 53, 10703 (1996) 161. W.Y. Liang, A.D. Yoﬀe, Phys. Rev. Lett. 20, 59 (1968) 162. D.W. Langer, R.N. Euwema, K. Era, T. Koda, Phys. Rev. B 2, 4005 (1970) 163. D.C. Reynolds, D.C. Look, B. Jogai, C.W. Litton, G. Cantwell, W.C. Harsch, Phys. Rev. B 60, 2340 (1999) 164. A. Mang, K. Reimann, S. R¨ ubenacke, Solid State Commun. 94, 251 (1995) 165. Y.P. Varshni, Physica 34, 149 (1967) 166. K.P. O’Donnel, X. Chen, Appl. Phys. Lett. 58, 2924 (1991) 167. C. Boemare, T. Monterio, M.J. Soares, J.G. Guilherme, E. Alves, Physica B 308–310, 985 (2001) 168. L. Wang, N.C. Giles, J. Appl. Phys. 94, 973 (2003) 169. R. P¨ assler, J. Appl. Phys. 89, 6235 (2001) 170. A. Manoogian, A. Leclerc, Can. J. Phys. 57, 1766 (1979) 171. J.M. Zhang, T. Ruf, R. Lauck, M. Cardona, Phys. Rev. B 57, 9716 (1998) 172. A. G¨ obel, T. Ruf, J.M. Zhang, R. Lauck, M. Cardona, Phys. Rev. B 59, 2749 (1998) 173. R. P¨ assler, Phys. Status Solidi B 216, 975 (1999) 174. S.V. Mel’nichuk, V.I. Sokolov, T.P. Surkova, V.M. Chernov, Sov. Phys. Solid State 33, 1833 (1991) 175. S. Serrano, F. Widulle, A.H. Romero, A. Rubio, R. Lauck, M. Cardona, Phys. Status Solidi B 235, 260 (2003) 176. R. P¨ assler, E. Griebl, H. Riepl, G. Lauther, S. Bauer, H. Preis, W. Gebhardt, B. Buda, D.J. As, D. Schikora, K. Lischka, K. Papagelis, S. Ves, J. Appl. Phys. 86, 4403 (1999) 177. P. Lautenschlager, M. Garriga, S. Logothetidis, M. Cardona, Phys. Rev. B 35, 9174 (1987) 178. H. Watanabe, M. Wada, T. Takahashi, Jpn. J. Appl. Phys. 3, 617 (1964) 179. J.E. Jaﬀe, R. Pandey, A.B. Kunz, Phys. Rev. B 43, 14030 (1991) 180. G.E. Jellison, Phys. Rev. B 65, 049902 (2001)

4 Surfaces and Interfaces of Sputter-Deposited ZnO Films A. Klein and F. S¨ auberlich

In this chapter chemical and electronic surface and interface properties of magnetron sputtered ZnO ﬁlms as determined from photoelectron spectroscopy are described. We particularly focus on interfaces that are important for Cu(In,Ga)Se2 thin ﬁlm solar cells. The use of in situ sample preparation utilizing integrated vacuum systems allows for systematic studies widely ruling out the inﬂuence of adsorbates on the electronic structure. No evidence for band bending at the surface is observed indicating the absence of surface states in the band gap. The ionization potential varies with deposition conditions, which can be attributed to changes in crystallographic orientation of the ﬁlms resulting in diﬀerent surface terminations. A variation of the Zn 2p and O 1s core level binding energies with respect to the valence band maximum is attributed to local disorder in ﬁlms deposited from ceramic ZnO targets at room temperature without addition of oxygen to the sputter gas. In the interface studies, no eﬀects attributable to sputter damage could be identiﬁed. The surface and interface chemistry of sputtered ZnO ﬁlms is rather governed by the ability of the surfaces to dissociate the oxygen molecules condensing from the gas phase. Poor oxygen dissociation on ZnO and CdS leads to the presence of peroxides on the ZnO surfaces and to a nonreactive interface between CdS and ZnO. Noticeable interface reactions are observed between Cu(In,Ga)Se2 or In2 S3 and ZnO, where no peroxide species is observed during initial ZnO growth. The variation of the corelevel binding energies with respect to the valence band maximum leads to an apparent variation of the band alignment at the CdS/ZnO1 interface with deposition conditions. The band alignment is strongly inﬂuenced by Fermi level pinning, which is particularly pronounced at In2 S3 /ZnO interfaces. The observed variation of band alignment at these interfaces can be related to observed eﬃciencies of Cu(In,Ga)Se2 solar cells using In2 S3 buﬀer layers. An amorphous nucleation layer of 2–3 nm thickness is observed for sputter deposition of ZnO ﬁlms on all investigated substrates. The amorphous layer leads to a modiﬁcation of the band alignment at the CdS/ZnO interface compared with the reverse deposition sequence (ZnO/CdS).

1

We generally use the notation substrate/overlayer throughout this chapter.

126

A. Klein and F. S¨ auberlich

4.1 Introduction 4.1.1 Semiconductor Interfaces Transparent conducting oxides are widely used as electrodes in thin ﬁlm optoelectronic devices as solar cells and light emitting diodes because of their transparency for visible light and their high electrical conductivity. Highest optical transparency and electrical conductivity are thus key aspects for such applications. Most work on TCO electrodes is, therefore, dedicated to ﬁnd deposition parameters, which improve these material parameters. In addition, contact properties are essential for the application of TCOs as electrodes. To elucidate the contact properties, the energy band diagrams of the devices have to be considered. Figure 4.1 shows energy band diagrams of two basic semiconductor contacts, a semiconductor/metal contact and a semiconductor p/n- heterocontact. Basic interface parameters are the Schottky barrier height ΦB = EF − EVB for p-type and ΦB = ECB − EF for n-type semiconductors, and the valence and conduction band oﬀsets ∆EVB and ∆ECB . The potential distribution across an interface is determined by the barrier heights and by the doping proﬁle. As doping of semiconductors is typically well controlled [1], the energy band diagrams of semiconductor contacts, which determines the function of the device, can usually be well predicted and modiﬁed if the barrier heights are known. Extensive research has been devoted in the past to understand the mechanisms governing barrier formation at semiconductor interfaces to enable prediction and possible modiﬁcation of barrier heights [2–8]. An essential feature of semiconductor interfaces is Fermi level pinning, which leads to barrier heights at semiconductor/metal contacts being almost independent on the metal. Fermi level pinning is especially pronounced for semiconductors with covalent bonding such as Si, Ge, and GaAs. It occurs even for defect-free semiconductor

ΦB

ECB Eg EVB

metal

semiconductor

Eg,1

semiconductor

∆ECB

EF

∆EVB

Eg,2

Fig. 4.1. Example energy band diagrams for a semiconductor/metal contact and and a semiconductor p/n-heterocontact. The Schottky barrier height for electrons ΦB,n is given by the energy diﬀerence of the conduction band minimum ECB and the Fermi energy EF . The valence and conduction band oﬀsets ∆EVB and ∆ECB are given by the discontinuities in the valence band maximum EVB and the conduction band minimum, respectively

4 ZnO Surfaces and Interfaces

127

atomically abrupt junctions and is explained by (metal) induced gap states, which have ﬁrst been introduced by Heine [9]. They are the result of wave function matching at the interface and are clearly illustrated by electronic structure calculations [10]. Fermi level pinning is reported to be less strong for polar-bonded materials as oxides [2, 11]. However, the experimental basis concerning barrier heights at oxide semiconductor interfaces is rather limited and is mostly restricted to electrical measurements. Electrical determinations of barrier heights on ZnO are, e.g., presented in Chap. 7 of this book. Other reviews are given in [12, 13]. The issue of Schottky barrier formation to ZnO is not treated in this chapter as such contacts are not of big importance in thin-ﬁlm solar cells. This is related to the fact that in thin ﬁlm solar cells metals are only used to contact highly-doped ﬁlms. For degenerately doped semiconductors, the barrier heights become very small because of the large space charge associated with depletion layers in such materials. 4.1.2 ZnO in Thin-Film Solar Cells Compared with the interfaces illustrated in Fig. 4.1, the role played by ZnO in a Cu(In,Ga)Se2 thin-ﬁlm solar cell is considerably more complex (see also Chap. 9). The basic structure and the commonly used energy band diagram are shown in Fig. 4.2 [14]. It is evident that ZnO does not only provide a transparent contact, but is also an essential part of the p/n-junction of the device. Best solar cells with eﬃciencies of almost 20 % are achieved if a ZnO bilayer is used, which consists of a combination of a nominally undoped ZnO and a highly doped ZnO layer [15]. ZnO is deposited onto a CdS layer, which is prepared mostly by a chemical bath. The lattice constants of CdS are signiﬁcantly larger than those of ZnO leading to a strongly lattice mismatched system. At such an interface, crystallographic defects are unavoidable. The electronic defect states associated with such defects typically have energy positions within the band gap of the smaller gap semiconductor and Mo

solar illumination

ZnO CdS CIGS

ΦB

CdS ZnO i n+ ∆ECB EF

p-CIGS ECB

EVB ∆EVB

Mo glass

∆EVB

Fig. 4.2. Structure and energy band diagram of a Cu(In,Ga)Se2 (CIGS) thin-ﬁlm solar cell. The ZnO window layer typically consists of a combination of a nominally undoped ZnO and a highly doped ZnO layer

128

A. Klein and F. S¨ auberlich

can be electrically charged. They might act as carrier recombination centers but will also modify the electrical potential distribution. Another signiﬁcant diﬀerence of the interface of ZnO in the Cu(In,Ga)Se2 solar cell compared with conventional semiconductor systems lies in the fact that the ZnO layer is typically prepared by magnetron sputtering. This is partially dictated by technological requirements (fast and large area coating) but also due to advantages of the technique that allows to prepare highly doped ﬁlms at low substrate temperature (see Chaps. 2 and 5). The latter issues are particularly related to the energetic particles involved in sputter deposition. However, it is generally believed that these energetic particles lead to a damage of the substrate surface by introducing defects. Avoiding such a sputter damage of the Cu(In,Ga)Se2 semiconductor might, e.g., explain why a CdS buﬀer layer is necessary to achieve highest conversion eﬃciencies. However, no clear evidence for the inﬂuence of sputter damage has been presented so far. A less complex situation is present if ZnO ﬁlms are used as substrates for amorphous Si thin-ﬁlm solar cells. These are described in chapters 8 and 6 of this book. In this case the problem of possible sputter damage and, due to the amorphous nature of the semiconductor, also of lattice mismatch are not an issue. For thin ﬁlm silicon solar cells, light trapping is of particular interest. This requires a dedicated surface morphology, which can be either achieved directly by depositing the ZnO layer using chemical vapor deposition (see Chap. 6) or by a suitable etching process of sputter deposited ﬁlms as described in Chap. 8 of this book. The morphology obtained by etching is strongly dependent on the sputter deposition parameters. It is well known that etching of semiconductors is highly anisotropic and depends on surface orientation and termination [16, 17]. Thus, the surface properties of sputter deposited ZnO ﬁlms are also of importance for thin ﬁlm silicon solar cell devices. 4.1.3 Photoelectron Spectroscopy (PES) Electronic properties of semiconductor surfaces and interfaces can be probed by diﬀerent techniques, including scanning probe techniques [18, 19], Kelvin probe [20] and photoelectric yield [21] measurements, electrical techniques such as current–voltage and capacitance–voltage measurements [1, 12, 13, 22], DLTS2 and admittance spectroscopy [23–27], internal photoemission [8, 22, 28], cathodoluminescence [29, 30], and others. A versatile tool that has contributed signiﬁcantly to the understanding of semiconductor interfaces is X-ray and ultraviolet photoelectron spectroscopy (XPS and UPS). Both most frequently used (standard) applications of photoelectron spectroscopy are elemental and chemical surface analysis [31, 32]. Binding energies in photoelectron spectroscopy (PES) are measured with respect to the Fermi level. Since samples are electrically connected to the spectrometer 2

Deep level transient spectroscopy.

4 ZnO Surfaces and Interfaces

129

system, the Fermi level is at a constant energy, which can be determined by a calibration measurement using a metallic sample. This allows to observe changes of the Fermi level within the band gap. If, for a given sample, the binding energy of a core level (CL) is known with respect to the valence band maximum (BEVB (CL)), the core-level binding energy itself can be directly used as a measure for the position of the Fermi level in the band gap (EF − EVB = BE(VB)). During an interface experiment, an overlayer is stepwise deposited onto a substrate. By monitoring the substrate and overlayer core-level binding energies during deposition, the evolution of the valence band maxima of the substrate and of the overlayer can be followed during interface formation [33–36]. The procedure is outlined in Fig. 4.3. As will be shown in Sects. 4.2.3.3 and 4.3.3, care has to be taken when applying this standard procedure to study surfaces and interfaces of sputter-deposited ZnO ﬁlms, as BEVB (CL) depends on the deposition parameters for this material. Photoelectron spectroscopy is a highly surface sensitive technique because of the inelastic mean free path of the photoelectrons λe , which depends on the electron kinetic energy Ekin and has typical values of 0.2–3 nm [31,37,38]. Determination of Schottky barrier heights ΦB , or valence band discontinuities ∆EVB , can be performed by following the evolution of the position of the valence band maxima with respect to the Fermi level of substrate and overlayer with increasing thickness of the overlayer. For layer-by-layer growth the attenuation of the substrate intensities is given by the inelastic

~ ~

energy band diagram substrate layer ∆ECB

ECB

∆EVB

BEVB (sub)

BEVB(sub)

EVB

eVb

EF

BEVB(layer) ∆ECL

~ ~

~ ~ 1020

eVb ~ ~

eVb

binding energy [eV]

intensity [arb. units]

Eg(layer)

∆ECL

1024

0

EF

Eg(sub)

~ ~

~ ~

layer (ZnO) substrate (CdS) Zn 2p3/2 Cd 3d5/2 valence bands core levels ∆EVB BEVB (layer)

406 404 6 binding energy [eV]

4

2

0

BE(CL)

Fig. 4.3. Experimental procedure for the determination of the band alignment using photoelectron spectroscopy

130

A. Klein and F. S¨ auberlich

mean free path of the photoelectrons. Hence, after a ﬁlm thickness of ∼5 nm, the substrate emissions are completely extinguished. The study of interface formation with PES thus requires control of the ﬁlm thickness in the subnanometer range. The rapid contamination of surfaces in air requires that the sample transfer between surface preparation or ﬁlm deposition to the photoelectron spectrometer is performed in vacuum. Consequently, the analysis system has to integrate all required thin ﬁlm preparation chambers and a surface analysis tool. The layout of such a system as it is used in Darmstadt is shown in Fig. 4.4. Experimental determinations of barrier heights on oxide semiconductor interfaces using photoelectron spectroscopy are rarely found in literature and no systematic data on interface chemistry and barrier formation on any oxide are available. So far, most of the semiconductor interface studies by photoelectron spectroscopy deal with interfaces with well-deﬁned substrate surfaces and ﬁlm structures. Mostly single crystal substrates and, in the case of semiconductor heterojunctions, lattice matched interfaces are investigated. Furthermore, highly controllable deposition techniques (typically molecular beam epitaxy) are applied, which lead to ﬁlms and interfaces with well-known structure and composition. The results described in the following therefore, for the ﬁrst time, provide information about interfaces with oxide semiconductors and about interfaces with sputter-deposited materials. Despite the rather complex situation, photoelectron spectroscopy studies of sputter-deposited ion source

UV source

x-ray source with monochromator

electron analysor

surface analysis sample preparation

load lock sputter deposition and MBE of oxides

manipulator measurement system central sample handling system

LEED

surface cleaning with ion etching CVD and MBE deposition

MBE of CdS, CdTe and CdCl2 CSS CdTe deposition UHV CdCl2 activation

electrochemistry

Fig. 4.4. Layout of the integrated surface analysis and preparation system DAISYMAT (Darmstadt integrated system for materials research). A photoelectron spectrometer is connected by a sample handling system to various deposition and surface treatment chambers. Preparation and analysis can be repeatedly performed under controlled ultrahigh vacuum conditions

4 ZnO Surfaces and Interfaces

131

ZnO ﬁlms can provide substantial information on chemical and electronic properties of ZnO surfaces and interfaces, which occur in real thin ﬁlm solar cell structures. In addition, general information on the interface formation of oxide materials can be extracted. In the following we describe: – Electronic surface properties including Fermi level positions, work functions, and ionization potentials of sputter-deposited ZnO and Al-doped ZnO ﬁlms in dependence on deposition parameters. The results provide insight into aspects of doping, surface chemistry, and terminations. – Interfaces of sputter-deposited ZnO and ZnO:Al ﬁlms with diﬀerent substrate materials (CdS, In2 S3 , and Cu(In,Ga)Se2 ) in dependence on deposition parameters. The band alignments and chemical interactions at the interfaces are discussed.

4.2 Surface Properties of ZnO 4.2.1 Crystallographic Structure of ZnO Surfaces The wurtzite lattice of ZnO and its low-index surfaces are shown in Fig. 4.5. The basic low index surface terminations are (0001), (000¯1), (10¯10), and (11¯ 20). The (0001) and (000¯ 1) represent the zinc and oxygen-terminated surfaces of the polar {0001} direction, which corresponds to the {111} direction of the cubic zincblende lattice. In contrast to the polar (100) and (¯100) surfaces of the zincblende lattice, the surfaces with threefold symmetry (111) c (0001)

(0001) b

zinc oxygen

a

(1120)

(1010) Fig. 4.5. Crystallographic structure of ZnO and its basic surfaces

132

A. Klein and F. S¨ auberlich

and (¯ 1¯ 1¯ 1) are not equivalent. There is thus a clear distinction between the two surfaces. The same is true for the basal plane (0001) and (000¯1) surfaces of the wurtzite lattice. These are important surface terminations for sputtered ZnO thin ﬁlms, which often grow with a (0001) texture (see e.g., [39] and other contributions to this book). The (10¯ 10) and (11¯20) surfaces are nonpolar, i.e., they contain the same number of zinc and oxygen atoms. These surfaces show no reconstruction [40, 41]. The wurtzite (11¯ 20) surface directly corresponds to the well-known nonpolar (110) surface of the zincblende lattice [42]. Surfaces of real crystals never adopt the bulk-truncated structures shown in Fig. 4.5. They reconstruct or relax (inwards or outward movement of the atoms) to minimize their surface energy [42]. Known surface structures of zincblende and wurtzite structure semiconductors are summarized in [43]. Nonpolar surfaces of wurtzite (11¯ 20) and (10¯10) surfaces show no lateral surface reconstructions and are supposed to have a structure similar to the well-known zincblende (110) surface, which is characterized by an inward relaxation of the surface cations and partial electron transfer from the surface cation dangling bond to the surface anion dangling bond [42, 43]. Polar surfaces of semiconductors cannot be bulk-truncated because of the alternating charge of atomic planes in polar directions, which will lead to a diverging electrostatic potential due to the large number of lattice planes [44]. Compensation of the diverging electrostatic potential is possible by a rearrangement of charges at the surface. For the basal (0001) and (000¯1) surfaces of ZnO this corresponds to a removal of ∼1/4 of the surface charge.3 In principle, the rearrangement of the charge can be achieved by: (1) creation of a metallic surface by introduction of surface states; (2) removal of surface atoms; and (3) charged impurities at the surface, as e.g. hydroxyl or hydroxide species. All mechanisms have been invoked for the basal surfaces of ZnO [45–49]. The uncertainty concerning the identiﬁcation of the stabilization mechanism on polar ZnO surfaces is partly due to the lack of atomically resolved STM images. Such images are possible for the nonpolar (10¯10) and (11¯20) surfaces [40, 41] but have, to our knowledge, not been reported for polar surfaces. The polar cation terminated (111) surface of zincblende compounds typically displays a 2 × 2 reconstruction associated with removal of every fourth surface cation [43,50–52]. This structure is ideally suited to match the charging condition for surface stabilization for this particular surface orientation. The 2 × 2 reconstruction and the missing surface atoms can directly be observed by STM [52]. In contrast to literature [53], a 2 × 2 reconstruction is also frequently observed in our group for the (0001) surface of wurtzite CdS.4 The reconstruction on the anion terminated (¯1¯1¯1) surfaces of III–V and II–VI zincblende compounds are considerably more complex. These surfaces 3

4

The factor of 1/4 results from the distance of the Zn and O lattice planes along the c-direction. B. Siepchen et al., Darmstadt University of Technology (unpublished results).

4 ZnO Surfaces and Interfaces

133

show a strong tendency for facet formation [43]. Nonfacetted GaAs(¯1¯1¯1) show complex superstructures associated with adsorbed As species [54]. In contrast to most of the zincblende and wurtzite materials, the ZnO (0001) and (000¯ 1) surfaces are usually not reconstructed [40,47]. Only Kunat et al. report the observation of a 1 × 3 reconstruction for the ZnO (000¯1) surface and attribute the unreconstructed surface to a hydrogen termination [49]. A new stabilization mechanism for polar ZnO (0001) surfaces has been introduced by Dulub et al. based on scanning tunneling microscopy [41, 45, 46]. This involves a large number of oxygen-terminated step edges on the Zn-terminated surface. This termination allows for ﬁeld compensation and is consistent with an unreconstructed surface and with the predominant Zn surface atoms found in low energy ion scattering (LEIS) [40]. 4.2.2 Chemical Surface Composition of Sputtered ZnO Films Examples of XPS spectra recorded from in situ sputter-deposited ZnO and ZnO:Al ﬁlms are shown in Fig. 4.6. Absence of contaminations is evident from the survey spectra, which show only emissions from Zn and O. Aluminiumdoped ZnO ﬁlms, deposited from a target with a nominal Al concentration of 2 wt %, also show a small Al signal. Detailed spectra of the Zn 2p3/2 , O 1s, Zn LMM Auger, and valence band emissions for largely diﬀerent deposition conditions are given in the lower half of Fig. 4.6. Despite the strongly varying deposition conditions of the ﬁlms shown in Fig. 4.6, all spectra represent the ZnO composition. Binding energy variations are largely due to changes of the Fermi level position (see Sect. 4.2.3.1). The O 1s and the Zn LMM Auger line show a noticeably diﬀerent shape for the most highly doped ﬁlms, which are those deposited from ZnO:Al targets without addition of oxygen to the sputter gas. The high doping level is reﬂected by the highest binding energy of the corresponding spectra. The diﬀerent shapes of the Zn LMM line can be explained by the changes in electron concentration: The electron gas in highly degenerate TCOs leads to screening of the core hole and to inelastic scattering of the photoelectron because of excitation of plasmons (see Fig. 4.17) [55–58]. The latter leads to shoulders on the high binding energy side of the core levels and to a second Auger emission, which is shifted to lower binding energies [55]. This explains the typical “smeared” appearance of the Zn LMM Auger spectra of the ﬁlm with the highest carrier concentration (spectra (d) in Fig. 4.6). 4.2.2.1 Oxygen and Aluminium Content The chemical composition of the surface can be evaluated from the integrated intensities of the core-level emissions [31]. Unfortunately, the accuracy of quantitative analysis with photoelectron spectroscopy is generally limited to a few percent even when good standards are available [31]. Therefore, even from a large number of measured samples it was not possible to observe a

134

A. Klein and F. S¨ auberlich

Al 2p Zn 2p

ZnO:Al

1400

1200

1000

800

Zn 2p

600

Zn 3s Zn 3p Zn 3d

O 1s

O KLL

ZnO

Zn LMM

Zn 2s

normalized intensity

Survey

400

200

0

Zn LMM

O 1s

VB

normalized intensity

f f

f

f e

e

e

e

d d

d

c

c

d

c c b b

b

b a

a 1025

a 1020

a

535

530

500

490

5

0

binding energy [eV] Fig. 4.6. X-ray photoelectron spectra of undoped ZnO (a–c) and of Al-doped ZnO (d–f ) prepared by magnetron sputtering. The spectra are excited with monochromatic Al Kα radiation (hν = 1486.6 eV). ZnO:Al ﬁlms are prepared from a target containing 2 wt % Al. The ﬁlms are prepared with 100 % Argon as sputter gas either at room temperature (a and d) or at a substrate temperature of 400◦ C (b and e). Spectra (c) and (f ) are recorded from ﬁlms deposited onto samples held at room temperature in a sputter gas mixture of 50 % argon and 50 % oxygen

systematic variation of the oxygen to zinc intensity ratio in dependence on deposition conditions. The oxygen content in ﬁlms deposited from undoped ZnO targets varies between 46 and 49 %. Sensitivity factors supplied by the manufacturer of the XPS system5 are used for quantiﬁcation. Nonappropriate 5

Physical Electronics: PHI 5700.

4 ZnO Surfaces and Interfaces

Al content [at-%]

12

135

calculation of the composition with

10

without

high binding energy component of the O1s emission lines

8 6 4 2 0

10

20

30

40

50

100

[O2 ]/([Ar]+[O 2 ]) [%]

300

500

temperature [°C]

Fig. 4.7. Aluminium content of sputter deposited ZnO:Al ﬁlms. A target with a nominal Al content of 2 wt % has been used. The shadowed regions indicate the general behavior. The atomic concentration is calculated with and without considering the high binding energy oxygen species, which contributes to the O 1s signal (see Sect. 4.2.2.2)

sensitivity factors are most likely the reason for the deviation of the determined composition from the nominal oxygen content of 50 %. Even larger deviations are observed for other oxides. It is thus not possible to derive an absolute number for the composition of the ﬁlms. A better reproducibility and quantiﬁcation of concentrations with XPS is possible for the relative cation concentrations of mixed cation systems as e.g. (Zn,Al)O, (Zn,Mg)O [59], and (In,Sn)O (ITO) [58]. Figure 4.7 shows the variation of the Al content of ZnO:Al ﬁlms deposited with diﬀerent Ar/O2 ratios in the sputter gas and at diﬀerent substrate temperatures. There is only a small or negligible dependence of the Al content if oxygen is added to the sputter gas. In contrast, a strong enrichment of Al is obtained at higher substrate temperature. This is explained by the high vapor pressure of Zn, which reevaporates from the surface at higher substrate temperatures before a ZnO compound can form. A comparable behavior has been observed for (Zn,Mg)O ﬁlms, where also a strong enrichment of Mg is observed at higher substrate temperatures [59]. 4.2.2.2 Oxygen-Related Surface Species and Initial Growth of ZnO Films The O 1s spectra in Fig. 4.6 show an additional small emission at higher binding energies. The energy diﬀerence between the ZnO-related emission at 530–531 eV and the high binding energy component amounts to 1.6–1.8 eV. Such a species is always observed on ZnO surfaces. In literature, it is mostly attributed to adsorbed species. These include water and hydroxides

136

A. Klein and F. S¨ auberlich

[49, 60–62], physisorbed or chemisorbed oxygen [63–66], and COx species (see [67] and references therein). The latter can be excluded as an explanation for the species observed here because of the absence of carbon. Chen et al. also mention a possible contribution of oxygen vacancies [68] and Stucki et al. suggest that the high binding energy oxygen species is related to oxygen interstitials [69]. ZnO is rather hygroscopic. It is, therefore, reasonable to expect hydroxide at the surfaces of samples that have been exposed to air. Even for in situ sputter-deposited ﬁlms, hydroxide species cannot be excluded since water and hydrogen is always present in the residual gas, in the ceramic ZnO target, and/or the sputter gas. However, a high binding energy O 1s component is also observed at UHV-cleaned single crystal ZnO surfaces [49] and is still present after heating in UHV to >500◦C [49, 70]. The high binding energy O 1s component of in situ sputter-deposited ZnO ﬁlms is related to a surface species. Photoelectron spectra recorded with diﬀerent photon energies at the synchrotron are shown in Fig. 4.8. Because of the diﬀerent photon energies, the photoelectrons have diﬀerent kinetic energies, which results in diﬀerent photoelectron escape depths. The highest surface sensitivity is obtained for a photoelectron kinetic energy of ∼50 eV, which is obtained for an excitation energy of 580 eV. With increasing surface

(b)

(a) 580 eV

0.8 600 eV

0.6 0.4

700 eV

relative intensity

normalized intensity

1.0

0.2 970 eV 5

0

rel. binding energy [eV]

0

500

1000

kinetic energy [eV]

Fig. 4.8. (a) O 1s core-level spectra of a sputter-deposited undoped ZnO ﬁlm. The spectra were recorded with diﬀerent photon energies at the synchrotron. Binding energies are given with respect to the ZnO component. The relative intensity of the high binding energy component is shown in (b). The dashed line represents the calculated dependency for a homogeneous surface layer using energy-dependent inelastic mean paths as provided by Tanuma, Powell, and Penn [37]. Using their material parameters for ZnS, the ﬁt to the experimental data reveals a thickness of the surface layer of 3.3 ± 0.1 ˚ A

4 ZnO Surfaces and Interfaces

137

sensitivity, the contribution of the high binding energy component to the total O 1s intensity increases. The intensity ratio of the high binding energy to the ZnO component of the O 1s level is shown in Fig. 4.8b. The dependence on electron kinetic energy is reasonably well reproduced by modeling the intensity ratio using a homogeneous surface layer with a diﬀerent oxygen species. The formula for the inelastic mean free path proposed by Tanuma, Powell, and Penn with the parameters for ZnS [37] has been used to calculate the curve. Fitting the model for the intensity to the experimental data reveals a thickness of the surface layer of 3.3±0.1 ˚ A. The diﬀerent oxygen component is, therefore, located mostly in the topmost surface layer. Observations made during the deposition of ZnO on diﬀerent substrate surfaces support this model and provide arguments for the chemical identiﬁcation of the surface oxygen species. This is illustrated in Fig. 4.9. The spectra show the evolution of the O 1s signal during stepwise deposition of ZnO or ZnO:Al onto CdS, In2 S3 and Cu(In,Ga)Se2 by dc magnetron sputtering. The growth is interrupted several times in order to follow the changes in the spectra with ZnO thickness. During deposition onto CdS, the high binding energy component dominates at low coverage over the ZnO component. It is reduced during further deposition until it reaches the intensity typically observed for thick ZnO ﬁlms. Although the O 1s spectra of ZnO and ZnO:Al are diﬀerent due to the diﬀerent free carrier concentrations, the behavior is

(b)

(c)

(d)

intensity [arb. units]

ZnO deposition time

(a)

534

530

534

530 534 binding energy [eV]

530

534

530

Fig. 4.9. O 1s spectra recorded during stepwise deposition of ZnO (a and b) or ZnO:Al (c and d) onto diﬀerent substrates. The surface species dominates at low coverage for CdS substrates (a and c). For Cu(In,Ga)Se2 (b) or In2 S3 (d) substrates the surface species occurs not until a thick ZnO ﬁlm is deposited

138

A. Klein and F. S¨ auberlich

the same for both interfaces. Also alloying of Mg with ZnO results in an identical evolution of the O 1s spectra [71]. A diﬀerent behavior is observed during deposition onto In2 S3 or CIGS (see Fig. 4.9c,d). At low coverage the high binding energy component is not present, although the deposition conditions are otherwise identical. The substrate, therefore, determines whether the high binding energy component is observed or not at low coverage. It is thus not likely that the surface component is due to a hydroxide species originating from a contamination of the deposition system, as this is the same in all cases. It is rather suggested that the surface component is related to a peroxide species. In a peroxide, the O2− ions are replaced by O2− ions. The situation resembles the dumbbell2 like oxygen interstitial defect recently described in literature [72,73] (see also Sect. 1.6 of this book). The presence of peroxide species at ZnO surfaces is not unreasonable. During deposition the surface is exposed mainly to Zn and O2 species. There are also other, more reactive, oxygen species in the gas phase. However, the percentage of dissociated and ionized gas species in a magnetron discharge is below 1 % [74] (a detailed description of sputter deposition of ZnO is given in Chap. 5 of this book). The growth of the oxide ﬁlm, therefore, requires dissociation of the O2 species. According to the spectra shown in Fig. 4.9 it is suggested that oxygen dissociation is not favorable on the surfaces of the II–VI compounds ZnO and CdS, while Cu(In,Ga)Se2 and In2 S3 obviously favor the dissociation of oxygen molecules. The diﬀerent behavior is summarized in a tentative model in Fig. 4.10.

O2 Zn

ZnO and CdS

O2 Zn

CIGS and In2S3

Fig. 4.10. Tentative model describing the initial growth of ZnO on diﬀerent substrates that is consistent with the diﬀerent evolution of the O 1s signal (Fig. 4.9) and the diﬀerent reactivitiy at the interface (see Sects. 4.3.2.1, 4.4.1, and 4.5.2). A peroxo-like surface species is observed during growth on ZnO and CdS substrates but not on In2 S3 and Cu(In,Ga)Se2 substrates. The diﬀerences are attributed to the abilities of the surfaces to dissociate the adsorbed O2 molecules

4 ZnO Surfaces and Interfaces

139

4.2.3 Electronic Structure of ZnO Surfaces The electronic potentials at a semiconductor surface are shown in Fig. 4.11. There are basically two independent quantities that can be determined using photoelectron spectroscopy: The distance between the Fermi energy and the valence band maximum (BE(VB) = EF − EVB ) and the work function (φ). The former quantity can change with doping and surface band bending, which is introduced by charged surface or interface states. A change of the surface Fermi level position changes the work function by the same amount. A change in work function can also be induced by changes of the surface dipole of the material by a modiﬁcation of the structure of the surface or by adsorbates [75]. Motivated by the application of ZnO in gas sensors and catalysis and by the more general desire to understand surface properties of ionically bonded solids, electronic properties of ZnO surfaces have been investigated for many years [20, 76–80]. An overview of the early work on ZnO surface properties is included in the book of Henrich and Cox [81]. An extensive investigation of the surface potentials at ZnO (0001), (000¯1), and (10¯ 10) surfaces using UPS is described by Jacobi et al. [79]. A noticeable change of Fermi level position after surface preparation (ion bombardment and annealing in vacuum) is observed with time for the oxygen terminated (000¯ 1) and to a lesser extend also for the nonpolar (10¯10) surface. These changes in band bending have been related to oxygen removal caused by UV irradiation. The Fermi level stabilizes after several hours close to the (a)

(b)

(c)

energy

Evac

χ

IP

φ

φ

φ

BE (PES)

ECB EF qVb EVB

Fig. 4.11. Surface potentials of an n-type semiconductor in ﬂat band condition (a). The work function φ can change either by modiﬁcation of the surface dipole preserving ﬂat bands but modifying the electron aﬃnity χ and ionization energy IP = χ+Eg (b). The work function might also change by bending of the bands at the surface qVb (c). Surface dipole and band bending could also change simultaneously. In photoelectron spectroscopy, binding energies are measured with respect to the Fermi energy

140

A. Klein and F. S¨ auberlich

conduction band minimum. For the (0001) surface termination, the Fermi level is close to the conduction band minimum directly after surface preparation (ion bombardment and annealing). In addition, a time-dependent reduction of work function is observed for all surface orientations. A similar eﬀect has also been reported by Moormann et al. [20, 77]. More recently, Meier has also observed time-dependent binding energy shifts of ZnO thin ﬁlms [82]. The shifts observed after heating magnetronsputtered ZnO thin ﬁlms are in the same direction and of the same magnitude as those described by Jacobi et al. In addition, Meier describes also binding energy shifts in the opposite direction, directly after deposition at low substrate temperatures. The shifts are related to changes in hydrogen and hydroxide concentrations near the surface [82]. In our studies, we have not observed time-dependent binding energies of the sputter-deposited ZnO ﬁlms. A possible reason for the diﬀerence might be the diﬀerent surface termination of the ﬁlms. As will be shown below, the ZnO ﬁlms used in this study were prepared by dc magnetron sputtering and show mainly a (0001) surface termination. For this particular surface orientation no changes in surface Fermi level position with time have been observed by Jacobi et al. [79]. The ZnO ﬁlms investigated by Meier et al. were deposited by rf magnetron sputtering [82], which might lead to a diﬀerent surface termination. 4.2.3.1 Surface Fermi Level Position The core and valence levels in Fig. 4.6 show comparable binding energy shifts in dependence on deposition conditions. The shifts are mainly due to shifts of the Fermi level position at the surface. The Fermi level position with respect to the valence band maximum is directly measured as the binding energy of the valence band maximum. Values for magnetron-sputtered ZnO and ZnO:Al thin ﬁlms are shown in Fig. 4.12 in dependence on oxygen content in the sputter gas and substrate temperature. Films deposited from the undoped and Al-doped target at room temperature without addition of oxygen to the sputter gas show a valence band maximum ∼2.8 and ∼3.7 eV below the Fermi energy, respectively. If we take the band gap of ZnO as 3.3 eV, the surface Fermi level of highly doped ZnO:Al is above the conduction band minimum, as expected for a degenerately doped n-type semiconductor. A Fermi level position ∼0.4 eV above the conduction band minimum is in good agreement with observed shifts of the optical transitions in ZnO because of ﬁlling of conduction band states, known as Burstein–Moss shift [83, 84]. Therefore, for degenerately doped ZnO:Al, the surface Fermi level position detected by XPS/UPS agrees with the bulk Fermi level position detected by optical measurements, which corresponds to the ﬂat band situation shown in Fig. 4.11a. This indicates that the ZnO surfaces are free of surface states in the fundamental band gap, in agreement with theoretical calculations of the electronic structure of ZnO surfaces [85]. Surface-sensitive electron energy loss spectroscopy (EELS) also indicates the

BE (VB) [eV]

4 ZnO Surfaces and Interfaces

3.8

i-ZnO

3.6

ZnO:Al

3.4

ECB

141

ECB

3.2 3.0 2.8 2.6 0

10

20

30

40

[O2]/([Ar]+[O2])[%]

50

100

300

500

temperature [˚C]

Fig. 4.12. Valence band maximum binding energies of magnetron sputtered ZnO and ZnO:Al ﬁlms in dependence on the oxygen content in the sputter gas at room temperature (left) and in dependence on substrate temperature for deposition in pure Ar (right). The binding energies are derived from X-ray excited valence band spectra. All ﬁlms were deposited using a total pressure of 0.5 Pa, a sputter power density of 0.74 W cm−2 and a substrate to target distance of 10 cm. The horizontal line indicates the position of the conduction band minimum

absence of electronic transitions with energies below the fundamental gap at diﬀerent surfaces [86]. The theoretical and the EELS investigations were performed on diﬀerent polar and nonpolar single crystal surfaces. Obviously, the results are also valid for polycrystalline sputtered ZnO ﬁlms. The absence of surface states in the fundamental gap allows to discuss the measured Fermi level positions in Fig. 4.12 in terms of bulk doping. This is particularly interesting for the addition of oxygen to the sputter gas. Addition of oxygen to the sputter gas does not change the Fermi level for the undoped material, but it leads to a considerable lowering of the Fermi level for ﬁlms deposited from the Al-doped target. The latter compares well with the reported reduction of conductivity of ZnO:Al ﬁlms with the addition of oxygen (see e.g., [87]). The origin of the reduced carrier concentration and the associated lowering of the Fermi level is not clear yet. It has been argued that addition of oxygen leads to an increased formation of Al2 O3 [87], thus removing Al atoms from the active dopant sites, which are provided by the regular Zn lattice sites (AlZn ). However, a reduced carrier concentration can also result from an increased compensation because of the introduction of intrinsic acceptor states. According to recent theoretical calculations, zinc vacancies or oxygen interstitials in a dumbbell or rotated dumbbell conﬁguration are possible candidates (see e.g., [72, 73] and Sect. 1.6). The reﬁned calculations indicate that the zinc vacancy has the lowest formation energy under oxygen-rich conditions [73]. In a recent publication by Lany and Zunger [S. Lany, A. Zunger, Phys. Rev. Lett. 98, 045501 (2007)], the zinc vacancy

142

A. Klein and F. S¨ auberlich

has been explicitly used to demonstrate the compensation mechanism with increasing oxygen partial pressure (see Fig. 1.15 in Sect. 1.6.1). However, this consideration holds for thermodynamic equilibrium while the deposition of ZnO by magnetron sputtering is far from equilibrium and also includes energetic particles. Therefore, also other defects will be present. In any case, the situation is again diﬀerent compared with ITO, where it is well accepted that the formation of neutral (2 SnIn Oi ) defect complexes is responsible for the reduction of conductivity under oxidizing conditions [88, 89]. For higher oxygen content in the sputter gas, the Fermi level stabilizes at ∼2.7 eV above the valence band maximum. This particular energy position seems to be related to an intrinsic defect level of ZnO. A similar Fermi level position has also been observed during interface experiments [90]. For a Fermi level position ∼0.6 eV below the conduction band, it can be expected that all AlZn donors are ionized. In principle, the Fermi level stabilization energy should then be a few kT lower than the energy level of an acceptor state, such that a considerable fraction of the acceptors are ionized. However, the identiﬁcation of the associated defect level is not straightforward, as direct experimental observations of intrinsic defect energy levels are, in most cases, not available. On the other hand, theoretical calculations can reveal defect levels but suﬀer from an incorrect determination of band gaps (see e.g., discussion in [73]). Within this uncertainty, both the zinc vacancy and the oxygen interstitials are, according to density functional theory calculations, reasonable candidates for the observed Fermi level stabilization energy under oxygen-rich conditions [73]. Both defects form rather deep acceptor states. 4.2.3.2 Work Function and Ionization Potential The work functions and ionization potentials of sputter-deposited ZnO and ZnO:Al ﬁlms are shown in Fig. 4.13. The diﬀerent Fermi level positions of ZnO and ZnO:Al for deposition at room temperature in pure Ar are also observed in the work function. The undoped ﬁlms prepared under these conditions have a work function of ∼4.1 eV, while the Al-doped ﬁlms show values of ∼3.2 eV. The diﬀerence is almost of the same magnitude as for the Fermi level position and, therefore, explained by the diﬀerent doping level. Also the ionization potentials are almost the same under these preparation conditions. The work function of the undoped material is close to the value reported by Moormann et al. for the vacuum-cleaved Zn-terminated (0001) surface [20]. The same authors report a work function of 4.95 eV for the oxygen terminated ZnO(000¯ 1) surface, which is in good agreement with the values obtained for ﬁlms deposited with 5 % oxygen in the sputter gas. Since the Fermi level position of the undoped ZnO ﬁlms does not depend on the oxygen content in the sputter gas (Fig. 4.12), the diﬀerent work functions correspond to diﬀerent ionization potentials. The ionization potentials of the undoped ZnO ﬁlms prepared at room temperature are ∼6.9 eV for ﬁlms deposited in pure Argon and raise to ∼7.7 eV

work function [eV]

4 ZnO Surfaces and Interfaces 5.1

i-ZnO

4.8

ZnO:Al

143

4.5 4.2 3.9 3.6

ionization potential [eV]

3.3

7.8

charging

7.6

statistical orientation non-polar surface O termination

7.4 7.2 7.0

Zn termination

6.8 0

10

20

30

40

[O2]/([Ar]+[O2])[%]

50

100

300

500

temperature [˚C]

Fig. 4.13. Work function and ionization potential of magnetron sputtered ZnO and ZnO:Al ﬁlms in dependence on oxygen content in the sputter gas for samples deposited at room temperature (left) and in dependence on substrate temperature for deposition in pure Ar (right). The values are derived from He I excited valence band spectra. All ﬁlms were deposited using a total pressure of 0.5 Pa, a sputter power density of 0.74 W cm−2 and a substrate to target distance of 10 cm

for ﬁlms deposited with 5 % oxygen in the sputter gas. The lower value corresponds well with literature data given for the electron aﬃnity of ZnO(0001) (3.7 eV) [79]. Ionization potentials of ∼7.82 eV have been determined by Swank et al. for the nonpolar ZnO(10¯ 10) surface [76] as well as by Jacobi et al. for ZnO(10¯ 10) and for oxygen terminated ZnO(000¯1). These values are in good agreement with the ionization potential of ZnO ﬁlms sputtered from the undoped target with more than 5 % oxygen in the sputter gas. The variation of the ionization potential with surface orientation evident from the literature data corresponds well with a systematic study by Ranke using a cylindrical GaAs single crystal [91]. This revealed that the electron aﬃnity of the cation terminated (111) surface of GaAs (corresponding to wurtzite (0001)) is 0.4–0.5 eV lower than those of other surface terminations. The same variation is observed at single crystal surfaces of CdTe and

144

A. Klein and F. S¨ auberlich

CdS [92]. The ionization potentials of CdTe(111), (¯1¯1¯1), and (110) amount to 5.3, 5.6, and 5.65 eV, those of epitaxial CdS ﬁlms deposited onto these surfaces are given by 6.25, 6.85, and 6.75 eV, respectively. For comparison, the ionization potentials of polycrystalline CdTe and CdS ﬁlms amount to 5.8 and 6.9 eV [92]. The lower ionization potential of the cation terminated (111) or (0001) surfaces is related to a smaller surface dipole. The diﬀerent ionization potentials can also aﬀect the band alignment at weakly interacting interfaces [93]. According to these observations, the low ionization potential of the undoped ZnO ﬁlms deposited at room temperature in pure Ar corresponds to a predominant Zn-terminated (0001) surface orientation. The surface orientation agrees with a preferred (0001) orientation observed in X-ray diﬀraction for identically prepared ﬁlms (see Fig. 4.146 ). A pronounced c-axis texture is almost exclusively observed for sputtered ZnO ﬁlms (see e.g., [39, 94, 95] and other chapters of this book). With increasing oxygen content, the increase of ionization potential indicates a change of surface orientation, which agrees with a reduced texture in XRD (see Fig. 4.14). No signiﬁcant change of ionization potential is observed for a substrate temperature of 200–300◦C (see lower right graph of Fig. 4.13). Apparently, Zn-terminated (0001) surfaces are the predominant surface termination under these deposition conditions, in agreement with a pronounced (0001) texture in XRD (Fig. 4.14). The Al-doped ﬁlms deposited at room temperature show an ionization potential of 7.0–7.1 eV, independent on oxygen concentration. Reliable work function measurements of ZnO:Al ﬁlms deposited with more than 10 % oxygen in the sputter gas were not possible due charging eﬀects during measurement. Apparently, the presence of Al in the ﬁlms stabilizes the (0001):Zn surface termination. ZnO:Al ﬁlms deposited with 10 % oxygen in the sputter gas also exhibited a pronounced (0001) texture [70]. In contrast, an increase of substrate temperature leads to a reduced texture and an associated increase of ionization potential. This change is most likely related to the increase of the Al-content in the ﬁlms with increasing substrate temperature (see Fig. 4.7). A higher Al content hence leads to a deviation from the pronounced (0001) texture, as also reported by Sieber et al. [96]. 4.2.3.3 Core-Level Binding Energies Because of the variation of the Fermi level position with respect to the band edges (see Fig. 4.12), the binding energies of the core levels will also change. Hence, the core level binding energies of semiconductors are not a good reference. However, the binding energy diﬀerence between the core levels and the 6

The diﬀraction patterns in Fig. 4.14 were recorded in gracing incidence and plotted on a logarithmic scale and, therefore, have a diﬀerent appearance compared with data taken in standard Bragg–Brentano (Θ–2Θ) mode and data plotted on a linear scale.

4 ZnO Surfaces and Interfaces

145

normailzed log. intensity

250°C/Ar

RT/10%O2

30

40

50

60

70

80

210

203

104

202

004

200 112 201

103

110

102

100 002 100

RT/Ar

90

2Θ [deg] Fig. 4.14. Gracing incidence (Φ0 = 2.2◦ ) X-ray diﬀraction patterns of undoped ZnO ﬁlms deposited with identical parameters as those of ﬁlms used for photoemission experiments. Intensities are plotted on a logarithmic scale to emphasize the low intensity features. The patterns were recorded using Cu Kα radiation (λ = 1.54060 ˚ A). The thickness of the ﬁlms is ∼1 µm

valence band maximum BEVB (CL) should be constant for a given material as they reﬂect the density of states. These values are essential for the determination of the band alignment at semiconductor interfaces [8,33]. Corresponding values for the Zn 2p3/2 and the O 1s core levels of sputter-deposited ZnO ﬁlms in dependence on deposition conditions are given in Fig. 4.15. It is evident that the values depend on deposition conditions and on the doping of the ﬁlms. An almost identical variation is observed for undoped (Zn,Mg)O ﬁlms [59]. The variation in BEVB (CL) and in the binding energy diﬀerence between the two core levels is explained by a superposition of two diﬀerent eﬀects. To begin with, undoped ZnO ﬁlms show the highest BEVB (CL) when the ﬁlms are deposited at room temperature with pure Argon as sputter gas. The

A. Klein and F. S¨ auberlich

BEVB (Zn2p3/2) [eV]

146

i-ZnO

1018.9

ZnO:Al

1018.8

charging

1018.7 1018.6 1018.5 1018.4 i-ZnO

BEVB (O1s) [eV]

527.7

ZnO:Al

527.6

charging

527.5 527.4 527.3

BE (Zn) - BE (O) [eV]

527.2 491.25 491.20 491.15 491.10

i-ZnO ZnO:Al

491.05 0

10

20

30

40

[O2]/([Ar]+[O2])[%]

50

100

300

500

temperature [°C]

Fig. 4.15. Binding energy diﬀerence between Zn 2p3/2 and O 1s core levels and between the core levels and the valence band maximum. Films are deposited by dc magnetron sputtering from undoped and 2 % Al-doped ceramic targets at room temperature in dependence of sputter gas composition (left) and in pure Ar in dependence on substrate temperature (right). All ﬁlms were deposited using a total pressure of 0.5 Pa, a sputter power density of 0.74 W cm−2 and a substrate to target distance of 10 cm

4 ZnO Surfaces and Interfaces

147

addition of oxygen to the sputter gas or the deposition at elevated substrate temperatures leads to the same values for BEVB (CL), independent on the target material (undoped or doped). It is reasonable to assume that the values obtained under such conditions BEVB (O 1s) = 527.4 ± 0.1 eV BEVB (Zn 2p3/2 ) = 1018.55 ± 0.1 eV BE(Zn 2p3/2 ) − BE(O 1s) = 491.15 ± 0.1 eV are representative for well-ordered crystalline ZnO. In contrast, ﬁlms deposited at room temperature without addition of oxygen have to be expected to contain a considerable number of structural defects as the formation of well-ordered crystalline ZnO may be kinetically prohibited. Structural disorder can lead to a modiﬁcation of the density of states in the valence band, which is well known for the extreme disorder in amorphous semiconductors (Urbach tails). However, X-ray diﬀraction patterns (see Fig. 4.14) of ﬁlms prepared in the same way as those used for the data in Fig. 4.15 show clearly the long range order of the wurtzite lattice and provide no evidence for an amorphous structure of the ZnO ﬁlms. Nevertheless, local disorder can be present as a high density of crystallographic point defects or a high density of stacking faults [96,97]. These do not aﬀect the long range order but modify the local chemical bonding and thereby inﬂuence the electron wave functions and the local charge distribution. Local disorder might aﬀect the valence band maximum energy of ZnO also via a modiﬁcation of orbital hybridization. The Zn 3d levels are low lying d-states, which hybridize with the O 2p states and thus contribute to the valence band density of states [98–100]. Any change in local symmetry will also aﬀect the hybridization between the Zn 3d and O 2p states. It is expected that the p–d hybridization leads to a lowering of the valence band maximum in ZnO (see discussion of band alignment between II–VI compounds in Sect. 4.3.1). A reduction of the p–d interaction by disorder would thus lead to an upward shift of the valence band maximum and consequently to an increase of BEVB (CL) as indeed observed in the experiments. Instead of modifying the density of states, local disorder might also lead to a poorer screening of the core hole as a result of reduced polarizability of the lattice. This might also account for the larger BEVB (CL) values of ﬁlms deposited at room temperature in pure Ar. The variation of band alignment on the deposition conditions suggests a modiﬁcation of the density of states as origin for the changes in BEVB (CL) (see Sect. 4.3.3). However, a detailed comparison of valence band spectra provides no hint for such an explanation as no changes in the shape and width of the spectra are observed (see Fig. 4.16). There are also no evident changes in the distance between Zn 3d level and the valence band maxima in these spectra. Hence, although the variation in BEVB (CL) for undoped ZnO suggests the presence of a strong local disorder for ﬁlms deposited at room

148

A. Klein and F. S¨ auberlich ZnO:Al

normalized intenstiy

i-ZnO

14

12

10

8

6

4

2

0

14

12

10

8

6

4

2

0

binding energy [eV]

Fig. 4.16. Comparison of He II (hν = 48.86 eV) valence band spectra of sputter deposited undoped (left) and Al-doped (right) ZnO ﬁlms. Diﬀerent curves belong to diﬀerent sputter conditions as substrate temperature and oxygen partial pressure. The binding energies of the spectra were shifted for better comparison of shapes of the spectra

temperature in pure Ar, a detailed description of the inﬂuence of the disorder cannot be given yet. In addition to the local disorder, the high free-electron concentration of the Al-doped ﬁlms deposited at low substrate temperature and with less than ∼2 % oxygen in the sputter gas also contributes to the variation in BEVB (CL). The free electrons lead to an increased screening of the positive charge of the core hole, which is created by the photoemission process. The additional screening reduces the apparent binding energy of the core hole. The eﬀect has been systematically observed for Sb-doped SnO2 [55, 56] and also for ITO [57, 58]. The screening of the core hole aﬀects not only the apparent binding energy of the core level and thereby BEVB (CL), but also the line shape of the core levels due to the occurrence of plasmon satellites, which has already been mentioned earlier. The processes and their inﬂuence on line shape are illustrated in Fig. 4.17. According to Fig. 4.15, a BEVB (Zn 2p3/2 ) of highly doped ﬁlms is ∼150 meV smaller than those of ﬁlms deposited at higher substrate temperatures or with addition of oxygen to the sputter gas. For ITO, the diﬀerence in BEVB (In 3d5/2 ) between highly doped and undoped material amounts to ∼500 meV, which is considerably larger than for ZnO. Most likely, the reduction of the apparent core-level binding energy in ZnO by screening is superimposed by an increase in binding energy, which is observed for identically prepared undoped ZnO ﬁlms. The diﬀerence in BEVB (Zn 2p3/2 ) for i-ZnO and ZnO:Al for ﬁlms prepared at room temperature with pure Ar as sputter gas amounts to ∼400 meV, which is comparable to the shift observed for ITO.

intensity [arb. units]

4 ZnO Surfaces and Interfaces In 3d5/2 screened component

O 1s screened component

plasmon satellites

448

149

plasmon satellites

446

444

442

534

532

530

528

binding energy [eV] core hole screening

e photoelectron

hωP

plasmon excitation

Fig. 4.17. Photoelectron line shape of the In 3d5/2 and O 1s core levels of a highly doped ITO ﬁlm [58]. The high concentration of free electrons leads to excitation of plasmons, which give rise to shoulders at the high binding energy side of the main emission. Furthermore, the polarization of the free electron gas leads to an additional screening of the core hole, which reduces the binding energy of the main emission component. The plasmon energy depends on the electron concentration and amounts to 0.5–1 eV for the highest doped ﬁlms [56, 101]

4.3 The CdS/ZnO Interface 4.3.1 Band Alignment of II–VI Semiconductors The CdS/ZnO interface is of particular importance in Cu(In,Ga)Se2 thin ﬁlm solar cells because it is used in the standard cell conﬁguration (Fig. 4.2). A ﬁrst experimental determination of the band alignment at the ZnO/CdS interface has been performed by Ruckh et al. [102]. The authors have used exsitu sputter-deposited ZnO ﬁlms as substrates. The interface formation was investigated by stepwise evaporation of the CdS compound from an eﬀusion cell. Photoelectron spectroscopy revealed a valence band oﬀset of ∆EVB = 1.2 eV. An identical value of 1.18 eV has been derived using ﬁrst-principles calculations [103]. With the bulk band gaps of CdS and ZnO of 2.4 and 3.3 eV, respectively, this leads to a conduction band oﬀset 0.3 eV, with the conduction band minimum of CdS being above the one of ZnO. This value is frequently used in the literature for modeling of the Cu(In,Ga)Se2 thin ﬁlm solar cells [14, 104, 105].

150

A. Klein and F. S¨ auberlich 8

theor. Wei et al. exp. various exp. Schulmeyer

6

–4

4

0

1,17

0

uncertainty 100 meV

EVB ZnO

0,60

0,18

0

0,53

2

1,26

–2 ECB

ZnS

ZnSe

ZnTe

CdS

CdSe

∆EVB (CuInSe2 /MX) [eV]

–6

–1,00

∆EVB (ZnS/MX) [eV]

Evac

2

CdTe

Fig. 4.18. Band alignment of II–VI compound heterointerfaces calculated ab-initio by density functional theory [103] (left section of each column). The valence band maxima are given with respect to the valence band maximum of ZnS as in the original publication by Wei et al. Available experimental values for the band alignments of individual interfaces are shown in the middle section of each column. Corresponding experimental details and references are given in Table 4.1. The right section contains valence band oﬀsets ∆EVB of Cu(In,Ga)Se2 /II–VI heterointerfaces as determined by Schulmeyer et al. [36, 106, 107] (right axis). The energy diﬀerence between the left and right axes corresponds to the valence band oﬀset at the CuInSe2 /ZnS interface. Vacuum levels as determined for the diﬀerent materials are shown at the top. Except for ZnS the vacuum energies are at approximately the same energy, indicating that only small interface dipoles occur

The calculated band alignments of the II–VI semiconductors including ZnO [103] are shown in Fig. 4.18. The energies of the valence band maxima are referenced to the valence band maximum of ZnS. Included in the ﬁgure is a selection of experimental results for band alignment at diﬀerent in situ prepared II–VI semiconductor interfaces (for detailed values and references see Table 4.1). All experimental results agree with the theoretical prediction within 100 meV. An additional justiﬁcation of the theoretical band alignments is given by the valence band oﬀsets between the II–VI compounds and Cu(In,Ga)Se2 . These interfaces are prepared by evaporation of II–VI compounds onto oxide-free device-grade Cu(In,Ga)Se2 surfaces [36,106,107]. Such surfaces can be made available using the decapping procedure by heating-oﬀ of a Se cap layer, which is deposited onto a freshly prepared Cu(In,Ga)Se2 layer without breaking vacuum. The procedure allows for storage and transport of the samples in air. The agreement between the calculations and the various experimental results is excellent. This indicates the wide applicability of the calculated band alignments. The general behavior also conﬁrms the original value given

4 ZnO Surfaces and Interfaces

151

Table 4.1. Valence band oﬀsets at interfaces of II–VI compounds determined by photoelectron spectroscopy Interface ZnO/CdS ZnO/ZnTe CdTe/ZnTe CdS/CdTe CdTe/ZnTe CdSe/ZnTe

∆EVB (eV) 1.2 2.37 0.1 1.01 0.1 0.64

Ref. [102] [108] [108] [109] [110] [111]

Growth MBE Sputtering Sputtering MBE MBE MBE

Crystallinity Poly Poly Poly Poly Single Single

Lattice Mismatched Mismatched Mismatched Mismatched Mismatched Matched

by Ruckh et al. [102] for the band alignment at the ZnO/CdS interface and its use in device simulation. It also provides a valuable tool for the selection of diﬀerent buﬀer layers in solar cells. In general, Zn-containing compounds as ZnS, ZnSe, and ZnTe result in a considerably larger conduction band oﬀset to ZnO compared with Cd-containing compounds as CdS, CdSe, and CdTe. This is related to the larger band gaps of the Zn- compared to the Cd-chalcogenides in combination with the similar valence band maximum energies for the same anions. Although the band alignment at the interfaces seems to be well established from the data presented in Fig. 4.18 and Table 4.1, the inﬂuence of the sputter deposition on the electronic properties of the interface has not yet been addressed systematically. Such studies are required to identify the fundamental parameters that govern interface chemistry and electronic properties. In the following we describe our recent results on the CdS/ZnO interfaces. Although being rather extensive, so far only CdS ﬁlms prepared by thermal evaporation have been used. In high eﬃciency solar cells, the CdS ﬁlms are prepared by chemical bath deposition [15], which might lead to diﬀerent interface properties. However, noticeable eﬃciencies can also be achieved using evaporated CdS ﬁlms [112]. Furthermore, Weinhardt et al. have studied the band alignment at the CdS/ZnO interface by sputter depth proﬁling of a layer structure used for solar cells, i.e., a chemical bath deposited CdS buﬀer layer and sputter-deposited undoped ZnO [113]. Valence and conduction band oﬀsets are determined as ∆EVB = 0.96 ± 0.15 eV and ∆ECB = 0.1 ± 0.15 eV. The valence band oﬀset lies well within the range of values presented in the following and therefore suggests that the electronic interface properties are only little aﬀected by the CdS deposition technique. 4.3.2 Sputter Deposition of ZnO onto CdS A determination of the band alignment at the CdS/ZnO interface where ZnO has been stepwise deposited by magnetron sputtering has been published by Venkata Rao et al. [71]. A more extended series of spectra recorded during ZnO deposition by dc magnetron sputtering onto CdS are presented in Fig. 4.19. During ZnO deposition the sample was held at room temperature.

152

A. Klein and F. S¨ auberlich Zn 2p

S 2p

Cd 3d

O 1s

ZnO deposition time

intensity [arb. units]

2948s

2048

1024 512 256 128 64 32 16 8 4 2 1 0

1024

1020 Cd MNN

534

530

406

404

Zn LMM

165

161

VBM 2948s 2048 1024

intensity [arb. units]

512 256 128 64 32 16 8 4 2 1 0

1114 1108 1102 504 500 496 492 6

4

2

0

binding energy [eV]

Fig. 4.19. Core levels (Zn 2p3/2 , O 1s, Cd 3d5/2 , S 2p), Auger levels (Cd MNN, Zn LMM) and valence bands (VB) recorded during stepwise sputter deposition of ZnO onto a CdS substrate. The deposition times are indicated in seconds. All spectra were excited using monochromatic Al Kα radiation (hν = 1486.6 eV). ZnO was deposited from an undoped target using pure Ar as sputter gas and a sputter power of 5 W (dc)

The bottom spectra are taken from a freshly evaporated CdS ﬁlm. The only emissions observed are from Cd and S core levels and Auger levels and from the valence band region. With increasing deposition time of ZnO, the Cd and S levels are attenuated, while the emissions from the growing ZnO ﬁlm

4 ZnO Surfaces and Interfaces

153

are increasing. The valence band spectra of CdS are gradually replaced by those of ZnO, which is evident from the higher binding energy position of the valence band maximum. It is recalled that the escape depth of the photoelectrons is ∼1–2 nm. The low attenuation of the substrate emissions after the ﬁrst deposition step corresponds to a nominal ZnO ﬁlm thickness of less than 1 ˚ A.7 To achieve such a low deposition rate by dc magnetron sputtering, a low sputter power of 5 W (2 in.-target) has been used. 4.3.2.1 Interface Chemistry and Film Growth A chemical reaction at the interface can, in principle, result in additional components of the core levels or Auger lines, or in a diﬀerent energy shift of the diﬀerent lines of the substrate or ﬁlm. In particular, an oxidation of the CdS substrate should result either in the occurrence of SOx emissions at binding energies of 166–169 eV [114], or in the change of the line shape of the Cd MNN Auger emission.8 No such eﬀects are observed. In addition, the substrate emission lines (Cd 3d, O 1s, Cd MNN) show only very small and nearly parallel binding energy shifts (see also Fig. 4.20). An oxidation of the substrate can, therefore, be excluded within the sensitivity of the experiment. Hence, even for sputter-deposited oxide ﬁlms, oxidation of the substrate is not inevitable. This might be explained by the fact that most of the oxygen involved in the growth process is present as stable O2 molecules. In contrast to the substrate emissions, the line shape of the growing O 1s emission changes signiﬁcantly with ﬁlm thickness. The spectra are the same as those shown in Fig. 4.9a. For low coverage, the high binding energy component dominates. The behavior has been explained in Sect. 4.2.2.2 by the formation of peroxide species due to the inability of the CdS surface to dissociate the adsorbed O2 molecules. This also explains why no oxidation of the substrate is observed during growth. It has been argued in Sect. 4.2.2.2 that dissociation of O2 is facilitated on In2 S3 and Cu(In,Ga)Se2 surfaces, since no peroxide is observed on these surfaces at low ZnO coverage (see Fig. 4.9b,d). In fact, during deposition of ZnO onto In2 O3 and Cu(In,Ga)Se2 , oxidation of the substrate is observed by XPS (see Sects. 4.5 and 4.4), which supports the interpretation given in Fig. 4.10. At low coverage, the binding energy of the ZnO O 1s level (low binding energy O 1s component) does not shift in parallel to the binding energy of the Zn 2p3/2 level. The binding energy diﬀerence BE(Zn 2p3/2 ) − BE(O 1s) is up to 0.7 ± 0.1 eV larger than the value observed for the thick ZnO ﬁlm. This is evident from the splitting of the two curves for ZnO at low deposition times in Fig. 4.20. An identical behavior is observed in all experiments where ZnO has been deposited onto CdS, In2 S3 , and Cu(In,Ga)Se2 (see Sects. 4.4 and 4.5, 7 8

The deposition rate is ∼5 nm min−1 . Chemical shifts are not very pronounced for the Cd 3d core level [114] and might therefore be hardly identiﬁed.

154

A. Klein and F. S¨ auberlich Zn 2p O 1s Cd 3d S 2p

EF - EVBM [eV]

2.0 2.4 1.2 eV

2.8 3.2 3.6 0

1000 2000 deposition time [s]

Fig. 4.20. Evolution of the CdS and ZnO valence band maxima as derived from the binding energies of the core levels by subtracting BEVB (CL) values determined from the CdS substrate and the thick ZnO ﬁlm, respectively [71]. The diﬀerent evolution of the Zn 3d and O 1s binding energies is attributed to an amorphous structure of the ZnO layer during the initial growth. The thickness of the amorphous layer is ∼2 nm. The ZnO ﬁlms were deposited by magnetron sputtering from an undoped ZnO target at room temperature using 5 W dc power

Figs. 4.30 and 4.37, and [70]) and also during deposition of (Zn,Mg)O onto CdS [71]. Therefore, the diﬀerent evolution of the oxygen and zinc core-level binding energies can not be related to the observation of the peroxide surface species, as no such species is observed during initial growth on In2 S3 and Cu(In,Ga)Se2 (see Fig. 4.9). For the reverse deposition sequence (CdS on ZnO), the binding energy diﬀerence between the Zn 2p and O 1s core levels remains constant during deposition [70, 90]. However, in these experiments the binding energy diﬀerence between the Cd 3d and S 2p core levels is diﬀerent at low coverage [70,90]. An identical behavior is observed during growth of CdS onto SnO2 [115] and during growth of CdTe onto In2 O3 [116]. It is, therefore, clear that the different binding energy diﬀerence between anion and cation core levels at low coverage cannot be due to a chemical interaction at the interface but is rather related to the structure of the growing ﬁlm. Yoshino et al. [117] have studied the growth of magnetron-sputtered ZnO ﬁlms on diﬀerent substrates using transmission electron microscopy and X-ray diﬀraction. A crystalline nucleation layer of ZnO is only observed on surfaces of crystalline materials such as sapphire and Au. In the case of unordered surfaces like glass or Al, which are either amorphous or develop an amorphous oxide layer during deposition, the ZnO nucleation layer is highly disordered or amorphous. Similar results were observed by Mirica et al. [97] who compared the growth of ZnO on oxide forming (Si) and nonoxide forming (Pt)

4 ZnO Surfaces and Interfaces

155

substrates. Sieber et al. [96] also found a highly disordered layer at the Si/ZnO interface consisting of an amorphous SiO2 and an nanocrystalline ZnO layer. The formation of an amorphous nucleation layer of ZnO on (poly-) crystalline chalcogenides as CdS, In2 S3 and Cu(In,Ga)Se2 could be related to the interface chemistry. In the case of CdS, oxygen dissociation and therefore formation of a crystalline layer is hindered. For In2 S3 and Cu(In,Ga)Se2 (see following sections), oxygen dissociation at the surface leads to a partial oxidation of the substrate surface, which also destroys the regular atomic arrangement at the surface. Following these studies, a microstructure of sputter-deposited ZnO ﬁlms on polycrystalline CdS substrates is outlined in Fig. 4.21. The diﬀerent evolution of the Zn 2p and O 1s binding energies can consequently be attributed to the amorphous ZnO nucleation layer with a diﬀerent chemical bonding between Zn and O. The model is also valid for polycrystalline In2 S3 and Cu(In,Ga)Se2 substrates and for deposition of (Zn,Mg)O ﬁlms, as these show the same behavior (see Figs. 4.20 and 4.24). It is not clear whether an amorphous nucleation layer occurs also when the ZnO is deposited by other techniques as MBE, CVD, or PLD, as no data are available for such interfaces. In addition, the inﬂuence of the polycrystallinity of the substrates is not clear so far. surface species

polycrystalline, preferred orientation transition depends on preparation parameters

polycrystalline, statistical orientation amorphous polycrystalline substrate

Fig. 4.21. Microstructure of ZnO ﬁlms on polycrystalline CdS substrates following models for amorphous substrates suggested in literature [96, 97, 117]. No chemical reaction between CdS and ZnO is identiﬁed suggesting a sharp interface. The initial growth of ZnO proceeds with a highly disordered or amorphous structure, giving rise to the diﬀerent evolution of the Zn 2p and O 1s core-level binding energies at low coverage, which is generally observed at low coverage during sputter deposition of ZnO onto CdS (see Figs. 4.20 and 4.24 and [71]) The thickness of the amorphous layer is ∼2 nm

156

A. Klein and F. S¨ auberlich

4.3.2.2 Band Alignment at the Interface The valence band oﬀset at the interface can be directly evaluated from the data in Fig. 4.20. For deposition times between 500 and 2 000 s, the diﬀerence in valence band maxima positions for CdS and ZnO is almost constant with an average value of ∆EVB = 1.2 ± 0.1 eV. This is in excellent agreement with the value from Ruckh et al. [102] for the ZnO/CdS interface and with the theoretical calculation by Wei et al. [103], suggesting that there is only little inﬂuence of the preparation conditions on the band alignment. However, given the highly asymmetric microstructure of the CdS/ZnO and the ZnO/CdS interfaces, which is expressed by Fig. 4.21, the agreement between the valence band oﬀsets for the two orientations cannot necessarily be expected. In addition the agreement turns out to be fortuitous when data for additional interfaces are taken into account (see Sect. 4.3.3). The energy band diagram of the interface as determined from the spectra is shown in Fig. 4.22. The positions of the vacuum energy, which is determined from UPS measurements of the CdS substrate and the thick ZnO layer are included. Taking these values, there appears to be a large discontinuity in the vacuum energy (dipole potential) at the interface, which could be attributed to a charge transfer at the interface. The large discontinuity in the vacuum level is a result of the low electron aﬃnity of the ZnO, which is due to the (0001):Zn surface termination obtained under the selected preparation conditions of the ZnO ﬁlm (see Sect. 4.2.3.2). However, when the details of the microstructure (Fig. 4.21) are taken into account, the ZnO layer close to the interface is not c-axis but rather randomly oriented. Therefore, a larger electron aﬃnity has to be used, which consequently leads to a smaller interface dipole potential as illustrated in the right part of Fig. 4.22. A similar model, although with a reversed order of electron aﬃnities has been used by L¨ oher et al. to describe the band alignment between II–VI compounds and layered transition metal dichalcogenides [93]. 4.3.3 Dependence on Preparation Condition To study the inﬂuence of the preparation conditions on the interface properties, a number of diﬀerent interfaces have been prepared. Details of the preparation and the determined valence band oﬀsets are listed in Table 4.2. The experiments include not only both deposition sequences, but also interfaces of Al-doped ZnO ﬁlms, which have been conducted to elucidate the role of the undoped ZnO ﬁlm as part of the Cu(In,Ga)Se2 solar cell. Details of the experimental procedures and a full set of spectra for all experiments are given in [70]. Table 4.2 includes a number of interfaces between substrates of undoped ZnO ﬁlms and evaporated CdS layers (ZOCS A-D). In a recent publication [90] diﬀerent values were given for the valence band oﬀsets, as the dependence of BEVB (CL) on the deposition conditions was not taken into account in this publication.

4 ZnO Surfaces and Interfaces Evac

157

Evac CdS

1.17

ZnO

CdS

0.27 ZnO

ZnO

4.45

4,45

4.5 3.6

0.32

ECB

3.6

0.32

ECB

EF

EF 2.42

2.42 3.3

EVB 1.20

3.3

3.3

random

(0001)

EVB 1.20

Fig. 4.22. Energy band diagram at the CdS/ZnO interface. All values are given in electronvolt. The left side shows the diagram where the vacuum energy of the ZnO layer is determined from the thick ZnO ﬁlm. For the selected preparation conditions (room temperature, 100 % Ar) the ﬁlm is (0001) textured with a Zn termination (see Sect. 4.2.3.2), which results in a large discontinuity of the vacuum energy at the interface. According to the microstructure of the interface (Fig. 4.21), the ZnO close to the CdS is not oriented leading to a larger electron aﬃnity of ∼4.5 eV and hence to a lower discontinuity in the vacuum energy. In the left part of the ﬁgure, the amorphous ZnO nucleation layer is indicated by the shaded area. The amorphous region is omitted in the right part for clarity

The experimentally determined valence band oﬀsets span quite a large range from ∆EVB = 0.84 − 1.63 eV. The variation of 0.8 eV is considerably larger than the experimental uncertainty, which is ±0.1 eV for most experiments with only a few exceptions. Experiments with a larger uncertainty have been omitted. The experimental procedure for the determination of the valence band oﬀsets directly relies on the core level to valence band maximum binding energy diﬀerences BEVB (CL) as described in Sect. 4.1.3 and Fig. 4.3. The corresponding values for the Zn 2p3/2 and the Cd 3d5/2 core level are therefore included in Table 4.2. These values are determined directly from the respective interface experiments. With two exceptions (CSZA-E and ZACS-C), the values for the Zn 2p3/2 core level show the same dependence on deposition conditions as given in Fig. 4.15. For these two exceptions, also the Fermi level position

158

A. Klein and F. S¨ auberlich

Table 4.2. Details of experiments on diﬀerent CdS/ZnO interfaces studied by photoelectron spectroscopya Label CSZO-A ZOCS-A ZOCS-C ZOCS-D Ref. [102] CSZA-A CSZA-B CSZA-D ZACS-A ZACS-D

Interface CdS/ZnO ZnO/CdS ZnO/CdS ZnO/CdS ZnO/CdS CdS/ZnO:Al CdS/ZnO:Al CdS/ZnO:Al ZnO:Al/CdS ZnO:Al/CdS

ZnO CdS (◦ C) ∆EVB (eV) Zn-VB (eV) Cd-VB (eV) 5W 250 1.20 1018.85 403.46 3W 25 0.84 1018.89 403.41 450◦ C 25 1.00 1018.74 403.45 25 1.11 1018.54 403.48 10 %O2 RF 25 1.10 1019.1 403.4 S 25 1.63 1018.44 403.48 250 1.17 1018.53 403.58 10 %O2 S 250 1.57 1018.44 403.53 S 25 1.43 1018.45 403.42 25 1.37 1018.55 403.52 10 %O2

a

Unless indicated, ZnO ﬁlms were sputter deposited using the standard parameters “S,” which are deﬁned as T = 25◦ C, p = 0.5 Pa (100 % Ar), P = 15 W (dc). All CdS ﬁlms were prepared by thermal evaporation from the compound. Uncertainties of valence band oﬀsets ∆EVB and core level to valence band maximum binding energy diﬀerences BEVB (CL) (“Zn-VB” for ZnO and “Cd-VB” for CdS) are typically ±0.1 eV. Experiments with larger uncertainties have been omitted. The ZnO:Al ﬁlms were deposited from targets with an Al content of 2 wt %. Details of the experimental procedures and a full set of spectra for all experiments are provided in [70]. The valence band oﬀsets given for the interfaces ZOCS-A to ZOCS-D diﬀer from those in a recent publication [90] in which a constant value of BEVB (CL) for Zn 2p and O 1s levels have been used. Data from Ruckh et al. [102] are also included. For a better comparison we have adopted their value determined from the core-level binding energy diﬀerence (1.1 eV) instead of the value determined directly from the valence band spectra (1.2 eV)

is much lower than expected (BE(VB) ≈ 3.2 eV instead of ∼3.6 eV). As the ZnO:Al ﬁlms are deposited at 450◦ C substrate temperature in these experiments, the deviating ﬁlm properties might be related to diﬀusion of species from the substrates. The BEVB (CL) for the Cd 3d5/2 level of the CdS ﬁlms are 403.46 ± 0.05, which is in good agreement with literature [36, 102, 115]. All CdS ﬁlms were deposited by thermal evaporation from the compound. No signiﬁcant dependence of the ﬁlm properties with substrate or source temperature has been noticed so far. To account for a possible inﬂuence, the experimental valence band oﬀsets are plotted vs. BEVB (Zn 2p3/2 ) in Fig. 4.23. For comparison, values obtained for interfaces between thermally evaporated CdS ﬁlms and undoped (Zn,Mg)O ﬁlms prepared by rf magnetron sputtering have been included [59]. The agreement between the valence band oﬀsets of ZnO (squares) and (Zn,Mg)O (triangles) is due to the fact that the larger band gap of (Zn,Mg)O is due to a higher conduction band minimum and a nearly zero valence band oﬀset between these two materials [71,118]. According to Fig. 4.23, the experimental valence band oﬀsets can be divided into diﬀerent groups, which are indicated by circled numbers.

4 ZnO Surfaces and Interfaces

1.6

159

3

0.2eV 4

1.4 ZnO/CdS

1.2

3

1.0

CdS/ZnO ZnMgO/CdS

1

0.35eV

∆EVB [eV]

? 2

CdS/ZnMgO ZnO:Al/CdS CdS/ZnO:Al ZnO/CdS (Ruckh)

0.8 1019.0

1018.8

1018.6

1018.4

BEVB (Zn2p3/2) [eV]

Fig. 4.23. Valence band oﬀsets for CdS/ZnO, CdS/(Zn,Mg)O and CdS/ZnO:Al interfaces as determined by photoemission experiments. Solid symbols are for sputter deposition of the oxides onto CdS, open symbols are for deposition of CdS onto the oxides. The value from Ruckh et al. [102] is included (diamonds). The circled numbers serve to classify the diﬀerent values as described in the text

1. The valence band oﬀsets for the interfaces where CdS was evaporated onto undoped ZnO (open squares) or (Zn,Mg)O (open triangles) show a linear dependence on BEVB (Zn 2p3/2 ). Using ordered ZnO ﬁlms as substrates (BEVB (Zn 2p3/2 ) ∼ 1018.5 eV; see Sect. 4.2.3.3) results in a valence band oﬀset of ∆EVB = 1.2 ± 0.1 eV, which agrees with the theoretical value of Wei and Zunger [103]. The larger values of BEVB (Zn 2p3/2 ) correspond to ZnO layers deposited at room temperature in pure Ar, which show a high local disorder. Using such ﬁlms as substrates for CdS deposition results in a valence band oﬀset of ∆EVB = 0.85 ± 0.1 eV. The disorder, therefore, shifts the valence band maximum of ZnO upwards in energy. This leads both to a smaller valence band oﬀset and to a larger binding energy of the core levels with respect to the valence band maximum BEVB (Zn 2p3/2 ). The valence band oﬀset determined for a ZnO ﬁlm prepared with standard deposition conditions (room temperature, 100 %Ar) is smaller than the one obtained when the ZnO ﬁlm is deposited with the addition of oxygen to the sputter gas. An inﬂuence of the ionization potential of the ZnO surface on the band alignment might therefore also be considered as origin for the variation in band alignment. The ionization potential of the ZnO ﬁlm deposited with oxygen is ∼0.8 eV larger for standard deposition conditions (see Sect. 4.2.3.2). However, a higher substrate temperature does

160

A. Klein and F. S¨ auberlich

not lead to a larger ionization potential (see Fig. 4.2.3.2) but also results in a larger valence band oﬀset. It is, therefore, not likely that the diﬀerent ionization potential is the origin for the variation in band alignment. This agrees with theoretical considerations [119], which indicate that a variation of band alignment with surface orientation is only possible for interfaces between nonisovalent semiconductors and, therefore, not for the interface between the two II–VI semiconductors ZnO and CdS. The experimental result of Ruckh et al. [102], which is obtained for comparable preparation conditions, deviates from the general behavior observed in our own data. The reason for this deviation is not clear. 2. The interfaces prepared by sputter deposition of ZnO (ﬁlled square) or (Zn,Mg)O (ﬁlled triangles) exhibit a valence band oﬀset of ∆EVB = 1.2 eV. The ZnO and (Zn,Mg)O ﬁlms were prepared at room temperature in pure Ar and therefore exhibit a large disorder and a large BEVB (Zn 2p3/2 ). Compared with the interface with reverse deposition sequence, the oﬀset is ∼0.35 eV larger. This indicates a rather strong inﬂuence of the deposition sequence on the band alignment at the CdS/ZnO interface, which is most likely related to the amorphous nucleation layer when ZnO is deposited onto CdS. Unfortunately, there are currently no band alignments for the CdS/ZnO interface available, where undoped ZnO is deposited using other deposition parameters or deposition techniques. Such data would be important to distinguish between diﬀerent inﬂuences on the band alignment. The use of higher substrate temperatures or oxygen in the sputter gas during ZnO deposition should lead to ordered ZnO ﬁlms with a smaller BEVB (Zn 2p3/2 ). It would be interesting to know if the resulting band alignment shows the same dependence as for the reverse deposition sequence. In this case the valence band oﬀsets should follow the upper dashed line in Fig. 4.23 (“?”) and a ∆EVB of ∼1.6 eV should result. 3. The valence band oﬀsets for deposition of ZnO:Al on CdS are ∼1.6 eV when the ZnO:Al ﬁlms are prepared using pure Ar, which leads to degenerately doped material (CSZA-A and CSZA-D). Deposition of ZnO:Al ﬁlms with a sputter gas containing 10 % O2 results in a low doped material (see Sect. 4.2.3.1). For such deposition conditions (CSZA-B) a valence oﬀset of ∼1.2 eV is obtained. The diﬀerence of 0.4 eV is related to Fermi level pinning in the CdS substrate, which becomes evident by plotting the evolution of the valence band maxima in dependence on deposition time (left graph in Fig. 4.24). The Fermi level pinning is expressed by the observation that the valence band maximum binding energy in CdS lies always between EF − EVB = 1.8 − 2.2 eV. These limits hold for a large set of experiments performed in the surface science group in Darmstadt. They are observed without exception for any used CdS source temperatures (growth rate), substrate temperatures, and substrate materials as ZnO, SnO2 , CdTe, and CuInSe2 .

4 ZnO Surfaces and Interfaces CdS/ZnO:Al

161

ZnO:Al/CdS

1.5

CdS

1.8eV 2.2eV

2.5

1.6eV

1.2eV

1.37eV 1.43eV

3.0

CSZA-A CSZA-B CSZA-D

3.5

ZACS-A ZACS-D

4.0 0

50 100 150 200 250 0 deposition time [sec]

10

20

30

ZnO:Al

EF - EVB [eV]

2.0

40

deposition time [min]

Fig. 4.24. Evolution of CdS and ZnO valence band maxima positions during sputter deposition of ZnO:Al onto CdS (left) and during evaporation of CdS onto ZnO:Al (right). The Fermi level in CdS is always within 1.8–2.2 eV above the valence band maximum (Fermi level pinning)

The Fermi level pinning is related to the defect structure of the material and therefore to the preparation of the CdS ﬁlms.9 The particular defects, which lead to the pinning of the Fermi level in CdS, are, however, not yet identiﬁed. Since the position of the Fermi level in the CdS substrate is obviously restricted to 1.8–2.2 eV, the band alignment is determined by the Fermi level position in the ZnO:Al ﬁlm at low coverage. This can also be extracted from Fig. 4.24. The Fermi level at very low coverage (≤2 nm) cannot be determined unambigously because of the diﬀerent evolution of the Zn 2p and O 1s binding energies, which has been attributed to the amorphous nucleation layer (see Sect. 4.3.2). A unique Fermi level position in the growing ZnO ﬁlm can be determined only for deposition times ≥100–150 s, which correspond to a ﬁlm thickness of ∼2 nm. The obtained values of EF − EVB ≈ 3.8 eV for degenerately doped ZnO:Al deposited in pure Ar and of EF − EVB ≈ 3.0 eV for undoped ZnO:Al deposited with 10 % oxygen are close to the values obtained for very thick ﬁlms. Accordingly, there are almost no further changes of the Fermi level position for larger deposition times. This corresponds to a very fast establishing of the bulk Fermi level position in growing ZnO:Al ﬁlms, which must be associated with a high defect concentration. It is suggested that this behaviour is related to the amorphous nucleation layer. 9

All CdS ﬁlms were evaporated from resistively heated Al2 O3 crucibles using 99.999 % purity CdS source material.

162

A. Klein and F. S¨ auberlich

The Fermi level pinning in CdS and the fast establishing of the Fermi level in ZnO:Al determine the band alignment at the interface: (1) The valence band oﬀset determined for the CdS/ZnO:Al interfaces with degenerately doped ZnO ﬁlms (EF − EVB ≥ 3.8 eV) cannot be lower than 3.8 − 2.2 = 1.6 eV. This is exactly the value derived from the corresponding experiments. (2) The valence band oﬀset for ZnO:Al ﬁlms deposited with addition of oxygen (EF − EVB ≤ 3.0 eV) cannot be larger than 3.0−1.8 = 1.2 eV, which is again the value determined in the experiment. 4. The experiments where ZnO:Al is used as substrate for the deposition of CdS give similar valence band oﬀsets of ∆EVB = 1.4 ± 0.1 eV for degenerately doped ZnO:Al and for ﬁlms where oxygen has been added to the sputter gas. Apparently, Fermi level pinning does not contribute to the band alignment in this case. This is related to the large change of the Fermi level in the ZnO:Al substrates. The large diﬀerence of the Fermi level position of degenerately doped ZnO:Al (EF − EVB = 3.8 eV) and of compensated ZnO:Al (2.8 eV) is strongly decreased during CdS deposition onto the substrates (see right part of Fig. 4.24). Hence the Fermi level is not pinned in the ZnO:Al substrates but can change considerably during interface formation. This supports that the fast establishing of the Fermi level position in growing ZnO:Al ﬁlms described in point 3 is related to the amorphous nucleation layer. The valence band oﬀsets determined for the ZnO:Al/CdS interfaces (1.4 ± 0.1 eV) are 0.2–0.4 eV larger than the values obtained for interfaces where undoped ZnO or (Zn,Mg)O ﬁlms have been used as substrate. This points toward an inﬂuence of the Al content in the ZnO ﬁlm on the band alignment. An explanation for this cannot be given yet. 4.3.4 Summary of CdS/ZnO Interface Properties No interface reaction is observed at the CdS/ZnO interfaces. Even the sputter deposition of ZnO onto CdS with or without oxygen in the sputter gas does not lead to an increased reactivity. This is most likely related to the poor ability of the CdS and ZnO surfaces to dissociate oxygen atoms (compare Sect. 4.2.2.2 and discussion of the reactivity in Sects. 4.4 and 4.5). The results presented in this section further illustrate that there is a considerable dependence of the band alignment at the CdS/ZnO interface on the details of its preparation. An important factor is the local structure of the ZnO ﬁlm. There is considerable local disorder when the ﬁlms are deposited at room temperature in pure Ar, deposition conditions that are often used in thin ﬁlm solar cells. It is recalled that the disorder is only on a local scale and does not aﬀect the long range order of the ﬁlms, as obvious from clear X-ray diﬀraction patterns recorded from such ﬁlms (see discussion in Sect. 4.2.3.3). Growth of sputter deposited ZnO on CdS always results in an amorphous nucleation layer at the interface. The amorphous nucleation layer aﬀects the valence band oﬀset.

4 ZnO Surfaces and Interfaces

163

To give an individual value for the band alignment is not possible. Structurally well-ordered interfaces, which are obtained e.g., by deposition of CdS onto ZnO layers deposited at higher temperatures and/or with the addition of oxygen to the sputter gas, show a valence band oﬀset of ∆EVB = 1.2 eV in good agreement with theoretical calculations [103]. Sputter deposition of undoped ZnO at room temperature in pure Ar onto CdS also leads to a valence band oﬀset of 1.2 eV. In view of the observed dependencies of the band oﬀsets this agreement is fortuitous, as the inﬂuence of the local disorder and of the amorphous nucleation layer most likely cancel each other. The amorphous nucleation layer has the consequence that the Fermi level of the growing ZnO ﬁlms reaches its equilibrium value already at very low thickness (∼2 nm). This is particularly important for ZnO:Al ﬁlms, where the Fermi level changes by more than 1 eV upon the addition of oxygen to the sputter gas. The amorphous nucleation layer, therefore, substitutes the space charge layer, which is usually necessary for charge equilibration at the interface. This important eﬀect is illustrated in Fig. 4.25. The absence of large band bendings is also evident for the CdS substrates where the Fermi level is restricted to a narrow energy region in the upper part of the band gap. As there is no (large) bending of the bands in the CdS substrate and in the ZnO ﬁlm, the band alignment at the interface is mainly determined by the Fermi level in the materials and no longer by the chemical bonds at the interface as in the case of structurally well-ordered interfaces. An important technological consequence of the amorphous nucleation layer is, therefore, that it is essential to control the Fermi level position in the CdS

ZnO

CdS

ZnO

CdS

ZnO

∆ECB

ECB EF

amorphous nucleatio n layer

2.4eV 3.3eV

EVB ∆EVB (a)

(b)

(c)

Fig. 4.25. Inﬂuence of the amorphous nucleation layer of the ZnO ﬁlm on the band alignment at a hypothetical CdS/ZnO interface: (a) CdS and ZnO before contact; (b) in contact with charge equilibrium established by space charge layers; (c) in contact with equilibrium established by charges localized in an amorphous ZnO nucleation layer

164

A. Klein and F. S¨ auberlich

substrate and in the ﬁlm to establish a desired band alignment. This might be a reason why chemical bath deposited CdS layers lead to superior solar cells compared with evaporated CdS ﬁlms. However, Fermi level pinning might also be present in chemical bath deposited CdS ﬁlms. The contribution of Fermi level pinning to the band alignment is one of the most important results of the performed studies. It is also very pronounced when CdS is replaced by In2 S3 . Corresponding results are presented in Sect. 4.5.

4.4 The Cu(In,Ga)Se2 /ZnO Interface In principle, a Cu(In,Ga)Se2 thin-ﬁlm solar cell should be possible without the use of so-called buﬀer layers like CdS. The necessary p-n junction might be provided by the p-type Cu(In,Ga)Se2 absorber and the n-type TCO. Such a cell structure is also advantageous as it requires less production steps. Consequently, there has been considerable eﬀort to prepare Cu(In,Ga)Se2 thinﬁlm solar cells without a chalcogenide buﬀer layer (see Chap. 9 of this book and [120]). Conversion eﬃciencies above 16 % have yet been achieved [121]. For interface studies, it is recommended to use well-deﬁned substrate surfaces. In particular, if the inﬂuence of substrate oxidation is investigated, it is essential to use an unoxidized surface. Oxide-free surfaces of Cu(In,Ga)Se2 can be prepared using the decapping procedure, which is described in detail elsewhere [36, 122]. Basically a Se cap layer is deposited onto the Cu(In,Ga)Se2 surface directly after preparation of the layer before the sample is removed from the vacuum system. The cap layer eﬀectively protects the Cu(In,Ga)Se2 surface against oxidation and can be easily removed from the surface by heating at 300◦ C in the vacuum system at the beginning of the interface experiment. It has been shown that the capping and decapping does not deteriorate the surface as the ﬁlms allow for preparation of solar cells with the same eﬃciency as obtained from identically prepared uncapped ﬁlms [123]. The Se-capped Cu(In,Ga)Se2 ﬁlms used for the present studies were prepared at the Zentrum f¨ ur Sonnenenergie und Wasserstoﬀforschung in Stuttgart, Germany with 30 % of the In substituted by Ga. The ﬁlms are also used for solar cell preparation and yield an energy conversion eﬃciency of ∼14 % [36,123]. Good conversion eﬃciencies are obtained from ﬁlms, which are prepared with a slight Cu deﬁciency (∼22 % Cu instead of the nominal 25 % of Cu in stoichiometric chalcopyrites) [124]. Surfaces of such materials are, however, considerably depleted of Cu and show a surface composition that corresponds to the Cu(In,Ga)3 Se5 vacancy compound with a typical Cu concentration of 11 − 13 % [36, 123, 125]. The importance of this compound for the Cu(In,Ga)Se2 surfaces and interfaces has been pointed out ﬁrst by Schmid et al. [126, 127].

4 ZnO Surfaces and Interfaces

165

4.4.1 Chemical Properties A set of spectra recorded during stepwise deposition of ZnO onto a decapped Cu(In,Ga)Se2 surface is shown in Fig. 4.26. The ZnO ﬁlm has been sputtered from an undoped ZnO target using 15 W dc power but otherwise the same “standard” deposition conditions, which have been used for investigation of the CdS/ZnO interface. On a ﬁrst inspection no changes in the shape of the peaks is observed during deposition. A chemical reaction between Cu(In,Ga)Se2 and ZnO is, therefore, not evident. The O 1s spectra do not show the surface species at low coverage in contrast to the CdS/ZnO interface (see Fig. 4.19). The diﬀerence has already been discussed in Sect. 4.2.2.2, where the absence of the surface species has been attributed to the propensity of the Cu(In,Ga)Se2 surface to dissociate oxygen. However, in case the oxygen can easily dissociate on the surface, an Ga 2p

Cu 2p

Se 3d

In 3d 364

intensity [arb. units]

64 24 8 4 2 1 0

1120 1116 Zn 2p

934

930

O 1s

446

442

56

52

Zn LMM

VB

intensity [arb. units]

364 64 24 8 4 2 1 0

1024

1020

534

530

500 490 binding energy [eV]

4

0

Fig. 4.26. Core levels, Zn LMM Auger level and valence bands recorded using monochromatic Al Kα radiation during sputter deposition of undoped ZnO onto a decapped Cu(In,Ga)Se2 sample. The deposition times are indicated in seconds

166

A. Klein and F. S¨ auberlich Cu 2p

normalized FWHM

1.5

Ga 2p In 3d

1.4

Se 3d

1.3 1.2 1.1 1.0 0

20

40

60

deposition time [s] Fig. 4.27. Normalized full widths at half maximum of substrate core levels in dependence on deposition time for the spectra shown in Fig. 4.26

oxidation of the substrate is to be expected. A ﬁrst indication of a stronger chemical interaction between the deposited ZnO and the Cu(In,Ga)Se2 surface is provided by the full widths at half maximum (FWHM) of the substrate core-level emissions. These are given in Fig. 4.27. The values are normalized for better comparison. A considerable increase of the FWHM is particularly observed for the Ga 2p and to a lesser extend also for the In 3d level. The small changes of the Ga and In core levels with changes in chemical environment is well known in surface science. A higher sensitivity to chemical interactions is usually obtained by the Auger levels. The Zn LMM Auger level of the growing ZnO ﬁlm is included in Fig. 4.26. It shows the same structure independent on ZnO ﬁlm thickness, indicating no substantial changes in chemical environment. In particular, there is no evidence for the formation of Zn–Se bonds, which are expected at ∼2 eV lower binding energies and which have been observed by Loreck et al. for a diﬀerent absorber composition and interface preparation [128]. Figure 4.28 shows the substrate Auger emissions. No noticeable changes during ZnO growth are observed in the Cu LMM and the Se MNN levels. A small additional emission occurs at higher binding energy with increasing ZnO thickness in the In MNN level. This is a clear indication for the formation of In–O bonds. An additional emission also grows in the Ga LMM emission. Hence the formation of Ga–O bonds is also indicated. The formation of Ga–O and In–O bonds is thermodynamically favorable compared with Zn–O bonds because of the larger formation enthalpies of Ga2 O3 (6.15 eV/Ga atom [129]) and In2 O3 (4.8 eV/In atom [130]) compared with 3.6 eV for ZnO. Nevertheless, the oxidation of the Cu(In,Ga)Se2 substrate is not very pronounced and limited to the topmost surface layers as estimated from the given spectra.

4 ZnO Surfaces and Interfaces In MNN

Na 1s

Cu LMM

ox

Ga LMM

167

Se MNN

intensity [arb. units]

ox

1090 1080 1070 575

570

565

425

415

180

160

binding energy [eV] Fig. 4.28. Auger levels recorded using monochromatic Al Kα radiation during sputter deposition of undoped ZnO onto a decapped Cu(In,Ga)Se2 sample. Vertical dashed lines indicate line positions from the clean substrate and from additional components occurring during ZnO deposition. The latter are indicated by “ox.” The In MNN spectra also includes the Na 1s signal, which is due to diﬀusion of Na from the soda lime glass substrate

Chemical interactions at the interface are further reﬂected in the decay of the substrate emissions with increasing ZnO ﬁlm thickness. Because of the dependence of the photoelectron escape depth on kinetic energy [31], a faster attenuation of deeper bound levels is to be expected. This contrasts with the experimental observation, which is presented in Fig. 4.29. As expected the slowest attenuation with increasing ZnO deposition is obtained for the Se 3d level, followed by sligthly faster attenuation of the In 3d level. The strongest attenuation is observed for the Cu 2p level and not for Ga 2p as expected from its lower kinetic energy. In general, the fast attenuation of the substrate core levels suggests a layer-by-layer growth mode and no substantial intermixing between the Cu(In,Ga)Se2 substrate and the ZnO ﬁlm. Such an intermixing occurs only at higher substrate temperatures [131]. The decay length obtained from ﬁtting exponential curves to the integrated intensities are plotted in the right graph of Fig. 4.29 in dependence on the kinetic energy of the photoelectrons. Escape depths calculated according to the formula given by Tanuma, Powell, and Penn [37] and by Seah and Dench [38] are shown for comparison. The curves are scaled to match the values of the In 3d and Se 3d levels. The decay length of the Cu 2p level is considerably smaller than the theoretical value, while the one of Ga is considerably larger. This suggests that two mechanisms are responsible for the faster attenuation of the Cu 2p compared with the Ga 2p level: (1) a depletion

A. Klein and F. S¨ auberlich

normalized intensity

1

Cu 2p Ga 2p In 3d Se 3d

6 5 4 3 2

0

20

40

60

deposition time [s]

att. length [arb. units]

168

40

Se3d

35 30

Ga2p

E1/2

In3d

25 TPP

20 15

Cu2p

400

800

1200

kinetic energy [eV]

Fig. 4.29. Normalized integrated intensities (left) of substrate core levels in dependence on deposition time for the spectra shown in Fig. 4.26. The deposition rate is estimated to be ∼2 nm min−1 . The lines in the left graph are obtained by curve ﬁtting of the data to an exponential decay. The derived attenuation times are displayed in the right graph in dependence on electron kinetic energy together with theoretical energy-dependent escape√depth calculated using the formula by Tanuma, Powell, and Penn [37] and using a E law [38]

of Cu from the surface and (2) an enrichment of Ga at the surface. The latter might be explained by the formation of the Ga-oxide species, which could lead to some segregation of Ga. The ﬁrst observation would be in-line with observed Cu depletion at Cu(In,Ga)Se2 surfaces as a result of interface formation [36,132–136], which is related to an upward shift of the Fermi level [134]. An upward shift of the Fermi level is also observed during deposition of ZnO onto de-capped Cu(In,Ga)Se2 (see Sect. 4.4.2), supporting this correlation. However, as the Cu(In,Ga)Se2 surfaces used for this study are already considerably depleted of Cu, a strong further Cu depletion is not expected. In contrast to the tendency for a Cu depletion in the course of ZnO deposition, Lauermann et al. have reported Cu accumulation at the interface between Cu(In,Ga)(S,Se)2 and rf magnetron sputtered (Zn,Mg)O ﬁlms [137]. The observation was attributed to the electric ﬁeld, which is present during the sputter deposition process. For rf sputtering, the electric ﬁeld at the substrate might be considerably larger than for dc sputtering [138], which is used in the experiment presented in Figs. 4.26–4.29. This diﬀerence might explain the discrepancy between the two diﬀerent experiments. However, also the diﬀerent substrate composition and preparation as well as other sputter parameters might contribute to the diﬀerent behavior. Another species might contribute to the chemistry and electronic properties at the interface. As evident from the In MNN spectra shown in Fig. 4.28, there is also sodium present at the surface. The sodium diﬀuses from the soda lime glass substrate during deposition of the Cu(In,Ga)Se2 ﬁlm and has a beneﬁcial eﬀect on the solar cell conversion eﬃciency [139]. As mentioned

4 ZnO Surfaces and Interfaces

169

by Platzer-Bj¨ orkman et al., the Na content might be directly related to the conversion eﬃciency of Cu(In,Ga)Se2 /ZnO solar cells [140]. The spectra in Fig. 4.28 indicate that the Na 1s core level exhibits a slower attenuation compared with the nearby In MNN Auger level. A comparable behavior has been observed at interfaces between CdS and single crystal CuInSe2 substrates, where Na monolayers have been deliberately inserted by vacuum deposition [133] and also between interfaces of CdS and decapped polycrystalline Cu(In,Ga)Se2 ﬁlms [141]. Apparently Na is at least partially dissolved in the growing ZnO ﬁlm. The eﬀect of Na might, therefore, at least partially be related to its inﬂuence on the doping of the ﬁrst ZnO layers. Na may act both as an acceptor and as a donor in ZnO [142]. If Na is predominantly inserted as an acceptor by substituting Zn or as a donor (interstitial Na) depends on the Fermi energy position in the ZnO. At the interface the Fermi energy is close to the conduction band and it is, therefore, more likely that Na is incorporated as an acceptor (see description in Chap. 1). The reduced doping of the ZnO can change the energy band diagram. 4.4.2 Electronic Properties To determine the valence band oﬀset at the interface, the binding energies of the core levels are plotted in dependence on deposition time in Fig. 4.30. Core ~ ~ Zn 2p O 1s Ga 2p Cu 2p In 3d Se 3d

0.8

BE(VB) [eV]

1.2 1.6 2.0 2.4 2.8 3.2 ~ ~

0

16

32

48

64

80

332 364

deposition time [s] Fig. 4.30. Evolution of valence band maxima in dependence on ZnO deposition time as derived from core-level binding energies of the spectra shown in Fig. 4.29. The ZnO ﬁlms were deposited by magnetron sputtering from an undoped ZnO target at room temperature using 15 W dc power. Core level to valence band maxima binding energy diﬀerences are comparable to those presented in Fig. 4.15 for ZnO and to those given in [36] for Cu(In,Ga)Se2 . The diﬀerent evolution of the Zn 2p and O 1s derived valence band positions for ZnO deposition times indicates the presence of an amorphous nucleation layer, as already discussed in Sect. 4.3.2

170

A. Klein and F. S¨ auberlich

level to valence band maximum binding energy diﬀerences are determined from the decapped Cu(In,Ga)Se2 substrate and the thick ZnO ﬁlm, respectively. The values correspond well with those given recently for Cu(In,Ga)Se2 [36] and with those in Fig. 4.15 for undoped ZnO ﬁlms deposited at room temperature in pure Ar. According to our experience, it is more diﬃcult to determine a reliable valence band oﬀset for the Cu(In,Ga)Se2 /ZnO interface than for the CdS/ZnO interface. This is related to the lower substrate core-level intensities because of the presence of multiple cations. The substrate intensity might, therefore, be already completely suppressed when the Zn 2p and the O 1s derived valence band maxima (see ﬁlled circles and squares in Fig. 4.30) reach the same value, and, therefore, reﬂect a proper ZnO valence band maximum (end of the amorphous nucleation layer). This diﬃculty is not present in the data set in Fig. 4.30 and for a deposition time of 64 s a valence band oﬀset of ∆EVB = 2.15 ± 0.1 eV can be determined. In another experiment, we have derived a slightly smaller valence band oﬀset of ∆EVB = 1.98 ± 0.2 eV [70]. The larger uncertainty is due to the above-mentioned diﬃculties. The valence band oﬀsets determined in our group are very close to values reported in literature. Platzer-Bj¨ orkman et al. have determined ∆EVB = 2.2 ± 0.2 eV for ALD10 -ZnO deposited onto CuInSe2 or Cu(In,Ga)Se2 [143, 144]. Weinhardt et al. give a valence band oﬀset for ILGAR11 -ZnO on CuIn(S,Se)2 substrates of ∆EVB = 1.8 ± 0.2 eV [60]. The comparable values for the diﬀerent interface preparation and substrate compositions suggest a rather small variation of the band alignment with these parameters. Figure 4.31 shows ultraviolet photoelectron spectra recorded during the same interface experiment shown in Fig. 4.26. A clear transition from the Cu(In,Ga)Se2 valence band structure with a valence band maximum at ∼0.8 eV binding energy to the ZnO valence band structure with a valence band maximum at ∼3 eV is observed with increasing ZnO deposition. The well-resolved valence band features are enabled by the in situ sample preparation. Also very sharp secondary electron cutoﬀs are obtained, which allow for an accurate determination of work functions. The work functions of Cu(In,Ga)Se2 and ZnO are determined as 5.4 and 4.25 eV, respectively. These result in ionization potentials of 6.15 and 7.15 eV for Cu(In,Ga)Se2 and ZnO. The ionization potential of ZnO is slightly larger than the one usually obtained for such deposition conditions, which lead to a predominant (0001) Zn-terminated surface. The deviation is, however, directly explained by the microstructure of the ﬁlm (see Fig. 4.21). At low coverage the (0001) orientation of the grains is not yet fully developed and a statistical orientation should lead to a larger ionization potential. The evolution of the orientation of the ZnO grains is reﬂected in the strong shift of the secondary electron cutoﬀ for deposition times ≥8 s. At this coverage the ZnO valence band structure is 10 11

Atomic layer deposition. Ion layer gas reaction.

intensity [arb. units]

4 ZnO Surfaces and Interfaces

171

364 64 24 8 4 2 1 0

18

16

14 12

10

8

6

4

2

0

binding energy [eV] Fig. 4.31. UPS valence bands recorded during deposition of undoped ZnO onto decapped Cu(In,Ga)Se2 showing the valence band structure (right) and the secondary electron cutoﬀ (left). The deposition times are indicated in seconds

almost completely developed but the work function still amounts to ∼4.8 eV. An energy band diagram of the Cu(In,Ga)Se2 /ZnO interface is shown in Fig. 4.32. The band alignment at the Cu(In,Ga)Se2 /ZnO interface can be compared with the band alignment expected for Cu(In,Ga)Se2 /CdS/ZnO sequence. The valence band oﬀset at the CdS/ZnO interface is taken as ∆EVB = 1.2 eV, which is obtained for the same ZnO deposition conditions as those used in the investigation of the Cu(In,Ga)Se2 /ZnO interface. The valence band oﬀset at the Cu(In,Ga)Se2 /CdS interface is taken from literature [36,106,123,125]. In the corresponding studies identically prepared decapped Cu(In,Ga)Se2 substrates have been used. CdS was deposited by thermal evaporation as also applied in the studies of the CdS/ZnO interface formation. The transitivity of the band alignment in the Cu(In,Ga)Se2 /CdS/ZnO sequence is excellently fulﬁlled as evident from Fig. 4.33. This indicates that a modiﬁcation of the band alignment by the introduction of the CdS buﬀer layer does not account for the superior conversion eﬃciencies typically obtained with CdS buﬀer layers. However, the high eﬃciencies are obtained with chemical bath deposited and not with evaporated CdS buﬀer layers. So far, Cu(In,Ga)Se2 thin-ﬁlm solar cells with a direct contact between the absorber and the TCO have achieved higher eﬃciencies than solar cells with evaporated CdS buﬀer layers [112, 120].

172

A. Klein and F. S¨ auberlich

Evac CIGS

1.13

ZnO

4.94 3.86

ECB EF EVB

0.05 1.2 3.3 2.15

Fig. 4.32. Energy band diagram of the Cu(In,Ga)Se2 /ZnO interface as determined from photoelectron spectroscopy. The shaded region indicates the amorphous nucleation layer of the ZnO ﬁlm

E ZnO

CdS

CIGS

ZnO

0.9 eV 2.15 eV 1.2 eV EVBM

Fig. 4.33. Energetic positions of valence band maxima in the Cu(In,Ga)Se2 / CdS/ZnO sequence showing the transitivity of band alignment. The valence band oﬀsets for CdS/ZnO and Cu(In,Ga)Se2 /ZnO are discussed in this chapter. The valence band oﬀset for the Cu(In,Ga)Se2 /CdS interface is taken from literature [36, 106, 123, 125]

4.5 The In2 S3 /ZnO Interface 4.5.1 Cu(In,Ga)Se2 Solar Cells with In2 S3 Buﬀer Layers In2 S3 or In2 S3 containing compounds are possible alternatives for the CdS buﬀer layer in Cu(In,Ga)Se2 thin-ﬁlm solar cells [120, 145–148]. The In2 S3 layers are prepared by various techniques as chemical bath deposition [145], thermal evaporation [146], atomic layer deposition (ALD) [147], and magnetron sputtering [148]. Energy conversion eﬃciencies above 16 % have been

4 ZnO Surfaces and Interfaces

absorber

double buffer

CIGS CdS

CIGS

front contact

Inx (OH,S)y i-ZnO

Inx (OH,S)y

173

ZnO:Al

CdS i-ZnO ZnO:Al

type A

type B

Fig. 4.34. Schematic arrangement of buﬀer layers used in the experiments carried out by Nguyen et al. [154, 155]

reached [147]. Depending on deposition technique and conditions Inx Sy shows a variety of compositions and structures [149]. The In–S phase diagram is given by Godecke and Schubert [150]. Here we present results using In2 S3 layers prepared by thermal evaporation of the In2 S3 compound. The substrates were held at room temperature or at 250◦ C during deposition. Using optical transmission we have determined an optical gap of the ﬁlms of Eg = 1.9 eV [107]. This value is smaller than the optical gap of Eg = 2.75 eV determined by Spiering et al. for ALD-In2 S3 [151] but agrees well with optical gaps of Eg = 1.98 eV or 2.0–2.2 eV determined for MOCVD In2 S3 ﬁlms by Nomura et al. [152] and for evaporated ﬁlms by Timoumi et al. [153]. A particular buﬀer layer experiment, carried out by Nguyen et al. [154, 155], is shown in Fig. 4.34. Two diﬀerent combinations of chemical bath deposited CdS and Inx (OH,S)y buﬀer layers were used to fabricate Cu(In,Ga)Se2 thin-ﬁlm solar cells. The experiment was deﬁned in order to identify the interface that leads to poor eﬃciencies if single Inx (OH,S)y buﬀer layers are used. The type A arrangement of the two buﬀer layers with a Cu(In,Ga)Se2 /CdS and an Inx (OH,S)y /ZnO interface results in poor eﬃciencies, while type B arrangement with a Cu(In,Ga)Se2 /Inx (OH,S)y and a CdS/ZnO interface results in a high eﬃciency. This observation strongly suggests that the interface between Inx (OH,S)y and ZnO limits the eﬃciency. 4.5.2 Chemical Properties X-ray photoelectron spectra recorded during interface formation of magnetron sputtered Al-doped ZnO with an evaporated In2 S3 substrate are shown in Fig. 4.35. The In2 S3 substrate has been deposited at 250◦ C substrate temperature and the ZnO:Al was deposited at room temperature in pure Ar, resulting in a degenerately doped ﬁlm. The valence band maximum after the last deposition step (not shown) is at EF − EVB = 3.9 ± 0.1 eV.

174

A. Klein and F. S¨ auberlich In 3d

512

S 2p

ox

In MNN

intensity [arb. units]

256 128 64 32 16 8 4 2 1 0

448 446 444 442

168

164

Zn 2p

160

1075 Zn LMM

intensity [arb. units]

O 2p

1085

1024

1020

534

530

505 500 495 490

binding energy [eV] Fig. 4.35. Core levels and cation Auger levels of an In2 S3 substrate during sputter deposition of Al-doped ZnO (experiment ISZA-B). The deposition times are indicated in seconds. In2 S3 was deposited at 250◦ C and ZnO:Al at room temperature in pure Ar. Reproduced with permission from [136]

Degenerate doping is further indicated by the broad structure of the Zn LMM Auger line after the last deposition step (compare Sect. 4.2.2). From the core level and Zn LMM Auger level shown in Fig. 4.35 no chemical reactions at the interface are evident. However, the In MNN Auger level shows an additional emission at higher binding energy with increasing ZnO deposition. Compared with the deposition of ZnO onto Cu(In,Ga)Se2 (Fig. 4.28), the oxidation of the In is more pronounced. The O 1s level shows no high binding energy species at low coverage, as also observed for the Cu(In,Ga)Se2 substrate (Fig. 4.26). According to the discussion in Sect. 4.2.2.2, this indicates that the In2 S3 surface facilitates oxygen

4 ZnO Surfaces and Interfaces

175

dissociation. This observation concurs well with substrate oxidation observed in the In MNN Auger level. Interface formation between In2 S3 substrates and ZnO:Al has also been studied with 10 % oxygen added to the sputter gas during ZnO:Al deposition. Corresponding X-ray photoelectron spectra are shown in Fig. 4.36. The In2 S3 substrate has been deposited at 250◦ C substrate temperature and the ZnO:Al was deposited at room temperature. This results in a highly compensated ﬁlm. The valence band maximum after the last deposition step (not shown) is at EF −EVB = 2.7±0.1 eV. A low doping of the ﬁlms is further indicated by the sharper features of the Zn LMM Auger line after the last deposition step compared with the spectra obtained for the highly doped ﬁlm in Fig. 4.35. 729

In 3d

S 2p

ox

In MNN

intensity [arb. units]

243 81 27 9 3 1 0

448 446 444 442

168

164

160 O 1s

1085

1075 Zn LMM

intensity [arb. units]

Zn 2p

1024

1020

536

532

528

505 500 495 490

binding energy [eV] Fig. 4.36. Core levels and cation Auger levels of an In2 S3 substrate during sputter deposition of Al-doped ZnO (experiment ISZA-C). The deposition times are indicated in seconds. In2 S3 was deposited at 250◦ C and ZnO:Al at room temperature in a sputter gas containing 10 % O2

176

A. Klein and F. S¨ auberlich

Again no evident chemical changes in the core level and Zn LMM Auger emissions are observed during ZnO deposition. The broadening of the In MNN Auger line is ascribed to the substrate oxidation. Compared with the interface where ZnO has been deposited without oxygen in the sputter gas, the broadening of the In MNN Auger level is not more pronounced. Hence, substrate oxidation is not enhanced when oxygen is added to the sputter gas. This further supports the conclusion that substrate oxidation is not governed by the energy of the deposited particles but rather by the ability of the substrate surface to dissociate oxygen. Interface formation between In2 S3 and ZnO has also been studied for the reverse deposition sequence with Al-doped ZnO ﬁlms used as substrates [70]. In this case, only degenerately doped substrates were used. Photoemission spectra indicate no chemical reactivity at the surface. 4.5.3 Electronic Properties The valence band oﬀsets between In2 S3 and ZnO are determined from the substrate and overlayer core-level binding energies in dependence on deposition time. Core-level binding energies with respect to the valence band maximum are determined from the substrate and the thick overlayer, respectively. The values for ZnO:Al are within the experimental uncertainty of ±0.1 eV the same as those presented in Fig. 4.15. Values for In2 S3 ﬁlms deposited at room temperature are 443.7 ± 0.1 eV for In 3d and 160.4 ± 0.1 eV for S 2p, respectively. Values for In2 S3 ﬁlms deposited at 250◦C are 444.0 ± 0.1 eV for In 3d and 160.85 ± 0.1 eV for S 2p, respectively. The diﬀerence between the ﬁlms deposited at diﬀerent substrate temperatures indicates a signiﬁcant change in the density of states, which could be related to a diﬀerent composition and/or structure of the ﬁlms. However, In 3d to S 2p intensity ratios for In2 S3 ﬁlms deposited at room temperature and at 250◦C are not signiﬁcantly diﬀerent. Therefore, structural diﬀerences most likely account for the variation of the core-level binding energies with respect to the valence band maximum. It has to be expected that the structural diﬀerences also aﬀect the band gaps of the In2 S3 ﬁlms. Band gaps were so far only determined for the ﬁlms deposited at room temperature [107]. The valence band oﬀsets between In2 S3 and ZnO:Al determined from Fig. 4.37 are ∆EVB = 2.78 ± 0.2 eV and ∆EVB = 1.86 ± 0.2 eV for the interface with degenerately doped and with oxygen compensated low doped ZnO:Al, respectively. There is almost no change of the valence band maximum binding energy in the substrate in the course of ZnO:Al deposition. No band bending is, therefore, introduced in the substrate by contact formation. As both substrates show comparable Fermi level positions, only the diﬀerent Fermi level positions in the ZnO:Al ﬁlms account for the diﬀerent valence band oﬀsets. This situation is comparable to those observed at the CdS/ZnO:Al interface (see left graph in Fig. 4.24). However, the Fermi level in In2 S3 changes even less than in CdS. Therefore, the diﬀerence in ∆EVB

4 ZnO Surfaces and Interfaces

1.0

ISZA-B

177

ISZA-C

BE(VB) [eV]

1.5 2.0

1.86±0.2 eV

2.78±0.2 eV In 3d Sp Zn 2p O 1s

2.5 3.0 3.5 4.0 0

100

200

300

0

100

200

300

deposition time [sec] Fig. 4.37. Evolution of valence band maxima of In2 S3 and ZnO:Al for the two experiments displayed in Figs. 4.35 (left) and 4.36 (right). The diﬀerence between the curves derived from the Zn 2p and O 1s level at low coverage indicates the presence of an amorphous nucleation layer. Reproduced with permission from [136]

between In2 S3 and highly doped or low doped ZnO:Al amounts to ∼0.9 eV, which is considerably larger than for CdS (∼0.4 eV). The larger diﬀerence is a result of the stronger Fermi level pinning in In2 S3 compared with CdS. Energy band diagrams for diﬀerent In2 S3 /ZnO and ZnO/In2 S3 interfaces are summarized in Fig. 4.38. Vacuum energies are not included, as no UPS was available during the experiments. Hence, it was not possible to determine work functions and ionization potentials for the deposited ﬁlms. Interface dipole potentials can, however, be estimated using ionization potentials of 7.1 eV for ZnO:Al according to Sect. 4.2.3.2 and an identical value for In2 S3 [107]. This results in interface dipole potentials of 1.8–2.8 eV for the diﬀerent investigated interfaces. These large values indicate a considerable density of states at the interface, which would explain the strong inﬂuence of the In2 S3 /ZnO interface on the solar cell performance [154, 155]. The largest diﬀerence in valence band oﬀset is observed for the two experiments presented in detail above. Valence band oﬀsets for the other three interfaces are very similar and amount to 2.3–2.4 eV. All three interfaces leading to this valence band oﬀset are prepared using the same deposition conditions: ZnO:Al deposition at room temperature in pure Ar leading to degenerately doped ZnO:Al and In2 S3 deposition at room temperature. The diﬀerence in ∆EVB compared with the interface in which In2 S3 has been deposited at 250◦ C substrate temperature (the second diagram from the left in Fig. 4.38) is the diﬀerent Fermi energy position in the In2 S3 ﬁlm. For all investigated interfaces the valence band oﬀset can be estimated by the alignment of the Fermi levels of thick In2 S3 and ZnO:Al prepared under the same conditions used in the interface experiment. This is related to the very small band bending observed at the interfaces (≤0.2 eV) and concurs

178

A. Klein and F. S¨ auberlich In2S3 ZnO RT

In2S3 ZnO

In2S3 ZnO

250˚C

250˚C +O2

ZnO

In2S3

ZnO

RT

In2S3 RT

ECB EF 1.1

1.6

1.1

1.3

1.4

2.7 EVB

3.9 2.4

3.9 2.8

1.9

3.8 4.1 2.3

2.4

Fig. 4.38. Energy band diagrams at In2 S3 /ZnO interfaces as determined from photoelectron spectroscopy. The material used as substrate during interface formation is shown to the left. In2 S3 ﬁlms were deposited by evaporation onto substrates held either at room temperature or at 250◦ C. ZnO:Al ﬁlms were prepared by dc magnetron sputtering at room temperature in pure Ar or with 10 % O2 in the sputter gas as indicated at the top. All values are given in electronvolt. Band bending at the interface is ≤0.2 eV in all experiments. Because of the uncertainty in the band gap, the conduction band positions of In2 S3 are given as dotted lines. Reproduced with permission from [136]

well with the large dipole potentials. The behavior is illustrated in the top part of Fig. 4.39. Figure 4.39 also shows the alignment of valence bands at CuGaSe2 /CdS/ZnO:Al and CuGaSe2 /In2 S3 /ZnO:Al interfaces. CuGaSe2 was chosen as no experimental determination of the band alignment between Cu(In,Ga)Se2 and In2 S3 is available. For the determination of the valence band oﬀset at the CuGaSe2 /CdS and CuGaSe2 /In2 S3 interface the respective contact materials were evaporated at room temperature onto decapped CuGaSe2 surfaces [106, 107]. The valence band oﬀsets at the CdS/ZnO:Al and In2 S3 /ZnO:Al are taken from this chapter for deposition of the ZnO:Al ﬁlms at room temperature in pure Ar. A deviation of the valence band oﬀsets from transitivity of ∼0.6 eV is evident. The deviation can be attributed to the Fermi level pinning at the In2 S3 /ZnO interface. Of the energy band diagrams given in Fig. 4.38, the band alignment most suitable for the Cu(In,Ga)Se2 thin-ﬁlm solar cell is provided by the interface with the smallest valence band oﬀset (middle diagram). Irrespective of the uncertainty in the In2 S3 band gap, this alignment has the smallest conduction band oﬀset and hence the largest separation between the In2 S3 valence band maximum and the ZnO conduction band minimum. This situation should be favorable with respect to the suppression of recombination

4 ZnO Surfaces and Interfaces

In2S3 RT

179

ZnO:Al

250°C

Ar

O2 ELB

ELB EF

EF 1,1

EVB

1,6 1,6 2,3

2,7

2,8

EVB

3,9

ZnO:Al

CdS 1 eV

1,6 eV EVB

CGS

In2S3

ZnO:Al

0,8 eV

2,4 eV 0,6 eV

Fig. 4.39. Top: Estimation of band alignment (middle section) from the Fermi level positions measured at the surfaces of thick In2 S3 and ZnO:Al ﬁlms in dependence on deposition conditions. Bottom: Energy level alignment in the system CuGaSe2 -In2 S3 -CdS-ZnO:Al. Valence band oﬀsets for CdS/ZnO:Al and In2 S3 /ZnO:Al are taken from the results presented in this Chap. for ZnO:Al ﬁlms deposited at room temperature in pure Ar. Valence band oﬀsets for CuGaSe2 /CdS and CuGaSe2 /In2 S3 are taken from literature [106, 107]

at the In2 S3 /ZnO interface. In addition, the smallest conduction band oﬀset also allows for larger open circuit voltages [156]. Hence, better solar cell eﬃciencies are expected for a In2 S3 deposition temperature around 250◦C and the use of undoped ZnO ﬁlms. This expectation agrees well with the preparation conditions for optimized Cu(In,Ga)Se2 solar cells applying In2 S3 buﬀer layers [146, 148]. Acknowledgement. The work presented here was supported by the German Federal Ministry of Education and Research (BMBF) in the framework of the ZnO

180

A. Klein and F. S¨ auberlich

network project (project No. 01SF0034). The discussions and collaborations with the contributing groups permanently stimulated this work. Our work would not have been possible without the continuous support and encouragement by Wolfram Jaegermann, head of the surface science group in Darmstadt. We further acknowledge the experimental contributions of Frauke R¨ uggeberg, Dr. Gutlapalli Venkata Rao, Christoph K¨ orber, and Juan Angel Sans. We also thank Robert Kniese and Michael Powalla from Zentrum f¨ ur Sonnenenergie und Wasserstoﬀforschung for providing the Se-capped Cu(In,Ga)Se2 thin ﬁlms.

References 1. S.M. Sze, Physics of Semiconductor Devices, 2nd edn. (Wiley, New York, 1981) 2. W. M¨ onch, Electronic Properties of Semiconductor Interfaces (Springer, Berlin Heidelberg New York, 2003) 3. A. Franciosi, C.G. van de Walle, Surf. Sci. Rep. 25, 1 (1996) 4. M. Peressi, N. Binggeli, A. Baldereschi, J. Phys. D 31, 1273 (1998) 5. E.T. Yu, J.O. McCaldin, T.C. McGill, Solid State Phys. 46, 1 (1992) 6. R.T. Tung, J. Vac. Sci. Technol. B 11, 1546 (1993) 7. H. Kroemer, J. Vac. Sci. Technol. B 11, 1354 (1993) 8. F. Capasso, G. Margaritondo (eds.), Heterojunction Band Discontinuities (North-Holland, Amsterdam, 1987) 9. V. Heine, Phys. Rev. 138, 1689 (1965) 10. S.G. Louie, J.R. Chelikowsky, M.L. Cohen, Phys. Rev. B 15, 2154 (1977) 11. S. Kurtin, T.C. McGill, C.A. Mead, Phys. Rev. Lett. 22, 1433 (1969) 12. K. Ip, G.T. Thaler, H. Yang, S.Y. Han, Y. Li, D.P. Norton, S.J. Pearton, S. Jang, F. Ren, J. Cryst. Growth 287, 149 (2006) 13. J.H. Lim, S.J. Park, in Zinc Oxide: Bulk, Thin Films and Nanostructures, ed. by C. Jagadish, S.J. Pearton (Elsevier, Oxford, 2006), pp. 267–283 14. U. Rau, H.W. Schock, Appl. Phys. A 69, 131 (1999) 15. M.A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner, A. Swartzlander, F.A. Hasoon, R. Nouﬁ, Prog. Photovolt. Res. Appl. 7, 311 (1999) 16. P.H.L. Notten, J.E.A.M. van den Meerakker, J.J. Kelly, Etching of III-V Semiconductors: An Electrochemical Approach (Elsevier, Oxford, 1991) 17. R. Hannemann, A. Mariano, J. Appl. Phys. 34, 384 (1963) 18. R. Wiesendanger, H.J. G¨ untherodt (eds.), Scanning Tunneling Microscopy, vol. I–III, 2nd edn. (Springer, Berlin Heidelberg New York, 1996) 19. C.S. Jiang, F.S. Hasoon, H.R. Moutinho, H.A. Al-Thani, M. Romero, M.M. Al-Jassim, Appl. Phys. Lett. 82, 127 (2003) 20. H. Moormann, D. Kohl, G. Heiland, Surf. Sci. 80, 261 (1979) 21. K. Winer, L. Ley, Phys. Rev. B 36, 6072 (1987) 22. R.C. Neville, C.A. Mead, J. Appl. Phys. 43, 4657 (1972) 23. M. Igalson, H.W. Schock, J. Appl. Phys. 80, 5765 (1996) 24. H.L. Tuller, J. Electroceram. 4, 33 (1999) 25. F. Greuter, G. Blatter, Semicond. Sci. Technol. 5, 111 (1990) 26. J.H. Werner, K. Ploog, H.J. Queisser, Phys. Rev. Lett. 57, 1080 (1986) 27. M. Turcu, I.M. K¨ otschau, U. Rau, J. Appl. Phys. 91, 1391 (2002) 28. K. Nishi, H. Ohyama, T. Siuzuki, T. Mitsuyu, T. Tomimasu, Appl. Phys. Lett. 70, 3585 (1997)

4 ZnO Surfaces and Interfaces

181

29. L.J. Brillson, S. Chang, J. Shaw, R.E. Viturro, Vacuum 41, 1016 (1990) 30. M.J. Hetzer, Y.M. Strzhemechny, M. Gao, M.A. Contreras, A. Zunger, L.J. Brillson, Appl. Phys. Lett. 86, 162105 (2005) 31. D. Briggs, M.P. Seah, Practical Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy (Wiley, New York, 1983) 32. S. H¨ ufner, Photoelectron Spectroscopy (Springer, Berlin Heidelberg New York, 1995) 33. J.R. Waldrop, R.W. Grant, S.P. Kowalczyk, E.A. Kraut, J. Vac. Sci. Technol. A 3, 835 (1985) 34. W. Jaegermann, A. Klein, C. Pettenkofer, in Electron Spectroscopies Applied to Low-Dimensional Materials, ed. by H.I. Hughes, H.P. Starnberg, Physics and Chemistry of Materials with Low-Dimensional Structure (Kluwer, Dordrecht, 2000), pp. 317–402 35. K. Durose, S.E. Asher, W. Jaegermann, D. Levi, B.E. McCandless, W. Metzger, H. Moutinho, P.D. Paulson, C.L. Perkins, J.R. Sites, G. Teeter, M. Terheggen, Prog. Photovolt. Res. Appl. 12, 177 (2004) 36. A. Klein, T. Schulmeyer, in Wide Gap Chalcopyrites, ed. by S. Siebentritt, U. Rau (Springer, Berlin Heidelberg New York, 2006) 37. S. Tanuma, C.J. Powell, D.R. Penn, Surf. Interface Anal. 17, 927 (1991) 38. M.P. Seah, W.A. Dench, Surf. Interface Anal. 1, 2 (1979) 39. H.L. Hartnagel, A.L. Dawar, A.K. Jain, C. Jagadish, Semiconducting Transparent Thin Films (Institute of Physics Publishing, Bristol, 1995) 40. O. Dulub, L.A. Boatner, U. Diebold, Surf. Sci. 519, 201 (2002) 41. U. Diebold, L.V. Koplitz, O. Dulub, Appl. Surf. Sci. 237, 336 (2004) 42. W. M¨ onch, Semiconductor Surfaces and Interfaces (Springer, Berlin Heidelberg New York, 1993) 43. G. Chiarotti (ed.), Physics of Solid Surfaces, Landolt-B¨ ornstein, Numerical Data and Functional Relationships in Science and Technology, NS III 24a (Springer, Berlin Heidelberg New York, 1993) 44. C. Noguera, J. Phys. Condens. Matter 12, R367 (2000) 45. G. Kresse, O. Dulub, U. Diebold, Phys. Rev. B 68, 245409 (2003) 46. O. Dulub, U. Diebold, G. Kresse, Phys. Rev. Lett. 90, 016102 (2003) 47. A. Wander, F. Schedin, P. Steadman, A. Norris, R. McGrath, T.S. Turner, G. Thornton, N.M. Harrison, Phys. Rev. Lett. 86, 3811 (2001) 48. B. Meyer, D. Marx, Phys. Rev. B 67, 035403 (2003) 49. M. Kunat, S.G. Girol, T. Becker, U. Burghaus, C. W¨oll, Phys. Rev. B 66, 081402 (2002) 50. D.J. Chadi, Phys. Rev. Lett. 52, 1911 (1984) 51. M. Moll, A. Kley, E. Pehlke, M. Scheﬄer, Phys. Rev. B 54, 8844 (1996) 52. K.W. Haberern, M.D. Pashley, Phys. Rev. B 41, 3226 (1990) 53. S.C. Chang, P. Mark, J. Vac. Sci. Technol. 12, 629 (1975) 54. D.K. Biegelsen, R.D. Bringans, J.E. Northrup, L.E. Swartz, Phys. Rev. Lett. 65, 452 (1990) 55. R.G. Egdell, J. Rebane, T.J. Walker, D.S.L. Law, Phys. Rev. B 59, 1792 (1999) 56. R.G. Egdell, T.J. Walker, G. Beamson, J. Electron Spectrosc. 128, 59 (2003) 57. V. Christou, M. Etchells, O. Renault, P.J. Dobson, O.V. Salata, G. Beamson, R.G. Egdell, J. Appl. Phys. 88, 5180 (2000) 58. Y. Gassenbauer, R. Schafranek, A. Klein, S. Zafeiratos, M. H¨avecker, A. Knop-Gericke, R. Schl¨ ogl, Phys. Rev. B 73, 245312 (2006)

182

A. Klein and F. S¨ auberlich

59. G. Venkata Rao, A. Klein, (unpublished results) 60. L. Weinhardt, M. B¨ ar, H.J. Muﬄer, C.H. Fischer, M.C. Lux-Steiner, T.P. Niesen, F. Karg, T. Gleim, C. Heske, E. Umbach, Thin Solid Films 431–432, 272 (2003) 61. L. Jing, Z. Xu, X. Sun, J. Shang, W. Cai, Appl. Surf. Sci. 180, 308 (2001) 62. M. Kunat, S.G. Girol, U. Burghaus, C. W¨oll, J. Phys. Chem. B 107, 14350 (2003) 63. G. Sch¨ on, J. Electron Spectrosc. 2, 75 (1973) 64. G.E. Hammer, R.M. Shemenski, J. Vac. Sci. Technol. A 1, 1026 (1983) 65. M.N. Islam, T.B. Ghosh, K.L. Chopra, H.N. Acharya, Thin Solid Films 280, 20 (1996) 66. J.F. Chang, W.C. Lin, M.H. Hon, Appl. Surf. Sci. 183, 18 (2001) 67. T. Becker, M. Kunat, C. Boas, U. Burghaus, C. W¨oll, J. Chem. Phys. 113, 6334 (2000) 68. M. Chen, Z.L. Pei, C. Sun, L.S. Wen, X. Wang, J. Cryst. Growth 220, 254 (2000) 69. F. Stucki, P. Br¨ uesch, F. Greuter, Surf. Sci. 189–190, 294 (1987) 70. F. S¨ auberlich, Ph.D. thesis, Technische Universit¨ at Darmstadt (Mensch und Buch, Berlin, 2006) 71. G. Venkata Rao, F. S¨ auberlich, A. Klein, Appl. Phys. Lett. 87, 032101 (2005) 72. P. Erhart, A. Klein, K. Albe, Phys. Rev. B 72, 085213 (2005) 73. P. Erhart, K. Albe, A. Klein, Phys. Rev. B 73, 205203 (2006) 74. K. Ellmer, J. Phys. D Appl. Phys. 33, R17 (2000) 75. A. Zangwill, Physics at Surfaces (Cambridge University Press, Cambridge, 1988) 76. R.K. Swank, Phys. Rev. 153, 844 (1967) 77. H. Moormann, D. Kohl, G. Heiland, Surf. Sci. 100, 302 (1980) 78. W. G¨ opel, U. Lampe, Phys. Rev. B 22, 6447 (1980) 79. K. Jacobi, G. Zwicker, A. Gutmann, Surf. Sci. 141, 109 (1984) 80. C.T. Campbell, K.A. Daube, J.M. White, Surf. Sci. 182, 458 (1987) 81. V.E. Henrich, P.A. Cox, The Surface Science of Metal Oxides (Cambridge University Press, Cambridge, 1994) 82. U. Meier, Ph.D. thesis, BTU Cottbus (2003) 83. C. Agashe, O. Kluth, J. H¨ upkes, U. Zastrow, B. Rech, M. Wuttig, J. Appl. Phys. 95, 1911 (2004) 84. H. Fujiwara, M. Kondo, Phys. Rev. B 71, 075109 (2005) 85. I. Ivanov, J. Pollmann, Phys. Rev. B 24, 7275 (1981) 86. R. Dorn, H. L¨ uth, M. B¨ uchel, Phys. Rev. B 16, 4675 (1977) 87. S. Brehme, F. Fenske, W. Fuhs, E. Nebauer, M. Poschenrieder, B. Selle, H. Sieber, Thin Solid Films 342, 167 (1999) 88. J.H. Hwang, D.D. Edwards, D.R. Kammler, T.O. Mason, Solid State Ionics 129, 135 (2000) 89. S.P. Harvey, T.O. Mason, Y. Gassenbauer, R. Schafranek, A. Klein, J. Phys. D Appl. Phys. 39, 3959 (2006) 90. F. S¨ auberlich, J. Fritsche, R. Hunger, A. Klein, Thin Solid Films 431–432, 378 (2003) 91. W. Ranke, Phys. Rev. B 27, 7807 (1983) 92. J. Fritsche, D. Kraft, A. Thissen, T. Mayer, A. Klein, W. Jaegermann, Mater. Res. Soc. Symp. Proc. 668, H6.6 (2001)

4 ZnO Surfaces and Interfaces

183

93. T. L¨ oher, A. Klein, Y. Tomm, C. Pettenkofer, W. Jaegermann, Semicond. Sci. Technol. 15, 1 (2000) 94. R. Cebulla, R. Wendt, K. Ellmer, J. Appl. Phys. 83, 1087 (1998) 95. G.J. Exarhos, S.K. Sharma, Thin Solid Films 270, 27 (1995) 96. I. Sieber, N. Wanderka, I. Urban, I. D¨ orfel, E. Schierhorn, F. Fenske, W. Fuhs, Thin Solid Films 330, 108 (1998) 97. E. Mirica, G. Kowach, P. Evans, H. Du, Cryst. Growth Des. 4, 147 (2004) 98. P. Schr¨ oer, P. Kr¨ uger, J. Pollmann, Phys. Rev. B 47, 6971 (1993) 99. C. McGuinness, C.B. Stagarescu, P.J. Ryan, J.E. Downes, D. Fu, K.E. Smith, R.G. Egdell, Phys. Rev. B 68, 165104 (2003) 100. C.L. Dong, C. Persson, L. Vayssieres, A. Augustsson, T. Schmitt, M. Mattesini, R. Ahuja, C.L. Chang, J.H. Guo, Phys. Rev. B 70, 195325 (2004) 101. I. Hamberg, C.G. Granqvist, J. Appl. Phys. 60, R123 (1986) 102. M. Ruckh, D. Schmid, H.W. Schock, J. Appl. Phys. 76, 5945 (1994) 103. S.H. Wei, A. Zunger, Appl. Phys. Lett. 72, 2011 (1998) 104. R. Klenk, Thin Solid Films 387, 135 (2001) 105. M. Gloeckler, A.L. Fahrenbruch, J.R. Sites, in Proceedings of the 3rd World Conference on Photovoltaic Solar Energy Conversion, Osaka, Japan, 2003, p. 2P.D3.52 106. T. Schulmeyer, R. Kniese, R. Hunger, W. Jaegermann, M. Powalla, A. Klein, Thin Solid Films 451–452, 420 (2004) 107. T. Schulmeyer, A. Klein, R. Kniese, M. Powalla, Appl. Phys. Lett. 85, 961 (2004) 108. B. Sp¨ ath, J. Fritsche, F. S¨ auberlich, A. Klein, W. Jaegermann, Thin Solid Films 480–481, 204 (2005) 109. J. Fritsche, A. Klein, W. Jaegermann, Adv. Eng. Mater. 7, 914 (2005) 110. T.M. Duc, C. Hsu, J.P. Faurie, Phys. Rev. Lett. 58, 1127 (1987) 111. E.T. Yu, M.C. Phillips, J.O. McCaldin, T.C. McGill, J. Vac. Sci. Technol. B 9, 2233 (1991) 112. M. Rusu, T. Glatzel, A. Neisser, C.A. Kaufmann, S. Sadewasser, M.C. Lux-Steiner, Appl. Phys. Lett. 88, 143510 (2006) 113. L. Weinhardt, C. Heske, E. Umbach, T.P. Niesen, S. Visbeck, F. Karg, Appl. Phys. Lett. 84, 3175 (2004) 114. J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, Handbook of X-ray Photoelectron Spectroscopy (Physical Electronics, Eden Prairie, 1995) 115. J. Fritsche, S. Gunst, A. Thissen, R. Gegenwart, A. Klein, W. Jaegermann, Mater. Res. Soc. Symp. Proc. 668, H5.1 (2001) 116. F. R¨ uggeberg, A. Klein, Appl. Phys. A 82, 281 (2006) 117. Y. Yoshino, K. Inoue, M. Takeuchi, K. Ohwada, Vacuum 51, 601 (1998) 118. A. Ohtomo, M. Kawasaki, I. Ohkubo, H. Koinuma, T. Yasuda, Y. Segawa, Appl. Phys. Lett. 75, 980 (1999) 119. W.R.L. Lambrecht, B. Segall, Phys. Rev. B 41, 2832 (1990) 120. D. Hariskos, S. Spiering, M. Powalla, Thin Solid Films 480–481, 99 (2005) 121. T. Negami, T. Aoyagi, T. Satoh, S. Shimakawa, S. Hayashi, Y. Hashimoto, in Proceedings of the 29th IEEE Photovoltaic Specialists Conference, New Orleans, USA, 2002, p. 656 122. R. Hunger, T. Schulmeyer, A. Klein, W. Jaegermann, K. Sakurai, A. Yamada, P. Fons, K. Matsubara, S. Niki, Surf. Sci. 557, 263 (2004)

184

A. Klein and F. S¨ auberlich

123. T. Schulmeyer, R. Hunger, W. Jaegermann, A. Klein, R. Kniese, M. Powalla, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 2O.C10.05 124. R. Herberholz, U. Rau, H.W. Schock, T. Haalboom, T. G¨odecke, F. Ernst, C. Beilharz, K.W. Benz, D. Cahen, Eur. Phys. J.: Appl. Phys. 6, 131 (1999) 125. T. Schulmeyer, R. Hunger, M. Lebedev, W. Jaegermann, A. Klein, R. Kniese, M. Powalla, Thin Solid Films 480–481, 110 (2005) 126. D. Schmid, M. Ruckh, F. Grunwald, H.W. Schock, J. Appl. Phys. 73, 2902 (1993) 127. D. Schmid, M. Ruckh, H.W. Schock, Appl. Surf. Sci. 103, 409 (1996) 128. C. Loreck, I. Lauermann, A. Grimm, R. Klenk, M. B¨ar, S. Lehmann, S. Sokoll, M.C. Lux-Steiner, F. Erfurth, L. Weinhardt, C. Heske, S. Visbeck, T.P. Niesen, C. Jung, C.H. Fischer, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, pp. 1874–1877 129. R.C. Weast (ed.), CRC Handbook of Chemistry and Physics, 65th edn. (CRC, Boca Raton, 1985) 130. M. Chase, NIST-JANAF Thermochemical Tables, 4th edn. J. Phys. Chem. Ref. Data, Monograph 9 (1998) 131. R. Hunger, P. Fons, K. Iwata, A. Yamada, K. Matsubara, S. Niki, K. Nakahara, H. Takasu, Mater. Res. Soc. Symp. Proc. 668, H8.21.1 (2002) 132. A. Klein, T. L¨ oher, C. Pettenkofer, W. Jaegermann, J. Appl. Phys. 80, 5039 (1996) 133. A. Klein, T. L¨ oher, C. Pettenkofer, W. Jaegermann, in Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 1997, pp. 2068–2071 134. A. Klein, W. Jaegermann, Appl. Phys. Lett. 74, 2283 (1999) 135. A. Klein, J. Fritsche, W. Jaegermann, J.H. Sch¨ on, C. Kloc, E. Bucher, Appl. Surf. Sci. 166, 508 (2000) 136. A. Klein, F. S¨ auberlich, B. Sp¨ ath, T. Schulmeyer, D. Kraft, J. Mater. Sci. 42, 1890 (2007) 137. I. Lauermann, C. Loreck, A. Grimm, R. Klenk, H. M¨onig, M.C. Lux-Steiner, C.H. Fischer, S. Visbeck, T.P. Niesen, Thin Solid Films 515, 6015 (2007) 138. B. Chapman, Glow Discharge Processes (Wiley, New York, 1980) 139. J. Hedstr¨ om, H. Ohlsen, M. Bodeg˚ ard, A. Kylner, L. Stolt, D. Hariskos, M. Ruckh, H.W. Schock, in Proceedings of the 23rd IEEE Photovoltaic Specialists Conference, Louisville, USA, 1993, p. 364 140. C. Platzer-Bj¨ orkman, T. T¨ orndahl, J. Kessler, L. Stolt, in Proceeding of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 2005 141. T. Schulmeyer, A. Klein (unpublished results) ¨ Ozg¨ ¨ ur, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, 142. U. V. Avrutin, S.J. Cho, H. Morko¸c, J. Appl. Phys. 98, 041301 (2005) orkman, J. Lu, J. Kessler, L. Stolt, Thin Solid Films 431–432, 143. C. Platzer-Bj¨ 321 (2003) 144. C. Platzer-Bj¨ orkman, T. T¨ orndahl, D. Abou-Ras, J. Malmstr¨ om, J. Kessler, L. Stolt, J. Appl. Phys. 100, 044506 (2006) 145. D. Hariskos, M. Ruckh, U. R¨ uhle, T. Walter, H.W. Schock, J. Hedstr¨ om, L. Stolt, Sol. Energ. Mater. Sol. Cell. 41–42, 345 (1996) 146. A. Strohm, L. Eisenmann, R.K. Gebhardt, A. Harding, T. Schl¨otzer, D. Abou-Ras, H.W. Schock, Thin Solid Films 480–481, 162 (2005)

4 ZnO Surfaces and Interfaces

185

147. N. Naghavi, S. Spiering, M. Powalla, B. Cavana, D. Lincot, Prog. Photovolt. Res. Appl. 11, 437 (2003) 148. S. Spiering, A. Eicke, D. Hariskos, M. Powalla, N. Naghavi, D. Lincot, Thin Solid Films 451–452, 562 (2004) 149. R. Diehl, R. Nitsche, J. Cryst. Growth 28, 306 (1975) 150. T. Godecke, K. Schubert, Z. Metallkunde 76, 358 (1985) 151. S. Spiering, D. Hariskos, M. Powalla, N. Naghavi, D. Lincot, Thin Solid Films 431, 359 (2003) 152. R. Nomura, K. Konishi, H. Matsuda, Thin Solid Films 198, 339 (1991) 153. A. Timoumi, H. Bouzouita, M. Kanzari, B. Rezig, Thin Solid Films 480, 124 (2005) 154. Q. Nguyen, U. Rau, M. Mamor, K. Orgassa, H.W. Schock, J.H. Werner, in Proceedings of the 17th European Photovoltaic Solar Energy Conference, M¨ unchen, Germany, 2001, p. 1107 155. Q. Nguyen, K. Orgassa, I. K¨ otschau, U. Rau, H.W. Schock, Thin Solid Films 431–432, 330 (2003) 156. A. Yamada, K. Matsubara, K. Sakurai, S. Ishizuka, H. Tampo, P.J. Fons, K. Iwata, S. Niki, Appl. Phys. Lett. 85, 5607 (2004)

5 Magnetron Sputtering of ZnO Films B. Szyszka

5.1 Introduction Glow discharge sputtering is one of the oldest deposition techniques that utilizes an energy input to promote surface diﬀusion at the substrate and thus, to achieve dense and well-adhering coatings at low substrate temperatures. The term “sputtering” means the ejection of atoms from a usually solid target material due to the impact of highly energetic species. These highly energetic species are usually positive ions, which can either be accelerated in the cathode sheath of a plasma discharge or in an ion source. The simplest approach for the deposition of ZnO ﬁlms by sputtering is sketched in Fig. 5.1: A DC glow discharge is ignited between a cathode, which is a planar Zn target, and the anode, which is the chamber of the vacuum system. The system is pumped to a pressure of ∼10 Pa and Ar and O2 are introduced into the system. The metallic target is oxidized, so that Zn and O atoms are sputtered from the target and condense on the substrate, where the ZnO ﬁlm is formed. However, this concept has many drawbacks in terms of ﬁlm properties, deposition rate, and process stability. The development of the so-called “magnetron” sputter sources by Chapin [1] was a breakthrough in the 1970s toward large area high-rate deposition. The key was to increase the sputter current by magnetic conﬁnement of the plasma in front of the target. This feature allows both the deposition rate to be increased and the pressure to be decreased. Both features are crucial for cost eﬀective sputter deposition of high-quality ZnO ﬁlms. For transparent and conductive ZnO-based TCO ﬁlms, however, several further breakthroughs have been necessary: The most crucial point is the control of stoichiometry and phase composition, which are key parameters for eﬃcient doping. Many approaches have been realized: Reactive sputtering from alloy targets allows growth conditions to be varied to a large extent. The metallic targets used are cheap compared to ceramic targets but the drawback is the need for a precise control of the reactive process, which is a delicate but solvable task using advanced process control techniques. Ceramic target sputtering on the other hand permits more robust processes since the metalto-oxygen ratio is deﬁned by the target stoichiometry up to a certain extent. The following sections review the work on magnetron sputtering of ZnO ﬁlms focusing on TCO properties. Section 5.2 outlines the history of ZnO

188

B. Szyszka power supply (cathode)

Zn target +

Ar O+ reactive gas

plasma

Zn O2 O Zn

O Ar ZnO film

substrate

vacuum chamber (anode) pump

Fig. 5.1. Setup for glow discharge sputter deposition of ZnO by reactive DC sputtering of a metallic Zn target in an Ar/O2 atmosphere

sputtering. Section 5.3 is about the basics of magnetron sputtering and the ZnO ﬁlm properties, which can be achieved by magnetron sputtering. Section 5.4 treats the manufacturing technology for large area deposition and Sect. 5.5 gives an overview of emerging developments toward more advanced and lower-cost ZnO sputtering technology.

5.2 History of ZnO Sputtering Sputtering of ZnO ﬁlms has a history covering more than three decades. The technique is now one of the most versatile deposition processes for industrial production of ZnO ﬁlms. First sputtering processes for ZnO deposition were developed in the late 1960s for manufacturing surface acoustic wave devices [2]. The piezoelectric properties of ZnO ﬁlms are crucial for that application and major eﬀorts were made to develop ZnO sputtering processes which enabled c-axis oriented growth, high resistivity and unique termination of the ZnO crystallites [3, 4]. Large-area sputtering of ZnO was established in the ﬁeld of energyeﬃcient glazing in the early 1980’s. At that time, ZnO was used as a dielectric ﬁlm for Ag-based low emissive (low-E) coatings. Coating designs such as ﬂoat glass/ZnO/Ag/blocker/ZnO were implemented by planar cathode reactive sputtering onto large-area glass panes [5,6]. ZnO was chosen as dielectric material because of its high sputtering rate and its suitability for reactive DC sputtering. Even today, ZnO ﬁlms are key components in modern Ag-based coatings for architectural glazing. Nowadays, the most challenging application

5 Magnetron Sputtering of ZnO Films

189

of ZnO for architectural glazing is its use as a seeding layer for the growth of highly conductive Ag and also as a blocker ﬁlm on top of the Ag. Two diﬀerent aspects are important for this application: For the seeding layer, it is the wurtzite phase crystal structure: Appropriately deposited ZnO ﬁlms reveal c-axis oriented textured growth, which promotes the heteroepitaxial growth of Ag. The state of the art is to use a ZnO-based ﬁlms as thin seeding layer (∼2–5 nm thickness) for the silver ﬁlm [7, 8]. The use of ZnO ﬁlms as a nonabsorbing blocker on top of the Ag is enabled by ceramic target magnetron sputtering [9]. Compared to reactive magnetron sputtering, this process allows the particle energy at the surface of the growing ﬁlm to be decreased signiﬁcantly, which is crucial to maintaining the metallic conductivity of the Ag ﬁlm underneath. Transparent and conductive, sputtered ZnO-based ﬁlms with resistivity below 1,000 µΩ cm were ﬁrst reported by Brett and coworkers in the early 1980’s [10]. Breakthroughs such as high-rate magnetron sputtering of ZnO:Al [11] guided the developments toward large-area manufacturing technology, with main emphasize on thin-ﬁlm photovoltaics [12, 13].

5.3 Magnetron Sputtering Magnetron sputtering [14] is a production proven high precision and high-rate coating technology based on simple and rugged components. The technology was developed in the early 1970’s [1]. The ﬁrst applications were the metallization of polymer parts for the automotive industry, the metallization of wafers for microelectronics [15], and the large-area deposition of energyeﬃcient coatings for architectural glazing. Further applications are numerous; major examples are coatings on polymeric web [16] and display glass [17,18], as well as coatings for data storages, such as CD’s, DVD’s [19] and hard disk drives [20]. Magnetron sputtering is also a key technology for wear-resistant coatings on tools and components. Furthermore, magnetron sputtering processes are essential to the industrial manufacturing of thin-ﬁlm solar cells, where transparent electrodes and back contacts are deposited by magnetron sputtering and where major R&D eﬀorts are concentrating on the deposition of semiconductors using this technique. It is a high-vacuum deposition process where the ﬁlm-forming atoms are generated by sputtering from a metallic or compound target plate, which is the cathode of a glow discharge process. The sputtered atoms are transported to the substrate through a low pressure plasma environment. The condensation of mostly neutral atoms and ﬁlm growth is under concurrent bombardment by energetic species from the plasma [21], which promotes nucleation, compound formation, and ﬁlm growth on the substrate [22, 23]. Since the coating material is passed into the vapor phase via momentum exchange caused by energetic particle impact rather than by a chemical or thermal process, virtually any material is a candidate for coating. Films

190

B. Szyszka

containing almost every solid element in the periodic table have been prepared by sputtering. Alloys and compounds can be sputtered such their stoichiometry is preserved, so oxide ﬁlms such as ZnO:Al can be sputtered either reactively from a metallic alloy target, where reactive gas is introduced into the process chamber, or from ceramic compound targets in a pure Ar atmosphere. The process technology of magnetron sputtering has been described extensively in the literature [24, 25]. A noteworthy introduction into magnetron sputtering of optical coatings is given by Herrmann et al. [26]. 5.3.1 Glow Discharge Characteristics The sputtering target is the cathode of a low pressure glow discharge operating in either in a non-reactive (compound target sputtering in Ar) or reactive atmosphere (elemental target sputtering in Ar/O2 mixture). High pressure in the order of 5 − 15 Pa is necessary for sputtering without the support of a magnetic ﬁeld. The target is subject to an ion bombardment by Ar+ and O+ ions, which are generated by electron impact ionization and which are accelerated in the cathode sheath toward the target. Plasma formation and the characteristics of glow discharge processes are subjects of extensive monographs [27, 28]. 5.3.2 Processes at the Target Surface Sputtering of the target material is the fundamental process for the generation of thin ﬁlm forming species. Sputtering is a process whereby material is dislodged and ejected from the surface of a solid target material as a result of momentum exchange associated by energetic particle impact. The impact of particles with energy of several 100 eV gives rise to a collision cascade in the target, which ends up in the emission of a certain amount of target atoms at energies of a few electronvolt as neutral atoms mostly. An overview on the processes at the target surface is given in Fig. 5.2. The properties of the emitted material depend on the bombarding ions, their kinetic energy, incidence angle, atomic mass as well as on the target material and its structure. The most important parameter of the sputtering process is the so-called sputter yield Y which deﬁnes the number of emitted target atoms Ze per incident particles Zi . Y =

Ze Zi

(5.1)

The dependence of sputtering yield on ion energy is shown in Fig. 5.3 for diﬀerent elemental targets. At moderate ion energies in the order of 30 to 1,000 eV, sputtering is characterized by knock on eﬀects where the incident ions collide with a surface atom and these atoms further react with additional atoms. These events may eventually lead to a release of target material atoms.

5 Magnetron Sputtering of ZnO Films

191

Fig. 5.2. Processes occurring at the target surface due to the impact of highly energetic particles (reprinted from [27])

Fig. 5.3. The sputtering yield of Si, Zn, Ti, Cu, and Al for Ar+ bombardment as a function of ion energy. The sputtering yield for Zn at 300, 500, and 1,000 eV is 3.7, 5, and 7, respectively (reprinted from [30])

In this regime, the sputtering depends strongly on the details of the interaction such as the geometric position of the impinging ions on the target surface and the local binding energies. The sputtering yields and the eﬀect of target poisoning due to oxide formation on the target surface as well as oxide

192

B. Szyszka

implantation can be modeled using Monte Carlo simulations implemented in tools such as TRIDYN [29]. At low energy, a threshold is observed. This is due to the fact that the surface binding energy barrier has to be overcome for sputtering to occur. An important consequence is that the sputter yield of an oxidized target surface is much smaller than for a metallic surface. At high energy, the impinging ion has enough energy to break all bonds in its local environment. This is the collision cascade regime, which can be accurately modeled. For the application of ZnO sputtering using Ar as sputter gas and typical deposition conditions with impinging ion energy in the order of 100–1,000 eV, it is remarkable that Zn has the highest sputtering rate known for any elemental target material. Further information on the details of the sputtering process can be found in the extensive literature [31, 32]. 5.3.3 Magnetron Operation The conventional conﬁguration for reactive glow discharge sputtering described in Fig. 5.1 suﬀers from the low ionization eﬃciency of the electrons crossing the gap between the cathode and the anode. Pressures on the order of 10 Pa are necessary to operate the discharge. The discharge current is small and therefore, the growth rate is low. Furthermore, the high pressure gives rise to an unwanted thermalization of the low-energy sputtered target material atoms. High-energy species, on the other hand, such as fast, reﬂected Ar neutrals and negatively charged oxygen ions are not thermalized to the same extent, since the scattering cross sections depends strongly on the particle energy [33]. This eﬀect gives rise to an increased defect density in the growing ﬁlm since the surface diﬀusion promoting low-energy species with kinetic energy up to a few tenth of electronvolt [34] are suppressed, while the unwanted high-energy and thus defect-generating species with higher energies still reach the substrate. Furthermore, the substrate is subject to intense electron bombardment, causing a substantial heat load which is unwanted in many applications. These drawbacks can be circumvented using the magnetron conﬁguration of the sputtering cathode. Magnetron sputtering utilizes magnetic trapping of the electrons to conﬁne the plasma close to the cathode [25]. The magnetic ﬁeld is formed parallel to the cathode surface and perpendicular to the electric ﬁeld, as shown in Fig. 5.4. As a consequence, the electrons which are accelerated in the cathode sheath are forced onto a closed loop drift path parallel to the target surface because of the Lorentz Force. This magnetic trapping of the electrons and the corresponding ambipolar diﬀusion of the ions raises the plasma density in front of the target. A much higher ion current and therefore deposition rate is possible. Furthermore, the pressure can be decreased, which improves

5 Magnetron Sputtering of ZnO Films

a)

193

b)

Fig. 5.4. (a) Operation principle of a planar magnetron cathode. The secondary electrons emitted from target surface are trapped by the magnetic ﬁeld due to the Lorentz force. The result is a plasma torus in front of the target. (b) Planar magnetron cathode of 3.75 m length (from [35])

the ﬁlm properties since less scattering in the gas phase occurs. Also, the thermal load on the substrate is decreased. Magnetron sputtering therefore allows for coating on temperature-sensitive substrates, such as polymers or organic coatings. The plasma discharge can be excited using DC, MF (some tens of kHz) or RF (13.56 MHz) excitation. The positive ions bombard the negativelycharged target electrode which serves as the cathode of the discharge. The ﬁrst use of such a conﬁguration was reported by Penning et al. for sputtering a cylindrical cathode in a coaxial magnetic ﬁeld [36, 37]. However, it took more than 30 years for the invention of the planar magnetron by Chapin [1]. The plasma conﬁnement allows low pressure/high current operation of the discharge. The total pressure can be as low as 0.1 Pa. For a target-tosubstrate distance in the order of 100 mm, direct transport of target material to the substrate occurs in the line of sight at that low pressure. Interactions of sputtered material with the gas phase are limited to a few collisions. As a consequence, much of the kinetic energy of sputtered material can be transferred to the substrate, enhancing surface diﬀusion and reactivity and improving the ﬁlm properties compared to conventional sputtering or evaporation processes. The energetic bombardment of the substrate can be changed using a variety of modiﬁed magnetron sputter processes such as dual magnetron operation with MF plasma excitation [38], RF superimposed DC magnetron sputtering [39] and unbalanced magnetron sputtering [40]. The basic concept of these techniques is to utilize intense low-energy ion bombardment from a dense plasma near the substrate to improve the thin-ﬁlm properties.

194

B. Szyszka

5.3.4 Magnetron Sputtering of ZnO As outlined in Sect. 5.2, many attempts have been made to deposit ZnO-based ﬁlms by sputtering. The compilation of papers in Table 5.1 may serve as a reference to identify useful papers. The classiﬁcation criteria are: plasma excitation used (DC/MF/RF); reactive or ceramic deposition; material deposited and ﬁlm properties relevant for TCO applications. 5.3.4.1 Reactive Magnetron Sputtering In reactive magnetron sputtering [66], metal targets are sputtered in a reactive atmosphere. This approach is not only cost-eﬀective due to the use of metal alloys instead of ceramics but also allows the ﬁlm properties to be varied to a large extent. The fundamental operating principle for reactive sputtering is sketched in Fig. 5.1. The metallic sputter target serves as the cathode of a glow discharge operating in an Ar/O2 gas mixture and the anode can be considered to be the chamber. However, there are other ways to operate the discharge such as medium frequency (MF) dual magnetron and radio frequency (RF), which are described later on. The target is bombarded by high-energy ions which are accelerated toward the cathode. This process releases metal atoms (Zn) from metallic areas of the targets as well as atomic oxygen (O) from oxidized areas of the target surface. The sputtered material condenses on the substrate, where it reacts with the reactive gas and forms and oxide ﬁlm. The reactive sputter process can be considered as a getter pump where the reactive gas is gettered by the reaction with the target material. The pumping speed of this getter pump, however, depends strongly on the state of the target surface since the release of metal atoms by sputtering depends on the sputtering yield. This eﬀect gives rise to the nonlinear process characteristics of reactive magnetron sputtering, as shown in Fig. 5.5 for reactive magnetron sputtering of ZnO:Al ﬁlms. The discharge is operated at constant power using the reactive MF magnetron sputtering process described in [38]. The discharge voltage and total pressure are shown as a function of the oxygen ﬂow rate. A continuous dependence of total pressure and discharge voltage on the oxygen ﬂow rate is observed only for low or high oxygen ﬂow rates, whereas an abrupt noncontinuous change of process parameters is observed at intermediate ﬂow rates. Furthermore, strong hysteresis is observed between increasing and decreasing oxygen ﬂow rates. The three process regimes are related to the state of the target. The nomenclature was proposed by Schiller et al. [67]: Elemental or metallic mode, transition mode, and compound or oxide mode. Metallic Mode – When the partial pressure of the reactive gas is low, the target surface is metallic or partially oxidized. The target coverage is a continuous function of the oxygen partial pressure.

5 Magnetron Sputtering of ZnO Films

195

Table 5.1. Literature survey on magnetron sputtering of ZnO-based TCO ﬁlms Ref. Scope Process Material Goal Year

[41] RDCMS ZnO1−x fd 1985 [42] RDCMS ZnO:Al fd 1999 [40] RDCMS ZnO:Al ZnO:In fd 1991 [11] RDCMS ZnO:Al fd 1994

[43] RDCMS, RMFMS ZnO:Al lac, dd 2003

[44] RDCMS ZnO:In fd 1992 [45] DCMS ZnO1−x dev 1991

Growth conditions Coater and magnetron Process mode and control Target Substrate, TS (◦ C) P/A (W cm−2 ) as (nm s−1 )/ad (nm m min−1 ) Batch, 150 mm planar Baﬄe, substrate RF bias Zn Glass, PET, 80◦ C Batch, 90 mm planar Low power Zn:2 wt% Al Glass, RT 1.6 W cm−2 , 0.8 − 2 nm s−1 Batch Low power Zn target, Al or In pieces Glass, 180 − 220◦ C <2 nm s−1 Batch, 3 inch planar Low power Zn:2 wt% Al Glass, <150◦ C 2.2 W cm−2 <2 nm s−1 In-line, dual magnetron tm, q(O2 ) = f(PEM) 1 400 × 100 mm2 each Zn:2.0 wt% Al Glass, RT <5W cm−2 125 nm m min−1 Batch, 70 mm planar tm, q(O2 ) = f (U) Zn:2.5 at.% In Glass, RT 1.2 nm s−1 Batch, 4 inch planar ZnO, H2 O gas Glass, 300–400◦ C I = 500 mA 0.3–0.6 nm s−1 dev a-Si, pav

Layer properties d (nm), cd (at.%) ρ (Ω cm), ne (cm−3 ), µ (cm2 V−1 s−1 )

700 nm, x = 0.05 1.9 × 10−3 Ω cm, 1.8 × 1020 cm−3 , 18 cm2 V−1 s−1 170–400 nm, 1.4 − 1.8 at.% 6 × 10−4 Ω cm, 1.0 × 1021 cm−3 , 10 cm2 V−1 s−1 700 nm 2 at.% (Al), ∼11 at.% (In) 5.0 × 10−4 Ω cm (Al) 1.4 × 10−3 Ω cm (In) ∼500 nm 1.8–2.2 at.% 4.5 × 10−4 Ω cm 8 × 1020 cm−3 17 cm2 V s−1 1,040 nm, 9.1 × 10−4 Ω cm

∼250 nm, ∼10−2 − 10−3 Ω cm

2,000 nm pav: 1.8 × 10−3 Ω cm

196

B. Szyszka Table 5.1. (continued)

Ref. Scope [46] DCMS ZnO:Al fd 1990 [47] DCMS ZnO:Ga fd 1997 [48] DCMS ZnO:Ga fd 2002 [49] DCMS ZnO:Sc, ZnO:Y fd 2000 [50] RMFMS ZnO:Al fd 2001

[51] RMFMS ZnO:Al fd 1999

[52] RMFMS ZnO:Al lac, dd 2003 [53] RRFS ZnO:Al fd 1988

Growth conditions Batch, 120 mm planar Zn0:2–3 wt% Al2 O3 Glass, 250◦ C

Layer properties 300–600 nm 2.7 × 10−4 Ω cm

Batch, 3 inch planar ZnO:3–10 wt% Ga2 O3 , sintered Glass, 100 − 250◦ C 1.1 W cm−2 Batch, ∼80 mm planar Zn:5.7 wt% Ga2 O3 Gas: Ar, Ne, Kr Glass, RT ∼2 W cm−2 , 0.2 − 0.6 nm s−1 Batch, 150 mm planar ZnO:Sc2 O∗3 , ZnO:Y2 O†3 Glass, 200◦ C 0.6 W cm−2

300 nm, 2.2 × 10−4 Ω cm, 9 × 1020 cm−3 , 32 cm2 V−1 s−1 (5.7 wt% Ga2 O3 , 250◦ C)

Batch, dual magnetron, 500 × 88 mm2 tm, P = f (U ) Zn:1.5 wt% Al Glass, RT∗ , 250◦ C† ∼6 W cm−2 , 7 nm s−1 Batch, 500 × 88 mm2 DMS Met. mode, Zn desorption Zn:0.9–1.5 wt% Al (segmented) 100◦ C∗ , 250◦ C† ∼10 W cm−2 8.2∗ , 8.8† nm s−1 In-line, 400 × 120 mm2 DMS tm, q(O2 ) = f (PEM) Zn:1.5 wt% Al RT∗ , 300◦ C† 2.9 W cm−2 Batch, 125 mm planar Zn, Al wires Glass, 100◦ C 0.01 nm s−1

∼200 nm

∼1,000 nm 2 wt%∗ , 4 wt%† 3.1 × 10−4 Ω cm∗ , 7.9 × 10−4 Ω cm† , ∗ 6.9 × 1020 cm−3 , ∗ 29 cm2 V−1 s−1 500 nm ∼3 at.% 7.5 × 10−4 Ω cm∗ , 2.5 × 10−4 Ω cm†

∼500 nm 3.6 at.%∗ , 2.2 at.%† 4.8∗ , 3.0† × 10−4 Ω cm 8.6∗ , 8.5† × 1020 cm−3 15∗ , 25† cm2 V−1 s−1 ∼10∗ , 4.4† × 10−4 Ω cm † 5.8 × 1020 cm−3 2 −1 −1 † 24 cm V s

540 nm 4.9 × 10−4 Ω cm 4.3 × 1020 cm−3 , 24 cm2 V−1 s−1

5 Magnetron Sputtering of ZnO Films

197

Table 5.1. (continued) Ref. Scope [54] RFMS ZnO1−x G: fd 1981

Growth conditions Batch, 4 inch planar ZnO, H2 gas RT−150◦ C ∼1.2 W cm−2 , ∼0.5 nm s−1

[55] RFMS ZnO1−x fd 1982 [56] RFS ZnO1−x fd 1983 [57] RFMS ZnO:Al fd 1991

Batchc) , 80 mm planar ZnO Glass, ∼90◦ C <0.3 nm s−1 Batch, 75 mm planar ZnO ∼100◦ C, ∼1.4 W cm−2 ,

[58] RFMS ZnO:Al fd 1996 [59] RFMS ZnO:X fd 1996

[60] RFMS ZnO:X fd 1985

[61] RFMS ZnO:Al fd 1997 [62] RFMS ZnO:(Al, H) fd 1999

Batch, 125 mm magnetron ZnO:2.0 wt% Al2 O3 Sapphire, 230◦ C, <1.4 W cm−2 , 0.5 nm s−1 Hetero epitaxial growth Batch, 4 inch planar ZnO:2.0 wt% Al2 O3 sintered Glass, 150◦ C Batch, 1 inch planar ZnO:2.5 wt% Al2 O3 ∗ ZnO:6.9 wt% Ga2 O3 † ZnO:10.5 wt% In2 O3 ‡ 150◦ C, 11 W cm−2 , 0.15 nm s−1 Batch, 80 mm planarb ZnO∗ ZnO:B2 O3 † ZnO:Al2 O3 ‡ ZnO:Ga2 O3 × ZnO:In2 O3 ◦ Glass, <150◦ C 2 W cm−2 , 0.5 nm s−1 su Batch, 4 in. planar ZnO:0.5 wt% Al2 O3 , sintered Glass, 150◦ C Batch ZnO:2.0 wt% Al2 O3 , sintered RT, 2 W cm−2

Layer properties ∼1,000 nm 7 × 10−3 Ω cm 1.1 × 1020 cm−3 , 8 cm2 V−1 s−1 2 × 10−3 Ω cm: 2nd plasma at substrate 200 nm 5 × 10−4 Ω cm 1.0 × 1020 cm−3 , 120 cm2 V−1 s−1 500 nm 5 × 10−3 Ω cm

380–670 nm 2 at.% 1.2 nm min−1 : 140 µΩ cm 1.3 × 1021 cm−3 , 34 cm2 V−1 s−1 22.3 nm min−1 : 300 µΩ cm ∼700 nm 6.9 × 10−4 Ω cm 400 − 800 nm, 2.6∗ , 4.0† , 4.2‡ at.% 8.3∗ , 5.6† , 9.1‡ × 10−4 Ω cm 3.4∗ , 5.5† , 5.7‡ × 1020 cm−3 22∗ , 20† , 12‡ cm2 V−1 s−1 425–650 nm 3.2∗ , 6.4† , 1.9‡ , 5.1× , 8.1◦ × 10−4 Ω cm 0.51∗ , 2.5† , 15‡ , 4.4× , 4.0◦ × 1020 cm−3 38∗ , 39† , 22‡ , 28× , 20◦ cm2 V−1 s−1

140 nm 4.7 × 10−4 Ω cm, 7.5 × 1020 cm−3 , 15 cm2 V−1 s−1 500 nm 4 × 10−4 Ω cm, 6 × 1020 cm−3 , 27 cm2 V−1 s−1

198

B. Szyszka Table 5.1. (continued)

Ref. Scope [63] RFMS ZnO:Al CIS superstrate, pa 550◦ C 2001 [39] RFDCMS M: ZnO:Al fd 1998

Growth conditions Batch, 100 mm magnetron ZnO:2.0 wt% Al2 O3 200–250◦ C, 3.2 W cm−2 , 0.5 nm s−1

Batch, 3 inch magnetron ZnO:2.0 wt% Al2 O3 RT 0.5 W cm−2 , <0.5 nm s−1 dev CIS [64] RFMS Batch, 126 mm planar ZnO:Ga ZnO:5.0 wt% Ga2 O3 Glass, RT, 0.9 W cm−2 , fd 0.1 nm s−1 1990 [65] Co-deposition Batch ZnO:Al Zn:2 wt% Al energy eﬀ. Glass, <100◦ C 0.2 nm s−1 windows 1988

Layer properties ∼10−3 Ω cm

10–370 nm 3.3–4.0 at.% ∼7 × 10−4 Ω cm, 5–7×1020 cm−3 , ∼15 cm2 V−1 s−1 300 nm ∼10−3 Ω cm, 1.5 × 1021 cm−3 , 7–20 cm2 V−1 s−1 ∼300 nm 2.14 at.% 5.4 × 10−4 Ω cm, 4.5 × 1020 cm−3 , 26 cm2 V−1 s−1

(Process (DCMS: DC magnetron sputtering from a ceramic target; RDCMS: reactive DC magnetron sputtering; RMFMS: reactive MF magnetron sputtering; RFMS: RF magnetron sputtering from a ceramic target; RRFMS: reactive RF magnetron sputtering)); (Goal (fd: fundamental development; dev: device-speciﬁc; lac: large-area coating; dd: dynamic deposition)); (Remarks (su: substrates perpendicular to the target (otherwise parallel); hetero-epitaxial growth; pav, post-annealed in vacuum; dev a-Si, TCO front electrode for a-Si:H solar cells; dev CIS, TCO front electrode for chalcopyrite solar cells))

– The deposition rate is high because of the high sputter yield of the metallic target surface. – The deposited ﬁlms are usually metal-rich and thus absorbing and not suitable for optical applications. – In the metallic mode, the pumping speed of the sputter process is high compared to the pumping speed of the turbo pump while the opposite is true in the oxide mode. These changes of the systems pumping speed due to target oxidation are even more severe in production systems. Oxide Mode – When the oxygen partial pressure is raised to a critical value, the process becomes unstable: An avalanche eﬀect occurs since the rise in oxygen partial pressure decreases the pumping speed for the reactive gas. The target surface becomes fully oxidized.

5 Magnetron Sputtering of ZnO Films

199

Fig. 5.5. Hysteresis curve for total pressure (ptot = p(Ar) + p(O2 ) and target voltage U as a function of oxygen ﬂow q(O2 ) for reactive MF magnetron sputtering of ZnO:Al at constant power, P = 4 kW. The eﬀective pumping speed of the turbo pump is 380 l s−1 , the eﬀective pumping speed of the sputter process for reactive gas is 1,230 l s−1 in the metallic mode and 200 l s−1 in the oxide mode (Ar ﬂow q(Ar) = 60 sccm). The labeling of the process modes refers to the increase of the oxygen ﬂow (reprinted from [68])

– The deposition rate is low because of the small sputter yield of the compound surface. – Film growth is under oxygen-rich conditions so the ﬁlms are stoichiometric. The high partial pressure of the reactive gas limits the surface mobility, so ﬁlm properties are not ideal. Furthermore, dopants are oxidized and thus, TCO ﬁlms become non-conductive. – When the oxygen partial pressure is decreased to a critical limit, the cathode surface becomes partially metallic. The sputter yield is increased and the 2nd avalanche eﬀect occurs for the transition from the oxide to the metallic mode. Transition Mode – The transition mode is an unstable process regime for conventional deposition systems. Closed-loop control concepts or modiﬁed chamber designs outlined in Sect. 5.3.4.2 are necessary for transition mode process control. – In transition mode, the oxygen to metal ratio on the substrate is a continuous function of the oxygen partial pressure. It is, therefore, possible to control the stoichiometry of ﬁlm and thus to optimize doping, morphology, and phase composition. The non-continuous change of reactive gas partial pressure is the primary problem in reactive sputtering of ZnO-based TCO’s. The hysteresis characteristics of the reactive process can be modeled within the framework of the Berg model, where the condensation of reactive gas atoms on the surfaces

200

B. Szyszka Θt

target (A) (1-Θt)

Θt

At j

j

FA(1-ΘC)

Qc

FAB(1-ΘC)

to pump Qp

F AΘ C

reactive gas (B) supply Qtot

FABΘC

Qt

QC

Ac Θc a)

Θc

(1-Θc)

collecting area: substrate and chamber

b)

Fig. 5.6. Reactive sputter process for depositing the compound ﬁlm AB. (a) Balance of reactive gas ﬂow Qtot , which is partially gettered at the target (QT ) and at the substrate (QC ) and partially pumped by the vacuum pump (QP ). The fraction of the target surface At that is covered by the compound AB is Θt . The fraction of the collecting area Ac covered is Θc . j is the sputter current density. (b) Deﬁnition of particle ﬂuxes that alter the target and collecting area coverage fractions Θt and Θc (see text), (modiﬁed from [70])

of the target and chamber and the corresponding changes of pumping speed and sputter yield are taken into account [69, 70]. The basic concept of that model is shown in Fig. 5.6. Within the framework of the Berg model, the reactive process is described as a gettering process operating on two surfaces, the target and the substrate (including the chamber). Those areas of substrate and target, which are metallic, getter the impinging reactive gas and turn therefore to an oxidized state. The target is sputtered with the ion current of the discharge. The oxidized areas of the target change to a metallic state due to sputtering. At the substrate, by contrast, a change from the oxidized to the metallic state is only possible due to the impinging metal ﬂux from the target. This simple approach allows the balance equation for the reactive gas partial pressure and deposition rate to be formulated. An example is given in Fig. 5.7 for the stability analysis of process control loops. The second problem is the decrease of the deposition rate in the oxide mode, which is most signiﬁcant for highly reactive oxides such as MgO and TiO2 , where the deposition rate changes by more than one order of magnitude [72, 73]. For ZnO on the other hand, the deposition rate decreases only by a factor of 2, as shown in Sect. 5.3.4.3. The third problem is process stability at high power levels during longterm operation: In DC operation, the oxidized area of the target is subject to continuous ion bombardment. The oxidized surface exhibits some conductivity, which allows for charge neutralization to a certain extent. When the power is increased further, the dielectrics charge up until the electrical ﬁeld

5 Magnetron Sputtering of ZnO Films

201

Fig. 5.7. Stability analysis of a process control loop for the reactive magnetron sputtering of high-index metal oxides. The control of discharge power to stabilize the oxygen partial pressure set point is modeled within the framework of the Berg model. A cycle time of 100 ms and process uncertainties for discharge current and oxygen partial pressure measurements are assumed. (from [71])

strength for dielectric breakdown is reached. The stored electric energy is ﬁnally transferred into an arc discharge. Arcs can be either micro-arcs when the surface charge of a small area of the target is transferred into an arc or hard arcs when an arc is ignited between the cathode and ground potential. Micro-arcs may develop rapidly into hard arcs if the power supply is not appropriately controlled. Even micro-arcs give rise to pinhole defects because of local target melting while hard arcs may also damage the target substantially. Arcing is therefore highly undesired in reactive sputtering. Advanced power supplies oﬀer fast arc handling features whereby the discharge power is interrupted when arcing is detected from the current–voltage characteristic of the discharge. Fast arc handling and thus low arc energy is, therefore, an important criterion for choosing a power supply for reactive sputtering [74]. Pulsed plasma excitation instead of DC operation of the discharge opens up further opportunities for stable long-term processes at high power levels. In single magnetron pulse mode deposition, the discharge voltage can either be switched oﬀ during pulsing (unipolar pulsing) or the discharge voltage can be applied as a reverse voltage to the target (bipolar pulsing). In bipolar pulsing, the positive charged area is neutralized by the electrons accelerated to the cathode in the afterglow of the plasma discharge. Pulsed operation allows charge neutralization of the target surface and thus prevents the formation of arcs. A further eﬀect of bipolar pulsed power plasma excitation addresses the disappearing anode eﬀect and thus the long-term stability of reactive sputtering. In conventional reactive DC magnetron operation, the anode gets coated with oxide ﬁlms. These insulating ﬁlms give rise to plasma ﬂuctuations

202

B. Szyszka

Fig. 5.8. Schematic diagram of the reactive dual magnetron sputter process using MF plasma excitation. ptot = 10−3 − 10−2 mbar. Cathode length up to 3.75 m, power density up to 20 W cm−2 , plasma excitation frequency 10–100 kHz

since the current path in the plasma is subject to changes which end up in inhomogeneous, non-reproducible coating conditions. This problem is most severe for highly insulating dielectrics such as SiO2 and MgO but is also relevant for ZnO deposition. The method of choice is dual magnetron sputtering, where two magnetrons operate side by side, connected to a medium frequency (some tenths of kilohertz) power supply, as shown in Fig. 5.8. One target serves as the anode of the process, while the other target is the actual cathode. The cathode is subject to sputtering and thus, dielectric areas may charge up. The anode, however, is subject to electron bombardment and thus, positive surfaces charges are neutralized in anode operation. Calculations reveal [75] that plasma excitation with a few tenths of kilohertz is suﬃcient to prevent the dielectric breakdown of these charged surfaces and thus, stable long-term processes can be achieved with a very low arc level. As a consequence, MF plasma excitation allows for higher power levels because of the superior process stability and also improved ﬁlm properties because of the more intense ion bombardment of the substrate [38]. Reactive MF magnetron sputtering was developed in the early 1990’s. Today, it is the state-of-the-art production technology for large-area optical coatings [76] and beside ceramic target magnetron sputtering a promising candidate for the mass production of ZnO-based TCO’s by sputtering for thin-ﬁlm photovoltaics. 5.3.4.2 Control of Stoichiometry and Phase Composition The properties of ZnO ﬁlms depend strongly on stoichiometry and phase composition. This holds for TCO ﬁlms as well as for dielectric ﬁlms: Dielectric ZnO Films – For ZnO ﬁlms with good piezoelectric properties, it is crucial to achieve c-axis oriented growth as well as a unique termination of the wurtzite phase ZnO crystallites [3].

5 Magnetron Sputtering of ZnO Films

203

TCO Films Based on Substoichiometric ZnO – n-doping can be achieved for substoichiometric ZnO ﬁlms where two electrons are generated for each oxygen vacancy. ZnO1−x ﬁlms with x < 0.05 exhibit small absorption in the visible range as well as a carrier density up to 2 × 1020 cm−3 [41]. – Substoichiometric ﬁlms can be obtained either by substoichiometric growth conditions [10] or by subsequent reduction. Annealing in H2 atmosphere decreases the resistivity considerably for ZnO ﬁlms deposited by spray pyrolysis [77]. Sputtered ZnO:Al ﬁlms do not show a substantial eﬀect [78]. TCO Films Based on Doped ZnO – Doping by group III elements allows for the substitution of Zn and thus, for the creation of one electron per dopant atom. – TCO ﬁlms such as ZnO:Al, ZnO:In, and ZnO:Ga are meta-stable materials since the phase segregation (e.g., formation of the oxides of the dopants instead of substitutional incorporation) is favored by thermodynamics. The doping eﬃciency for magnetron sputtered ZnO:Al ﬁlms is usually below 50%. Doping eﬃciency up to 100% has been reported for ﬁlms by PLD [79]. – Hydrogen doping is another method for n-doping of ZnO [80]. The incorporation of hydrogen is on interstitials. First principle simulations reveal that hydrogen is bonded to oxygen atoms as H+ , contributing one electron to the conduction band [81]. To summarize, it is mandatory to achieve growth conditions where the metal-to-oxygen ratio at the substrate is suﬃcient not only for transparent ﬁlm formation but also for appropriate doping. In reactive magnetron sputtering, the technical implementation of high doping eﬃciency is diﬃcult due to the hysteresis problem described earlier. The most important deposition parameters for reactive deposition are (1) the rate and the composition of the sputtered material at the target, (2) oxygen partial pressure p(O2 ), (3) total pressure ptot , (4) and substrate temperature TS . Optimization of these parameters is a diﬃcult task, which is still a burden for broad use of that technology: The core problem in reactive sputtering of not only ZnO but also for other compounds is the strong coupling of growth processes at the substrate with the oxidization state of the target material. There are several approaches to solve this particular problem: – – – – –

Baﬄing of the cathode Separation of the gas inlets for the sputtering and reactive gases Desorption of surplus zinc Separation of deposition and reaction zones Transition mode process control

204

a)

B. Szyszka

b)

Fig. 5.9. Conﬁguration for a baﬄed reactive magnetron used by Brett et al. for growing substoichiometric transparent and conductive ZnO ﬁlms. (a) Schematic diagram of the coater. (b) reactive gas baﬄe used for metallic mode ZnO deposition (reprinted from [10])

An overview of these topics is given in [26] for optical ﬁlms in general. The following subsections refer to ZnO deposition in more detail, taking also the speciﬁc property of Zn, its extraordinarily high vapor pressure, into account. In ceramic target sputtering, by contrast, the situation is not as diﬃcult since the most important parameter, the metal-to-oxygen ratio, is deﬁned by the stoichiometry of the ceramic target material to a certain extent. Ceramic target sputtering is described in Sect. 5.3.5. Metallic Mode Sputtering Using Baﬄes The use of a baﬄe between the target and the substrate creates a reactive gas partial pressure gradient when the reactive gas is introduced near the substrate. In this way, the cathode can be operated in the metallic mode even with a high reactive gas partial pressure at the substrate. The concept of baﬄed reactive magnetron sputtering dates back to the early 1980s [10]. The basic conﬁguration is shown in Fig. 5.9. A reactive gas baﬄe is mounted in front of the target. Furthermore, RF biasing of the substrate has been used to achieve desorption of surplus Zn activated by energetic bombardment from the RF discharge at the substrate. This technology allows for ﬁne tuning of the stoichiometry. Transparent substoichiometric ZnO1−x ﬁlms with x = 0.01 were grown with ne = 1.6 × 1020 cm−3 and µ = 19 cm2 V−1 s−1 corresponding to ρ = 2.1 × 10−3 Ω cm. A similar technology has also been reported by Maniv et al. [82] using RF biased substrate rotation and larger magnetrons. The lowest resistivity

5 Magnetron Sputtering of ZnO Films

205

achieved was 2 × 10−2 Ω cm in this case because of the inferior stoichiometry control using a large-area deposition system. A comparison of the inﬂuence of the substrate potential on Zn desorption during reactive DC magnetron sputtering of ZnO:Al has been given by Wendt et al. [83]. ZnO:Al deposition on ﬂoating substrates revealed substoichiometric growth, while ﬁlms deposited on conductive substrates kept at ground potential were fully oxidized. Stoichiometry analysis by Rutherford back scattering (RBS) revealed a substantial increase of Al content by a factor up to 4 due to desorption of Zn. The eﬀect is attributed to the higher electron ﬂux for the substrates at ground potential. The disadvantages of baﬄed sputter processes are the substantial decrease in deposition rate and the poor long-term stability of the process because of severe ﬂaking by material deposited on the baﬄe. Metallic Mode Sputtering Using Zn Desorption The high vapor pressure of Zn allows even metallic mode operation without reactive gas baﬄing since desorption of surplus Zn can be achieved by suﬃcient substrate heating. The feasibility of this approach has been shown using high-rate reactive MF sputtering for ZnO:Al ﬁlms with a resistivity of 300 µΩ cm at a growth rate of 9 nm s−1 . The process parameters are summarized in Table 5.2 [51]. Table 5.2. Process parameters for reactive MF magnetron sputtering of ZnO:Al ﬁlms in the metallic mode using Zn desorption Process

Reactive MF magnetron sputtering. Sinusoidal plasma excitation (40 kHz, Advanced Energy PEII). Static deposition. Boxcoater Pfeiﬀer PLS 580. Process operation in the metallic mode of the discharge. System parameters Target–substrate distance dST 90 mm Target dimensions Target material

Ar ﬂow O2 ﬂow Substrate temperature Process parameters Discharge power Ar partial pressure Process control Substrate material Deposition rate Film thickness

480 × 88 mm2 A: segmented targets 0.9 − 1.5 wt% Al B: homogenous targets 1.5 and 2.0 wt% Al q(Ar) 60 sccm q(O2 ) A: 163 sccm B: 140 sccm TS A: 200◦ C B: RT − 300◦ C P 8 kW p(Ar) 150 mPa – – borosilicate glass AF45, Si as ∼9 nm s−1 d ∼500 nm

206

B. Szyszka metallic mode

600

oxide mode

metallic mode

oxide mode 100

10 550

a [nm/s]

450

U(q(O2)) setpoints choosen for deposition

400

8

10–1

6

10–2

4

[cm]

U [V]

500

10–3

350 = 300 cm

2 300 100

(a)

120

140

160

180

q(O2) [sccm]

200

140

(b)

160

180

200

10–4

q(O2) [sccm]

Fig. 5.10. Inﬂuence of oxygen ﬂow q(O2 ) on target voltage U (a) and on deposition rate a and resistivity ρ, respectively, (b) for reactive MF magnetron sputtering of ZnO:Al at TS = 200◦ C. Segmented targets were used (Zn:0.9–1.5 wt% Al). Results shown are for Zn:1.5 wt% Al (reprinted from [51])

Transparent ﬁlms have been achieved at substrate temperatures exceeding 100◦ C. The dependence of discharge voltage, deposition rate, and resistivity for deposition at a substrate temperature of TS = 200◦C on the oxygen ﬂow rate is shown in Fig. 5.10. Stable operation of the discharge has been achieved in metallic mode for oxygen ﬂow up to 165 sccm. A low deposition rate and high resistivity has been observed at low oxygen ﬂow rates due to the formation of Al-rich ﬁlms giving rise to Al2 O3 segregation. An increase of the deposition rate from 2 to 9 nm s−1 was achieved when the oxygen ﬂow was raised from 140 to 163 sccm. These ﬁlms exhibit a low resistivity of ρ = 300 µΩ cm. Films grown in the oxide mode reveal high resistivity because Al dopants are oxidized under these conditions. The drawback of this concept is the severe contamination of the coating equipment because of desorption of surplus Zn that condenses on all lowtemperature interior surfaces of the coater. Furthermore, the Al doping level increases when the substrate temperature is raised since the highly reactive Al does not desorb. This is shown in Fig. 5.11 where the dependence of the elemental composition on substrate temperature is shown. The Al doping level corresponding to the Al alloy of the targets is 2.1 at.%. However, the observed Al concentration is more than 5 at.% for ﬁlms deposited at TS = 300◦ C. Even at 100◦ C, the Al concentration is 3.5 at.%. This eﬀect allows transparent ZnO:Al ﬁlms to be deposited at a low substrate temperature. A resistivity of 480 µΩ cm has been achieved at TS = 100◦ C. Transition Mode Process Control The processing techniques outlined earlier utilize either Zn desorption in the metallic mode or reactive gas partial pressure gradients for the deposition of transparent and conductive ZnO ﬁlms. From a manufacturing viewpoint,

5 Magnetron Sputtering of ZnO Films

207

Fig. 5.11. Dependence of Zn, Al, and O content of the ZnO:Al ﬁlms and deposition rate a on substrate temperature TS for metallic mode deposition of ZnO:Al (Process parameter set “B” in Table 5.2). The Al concentration without Zn desorption would be 2.1 at.% (reprinted from [51])

these are non-ideal conditions due to the severe reduction in deposition rate and equipment contamination. Therefore, considerable eﬀorts were made on deposition in the transition mode of the reactive magnetron discharge, where nonstable conditions are used and where closed-loop feedback control is mandatory for process stabilization. Figure 5.12 gives an overview of some process control techniques that are applicable for the stabilization of reactive sputter processes in the transition mode. The simplest approach to determine the state of a Zn:Al target in reactive sputtering is the measurement of the target voltage, since the plasma impedance decreases when the target is oxidized, as shown in Fig. 5.10. However, the discharge voltage is subject to changes when target erosion or substrate movement occurs. Therefore, more advanced techniques have been developed, such as plasma emission monitoring (PEM) via optical emission spectroscopy or monitoring the oxygen partial pressure via λsensors. The controlled variable of the process can be the discharge power or the reactive gas ﬂow. Other techniques utilize the harmonic analysis of the electrical signal waveforms to determine the target state [84]. Furthermore, the optical properties of ﬁlms can be measured using in-situ photometry and in-situ ellipsometry for process control [85]. A comparison of the diﬀerent technologies for transition mode process control is given in [86, 87]. Large-area coaters are usually equipped with either plasma emission monitoring (PEM) of the target coverage to control the oxygen ﬂow [88] or measurement of the oxygen partial pressure to control the discharge power using the λ-probe adapted from the automotive industry [89].

208

B. Szyszka discharge power process control unit

oxygen flow discharge voltage λ-probe

reactive sputter process

film

O2 partial pressure

mass spectroscopy

p: O2, Ar, Met.

optical emission spectroscopy

p: O2, Ar, Met.

harmonic analysis of voltage and current by FFT

Fk: Ui / Ii

ellipsometry

optical constants, film thickness, growth rate

RT-photometry

Fig. 5.12. Process control techniques for the in situ characterization of ﬁlm and process properties and closed-loop control strategies for optical ﬁlm deposition by reactive sputtering

5.3.4.3 Reactive MF Magnetron Sputtering of ZnO This section is on the reactive MF magnetron sputtering of ZnO ﬁlms in the transition mode. The ﬁrst part is on the deposition of dielectric ZnO ﬁlms. The reactive sputtering of ZnO:Al is discussed in subsequent sections. Dielectric Films Figure 5.13 shows the dependence of O2 partial pressure on discharge power and the growth rate achieved for ZnO deposition at diﬀerent substrate temperatures using reactive MF magnetron sputtering. The O2 partial pressure was measured with a λ-probe and the discharge power was controlled at constant gas ﬂow. The experimental conditions are summarized in Table 5.3. In Fig. 5.13a at high power (point A) a low O2 partial pressure of p(O2 ) = 10 mPa is achieved. The decrease in discharge power (transition from A to B) shows a continuous increase in reactive gas partial pressure. Beyond point B, a further continuous increase in reactive gas partial pressure can be achieved only by increasing the discharge power. An abrupt transition from the metallic to the oxide mode (point D) would be observed for a further decrease of discharge power. The closed loop process control permits the process to be stabilized in the region of the unstable transition mode between B and C, allowing compensation for the dynamic behavior of the target oxidation by corresponding

5 Magnetron Sputtering of ZnO Films

209

control of the discharge power. The scattering of the characteristic curve in Fig. 5.13 indicates the power changes that occur to stabilize the unstable operating points. The symbols indicate the process parameters with which samples were prepared at diﬀerent substrate temperatures. oxide mode

characterisitic

process hysteresis without closed-loop control

transition mode non transparent film desorption of Zn

transition metallic mode mode

Fig. 5.13. (a) Process characteristic for reactive MF magnetron sputtering of ZnO showing the dependence of reactive gas partial pressure p(O2 ) on discharge power P and process set points chosen for deposition, (b) dependence of growth rate a on reactive gas partial pressure p(O2 ) for diﬀerent substrate temperatures (reprinted from [90]) Table 5.3. Process parameters for reactive MF magnetron sputtering of ZnO ﬁlms (from [90]) Process

Reactive MF magnetron sputtering. Sinusoidal plasma excitation (40 kHz, advanced energy PEII) Static deposition. Boxcoater Pfeiﬀer PLS 580. Transition mode process control via λ–probe measurement of O2 -partial pressure.

System parameter

Target–substrate distance Target dimension Target material Ar ﬂow O2 ﬂow Substrate temperature Discharge power O2 partial pressure Ar partial pressure Process control Substrate material Deposition rate Film thickness

Process parameters

dST

90 mm 480 × 88 mm2 Zn 99.99 % purity q(Ar) 2 × 30 sccm q(O2 ) 2 × 40 sccm TS RT, 120◦ C, 200◦ C P 2–5 kW p(O2 ) 20 − 100 mPa p(Ar) 150 mPa −P = f(p(O2 )) borosilicate glass AF45, Si as 2–8 nm s−1 d ∼1,000 nm

210

B. Szyszka

The choice of the deposition parameters was based upon experiments on the reactive sputtering of transparent and conductive ZnO:X ﬁlms [51, 68] described in the previous section. ZnO ﬁlms with a thickness of d ≈ 1,000 nm were deposited at various reactive gas partial pressures onto unheated substrates and substrates at temperatures of 120 and 200◦ C. Figure 5.13b shows the inﬂuence of the reactive gas partial pressure p(O2 ) on the deposition rate a during the reactive MF magnetron sputtering of ZnO coatings on quartz. The parameter space shows two regions, describing the growth of ZnO under an excess of either zinc or oxygen. For growth under an excess of oxygen, an increase of the reactive gas partial pressure leads to a reduction in the deposition rate, which is independent of the substrate temperature. This results from the oxidization of the target, since the metallic particle current density j(Zn) is reduced by the low sputtering yield of the oxidized target. When ﬁlm growth occurs under an excess of zinc, the inﬂuence of substrate temperature becomes signiﬁcant. The expected steady decrease in deposition rate with increasing reactive gas partial pressure is only observed at low substrate temperatures – an increase in rate can already be seen at TS = 200◦ C when the reactive gas partial pressure is increased. As a ﬁrst approximation, the deposition rate in this region is directly proportional to the reactive gas partial pressure, so that for p(O2 ) = 0 mPa, no coating would result. Qualitatively, this observation can be attributed to the high Zn vapor pressure. Once a critical substrate temperature is exceeded, sputtering at a low reactive gas partial pressure leads to the desorption of excess Zn, so the deposition rate is limited by the reactive gas partial pressure p(O2 ) as described earlier. Table 5.4 shows that the Zn vapor pressure at a substrate temperature of 300◦C is still too low to explain the observed relationship. The Zn desorption is, however, supported by the intensive, high-energy ion bombardment of the MF excited magnetron discharge [38], so that desorption already takes place at lower substrate temperatures. Similar relationships between the deposition rate, O2 partial pressure, and substrate temperature have been noted in the literature [92]. Systematic investigations of the inﬂuence of ion bombardment on Zn desorption are reported in [93]. Figure 5.14 shows the X-ray diﬀraction patterns of ZnO ﬁlms deposited at diﬀerent substrate temperatures at p(O2 ) = 33 mPa, where the maximum deposition rate is achieved. Only the (002) reﬂection of the wurtzite phase can be identiﬁed, which indicates textured growth of the polycrystalline ﬁlms with the c-axis perpendicular to the substrate even at room temperature. Table 5.4. Zn vapor pressure (from [91]) T p(Zn) (Pa)

25 2.5 × 10−8

100 1.1 × 10−7

200 7.3 × 10−4

300 2.1 × 10−1

5 Magnetron Sputtering of ZnO Films

211

Fig. 5.14. XRD spectra measured in Bragg–Brentano geometry (Θ − 2Θ or ω scans) for ZnO ﬁlms deposited at the maximum rate (p(O2 ) = 33 mPa) and diﬀerent substrate temperatures

Fig. 5.15. (002) peak intensity I, full width at half maximum FWHM of the (002) peak and FWHM of the rocking curve as a function of oxygen partial pressure for three substrate temperatures

The intensity of the (002) reﬂections diﬀers for ﬁlms with comparable layer thickness by two orders of magnitude. Figure 5.15 summarizes the analyses of the XRD measurements for the ZnO series deposited at diﬀerent substrate temperature. The structural properties depend considerably on the substrate temperature and the reactive gas partial pressure. Films deposited at TS = 200◦ C in transition mode reveal the optimum properties. From Figure 5.13b it follows that the transition between the two coating regimes takes place at a reactive gas partial pressure of p(O2 ) ≈ 33 mPa. At that partial pressure, the maximum rate is achieved for high substrate

212

B. Szyszka

temperatures. At low substrate temperatures, by contrast, the ﬁlms are already fully transparent. The maximum rate at TS = 200◦ C is only 80% of the rate which is achieved for substrates at room temperature. The index of refraction measured at 550 nm using spectroscopic ellipsometry is between 2.03 and 2.04. The density is therefore similar. This indicates that the eﬀect of Zn desorption is substantial for the high-temperature high-rate reactive magnetron sputtering of ZnO. Figure 5.16 shows atomic force micrographs of samples deposited at TS = 200◦ C with diﬀerent values of the reactive gas partial pressure p(O2 ) (minimum pressure, pressure corresponding to the maximum deposition rate and maximum pressure). The AFM micrographs reveal that the surface morphology for the samples deposited in the metallic mode and those deposited at the maximum rate is similar, whereas that of the samples deposited at high p(O2 ) is signiﬁcantly diﬀerent. For the sample deposited at p(O2 ) = 17 mPa lateral structures about 100 nm in diameter, corresponding to coherently scattering crystallites, are noticeable. With the transition to a higher reactive gas partial pressure of p(O2 ) = 27 mPa an increase of the crystallite size is visible, whereas the surface roughness decreases from 30.6 to 22.4 nm. The crystallites in these samples show predominantly circular boundaries. At p(O2 ) = 97 mPa the layers are smooth, the RMS roughness is only 10.3 nm. For these samples long crystallites haphazardly arranged on the surface are observed, which overlap each other randomly (see the white arrow). The crystallites have a height of about 50 nm normal to the substrate (see arrow) and thus correlate with the grain size of 53 nm determined by XRD for these samples. The transition in the morphology from p(O2 ) = 27 mPa to p(O2 ) = 97 mPa is due to reduced surface diﬀusion of Zn atoms under an excess of oxygen. At a high reactive gas partial pressure, the impinging species condense immediately, while at a lower reactive gas partial pressure the surface

150 nm

110 nm

100 nm

a: p(O2) = 17 mPa RRMS = 30.6 nm

100 nm

b: p(O2) = 27 mPa RRMS = 22.4 nm

51 nm

100 nm

c: p(O2) = 97 mPa RRMS = 10.3 nm

Fig. 5.16. AFM images (1×1 µm2 ) of the surface morphology and root mean square value of the surface roughness RRMS for 1 µm ZnO ﬁlms deposited at TS = 200◦ C at diﬀerent oxygen partial pressure

5 Magnetron Sputtering of ZnO Films

213

energy of the respective crystallites is minimized by surface diﬀusion of the Zn atoms, resulting in the formation of larger spherical aggregates. Besides crystal orientation and surface morphology, the termination of the ZnO crystallites is another important parameter for the optimization of ZnO ﬁlms: High piezo-electric coupling coeﬃcients can only be achieved if the ZnO crystallites have a uniform polarity. There are only a few processes that allow the polarity of the ZnO ﬁlms to be assessed. For example, [94] reports a direct process for investigating ZnO polarity based on the analysis of time-of-ﬂight spectra in ion-scattering experiments: low-energy ion beams are scattered over the surface of the layer, causing it to be sputtered by the ion bombardment. The detector allows the time-resolved detection of the Zn-particle stream and of the scattered primary ions. The extent of O termination and Zn termination for the layer can be calculated from the diﬀerence between these signals. A simple process was used by Hickernell to determine polarity [3, 95]: the criterion is the etching behavior of the ZnO layer since the (00¯1)-surface (Oterminated) shows an etching rate a factor of ten times greater than that of the (001)-surface (Zn-terminated). Perpendicular to the c-axis, the etching rate is 40–50 times the etching rate of the (001)-surfaces. Good SAW (surface acoustic wave) characteristics are reported for Znterminated ZnO ﬁlms homogeneously etched at low rates [95]. O-terminated crystallites lead to the development of craters which are identiﬁable in SEM (scanning electron microscopy) due to their higher etching rates. Figure 5.17 compares SEM images of untreated and etched ZnO samples where etching has been performed for 5 s in 25% nitric acid at room temperature in an analogous manner to that reported by Hickernell [95]. In the untreated condition, a signiﬁcant columnar structure is clearly recognizable in the breaking edge. The images of the etched samples show that almost no etching occurs at p(O2 ) = 17 mPa. At p(O2 ) = 20 mPa, on the other hand, the SEM reveals a rough surface morphology with deep etch craters. Etched samples deposited at higher oxygen partial pressure are not shown, since those ﬁlms were fully etched away. The conclusion in view of the results from Hickernell is that samples deposited at a low reactive gas partial pressure under Zn-excess conditions show a unique Zn termination, while the transition to deposition under oxygen excess leads to the development of O-terminated crystallites, so that for these samples unfavorable SAW characteristics are expected. The transition mode process described here diﬀers fundamentally from conventional reactive sputtering processes for ZnO coatings which do not permit optimization of the ﬁlm characteristics since they are hysteresis-based. An example of this is the work of Jacobson et al. [96] on reactive DC-Magnetron sputtering of ZnO coatings. The RF Magnetron sputtering of ceramic ZnO targets [97,98] oﬀers fewer degrees of freedom for process optimization compared to the technology

214

B. Szyszka

a: p(O2) = 20 mPa, as prepared

b: etched

a: p(O2) = 17 mPa, as prepared

b: etched

Fig. 5.17. SEM images of untreated and etched ZnO samples deposited under excess Zn conditions at TS = 200◦ C

described here, because the advantageous deposition under Zn-excess conditions can be achieved only up to a certain extent due to the given stoichiometry of the ceramic target material. The investigations described here show that the reactive MF magnetron sputtering of ZnO layer systems oﬀers new degrees of freedom for process optimization, since the reactive gas partial pressure can be altered steadily even at high deposition rates. Films with high-quality structural characteristics were achieved particularly in the region of the unstable operating points that are not accessible with conventional reactive sputtering processes. The ability to adjust the ﬁlm-forming particle ﬂuxes to the substrate continuously and to achieve plasma activation of the growth via ion bombardment from the plasma are key parameters for the deposition of doped ZnO ﬁlms, as described in the subsequent sections.

5 Magnetron Sputtering of ZnO Films

215

Table 5.5. Literature survey of the structural data (FWHM of Bragg–BrentanoScan and rocking curve) of ZnO layers for SAW applications Process

T (◦ C)

FWHM (Θ − 2Θ) (◦ )

FWHM (Ω) (◦ )

Ref.

RMFMS

200 200 590 500 300 200 300 350

0.08 0.14 0.22 0.24

1.78 1.19 2.85

[90] [90] [99] [4] [98] [100] [101] [102]

RRFMS RFMS RF-Mag-ECR PLD

0.4 0.13

1.77 1.4 2.9 2.07

RMFMS: reactive MF magnetron sputtering; RRFMS: reactive RF magnetron sputtering; RFMS: RF magnetron sputtering from a ceramic target; RF-Mag-ECR: RF magnetron mode ECR sputtering; PLD: pulsed laser deposition

Transparent Conductive Films The transition mode process control described above is the key to reactive magnetron sputtering of ZnO:Al ﬁlms. Several approaches have proven to be useful, either adjusting the reactive gas ﬂow or the discharge power as a function of appropriate process variables. As an example, Fig. 5.18 shows the dependence of ﬁlm resistivity on oxygen partial pressure for the MF reactive magnetron sputtering of ZnO:Al ﬁlms. The ﬁlms were sputtered with the process control scheme described above for dielectric ﬁlms (oxygen partial pressure measurement with a λsensor and adjustment of the discharge power). The process was implemented in an in-line coater using a dual cathode arrangement with 750 mm target length. Zn:2.0 wt% Al targets were used. Details are described in [103]. Substoichiometric ﬁlms were obtained for p(O2 ) < 30 mPa at low substrate temperatures up to TS = 180◦ C. For higher substrate temperatures, Zn desorption occurs and thus, Al-rich ﬁlms with poor electrical properties were grown with Al content up to 6%. At higher reactive gas partial pressure, constant dopant concentration is observed with cAl = 1.8−2.5 at.%. For p(O2 ) = 30 − 35 mPa, the ﬁlms exhibit a low resistivity of < 1,500 µΩ cm. For TS > 100◦ C, ﬁlms reveal resistivity of ρ = 340−890 µΩ cm. At higher reactive gas partial pressure, excess reactive gas causes the oxidization of dopants and thus Al2 O3 incorporation. These ﬁlms show poor conductivity. A compilation of the literature on transition mode process control for ZnO deposition is given within Table 5.1. 5.3.5 Ceramic Target Magnetron Sputtering Ceramic target sputter processes utilize ceramic targets made by sintering or hot isostatic pressing (HIP). For lab-scale R&D work, another option is to use powder targets for sputtering [104].

216

B. Szyszka

Fig. 5.18. Transition mode process stabilization and ﬁlm resistivity for ZnO:Al ﬁlms by reactive MF magnetron sputtering at diﬀerent substrate temperature (reprinted from [103])

The main advantage of ceramic target sputtering is the fact that the oxygen for ﬁlm formation is released from the target surface during sputtering. It is much easier to achieve stable deposition conditions since the diﬃcult transition mode process control is not necessary. Furthermore, sputter processes from ceramic targets avoid the problem of Zn desorption when deposition is performed onto heated substrates. The main draw backs of ceramic target sputtering are the high costs for powder metallurgy and manufacturing of the ceramic targets. RF sputtering is necessary when dielectric compounds such as sintered ZnO:Al2 O3 disks are sputtered [58]. Since about 1995, it has been possible to make conductive sintered ceramics due to reducing annealing of the ceramic or due to incorporation of special dopants [80] on a large scale. When the conductivity of the target material is suﬃcient, it is possible to use either MF, pulsed, or DC plasma excitation for sputtering. Properties such as high density and homogenous conductivity deﬁne the quality of ceramic targets. Ceramics with nonconductive inclusions exhibit arcing and nodule growth on the target surface, while targets with low density exhibit inferior ﬁlm properties due to contaminations with diﬀerent species. Ceramic target magnetron sputtering has been used for fundamental materials development in several groups. The results from RF sputtering serve as a reference for the material properties that can be achieved by sputtering. ZnO:Al ﬁlms with a resistivity as low as 140 cm were prepared by RF ceramic target magnetron sputtering at deposition rate of 1.2 nm min−1 at TS = 230◦ C [57]. The properties of ZnO:Al ﬁlms for the application as the front electrode in a-Si:H p–i–n solar cells have been compared by M¨ uller et al. [106]. A strong advantage from a process viewpoint is the fact that the Al enrichment for ZnO:Al ﬁlms deposited by reactive magnetron sputtering at

5 Magnetron Sputtering of ZnO Films Reactive sputtering

217

Ceramic targets

Metallic targets Accurate process control required

Sintered ZnO targets

Cheap alloy targets

Reasonable costs

Cost intensive production process

Excellent film properties are obtained with good control

Electrical properties not optimal

Film properties superior to sintered target

HIP ZnO targets

Conventional DC- or Pulsed DC-Sputtering stable process

Increasing target costs relative target costs ~ 60%

100%

~ 300%

Fig. 5.19. Comparison of reactive and ceramic target sputtering processes. The relative target costs are a rough estimate for planar magnetron targets (from [105])

Fig. 5.20. Dependence of ﬁlm composition of ZnO:Al ﬁlms deposited by ceramic target magnetron sputtering on substrate temperature (from [107])

high temperature is not observed for ceramic target sputtering, as shown in Fig. 5.20 where no Al enrichment occurs for ﬁlms deposited at substrate temperatures up to TS = 200◦ C. 5.3.6 Other Technologies for Sputter Deposition of ZnO The various magnetron sputter processes described above utilize the same plasma process for sputtering and for activating ﬁlm growth on the substrate. There are also other techniques available, which permit a certain separation of sputtering and ﬁlm growth activation.

218

B. Szyszka

The most prominent technique is ion beam sputtering, where one ion source is used for sputtering the target material and a second ion source may be used for activation of the ﬁlm growth on the substrate surface. This approach allows precise adjustment of particle energy and particle ﬂuxes to the substrate. Ion beam sputtering of ZnO-based ﬁlms has been described in several publications [108–111]. The ﬁlm properties of ion beam sputtered ZnO ﬁlms are comparable to those obtained by magnetron sputtering. Nevertheless, industrial use of this technique is limited by the diﬃculty of up-scaling and the low deposition rate in the order of a few tens of nanometer per second. Combining an electron cyclotron resonance (ECR) plasma with a sputtering process is another option to seperate sputtering and plasma activation. It has been shown to be an attractive technique for the deposition of dielectric ZnO ﬁlms [101]. The combination of magnetron sputtering and inductively coupled plasma excitation (ICP) is a technique which allows enhanced ionization of the sputtered material. The combination of transition mode process control of the reactive sputtering and ICP plasma excitation is described in [112]. However, the resistivity of ZnO:Al ﬁlms sputtered from a Zn:1.5 wt% Al target is in the of 1,000 µΩ cm at TS = 150 ◦C, which is inferior to results from conventional sputter processes under similar conditions.

5.4 Manufacturing Technology 5.4.1 In-Line Coating The in-line manufacturing of ZnO-based TCO ﬁlms is a highly eﬃcient production technology, which has been adopted from the ﬂat-panel display industry. Figure 5.21 shows a typical vertical in-line system used to deposit ZnO:Al over large areas for a-Si:H/µc-Si:H superstrate thin ﬁlm photovoltaics by ceramic target or reactive magnetron sputtering. For this application, a rough and well-deﬁned surface morphology is necessary. It is achieved by post-deposition wet-chemical etching (see [113] and Chaps. 6 and 8). The ﬁlm thickness to be deposited is in the order of 1,000 nm and about 300 nm of layer thickness is removed by the wet-chemical etching procedure. A crucial parameter for the productivity of an in-line sputtering system is the cycle time, with the deposition rate of the magnetron sputter process being an important parameter. The ﬁlm thickness d [nm] at a given substrate transport speed vc [m min−1 ] can be calculated from the dynamic deposition rate ad [nm m min−1 ] which is the ﬁlm thickness when the substrate speed is 1 m min−1 . The dynamic rate can be derived from the static rate as [nm min−1 ] on assuming the width b [m] of the coating zone. d = ad /vc

with

ad = as b

5 Magnetron Sputtering of ZnO Films

219

b)

a)

transfer M8

1 xTMP 2000

Roots 9000

process M7 S4

1 x TMP 2000

1 x TMP 2000

process process S3 M6 M5

2 x TMP 2000

process M4

1 x TMP 2000

transfer process S2 M2 M3

2 x TMP 2000

S1

1 x TMP 2000

M1

8800

Fig. 5.21. Schematic diagram (from [103]) (a) and photo (b) of a large-area vertical in-line coater used for ZnO:Al-deposition (substrate dimensions up to 1,000 × 600 × 150 × mm3 )

Dynamic deposition rates in the order of 60 nm m min−1 have been reported for dual magnetron reactive MF magnetron sputtering of highquality ZnO:Al ﬁlms at a moderate power density of P/A ≈ 4 W cm−2 [89]. In this case, the power density was limited by the power supply and the process control algorithms used. For an optimization towards higher deposition rates, dynamic deposition rates of ad > 100 nm · m/min can be reached [43]. The dependence of the dynamic rate and resistivity on total pressure and substrate temperature is shown in Fig. 5.22. It is important to notice the increase of the deposition rate at a high reactive gas partial pressure, which is due to the higher reaction eﬃciency and thus suppressed Zn desorption at higher oxygen partial pressure. Similar deposition rates have been achieved using single magnetron ceramic target magnetron sputtering at high power density of 9 W cm−2 . Sintered sputtering targets allow such high power densities even for longterm operation [80].

220

B. Szyszka

Fig. 5.22. Resistivity (a) and dynamic deposition rate (b) for large area reactive MF transition-mode magnetron sputtering at diﬀerent total pressures and substrate temperatures. The power density is ∼4 W cm−2 (from [89])

5.4.1.1 Static Deposition The deposition of ZnO coatings by in-line sputtering is basically a multilayer approach since the ﬁlm grows under strongly varying conditions. The analysis of static deposition experiments in terms of particle ﬂux and ﬁlm property distribution is, therefore, helpful to understand and hence to improve the deposition process. Coating by magnetron sputtering involves a variety of species impinging on the substrate. Low-energy bombardment (rough estimation: E < 30 eV) promotes the surface diﬀusion and enhances the reactivity of chemical species. High-energy particles, on the other hand, increase the ﬁlm stress [114] and degrade the crystal structure, resulting in a high resistivity of the grown TCO ﬁlm. The origin of the energetic ﬂux is either the plasma sheath near the target (fast oxygen ions), the plasma sheath near the substrate (bombardment with positive ions), or the sputtering process itself (sputtered particles, reﬂected neutrals). As a consequence, the distribution of ﬁlm properties may reveal quite substantial inhomogeneity. Early work on the inhomogeneity of ﬁlm properties was carried out by Tominaga et al. [116] and by Minami et al. [115]. Figure 5.23 shows the resistivity proﬁle obtained for ZnO:Al ceramic target DC magnetron sputtering [115]. A low resistivity is observed for all temperatures in the center of the target. Opposite the racetrack, however, a high ﬁlm resistivity is observed for substrate temperatures up to 325◦ C. At 350◦ C, on the other hand, the situation changed. The resistivity is homogenous due to the thermally induced healing of the crystal defects, which are caused by bombardment with fast oxygen ions. Furthermore, the inhomogeneity due to the particle damage depends also on the erosion of the target used. This is shown in Fig. 5.24, where experiments with old and new targets have been compared [117, 118]. The new

5 Magnetron Sputtering of ZnO Films L=-5 Thermocouple

L=0

221

L=5

Heater Anode (holder)

Substrate Erosion Area Shield

Target Yoke

a)

Magnet

DC Generator

b)

Fig. 5.23. Deposition system (a) and distribution of resistivity (b) for ZnO:Al ﬁlm deposition by ceramic target (sintered ZnO:2.0 wt% Al2 O3 ) DC magnetron sputtering. TS = 150◦ C (open triangle), 300◦ C (open circle), 325◦ C (open square) and 350◦ C (bullet) (reprinted from [115])

a)

b)

Fig. 5.24. Sheet resistance and ﬁlm thickness distribution for static deposition DC magnetron sputtering of ZnO:Al on glass from sintered ZnO:Al2 O3 targets using a new or an eroded sputtering target (reprinted from [117])

target exhibits a stronger eﬀect since an intense oxygen ion particle beam is formed by the ﬂat target surface. In reactive magnetron sputtering, the situation is diﬀerent since the formation of charged oxygen ions can be suppressed up to a certain extent by using an appropriate cathode conﬁguration. Figure 5.25 shows a versatile layout for transition-mode dual magnetron sputtering which uses a plasma emission monitor to control the oxygen ﬂow. The oxygen gas inlet connected to the piezo-electric valve is located between the targets, the sputter gas Ar is introduced via the outer gas lines. The experiment was intended to compare the resistivity proﬁle when (1) all the reactive gas passes through the centeral gas inlet or when (2) a larger amount is introduced via the sputter gas inlet system.

222

B. Szyszka Heater Substrate carrier Shielding PEM sensor Ar inlet, Ar/O2 inlet Target Cathode Magnets O2 inlet controled by PEM MF Generator (40 kHz)

Fig. 5.25. Dual magnetron conﬁguration for MF reactive magnetron transitionmode sputtering. Fast PEM control of the reactive gas ﬂow inlet is applied via a piezoelectric valve for the gas inlet between the targets is used. Either all oxygen can be introduced between the cathodes or the total ﬂow can be divided such that most of the reactive gas is introduced via the outer gas lines

Fig. 5.26. Proﬁle of resistivity obtained for static deposition using the conﬁguration shown in Fig. 5.25

The resistivity proﬁle observed is shown in Fig. 5.26. The well-known inhomogeneity is observed when the centeral gas inlet is used. Using the outer gas inlet to introduce some of the oxygen improves the resistivity considerably. Almost no damage due to energetic particles can be observed. As a consequence, it must be emphasized that the generation of undesired high-energy species depends strongly on the conﬁguration of the sputter system. Pressure gradients achieved by separated gas inlets may be useful for suppressing the formation of energetic oxygen ions. The ultimate goal is to prevent the interaction of oxygen with the high-density magnetron plasma near the target. One option to achieve this is to use ozone as the reactive gas directed toward the substrates. High sticking coeﬃcients in the order of 100% have been reported for reactive DC magnetron sputtering of ITO using this approach [119].

5 Magnetron Sputtering of ZnO Films

223

5.4.1.2 Process Control Achieving reproducible and homogenous ﬁlm properties over large area substrates is a key requirement for industrial manufacturing. However, the conditions of the plasma discharge may vary over large area substrates since process perturbations are introduced during sputtering. An important issue in in-line magnetron sputtering is the impact of the substrate transport on the pressure and plasma distribution. Appropriate conditions are achieved when the distance between the substrates in the coater is small (few centimeter). Under these conditions, the discrete substrates approximate a continuous ribbon and the impact of the substrate transport on the pressure distribution is small. In lab-scale or small-scale deposition systems, when only one carrier is passed through the system at a given time, the situation is non-ideal. The substrate transport gives rise to severe pressure changes, which can be modeled using direct simulation Monte-Carlo techniques [120]. A second issue is the desorption of adsorbents, in particular desorption of water, when substrates are transported into the vacuum system. These factors can be minimized by engineering the equipment appropriately. The plasma process itself is subject to constant changes, at least due to the erosion of the target. In DC or pulsed magnetron sputtering using a single cathode, the anode is subject to coating. The build-up of dielectric ﬁlms on the interior of the coater will therefore change the plasma conﬁguration. The lack of process stability due to the disappearing anode eﬀect is most critical for dielectric ﬁlms such as SiO2 , but also has to be taken into account for sputtering of ZnO-based ﬁlms. The most sensitive issue, however, is the process stabilization of transitionmode reactive magnetron ZnO:Al sputtering for large substrates. For this application, it is not only mandatory to stabilize the reactive gas partial pressure at the center of the cathode. The second issue is to achieve symmetry control along the length of the targets, as the state of the targets may change due to local oxidization of the surface. This might occur on one end of the cathode, while the other end of the cathode operates still in the metallic mode. An example of this problem is shown in Fig. 5.27, where the transitionmode closed-loop control of central oxygen partial pressure is performed by adjusting the discharge power. The graph shows that the central oxygen partial pressure is kept constant with high accuracy. Additional λ-sensors have been used to monitor the top and the bottom O2 partial pressure. These pressures vary for the two consecutive runs by several millipascal. As a consequence, the ﬁlms are neither homogeneous nor reproducible. Therefore, the control loop shown in Fig. 5.28 was developed to solve the problem of symmetry control [121]. Two additional PID control loops are used to control the homogeneity of the reactive gas partial pressure because of appropriate regulation of the threefold gas inlet (top/center/bottom). The

224

B. Szyszka

Fig. 5.27. Time dependence of the oxygen partial pressures and the discharge power during the coating process of substrates of 1,000 × 600 mm2 using identical deposition conditions. Without symmetry control the oxygen partial pressures in the upper and lower part of the discharge may vary from one run to the next or change in an unpredictable manner during the process

Fig. 5.28. Operating principal of the symmetry control for reactive MF magnetron sputtering of ZnO:Al. Three λ-sensors are used to monitor oxygen partial pressure at diﬀerent positions along the target

measured p(O2 ) values are used as input for a control loop controlling the threefold oxygen inlet. An example for the performance of the control loop is shown in Fig. 5.29, where changes of p(O2 ) and p(tot) have been made. Due to the symmetry control, it was possible to achieve constant p(O2 ) over the length of the cathodes.

5 Magnetron Sputtering of ZnO Films

225

Fig. 5.29. Example of the response of the control loop to perturbations such as a changed setting for p(O2 ) at 280 s or a change of total pressure in the chamber (at 500 s) for statistic deposition. The symmetry control was set to equalize the partial pressures at all three positions Transmittance τv [%] 10

75.0

Sheet resisancet [Ω]

69.0

20

10 20

Sample 4 66.0 69.0 72.0 75.0 78.0 81.0 84.0 87.0 90.0

81.0

50 10 20 69.0 75.0

40 50

position [cm]

position [cm]

40

30

4.0

Sample 4

30

30

40

60

40

position [cm]

20

8.0

6.0

10 20

10.0 4.0

30

Sample 6

40 50

80

6.0

50

8.0 6.0 80

6.0 60

40

3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0

Sample 6

20

position [cm]

Fig. 5.30. Optical transmittance and sheet resistivity of samples deposited under identical deposition conditions with active symmetry control. The highly absorbing region on sample 6 (entrance region) is an artefact due to incorrect positioning of the carrier

The impact of symmetry control on the homogeneity of ﬁlm properties for large-area ZnO:Al coating is shown in Fig. 5.30. Homogeneous coatings with acceptable variation of transmittance and sheet resistance have been achieved. The concept of symmetry control described here is well known in the glass industry. Other implementations have been realized using multiple PEM sensors [43]. 5.4.2 Rotatable Target Magnetron Sputtering The concept of rotatable target magnetron (C-Mag) sputtering is shown in Fig. 5.31a. The target material is a water-cooled tube, which is attached to

226

B. Szyszka

a)

b)

Fig. 5.31. (a) Cross-section (reprinted from [123]) and industrial construction (b) of rotatable magnetron sputter sources

a)

severe debris formation

b)

clean target surface

Fig. 5.32. Photographs of double magnetron planar (a) and rotatable (b) targets after long-term reactive sputtering of ZnO:Al. The planar target is shown after 1,850 kW h, the rotatable target after 11,500 kW h (source: Von Ardenne Anlagentechnik, Dresden)

the end blocks containing the sealing, the water and power feedthroughs and the rotary drive [122]. A key advantage of ZnO C-Mag sputtering is the decrease of debris and defect formation since the target rotates, so that the redeposition zones on the target are sputtered continuously. Figure 5.32 compares the target surface for planar and rotatable reactive magnetron sputtering of ZnO:Al. Severe formation of debris occurs on the surfaces of the planar cathodes, while there is no debris on the rotatable target. Similarly, redeposition zones can be prevented by using planar magnetrons with moving magnets [89]. Further advantages are the higher degree of material utilization and the large material stock, which is raised by a factor of 4 compared with planar targets. Both factors enable long-term operation (up to 2 weeks uninterrupted) without maintenance, while planar cathodes have to be maintained each week.

5 Magnetron Sputtering of ZnO Films

227

5.5 Emerging Developments 5.5.1 Ionized PVD Techniques Process techniques such as ﬁltered arc [124–126] deposition, or pulsed laser deposition [79, 127–130] generate outstanding ﬁlm properties for a variety of TCO materials. The main feature is that the ﬁlm formation is achieved with ions instead of using mostly neutral atoms. The state of the art is the pulsed laser deposition of ZnO:Al ﬁlms, where resistivity of 85 µΩ cm has been reported at a substrate temperature of 230◦ C [79]. This result can be attributed to the high doping eﬃciency because of the suppression of Al2 O3 formation when ionized species are used for ﬁlm formation. The new concept of high-power pulsed magnetron sputtering (HPPMS) allows for similar plasma characteristics using large-area magnetron sputtering [131]. Power densities in the order of 1 kW cm−2 are applied in short pulses. This mode of operation leads to a substantial ionization of the sputtered material. Ionization on the order of 80% has been shown for Ag sputtering. ITO ﬁlms deposited by HPPMS have shown improved performance compared with conventional sputtering in terms of surface morphology and resistivity [132]. For ZnO:Al, the relevant parameter space has not been achieved up to now due to unstable process caused by arcing at high levels of ionization. However, initial promising results have been reported for room temperature deposition of ZnO:Al using reactive HPPMS [133]. 5.5.2 Hollow Cathode Sputtering Hollow cathode gas ﬂow sputtering is an alternative approach for large-area sputter deposition of oxides. A schematic diagram of the hollow cathode gas ﬂow sputter process is shown in Fig. 5.33. A hollow cathode glow discharge with high plasma density is formed between the targets. The sputtered material is transported to the substrate in an Ar ﬂow via forced convection. Reactive gas is introduced in the afterglow region of the discharge in a way that the metallic mode operation of the targets is not disturbed. Hollow cathode gas ﬂow sputtering operates in the pressure range of 0.1–1 mbar. The particles are thermalized at the substrate. Plasma activation of growth processes can be achieved either by applying a substrate bias or by pulsed mode operation of the discharge. Hollow cathode gas ﬂow sputtering is a promising technique for gentle growth of ZnO ﬁlms on sensitive substrate materials (e.g., organic photovoltaics and OLEDs). Furthermore, the process can be operated as a PVD/PECVD hybrid process for organic modiﬁcation of ceramic ﬁlms. The third process regime is the deposition of nanoparticles, since clusters are formed in the gas phase in high-pressure and high-power regime [134].

228

B. Szyszka

Fig. 5.33. Schematic diagram of hollow cathode gas ﬂow sputtering. Total pressure: ptot ≈ 0.1 − 1 mbar. Gas ﬂow: q(Ar) ≈ 1 − 5 slm for 75 cm target length

5.5.3 Model-Based Process Development Understanding the dependence of ﬁlm structure and morphology on system layout and process parameters is a core topic for the further development of ZnO technology. Work is being performed on in situ characterization of deposition processes. Growth processes are simulated using Direct Simulation Monte-Carlo (DSMC) techniques to simulate the gas ﬂow and sputter kinetics simulation and Particle-In-Cell Monte-Carlo (PICMC) techniques for the plasma simulation [132]. An example is given in Fig. 5.34, where a photo of a MF sputter process taken with a high speed camera is shown in comparison to the corresponding simulation of plasma density. The right-hand target is shown in cathode operation, while the left-hand target serves as the anode. The plasma is conﬁned between the targets. The anode is shielded by the magnetron’s magnetic ﬁeld and thus, the plasma reaches the anode in the area between the racetracks where the magnetic ﬁeld is weak. The racetrack area itself is subject only to a weak electron current drawn from the plasma. This explains why arcing is still observed in dual magnetron MF sputtering to some extent, in contrast to early models, which predicted that some 10 kHz of plasma excitation should be suﬃcient to prevent arcing. Acknowledgements. The work presented here was supported by the German Ministry for Science and Education (BMBF) within the framework of the OSTec project on development of energy-eﬃcient glazing (project No. 13N6520), the ZnO network project (project No. 01SF00314) and the 2+2 project transparent conductive

5 Magnetron Sputtering of ZnO Films

229

a)

b)

Fig. 5.34. (a) High-speed photograph (source: Fraunhofer FEP, Dresden) and (b) plasma simulation of an MF magnetron in-line sputter process (source: Fraunhofer IST, Braunschweig)

oxide ﬁlms (project No. NMT/03X2303A). The work would not have been possible without the constantly stimulating discussions and the fruitful cooperation of the project partners. I acknowledge the strong commitment of my coworkers Prof. Dr. Xin Jiang, Dr. Volker Sittinger, Dr. Rujang Hong, and Florian Ruske to the success of our work. Furthermore, I also thank Thomas H¨ oing, Udo Bringmann, Ralf P¨ ockelmann, Claudia Jacobs, Andr´e Kaiser, and Wolfgang Werner for various discussions and their outstanding support in the course of the projects.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

J.S. Chapin, Res. Dev. 25, 37 (1974) F.S. Hickernell, J. Appl. Phys. 44, 1061 (1973) F.S. Hickernell, Rev. Phys. Appl. 20, 319 (1985) F.C.M. van de Pol, F.R. Blom, T.J.A. Popma, Thin Solid Films 204, 349 (1991) H.J. Gl¨ aser, Glass Technol. 21, 254 (1980) R.E. Laird, J.D. Wolfe, SVC Ann. Tech. Conf. Proc. 37, 428 (1994) M. Arbab, Thin Solid Films 381, 15 (2001) Y. Tsuda, H. Omoto, K. Tanaka, H. Ohsaki, Thin Solid Films 502, 223 (2006) J. Kawaguchi, E. Kusano, Patent de 3628057a1, 1986 M.J. Brett, R.W. McMahon, J.D. Aﬃnito, R. Parsons, J. Vac. Sci. Technol. A 1, 352 (1983) K. Ellmer, F. Kudella, R. Mientus, R. Schieck, S. Fiechter, Thin Solid Films 247, 15 (1994)

230

B. Szyszka

12. R. Menner, R. Sch¨ aﬄer, B. Sprecher, B. Dimmler, in Proceedings of the 2nd World Conference on Photovoltaic Energy Conversion, Vienna, 1998 13. J. M¨ uller, G. Sch¨ ope, O. Kluth, B. Rech, V. Sittinger, B. Szyszka, R. Geyer, P. Lechner, H. Schade, M. Ruske, G. Dittmar, H.-P. Bochem, Thin Solid Films 442, 158 (2003) 14. R.K. Waits, J. Vac. Sci. Technol. 15, 179 (1978) 15. R.K. Waits, J. Vac. Sci. Technol. A 18, 1736 (2000) 16. D.M. Mattox, SVC Ann. Tech. Conf. Proc. 47, 23 (2004) 17. R. Latz, K. Michael, M. Scherer, Jpn. J. Appl. Phys. 30, L149 (1991) 18. M. Katayama, Thin Solid Films 341, 140 (1999) 19. P. Binkowska, B. Cord, P. Wohlfart, Microsyst. Technol. 13, 139 (2007) 20. J. Schroeder, B. Cord, W. Maaß, K.H. Schuller, U. Patz, Thin Solid Films 175, 287 (1989) 21. J.S. Colligon, J. Vac. Sci. Technol. A 13, 1649 (1995) 22. D.W. Hess, J. Vac. Sci. Technol. A 8, 1677 (1990) 23. N. Kaiser, Appl. Opt. 41, 3053 (2002) 24. J.A. Thornton, Surface Eng. 2, 283 (1986) 25. P.J. Kelly, R.D. Arnell, Vacuum 56, 159 (2000) 26. R. Herrmann, G. Br¨ auer, in Thin Films for Optical Coatings, ed. by R.E. Hummel, K.H. Guenter. Handbook of Optical Properties, vol. 1 (CRC, Boca Raton, 1995) pp. 135–187 27. B. Chapman, Glow Discharge Processes (Wiley, New York, 1980) 28. Y.P. Raizer, Gas Discharge Physics (Springer, Berlin Heidelberg New York, 1991) 29. D. Rosen, I. Katardjiev, S. Berg, W. M¨ oller, Nucl. Instrum. Meth. B 228, 193 (2005) 30. S.M. Rossnagel, in Handbook of Vacuum Science and Technology, ed. by D. Hoﬀman, B. Singh, J. Thomas (Academic Press, San Diego, 1998) pp. 609–627 31. R. Behrisch (ed.), Sputtering by Particle Bombardment I (Springer, Berlin Heidelberg New York, 1981) 32. R. Behrisch (ed.), Sputtering by Particle Bombardment II (Springer, Berlin Heidelberg New York, 1983) 33. R.E. Somekh, J. Vac. Sci. Technol. A 2, 1285 (1984) 34. T. Takagi, J. Vac. Sci. Technol. A 2, 382 (1984) 35. H.J. Gl¨ aser (ed.), Large Area Glass Coating (Von Ardenne Anlagentechnik GmbH, Dresden, 2002) 36. F.M. Penning, U.S. Patent 2,146,025, 1936 37. F.M. Penning, J.H. Moubis, Proc. K. Ned. Akad. Wet. 43, 203 (1940) 38. S. J¨ ager, B. Szyszka, J. Szczyrbowski, G. Br¨ auer, Surf. Coat. Technol. 98, 1304 (1998) 39. R. Cebulla, R. Wendt, K. Ellmer, J. Appl. Phys. 83, 1087 (1998) 40. G.L. Harding, B. Window, E.C. Horrigan, Sol. Energ. Mater. 22, 69 (1991) 41. M.J. Brett, R.R. Parsons, Solid State Commun. 54, 603 (1985) 42. S. Brehme, F. Fenske, W. Fuhs, E. Nebauer, M. Poschenrieder, B. Selle, I. Sieber, Thin Solid Films 342, 167 (1999) 43. T. Wallendorf, S. Marke, C. May, J. Str¨ umpfel, Surf. Coat. Technol. 174–175, 222 (2003) 44. H. Czternastek, A. Brudnik, M. Jachimowski, E. Kolawa, J. Phys. D Appl. Phys. 25, 865 (1992)

5 Magnetron Sputtering of ZnO Films

231

45. T. Nakada, Y. Ohkubo, A. Kunioka, Jpn. J. Appl. Phys. 30, 3344 (1991) 46. T. Minami, K. Oohashi, S. Takata, T. Mouri, N. Ogawa, Thin Solid Films 193–194, 721 (1990) 47. M. Miyazaki, K. Sato, A. Mitsui, H. Nishimura, J. Non-Cryst. Solids 218, 323 (1997) 48. M. Kon, P.K. Song, A. Mitsui, Y. Shigesato, Jpn. J. Appl. Phys. 41, 6174 (2002) 49. T. Minami, T. Yamamoto, T. Miyata, Thin Solid Films 366, 63 (2000) 50. N. Malkomes, M. Verg¨ ohl, B. Szyszka, J. Vac. Sci. Technol. A 19, 414 (2001) 51. B. Szyszka, Thin Solid Films 351, 164 (1999) 52. M. Kon, P.K. Song, Y. Shigesato, P. Frach, S. Ohno, K. Suzuki, Jpn. J. Appl. Phys. 42, 263 (2003) 53. Y. Igasaki, M. Ishikawa, G. Shimaoka, Appl. Surf. Sci. 33–34, 926 (1988) 54. J.B. Webb, D.F. Williams, M. Buchanan, Appl. Phys. Lett. 39, 640 (1981) 55. T. Minami, H. Nanto, S. Takata, Appl. Phys. Lett. 41, 958 (1982) 56. K. Ito, T. Nakazawa, Jpn. J. Appl. Phys. 22, 245 (1983) 57. Y. Igasaki, H. Saito, J. Appl. Phys. 70, 3613 (1991) 58. S. Bose, S. Ray, A.K. Barua, J. Appl. Phys. 29, 1873 (1996) 59. R. Wang, L.L.H. King, A.W. Sleight, J. Mater. Res. 11, 1659 (1996) 60. T. Minami, H. Sato, H. Nanto, S. Takata, Jpn. J. Appl. Phys. 24, L781 (1985) 61. K.C. Park, D.Y. Ma, K.H. Kim, Thin Solid Films 305, 201 (1997) 62. M.L. Addonizio, A. Antonaia, G. Cantele, C. Privato, Thin Solid Films 349, 93 (1999) 63. F.-J. Haug, Z. Geller, H. Zogg, A.N. Tiwari, C. Vignali, J. Vac. Sci. Technol. A 19, 171 (2001) 64. B.H. Choi, H.B. Im, J.S. Song, K.H. Yoon, Thin Solid Films 193, 712 (1990) 65. Z.C. Jin, I. Hamberg, C.G. Granqvist, Thin Solid Films 164, 381 (1988) 66. W.D. Westwood, in Physics of Thin Films, vol. 14, ed. by M.H. Fracombe, J.L. Vossen (Academic Press, New York, 1989) pp. 1–79 67. S. Schiller, G. Beister, W. Sieber, Thin Solid Films 111, 259 (1984) 68. B. Szyszka, Ph.D. thesis, Justus-Liebig-Universit¨at (1999) 69. C. Nender T. Larsson, H.O. Blomm, S. Berg, J. Vac. Sci. Technol. A 6, 1832 (1988) 70. S. Berg, T. Nyberg, Thin Solid Films 476, 215 (2005) 71. N. Malkomes, M. Verg¨ ohl, J. Appl. Phys. 89, 732 (2001) 72. V. Kirchhoﬀ, T. Kopte, T. Winkler, M. Schulze, P. Wiedemuth, Surf. Coat. Technol. 98, 828 (1998) 73. H. Kupfer, R. Kleinhempel, F. Richter, C. Peters, U. Krause, T. Kopte, Y. Cheng, J. Vac. Sci. Technol. A 24, 106 (2006) 74. D.J. Christie, H.V. Walde, T.J. Ash, SVC Ann. Tech. Conf. Proc. 48, 39 (2005) 75. J. Szczyrbowski, G. Teschner, SVC Ann. Tech. Conf. Proc. 38, 389 (1995) 76. G. Br¨ auer, M. Ruske, J. Szczyrbowski, G. Teschner, A. Zmelty, Vacuum 51, 655 (1998) 77. J. Aranovich, A. Ortiz, R.H. Bube, J. Vac. Sci. Technol. A 16, 994 (1979) 78. T. Minami, H. Sato, H. Nanto, S. Takata, Thin Solid Films 176, 277 (1989) 79. H. Agura, A. Suzuki, T. Matsushita, T. Aoki, M. Okuda, Thin Solid Films 445, 263 (2003) 80. F. Ruske, V. Sittinger, W. Werner, B. Szyszka, K. Dietrich, R. Rix, K. U. van Osten, Surf. Coat. Technol. 200, 236 (2005)

232 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.

B. Szyszka C.G. Van de Walle, Phys. Rev. Lett. 85, 1012 (2000) S. Maniv, C.J. Miner, W.D. Westwood, J. Vac. Sci. Technol. A 1, 1370 (1983) R. Wendt, K. Ellmer, Surf. Coat. Technol. 93, 27 (1997) N. Malkomes, B. Szyszka, J. Vac. Sci. Technol. A 19, 2368 (2001) M. Verg¨ ohl, N. Malkomes, T. Matth´ee, G. Br¨ auer, Thin Solid Films 377–378, 43 (2000) N. Malkomes, A. Pﬂug, B. Szyszka, M. Verg¨ ohl, SVC Ann. Tech. Conf. Proc. 45, 41 (2002) W.D. Sproul, D.J. Christie, D.C. Carter, Thin Solid Films 491, 1 (2005) C. May, R. Menner, J. Str¨ umpfel, M. Oertel, B. Sprecher, Surf. Coat. Technol. 169–170, 512 (2003) B. Szyszka, V. Sittinger, X. Jiang, R.J. Hong, W. Werner, A. Pﬂug, M. Ruske, A. Lopp, Thin Solid Films 442, 179 (2003) B. Szyszka, Thin Solid Films, submitted O. Kubaschewski, C.B. Alcock, Metallurgical Thermochemistry (Pergamon, Oxford, 1979) N. Tsuji, H. Komiyama, K. Tanaka, Jpn. J. Appl. Phys. 9, 835 (1990) M.J. Brett, R.R. Parsons, Can. J. Phys. 63, 819 (1985) T. Ohnishi, A. Ohtomo, K. Takahashi, M. Yoshimoto, H. Koinuma, Appl. Phys. Lett. 72, 824 (1998) F.S. Hickernell, T. Hickernell, in IEEE Ultrasonics Symposium, 373–376, 1992 E. Jacobsohn, D. Shechtman, Mat. Res. Soc. Symp. Proc. 242, 779 (1992) K.B. Sundaram, A.I. Khan, Thin Solid Films 295, 87 (1997) S. Takada, J. Appl. Phys. 73, 4739 (1993) J.G. Gardeniers, Z.M. Rittersma, G.J. Burger, J. Appl. Phys. 83, 7844 (1998) M. Kadota, M. Minakata, Jpn. J. Appl. Phys. 37, 2923 (1998) T. Inukai, M. Matsuoka, K. Ono, Thin Solid Films 257, 22 (1995) V. Craciun, J. Elders, J.G. Gardeniers, I.W. Boyd, Appl. Phys. Lett. 65, 2963 (1994) B. Szyszka, T. H¨ oing, X. Jiang, A. Bierhals, N. Malkomes, M. Verg¨ ohl, V. Sittinger, U. Bringmann, G. Br¨ auer, SVC Ann. Tech. Conf. Proc. 44, 272 (2001) P.J. Kelly, Y. Zhu, A. Postill, Thin Solid Films 426, 111 (2003) K.U. van Osten, (gfe) company information, 2007 J. M¨ uller, G. Sch¨ ope, O. Kluth, B. Rech, M. Ruske, J. Trube, B. Szyszka, X. Jiang, G. Br¨ auer, Thin Solid Films 392, 327 (2001) F. Ruske, Ph.D. thesis, Justus-Liebig-Universit¨at (2007) Y. Qu, T.A. Gessert, K. Ramanathan, R.G. Dhere, R. Nouﬁ, T.J. Coutts, J. Vac. Sci. Technol. A 11, 996 (1993) Y. Qu, T.A. Gessert, T.J. Coutts, R. Nouﬁ, J. Vac. Sci. Technol. A 12, 1507 (1994) S. Kohiki, M. Nishitani, T. Wada, Jpn. J. Appl. Phys. 33, 6706 (1994) J.W. Seong, K.-H. Kim, Y.W. Beag, S.K. Koh, K.H. Yoon, J. Vac. Sci. Technol. A 22, 1139 (2004) S.J. Jung, Y.H. Han, B.M. Koo, J.J. Lee, J.H. Joo, Thin Solid Films 475, 275 (2005) B. Rech, O. Kluth, T. Repmann, T. Roschek, J. Springer, J. M¨ uller, F. Finger, H. Stiebig, H. Wagner, Sol. Energ. Mater. Sol. Cell. 74, 439 (2002) H. Windischmann, Crit. Rev. Solid State 17, 547 (1992)

5 Magnetron Sputtering of ZnO Films

233

115. T. Minami, H. Sato, H. Imamoto, S. Takata, Jpn. J. Appl. Phys. 31, 257 (1992) 116. K. Tominaga, T. Yuasa, M. Kume, O. Tada, Jpn. J. Appl. Phys. 24, 944 (1985) 117. R. Mennerm, M. Powalla, in Forschungsverbund Sonnenenergie Tagungsband, 71–78, 2005 118. D. Severin, O. Kappertz, T. Nyberg, S. Berg, M. Wuttig, Thin Solid Films 515, 3554 (2007) 119. A. Alam, K. Sasaki, T. Hata, Thin Solid Films 281, 209 (1996) 120. A. Pﬂug, N. Malkomes, V. Sittinger, B. Szyszka, SVC Ann. Tech. Conf. Proc. 45, 16 (2002) 121. F. Ruske, A. Pﬂug, V. Sittinger, W. Werner, B. Szyszka, Thin Solid Films 502, 44 (2006) 122. J. Rietzel, S. Nadel, SVC Ann. Tech. Conf. Proc. 47, 224 (2004) 123. M.W. McBride, SVC Ann. Tech. Conf. Proc. 33, 250 (1990) 124. A. Ben-Shalom, L. Kaplan, R.L. Boxman, S. Goldsmith, M. Nathan, Thin Solid Films 236, 20 (1993) 125. L. Kaplan, A. Ben-Shalom, R.L. Boxman, S. Goldsmith, U. Rosenberg, M. Nathan, Thin Solid Films 253, 1 (1994) 126. T. David, S. Goldsmith, R.L. Boxman, Thin Solid Films 447–448, 61 (2004) 127. A. Suzuki, T. Matsushita, N. Wada, Y. Sakamoto, M. Okuda, Jpn. J. Appl. Phys. 35, L56 (1996) 128. H. Hiramatsu, K. Imaeda, N. Horio, M. Nawata, J. Vac. Sci. Technol. A 16, 669 (1998) 129. H. Hosono, H. Ohta, M. Orita, K. Ueda, M. Hirano, Vacuum 66, 419 (2002) 130. H. Tanaka, K. Ihara, T. Miyata, H. Sato, T. Minami, J. Vac. Sci. Technol. A 22, 1757 (2004) 131. U. Helmersson, M. Lattemann, J. Bohlmark, A.P. Ehiasarian, J.T. Gudmundsson, Thin Solid Films 513, 1 (2006) 132. M. Siemers, A. Pﬂug, B. Szyszka, SVC Ann. Tech. Conf. Proc. 49, 60 (2006) 133. F. Ruske, A. Pﬂug, V. Sittinger, W. Werner, B. Szyszka, D.J. Christie, in Proceedings of the 6th International Conference on Coatings on Glass and Plastics, Dresden, Germany, 2006 134. M. Birkholz, U. Albers, T. Jung, Surf. Coat. Technol. 179, 279 (2004)

6 Zinc Oxide Grown by CVD Process as Transparent Contact for Thin Film Solar Cell Applications S. Fa¨ y and A. Shah

6.1 Introduction 6.1.1 History of ZnO Growth by CVD Process Metalorganic chemical vapor deposition of ZnO ﬁlms (MOCVD) [1] started to be comprehensively investigated in the 1980s, when thin ﬁlm industries were looking for ZnO deposition processes especially useful for large-scale coatings at high growth rates. Later on, when TCO for thin ﬁlm solar cells started to be developed, another advantage of growing TCO ﬁlms by the CVD process has been highlighted: the surface roughness. Indeed, a large number of studies on CVD ZnO revealed that an as-grown rough surface can be obtained with this deposition process [2–4]. A rough surface induces a light scattering eﬀect, which can signiﬁcantly improve light trapping (and therefore current photo-generation) within thin ﬁlm silicon solar cells. The CVD process, indeed, directly leads to as-grown rough ZnO ﬁlms without any post-etching step (the latter is often introduced to obtain a rough surface, when working with as-deposited ﬂat sputtered ZnO). This fact could turn out to be a signiﬁcant advantage when upscaling the manufacturing process for actual commercial production of thin ﬁlm solar modules. The zinc and oxygen sources for CVD growth of ZnO ﬁlms are given in Table 6.1. The reactions with N2 O and CO2 lead to growth rates that are much lower than those obtained when using O2 or H2 O. On the other hand, the high reactivity of O2 and H2 O with DEZ and DMZ can induce premature reactions that may cause diﬃculties in controlling the deposition of ZnO layers. This aspect became especially important when deposition on large substrates was considered. To overcome this problem, less reactive agents have also been used, like tetrahydrofuran and alcohols. The CVD process was carried out ﬁrst at Atmospheric Pressure (APCVD), as illustrated by the typical schematic diagram shown in Fig. 6.1. Later on, the pressure was reduced to 0.1–10 Torr, in order to improve further the control of the chemical reactions involved in the deposition process, and thus, also improve the thickness homogeneity of the resulting layers [2, 4, 9, 16–18]. A further advantage of the low-pressure CVD (LP-CVD) process is that the

236

S. Fa¨ y and A. Shah Table 6.1. Precursors used in chemical vapor deposition of zinc oxide

Precursors Dimethylzinc Diethylzinc Zinc acetyl-acetonate Oxygen Water

Abbreviation DMZ DEZ

Tetrahydrofuran Alcohols

Formula (CH3 )2 Zn (C2 H5 )2 Zn Zn(C5 H7 O2 )2 O2 H2 O N2 O, CO2 C4 H 8 O

Source [5] [6] [7] [6, 8, 9] [4, 10] [11] [12, 13] [14, 15]

vapors are mixed very near to the substrate. Fig. 6.2 depicts the LP-CVD system developed at IMT Neuchˆatel. 6.1.2 Extrinsic Doping of CVD ZnO Extrinsic doping of ZnO ﬁlms deposited by the CVD process has been achieved by using aluminium, boron, ﬂuorine, gallium, or indium as foreign atoms. The CVD process allows one to tune very easily the doping level of the ZnO ﬁlms. This is done basically by varying the ﬂow of the doping gas that is introduced into the chamber during deposition. Such ﬁne tuning of the doping level of ZnO ﬁlms can lead to optimized electrical and optical properties as needed for the rough TCO used within thin ﬁlm solar cells. More recently, hydrogen has been identiﬁed as a doping agent in ZnO [19] (see also Sect. 2.1.1.1). Indeed, Myong et al. observed an increase of the free carrier concentration while introducing a ﬂow of H2 through a mercury bath during growth of photo-MOCVD ZnO from DEZ and H2 O [20]. Finally, it has to be noted that, all along this chapter, the expression undoped ﬁlms refers to ﬁlms that are not intentionally doped with an extrinsic dopant. Indeed, an electrical conductivity due to free carriers is often observed within undoped TCO ﬁlms, even if the latter are not intentionally doped. This is often attributed to the non-stochiometry of these ﬁlms or, alternatively, to the presence of defects incorporated into the ﬁlms during their growth (see Sect. 1.6.1). The following section describes in detail the important steps in the development of polycrystalline doped ZnO layers as deposited by CVD process. These steps consist in optimizing the following: – – – – –

Process pressure Growth mechanisms of ZnO layers Substrate temperature O/Zn ratio Dopant introduction into the CVD reaction

6 CVD ZnO

237

Fig. 6.1. An AP-CVD system for ZnO:F deposition, as developed by Hu and Gordon. DEZ and ethanol vapors are used as precursors for the ZnO formation. These vapors are transported with the help of the carrier gas Helium from their respective bubblers and then further diluted by pure helium from dilution lines. Adhesion of the ﬁlms on the substrate can be strongly improved by introducing in the reactor a small amount of water. Hexaﬂuoropropene is used as doping source. The total pressure is around 1 atm and the substrate is heated around 400◦ C. Reprinted with permission from [15]

In the following, each step will be described, both for the AP-CVD and for the LP-CVD cases. The results presented are focused on those properties of ZnO ﬁlms that are useful for thin ﬁlm solar cell applications, i.e., transparency, conductivity, and light scattering capability. The last part of paragraph 6.2 comments brieﬂy on alternative methods of CVD processes that have been investigated for ZnO deposition (PE-CVD, photo-CVD, . . . ). Finally, the last paragraph of this chapter summarizes the work so far published on thin ﬁlm solar cells using CVD-deposited ZnO layers as transparent contacts.

238

S. Fa¨ y and A. Shah

Fig. 6.2. The LP-CVD system for ZnO:B deposition, as developed at IMT Neuchˆ atel. DEZ and water vapors are used as precursors, and directly evaporated in the system. In this case, the vapors are not diluted in a carrier gas. Diborane (B2 H6 ) is used as doping gas, 1% diluted in Argon. The total pressure is set to 0.5 mbar (∼0.37 Torr), and the substrate is heated to 155◦ C

6.2 Development of Doped CVD ZnO Films 6.2.1 Inﬂuence of Deposition Pressure Here, we have (as already stated in Sect. 6.1.1) to distinguish between two diﬀerent growth regimes, according to their pressure ranges: – A high-pressure regime, where the AP-CVD (atmospheric-pressure chemical vapor deposition) process takes place. In general, most of the AP-CVD systems work at atmospheric pressure. In some cases, the pressure has been reduced down to ∼20 Torr, in order to try to have a better control of the chemical reactions that result in the ﬁnal ZnO ﬁlms. However, as the results obtained at pressures in-between the atmospheric pressure and ∼20 Torr are similar than those obtained at atmospheric pressure, they will be associated in this chapter with the AP-CVD process. The APCVD process is more diﬀusion-limited than kinetics-limited : the transport of the reactants to the growing surface is the limiting factor here. The growth rate of the ZnO ﬁlm increases when the pressure is reduced. Indeed, at lower pressures, the diﬀusivity of the reactants increases, reducing the time necessary for their transport to the growing surface. The data concerning doped AP-CVD ZnO presented in this paragraph comes

6 CVD ZnO

239

mainly from research carried out by the MIT group (Hu, Gordon et al.). Apart from this work, the present paragraph refers also to many preliminary studies carried out on undoped AP-CVD ZnO [5, 6, 8, 9, 14, 21, 22]. – A low-pressure regime, where the LP-CVD process takes place and which is associated with the pressure range from less than 1 Torr to about 10 Torr. In this case, the CVD process is more kinetics-limited than diﬀusion-limited. This means that the transport of the precursors from the inlets of the chamber to the growing surface is very fast, and that the growth of ZnO ﬁlms is mainly limited here by the kinetics of the reactions that occur at the growing surface. As an illustration, the DEZ consumption CDEZ (deﬁned as the percentage of DEZ eﬀectively used for ZnO growth, the rest of DEZ being pumped out of the chamber) is shown in Fig. 6.3, as a function of the total pressure in an LP-CVD reactor. CDEZ increases when the pressure is raised. Indeed, the residence time of the vapors inside the chamber is increased when the pressure is augmented, allowing thereby the growing surface to “absorb” more material. This leads, in its turn, to an increase of the growth rate with chamber pressure (at least within the pressure range of 1–10 Torr). The data concerning doped LP-CVD ZnO presented in this paragraph comes mainly from research carried out by the authors’ group (IMT Neuchˆ atel [17]), and by the Tokyo Institute of Technology (TIT) group (Yamada et al.). The upper pressure limit of the Low-Pressure regime is estimated around 10–15 Torr, because it corresponds to the pressure for which one may observe 65 60

CDEZ [%]

55 50 45 40 35 30

0

0.5

1

1.5

2

2.5

3

3.5

4

process pressure [mbar] Fig. 6.3. DEZ consumption CDEZ , i.e., percentage of DEZ eﬀectively used for ZnO growth, as a function of the pressure in an LP-CVD chamber, for which DEZ and water vapors are used as precursors

240

S. Fa¨ y and A. Shah

premature reactions in the vapor phase. These premature reactions, which indicate that the diﬀusion-limited regime has been reached, could limit the growth rate of ZnO ﬁlm on the substrate. This is illustrated by the curve of growth rate variation as a function of pressure, for an LP-CVD process, published by Yamada et al. [23] and shown in Fig. 6.4. One sees there that, as predicted for the LP-CVD regime, the growth rate increases with pressure up to 10 Torr, and decreases thereafter for higher pressures. However, at lower deposition temperature, the growth rate stays constant for pressure higher than 10 Torr. This is probably because, in this speciﬁc case, the surface reactions are still the limiting factor, even at higher pressures. Finally, it has to be noted that the pressure limit between the LP-CVD regime (kinetics-limited ) and the AP-CVD regime (diﬀusion-limited ) is not identical for all CVD systems. Indeed, it strongly depends on the reactor geometry (e.g., gas inlet, gas shower, laminar ﬂow), on the thermal emissivity of the chamber components (walls, substrate holder, . . . ), and on the reactivity of the precursors used for the chemical deposition of ZnO. The limit of 10 Torr has been chosen here based on the results published in the literature, and should be taken only as an indication.

Fig. 6.4. Variation of growth rate as a function of chamber pressure for ZnO:B ﬁlms deposited by CVD using DEZ and H2 O as precursors and B2 H6 as doping gas. Reprinted with permission from [23]

6 CVD ZnO

241

6.2.2 Growth Mechanisms for CVD ZnO 6.2.2.1 Growth of Polycrystalline CVD ZnO Films For both AP-CVD and LP-CVD processes, deposition conditions have been found, which result in polycrystalline ZnO ﬁlms that possess a strong preferential crystallographic orientation: Fluorine-doped AP-CVD ZnO, as developed by Hu and Gordon [24]: Figure 6.5 shows that the corresponding ﬁlms, if deposited above 350◦ C, are highly oriented, with the c-axis perpendicular to the substrate,1 i.e., with the same orientation as is usually observed for sputtered ZnO layers (i.e., one has here a (0002) X-ray diﬀraction peak). In Fig. 6.5, the bright-ﬁeld transmission electron micrograph (TEM) shows that these ﬁlms are composed of multiple grains. This is conﬁrmed by the electron diﬀraction pattern also shown in Fig. 6.5; it consists of large rings characteristic for a multicrystalline structure. This structure possesses a rough surface, as shown in Fig. 6.5. Such a surface can indeed induce a light scattering eﬀect that is particularly suitable for thin ﬁlm silicon solar cells. Boron-doped LP-CVD ZnO, as developed by Fa¨ y et al. [17]: the corresponding XRD graph for an LP-CVD ZnO:B layer is shown in Fig. 6.6. A strong preferentially oriented growth is observed here also, but it is, in this case, perpendicular to the (11¯ 20) crystallographic planes. This means that the c-axis is parallel to the substrate, in contrast with the orientation more commonly observed for AP-CVD and sputtered ZnO layers [15, 25, 26]. The TEM micrograph of a cross-section of a 2 µm-thick LP-CVD ZnO:B ﬁlm covered by a µc-Si:H solar cell is shown in Fig. 6.6 (top left). One can observe in this micrograph a ﬁrst 500 nm-thick initial layer (also called nucleation layer ); it is composed of small grains, with no discernible preferential orientation. After this initial layer or nucleation layer, columnar grains are formed and extend up to the surface of the ZnO ﬁlm. The upper ends of the columnar grains appear at the ZnO surface as pyramids. This can be observed in the SEM micrograph of the ZnO ﬁlm surface shown in Fig. 6.6 (top right). It is precisely such a pyramidal structure that gives to these ZnO layers an as-grown surface texture, which eﬃciently scatters the light that enters into the solar cell (see diﬀuse transmittance curves in Fig. 6.17). For a 2 µm-thick ZnO ﬁlm one obtains a basis size for the emerging pyramids that is roughly 400 nm. However, the grain size calculated with Sherrer’s equation from the XRD spectrum yields a value of only about 30 nm. This discrepancy can be understood by looking at the cross-sectional TEM micrograph given in Fig. 6.6 (top left): it reveals the presence of small crystallites at the bottom of the ﬁlms, followed by the growth of larger crystals. As crystals of dimensions larger than 0.5 µm do not contribute to a signiﬁcant broadening of the XRD peak, we can safely assume that the smallest crystals at the bottom of 1

This means that these ﬁlms are c-axis oriented.

242

S. Fa¨ y and A. Shah

Fig. 6.5. X-Ray diﬀraction of AP-CVD ZnO:F ﬁlm deposited at 400◦ C [15] (top); Transmission electron micrograph and electron diﬀraction patterns of an AP-CVD ZnO:F ﬁlm deposited at 400◦ C [24] (bottom left); SEM micrograph of an AP-CVD ZnO:F ﬁlm deposited at 400◦ C [15] (bottom right). Reprinted with permission from [15] and [24]

the ﬁlm are responsible for the observed broadening of the XRD peak, and thus, for the data that is used in Sherrer’s equation. To obtain local information about the crystallographic orientation of these columnar grains, a selected area diﬀraction (SAD) pattern has been obtained with the TEM electron beam focused on one of these columnar grains, as represented by the black circle in Fig. 6.6 (top left). The diﬀraction pattern thereby obtained is shown in Fig. 6.6 (bottom right). In contrast to AP-CVD ZnO:F, for which a multicrystalline structure has been observed (see Fig. 6.5), the regular spot pattern observed in Fig. 6.6 (bottom right) for LP-CVD ZnO:B indicates that each of the large columnar grains consists of a single crystallite. A closer view of the ZnO surface (see Fig. 6.7) reveals that the pyramids have nonregular facets, and are actually made up of a multitude of micro-steps. We can explain these steps as the successive stacking-up of (11¯ 20) and (0002) atomic planes, which are perpendicular to each other. Indeed, these planes possess the minimum surface energy density, which

6 CVD ZnO

243

5000

(1120)

[counts/s]

4000

3000

2000

1000

(1010)(0002) 0 20

30

40

50

60

70

80

2Θ [°]

Fig. 6.6. TEM micrograph cross-section of a 2 µm thick LP-CVD ZnO:B layer covered by a µc-Si:H solar cell (top left); SEM micrograph of the surface of a 2 µm thick LP-CVD ZnO:B layer deposited at 155◦ C and 0.5 mbar (top right); X-Ray diﬀraction of a 2 µm thick LP-CVD ZnO:B layer deposited at 155◦ C and 0.5 mbar (bottom left); Indexed selected-area diﬀraction pattern performed with the TEM electron beam focused on one columnar grain (represented by the black circle on the TEM micrograph shown at top left part of the ﬁgure). In the diﬀraction pattern the zone axis B = z = [1¯ 21¯ 3] (bottom right)

Fig. 6.7. SEM micrograph of the pyramidal grains emerging at the surface of a 2 µm-thick LP-CVD ZnO:B layer deposited at 155◦ C and 0.5 mbar

244

S. Fa¨ y and A. Shah

means that the adatoms arriving at the growing surface will arrange themselves preferentially along these crystallographic planes [26]. 6.2.2.2 Inﬂuence of Thickness Variation in the Structural Properties with Film Thickness AP-CVD ZnO: One ﬁnds in literature only a few comments on the evolution of structural properties with the thickness of AP-CVD ZnO ﬁlms. The main observation is that the grain size increases with ﬁlm thickness, as illustrated by the SEM micrographs of Fig. 6.8 for two AP-CVD ZnO:Al ﬁlms with diﬀerent thickness. Oda et al. observed that the spots of the diﬀraction patterns of AP-CVD ZnO ﬁlms grown with DEZ and alcohols become sharper as the thickness of the ZnO ﬁlm increases [14]. This is a further indication that the crystallites become larger as the thickness of the ZnO ﬁlms is increased (in this case, also, only the (0002) peaks are visible in the XRD diﬀraction pattern). LP-CVD ZnO: The evolution of the structural properties of LP-CVD ZnO thin ﬁlms with their thickness has been comprehensively investigated in [17] and is presented in detail hereafter. The thickness d of LP-CVD ZnO:B ﬁlms is represented in Fig. 6.9 as a function of deposition time t. For d > 500 nm, d is linearly dependent on t. For d < 500 nm, the growth of ZnO is slower than that for higher thickness, indicating an incubation phase during which nuclei are formed on the glass

Fig. 6.8. SEM micrographs of (a) a 270 nm-thick AP-CVD ZnO:Al and (b) a 840 nm-thick AP-CVD ZnO:Al deposited at 400◦ C from DEZ, ethanol, and triethylaluminum. Reprinted with permission from [27]

6 CVD ZnO

245

3

Thickness [mm]

2.5 2 1.5 1 0.5 0

0

5

10

15

20

deposition time [min] Fig. 6.9. Variation of ﬁlm thickness in dependence on deposition time for LP-CVD ZnO:B deposited at 155◦ C and 0.5 mbar

surface. Indeed, 500 nm corresponds to the thickness of the nucleation layer observed on TEM micrographs (see Fig. 6.6). When the glass surface is totally covered by the nuclei, the growth rate of ZnO becomes constant. SEM micrographs of the surface of four LP-CVD ZnO:B samples having diﬀerent thickness values are shown in Fig. 6.10, together with their respective XRD patterns. The SEM micrographs show clearly that the grain size increases with the ﬁlm thickness. For d < 500 nm, the ZnO ﬁlm is oriented along three directions perpendicular to the (11¯20), (0002), and (10¯10) crystallographic planes, which are, indeed, the planes possessing the minimum surface energy. The thickness of 500 nm corresponds to the nucleation layer observed in the TEM micrographs of Fig. 6.6. The nucleation layer can therefore be considered to be composed of small grains, which are oriented along three diﬀerent crystallographic directions. For d > 500 nm, the grains start to be preferentially oriented along a single crystallographic direction, which is perpendicular to the (11¯ 20) planes. This part of the ZnO layer corresponds to the conically shaped crystallites described in the previous paragraph. Geometrical considerations based on surface observations for LP-CVD ZnO:B ﬁlms: To quantify the size of the grains that are visible as pyramids on the surface of these particular ZnO ﬁlms, a dimensional parameter δ has been introduced in the following manner: the mean of the projected area of the pyramids has ﬁrst been evaluated from the SEM micrographs, as illustrated by the white drawings in Fig. 6.6 (top right). The dimensional parameter δ

246

S. Fa¨ y and A. Shah 500 d=0.44 µm

[counts/s]

400 300 200

(1120) (1010)

(0002)

100 0 20

30

40

50

60

70

80

70

80

70

80

70

80

1000 d=0.93 µm

[counts/s]

800 (1120)

600 400 200

(1010) (0002)

0 20

30

40

50

60

2500 d=1.91 µm

(1120)

[counts/s]

2000 1500 1000 500 (1010) (0002)

0 20

30

40

50

60

5000 d=2.99 µm

(1120)

[counts/s]

4000 3000 2000 1000 (1010) (0002)

0 20

30

40

50

60

2Θ [˚]

Fig. 6.10. SEM micrographs of surface of LP-CVD ZnO:B ﬁlms deposited at 155◦ C with various thicknesses, shown here with their corresponding XRD patterns

6 CVD ZnO a)

247

b) 1200

S ~ δ2

1000

δ [nm]

800

d

600 400 200 0

substrate

0

2

4 6 8 thickness d [µm]

10

12

Fig. 6.11. (a) Growth scheme of the single crystallites that constitute the LP-CVD ZnO:B ﬁlms shown in Fig. 6.6 (top left). (b) Grain size parameter δ in function of the thickness d of the LP-CVD ZnO:B ﬁlms

is then chosen as the square root of the mean projected area; it is, thus, a one-dimensional parameter. δ is plotted in Fig. 6.11b in dependence on ZnO ﬁlm thickness d. It increases linearly with d, conﬁrming the growth scheme described in Fig. 6.11a, where crystallites grow vertically from the substrate, widening progressively as d is increased. Indeed, the crystallites grow in such a manner that their projected area at the surface of the ZnO ﬁlm can be deduced by simple similarity. Furthermore, extrapolation of the linear regression of Fig. 6.11b to δ = 0 gives us the value d = 0. This means that the large crystallites observed for d ≥ 500 nm, which extend right up to the surface of the entire ZnO ﬁlm, actually start to grow from the substrate itself and not from the top of the 500 nm-thick nucleation layer. The root mean square values of the roughness Srms that have been deduced from AFM images of the surface of ZnO layers are plotted in Fig. 6.12a in dependence on ﬁlm thickness d, and in Fig. 6.12b in dependence on grain size δ. Srms , which corresponds to the mean height of the pyramids present at the ZnO surface, increases linearly with d and δ. This means that as the ZnO ﬁlm thickness increases, the crystallites become larger and the pyramids on the surface higher. The increase of the average base value of the pyramids leads also to an increase of their mean height. One can deduce from the scheme of Fig. 6.13 that the angle θ (formed by the facets of the pyramids) remains constant when the thickness of the ZnO layers increases. Furthermore, it is possible to evaluate the angle θ from the slope tan α of the linear ﬁtting curve of Fig. 6.12b. Srms is deﬁned as follows: 1 L 2 Srms = z dz, (6.1) L 0 L being the scanned length and z the height of the points measured with the AFM probe. To simplify the calculation, the three-dimensional AFM scan,

248

S. Fa¨ y and A. Shah 90 80

Srms [µm]

70 60 50

α

40 30 20 10

0

500

1000

1500

2000

2500

3000 50

100 150 200 250 300 350 400 450

δ [nm]

thickness d [nm]

Fig. 6.12. Surface roughness Srms of LP-CVD ZnO:B ﬁlms in function of (a) their thickness d and (b) their grain size δ

θ

Srms2 θ

Srms1

δ1

δ2

Fig. 6.13. Schematic description of the increase of both the width (from δ1 to δ2 ) and the height (from Srms1 to Srms2 ) of the crystallites that constitute the LP-CVD ZnO:B ﬁlms, when d is increased

from which the value of Srms is deduced, is reduced to a two-dimensional scan, as schematically drawn in Fig. 6.14. Indeed, Fig. 6.14 is similar to a single linear AFM scan passing through all the highest points of the pyramids. Based on the scheme presented in Fig. 6.14 and (6.1), Srms can be calculated as δ 2 2 2 δ 2 2 Srms = x tan θ dx + (δ − x)2 tan2 θ dx. (6.2) δ 0 δ δ2

6 CVD ZnO

θ

θ

249

Srms

δ Fig. 6.14. Scheme of a single linear AFM scan passing through all the highest points of the pyramids that constitute the surface of the LP-CVD ZnO ﬁlms

From (6.2) one can deduce the relation between tan α and tan θ according to tan α =

tan θ Srms = √ . δ 6

(6.3)

This leads to an average value for θ of 60◦ , which is in good accordance with values evaluated from TEM pictures and AFM scans. The curve in Fig. 6.12b conﬁrms, thus, the model proposed in Fig. 6.13, which is based on the growth of single crystallites that extend through the whole thickness of the LP-CVD ZnO ﬁlm. In conclusion, the above-proposed growth model allows us to deﬁne for LP-CVD ZnO ﬁlms a one-dimensional parameter δ that describes both the width of the crystallites constituting the bulk of the ﬁlms, as well as the height of the pyramids present at the surface of these ﬁlms. Variation in the Electrical Properties with Film Thickness AP-CVD ZnO: Hu and Gordon [10,28] observed an increase of electron mobility µ with increasing ﬁlm thickness d, both for boron-doped and for galliumdoped AP-CVD ZnO ﬁlms, with a more pronounced slope for small thickness. They attributed this behavior to the inﬂuence of the grain boundary scattering eﬀect, which is dominant for thinner ﬁlms that are composed of smaller crystallites. LP-CVD ZnO: The variation of the electrical properties for a boron-doped LP-CVD ZnO ﬁlm in dependence on its thickness d is presented in Fig. 6.15. The electrical properties are: resistivity ρ, carrier density N , and mobility µ. The variation of ρ, N , and µ as a function of d conﬁrms the hypothesis of two diﬀerent growth regimes, one below and another above 500 nm of ﬁlm thickness. The presence of small grains within the ﬁrst 500 nm of LP-CVD ZnO ﬁlms (see TEM micrograph of Fig. 6.6 (top left)) limits the doping eﬃciency, leading to a reduced value of N for d < 500 nm. For d > 500 nm, the doping eﬃciency is no more limited by the presence of small grains and N remains constant (for given deposition conditions) when d is further increased. On the other hand, µ increases drastically within the ﬁrst

S. Fa¨ y and A. Shah

N [cm–3]

ρ [Ωcm]

10–1

10–2

5x1020

20

4x1020

15

3x1020

10

2x1020

µ [cm2/Vs]

250

5

N µ

10–3

0

0.5

1 1.5 2 thickness d [µm]

2.5

3

1x1020

0 0

0.5

1 1.5 2 thickness d [µm]

2.5

3

Fig. 6.15. (a) Resistivity ρ, (b) carrier density N , and mobility µ for a boron-doped LP-CVD ZnO ﬁlm, in dependence on thickness d

500 nm, highlighting here the eﬀect of electron scattering induced by grain boundaries: This eﬀect is dominant in ﬁlms composed of small and multiple grains. For d > 500 nm, µ continues to increase but more slowly. The further increase in µ is certainly due to the increase in the widths of the crystallites as d is increased. Variation in the Optical Properties with Film Thickness AP-CVD ZnO: Hu and Gordon [10,27,29] deﬁned the average visible absorption of ZnO ﬁlms for solar cell applications as A=

dλΨ (λ)A(λ) , dλΨ (λ)

(6.4)

where Ψ (λ) is the solar irradiance spectrum for AM 1.5,2 and A(λ) is the wavelength-dependent absorbance of the ZnO ﬁlms obtained from reﬂectance and transmittance spectra. Hu and Gordon observed an increase of A with ﬁlm thickness for boron-doped (see Fig. 6.16), aluminum-doped, and galliumdoped AP-CVD ZnO ﬁlms. Because of this increase in absorption, it is not feasible to grow AP-CVD ZnO ﬁlms that are thicker than 1 µm. LP-CVD ZnO: Total and diﬀuse transmittance (TT and DT) for borondoped LP-CVD ZnO ﬁlms are shown in Fig. 6.17 as a function of the ﬁlm thickness. TT remains superior to 85% in the spectral range [500–900 nm] for ZnO layers with d < 1.5 µm. Because ∼15% of the incident light is reﬂected due to the change of refractive index at the air/ZnO and glass/air interfaces, this means that, for d < 1.5 µm, the absorption of the LP-CVD ZnO:B itself is too low to be measured by the spectrometer. TT is reduced by only about 5% for a thickness d = 3 µm. This means that 3 µm-thick ZnO ﬁlms still have a high transparency, in spite of their relatively high thickness. For λ > 900 nm, 2

AM denotes air mass, which characterizes the intensity and spectral distribution of sun light (see e.g. [30]).

6 CVD ZnO

251

Fig. 6.16. Variation of the average visible absorption of AP-CVD ZnO:B ﬁlms as a function of ﬁlm thickness. The ﬁlms were deposited with diﬀerent times at 360◦ C, using DEZ and tert-butanol as precursors and B2 H6 as doping gas. Reprinted with permission from [29] 100 d=440nm

transmittance [%]

80 TT

60 d=2990nm DT d=2990nm

40

20 d=440nm

0

400

600

800

1000

1200

1400

1600

1800

2000

wavelength [nm] Fig. 6.17. Total and diﬀuse transmittance (TT and DT) of LP-CVD ZnO:B ﬁlms grown at 155◦ C and 0.5 mbar, in function of their thickness d

in the near infra-red (NIR) range, TT is signiﬁcantly reduced due to the phenomenon of free carrier absorption (FCA) [31]. Thin LP-CVD ZnO ﬁlms having d < 500 nm practically do not scatter the light at all (i.e., DT is almost equal to 0). For d > 500 nm, the light scattering capability of LP-CVD ZnO ﬁlms increases with thickness, as illustrated by the progressive increase of DT with d observed in Fig. 6.17. This behavior is

252

S. Fa¨ y and A. Shah

certainly due to the progressive increase of the grain size with thickness (see Fig. 6.11b), an eﬀect that yields rougher ZnO surfaces and therefore higher light scattering capability for thicker ﬁlms. 6.2.3 Inﬂuence of Substrate Temperature In the CVD process, the substrate temperature is a key parameter: by heating the substrate, one supplies, directly to the growing surface, additional energy that favorably inﬂuences the speciﬁc chemical reactions that lead to ZnO growth. 6.2.3.1 Growth Rate AP-CVD ZnO: As has been explained in Sect. 6.2.1, AP-CVD processes are basically diﬀusion-limited processes rather than kinetics-limited processes. Indeed, here, the time for the transport of the reactants to the growing surface is the main limiting factor. Hence, the growth rate should be almost independent of substrate temperature. Actually, Roth and Williams [9], who carried out studies on ZnO growth by the AP-CVD process, observed a temperature independent growth rate for temperatures below 350◦C, but not for higher temperatures. They explained this behavior by the fact that the formation of ZnO depends not only on simple reactions that occur directly above the growing surface, but also on thermally activated branching reactions that contribute to ZnO formation at higher temperatures. These branching reactions can be activated during the transport of the reactant to the growing surface, provided that the temperature of the chamber is high enough. This also means that ZnO growth within an AP-CVD chamber strongly depends on the geometrical conﬁguration of the deposition system, in particular, on the way the reactants are transferred to the growing surface. Hu and Gordon [27] carried out ZnO deposition in an atmospheric-pressure laminarﬂow horizontal reactor, i.e., in a reactor where the ﬂow of precursors is parallel to the substrate (see Fig. 6.18a). As the reactant gas mixture moves downstream, there is a net increase in the density of ﬁlm precursors (in the gas phase), which diﬀuse to the substrate to form the ZnO ﬁlm. This means that the growth rate is lower on that edge of the substrate which is close to the entry point of the reactants, and that there is a peak in the growth rate for the speciﬁc position D on the substrate corresponding to the point where a maximum of ﬁlm precursors are created in the gas phase. Hu and Gordon obtained growth rates of ∼20–25 ˚ A s−1 with this laminar-ﬂow reactor. Lower growth rates around 5–15 ˚ A s−1 have been reported for a vertical reactor conﬁguration (see Fig. 6.18b). In this case, the reactant mixture always reaches the substrate at the same location. The growth rate now depends on the quantity of ﬁlm precursors created in the gas phase during the transfer of the reactants from the gas inlet down to the growing surface.

6 CVD ZnO

253

Gas inlet D

Substrate

Gas inlet Exhaust

D Substrate

Low growth rate

Low growth rate Peak growth rate

Exhaust (a) Laminar-flow horizontal reactor

(b) Vertical reactor

Fig. 6.18. Schematic drawing of two diﬀerent AP-CVD reactor conﬁgurations

Therefore, in the vertical reactor conﬁguration, the distance D between the gas inlet and the substrate, as well as the temperature of the whole chamber, have both a strong inﬂuence on the growth rate. Shealy et al. [8] observed a rapid increase of the growth rate when they reduced D in their vertical reactor. However, they also stressed the fact that one has to make a compromise between high growth rate and ﬁlm uniformity over the entire substrate area when selecting this distance. Indeed, reducing the distance between the gas inlet and the substrate will have an adverse eﬀect on the uniformity of the resulting ZnO ﬁlms. LP-CVD ZnO: In contrast to AP-CVD processes, LP-CVD processes are kinetically limited, i.e., the growth rate of the ﬁlm depends mainly on the rate of reactions at the growing surface. These reactions become the limiting factor, because the transport time necessary to carry the reactant from the gas inlet to the growing surface is strongly reduced by choosing the low pressure regime. As the speed of the reactions at the growing surface is increased when the temperature is raised, a temperature-dependent growth rate should basically be observed for LP-CVD process. In Fig. 6.19, the deposition rate and the consumption of DEZ (deﬁned as the percentage of DEZ eﬀectively used for ZnO growth, the rest of DEZ being pumped out of the chamber) are shown as a function of the substrate temperature for the LP-CVD process developed at IMT Neuchˆatel, which operates at 0.5 mbar and uses DEZ and water vapors as growth precursors. More DEZ is consumed by the growing surface at higher temperatures, leading to an increased growth rate, from less than 15 ˚ A s−1 at 130◦C to more than 40 ˚ A s−1 at 180◦ C. Wenas et al. [4], who worked at higher pressures, at the limit between the AP-CVD and LP-CVD regimes (i.e., they worked at

60

100

50

80

40

60

30

40

CDEZ[%]

S. Fa¨ y and A. Shah

Growth Ratetd [As-1]

254

td 20

10 120

CDEZ

140

160

180

200

20

0 220

Substrate Temperature [˚C] Fig. 6.19. Deposition rate and DEZ consumption for LP-CVD ZnO ﬁlms deposited at various substrate temperatures

pressures between 10 and 20 Torr), also observed an increase of the growth rate with temperature. However, in their case a saturation occurred when the temperature was increased even further. The saturation eﬀect was explained by the presence of preliminary reactions that take place prior to the precursors reaching the growing surface. This fact illustrates the fundamental diﬀerence between the two pressure regimes of the CVD process as described in Sect. 6.2.1. However, it has to be noted that the typical growth rates obtained by Wenas et al. are only around 7 ˚ A s−1 , i.e., much lower than those obtained by IMT Neuchˆ atel (i.e., around 25 ˚ A s−1 ). This diﬀerence between the two growth rates is surprising because at IMT Neuchˆatel, ZnO is deposited at pressures lower than 1 Torr, i.e., in a regime where low growth rates are expected. One explanation may be the diﬀerent DEZ ﬂow rates used for the ZnO deposition. Indeed, we will see in Sect. 6.2.4.1 that the growth rate of the CVD ZnO ﬁlms depends also strongly on the metal-organic precursor ﬂow. 6.2.3.2 Variation of Structural Properties with Substrate Temperature For both AP-CVD and LP-CVD systems, the structural properties of the resulting ﬁlms have been found to strongly depend on the substrate temperature used during deposition. In many cases, AP-CVD ZnO ﬁlms are found to acquire a pronounced preferential orientation along the (0002) direction (i.e., with the c-axis perpendicular to the substrate), when the substrate temperature is increased up

6 CVD ZnO

255

Fig. 6.20. X-Ray diﬀraction pattern of AP-CVD ZnO ﬁlms grown at diﬀerent substrate temperatures. (a) 320◦ C, (b) 350◦ C, (c) 400◦ C, (d) 420◦ C. Reprinted with permission from [22]

to 300–400◦C. For lower temperatures, the ﬁlms are either amorphous [8] or show various diﬀraction peaks, as illustrated in Fig. 6.20, indicating that they are randomly oriented [5,10,22]. In some cases, almost no ZnO formation was observed at the growing surface for lower temperatures. Note that, as one increase the substrate temperature, not only is the structural orientation of the resulting ZnO ﬁlms modiﬁed, but there is evidently also a modiﬁcation of surface morphology, and in general, an increase of grain size [5, 10, 14, 15, 32]. For LP-CVD ZnO, similar variations in the structural properties of ZnO ﬁlms occur with an increase of substrate temperature, but this takes place at a lower temperature, i.e., around 160◦C, and with diﬀerent crystallographic orientations. Indeed, the preferential orientation is (11¯20) for LP-CVD ZnO, instead of (0002) for the AP-CVD ZnO [3, 16, 23]. Figure 6.21 shows XRD patterns of undoped LP-CVD ZnO ﬁlms deposited with increasing substrate temperature, along with the corresponding SEM micrographs of their surface. The XRD patterns show a clear change in the growth direction, when the deposition temperature is increased: the ﬁlms change from a preferential orientation perpendicular to the (0002) crystallographic planes to a preferential orientation perpendicular to the (11¯ 20) crystallographic planes. This

256

S. Fa¨ y and A. Shah 12000

T=130˚C

(0002)

counts/s

10000 8000 6000 4000 2000

(1011)

(0004)

0 25000

T=146˚C

(0002)

counts/s

20000 15000 10000 5000

(1011)

T=164˚C

3500

(1120)

3000

counts/s

(0004)

(1120)

0 4000

2500 2000 1500 1000 500

(1010)

0 4000

T=180˚C

3500

counts/s

3000

(1120)

2500 2000 1500 1000 500

(1010)

0 500

T=236˚C

(1011)

(1120)

counts/s

400 300

(1010) 200

(1122) (1012)

100

(1013)

(0002) 0

20

30

40

50

60

70

80

2Θ [˚]

Fig. 6.21. XRD patterns and corresponding SEM micrographs of LP-CVD ZnO ﬁlms deposited at various substrate temperatures

6 CVD ZnO

257

is an abrupt transition in the crystallographic preferential growth direction and is accompanied by a morphological change of the growing surface, as observed in the SEM micrographs of the ZnO ﬁlm surface. Above 150◦ C, large tetragonal grains having their growth axis preferentially oriented perpendicular to the (11¯ 20) crystallographic planes emerge out of the surface. If the temperature is further increased (i.e., up to 180◦ C), the size of these large grains decreases, and at even higher temperatures (T ≥ 230◦ C), the growth of the ZnO layers becomes disordered, with the apparition, in the XRD pattern, of multiple growth orientations. 6.2.3.3 Variation of Electrical Properties with Substrate Temperature AP-CVD ZnO: The general trend of the conductivity in dependence on temperature, as observed for doped AP-CVD ZnO ﬁlms is (a) either an increase of the conductivity when the temperature is increased up to ∼460◦C, or (b), an increase followed by a decrease, leading to a maximum value of conductivity. The latter is often observed around 400 − 420◦ C. These two trends are induced by a corresponding variation of the carrier density N . Indeed, in some cases, N is continuously augmented when one increases the temperature [10, 24, 28]. This can be explained by a more eﬃcient dopant incorporation at higher substrate temperatures. However, in some cases N has also been found to decrease when the temperature was increased (e.g., growth of AP-CVD ZnO:F using DEZ and ethanol as reactants [15, 27]). Hu and Gordon attributed this decrease of carrier density N with temperature to the hypothesis that, in this temperature range, the ﬂuorine incorporation rate increases less rapidly than the deposition rate when the temperature is augmented. They also highlighted the fact that, with an horizontal conﬁguration of the AP-CVD reactor (see Fig. 6.18a), doping uniformity is not maintained along the substrate (just as growth rate uniformity is also not maintained) and that the values of N measured at individual points on the substrate do not necessarily reﬂect the overall value of N . The electron mobility µ usually always increases with temperature. This behavior has been correlated with the increase in grain size with temperature, as is usually observed for AP-CVD ZnO (see Fig. 6.22). Indeed, when the size of the grains is increased, the electron scattering eﬀect by grain boundaries is reduced, resulting generally in higher electron mobility µ. In some cases, however, a saturation or even a decrease of µ is observed for higher temperatures. In fact, if N is augmented with increasing substrate temperature, the scattering eﬀect by ionized impurities is also enhanced and therefore may reduce µ again (see chapter 2). LP-CVD ZnO: For LP-CVD ZnO, the usually observed increase of N with increasing substrate temperature suggests also a more eﬃcient dopant incorporation at higher temperatures [18]. However, the mobility µ does not follow the same increase as in AP-CVD, when the substrate temperature is

258

S. Fa¨ y and A. Shah

Fig. 6.22. Variation of the electron mobility in function of the grain size for APCVD ZnO ﬁlms. Reprinted with permission from [15] 40

101

30

0

µ [cm2/Vs]

resistivity ρ [Ωcm]

35

Morphological Transition

10

10–1

400

µ

250

20

200

15

150 100 Morphological Transition

5 –3

120

300

25

10

10–2

10

350

δ

130

140 150 160 170 180 substrate temperature [°]

190

0 120

δ [nm]

102

130

140 150 160 170 180 substrate temperature [°]

50 0 190

Fig. 6.23. Variations of (a) resistivity ρ, (b) Hall mobility µ, and grain size δ as a function of substrate temperature for ∼2.4 µm-thick undoped LP-CVD ZnO ﬁlms

increased. Indeed, in the case of LP-CVD ZnO, the trend of µ vs. temperature is strongly related to the abrupt morphological changes observed around 155◦ C (see Fig. 6.21). In Fig. 6.23, the variations of the resistivity ρ and Hall mobility µ are shown for undoped LP-CVD ZnO developed at IMT Neuchˆ atel. One sees in Fig. 6.23a that ρ drops almost two orders of magnitude around 150◦ C, and increases again after 170◦ C. The lowest value of ρ, respectively the highest value of µ, is obtained just after the morphological transition, i.e., in the range of temperatures where one obtains ZnO layers with preferential (11¯ 20) oriented growth and where large single crystallites emerge at the surface. Furthermore, for higher temperatures (235◦ C), for which the deposited ZnO ﬁlms are no longer oriented along a single preferential direction, the resistivity becomes very high due to a sharp drop in µ. Quantitative image analysis of the SEM micrographs given in Fig. 6.21 yields the average size of the emerging crystals δ: this average crystallite size is also shown in Fig. 6.23b as a function of the deposition temperature. δ goes through a maximum when the substrate temperature is increased, as does µ. Such a behavior can

6 CVD ZnO

259

be expected if one postulates that the electron mobility µ decreases when the density of grain boundaries increases. This demonstrates once again the importance of grain boundary scattering eﬀects in polycrystalline CVD ZnO ﬁlms. 6.2.3.4 Variation of Optical Properties with Substrate Temperature For both AP-CVD and LP-CVD processes, the main observation concerning the variation of the transparency of the ZnO ﬁlms is a reduction of transmittance in the near-infrared (NIR) range when the substrate temperature is increased. Indeed, N is usually increased with temperature, and this leads to a stronger free carrier absorption (FCA) eﬀect for higher substrate temperatures. AP-CVD ZnO: Figure 6.24 shows the diﬀuse transmittance (DT) of AP-CVD ZnO:F deposited at various substrate temperatures. DT clearly increases with increasing substrate temperature, indicating that the light scattering capability of AP-CVD ZnO ﬁlms is increased with temperature. This can be correlated with the increase of the grain size with temperature observed for AP-CVD ZnO ﬁlms. LP-CVD ZnO: Optical total and diﬀuse transmittance spectra (TT and DT spectra) of a temperature series of undoped LP-CVD ZnO ﬁlms are shown in Fig. 6.25: TT does not vary strongly with substrate temperature. Indeed, as Fig. 6.25 is related to a series of undoped samples, the values of carrier density N are too low to produce an observable free carrier absorption eﬀect

Fig. 6.24. Diﬀuse transmittance (DT) of AP-CVD ZnO:F ﬁlms deposited at various substrate temperatures. Reprinted with permission from [15]

260

S. Fa¨ y and A. Shah

100 TT 80 130˚C 60

146˚C 155˚C

40

164˚C

DT

180˚C 20

0

400

600

800

1000

1200

1400

1600

1800

2000

Fig. 6.25. Total and diﬀuse transmittance (TT and DT) of ∼2.4 µm-thick undoped LP-CVD ZnO ﬁlms deposited at various substrate temperatures. Reprinted with permission from [3]

in the near infrared. DT, on the other hand, varies considerably with substrate temperature (even though the ﬁlms all have the same thickness). DT increases up to the morphological transition described above (i.e., up to a substrate temperature of about 155◦ C), and decreases again for higher substrate temperatures. Thereby it follows a trend similar to the one observed for the electron mobility µ. This means that the optimum conditions for light scattering are obtained at the same temperature, for which the optima for the mobility µ and for the resistivity ρ are obtained, i.e., at a substrate temperature just above the temperature needed for the morphological transition. Comparing this trend for DT with the SEM micrographs of Fig. 6.21, it becomes evident that the large grains obtained just after the transition, at a substrate temperature of 155◦ C, are responsible for the high value of the DT spectrum obtained here. For higher substrate temperatures (i.e., from ∼160 to ∼180◦ C), the size of the grains observed in the SEM micrographs decreases again, as does the diﬀuse optical transmittance of the LP-CVD ZnO ﬁlms. 6.2.3.5 Summary: Optimized Substrate Temperature for AP-CVD and LP-CVD ZnO Films For both AP-CVD and LP-CVD processes, the highest substrate temperature will lead to the highest growth rate. Sometimes, in the case of AP-CVD systems, saturation will occur at a higher temperature, but this depends strongly on the reactor conﬁguration. The optimum temperature to obtain the best electrical properties for AP-CVD ZnO ﬁlms is between 400 and 450◦C. Indeed, if in the majority of

6 CVD ZnO

261

cases, both carrier density N and mobility µ are still increased after 400◦ C, a saturation or even decrease of these two parameters for T > 450◦ C has been reported. For the LP-CVD process, the optimum temperature that leads to the best electrical properties is located just after the morphological transition, i.e., around 160◦ C. In fact, below and above this temperature, µ is reduced due to a stronger scattering at grain boundaries: this eﬀect is induced by grain size and/or by the absence of preferential crystallographic orientation within the ﬁlms. Finally, the optimum temperature that leads to the best optical properties is also around 400–450◦C for the AP-CVD process. An increased temperature produces larger grains at the ZnO surface and therefore a better light scattering capability. However, if N becomes too high, i.e., usually for T > 450◦ C, free carrier absorption is increased and therefore the absorbance of the ZnO ﬁlms in the near infrared increases, too. Concerning the LP-CVD process, the same considerations can be made concerning carrier density N and free carrier absorption. Furthermore, the light scattering capability of LP-CVD ZnO ﬁlms is the highest just after the morphological transition, i.e., around 160◦ C, the temperature at which large grains appear on the ﬁlm surface. In conclusion, for both AP-CVD and LP-CVD processes, only a narrow range of temperatures can be identiﬁed for optimum performance (a range that is typically 40◦ C-wide). Within this narrow temperature range highly oriented ﬁlms are obtained that have electrical and optical properties suitable to act as transparent conductors in solar cells. The typical substrate temperature is around 400◦ C for the AP-CVD process, whereas it is around 160◦ C for the LP-CVD process. The two processes yield ﬁlm orientations that are perpendicular to each other. 6.2.4 Inﬂuence of Precursor Flux 6.2.4.1 Stoichiometry For the chemical vapor deposition of ZnO, the ratio of the various precursors that participate in the chemical reaction leading to ZnO formation is an important parameter: it inﬂuences the stoichiometry of the deposited ﬁlms and therefore, also, their properties. DEZ and DMZ, which are the metalorganic precursors mostly used for ZnO formation, react in the presence of oxidizing agents like O2 or H2 O. The equation for the complete oxidation reaction of DEZ as well as the equation for the complete reaction of DEZ with H2 O are given here as examples ((6.5) and (6.6)): (C2 H5 )2 Zn + 7O2 −→ ZnO + 4CO2 + 5H2 O

(6.5)

(C2 H5 )2 Zn + H2 O −→ ZnO + C2 H6

(6.6)

Actually, the situation that prevails in a CVD reactor is more complex: the gaseous oxidation of precursors like DEZ probably involves many intermediate reactions, such as those described by Roth and Williams [9]. The

262

S. Fa¨ y and A. Shah

probability for these intermediate reactions to occur, as well as the location where they take place within the reactor, strongly depend on the deposition conditions and also on the reactor conﬁguration. However, one common observation that can be made for all CVD processes is that an excess of the oxidizing agent is necessary to grow ZnO ﬁlms that can be used as TCO layers. Indeed, Smith [5] observed that for O2 /DMZ ratios below 1, ZnO ﬁlms deposited by AP-CVD become optically absorbing. The same observation has been made by the authors who have studied in detail the LP-CVD process for ZnO deposition, by carefully monitoring, here, the H2 O/DEZ ratio. The reason for the transparency degradation at low H2 O/DEZ ratios is unclear. It is, however, suspected that the formation of other species like zinc–ethyl groups or elemental zinc occurs when the ratio is lowered below 1, leading, then, to a poorer transparency of the deposited ZnO ﬁlms. Groups working on AP-CVD ZnO observed that the growth rate of the process increases with an increase in the ﬂow rate of the oxidizing agent, but only up to a certain value of the latter, a value that is given by the ratio of O/Zn, which is necessary to ensure a complete reaction. This threshold ratio is between 4 and 10 for reactions using O2 as oxidizing agent [5,8]. It is around 60 for reactions using DEZ and ethanol as precursors [15]. For O/Zn ratios higher than this threshold, the growth rate is independent of the ﬂow rate of the oxidizing agent. In the case of LP-CVD ZnO deposited from the reaction of DEZ and H2 O, the growth rate is independent of the H2 O ﬂow rate, as soon as the H2 O/DEZ ratio becomes greater than 1. All these results are in good accordance with (6.5) and (6.6), which state that 7 mol of O2 are necessary to oxidize 1 mol of DEZ for ZnO formation, whereas only 1 mol of H2 O is necessary to react with DEZ to form ZnO. Furthermore, once the threshold ratio for O/Zn is reached, i.e., once one has reached the ratio that ensures complete ZnO reaction, the growth rate becomes linearly dependent on the ﬂow rate of the metal-organic precursor. This is illustrated in Fig. 6.26 for the growth of AP-CVD ZnO from DMZ and O2 , and in Fig. 6.27 for the growth of LP-CVD ZnO from DEZ and H2 O. The following paragraph discusses the particular case of boron-doped LPCVD ZnO developed at IMT Neuchˆ atel, for which DEZ and H2 O are used as growth precursors. It describes in more detail the structural, optical, and electrical properties of the resulting ZnO ﬁlms when the gas ﬂow ratio H2 O/DEZ is varied. 6.2.4.2 Inﬂuence of the H2 O/DEZ Ratio To obtain TCO layers with suﬃcient optical transparency, the H2 O/DEZ ratio has to be kept higher than 1. Indeed, as mentioned above, a small excess of DEZ compared to H2 O leads to ZnO layers having a dark visual appearance. Keeping the DEZ and B2 H6 ﬂows constant while increasing the H2 O ﬂow rate increases the total gas ﬂow. In this way, the H2 O/DEZ ratio

6 CVD ZnO

263

Fig. 6.26. Growth rate in dependence of metal-organic precursor ﬂow (DMZ) for AP-CVD ZnO:F ﬁlms deposited at a substrate temperature of 300◦ C. These ﬁlms are deposited from DMZ, oxygen, and ﬂuorocarbon doping gas. Reprinted with permission from [24]

growth rate td [Å/s]

30

25

20

15

10

6

8

10

12

14

DEZ flow [sccm]

Fig. 6.27. Growth rate in function of metal–organic precursor ﬂow (DEZ) for LP-CVD ZnO:B ﬁlms deposited at 155◦ C and 0.5 mbar

was varied from 1 to 4, while keeping the doping ratio, i.e., the B2 H6 /DEZ ratio, constant (the inﬂuence of a variation in the doping ratio is described in detail in Sect. 6.2.5). All samples of this series have a thickness around 2.5 µm. This allows one to minimize the inﬂuence of thickness variation on the properties of the ZnO ﬁlms (see Sect. 6.2.2.2), and thus, observe only the inﬂuence of a variation in the H2 O/DEZ ratio. Electrical Properties The resistivity ρ, the Hall mobility µ, and the carrier concentration N of the ZnO ﬁlms deposited for this H2 O/DEZ series are represented in Fig. 6.28. The resistivity ρ increases by about 20% with an increase in the H2 O/DEZ

S. Fa¨ y and A. Shah 5x1020

2x10–3

30 N µ

1x10

N [cm–3]

ρ[Ωcm]

4x1020

–3

8x10

28 26

3x1020 24 2x1020

–4

22 (b)

(a) 6x10–4 0.5

µ [cm2/Vs]

264

1

1.5

2

2.5

3

3.5

4

1x1020 0.5

4.5

1

1.5

2

H2O/DEZ

2.5

3

3.5

20 4.5

4

H2O/DEZ

Fig. 6.28. Variations of (a) resistivity ρ, (b) Hall mobility µ, and carrier density N as a function of the H2 O/DEZ ratio, for 2.5 µm-thick LP-CVD ZnO:B ﬁlms deposited at 155◦ C, 0.5 mbar, and B2 H6 /DEZ = 0.6 40

100 H2O/DEZ=0.8

80

H2O/DEZ=1.2

60

H2O/DEZ= 4

H2O/DEZ= 2 TT 40 20

(b) 35

haze600nm [%]

transmittance [%]

(a)

DT

2.44µm 2.6µm

30

25 2.27µm 20

2.54µm

2.68µm 2.41µm

15

2.56µm 0

400

800

1200

wavelenght [nm]

1600

2000

10 0.5

1

1.5

2

2.5

3

3.5

4

4.5

H2O/DEZ

Fig. 6.29. Variation of (a) the total and diﬀuse transmittance curves (TT and DT), (b) the haze factor measured at λ = 600 nm (DT/TT at 600 nm), as a function of the H2 O/DEZ ratio, for 2.5 µm-thick LP-CVD ZnO:B ﬁlms deposited at 155◦ C, 0.5 mbar and with B2 H6 /DEZ = 0.6

ratio from 1 to 4. This increase is mainly due to the decrease of µ, whereas N remains constant at a value around 2 × 1020 cm−3 . Optical Properties Total and diﬀuse transmittance spectra (TT and DT, resp.) and haze factor (i.e., DT/TT measured at 600 nm) are presented in Fig. 6.29 as a function of the H2 O/DEZ ratio. As the thickness d of the ZnO samples does not vary signiﬁcantly within this series, we may assume that the trends observed hereafter are not due to a variation of d. TT does not vary within this series, with the single exception of the TT curve for the ZnO sample deposited with a H2 O/DEZ ratio of 0.8, i.e., the only sample deposited with an excess of DEZ. The TT of this sample is systematically lower than the TT curves of the ZnO samples deposited with an excess of water. The reduction of TT in the NIR area is similar for all the curves. This indicates that free carrier

6 CVD ZnO

265

absorption, which depends on the free carrier density of the ZnO ﬁlms, is not changed by a variation of the H2 O/DEZ ratio. This is in accordance with the behavior of the carrier density N (see Fig. 6.28b), which does not vary with the H2 O/DEZ ratio. DT is reduced when the H2 O/DEZ ratio is increased, leading thus to a reduction in the scattering capability of the ZnO layers. Structural Properties SEM micrographs of the LP-CVD ZnO:B surface show that the pyramidalshape of the microstructure described in Sect. 6.2.2 is always present, when the H2 O/DEZ ratio is varied from 1 to 4. However, the size of the pyramids is decreased. This is documented by a corresponding decrease in the parameter δ shown in Fig. 6.30 (remember that δ is the square root of the mean projected area of the pyramids present at the ZnO surface). The reduction in δ corresponds to the reduction in µ and in the haze factor as observed in Figs. 6.28b and 6.29b, respectively. As the carrier density N does not vary with H2 O/DEZ, we can safely assume that the density of ionized impurities within the ZnO ﬁlms also remains constant when the H2 O/DEZ ratio is increased. The observed decrease in µ would therefore be induced here solely by an increase in grain boundary scattering, i.e., by an increase in the density of grain boundaries. This increase of grain boundary density evidently occurs when the grains become smaller (i.e., when δ is reduced). In conclusion, the increase of the H2 O/DEZ ratio leads to a reduction of the mean width of the crystallites that constitute the ZnO ﬁlms. This reduction in grain size induces a diminution of the electron mobility as well as a reduction of the light scattering capability of the ZnO layers. It is therefore important, in the LP-CVD process, to keep the H2 O/DEZ ratio close to 1, in order to obtain ZnO ﬁlms with the highest conductivity and with the best 340 320

δ [nm]

300 280 260 240 220 0.5

1

1.5

2

2.5

3

3.5

4

4.5

H2O/DEZ

Fig. 6.30. Variation of grain size δ as a function of the H2 O/DEZ ratio for 2.5 µmthick LP-CVD ZnO:B ﬁlms deposited at 155◦ C and with B2 H6 /DEZ = 0.6

266

S. Fa¨ y and A. Shah

light scattering capability. However, it has been also experimentally veriﬁed that a small excess of DEZ in the chemical reaction immediately leads to a drop in the total transmission of the ZnO layers. To be sure not to fall into this regime, the H2 O/DEZ ratio should be set slightly larger than 1. 6.2.5 Doping of CVD ZnO Films 6.2.5.1 The Choice of the Doping Gas The choice of a dopant is driven by fundamental considerations, as well as by practical criteria. From a fundamental or theoretical point of view, ﬂuorine has been identiﬁed as an ideal dopant, because it can result, if adequately integrated into ZnO ﬁlms, in higher electron mobilities than group III metals (B, Al, Ga, In), which are commonly used as dopants for ZnO ﬁlms [24]. Indeed, as explained by Hu and Gordon [15], there is less electron scattering if the dopant atoms substitute within the ZnO lattice for oxygen atoms rather than for zinc atoms. The reasons for this are as follows: ZnO is a relatively ionic semiconductor, and the conduction band is derived mainly from the zinc orbitals. Thus, perturbing the arrangement of zinc atoms will also cause a strong local perturbation to the conduction band. This will lead, in turn, to a stronger electron scattering that will provoke a reduction of electron mobility. Group III metals substitute for zinc, whereas halogen atoms substitute for oxygen. The latter would therefore perturb mainly the valence band, and leave the conduction band relatively free of scattering. One would, thus, continue to have relatively high electron mobility. Fluorine atoms are those halogen atoms whose size shows the best ﬁt to the size of the oxygen atoms, making them, thus, ideal candidates for substituting oxygen atoms. Another advantage of using ﬂuorine atoms for doping ZnO ﬁlms in view of their use within thin ﬁlm silicon solar cells is that ﬂuorine atoms are electrically not active within thin ﬁlm silicon material, in contrast to the atoms of group III metals, which are p-type dopants for silicon. The risk of deteriorating the quality of the thin ﬁlm silicon layers by out-diﬀusion of the dopant atoms from the ZnO layer into the silicon layer is therefore reduced by using ﬂuorine as dopant. From an experimental point of view, a criterion for the selection of the dopant is the ready availability of a corresponding vapor or gas source containing the dopant, as well as the capacity of the latter to decompose and react with the other precursors used for ZnO growth. This should occur in such a way that it eﬀectively allows the dopant atom to substitute either the Zn atoms, or (preferably) the O atoms. Hu and Gordon [10, 24, 27, 29, 32] tested, in the context of the AP-CVD process, various dopants (Al, Ga, B, In, and F) for their incorporation into ZnO layers. They found higher mobility values for ZnO:F ﬁlms (∼35 cm2 (V s)−1 ) than for ZnO:Al ﬁlms (∼25 cm2 (V s)−1 ), conﬁrming, thus, the predicted higher suitability of

6 CVD ZnO

267

fluorine as dopant for ZnO deposited by CVD. However, they also observed that for deposition temperatures below 375◦ C, the ﬂuorine in the ﬁlm is not activated, leading to more resistive ZnO ﬁlms. This reduces the possibility to obtain a high doping eﬃciency with ﬂuorine atoms when using the LP-CVD process, for which typical deposition temperatures are around 160◦C. Indeed, the most promising results for the doping of LP-CVD ZnO ﬁlms at relatively low temperatures have so far been obtained with boron atoms, derived from the decomposition of diborane (Note that diborane decomposes already at ∼200◦C). Furthermore, it has to be remarked that, when boron atoms are used as dopant, mobility values of up to 35 cm2 (V s)−1 have also been obtained. These values are just as high as the mobility values obtained with ﬂuorine dopants in the case of AP-CVD [17, 33]. 6.2.5.2 Inﬂuence of Doping on the Properties of CVD ZnO Films Structural Properties Inﬂuence on grain size: The majority of groups that have tested the doping of ZnO for the CVD process have reported that, as the doping level is increased, the grain size (as observed at the ZnO ﬁlm surface) is reduced. This has been observed on ZnO ﬁlms for both the AP-CVD [10,27,28] (see Fig. 6.31) and the LP-CVD process [18, 34–36] (see Figs. 6.32 and 6.33), even though these two processes result in ZnO ﬁlms that have totally diﬀerent growth orientations (as described earlier), i.e., along the (0002) axis for the AP-CVD layers and along the (11¯ 20) axis for the LP-CVD layers. To understand why two ZnO layers having a similar thickness can possess such a large diﬀerence in grain size, TEM micrographs have been compared (Fig. 6.34), one for the slice of an undoped LP-CVD ZnO ﬁlm (i.e., a ﬁlm with a large grain size) and the other one for the slice of a heavily boron-doped LP-CVD ZnO ﬁlm (i.e., a ﬁlm with a small grain size).

Fig. 6.31. SEM micrographs of surface of AP-CVD ZnO:Ga ﬁlms deposited at 350◦ C. Sample (a) is undoped, and sample (b) is doped with 0.0032% thriethylgallium in the gas phase. Reprinted with permission from [10]

268

S. Fa¨ y and A. Shah

Fig. 6.32. SEM micrographs of surface of LP-CVD ZnO:B ﬁlms deposited at 155◦ C and with (a) B2 H6 /DEZ = 0 (undoped), (b) B2 H6 /DEZ = 0.6, and (c) B2 H6 /DEZ = 1.3 (highly doped) 550 500

δ [nm]

450 400 350 300 250 200

0

0.2

0.4

0.6 0.8 B2H6/DEZ

1

1.2

1.4

Fig. 6.33. Variation of grain size δ as a function of the B2 H6 /DEZ ratio for ∼2.4 µm-thick LP-CVD ZnO:B ﬁlms deposited at 155◦ C and with H2 O/DEZ = 1.2

The main diﬀerence between the undoped and the heavily boron-doped ﬁlms, as seen in these two micrographs, is given by the growth angle of the conical crystallites that grow from the substrate up to the ZnO surface: the growth angle is signiﬁcantly larger for the undoped ZnO sample. Indeed, although the nucleation layer (deﬁned in Sect. 6.2.2.1) has a similar thickness for both samples, the density of nucleation sites is higher for the heavilydoped sample. Therefore, when the doping ratio is increased, the density of grains oriented perpendicularly to the (11¯ 20) planes, i.e., of grains that can start to grow from the nucleation layer onwards, is also increased. There is then more competition between the growing grains and this leads to smaller growth angles. As a consequence, the basis of the pyramidal grains observed at the surface of the ZnO ﬁlm is, on an average, also smaller if the ﬁlms are heavily doped. Furthermore, the growth model for the LP-CVD process, as presented in Sect. 6.2.2.2, tells us that the wider the grains are in the bulk of the LP-CVD ZnO ﬁlm, the higher the pyramids become at the ZnO

6 CVD ZnO

(a)

269

(b)

Fig. 6.34. TEM micrographs of slices of LP-CVD ZnO:B ﬁlms deposited at 155◦ C with (a) B2 H6 /DEZ = 0 (undoped) and (b) B2 H6 /DEZ = 1.3 (highly doped)

surface. If we do accept this growth model, it becomes evident that the surface roughness of ZnO:B ﬁlms deposited by the LP-CVD process is reduced when B2 H6 is introduced into the chemical reaction. Inﬂuence on Crystal Orientation: Hu and Gordon [10] reported that, for ZnO:Ga layers deposited by AP-CVD at 370◦ C, dopant introduction (i.e., introduction of triethylgallium) leads not only to a reduction in grain size, but also to ZnO ﬁlms that have a less pronounced orientation along the caxis. In addition, the growth rate is, in this range of deposition temperatures (around 370◦ C), dependent on the dopant concentration. In contrast with this situation, when the deposition temperature is above 430◦ C, both undoped and doped ﬁlms are highly oriented with their c-axis perpendicular to the substrate. Moreover, Hu and Gordon observed a lower dependency of the growth rate on the dopant concentration for AP-CVD ZnO:Al and AP-CVD ZnO:F ﬁlms deposited at higher temperatures, i.e., at 400 and 435◦ C, respectively. A reduction in the preferential growth orientation when increasing the doping level has also been observed by Wenas et al. [18] for LP-CVD ZnO:B ﬁlms. Indeed, as can be seen in Fig. 6.35a, the (11¯20) peak (indicated as (110) in the graph) is predominant only for low doping levels, and the (10¯10) peak (indicated as (100) in the graph) becomes more and more pronounced when the doping level is increased. Wenas et al. used pressures around 6 Torr and B2 H6 as dopant gas. Fa¨ y et al. [3] used a pressure lower than 1 Torr, but also B2 H6 as dopant gas. They, however, did not observe any substantial modiﬁcation of crystal orientation by doping. Indeed, as illustrated in Fig. 6.35b, the (11¯20)/(10¯10) peak ratio observed here is quite high for undoped ZnO ﬁlms deposited at 155◦ C, and remains at a high value even when the dopant concentration is increased. Moreover, in this case, the deposition rate does not vary signiﬁcantly with the introduction of B2 H6 , as is shown in Fig. 6.36.

270

S. Fa¨ y and A. Shah 102

(1120)/(1010)

peak intensity ratio

102

101

101

(b)

(a) 10

100

0

0.1 0.5 0.9 B2H6 flow rate (µmol/min)

0

0.2 0.4 0.6 0.8 1.0 1.2 1.4 B2H6 /DEZ

¯ ¯ (corresponding to (110)/(100) ratio in Bravais notation) Fig. 6.35. (1120)/(10 10) XRD peak ratio in dependence on doping level: (a) LP-CVD ZnO:B ﬁlms deposited at 150◦ C and 6 Torr with varying B2 H6 ﬂow rate. Reprinted with permission from [18], (b) LP-CVD ZnO:B ﬁlms deposited at 155◦ C and 0.5 mbar (deposited at IMT Neuchˆ atel) as a function of B2 H6 /DEZ ratio 40

growth rate td [A/s]

35 30 25 20 15 10

0

0.2

0.4

0.6

0.8

1

1.2

1.4

B2H6/DEZ

Fig. 6.36. Growth rate as a function of the doping level, i.e., the B2 H6 /DEZ ratio for LP-CVD ZnO:B ﬁlms deposited at a substrate temperature of 155◦ C and a process pressure of 0.5 mbar

In conclusion, it seems that for both the AP-CVD and the LP-CVD processes, deposition parameters have been found, for which the introduction of a dopant component does not signiﬁcantly perturb the crystallographic orientation of the ZnO ﬁlms. However, it is, at this stage, not possible to identify the mechanisms and critical parameters governing the inﬂuence of dopants on ﬁlm structure. Nevertheless, some speculative statements can be made: in the case of the AP-CVD process, the deposition temperature seems to be the critical factor, i.e., above ∼400 ◦ C the growth orientation of ZnO ﬁlms is no more perturbed by the introduction of a dopant. For the LP-CVD

6 CVD ZnO

271

process, pressure seems to be the critical factor, i.e., below a certain pressure, the growth orientation becomes also insensitive to dopant introduction. Electrical Properties Figures 6.37 and 6.38 show the variation of electrical properties as a function of the dopant content of ZnO ﬁlms. Figure 6.37 shows the case of AP-CVD ZnO:F with ﬂuorine as dopant (here, the ﬂuorine atomic fraction is considered as dopant content ). Figure 6.38 shows the case of LP-CVD ZnO:B with boron as dopant (here, the B2 H6 /DEZ ratio is considered as dopant content ). The electrical properties taken into consideration are the conductivity σ, the resistivity ρ, the mobility µ, and the free carrier density N . 2000

electron density [cm–3]

conductivity [S/cm]

(a) 1000 1000 500 0 0.0 6x1020 (b) 5x1020

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

0.2

0.4

0.6

0.8

1.0

4x1020 3x1020 2x1020 1x1020 0 0.0 30 (c)

mobility [cm2/Vs]

0.2

25 20 15 10 0.0

fluorine content [at %]

Fig. 6.37. Variation of (a) conductivity σ, (b) carrier density N , and (c) mobility µ as a function of the ﬂuorine content for AP-CVD ZnO:F ﬁlms deposited at 400◦ C. Reprinted with permission from [15]

272

S. Fa¨ y and A. Shah

The introduction of active dopants leads to a corresponding increase in N and, as a consequence, to a strong drop of ρ by an order of magnitude, resulting in resistivities around 10−3 Ω cm (or conductivity values around 103 S cm−1 ). However, one can also observe that, above a certain doping level (ﬂuorine content >0.5 at. % and B2 H6 / DEZ gas doping ratio >0.6), a further increase of the dopant concentration does not yield a further decrease of ρ. This is due to (a) a stabilization of N and even, in some cases, a diminution of N at these high levels of dopant concentration, and (b) a steady decrease of µ. Hu and Gordon calculated the doping eﬃciency ηDE from the gallium content in the ZnO ﬁlms (determined by electron microprobe analysis) and the electron density, for the case of AP-CVD ZnO:Ga. They observed that ηDE steadily decreases with an increase of the gallium content (see Fig. 6.39). 40

4.8 1020

N µ

4 1020

3.2 10–3

30

N [cm–3]

ρ [Ωcm]

20

–3

2.4 10

1.6 10–3

3.2 10

2.4 1020

25

1.6 1020

20

8 1019

15

–4

8 10

0

0.2

0.4

0.6

0.8

1

1.2

35

0

1.4

0.2

0.4

B2H6/DEZ

0.6

0.8

1

1.2

[cm2/Vs]

4 10–3

10 1.4

B2H6/DEZ

Fig. 6.38. Variation of (a) resistivity ρ, (b) carrier density N , and mobility µ, in function of the B2 H6 /DEZ ratio, for LP-CVD ZnO:B ﬁlms deposited at 155◦ C and 0.5 mbar 70

Ne ηDE

60

8.0 50 6.0

40 30

4.0

20 2.0

Doping efficiency [%]

Electron density [1020cm–3]

10.0

10 0.0

0

1

2

3

4

5

6

7

8

9

0 10

Gallium content [at %]

Fig. 6.39. Variations of electron density Ne and doping eﬃciency ηDE (calculated from the gallium content and the electron density) as a function of the gallium content in AP-CVD ZnO:Ga ﬁlms deposited at 370◦ C. Reprinted with permission from [10]

6 CVD ZnO

273

This means that the fraction of gallium atoms that are electronically active within the ZnO:Ga ﬁlms decreases when the doping level increases. This can explain that N does not continue to increase for higher doping levels. The same conclusion has been made for LP-CVD ZnO:B by Yamada and Wenas [18], who performed SIMS measurements. These showed that, at a boron concentration of 2× 1021 cm−3 , only 10% of the boron atoms contained within the resulting LP-CVD ZnO:B ﬁlms were electronically active. Concerning the free carrier mobility, it has been observed that, except for very low doping ratios (e.g., ﬂuorine content <0.4 at. % for AP-CVD ZnO:F or gas doping ratio B2 H6 /DEZ < 0.2 for the LP-CVD ZnO:B), µ continuously decreases when the doping ratio is increased. This is attributed to the following: – An enhancement of the electron scattering eﬀect by ionized impurities. – An increase in the density of inactive neutral defects with the increase in doping level; these inactive neutral defects may possibly also introduce new states in the band gap, and thus, lead to a higher density of scattering centers. – An increase in electron scattering by grain boundaries. Indeed, in the majority of cases, the grain size is reduced when the doping ratio is increased, augmenting in this way the density of grain boundaries in the ZnO ﬁlms. For very low doping levels, an increase of µ has been observed when the doping level is augmented (it can be seen in Fig. 6.37 for ﬂuorine contents <0.4 and in Fig. 6.38 for gas doping ratios B2 H6 /DEZ < 0.2). Such an increase in electron mobility with an augmentation of the doping level (at very low doping levels) has also been observed by groups working on other TCO materials [31, 37, 38]. It is often explained by the following two hypotheses: (1) at low doping levels, electron scattering by grain boundaries is the main limiting eﬀect and (2) the barrier height at the grain boundaries may be reduced by the introduction of dopants in the TCO ﬁlms. Such a decrease of the barrier height would then lead to an increase of the electron mobility. Optical Properties The transmittance and reﬂectance spectra of an undoped AP-CVD ZnO ﬁlm and of a doped AP-CVD ZnO:Al ﬁlm are shown in Fig. 6.40. While the transmittance of the undoped ﬁlm stays over 80% along the whole visible range, the transmittance of the doped ﬁlm displays a pronounced drop in the near-infrared wavelength range. The drop corresponds to a minimum in the reﬂectance curve, as well as to a maximum (peak) in the absorbance curve. This occurs close to the so-called plasma frequency. These eﬀects are due to free carrier absorption. When N is increased, the plasma frequency is shifted towards shorter wavelengths, and the drop in optical transmittance becomes more pronounced. This is illustrated for the case of LP-CVD ZnO:B ﬁlms in

274

S. Fa¨ y and A. Shah

Fig. 6.40. Transmittance and reﬂectance spectra of (a) an undoped AP-CVD ZnO ﬁlm and (b) an AP-CVD ZnO ﬁlm doped with aluminum. Reprinted with permission from [27]

Fig. 6.41, where one can note a progressive drop of the transmittance curves with increase in B2 H6 /DEZ gas doping ratios. A closer look at the short-wavelength range in Fig. 6.41 reveals an increase of the gap energy Eg with increasing B2 H6 /DEZ ratios. Such an increase can be attributed to band ﬁlling by the additional free electrons, which occupy the energy states located above the conduction band minimum, leading, thus, to a widening of the optical band gap. This eﬀect is known as the Burstein– Moss eﬀect [39, 40]. The Burstein–Moss model predicts a N 2/3 dependency of ∆Eg (∆Eg is deﬁned as the diﬀerence between the measured optical gap of the doped ZnO ﬁlm and the optical gap of the intrinsic ZnO). Such a dependency has been observed for both LP-CVD ZnO and AP-CVD ZnO ﬁlms (see Fig. 6.42 for the case of AP-CVD ZnO). However, in some cases, a discrepancy with respect to Burstein–Moss theory has been experimentally observed: Roth et al. [41, 42] have studied absorption eﬀects in heavily-doped ZnO ﬁlms. They observed that for N > 3×1019 cm−3 , a shrinkage eﬀect of the gap occurs in addition to the Burstein– Moss eﬀect. This shrinkage of the gap is due to the merging of the donor and

6 CVD ZnO

B2H6/DEZ=0 B2H6/DEZ=0.85

B2H6/DEZ=0.3 B2H6/DEZ=1.3

275

B2H6/DEZ=0.5

80 60

TT

40

DT 20 0

400

600

800

1000 1200 1400 1600 1800 2000

Fig. 6.41. Total and diﬀuse transmittance curves for LP-CVD ZnO:B ﬁlms deposited at 155◦ C and 0.5 mbar as a function of the doping level, i.e., of the B2 H6 /DEZ ratio

Fig. 6.42. Energy band gap as a function of N 2/3 (N being the carrier density) for AP-CVD ZnO:B ﬁlms deposited at 375◦ C from DEZ, ethanol, and diﬀerent diborane concentrations. The linear ﬁtting curve is in good accordance with the Burstein–Moss model. Reprinted with permission from [28]

conduction bands, which occurs when the carrier density is increased above a critical value, leading to a so-called semiconductor/metal transition [43]. It can be set proportional to N 1/3 in a ﬁrst approximation [44, 45]. The superposition of both eﬀects (Burstein–Moss eﬀect and band gap shrinkage eﬀect) lead to absolutes values of ∆Eg that diﬀer from those predicted by the simple Burstein–Moss model. Figure 6.43 shows experimental values of the optical band gap Eg plotted vs. the free carrier density N for LP-CVD ZnO ﬁlms developed at IMT Neuchˆ atel. The ﬁlms have been deposited with

276

S. Fa¨ y and A. Shah

B2H6/DEZ 3.8

0

0.3

0. 6

1

1.5 2

1.51020

21020

Optical bandgap Eg [eV]

Eg (Experimental) 3.7

E0 + ∆EBM E0 + ∆EBM-∆EBGN

3.6 3.5 3.4 3.3 3.2

51019

11020

Carrier density N

[cm-3]

Fig. 6.43. Variation of band gap Eg measured on LP-CVD ZnO:B ﬁlms deposited at 155◦ C and 0.5 mbar and with various doping levels, in function of the carrier density N . The dashed line and the full line are the predicted variations of Eg taking into account the Burstein–Moss eﬀect alone, and both the Burstein–Moss and the band gap narrowing eﬀects, respectively. E0 is the band gap of undoped ZnO and is set for the evaluation here to 3.3 eV. Reprinted with permission from [33]

various B2 H6 /DEZ ratios, i.e., various doping levels. The dashed line and the full line are the predicted variations of Eg taking into account the Burstein– Moss eﬀect alone, and both the Burstein–Moss and the band gap narrowing eﬀects, respectively. The predicted values ﬁt well with the experimental data if both models are taken in account. The diﬀuse transmittance (DT) of LP-CVD ZnO:B layer is also shown in Fig. 6.41 for various B2 H6 /DEZ ratios. DT decreases when the doping ratio is increased. This behavior, which has also been observed for AP-CVD ZnO ﬁlms, is in accordance with the observed diminution of the surface roughness with an increase in doping ratio (see Sect. 6.2.5.2). It shows that the light scattering capability of CVD ZnO ﬁlms can be drastically reduced if too high dopant gas concentration is added into the CVD reactor. This emphasizes, once again, the necessity to improve, as much as possible, the electron mobility µ within the ZnO ﬁlms, in order to minimize the amount of dopant necessary to make the ZnO ﬁlms conductive enough for their application within solar cells.

6 CVD ZnO

277

6.2.6 The Eﬀect of Grain Size on Electrical and Optical Properties of CVD ZnO Layers As has been discussed above, there is a strong inﬂuence of grain size on the electrical and optical properties of ZnO ﬁlms. Take as a ﬁrst example, ZnO ﬁlms grown by the LP-CVD process in the substrate temperature range between 155 and 180◦ C: they have a microstructure as described in Sect. 6.2.2.1 with conical crystallites that form pyramids at the surface (see Fig. 6.6). This microstructure has, via the pyramidal structure of the surface, a pronounced inﬂuence on the optoelectronic properties of the ﬁlms, speciﬁcally on their light scattering capability. In Fig. 6.44 haze values of various LP-CVD ZnO samples deposited at IMT Neuchˆatel are presented as a function of the parameter δ (δ is deﬁned in Sect. 6.2.2.2: it is the dimensional parameter that describes the mean width of the grains appearing at the surface of the ZnO ﬁlms). There is a straightforward correlation between the haze value of the ﬁlms and their dimensional parameter δ: the larger the grains, the higher are the pyramids at the surface of the ZnO ﬁlms, the higher is the haze value (i.e., the higher is the light scattering capability of the ZnO ﬁlms). These observations highlight the importance to obtain an appropriate microstructure for the ZnO ﬁlms that are to be used within thin ﬁlm silicon solar cells. In this case, one wishes, in fact, to obtain ZnO ﬁlms with large crystallites in order to enhance their light scattering capability. A similar exercise has been carried out in Fig. 6.45, for the electron mobility values of LP-CVD ZnO ﬁlms. Here again, a clear correlation between the electron mobility and the dimensional parameter δ can be seen. Indeed, the 100

Haze600nm [%]

80

60

40

20

0

0

100

200

300

400

500

600

700

800

δ [nm] Fig. 6.44. Haze values (DT/TT measured at 600 nm) of various LP-CVD ZnO:B samples deposited at IMT Neuchˆ atel, as a function of the grain size parameter δ

278

S. Fa¨ y and A. Shah 40 35

Mobility [cm2/Vs]

30 25 20 15 Doped ZnO series 10 Doping series of ZnO 5 0

Undoped ZnO series 0

100

200

300

400

500

600

700

800

δ [nm]

Fig. 6.45. Carrier mobility µ of various LP-CVD ZnO:B samples deposited at IMT Neuchˆ atel as a function of the grain size parameter δ. The “doping” series is the one for which the gas doping ratio B2 H6 /DEZ has been varied. The “doped” and “undoped” series are those for which doped and undoped ZnO ﬁlms were deposited under a variation of deposition parameters, like substrate temperature or gas ratio H2 O/DEZ

following interpretation can be given to explain this correlation: the larger the grains, the lower is the density of grain boundaries, the less pronounced is electron scattering by grain boundaries, and the higher are the values of electron mobility. The electron mobility µ reaches a maximum value of ∼40 cm2 (V s)−1 . Above this value, µ does not continue to increase with δ. Indeed, for δ > 500 nm, the crystallites that constitute the ZnO ﬁlms are so large that the grain boundary scattering eﬀect is no more the limiting factor for µ. A similar dependency of both light scattering capability and electron mobility on grain size has also been observed for ZnO ﬁlms deposited by the AP-CVD process. Indeed, the highest values of µ (i.e., 36 cm2 (V s)−1 ) and of diﬀuse transmittance reported for the AP-CVD process have been obtained for ZnO:F ﬁlms grown at 450◦C, and it is precisely for these ﬁlms that the largest grain size has been measured [15]. In conclusion, it has been proven that grain size plays an important role in determining the properties of polycrystalline CVD ZnO ﬁlms. Larger grains lead to an improved light scattering capability, and this allows one to achieve better light trapping within thin ﬁlm silicon solar cells. Furthermore, larger grains lead also to a reduced density of grain boundaries, and this yields higher values of electron mobility. Finally, it has to be noted that a reduction in the density of grain boundaries may also have a beneﬁcial eﬀect on the stability of the ZnO ﬁlms, as well as on the performance of thin ﬁlm silicon solar cells deposited on top of

6 CVD ZnO

279

the ZnO ﬁlms. Indeed, it is suspected that these grain boundaries act also as charge traps and as sites where a multitude of diﬀerent molecules can diﬀuse and change the local properties of the ﬁlms. This can aﬀect the intrinsic properties of the ZnO, mainly the electrical properties (by widening the depletion barrier of the grain boundaries and therefore further reducing µ). Additionally, the out-diﬀusion of species trapped at grain boundaries may also lead to deterioration of the properties of the semiconductor material deposited on top of the ZnO layer. 6.2.7 Alternative CVD Methods for Deposition of Thin ZnO Films Apart from the classical processes of AP-CVD and LP-CVD described in this chapter, some alternatives of CVD processes have also been tried in the laboratory for ZnO deposition: – – – –

Laser-induced CVD Photo-CVD Plasma-enhanced CVD (PE-CVD) Expanding-Thermal-Plasma CVD (ETP-CVD)

Laser-induced LP-CVD of ZnO has been studied ﬁrst with the goal of improving the epitaxial growth of c-axis oriented ZnO (Note: c-axis oriented means that the c-axis is perpendicular to the substrate) [46, 47]. The growth of monocrystalline ZnO has been achieved under certain deposition and UV illumination conditions. This technique has then been applied by Yamada et al. [23, 48] to study the impact of UV irradiation during the Photo-CVD growth of polycrystalline ZnO, using a UV lamp added to the standard LPCVD process. No signiﬁcant change in the polycrystalline growth of the ZnO has been observed, compared to the growth of LP-CVD ZnO ﬁlms. However, a signiﬁcant increase of ﬁlm conductivity has been observed, coming mainly from the increase in free carrier mobility. Indeed, undoped LP-CVD ZnO ﬁlms had their mobility increased from 22 to 60 cm2 (V s)−1 by UV light irradiation. Furthermore, Yamada et al. deposited boron-doped photo-CVD ZnO ﬁlms with mobility values around 50 cm2 (V s)−1 and carrier densities (which were also increased by UV light irradiation) around 2 × 1020 cm−3 ; this leads to resistivity values of 6 × 10−4 Ω cm. In comparison, the same LP-CVD ZnO:B ﬁlms grown without UV light irradiation have a resistivity around 2 × 10−3 Ω cm. In the same manner, PE-CVD of ZnO has been ﬁrst studied with the goal of improving the crystallinity of c-axis oriented ZnO ﬁlm growth at low substrate temperature (i.e. between 150 and 300◦ C) [49, 50]. More recently, expanding thermal plasma has been coupled with an LP-CVD chamber in order to deposit natively rough ZnO ﬁlms [51–53]. Fig. 6.46 shows the expanding thermal plasma aixtron reactor system used at Eindhoven University of Technology to deposit PE-CVD ZnO ﬁlms.

280

S. Fa¨ y and A. Shah

Fig. 6.46. Schematic diagram of the expanding thermal plasma aixtron reactor system used at Eindhoven University of Technology to deposit PE-CVD ZnO ﬁlms [51]. DEZ and oxygen are used as precursors for ZnO formation. Trimethyl-aluminium (TMA) is used as doping source. The total pressure is ∼2.5 mbar and the substrate temperature is varied from ∼200 to 350◦ C. Reprinted with permission from [85]

As none of these alternative processes have so far been used industrially for deposition of large area ZnO as TCO for thin ﬁlm solar cells, they will not be treated in details here.

6.3 CVD ZnO as Transparent Electrode for Thin Film Solar Cells 6.3.1 Characteristics Required for CVD ZnO Layers Incorporated within Thin Film Solar Cells The present paragraph investigates in more detail the various ways in which CVD ZnO layers have been incorporated into thin ﬁlm solar cells. It, thus, attempts to deﬁne a framework that can help the solar cell designer to choose the most appropriate CVD deposition parameters for producing his ZnO layers. The properties of ZnO layers that have an importance for their application within the context of thin ﬁlm solar cells are (a) transparency,

6 CVD ZnO

281

(b) conductivity, and (c) especially in the case of thin ﬁlm silicon solar cells, surface roughness. The conductivity of ZnO layers can be augmented by increasing either the free carrier density or their mobility. However, a too high free carrier density implies also a high free carrier absorption in the near-infrared spectral region; the latter may have a harmful eﬀect on the spectral response of the solar cell, especially in the case of microcrystalline silicon and other materials with low band gaps. It is, therefore, in general advisable to ﬁrst optimize the process parameters with a view of obtaining the highest possible mobility values, and only in a second step to increase the carrier density, in order to arrive at the conductivity level required by the solar cell speciﬁcations. In the previous paragraphs, the authors have described in detail the development of AP-CVD and LP-CVD processes for ZnO ﬁlms. They have shown thereby that, for both processes, there is an optimum temperature for which the ZnO ﬁlms are strongly oriented along a single, preferential orientation, and that this generally leads to large grains. The optimum deposition temperature is around 400◦C for AP-CVD process and produces ZnO ﬁlms with a strong orientation perpendicular to the (0002) crystallographic planes. The optimum deposition temperature is around 160◦C for the LP-CVD process, and produces ZnO ﬁlms with a strong orientation perpendicular to the (11¯20) crystallographic planes. The CVD ZnO ﬁlms deposited at these optimum values of temperature are constituted of large grains that are associated with a high surface roughness, and which become wider as the thickness of the ZnO ﬁlms is increased. These large grains lead to a reduced density of grain boundaries; and this, in its turn, minimizes the inﬂuence of the grain boundary scattering eﬀect on electron mobility. Indeed, it is in the same optimum deposition temperature range that the highest mobility values have been obtained. Finally, in the very same temperature range, a high optical transparency is regularly observed, associated with a high light scattering capability. The latter is a direct consequence of the large grains appearing at the surface and the resulting as-grown surface roughness. It is therefore well proven now that these two optimum deposition temperatures (i.e., 400◦ C for AP-CVD and 160◦ C for LP-CVD) will produce those speciﬁc ZnO layers that are most adapted for thin ﬁlm solar cell applications. Based on the speciﬁc parameter ﬁeld (for µ, ρ, absorption, light scattering, . . . ) obtained within the optimum temperature range, one has thereafter to choose the adequate thickness and gas doping ratio for the application envisaged. This will depend on the speciﬁc use of CVD ZnO within the various possible thin ﬁlm solar cell conﬁgurations. 6.3.1.1 CIS (and CIGS) Thin Film Solar Cells Thin ZnO ﬁlms can be used either as a transparent and conductive window layer, or as a buﬀer layer, within CuInS2 (CIS) and Cu(In,Ga)Se2 (CIGS) thin ﬁlm solar cell devices (see Chaps. 4 and 9). In both cases, the ZnO layers

282

S. Fa¨ y and A. Shah Sunlight

Front TCO Buffer layer CIS Molybdenum Glass

Fig. 6.47. A CIS solar cell that uses ZnO layer as window layer (front TCO) and/or buﬀer layer

are deposited after the CIS/CIGS solar cell (see Fig. 6.47). For this reason, the LP-CVD technique is the more advantageous method of the two main CVD processes. Indeed, the low deposition temperature of the LP-CVD process avoids, or at least reduces, the damage to the underlying solar cell. CVD ZnO as a Transparent and Conductive Window Layer The front window layer in a CIS or CIGS solar cell is the last-deposited ﬁlm of the whole structure. It has to be transparent along the whole absorption spectrum of the solar cells, i.e., up to wavelengths as long as ∼1 300 nm. Furthermore, its conductivity has to be high enough to transport the large photogenerated current resulting from CIS cells (around 30 mA cm−2 ) to the metallic collection grid. The desired sheet resistance depends, of course, on the spacing of the metallic contacts used for the collection grid of the solar cell (see also Sect. 9.2.1), but it is usually between 5–10 Ωsq . CIS or CIGS is a material with a direct band gap; one of its main advantages is that it possesses a very high absorption coeﬃcient α; in fact α is high enough, so that under normal circumstances, no light trapping at all is necessary, i.e., no additional light scattering needs to be obtained by the front window layer. In practice, various kinds of TCO are used as front TCO here: ITO, sputtered ZnO, and CVD ZnO. The roughness of the CVD-type ZnO is not a decisive advantage for CIS/CIGS solar cells. However, several groups have reported that a rougher front ZnO actually leads to a higher photogenerated current [54,55]. Indeed, the roughness of the front ZnO probably reduces the primary reﬂection, and therefore lets more light enter into the solar cell. An ideal front CVD ZnO layer would be a lightly doped, or even an undoped layer thick enough to lower the sheet resistance down to 5 Ωsq . Such a ZnO ﬁlm has almost no free carrier absorption, which makes it highly transparent over the whole visible range, right up to the near infra-red range. Indeed, the free carrier absorption is proportional to the carrier density and

6 CVD ZnO

283

the thickness, but is also inversely proportional to the electron mobility [31]. On the one hand, enhancing the thickness of a lightly doped ZnO ﬁlm will increase free carrier absorption, but as the mobility will also be enhanced (see Sect. 6.2.2.2), this will limit the overall increase. On the other hand, adding more dopant within the ZnO ﬁlms will increase the carrier density and decrease the mobility (see Sect. 6.2.5.2). Combined together, the two trends will induce an overall enhancement of free carrier absorption within CVD ZnO ﬁlms. For this reason, one would rather increase the thickness than increase the doping level of a TCO ﬁlm, in order to decrease the sheet resistance of the ﬁlm while maintaining free carrier absorption as low as possible. Undoped LP-CVD ZnO ﬁlms with a thickness around 6 µm have been deposited by the group of IMT Neuchˆ atel [56]; they have a very high optical transparency and a sheet resistance of just 10 Ωsq , a value which is low enough that these ZnO layers could, in fact, act as electrical contacts for solar cells. In practice, a thickness of 6 µm can be considered to be too high from the point of view of production costs. Cracks may also appear in such very thick ZnO ﬁlms, leading to peeling and adhesion problems. Finally, very thick ZnO ﬁlms may create supplementary diﬃculties while patterning the ZnO during the production step of series connection for module production. It would therefore certainly be desirable to add a small amount of dopant, in order to reduce the sheet resistance of the resulting ZnO ﬁlms, without increasing the thickness to such high values. Therefore, in a practical situation, one may ﬁrst determine the maximum thickness allowed, and then evaluate how much dopant should be added, so as to lower the sheet resistance to just the desired limit value. An additional point in favor of doping the ZnO ﬁlms, even only slightly, is that adding just a very small amount of dopant (e.g., B2 H6 /DEZ ∼0.1 for LP-CVD ZnO) can already help to stabilize the electrical properties of these ﬁlms. CVD ZnO as a Buﬀer Layer A thin i-ZnO (intrinsic ZnO) layer is often used as a buﬀer layer in CIS/CIGS solar cells, between the absorber part of the cell and the front TCO. The role of this resistive (buﬀer) layer is mainly to provoke suitable ﬁeld-assisted hole collection at the contact interface, reducing, thus, the recombination rate at and near the ZnO/CIS or ZnO/CIGS interface (see also Chap. 9). Olsen et al. observed that if the resistivity of such a i-ZnO buﬀer layer is too low, one obtains a reduced eﬃciency of the CIS solar cell, mainly due to a drop in solar cell Fill Factor [57]. This kind of i-ZnO ﬁlm should therefore be resistive enough (ρ ∼ 1, 000 Ω cm). No dopant has to be used in this case. As the thickness of this ﬁlm is quite low (typically 30–50 nm), the ﬁlm will have a low mobility and it is therefore quite easy to obtain the high value of resistivity mentioned above.

284

S. Fa¨ y and A. Shah

6.3.1.2 Thin Film Silicon Solar Cells3 In thin ﬁlm silicon solar cells, one generally distinguishes between the superstrate or p–i–n conﬁguration and the substrate or n–i–p conﬁguration. In the ﬁrst case the deposition sequence is as follows: superstrate (i.e., glass), front contact (e.g., ZnO), p-layer, i-layer, n-layer, and then back reﬂector and back contact (e.g., ZnO + metal). Here, light enters into the solar cell through the glass substrate (which is therefore also called a superstrate). In the second case the deposition sequence is as follows: substrate (often metal sheet or plastic foil), back contact and back reﬂector (e.g., metal + ZnO), n-layer, i-layer, p-layer and, ﬁnally, front transparent contact layer (e.g., ZnO layer). Here, light enters into the solar cell through the top (last deposited) ZnO layer (just like in the case of CIS/CIGS solar cells). The second conﬁguration has to be used when an opaque (nontransparent) substrate (i.e., metal sheet or plastic foil4 ) is chosen as support for the solar cell deposition. Furthermore, one has to distinguish between amorphous silicon solar cells, on the one hand, and microcrystalline (and micromorph) solar cells on the other hand. Indeed, these two categories of solar cells have diﬀerent spectral ranges for absorption and photogeneration: the spectral range for absorption and photogeneration in amorphous silicon extends only up to about 800 nm, whereas microcrystalline silicon can absorb and convert light up to at least 1,000 nm. Finally, amorphous silicon is less sensitive to the quality and roughness of the surface on which it is deposited, whereas it is, in general, much more diﬃcult to deposit good microcrystalline silicon solar cells on a too rough substrate. ZnO can be advantageously used as TCO layer in various kinds of thin ﬁlm silicon solar cells, either as back or as front contact, or even as an intermediate reﬂector between the amorphous and the microcrystalline p–i–n junctions, within the micromorph tandem solar cell [58] (see also Chap. 8). Figure 6.48 illustrates the various possibilities for using a ZnO layer within a thin ﬁlm silicon solar cell. In the present paragraph, we will comment about the use of CVD ZnO both as front contact (or window layer ) and also as back contact (or part of the back reﬂector ). Conditions for Transparency and Conductivity Front ZnO layers: Somewhat similarly to the case of ZnO ﬁlms that are used as window layers in CIS/CIGS solar cells, the front ZnO in a thin ﬁlm silicon solar cell has to fulﬁll the following criteria: – A high transparency over the whole range of absorption of the solar cell. This condition is quite easy to fulﬁll for amorphous silicon solar cells, as 3 4

See also Sect. 8.2. Plastic foil has to be considered as being nontransparent because it can loose transparency and become “yellow” during UV exposure.

6 CVD ZnO

285

Fig. 6.48. A micromorph tandem cell, for which a ZnO layer can be used as front contact (i.e., window layer), back contact, and/or intermediate reﬂector [58]

the free carrier absorption present on doped CVD ZnO layers usually does not appear in the range of absorption of amorphous silicon. On the other hand, free carrier absorption can appear in the [800–1,000 nm] wavelength range, which is indeed a part of the absorption range of microcrystalline (and micromorph) solar cells. For the latter types of solar cells, it is therefore mandatory to minimize, as much as possible, the density of dopants that one introduces in order to lower the sheet resistance of the ZnO ﬁlm. – The sheet resistance, in its turn, must be low enough so as to keep the resistive series losses below a certain limit. The appropriate value depends on the geometry of the metallic contacts of the connection grid (see also Sect. 8.2.5); for thin ﬁlm silicon solar cells and modules, the sheet resistance is generally in the range between 5 and 10 Ωsq . The sheet resistance can be reduced by either adding more dopant within the CVD reaction or by increasing the layer thickness. As for the case of window layers for CIS/CIGS solar cells (discussed in Sect. 6.3.1.1), it is better to increase the thickness, especially for microcrystalline (and micromorph) solar cells. Indeed, free carrier absorption within ZnO ﬁlms is more rapidly increased by adding more dopant atoms than by increasing the ﬁlm thickness. Back ZnO layers: The transparency and conductivity criteria for the back ZnO layer, in a thin ﬁlm silicon solar cell can be formulated as follows: – As for the useful spectral range to be taken in account, similar considerations can be made as for the front ZnO layer, diﬀerentiating thereby between amorphous solar cells, on the one hand, and microcrystalline (micromorph) solar cells, on the other hand. However, one may note that the short wavelengths usually do not reach the back ZnO at all, because they are absorbed during the ﬁrst transit of the light through the solar cell. Therefore, the transparency of the back ZnO layer in the short spectral range is not a critical point.

286

S. Fa¨ y and A. Shah

– The sheet resistance of the back ZnO layer is a critical point only if the reﬂector used is nonconductive. In this case, the same considerations apply as for the front ZnO. If the reﬂector itself is a metallic layer (e.g., silver or aluminum), the latter will collect all the current that is photogenerated within the cell. In this case, the back ZnO is not used for current collection, but for optical reasons. Conditions for Obtaining a High Light Scattering Capability of CVD ZnO Because of the low absorption coeﬃcient of amorphous and microcrystalline silicon, it is mandatory to optimize light scattering within thin ﬁlm silicon solar cells by the use of suitably textured (rough) interfaces and surfaces. This paragraph comments about the ideal surface roughness for ZnO layers deposited by CVD, and used as front or back contacts within amorphous, microcrystalline (and micromorph) solar cells. We mentioned in Sect. 6.2 that to obtain the highest surface roughness with a CVD process, one has to use thick, undoped (or lightly doped) ZnO ﬁlms. Therefore, in a similar manner as described above, one should reduce the doping level and increase the thickness of the ZnO ﬁlms used, so as to obtain not only the highest optical transparency possible, but also the most pronounced light scattering eﬀect. Here again, the highest thickness that can be deposited is limited by other factors like production costs, adherence of ZnO ﬁlms on glass or on silicon, the ability to do subsequent laser patterning, and so on. Furthermore, if the underlying ZnO layer is too rough, growth problems for the subsequent silicon layers may appear. As an example, in [56] we report on experiments carried out at IMT Neuchˆ atel, where an increase in the surface roughness of the front ZnO layer (within a p–i–n-type microcrystalline cell) lead (as expected) to an increase of the photogenerated current current density Jsc , but also to a systematic drop in ﬁll factor FF and open circuit voltage Voc .5 This critical point highlights the fact that the growth of silicon, especially of microcrystalline Si (µc-Si:H), on rough ZnO layers is not a straightforward process and may lead to unexpected problems. There are yet other, purely optical considerations regarding the optimal roughness value for ZnO layers (some of these points are also discussed in Sect. 8.2.4.1): – It is not desirable that the front TCO layer has a too high light scattering eﬀect in the short wavelength range; indeed this may lead to this part of the solar spectrum being trapped within the ZnO layer itself (or within the p-layer of the cell) and never reaching the photovoltaically active i-layer. It is therefore quite diﬃcult to determine exactly the optimum surface roughness for the front ZnO layer: this layer should, in fact, scatter enough light without absorbing too much. To obtain suﬃcient light scattering for the longer wavelengths of the solar spectrum (i.e., for the near-infrared 5

Explanations for FF, Voc , and Jsc are given in Sect. 8.2.1.

6 CVD ZnO

287

(NIR) spectral range), it is therefore better to rely on a very rough back ZnO layer; indeed the shorter wavelengths never reach the back ZnO layer and therefore do not constitute a problem in this case. – The diﬀuse transmittance (DT) of a ZnO ﬁlm is, in general, evaluated with the help of a spectrophotometer, in a simpliﬁed glass/ZnO/air conﬁguration and not within an actual solar cell conﬁguration. The real path of the light within the whole solar cell structure (a structure containing a multitude of layers with diﬀerent indices of refraction and forming a multitude of optical interfaces), can therefore not be directly deduced from the just described simpliﬁed measurement of the diﬀuse transmittance (DT). On the other hand, measuring the transmittance of the whole solar cell structure does not allow one to identify the contributions of the individual layers contained within this structure. – A useful supplementary measurement is that of the angular distribution function (ADF) of the scattered light. This can be helpful to determine which kind of surface roughness (i.e., pyramids, craters, steps, . . . ) produces the most adequate light scattering eﬀect. One generally desires to have a light scattering eﬀect over large angles. – Finally, as it is not possible to experimentally test all the various kinds of surface textures within actual solar cell conﬁgurations, it can be useful to use numerical simulations, in order to evaluate the best combination of surface textures and roughness for both front and back TCO layers. The method usually applied for such simulations is to take the main optical properties of each layer of the solar cell (absorption, thickness, haze factor, ADF, surface roughness, . . . ), and then to put them all together in order to compute the quantum eﬃciency curve of the resulting solar cell. Such a task of optically simulating solar cells is very complex and beyond the scope of the present chapter. However, it is important to note here that a numerical simulation is always only an imperfect tool and can in no way fully replace experimental work and measurements on actual solar cells. To conclude we may say that, for thin ﬁlm silicon solar cells (just as in the case of CIS/CIGS solar cells), the low-temperature LP-CVD process is preferred over the AP-CVD process, in all cases for where ZnO is deposited as the last layer: this is the bottom layer ZnO for the p–i–n conﬁguration and the top layer ZnO for the n–i–p conﬁguration. 6.3.1.3 The Use of Figures of Merit to Predict the Ideal CVD ZnO for Thin Film Solar Cells As an alternative to numerical simulation, we may also use experimental measurements and evaluate the latter with an appropriate ﬁgure of merit (FoM). Such a FoM will allow us to compare the performance of diﬀerent transparent conductors.

288

S. Fa¨ y and A. Shah

Gordon et al. [59] proposed a ﬁgure of merit deﬁned by the ratio of the electrical conductivity over the optical absorption coeﬃcient in the visible spectral range. They tested many dopants for AP-CVD ZnO ﬁlms, and obtained the highest ﬁgure of merit for ﬂuorine-doped ZnO, which they used as TCO for amorphous silicon solar cells. It is, however, certainly useful to include the light scattering property of TCO ﬁlms in the ﬁgure of merit. Furthermore, one should also take into account the exact spectral range where the TCO has to operate (i.e., a diﬀerentiation is here necessary between amorphous and microcrystalline silicon solar cells). With this in mind, Fa¨ y et al. proposed a new, wavelengthdependent Figure of Merit (FoM(λ)) [35]: FoM(λ) = H(λ) ∗ (TT(λ) + TR(λ)) ∗ Rdev ! 1 if Rsq < Rref Rdev = Rref if Rsq > Rref Rsq

(6.7)

Here, Rref is the reference value of sheet resistance: it is the maximum value of sheet resistance that can be tolerated for the solar cell, taking into consideration the spacing between the metallic contacts of the collection grid. This new ﬁgure of merit FoM(λ) consists of successive multiplications of factors comprised between 0 and 1: the haze factor, the factor TT + TR (total transmittance and reﬂectance), and the factor Rdev , these are weighting coefﬁcients that describe the light scattering capability, the transparency, the reﬂectivity, and the electrical resistance, respectively. Absorbance and resistance, which are the physical parameters that have a direct bearing on solar cell performance, have been chosen here, rather than absorption coeﬃcient and resistivity, which are intrinsic material properties of the TCO ﬁlms. To illustrate this new ﬁgure of merit for the case of rough TCO ﬁlms, the value of FoM(λ) is shown in Fig. 6.49, for LP-CVD ZnO:B layers deposited with diﬀerent gas doping ratios B2 H6 /DEZ (i.e., with diﬀerent doping levels), with the reference value of sheet resistance ﬁxed at Rref = 10 Ωsq and Rref = 5 Ωsq . The result can be resumed as follows: if the geometry of the collection grid of the solar cell allows one to use a TCO layer with Rsq = 10 Ωsq , then the optimum TCO would be the one deposited with a B2 H6 /DEZ ratio of 0.3. On the other hand, if a Rsq value of only 5 Ωsq is required, the optimum TCO would be the one deposited with a B2 H6 /DEZ ratio of 0.5. Furthermore, a series of ZnO:B ﬁlms with various doping levels has been deposited, and the thickness of the ﬁlms has been adapted so as to keep the sheet resistance close to 10 Ωsq . This means that the thickness of the ZnO ﬁlms was increased as their doping level was decreased. For this series, the highest FoM(λ) of LP-CVD ZnO ﬁlms on the whole visible spectral range has been obtained for the 6 µm-thick undoped ﬁlm. This prediction has then been experimentally conﬁrmed: the same microcrystalline silicon solar cell was deposited on this ZnO series, and the highest photogenerated current

1

B2H6/DEZ = 0 B2H6/DEZ = 0.29

0.8

B2H6/DEZ = 0.5 B2H6/DEZ = 0.85

0.6

B2H6/DEZ = 1.3 Rref = 10 Ωsq

0.4

0.8

289

B2H6/DEZ = 0 B2H6/DEZ = 0.29

0.7

Figure of Merit

Figure of Merit

6 CVD ZnO

0.6

B2H6/DEZ = 0.5 B2H6/DEZ = 0.85

0.5

B2H6/DEZ = 1.3

0.4 Rref = 5 Ωsq

0.3 0.2

0.2 0.1 0

400

600

800

1000

1200

0

400

Wavelength [nm]

600

800

1000

1200

Wavelength [nm]

Fig. 6.49. Wavelength-dependent Figure of Merit FoM(λ) deﬁned in (6.7) of LP-CVD ZnO:B ﬁlms deposited with various doping levels, i.e., B2 H6 /DEZ ratios. The reference value of sheet resistance Rref has been set to (a) 10 Ωsq and (b) 5 Ωsq

was actually measured for the 6 µm-thick undoped layer [56, 60]. However, the values of FF and Voc were reduced, in this case, suggesting that growth problems occurred with the microcrystalline silicon layers deposited on this speciﬁc, very rough ZnO layer. This highlights again the importance of actually fabricating complete solar cells with the various ZnO ﬁlms one chooses to develop. This ﬁgure of merit, which takes in account the light-scattering capability of rough TCO, can be taken as a basis for TCO integration within thin-ﬁlm silicon solar cells. It can also be reﬁned by adding in (6.7) new weighting factors that could describe, for example, the Angular Distribution Function of the rough TCO. Indeed, the haze factor is suﬃcient to compare the roughness variation of similar surface morphologies, but it may be not suﬃcient enough when too diﬀerent surface morphologies are compared. In this case, it would be helpful to know, for example, the angle along which the light is preferentially scattered. 6.3.2 Experimental Results 6.3.2.1 CIS and CIGS Thin Film Solar Cells Concerning the use of CVD ZnO ﬁlms as window layers in CIS/CIGS solar cells (see paragraph 6.3.1.1), the most frequently used precursors for the deposition, by CVD, of such ZnO window layers are DEZ, H2 O, and, as doping source, diborane [54, 55, 61, 62]. The as-grown rough surface of LP-CVD ZnO layers allows these layers to act both as electrodes and as (partial) antireﬂection coatings. Sang et al. experimentally proved the advantage of using a rough front TCO instead of a ﬂat one. Figure 6.50 shows I–V characteristics of CIGS-submodules with an aperture area of 864 cm2 , which use either

290

S. Fa¨ y and A. Shah

Fig. 6.50. I–V characteristics of CIGS-based submodules (aperture area: 864 cm2 ) with diﬀerent windows. Reprinted with permission from [55]

a rough CVD ZnO:B layer or a ﬂat sputtered ZnO:Ga layer as front TCO window. The rough surface of CVD ZnO:B induces a better light absorption within the solar module. This leads to an increase of the photogenerated current from 29.90 to 33.32 mA cm−2 , and an increase of the eﬃciency from 10.82 to 12.03%. Shell Solar in Camarillo (California, USA), which patented, in 1988, the original CVD process for ZnO deposition [63–65], reported in 2002 on the status of their CIS module production unit, where LP-CVD ZnO layers were used as transparent window layers [66]: this unit was started in 1988 and was until recently producing 1 MWp of CIS modules per year, with up to 40 W per individual module. They reported an average aperture eﬃciency of 11%. Furthermore, they studied the drift of the growth rate of CVD ZnO between two cleaning processes of the reactor: Fig. 6.51a shows that for a same deposition time, the thickness of the CVD ZnO ﬁlms is continuously decreased from ∼2.2 to ∼1.8 µm over the course of around one thousand process runs of ZnO deposition. This drift is attributed to a progressive increase of parasitic ZnO coating on the reactor walls, which induces an increase of the surface area having a higher thermal emissivity. The deposition of ZnO on the walls will be progressively increased, implying, thus, less deposition on the substrate (i.e., on the CIS module). Fortunately, Fig. 6.51b shows that CIS module performance is completely insensitive to this thickness variation of the CVD ZnO ﬁlms. Concerning the use of CVD ZnO ﬁlms as buﬀer layers between the absorber part of the cell and the front TCO (see Sect. 6.3.1.1), highly resistive

6 CVD ZnO

291

(a)

(b) Fig. 6.51. (a) Deposition rate control chart for CVD of ZnO. (b) 1 × 4 square feet laminated module performance during the ZnO CVD reactor cycle for which the data of (a) were collected. Reprinted with permission from [66]

CVD ZnO layers grown by reacting either a zinc adduct with tetrahydrofuran or DEZ and H2 O vapors have proven to be eﬃcient buﬀer layers [55, 67, 68]. Terzini et al. [68] compared the use of sputtered ZnO with the use of LPCVD ZnO as buﬀer layer for CIS solar cells. They pointed out that, even with identical ZnO layers characteristics, the deposition conditions can modify the solar cell performances through a possible eﬀect on the CdS layer on which the ZnO buﬀer layer is deposited. By carefully optimizing the H2 O/DEZ ratio and the substrate temperature during the deposition of the CVD ZnO layer, they obtained CIS solar cells with Fill Factors of 70% and Voc of 708 mV, higher than the values of FF and Voc obtained by optimizing the sputtering deposition conditions of ZnO for its use as buﬀer layer (FF = 60.3% and Voc = 675 mV). These results highlight one advantage of the CVD process, it is a smoother deposition process. Indeed, the CVD process could, in certain cases, induce less defects on the layer on which the ZnO is deposited.

292

S. Fa¨ y and A. Shah

To improve the junction property between CIGS absorber/Zn(O,S,OH)x and ZnO:B window, Sang et al. [55] inserted an undoped thin CVD ZnO layer between the Zn(O,S,OH)x buﬀer and the ZnO:B window. They observed an increase of both the Voc and FF, leading to a 30×30 cm2 submodule (aperture area = 864 cm2 ) with 12.93% eﬃciency. Finally, Olsen et al. [67] reached an eﬃciency of 13.95% (aperture area: 0.13 cm2 ) for CIGS solar cells using a CVD ZnO buﬀer layer (see also Chaps. 4 and 9). 6.3.2.2 Thin Film Silicon Solar Cells Amorphous Silicon

Jsc [mA/cm2]

Voc [V]

Cells: Gordon [59] and Hegedus et al. [69] compared the performance of a-Si:H solar cells grown on various rough front TCO substrates, like SnO2 , LP-CVD ZnO:B, and AP-CVD ZnO:F, having diﬀerent haze factors. Systematically, higher values of photogenerated current in the a-Si:H solar cell have been obtained for CVD ZnO front TCO. However, lower values of FF and Voc were measured in those cases, this being attributed to the degradation of the CVD ZnO/p-layer interface. Wenas et al. [70] suggested that desorption of H2 O from LP-CVD ZnO:B deposited at low temperature during the subsequent growth of the amorphous silicon layer could damage the ZnO/p-layer interface. Indeed, as illustrated in Fig. 6.52, they observed an increase of all the a-Si:H solar cell

0.88 0.87 0.86 18 17 16

FF

0.74 0.72

Eff. [%]

0.70 12 11 10 as grown 200

250

300

350

Annealing temperature [˚C]

Fig. 6.52. Dependence of solar cells performance on the front CVD ZnO preannealing temperature. Reprinted with permission from [70]

6 CVD ZnO

293

parameters after an annealing of the ZnO layer at 250◦C prior to the solar cell deposition. Furthermore, by carefully optimizing the roughness of the front CVD ZnO (by varying its thickness, its doping level, and the growth rate), and by combining it with a highly reﬂective ZnO/AgAl rear contact, Wenas et al. [36, 71] reached a high initial conversion eﬃciency of 12.5% (Voc = 0.889 V, Jsc = 19.46 mA cm−2 , FF = 0.721; area: 3 × 3 mm2 ). Finally, Wenas et al. [72] also proposed a two-step LP-CVD process in order to smoothen the “too sharp” surface morphology of as-grown LPCVD ZnO by depositing a thin layer of ﬂat LP-CVD ZnO layer grown with DEZ/ethanol/H2 O reactant. By varying the thickness of both layers, various surface morphologies have been obtained, which can be suitable for thin ﬁlm silicon solar cell deposition. Sang et al. [73] used also a bi-layer structure for the ZnO deposition, but this structure consisted in their case of a rough LP-CVD ZnO layer and a thin ZnO layer deposited by atomic layer deposition (ALD). For cells with a small area of 0.09 cm2 , not much diﬀerence was observed in the performance between the cells using LP-CVD ZnO layers and those using the bi-layer structure. However, for cells with a larger area of 1 cm2 , improvement of the performances, especially the Fill Factor, has been observed for the cells deposited on the bi-layer ZnO structure. Sang et al. reached an initial eﬃciency of 9.72% (Voc = 0.915 V , Jsc = 14.44 mA cm−2, FF = 0.736; area: 1 cm2 ), leading to a stabilized eﬃciency of 8.2%. Addonizio et al. [16] studied the use of LP-CVD ZnO:B deposited with DEZ, H2 O, and B2 H6 with various solar cell conﬁgurations, as front or back TCO. They also observed a slight drop of Voc and FF together with an increase of Jsc when using LP-CVD ZnO instead of commercial SnO2 as front TCO for a-Si:H solar cells. Furthermore, they highlighted the advantage of the LP-CVD process, which provides less bombardment of the solar cell surface than the usual ITO sputtering process, when using LP-CVD ZnO as back contact within the p–i–n solar cell conﬁguration, or as front contact within the n–i–p solar cell conﬁguration. L¨ oﬄer et al. successfully deposited a-Si:H solar cells on their in-house ZnO grown by expanding thermal plasma CVD [52]. They reached initial eﬃciencies of around 10%, and supported also the need to improve the ZnO/p-layer interface. The IMT Neuchˆatel group experimentally proved the high potential of as-grown rough LP-CVD ZnO used as front TCO in a-Si:H solar cells [2]. Comparison of the relative spectral response of a-Si:H pin cells shown in Fig. 6.53 highlights the more eﬃcient light-trapping that takes place when using LP-CVD ZnO instead of SnO2 as front TCO. Indeed, a clear increase of the photogenerated current is observed in the wavelength range above 500 nm for p–i–n a-Si:H solar cells deposited on LP-CVD ZnO. The IMT Neuchˆatel group achieved a stabilized eﬃciency of 9% (cell area: 1 cm2 ) for a 0.25 µm-thick a-Si:H p–i–n solar cell deposited on LP-CVD

S. Fa¨ y and A. Shah

Spectral response [arb. units]

294

ZnO SnO2

glass/a-Si:H p-i-n

400

600

800

Wavelength [nm]

Fig. 6.53. Comparison of the relative spectral response of 0.35 µm thick a-Si:H p–i–n cells deposited on glass substrates covered with LP-CVD ZnO and SnO2 . Reprinted with permission from [74]

ZnO with FF and Voc values as high as those obtained with amorphous cells deposited on commercially available SnO2 ﬁlms, but with higher photogenerated current [74, 75]. They demonstrated the beneﬁcial eﬀect of the rough surface of LP-CVD ZnO:B layers on the photogenerated current, due to a better light trapping and reduced reﬂection loss of the LP-CVD ZnO cell system. Furthermore, this enhanced light trapping allowed them to limit the thickness of the amorphous silicon layers to 0.25 µm, reducing thereby the light-induced degradation within the a-Si:H solar cell. In 2003, the IMT Neuchˆatel group published [76] a record stabilized eﬃciency of 9.47% for a 1 cm2 single-junction a-Si:H solar cell using an LP-CVD ZnO layer as front TCO with an antireﬂective coating. The characteristics of this cell, independently conﬁrmed by NREL, are shown in Fig. 6.54. Modules: In 1993 already, the Siemens group was the ﬁrst group to deposit successfully amorphous silicon (a-Si:H) solar modules on ZnO layers grown by the CVD method. They reported [77] modules of 1 × 1 ft2 with an initial eﬃciency of 10.7% and a stabilized eﬃciency of 8.4%. More recently, Oerlikon Solar started the development of large-area R&D equipment (1.4 m2 ) for the deposition of ZnO by the LP-CVD process. The results obtained so far are promising: Oerlikon Solar fabricated 1.4 m2 a-Si:H modules with 112.4 W of initial power (see Fig. 6.55) [78]. These results are already above the results obtained on 1.4 m2 a-Si:H modules that use commercial SnO2 from the AFG company as front TCO. Furthermore, Oerlikon Solar developed also an LP-CVD ZnO/white back reﬂector contact concept, which showed a high light-trapping potential [79].

6 CVD ZnO

295

Fig. 6.54. NREL AM1.5 I–V characteristics of an amorphous p–i–n single junction solar cell deposited on LP-CVD ZnO coated glass after light-soaking of 800 h. The front of the glass substrate is covered by a broadband AR-coating. Reprinted with permission from [76]

Microcrystalline Silicon and Micromorph Tandem Solar Cells Up to now, results concerning the development of microcrystalline silicon solar cells deposited on CVD ZnO have been mainly reported by the IMT Neuchˆatel group, which use LP-CVD ZnO:B material as TCO. This work is done with a view of optimizing the deposition of the in-house developed micromorph tandem solar cells, described in Fig. 6.48. There is, as we have documented throughout this chapter, a great ﬂexibility in the CVD process. This ﬂexibility allows one to vary thickness and doping levels of front ZnO layers, so as to adapt their range of high transparency to the wider absorption spectrum of microcrystalline silicon. This was basically achieved by reducing the doping concentration and, thus, reducing the free carrier absorption.

296

S. Fa¨ y and A. Shah 1.2 1.0

Ico = 1.121A Vco = 150.2 V FF = 66.8 %

I [A]

0.8 0.6

P = 112.4 W (+/- 4.1 %)

0.4

ηap = 8.8 % (aperture area determined at Oerlikon Solar) substrate size 110 x 125 cm2

0.2 0.0 0

20

40

60

80

100

120

140

160

V [A] Fig. 6.55. Initial I–V characteristics of an industrial size a-Si:H p–i–n module prepared on LP-CVD front-ZnO. Reprinted with permission from [78]

At the same time, the thickness of the LP-CVD ZnO ﬁlms was increased, with the following two consequences: (a) the sheet resistance of these very lightly doped ZnO ﬁlms could be lowered to around 10 Ωsq , and (b) the surface roughness of the ZnO ﬁlms could be increased signiﬁcantly, so as to produce an improved and eﬃcient light scattering eﬀect in the longer wavelength range [34,35,56]. However, it was found that, as the Jsc of the µc-Si:H solar cells was drastically increased by using much rougher ZnO surfaces, the values of FF and Voc were also involuntarily reduced [60]. Indeed, the growth of µc-Si:H layers on very rough TCO layers is a well-known problem, which has been already investigated in several studies [80–82]. The drop in FF and Voc can thereby be attributed to an unsatisfactory growth of microcrystalline silicon layers on the sharp pyramids (with deep valleys between them) that are characteristic of LP-CVD ZnO surface morphology (see Fig. 6.6). Recently, IMT Neuchˆatel group introduced a novel post-treatment of the surface of highly rough LP-CVD ZnO:B ﬁlms to fully overcome this problem [83]. This surface treatment profoundly changes the morphology of the LP-CVD ZnO ﬁlms, transforming V-shaped valleys to U-shaped valleys that are apparently best suited for the subsequent growth of µc-Si:H solar cells. Indeed, after already 20 min of post-treatments, FF and Voc are drastically improved, as shown in Fig. 6.56. Furthermore, the photogenerated current Jsc is maintained at a high value until 40 min of post-treatment, and it is slightly reduced for longer post-treatment. This shows that this surface treatment of LP-CVD ZnO does not deteriorate the high light scattering capability of these highly rough ZnO ﬁlms. This led, together with an adequate optimization of the µcSi:H solar cells themselves, to an eﬃciency that is close to 10% (Voc = 545 mV; FF = 74.1%; Jsc = 24.7 mA cm−2 ; η = 9.9%).

6 CVD ZnO

297

Fig. 6.56. Evolution of the electrical characteristics of the solar cells as a function of the duration of the surface treatment (0 min stands for no surface treatment): (a) V oc and FF, (b) Jsc and reverse current–density (Irev ) measured at −2 V, (c) eﬃciency η. Reprinted with permission from [83]

Results obtained with individual a-Si:H and µc-Si:H solar cells are now being used to develop micromorph solar cells. Stabilized eﬃciency of 10.8% has so far been obtained for a micromorph tandem cell deposited on LP-CVD ZnO:B front TCO, with an initial eﬃciency of 12.3% (see Fig. 6.57). As neither ﬁne optimization of front ZnO surface roughness especially adapted to micromorph solar cell, nor intermediate reﬂector have been used to obtain this result, there is still a promising room for improvement in this ﬁeld.

298

S. Fa¨ y and A. Shah

Fig. 6.57. I–V characteristics (AM 1.5) of a micromorph tandem test cell on LP-CVD ZnO in the initial state and after 1, 000 h of light-soaking (1 sun at 50◦ C). The µc-Si:H bottom cell has a thickness of 2 µm. Reprinted with permission from [76]

6.4 Conclusions This chapter, which is dedicated to ZnO ﬁlms deposited by the CVD process, highlights the ﬂexibility of this process, which allows one to ﬁnely tune the deposition parameters, in order to obtain ZnO ﬁlms with the appropriate electrical and optical properties for thin ﬁlm solar cell applications. In particular, grain size can be signiﬁcantly increased either by depositing thicker ZnO ﬁlms or by decreasing the doping level. This possibility to obtain large grain sizes leads to ZnO ﬁlms that have high values of electron mobility, and can therefore attain high conductivity while keeping the disturbing free carrier absorption eﬀect limited. It also allows one to obtain light scattering features that are particularly well adapted to the requirements of thin ﬁlm silicon solar cells. The promising results obtained so far on thin ﬁlm silicon solar cells, which use CVD ZnO as TCO layers, have certainly demonstrated the high potential of this material. Finally, considering that, in general, ZnO is known to be particularly sensitive to a humid environment (see paragraph 9.4.1.2), this feature should also be checked, in more detail, for the case of ZnO ﬁlms deposited by CVD. In fact, Sang et al. [84] reported that in a humid environment, ZnO:B ﬁlms deposited by LP-CVD showed higher degradation than sputtered ZnO:Ga ﬁlms. On the other hand, Oerlikon Solar (formerly Unaxis Solar) [79] has proven that thin ﬁlm silicon solar modules using LP-CVD ZnO as TCO layers can successfully pass the standard damp-heat test, provided they are encapsulated in an appropriate manner.

6 CVD ZnO

299

Acknowledgement. The authors would like to especially thank Evelyne SauvainVallat, Ulrich Kroll, Romain Schl¨ uchter, J´erˆ ome Steinhauser and Christophe Ballif for their technical and scientiﬁc support during the study of LP-CVD ZnO:B layers at the Institute of Microtechnology (IMT) of the University of Neuchˆ atel. They gratefully acknowledge ﬁnancial support by the Swiss Federal Department of Energy and the Swiss Federal Commission for Technology and Innovation.

References 1. W. Kern, R.C. Heim, J. Electrochem. Soc. 117, 562 (1970) 2. S. Fa¨ y, S. Dubail, U. Kroll, J. Meier, Y. Ziegler, A. Shah, in Proc. of the 16th European Photovoltaic Solar Energy Conference (Glasgow, UK, 2000), pp. 361–364 3. S. Fa¨ y, U. Kroll, C. Bucher, E. Vallat-Sauvain, A. Shah, Sol. Energy Mater. & Sol. Cells 86, 385 (2005) 4. W.W. Wenas, Jpn. J. Appl. Phys. 30, L441 (1991) 5. F.T.J. Smith, Appl. Phys. Lett. 43, 1108 (1983) 6. S.K. Ghandhi, R.J. Field, J.R. Shealy, Appl. Phys. Lett. 37, 449 (1980) 7. Y. Kashiwaba, F. Katahira, K. Haga, T. Sekiguchi, H. Watanabe, J. Cryst. Growth 221, 431 (2000) 8. J.R. Shealy, B.J. Baliga, R.J. Field, S.K. Ghandhi, J. Electrochem. Soc. 128, 558 (1981) 9. A.P. Roth, D.F. Williams, J. Appl. Phys. 52, 6685 (1981) 10. J. Hu, R.G. Gordon, J. Appl. Phys. 72, 5381 (1992) 11. C.K. Lau, S.K. Tiku, K.M. Lakin, J. Electrochem. Soc. 127, 1843 (1980) 12. P.J. Wright, R.J.M. Griﬃths, B. Cockayne, J. Cryst. Growth 66, 26 (1984) 13. P. Souletie, S. Bethke, B.W. Wessels, H. Pan, J. Cryst. Growth 86, 248 (1988) 14. S. Oda, H. Tokunaga, N. Kitajima, J.I. Hanna, I. Shimizu, H. Kokado, Jpn. J. Appl. Phys. 24, 1607 (1985) 15. J. Hu, R.G. Gordon, Solar Cells 30, 437 (1991) 16. M.L. Addonizio, A. Antonaia, S. Aprea, R.D. Rosa, G. Nobile, A. Rubino, E. Terzini, in Proc. of the 2nd World Conference and Exhibition on Photovoltaic Solar Energy Conversion (Vienna, Austria, 1998), pp. 709–712 17. S. Fa¨ y, L’Oxyde de Zinc par D´epˆ ot Chimique en Phase Vapeur comme Contact Electrique Transparent et Diﬀuseur de Lumi`ere pour les Cellules Solaires. Ph.D. thesis, Ecole Polytechnique F´ed´eral de Lausanne (2003) 18. W.W. Wenas, A. Yamada, K. Takahashi, M. Yoshino, M. Konagai, J. Appl. Phys. 70, 7119 (1991) 19. C.G.V. de Walle, Phys. Rev. Lett. 85, 1012 (2000) 20. S.Y. Myong, K.S. Lim, Appl. Phys. Lett. 82, 3026 (2003) 21. K. Adachi, K. Sato, Y. Gotoh, H. Nishimura, in Proc. of the 22nd IEEE Photovoltaic Specialists Conference (Las Vegas, USA, 1991), pp. 1385–1388 22. N.D. Kumar, M.N. Kamalasanan, S. Chandra, Appl. Phys. Lett. 65, 1373 (1994) 23. A. Yamada, W.W. Wenas, M. Yoshino, M. Konagai, K. Takahashi, in Proc. of the 22nd IEEE Photovoltaic Specialists Conference (Las Vegas, USA, 1991), pp. 1236–1241

300

S. Fa¨ y and A. Shah

24. J. Hu, R.G. Gordon, Mater. Res. Soc. Symp. Proc. 202, 457 (1991) 25. K. Haga, P.S. Wijesena, H. Watanabe, Appl. Surf. Sci. 169/170, 504 (2001) 26. J.A.A. Selvan, H. Keppner, U. Kroll, J. Cuperus, A. Shah, T. Adatte, N. Randall, Mater. Res. Soc. Symp. Proc. 472, 39 (1997) 27. J. Hu, R.G. Gordon, J. Appl. Phys. 71, 880 (1992) 28. J. Hu, R.G. Gordon, Mater. Res. Soc. Symp. Proc. 242, 743 (1992) 29. J. Hu, R.G. Gordon, J. Electrochem. Soc. 139, 2014 (1992) 30. A.L. Fahrenbruch, R.H. Bube, Fundamentals of Solar Cells (Academic Press, New York, 1983) 31. K.L. Chopra, S. Major, D.K. Pandya, Thin Solid Films 102, 1 (1983) 32. J. Hu, R.G. Gordon, Mater. Res. Soc. Symp. Proc. 283, 891 (1993) 33. J. Steinhauser, S.Y. Myong, S. Fa¨ y, R. Schl¨ uchter, E. Vallat-Sauvain, A. R¨ ufenacht, A. Shah, C. Ballif, Mater. Res. Soc. Symp. Proc. 928, GG12.05 (2006) 34. S. Fa¨ y, L. Feitknecht, R. Schl¨ uchter, U. Kroll, E. Vallat-Sauvain, A. Shah, Solar Energy Materials and Solar Cells 90, 2960 (2006) 35. S. Fa¨ y, J. Steinhauser, R. Schl¨ uchter, L. Feitknecht, C. Ballif, A. Shah, in Proc. of the 15th International Photovoltaic Science and Engineering Conference (Shanghai, China, 2005), pp. 559–560 36. K. Tabuchi, W.W. Wenas, M. Yoshino, A. Yamada, M. Konagai, K. Takahashi, in Proc. of the 11th European Photovoltaic Solar Energy Conference (Montreux, Switzerland, 1992), pp. 529–532 37. K. Ellmer, J. Phys. D: Appl. Phys. 34, 3097 (2001) 38. T. Minami, MRS Bulletin 25(Aug), 38 (2000) 39. E. Burstein, Phys. Rev. 93, 632 (1954) 40. T.S. Moss, Proc. Phys. Soc. London B 76, 775 (1954) 41. A.P. Roth, J.B. Webb, D.F. Williams, Solid State Commun. 39, 1269 (1981) 42. A.P. Roth, J.B. Webb, D.F. Williams, Phys. Rev. B 25, 7836 (1982) 43. N. Mott, Metal Insulator Transitions (Barns and Noble Books, New York, 1974) 44. S.C. Jain, J.M. McGregor, D.J. Roulston, J. Appl. Phys. 68, 3747 (1990) 45. B.E. Sernelius, K.F. Berggren, Z.C. Jin, I. Hamberg, C.G. Granqvist, Phys. Rev. B 37, 10244 (1988) 46. M. Shimizu, H. Kamei, M. Tanizawa, T. Shiosaki, A. Kawabata, J. Cryst. Growth 89, 365 (1988) 47. M. Shimizu, T. Katayama, Y. Tanaka, T. Shiosaki, A. Kawabata, J. Cryst. Growth 101, 171 (1990) 48. A. Yamada, W.W. Wenas, M. Yoshino, M. Konagai, K. Takahashi, Jpn. J. Appl. Phys. 30, L1152 (1991) 49. Y.J. Kim, H.J. Kim, Mater. Lett. 21, 351 (1994) 50. T. Shiosaki, T. Yamamoto, M. Yagi, A. Kawabata, Appl. Phys. Lett. 39, 399 (1981) 51. R. Groenen, J. L¨ oﬄer, P.M. Sommeling, J.L. Linden, E.A.G. Hamers, R.E.I. Schropp, M.C.M. van de Sanden, Thin Solid Films 392, 226 (2001) 52. J. L¨ oﬄer, R.E.I. Schropp, R. Groenen, M.C.M. van de Sanden, in Proc. of the 28th IEEE Photovoltaic Specialists Conference (Anchorage, USA, 2000), pp. 892–895 53. J. L¨ oﬄer, Transparent Conductive Oxides for Thin-Film Silicon Solar Cells. Ph.D. thesis, University of Utrecht (2005) 54. D. Pier, K. Mitchell, in Proc. of the 9th European PV Solar Energy Conference (Freiburg, Germany, 1989), pp. 488–489

6 CVD ZnO

301

55. B. Sang, Y. Nagoya, K. Kushiya, O. Yamase, Sol. Energy Mater. & Solar Cells 75, 179 (2003) 56. J. Steinhauser, L. Feitknecht, S. Fa¨ y, R. Schl¨ uchter, J. Springer, A. Shah, C. Ballif, in Proc. of the 20th European Photovoltaic Solar Energy Conference (Barcelona, Spain, 2005), p. 1608 57. L.C. Olsen, H. Aguilar, F.W. Addis, W. Lei, J. Li, in Proc. of the 25th IEEE Photovoltaic Specialists Conference (Wahington D.C., USA, 1996), pp. 997–1000 58. J. Meier, S. Dubail, D. Fischer, J.A.A. Selvan, N. Pellaton-Vaucher, R. Platz, C. Hof, R. Fl¨ uckiger, U. Kroll, N. Wyrsch, P. Torres, H. Keppner, A. Shah, K.D. Ufert, in Proc. of the 13th European Photovoltaic Solar Energy Conference (Nice, France, 1995), pp. 1445–1450 59. R.G. Gordon, in NREL/SNL Photovoltaics Program Review, ed. by C.E. Witt, M. Al-Jassim, J.M. Gee (AIP Press, New York, 1997), pp. 39–48 60. L. Feitknecht, J. Steinhauser, R. Schl¨ uchter, S. Fa¨ y, D. Domin´e, E. Vallat-Sauvain, F. Meillaud, C. Ballif, A. Shah, in Proc. of the 15th International Photovoltaic Science and Engineering Conference (Shanghai, China, 2005), pp. 473–474 61. R.R. Arya, T. Lommasson, B. Fieselmann, L. Russell, L. Carr, A. Catalano, in Proc. of the 22nd IEEE Photovoltaic Specialists Conference (Las Vegas, USA, 1991), pp. 903–906 62. R.G. Dhere, K. Ramanathan, T.J. Coutts, B.M. Basol, V.K. Kapur, in Proc. of the 22nd IEEE Photovoltaic Specialists Conference (Las Vegas, USA, 1991), pp. 1077–1081 63. D. Pier. U.S. Patent 5,078,803 (1992) 64. P.S. Vijayakumar, K.A. Blaker, R.D. Wieting, B. Wong, A. Halani, C. Park. U.S. Patent 4,751,149 (1988) 65. R.D. Wieting, R.R. Potter. U.S. Patent 4,612,411 (1986) 66. R.D. Wieting, in Proc. of the 29th IEEE Photovoltaic Specialist Conference (New Orleans, USA, 2002), p. 478 67. L.C. Olsen, W. Lei, F.W. Addis, W.N. Shfarman, M.A. Contreras, K. Ramanathan, in Proc. of the 26th IEEE Photovoltaic Specialists Conference (Anaheim, USA, 1997), p. 363 68. E. Terzini, A. Antonaia, P. Thilakan, S. Aprea, I. Luck, in Proc. of the 16th European Photvoltaic Solar Energy Conference (Glasgow, UK, 2000), p. 706 69. S. Hegedus, W. Buchanan, X. Liu, R.G. Gordon, in Proc. of the 25th IEEE Photovoltaic Specialists Conference (Washington D.C., USA, 1996), pp. 1129–1132 70. W.W. Wenas, A. De, A. Yamada, M. Konagai, K. Takahashi, Sol. Energy Mater. & Solar Cells 34, 313 (1994) 71. W.W. Wenas, K. Dairiki, A. Yamada, M. Konagai, K. Takahashi, J.H. Jang, K.S. Lim, in Proc. of the 1st World Conference on Photovoltaic Energy Conversion (Waikaloa, USA, 1994), pp. 413–416 72. W.W. Wenas, M. Konagai, in Proc. of the 29th IEEE Photovoltaic Specialist Conference (New Orleans, USA, 2002), pp. 1130–1133 73. B. Sang, K. Dairiki, A. Yamada, M. Konagai, Jpn. J. Appl. Phys. 38, 4983 (1999) 74. J. Meier, U. Kroll, S. Dubail, S. Golay, S. Fa¨ y, J. Dubail, A. Shah, in Proc. of the 28th IEEE Photovoltaic Specialists Conference (Anchorage, USA, 2000), pp. 746–749

302

S. Fa¨ y and A. Shah

75. J. Meier, S. Dubail, S. Golay, U. Kroll, S. Fa¨ y, E. Vallat-Sauvain, L. Feitknecht, J. Dubail, A. Shah, Sol. Energy Mater. & Solar Cells 74, 457 (2002) 76. J. Meier, J. Spitznagel, U. Kroll, C. Bucher, S. Fa¨ y, T. Moriarty, A. Shah, in Proc. of the 3rd World Conference and Exhibition on Photovoltaic Solar Energy Conversion (Osaka, Japan, 2003), pp. 2801–2805 77. J. Bauer, H. Calwer, P. Marklstorfer, P. Milla, F.W. Schulze, K.D. Ufert, J. Non-Cryst. Solids 164-166, 685 (1993) 78. U. Kroll, in Proc. of the 21st European Photovoltaic Solar Energy Conference (Dresden, Germany, 2006), pp. 1546–1551 79. J. Meier, U. Kroll, J. Spitznagel, S. Benagli, A. H¨ ugli, T. Roschek, C. Ellert, M. Poppeller, G. Androutsopoulos, D. Borello, W. Stein, O. Kluth, M. Nagel, C. B¨ ucher, L. Feitknecht, G. B¨ uchel, J. Springer, A. B¨ uchel, in Proc. of the 20th European Photovoltaic Solar Energy Conference and Exhibition (Barcelona, Spain, 2005), p. 1503 80. J. Bailat, E. Vallat-Sauvain, L. Feitknecht, C. Droz, A. Shah, J. Non-Cryst. Solids 299-302, 1219 (2002) 81. E. Vallat-Sauvain, S. Fa¨ y, S. Dubail, J. Meier, J. Bailat, U. Kroll, A. Shah, Mater. Res. Soc. Symp. Proc. 664, A15.3.1 (2001) 82. Y. Nasuno, M. Kondo, A. Matsuda, Sol. Energy Mater. & Solar Cells 74, 497 (2002) 83. J. Bailat, D. Domin´e, R. Schl¨ uchter, J. Steinhauser, S. Fa¨ y, F. Freitas, C. Bucher, L. Feitknecht, X. Niquille, R. Tscharner, A. Shah, C. Ballif, in Proc. of the 4th World Conference on Photovoltaic Energy Conversion (Waikaloa, USA, 2006), pp. 1533–1536 84. B. Sang, K. Kushiya, D. Okumura, O. Yamase, Sol. Energy Mater. & Solar Cells 67, 237 (2001) 85. R. Groenen, J.L. Linden, H.R.M. Van Lierop, D.C. Schram, A.D. Kuypers, M.C.M. Van de Sanden, Applied Surface Science 173, 40 (2001)

7 Pulsed Laser Deposition of ZnO-Based Thin Films M. Lorenz

Pulsed laser deposition (PLD) is a growth method for thin ﬁlms by condensation of a laser plasma ablated from a single target, excited by the high-energy laser pulses far from equilibrium. First, the PLD technique is brieﬂy described beginning with the history and the fundamental processes. In the main part, the suitability of PLD as a fast and ﬂexible exploratory research technique for high-quality ZnO-based thin ﬁlm heterostructures is demonstrated by reviewing recent results. Finally, the innovative potential inherent to PLD will be demonstrated by mentioning advanced PLD techniques, including a high-pressure PLD process for free-standing ZnO-based nanowire arrays.

7.1 Brief History and Basics Pulsed laser deposition (PLD) [1–3] uses high-power laser pulses with an energy density of more than 108 W cm−2 to melt, evaporate, excite, and ionize material from a single target. This laser ablation produces a transient, highly luminous plasma plume that expands rapidly away from the target surface. The ablated material is collected on an appropriately placed substrate surface upon which it condenses and a thin ﬁlm nucleates and grows. The ﬁrst demonstration of PLD by Smith and Turner in 1965 was induced by the development of the ruby lasers [1]. The technique remained dormant for the next 20 years, and only about 100 PLD papers were published until 1986. The breakthrough of PLD as an accepted growth technique was made possible by the development of high-power lasers with suﬃciently high pulse energy and short pulse length, i.e., gas lasers with high-power thyratron switches or Q-switched solid state lasers [4]. In addition, with the discovery of the high-Tc oxide superconductors a complex oxide material of a high technological relevance was found [5, 6], which was very well suited for PLD. Concerning this, Dijkamp and Venkatesan demonstrated in 1987 the superior quality of YBa2 Cu3 O7−δ ﬁlms grown by PLD compared to those previously grown by other deposition methods [1]. The considerable research eﬀorts concentrated in the 1990s on the high-Tc superconductor thin ﬁlms pushed the devolopment of the PLD in terms of reproducibility [2, 7, 8], scaling to larger substrate areas [9,10], and deposition of heterostructures and multilayers [11]. Present day, PLD is an established growth technique for a variety of

304

M. Lorenz

materials [12]. Advantages of PLD compared to other established techniques are as follows [2, 5]: – The capability for stoichiometric transfer of multielement compounds from a single target to the substrate, i.e., the chemical composition of complex materials such as YBa2 Cu3 O7−δ can be reproduced nearly unchanged in the deposited ﬁlms. However, as will be shown later, exceptions from this general rule exist. – PLD is a reliable, versatile, and fast process. The deposition rate is in the order of tens of nm min−1 on small substrate areas of 1 cm2 . The ﬁlm thickness can be easily controlled by the number of applied laser pulses. – The laser as the external energy source for materials vaporization and the deposition chamber are spatially separated, resulting in an extremely clean process. The PLD process requires no ﬁlament or plasma gas inside the growth chamber as in contradiction thermal evaporation and sputtering do. Thus, an inert or reactive background gas can be applied during PLD growth with nearly no limitation of pressure, which can be controlled over orders of magnitude from the 10−5 mbar up to the 1 mbar range. – The synthesis of metastable materials and the formation of ﬁlms from species appearing only in the laser plasma are possible by PLD [2]. In spite of these advantages, industrial use of PLD has been slow [13] and nowadays there are only a few examples of smaller start-up companies using PLD for highly specialized applications [14–16]. Most PLD work up to now has been focused on the research ﬁeld and the reasons for that are listed here: – The volume deposition rate of PLD is only about 10−5 cm3 s−1 , that is much lower than that of other physical vapor deposition techniques as electron beam evaporation, magnetron sputtering, and vacuum arc deposition [17]. Furthermore, the energetical eﬃciency of high power lasers is only a few percent, which means that the overall eﬃciency of PLD is also low [18]. Consequently, upscaling of PLD to larger substrate areas is limited to about 5 in. diameter [9] due to the highly forward directed plasma plume [19]. Therefore, without additional lateral scanning of the substrate, a suﬃciently good thickness and composition homogeneity of the deposited ﬁlms is limited to an area of about 1 cm2 . – Depending on target density and material and on the deposition parameters, particulates and globules of molten material, the so called droplets, can be found on the deposited ﬁlms [20]. The size of the droplets is typically in the 1 µm range. The droplets are detrimental for some ﬁlm applications at the microscale, especially if lateral structuring in the micrometer range is required. For reduction or even suppression of the droplets, velocity ﬁlters, parallel oﬀ-axis conﬁgurations of plasma plume and substrate, and PLD setups with two colliding plasma plumes [21] have been used successfully. These additional precautions are based on

7 PLD of ZnO Films

305

target manipulator PLD chamber

(I) absorption of laser energy ablation of target material.

targets

(II) plasma expansion, heating partial absorption of laser radiation transfer of fast atoms, ions, clusters, and slow droplets, particles.

plasma plume UV window UV lens aperture

substrate substrate heater

(III) condensation of plasma nucleation and growth.

pulsed KrF excimer laser

Fig. 7.1. Scheme of a typical PLD setup for large-area ﬁlm growth. The main functional parts are designated on the left. The fundamental processes during (I) target ablation, (II) plasma expansion, and (III) growth are shortly described on the right, as introduction to the more detailed description in this section

the diﬀerent expansion dynamics of the small and fast atoms and clusters and of the much heavier and slower particulates and droplets in the laser plasma. However, the droplet reduction techniques often reduce the deposition rate and make the process much more complicated [1, 5]. – The fundamental processes of PLD, the laser ablation of material, the plasma creation and expansion, and the ﬁlm nucleation and growth (see Fig. 7.1), are not fully understood up to now [3, 22]. Modeling of these processes is diﬃcult because of their strong nonequilibrium character due to the high pulse energy coupled with short laser pulse lengths of typically some 10 ns. Thus, deposition of novel materials usually requires a period of empirical optimization of the PLD parameters by close interaction of growth and immediate ﬁlm characterization.

7.2 Fundamental Processes and Plasma Diagnostics The success of PLD has far surpassed the understanding of the fundamental laser ablation processes for the usually used high laser energies and short pulse lengths. PLD involves a wide range of physical phenomena and their investigation requires expensive diagnostics with nano- or picosecond time resolution [23]. However, as stated already in Sect. 7.1, for the optimization of the growth of a particular thin ﬁlm material, the detailed ablation mechanisms are of minor importance in many cases [2]. Relevant from a more practical viewpoint is that the ablation takes place on a short time scale in the nanosecond range to minimize the dissipation of the laser energy

306

M. Lorenz

beyond the volume of the melted and ablated surface layer of the target. Only within this condition, thermal destruction of the target together with phase segregation will be avoided. Furthermore, as well known as one of the major advantages of PLD and as mentioned earlier, the relative concentration of elemental species within the plasma plume corresponds to the chemical composition of the target material. Therefore, from this general consideration, PLD seems to be very well suited for the growth of granular crystalline or amorphous thin ﬁlms with complex stoichiometry [2]. To give an overview on the very diﬀerent basic eﬀects occurring in PLD, the physical processes will be brieﬂy reviewed in the following. Figure 7.1 shows schematically the main functional parts of a typical PLD setup and lists the fundamental processes during a PLD experiment. The phenomena during PLD can be divided into three main steps (I) ablation of the target material, (II) plasma expansion, and (III) condensation at the substrate surface [4, 24, 25]. Figure 7.2 shows a photograph of the inner parts of a PLD chamber suitable for large-area deposition up to 3-in. diameter as designed and built at the University Leipzig. (I) Absorption of the laser energy and ablation of the target material. According to [26], the laser ablation (also called photon induced sputtering) can be classiﬁed into the following primary and secondary mechanisms that take place simultaneously. The share of each particular process is hardly to determine.

Fig. 7.2. View into a PLD chamber built at University Leipzig with inner diameter of 405 mm, compromising a target manipulator (in front), a 3-in. diameter heater with a rotatable holder for 1 cm2 substrates (in the back). The laser entrance window into the chamber is visible left to the substrate heater. Compare Table 7.2 for technical details of this particular PLD chamber

7 PLD of ZnO Films

–

–

–

–

–

307

Thermal ablation by absorption of laser energy by the phonon system, melting, and evaporation of target material. The thermal sputtering can be described using the heat conduction equation, the thermodynamic materials parameters, the refractive indices (see [24]), and the laser parameters (wavelength, pulse duration, energy density). Beside evaporation, molten globules (droplets) can be expelled into the plasma by laser-induced recoil pressure or sub-surface superheating [20]. Such droplets can be found on the grown ﬁlms as shown below in Fig. 7.13. Exfoliational sputtering occurs when ﬂakes are detached from the target due to repeated thermal shocks. A high linear thermal expansion coeﬃcient, a high Young’s modulus, and a high melting point of the target material are necessary for exfoliational sputtering. In addition, the laser-induced temperature must approach, but not exceed the melting point to result in a shock wave induced cracking. Hydrodynamic sputtering means forming of droplets at the target as a result of transient melting and has no analog in ion sputtering. Asperities are formed at the target surface after melting. Nonthermal, photoinduced electronic sputtering is due to direct interaction of the laser photons with the electronic system of the target and involves for example photoablation, i.e., breaking of chemical bonds by high energy photons from the laser (5 eV at 248 nm wavelength). The photons can also create color centers that enhance thermal absorption of the target material [25]. Indirect collisional sputtering of the target by photon generated secondary ions and electrons from the laser plasma results in cone formation and target erosion [27]. Another secondary process is heating of the target by the laser-generated plasma.

(II) Expansion of the laser plasma including transfer of material. The laser radiation is partly absorbed by the vaporized target material and thereby a heated and excited laser-induced plasma is produced. The main absorption processes are photoabsorption and inverse Bremsstrahlung. The absorption due to inverse Bremsstrahlung dominates for long laser wavelengths, e.g., CO2 -lasers with wavelength of 10.6 µm, and is of little importance for KrF excimer lasers with wavelength of 248 nm. The reﬂection loss is less than 1% for 248 nm laser wavelength [25]. The modeling of the plasma expansion and the material transfer from the target to the substrate sensitively depends on the background gas pressure during PLD growth [25]: –

Under high vacuum conditions, i.e., pressure p < 10−2 mbar, the material transfer can be described using Monte Carlo simulations. Usually, inelastic collisions and collective phenomena as shock waves cannot be considered here. The so called Direct Simulation Monte Carlo method allows extension to slightly higher gas pressures.

308

M. Lorenz

–

– –

The hydrodynamic model is based on a suﬃciently high collision probability under thermodynamic equilibrium. This condition is fulﬁlled only at p > 1 mbar, i.e., at high pressures above the typical PLD ﬁlm deposition conditions, or at the beginning of plasma expansion, at high plasma density (small target to substrate distance). The model of an expanding shock wave according to Sedov and Taylor is valid at high pressures and large target-to-substrate distance. For the most important intermediate pressure range from 10−2 to 0.5 mbar exist only empirical friction models. In [25] an extension of the shock wave model is proposed, which is valid from 10−2 to 0.5 mbar at lower target-to-substrate distance from 20 to 50 mm.

The laser plume from a ZnO target was investigated experimentally [28, 29]. Time of ﬂight and quadrupole mass spectrometry and photoionization has been used to study the mass, charge, and kinetic energy of species ejected from a ZnO target in dependence on the energy density of a KrF laser [28]. Mostly, monatomic Zn and O atoms and ions were found. The desorption and ablation thresholds for ZnO were determined to be 0.25 and 0.7 J cm−2 , respectively. Neutral Zn atoms had energies from 1 to 4 eV, whereas Zn ions had maximum energy above 100 eV [28]. The time evolution of the laser plume of ZnO during ablation with an ArF laser in He gas was investigated in [29]. A typical luminescence spectrum of the ZnO plume in He gas at 4 mbar pressure at 30 mm distance from the target comprises several peaks ranging from about 250 up to 680 nm optical wavelength, which were assigned to neutral excited Zn atoms. (III) Condensation of material at the substrate surface, nucleation and ﬁlm growth. In contrast to the very complex processes of plasma expansion and transfer of target material to the substrate, the nucleation and growth at the substrate surface can be described by models, which also apply for other physical vapor deposition methods, such as molecular beam epitaxy (MBE), sputtering, or ion beam deposition [25]. Even the pulsed nature of PLD does not modify the well known three-dimensional island (Volmer–Weber), two-dimensional full monolayer (Frank–van der Merwe), and two-dimensional monolayer growth followed by three-dimensional islands (Stranski–Krastanov) growth modes, if the growth rate per laser pulse is less than one molecular building block, i.e., less than about 2˚ A per pulse [6]. For the growth of one atomic monolayer usually 5–10 laser pulses are required. The growth kinetics describes the nucleation processes on the atomic scale. Thermally activated processes as adsorption, desorption, and diffusion at the surface and in the volume, nucleation, and crystallization/ recrystallization determine the ﬁlm structure and can be controlled by the substrate temperature and the growth rate. Using a diagram ln(R) over 1/T, R being the deposition rate and T the growth temperature, three diﬀerent growth modes (epitaxial, polycrystalline, and amorphous) can be

7 PLD of ZnO Films

309

distinguished [30]. A similar structure zone model with the ratio of growth temperature to melting temperature and process pressure was already proposed by Thornton in 1977 for the sputtering of metallic ﬁlms [4]. In addition to thermally dominated processes for energies <0.1 eV, atoms, ions, molecules, and clusters with higher kinetic energy between 0.1 and 1,000 eV are involved in PLD growth. Four speciﬁc cases can be distinguished, depending on the kinetic energy Ekin of the arriving particles in the laser plasma. The energy of the species depends on the laser energy density at the target and the background gas pressure together with the target–substrate distance [4, 25, 30]: –

–

–

–

For Ekin ∼ = 0.01 – 1 eV, there is a thermal arrangement of particles, which dominates in MBE, laser-MBE [31], and electron beam evaporation. For Ekin ∼ = 0.2 – 50 eV, surface penetration into the ﬁrst monolayers reduces the temperature for epitaxy and yields a higher density of structures. These processes dominate in sputtering and PLD at higher background pressures of 0.05–0.5 mbar. For Ekin ∼ = 20 – 1,000 eV, local deposition of energy creates vacancies and interstitials and local heating. Deposition far from equilibrium is typical for PLD at lower background pressure. For Ekin ∼ = 100 eV – 10 keV, sputtering of the solid state ﬁlm surface probably can be element selective, and this reduces the growth rate of ion beam assisted processes and PLD.

7.3 PLD Instrumentation and Parameters for ZnO The main parts of a PLD system are the high-power pulsed laser and the deposition chamber including laser optics. For the laser, usually excimer gas lasers or frequency multiplied solid state Nd-YAG lasers are used. Excimer lasers have the advantage of much higher pulse energy in the ultraviolet spectral range, which enables higher growth rates and large-area PLD up to 4-in. diameter. Table 7.1 compares the main parameters of a high-power excimer laser for research purposes of Coherent Lambda Physik [32] with that of a pulsed Nd:YAG solid state laser by Quantel [33]. In contrast with excimer lasers with maximum pulse energy of 1,200 mJ at 248 nm wavelength [32], frequency multiplied solid state lasers reach maximum pulse energies of about 500 mJ at 355 nm wavelength and only about 150 mJ at 266 nm [33]. Consequently, the less expensive solid state lasers are suitable only for PLD of small area ﬁlms and most PLD groups use excimer lasers. The second main part of a PLD system is the vacuum chamber. The author’s group developed a large-area PLD process for the double-sided deposition of high-Tc superconducting thin ﬁlms (see for example [8, 10]). As substrate heater an arrangement of KANTHAL wire in ceramic tubes is used

310

M. Lorenz

Table 7.1. Comparison of main parameters of an excimer laser Coherent Lambda Physik LPX 305iF [32], and of a pulsed solid state Nd:YAG laser Quantel YG 981E [33], both suitable for research PLD systems

Parameter

Excimer LPX 305iF

Laser medium Basic wavelength Maximum pulse energy Harmonic wavelengths Max. pulse energy at shorter wavelength Maximum repetition rate Maximum average power Pulse duration Pulse-to-pulse stability

KrF or ArF gas 248 or 193 nm 1,200 mJ @ 248 nm – 650 mJ @ 193 nm

Beam dimensions Beam divergence Time jitter

12 × 30 mm2 (V × H) 1 × 3 mrad (V × H) ±2.5 ns

50 Hz 50 W @ 248 nm 25 ns ±3%

Pulsed Nd:YAG YG981E Nd:YAG solid state 1,064 nm 1,600 mJ @ 1,064 nm 532, 355, 266 nm 490 mJ @ 355 nm & 150 mJ @ 266 nm 10 Hz n.a. 8–11 ns ±2% (1,064 nm), ±8% (266 nm) 9 mm diameter <0.5 mrad ±0.5 ns

The initial investment cost of the Nd:YAG is less than half that of the excimer laser

(patent pending DE 10255453.7), which is suitable for double-sided deposition of double-side polished sapphire wafers. The main features of such a PLD chamber developed and built at University Leipzig are listed in Table 7.2. This PLD chamber with in situ ellipsometer ports is shown together with two other PLD chambers and the excimer laser LPX 305 in Fig. 7.3. In the last years, highly developed PLD and laser MBE chambers and systems with in situ reﬂection high-energy electron diﬀraction (RHEED, see Sect. 7.6.1) became available commercially [34–40]. Some of these developments are spin-oﬀs of PLD-based research groups, i.e., of T. Venkatesan at University of Maryland (USA) [34], H. Rogalla and D. Blank at University Twente (Netherlands) [35], and H. Koinuma and M. Kawasaki from Tokyo Institute of Technology and Tohuku University Sendai (Japan), respectively [36], and of James A. Greer [37]. Thus, these PLD systems represent the highest available standard of research experience. The typical parameters for the PLD of epitaxial ZnO-based thin ﬁlms on sapphire including information about target preparation are listed in Table 7.3. Within the range of these software controlled parameters, the properties of the deposited ﬁlms diﬀer widely, as will be shown in Sect. 7.4. Beside the parameters listed in Table 7.3, the ﬁlm properties will be inﬂuenced furthermore by a few more internal eﬀects, which will be listed and discussed in the following according to the scheme eﬀect/problem–cause–solution. Only the careful consideration of all these hidden eﬀects by experienced operators can ensure the highest quality and reproducibility of PLD grown ﬁlms.

7 PLD of ZnO Films

311

Table 7.2. Main features of a PLD chamber for large-area substrates up to 3-in. diameter Parameter Inner diameter × height Number of Cu-sealed ports Number of viton-sealed ports In situ ellipsometer ports Turbomolecular-drag-pump Base pressure Gas inlet Background gas pressure Laser focusing lens Laser entrance window Target manipulator Target-to-substrate distance Heater temperature Substrate temperature Substrate holder Substrate movement Computer control

Value 405 × 335 mm2 18 ports 2 ports 2 ports, angle of incidence 70◦ 260 l s−1 1 × 10−7 mbar 0–5,000 sccm 10−5 to 10 mbar UV fused silica uncoated, f = 300 mm UV fused silica uncoated, diameter 50×5 mm2 4 targets, max. diameter 25 mm, rotation and lateral translation 90–110 mm Max. 950◦ C Max. 800◦ C One wafer up to 3-in. diameter Rotation and lateral oﬀset Laser, gas pressure, heater, target, up to 11 consecutive steps

The chamber was designed by D. Natusch, University Leipzig (Figs. 7.2 and 7.3) and the process control software was written by H. Hochmuth

– Accuracy and reproducibility of laser pulse energy – depending on how the laser pulse energy is controlled, for example the state of laser resonator windows may inﬂuence the obtained pulse energy values – regular cleaning of laser resonator windows. – Deposition of the laser entrance window in the PLD chamber – reduced laser pulse energy – regular cleaning of the chamber entrance window after every growth run, shields between plume and window. – Ablation state of target – the deeper the ablation crater in the target is, the more the direction of plume expansion swings around the target normal – use of new and large-area targets. – Memory eﬀects of previously deposited material – contamination of deposited ﬁlm by material previously deposited in the chamber – use chamber only for one material, regular cleaning of all inner chamber parts. – Evaporation of heater material at high temperature and low pressure – contamination of the grown ﬁlm by material from the heater (see also Table 7.8) – use only high-temperature stable material for the hot parts of the heater.

312

M. Lorenz

Fig. 7.3. View into the PLD lab of the author in the Semiconductor Physics Group (head M. Grundmann) at University Leipzig, with three PLD chambers, excimer laser Lambda Physik LPX 305, and optical laser beam guiding system inside protection shielding. Two more PLD chambers are in the neighboring lab to the right

Table 7.3. Typical PLD parameters for epitaxial ZnO-based thin ﬁlms, using an excimer laser LPX 305 (see Table 7.1), and the PLD chamber described in Table 7.2 PLD parameter Background gases Background gas pressure Laser pulse energy Mask at laser exit Lens–target distance Laser focus size at target Laser energy density Target–substrate distance Substrates Substrate temperature Target material Target preparation ZnO ﬁlm growth rate Film thickness

Value O2 , N2 , N2 O 10−5 to 2 mbar 600 mJ 24 × 10 mm2 Lens focus length plus 0–40 mm 5 × 1.2 to 9 × 2.5 mm2 depending on lens position 5–1.3 J cm−2 depending on lens position 50–100 mm c-, a-, or r -plane sapphire, Si, SiC 300–750◦ C ZnO powder Alfa Aesar 99.9995% purity Pressing, sintering @ 900–1,150◦ C for 12 h in air 20–50 nm min−1 @ 10 Hz pulse repetition rate 2 nm to 2.5 µm

7 PLD of ZnO Films

313

7.4 Results on Epitaxial PLD ZnO Thin Films This section describes the potential of PLD to grow high quality epitaxial ZnO-based thin ﬁlms mainly on sapphire substrates , and in addition on silicon. Sapphire was chosen as the main substrate material because it ﬁts suﬃciently well the hexagonal structure of ZnO. The lattice mismatch of ZnO(0001) grown 30◦ in-plane rotated on sapphire(0001) is about 18%. As shown below, on c-plane and a-plane sapphire relaxed, granular ZnO thin ﬁlms with thickness of a few 100 nm and very good in-plane orientation can be grown. Furthermore, sapphire single crystals are chemically stable in air, hard and available reasonably priced also for larger areas. Substrates with zero or 0.09% [22] lattice mismatch as ZnO or ScAlMgO4 (SCAM), respectively, are a factor of 10–20 more expensive than sapphire. As will be shown in Sect. 7.5.4, ZnO thin ﬁlms with superior performance have been grown on SCAM single crystals by the Kawasaki group in Sendai. However, up to 2005, SCAM substrates were not freely disposable outside Japan. Of course, ZnO can be grown by PLD also on other substrates as for example GaAs, Si, 3C-SiC buﬀered Si, 6H-SiC. However, because of the oxidizing growth regime of the ZnO PLD, the surface of Si and SiC templates tends to oxidize and the resulting amorphous SiO2 layer prevents epitaxial ZnO growth and only c-axis textured ZnO ﬁlms without in-plane orientation can be prepared (see Sect. 7.5.4). A more extensive table of substrates used for the PLD of ZnO is given in the review paper entitled “Pulsed laser deposition of thin ﬁlms and superlattices based on ZnO” by Ohtomo et al. [22]. Beside this paper, there has recently been published a considerable number of extensive general or specialized reviews on ZnO, i.e., “A ¨ ur [41], “Epitaxcomprehensive review of ZnO materials and devices” by Ozg¨ ial growth of ZnO ﬁlms” by Tribulet [42], “P-type doping and devices based on ZnO” by Look [43], “Ferromagnetism of ZnO and GaN: A Review” by Liu [44], and a contribution to an Encyclopedia by Yao [45]. Because ZnO is a material with long-term research tradition with periodically increasing interest, the reader may compare also the older review of Hirschwald et al. [46]. In Sects. 7.4 and 7.5, we demonstrate the unique ﬂexibility of PLD in growing high-quality ZnO-based thin ﬁlms and heterostructures by presenting structural, electrical, optical, and device-related results of the Leipzig Semiconductor Physics Group headed by Marius Grundmann. More state of the art results of p-type and n-type conducting PLD ZnO ﬁlms on various substrates are reviewed in Sect. 7.5.4. An overview on advanced PLD techniques oﬀering extended possibilities in materials research is given in Sect. 7.6. Finally in Sect. 7.6.2, an innovative high-pressure PLD process for preparing ZnO-based nanostructures is demonstrated.

314

M. Lorenz

7.4.1 Structure of Nominally Undoped PLD ZnO Thin Films The structure of the PLD grown ZnO thin ﬁlms on c-plane, a-plane, and r -plane sapphire substrates will be explained by results of X-ray diﬀraction (XRD), transmission electron microscopy (TEM) with selected area diﬀraction patterns (SAD), and reﬂection high-energy electron diﬀraction (RHEED). Figure 7.4 shows XRD 2θ-ω scans measured in Bragg–Brentano geometry with Ni-ﬁltered Cu-Kα radiation on a Philips X’pert diﬀractometer with standard collimators and slits. Note the logarithmic intensity scaling in Fig. 7.4 which enhances the low-intensity background peaks as indicated. It is clearly seen that on c- and a-sapphire ZnO ﬁlms grow with their c-axis perpendicular to the substrate surface, and on r -plane sapphire with an a-axis perpendicular to the substrate surface. Therefore, on r -plane sapphire the ZnO c-axis is in-plane of the surface. More detailed high-resolution X-ray analysis of PLD grown ZnO thin ﬁlms has been done by Rahm [47]. Table 7.4 summarizes the out-of-plane and in-plane epitaxial relationships of ZnO ﬁlms and sapphire substrates, the c-axis and a-axis lattice constants of ZnO, the ZnO full peak widths at half maximum (FWHM) of 2θ-ω and ω scans, and the tilt of the ZnO structure along surface normal [47]. Because of the low intensity of the asymmetric (10¯ 14) reﬂection, the a-lattice constant has larger uncertainty compared to the c-axis lattice constant. The epitaxial relationships correspond to the results of Ohtomo (see Table 1 in [22]). The microscopic structure of the PLD ZnO thin ﬁlms and their interfaces to the sapphire or silicon substrates was investigated by TEM and HRTEM. Figure 7.5 shows a high resolution TEM cross-section of a nominally undoped ZnO thin ﬁlm grown on c-plane sapphire. The epitaxial growth of the ZnO on the c-plane sapphire substrate is obvious and the interface layer is only about two monolayers thin. Figure 7.6 shows bright ﬁeld TEM cross-section pictures of two epitaxial ZnO thin ﬁlms grown at about 600◦ C at diﬀerent oxygen partial pressure. Prior to the ZnO ﬁlm, a laterally homogeneous, about 10 nm thick and single crystalline MgO buﬀer layer was grown, which acts as diﬀusion barrier as proved in Sect. 7.4.3. Figure 7.6a nicely shows the decreasing density of dislocation lines in the ZnO ﬁlm from the interface to the ﬁlm surface due to relaxation of the lattice misﬁt. For the ZnO ﬁlms on MgO-buﬀered c-plane sapphire as shown in Fig. 7.6, the orientational relationship along growth direction is [0001]ZnO [111]MgO [0001]sapphire and 2110]sapphire [48]. in-plane [01¯ 10]ZnO [112]MgO [¯ Figure 7.7 shows TEM and HRTEM images, partly together with the corresponding SAD patterns, of PLD ZnO ﬁlms grown on Si(111) substrates [49]. Note that the lateral SAD image of the ﬁlm on Si(111) was taken from many columns as indicated in Fig. 7.7a. The ZnO columns on Si(111) (and also on Si(100), see [49]) show no preferential azimuthal in-plane orientation relating to the Si substrate. Figure 7.7b shows a HRTEM lattice image with intermediate amorphous SiOx layer of 2–4 nm thickness and an additional interdiﬀusion

7 PLD of ZnO Films

315

Fig. 7.4. XRD 2θ − ω scans of PLD grown, nominally undoped ZnO thin ﬁlms on c-plane (top), a-plane (center ), and r -plane (bottom) sapphire substrates, measured with Ni-ﬁltered Cu-Kα radiation. The ZnO ﬁlms were grown at 0.01 mbar O2 and about 650◦ C

zone of silicon suboxides with the ﬁrst ZnO monolayers. Figure 7.7c is a bright ﬁeld (110)Si TEM cross-section of ZnO on Si(111). The intermediate SiOx interface layer is clearly visible again. The existence of this layer between the substrate and the actual thin ﬁlm is most probably the reason for the very similar polycrystalline structure of the ZnO thin ﬁlms grown on Si(111) and on Si(100) substrates.

316

M. Lorenz

Table 7.4. Results of high-resolution XRD analysis of several series of epitaxial PLD ZnO thin ﬁlms grown on c-plane, a-plane, and r -plane sapphire at about 650◦ C substrate temperature, taken from [47] Parameter Epitaxial relations ZnO/sapphire Lattice constants FWHM ZnO(0002) Tilt along Surface normal

c-Sapphire ⊥ surface in-plane c (˚ A) a (˚ A) 2θ (◦ ) ω (◦ ) (◦ )

(0001)(0001) [10¯ 10][2¯ 1¯ 10] 5.204 3.244 0.012–0.027 0.042–0.128 None

a-Sapphire (0001)(11¯ 20) [11¯ 20][0001] 5.205 3.251 0.012–0.015 0.061–0.079 None

r-Sapphire (11¯ 20)(01¯ 12) [0001][0¯ 111] Not measured 3.254 Not measured Not measured 0.34

Fig. 7.5. High-resolution TEM cross section micrograph of a nominally undoped ZnO thin ﬁlm on c-plane sapphire, PLD grown at 0.05 mbar O2 and 700◦ C. The SAD pattern is taken from both ﬁlm and substrate area. Images by G. Wagner, Leipzig

Figure 7.8 shows a ZnO ﬁlm deposited on 3C-SiC buﬀered Si(111). The CVD-grown, only 3 nm thin 3C-SiC buﬀer layer does not improve the in-plane orientation of ZnO on Si. As already demonstrated in Fig. 7.7, polycrystalline ZnO without any preferential in-plane orientation was found. The HRTEM inset in Fig. 7.8 shows an amorphous layer at the ZnO/3C-SiC interface, which is formed by oxidation of the CVD grown 3C-SiC ﬁlm. Interestingly, these ZnO ﬁlms on SiC-buﬀered Si are c-axis textured and show slightly improved PL and CL characteristics compared to ZnO grown directly on silicon [50]. Figure 7.9 shows RHEED patterns obtained with 30 keV electrons impinging on clean surfaces of optimized ZnO thin ﬁlms grown on r -, a-, and c-plane sapphire. The azimuthal directions of the two types of RHEED images of the c-axis oriented ZnO ﬁlms (on a- and c-sapphire) are [1¯100] (top) and [¯2110] (bottom) [52].

7 PLD of ZnO Films

317

Fig. 7.6. TEM cross-sections of PLD ZnO thin ﬁlms grown on MgO buﬀered c-plane sapphire. (a) ZnO was grown at 0.016 mbar O2 and 600◦ C and shows decreasing density of dislocation lines from interface to surface. (b) ZnO was grown at 8 × 10−4 mbar O2 and 580◦ C. The thickness of the MgO buﬀer layer is about 10 nm. Images by W. Mader, Bonn

Fig. 7.7. TEM images and SAD patterns (insets) of a polycrystalline ZnO ﬁlm on silicon (111) PLD grown at 1 × 10−3 mbar O2 and about 540◦ C: (a) Bright ﬁeld Si(111) plane view observation, grain size is about 70 nm, (b) cross-section HRTEM lattice image with intermediate SiOx layer, and (c) weak beam Si(110) TEM cross-section. The area from which the SAD patterns were taken are within the white circles. Reprinted with permission from [49]

318

M. Lorenz

Fig. 7.8. TEM cross-section of an undoped PLD ZnO thin ﬁlm on 3C-SiC buﬀered Si(111), grown at 0.016 mbar O2 and 620◦ C. The SAD pattern (inset) was taken from the circled area. The HRTEM image (right) of the interface shows residual 3C-SiC and an amorphous interface layer. Images by G. Wagner, Leipzig

Fig. 7.9. RHEED images of optimized ZnO thin ﬁlm surfaces on r -plane, a-plane, and c-plane sapphire, in the two azimuthal orientations (top and bottom) separated by 45◦ (left) or 30◦ (middle and right), respectively. The RHEED patterns of the a-axis textured ﬁlm on r -plane sapphire (left) indicate an epitaxial and threedimensional, island-like growth. The ZnO ﬁlms on a-plane (middle) and c-plane sapphire (right) exhibit a smoother surface structure, as indicated by the streaky RHEED patterns and the observation of additional weak reﬂections in the top images due to 3 × 3 surface reconstruction [51]

7 PLD of ZnO Films

319

7.4.2 Surface Morphology of PLD ZnO Thin Films The surface morphology of the PLD grown ZnO-based ﬁlms is important for the interface quality of multilayer structures, including quantum wells with thickness of a few nanometer only, for the formation of metal–semiconductor Schottky contacts and for the optical emission properties. Therefore, the control and optimization of surface properties is essential for the successful application of ZnO thin ﬁlms in related device conﬁgurations. Figure 7.10 shows the surface roughness of ZnMgO ﬁlms deposited from a ZnMgO target with 4 wt. % MgO on a-plane sapphire substrates in dependence on the oxygen partial pressure during PLD. These investigations were done in view of optimization of ZnMgO/ZnO/ZnMgO quantum well structures (see Sect. 7.5.2), where indications for conﬁnement eﬀects were found in photoluminescence spectra [53]. Figure 7.10 shows minimum average roughness around 0.58 nm for 1 × 10−3 mbar oxygen partial pressure. Figure 7.11 shows AFM images of roughness optimized ZnO, ZnMgO (deposited from ZnO:4 wt. % MgO target on ZnO buﬀer layer), and MgO thin ﬁlms. The granular structure of all of these ﬁlms is obvious with decreasing grain size from ZnO on top to MgO on the bottom. ZnO ﬁlms optimized for superior electrical properties as for example high Hall mobility at room temperature show increased roughness as shown in Fig. 7.12. As a general observation, many granular oxide thin ﬁlm materials with optimized electrical or optical performance do not show low surface roughness simultaneously. Here, the mobility correlates with the grain size of the ﬁlms because of the limiting eﬀect of the grain boundary scattering, as

Fig. 7.10. Maximum, root mean square, and average roughness Rmax , Rq , and Ra , respectively (derived from 10 × 10 µm2 AFM scans) of 300 nm thick ZnMgO ﬁlms (PLD from a ZnO:4 wt. % MgO target) on a-plane sapphire in dependence on oxygen partial pressure during PLD growth. Lowest ﬁlm surface roughness is obtained around 10−3 mbar. The lines are drawn to guide the eye. Measured by G. Zimmermann

320

M. Lorenz

10nm 0 10

ZnO

10 0

10nm 0 10

ZnMgO 3nm 0

10 0

10

MgO

10 0

Fig. 7.11. AFM scan images (10×10 µm2 ) of roughness optimized ZnO (top, 1.4 µm thick), ZnO:4 wt. % MgO (center, 250 nm thick on ZnO buﬀer), and MgO (bottom, 400 nm thick) ﬁlms on a-plane sapphire with average roughness Ra of 1.45, 0.41, and 1.23 nm, respectively. Measured by G. Zimmermann

shown already in [51]. The hexagonal wurtzite structure of the ZnO grains is clearly visible in Fig. 7.12. The surface morphology of ZnO ﬁlms on sapphire, 3C-SiC/Si, and 6H-SiC substrates was also investigated by scanning electron microscopy (SEM) as

7 PLD of ZnO Films

321

150nm 0 10

10 0

ZnO µH = 97cm2/Vs

200nm 0 10

10 0

ZnO µH = 148cm2/Vs

Fig. 7.12. AFM scan images (10 × 10 µm2 ) of mobility optimized, 1.4 µm thick ZnO thin ﬁlms on a-plane sapphire. The Hall mobility at 300 K correlates with the grain size. The average roughness and the carrier concentration are 12.5 mm/1.6 × 1016 cm−3 (top) and 8.38 nm/3.1 × 1016 cm−3 (bottom), respectively. Measured by G. Zimmermann

shown in Fig. 7.13. The surface of an epitaxial ZnO ﬁlm on a-plane sapphire appears very smooth without any structure, except the single larger particle visible in Fig. 7.13a. Figure 7.13b shows a textured ZnO ﬁlm grown on 3C-SiC (3 µm)/Si(100) without visible crack formation, however with larger hill-like structure with diameter of several micrometer in the center and many surface crystallites in the 100 nm range. Because of the increased lattice deformation of ZnO on single crystalline 6H-SiC substrates (see [54]), microcracks are obvious at the ZnO surface in Fig. 7.13c,d. At higher substrate temperature, some droplet-like particles are collected at these cracks (Fig. 7.13d).

322

M. Lorenz

ZnO / 3C-SiC-Si(100)

ZnO / a-sapphire

(a)

1µm

ZnO / 6H-SiC

(c)

(b)

1µm

ZnO / 6H-SiC

1µm

(d)

1µm

Fig. 7.13. Typical SEM images of PLD ZnO ﬁlm surfaces: (a) ZnO (400 nm) on a-plane sapphire grown at 700◦ C, (b) ZnO (600 nm) on 3C-SiC (3 µm)/Si(100) grown at 640◦ C, (c) ZnO (300 nm) on 6H-SiC grown at 580◦ C, (d) ZnO (300 nm) on 6H-SiC grown at 700◦ C [54]

7.4.3 Electrical Properties of PLD ZnO Thin Films: Eﬀect of Buﬀer Layers The PLD grown ZnO-based ﬁlms on sapphire can be arranged as follows according to their growth- and doping-dependent electrical performance [55], for references giving details about the particular PLD ZnO ﬁlms see Table 7.9: (a) Intrinsically n-type doped ZnO ﬁlms deposited under various oxygen partial pressures show typical free electron concentrations from 1016 to 1018 cm−3 . Under particular deposition conditions, a low-temperature grown, thin ZnO nucleation layer may improve the electron mobility of the ZnO ﬁlms on c- and a-plane sapphire. (b) Unintentionally doped ZnO ﬁlms prepared at high growth temperatures exhibit Al indiﬀusion from the sapphire (Al2 O3 ) substrate, resulting in a background electron concentration of about 1016 cm−3 . (c) Clean ZnO ﬁlms with a CeO2 or MgO diﬀusion barrier layer to supress the Al diﬀusion from the substrate. Semiinsulating behavior with a low electron concentration ≤1014 cm−3 was obtained. In addition, also ZnMgO or ZnMnO ﬁlms with more than 5 at. % Mg or more than 2 at. % Mn, respectively, show semi-insulating behavior. (d) Intentionally doped ZnO layers with Ga or Al concentration up to a few at. % show electron concentrations above 1020 cm−3 , which is suitable for transparent conducting oxide (TCO) layers. (e) Attempts for p-type conducting ﬁlms by deposition in N2 O background gas, by codoping with N2 O and Ga, by doping with Li and N, and by group-V elements as Sb or P. A high Ga concentration up to 1019 cm−3

7 PLD of ZnO Films

323

could be compensated by codoping with N2 O, resulting in a semiinsulating behavior ≤1014 cm−3 , thus proving the acceptor-like incorporation of N in ZnO. Very low conductivity at the detection limit was obtained also for Li and P-doped ﬁlms, partially enhanced by post deposition annealing, but up to now no clear indication for p-type conductivity was obtained. Figure 7.14 depicts the ranges of resistivity vs. n-type carrier concentration of undoped and doped PLD ZnO thin ﬁlms on sapphire [56]. A six orders of magnitude range of carrier concentration and resistivity (2×1014 to 2 × 1020 cm−3 and 10−3 to 103 Ω cm, respectively) could be controlled with these epitaxial ﬁlm series deposited with variable PLD oxygen partial pressure. A nearly linear dependence of resistivity on carrier concentration was found. In the deposition chamber with the smaller 50 mm distance from PLD target to the substrate, we found a strong dependence of carrier concentration on oxygen partial pressure adjusted from 5 × 10−4 up to 0.04 mbar (see [56]). The Hall mobility at room temperature of nominally undoped ZnO ﬁlms on c-plane sapphire grown by a multistep PLD process with low-temperature nucleation layers [51] is shown in Fig. 7.15. The highest mobility values are obtained only in a narrow carrier concentration range of (2–5) × 1016 cm−3 . A much weaker dependence of carrier concentration and Hall mobility on the oxygen partial pressure was found for PLD growth at 100 mm target to substrate distance.

Fig. 7.14. Resistivity vs. carrier concentration at 300 K of undoped and doped PLD ZnO thin ﬁlms on sapphire [56]. The legend gives the maximum concentrations of the dopant oxides in the ZnO targets in weight-%. The optimized ZnO ﬁlms show lower intrinsic n-type conductivity. For each dopant element, series of ﬁlms grown at diﬀerent PLD oxygen pressures are included

324

M. Lorenz

Fig. 7.15. Peak behavior of Hall mobility at 300 K of undoped ZnO thin ﬁlms on cplane sapphire at carrier concentrations around 3 × 1016 cm−3 . For these particular growth conditions with a target-to-substrate distance of 50 mm, a low-temperature nucleation layer was used. Reprinted with permission from [51]

Fig. 7.16. Growth temperature-dependent reduction of carrier concentration (300 K) of ZnO ﬁlms on sapphire due to introduction of CeO2 buﬀer layers. The growth temperature is about 50◦ C lower as the given heater temperature. Hall measurements by H. von Wenckstern

For that, high Hall mobility values well above 100 up to 150 cm2 (V s)−1 were obtained reproducibly for carrier concentrations from about 6 × 1015 up to 3 × 1016 cm−3 . Buﬀer layers deposited on the sapphire substrates prior to ZnO deposition reduce the Al diﬀusion and thereby the n-type carrier concentration, as shown in Fig. 7.16. The use of CeO2 buﬀer layers is well established from the deposition of high-Tc superconducting thin ﬁlms on sapphire [7, 8, 10]. However, the eﬀect of the CeO2 buﬀer layer decreases with increasing growth temperature due to the limited thermal stability of the 10 nm thin layer as shown in Fig. 7.16. To give a clear evidence for the reduced interdiﬀusion, Fig. 7.17 shows the isotope intensity depth proﬁles of ZnO ﬁlms grown

7 PLD of ZnO Films

325

Fig. 7.17. SNMS isotope intensity sputter depth proﬁles of ZnO ﬁlms grown at 650◦ C without (top) and with (bottom) CeO2 buﬀer layer. The buﬀer layer with a thickness of about 50 nm reduces the interdiﬀusion of Al and Zn into ZnO and sapphire, respectively, and thereby the concentration of n-type carriers in the ZnO ﬁlm

without and with CeO2 buﬀer layer. The reduction of the Al and Zn concentrations within the ZnO ﬁlm and the Al2 O3 substrate, respectively, is obvious. Instead of CeO2 , MgO buﬀer layers show slightly improved barrier properties, so that mainly MgO buﬀer layers are used [57] to grow semi-insulating ZnO ﬁlms for conversion experiments into the p-type conducting state by acceptor doping [58]. A deeper insight into the lateral electrical homogeneity of the ﬁlms, the limiting mechanisms of the Hall mobility, and the thermal activation energies of shallow and deep defect levels can be gained by temperature-dependent Hall and deep level transient spectroscopy (DLTS) measurements [57,59,60]. To give an example, the temperature dependence of the Hall mobility and

326

M. Lorenz

of the carrier concentration of a series of optimized ZnO ﬁlms grown at different oxygen partial pressures is shown in Fig. 7.18. The ZnO ﬁlm thickness is around 1 µm. The temperature dependence of the mobility is similar for the ﬁlms grown at oxygen partial pressures from 3 × 10−3 to 3 × 10−2 mbar, whereas the ﬁlm grown at higher pressure of 10−1 mbar shows a lower mobility over the whole temperature range, a shifted maximum, and a decreasing temperature dependence for low temperatures. The corresponding carrier concentration in dependence on the inverse temperature is shown in Fig. 7.18. The three ﬁlms grown at lower PLD oxygen pressure show semiconductor-like decrease in carrier concentration with decrease in temperature. The cutoﬀ of the carrier concentrations with decrease in temperature is due to the detection limit of the Hall voltage in the 1014 cm−3 range. Therefore, the mobility

Fig. 7.18. Temperature dependence of the Hall mobility (top) and of the carrier concentration (bottom) of undoped PLD ZnO thin ﬁlms on a-plane sapphire grown at diﬀerent oxygen partial pressures (see legends). Note the diﬀerent temperature scales. The ﬁlm grown at highest pressure shows an unusual metal-like temperature dependence of the carrier concentration for T < 90 K. By H. von Wenckstern [59]

7 PLD of ZnO Films

327

Table 7.5. Energetic positions below the conduction band edge (Ec ) and densities of shallow (Hi , AlZn ) and deep (E1–E5) donor-like defect levels (traps) in ZnO identiﬁed by temperature-dependent Hall eﬀect and deep level transient spectroscopy, respectively, in undoped PLD ﬁlms and single crystals grown by seeded chemical vapor deposition (Eagle Picher), taken from H. von Wenckstern [57] PLD thin ﬁlm Defect Hi AlZn E1 E3 E4 E5

Ec − Et (meV) 65 ± 2 110 ± 20 290 ± 30

Single crystal −3

Nt (cm

)

(6.0 ± 1.0) × 10 (1.4 ± 0.2) × 1015 (6.2 ± 0.7) × 1015 16

Ec − Et (meV)

Nt (cm−3 )

37 ± 2

(1.3 ± 0.2) × 1017

100 ± 20 300 ± 30 540 ± 40 840 ± 50

(1.4 ± 0.2) × 1015 (2.2 ± 0.4) × 1014 (1.8 ± 0.4) × 1014 (4.4 ± 0.5) × 1014

and carrier concentrations can be given for T < 100 K for these ﬁlms only. In diﬀerence, the ﬁlm grown at 0.1 mbar could be measured down to about 50 K. The carrier concentration decreases down to about 100 K and than shows a metal-like increasing behavior at lower temperature. The origin of this behavior is not completely clear up to now. The Hall data could be modeled with the assumption of a very thin, highly n-type conductive degenerate layer near to the interface of the substrate [59]. By ﬁtting the temperaturedependent mobility data of PLD ZnO ﬁlms, the grain boundary scattering of the carriers was identiﬁed as the limiting mechanism in PLD ZnO thin ﬁlms on sapphire, thus opening potential for further ﬁlm optimization [61]. In addition to temperature-dependent Hall measurements, deep level transient spectroscopy (DLTS) was performed using high-quality Schottky contacts [60] (see also Sect. 7.5.3 with Fig. 7.31) to obtain thermal activation energies and densities of the electrically active defects in PLD ZnO thin ﬁlms and ZnO single crystals. To summarize these results, the obtained energetic positions and the densities of shallow and deep donor-like defect levels in a PLD ZnO thin ﬁlm and a ZnO single crystal from Eagle Picher are given in Table 7.5 [57, 61]. The shallow defect levels in ZnO can be assigned to the donors Hi at interstitial position and AlZn at Zn-position (compare [62]). However, the chemical identity of the deep donor levels is still under discussion; therefore, they are usually assigned as E1 to E5. 7.4.4 Luminescence of PLD ZnO Thin Films on Sapphire Photo- and cathodoluminescence (PL, CL) measurements at liquid helium temperature are sensitive tools for investigation of the optical recombination properties [62] of the PLD ZnO ﬁlms. In ZnO, the information depth of PL using a 325 nm He–Cd laser is only about 60 nm [63], corresponding to

328

M. Lorenz

Fig. 7.19. Photoluminescence spectra (2 K) of PLD ZnO thin ﬁlms on a-plane, c-plane, and r -plane sapphire substrates [63]. All ﬁlms were grown at about 650◦ C and at 1.6 × 10−2 mbar oxygen pressure. The FWHM of the most intense donor bound exciton peaks D0 X of the ZnO ﬁlms are 1.4 meV on a-plane sapphire, 1.7 meV on c-plane sapphire, and 2.6 meV on r -plane sapphire. The spectral resolution of the PL setup was 1 meV at 3.35 eV

the topmost surface layer even in PLD thin ﬁlms. CL has higher information depth, being about 250 nm for 5 keV acceleration voltage of the electron beam, and about 750 nm for 10 keV electron energy [54]. The diameter of the laser beam in PL, i.e., the lateral resolution, is usually a few 100 µm, whereas the electron beam in CL can be focused down to about 20 nm diameter at the surface, thus allowing, for example, CL line scans on cross-sections of micrometer thin ﬁlms, and CL intensity and CL wavelength scans at the 10 × 10 µm2 area range. Figure 7.19 shows PL spectra recorded at 2 K for 2.2, 0.7, and 1.5 µm thick PLD ZnO ﬁlms on a-plane, c-plane, and r -plane sapphire, respectively [63]. The full widths at half maximum (FWHM) of the most intense bound exciton peaks are 1.4, 1.7, and 2.6 meV for a-, c-, and r -sapphire, respectively. The ﬁlm on a-plane sapphire shows the narrowest FWHM among the ﬁlms under investigation and the free A-exciton (XA ) is most clearly resolved, thus indicating best structural properties of ZnO on a-plane sapphire. The ZnO ﬁlms on a- and c-plane sapphire grow c-axis textured, whereas ﬁlms on r -plane sapphire grow a-axis oriented with the ZnO c-axis being in-plane, as demonstrated already in Fig. 7.4. The PL spectrum of the ﬁlm on r -plane sapphire shows no phonon replica, probably due to the changed ZnO orientation.

7 PLD of ZnO Films

329

Fig. 7.20. Comparison of PL spectra at 2 K of a 2.2 µm thick ZnO ﬁlm on a-plane sapphire and of a ZnO bulk single crystal grown by seeded chemical vapor deposition (Eagle Picher), both (0001) oriented. The PLD ﬁlm was deposited by a 4-step PLD process [51] and shows a PL spectrum very similar to that of the single crystal. The energies of the assigned luminescence peaks are given in [eV]. The spectral resolution of the PL setup is 1 meV at 3.35 eV [63]

Figure 7.20 compares the PL spectra of the ZnO ﬁlm on a-plane sapphire with that of a ZnO bulk single crystal from Eagle Picher. General features, as the most intense lines of free and bound excitons, and the phonon replica , appear in both spectra and demonstrate the single crystalline-like structural quality of optimized PLD ZnO thin ﬁlms in the near-surface region. The most intense donor bound exciton peak of the ZnO ﬁlms on sapphire (Fig. 7.20) is the I6 peak, which can be attributed according to [62,64] to the Al donor. The FWHM of the donor bound exciton peaks I8..6 and I4 of the single crystal are less than 1 meV, compared to 1.4 meV for I6 of the ﬁlm. From the excitonic PL peak energies XA (n = 1), XA (n = 2) and the ﬁrst and second phonon replica of D0 X shown in Fig. 7.20, the band gap energies, the exciton binding energies, and the LO-phonon energies were calculated using the eﬀective mass model at k = 0 [63]. Table 7.6 summarizes these results for the ZnO single crystal and the PLD thin ﬁlm on a-plane sapphire. Within the error limits, the values for the single crystal, the PLD thin ﬁlm, and published values [62] are in good agreement, thus proving again the single crystal-like structural and optical properties of the near-surface region of the PLD thin ﬁlms.

330

M. Lorenz

CL intensity [1000 cps]

50 ZnO surface

40

30

20

interface D0X (I6-I9) T = 9K e-beam 10kV, 50pA

10

0 0

1 2 distance from ZnO interface [µm]

Fig. 7.21. CL line scan at 9 K of the intensity of the bound exciton transitions I6 to I9 measured at the cross-section of a 2.2 µm thick ZnO ﬁlm on a-plane sapphire from the sapphire substrate to the ZnO ﬁlm surface. The inset shows the corresponding SEM image of the cross section. Measured by J. Lenzner Table 7.6. Comparison of band gap energies Eg , exciton binding energies EXB , and 1LO and 2LO phonon energies ω1LO and ω2LO , calculated from the PL peak energies of a ZnO single crystal (Eagle Picher) and a PLD ZnO thin ﬁlm on a-plane sapphire at 2 K, as given in Fig. 7.20 [63]

Single crystal Eagle Picher PLD ZnO ﬁlm on a-sapphire

Eg (eV)

EXB (meV)

ω1LO (meV)

ω2LO (meV)

3.4381 ± 0.0015

61 ± 1

72.6 ± 1

73.4 ± 1

3.4364 ± 0.0015

60 ± 1

71.2 ± 1

72.9 ± 1

Thinner ZnO ﬁlms on sapphire, i.e., with thickness ≤0.5 µm, show often less intense PL spectra with broader excitonic peaks, which are due to the increasing density of dislocation lines near the interface, as shown in the TEM cross-section Fig. 7.6. To further investigate this phenomenon, a cross-section of a ZnO ﬁlm was prepared by simply breaking a 2.2 µm thick ﬁlm with the substrate and measuring a CL line scan of the excitonic peak intensity from the interface sapphire/ZnO as start point up to the surface of the ZnO ﬁlm, as shown in Fig. 7.21 together with the SEM view on the cross-section. The CL intensity increases from nearly zero at the sapphire/ZnO interface up to the maximum near the surface of the ZnO ﬁlm, thus supporting again the correlation of low defect density and high luminescence intensity. In summary, near the surface of PLD ZnO thin ﬁlms on a- and c-plane sapphire, structural quality and optical luminescence properties are very similar to those of ZnO bulk single crystals.

7 PLD of ZnO Films

331

7.4.5 Chemical Composition of Doped PLD ZnO Films and Doping Eﬀects The ﬂexible doping and alloying of ZnO thin ﬁlms is probably the most prominent advantage of PLD in comparison to other deposition techniques. Therefore, a variety of dopants has been investigated in PLD ZnO ﬁlms, as will be shown in the following. The addition of the dopant is done by mixing the dopant element in form of the appropriate oxide into the ZnO powder used for target preparation. By ball milling the powder mixture, a homogenous distribution of the dopant will be achieved. After pressing the powder, the target will be sintered at 900–1,150◦C in air for 12 h. Concerning the chemical composition of the thin ﬁlms grown from the multielement targets in the PLD process, usually a reproduction of the target composition in the thin ﬁlm is expected, as explained already in Sect. 7.1. However, depending on the properties of the deposited atomic species and the substrate temperature, this rule is not generally valid for all materials. Unfortunately, many of the doped and alloyed ZnO thin ﬁlms important for future optoelectronic applications show deviations of their chemical composition compared to that of the PLD target. The most pronounced examples ZnO:Li, MgZnO, and CdZnO of composition deviations are demonstrated in dependence on the oxygen partial pressure in Figs. 7.22 and 7.23 in contrast to ZnO:Mn, which shows a nearly 1:1 transfer. Averaged over the oxygen partial pressure, the dopant composition, and other deposition parameters, Table 7.7 summarizes the average composition transfer factors from the PLD target into the grown ﬁlm and calls the dopant oxide used for target preparation. The ﬁlm compositions were analyzed by ion beam analysis using combined Rutherford backscattering spectrometry (RBS) and particle induced X-ray emission (PIXE) with 1.2 MeV He+ and H+ ions, respectively [65]. For most dopant elements a transfer factor above 1 was determined (Table 7.7). That means that the chemical composition of the dopant element in relation to Zn is increased by the PLD process. However, some dopant elements, as P, Ni, and Cu, and Cd show transfer factors considerably below 1. To understand this diﬀerent behavior of the dopant elements in the PLD process (for the PLD parameters see Table 7.3), in Table 7.7 the evaporation temperature of Zn and of the dopant elements at normal conditions is given. The element with the lower evaporation temperature may evaporate preferably from the heated substrate, thus yielding a lower composition of this particular element in the ﬁlm. For example, P and Cd show lower evaporation temperatures as Zn, thus explaining the low transfer factors of these elements into the ZnObased ﬁlms. In the case of Cd, in addition a low solubility of CdO in ZnO of a few at. % is reported (see [56, 72]), thus explaining the extremely low transfer factor of Cd. However, the correlation of low transfer factor and low evaporation temperature does not apply for Ni and Cu as neighbor elements to Zn in the periodic system. In contrast, Mg has the highest transfer factor of all dopant elements. Probably, the preferential ablation of ZnO out of the

332

M. Lorenz

Fig. 7.22. Transfer of the chemical composition of Mn (top) and Li (bottom) from the PLD ZnO:Mn and ZnO:Li targets, respectively, into the thin ﬁlms, for diﬀerent target compositions and PLD gas pressures [65]. Further variation of concentrations is due to diﬀerent growth temperature and other PLD parameters. Li shows the largest scattering among all investigated dopant elements (see Table 7.7 and Fig. 7.23) due to the properties of the small Li atoms. Reprinted with permission from [65]

MgZnO target is an additional reason for the Mg-enriched ﬁlms because of the low optical absorption coeﬃcient of MgO with Eg around 8 eV for the excimer laser light at 248 nm wavelength. The observed continuous shift of the MgZnO ﬁlm composition with the ablation state of the MgZnO target supports this assumption of the preferential target ablation. Because the current research eﬀorts on the II–VI semiconductor ZnO are directed for example to blue and UV-optoelectronic devices, the cleanness requirements of the established semiconductor processes should be applied also for ZnO, if possible. To get a ﬁrst estimate about the trace element contamination of the PLD ZnO ﬁlms, we analyzed two undoped ZnO thin ﬁlms by PIXE/RBS for their trace element concentration [65] as given in Table 7.8. For most elements listed in Table 7.8, the detected trace element

7 PLD of ZnO Films

333

Fig. 7.23. Transfer of the chemical composition of Mg (top) and Cd (bottom) from the PLD MgZnO and CdZnO targets, respectively, into the thin ﬁlms, for diﬀerent compositions and PLD gas pressures [65]. Further variation of concentrations is due to diﬀerent growth temperature and other PLD parameters. Reprinted with permission from [65]

concentrations are near the detection limits of the ion beam analysis, which is around 10 ppm. At the edges of the 32.8 mm diameter wafers remarkable concentrations of the elements K, Ca, Ti, Cr, and especially Fe were found. Most of these elements are constituents of stainless steel and the stainless steel substrate holder and the hot parts of the substrate heater in the PLD chamber are most probably the source of the Fe and Cr contamination of the ﬁlm edges. In the center of both investigated ﬁlms, the impurity concentrations of nearly all elements are near the detection limit of the ion beam analysis, thus demonstrating the high potential of PLD to deposit contamination free ﬁlms. To provide an extended overview on the various PLD-based activities of the Leipzig Semiconductor Physics Group, Table 7.9 summarizes the published investigations on ZnO-based thin ﬁlms. The dopants were arranged in four groups, namely (1) the n-type donors Al and Ga, (2) the potential

334

M. Lorenz

Table 7.7. Transfer factor of dopant composition (in ZnO matrix) from the PLD target into the deposited thin ﬁlm as determined by RBS/PIXE by D. Spemann [65], averaged over the dopant composition and the oxygen partial pressure, compare also Figs. 7.22 and 7.23 dopant element x to ZnO

Evaporation temperature of the dopant element (◦ C)

Oxide of Dopant element in target

Undoped Li Mg Al P Ti Mn Fe Co Ni Cu Ga Cd Sb

Matrix:ZnO Li3 N MgO Al2 O3 P2 O5 TiO2 MnO2 , MnO Fe2 O3 , FeO CoO NiO CuO, Cu2 O Ga2 O3 CdO Sb2 O3

Transfer factor of dopant composition from PLD target into the ﬁlm cx (ﬁlm)/cx (target)

907 (Zn) 1,342 1,090 2,519 280 3,287 2,061 2,861 2,927 2,913 2,562 2,204 767 1,587

1.0 (Zn) 1.37 ± 0.72 1.86 ± 0.49 1.56 ± 0.44 0.50 ± 0.01 1.15 ± 0.43 1.03 ± 0.28 1.47 ± 0.16 1.23 ± 0.09 0.15 ± 0.02 0.74 ± 0.25 1.54 ± 1.04 0.09 ± 0.12 1.71 ± 0.32

The dopant oxide is used for target preparation and the evaporation temperatures of the dopant elements partially explain the deviations of the transfer factor from 1.0 Table 7.8. Typical atomic concentrations of trace elements (all values given in ppm) in large-area diameter ZnO thin ﬁlms on sapphire grown in a stainless steel PLD chamber with stainless steel substrate holder and KANTHAL-wire in Al2 O3 ceramic tubes as heater element at about 650◦ C Sample

Si

K

Ca

Ti

V

Cr

Mn

Fe

Ni

Cu

E391c E391e

1,342 <901

<16 30

24 28

<11 <13

<14 <16

<15 28

20 <19

35 1,069

<67 <69

<153 <166

E392c E392e

<971 <1,272

<30 54

<24 51

<21 27

<24 <29

50 85

<26 41

<33 206

<104 <103

<215 <243

901

16

12

11

14

15

17

20

67

153

DL

The PLD target was sintered from 99.9995 at. % ZnO powder. DL is the minimum detection limit of the combined PIXE/RBS analysis with 1.2 MeV protons. Each wafer (32.8 mm diameter) was analyzed at center (c) and edge (e) position. Measured by D. Spemann

p-type acceptors N, Li, P, and Sb, (3) the isoelectronic alloys ZnO:MgO and ZnO:CdO for tuning the electronic bandgap, and ﬁnally (4) the 3d-elements Sc, Ti, Mn, Fe, Co, Ni, Cu to obtain diluted magnetic semiconductors (DMS) with both ferromagnetic and semiconducting properties to employ the spin state for information transfer or storage in innovative spinelectronic devices.

7 PLD of ZnO Films

335

Table 7.9. Overview about published research results of the Leipzig Semiconductor Physics Group on ZnO-based PLD thin ﬁlms Dopant

Investigated eﬀect of doping

Al, Ga

n-type carrier concentration in dependence on PLD oxygen partial pressure Al-doped ZnO thin ﬁlm as ohmic back contact of ZnO ﬁlms with Pd Schottky contact on top, improved frequency response enables capacitance spectroscopy DLTS

Al

Mg

Mg

Mg Mg, Cd Cd

Infrared dielectric functions and phonon modes of wurtzite MgZnO, and Mg-rich cubic MgZnO ﬁlms by spectroscopic IR ellipsometry 380–1,200 cm−1 , 360–1,500 cm−1 , respectively. (See also Sects. 3.3 and 3.4.3) Dielectric functions, refractive indices, and band gap energies of wurtzite MgZnO, and Mg-rich rocksalt MgZnO ﬁlms by UV–vis ellipsometry 1–5 eV (See also Sects. 3.6.2 and 3.7.2) Quantum conﬁnement PL of MgZnO/ZnO hetero- and double-heterostructures grown by PLD blueshift, redshift of optical absorption edge in dependence on Mg, Cd content Incorporation of Cd by MOCVD (max. 8 at. %), XRD, PL in dependence of Cd content, comparison with Cd incorporation in sequential PLD of ZnO/CdO multilayers

Fe, Sb, Al, Phonon modes by Raman spectroscopy, comparison with Ga, Li published modes in ZnO:N (See also Sect. 3.4.2) Li, N Attempts for p-type conducting ZnO ﬁlms, photoluminescence N, P, Sb Electrical activity of group V acceptors in ZnO P Strong indication for p-type conducting surface areas on P-doped ZnO thin ﬁlms by scanning capacitance microscopy SCM Mn

Mn Co, Mn, Ti Ti, Co, Mn

Co, Al Gd, Nd

Magnetic hysteresis (SQUID), magnetic domains (MFM), luminescence (PL), PLD parameters of ferromagnetic ZnO:Mn ﬁlms, EPR of ZnO:Mn ﬁlms, valence state, hyperﬁne structure, ﬁne structure parameter Thermal activation energy and density of deep donor-like levels in 3d-doped ZnO ﬁlms, Schottky contacts (DLTS) Magnetoresistance in dependence of dopant composition from 0.02 up to 2 at. %, temperature and magnetic ﬁeld up to 6 T, for ZnO:Ti magnetic domains (MFM) Magnetoresistance in dependence of ﬁlm thickness, temperature, and magnetic ﬁeld up to 6 T Deep levels by DLTS, magnetoresistance, anomalous Hall eﬀect, magnetic domains for ZnO:0.1 at. % Nd by MFM

Refs. [56]

[57, 59]

[66–68]

[69–71]

[53] [56]

[72, 73] [74] [75] [76] [58]

[77, 78]

[79, 80] [81, 82] [78, 83] [84] [85]

336

M. Lorenz Table 7.9. (continued)

Dopant

Investigated eﬀect of doping

Refs.

ZnO/ BaTiO3 / ZnO

Electro-optical eﬀects of BTO, ZnO/BTO, and ZnO/BTO/ZnO heterostructures with ﬁxed/switchable polarization, IV and CV, ferroelectric hysteresis

[86–88]

The results are arranged according to the aim of doping or alloying, namely (1) n-type conductivity, (2) bandgap-tuning and quantum conﬁnement, (3) potential p-type conductivity, and (4) ferromagnetic behavior. In addition, (5) ZnO/BaTiO3 heterostructures as combination of ZnO with ﬁxed and BaTiO3 with electrically switchable polarization are included

Additionally, (5) ZnO/BaTiO3 heterostructures as combination of ZnO with ﬁxed polarization and BaTiO3 with electrically switchable polarization are included in Table 7.9. Highlights of the Leipzig research on doped and alloyed PLD ZnO ﬁlms include the measurement of the dielectric function of MgZnO ﬁlms over the whole mixing range from pure ZnO to pure MgO in the optical IR and in the vis–UV spectral range using spectroscopic ellipsometry, the observation of quantum conﬁnement eﬀects in MgZnO/ZnO/MgZnO heterostructures [53], and the successful observation of p-type conducting surface areas in P-doped ZnO ﬁlms grown by PLD [58]. With respect to ferromagnetism in ZnO ﬁlms doped with various 3d- and 4f-elements, the ﬂexibility and the high potential of PLD is demonstrated in Fig. 7.24, where stripe-like or bubble-like magnetic domains on the surface of Mn-, Co+Al-, and Ti-, and Nd-doped ZnO thin ﬁlms were imaged by magnetic force microscopy (MFM). As a direct proof of the magnetic eﬀect, the corresponding atomic force microscopy (AFM) image of the same surface area of each particular ﬁlm is given in Fig. 7.24. Finally, the investigation of shallow and deep defect levels in Mn-, Co-, and Ti-doped ZnO thin ﬁlms [77, 81, 83] should be mentioned, which became possible by the introduction of an advanced Schottky contact conﬁguration for the doped PLD ZnO thin ﬁlms using a degenerately doped ZnO:Al ﬁlm as ohmic back contact [57, 59]. As a reference, Table 7.5 in Sect. 7.4.3 compares the shallow and deep defect levels in undoped ZnO ﬁlms and a ZnO single crystal as determined by temperature-dependent Hall eﬀect and DLTS. Only the innovative front-to-back Schottky contact conﬁguration provides low series resistance and therefore improved frequency response up to 10 MHz, which allows for the ﬁrst time DLTS measurements of deep level activation energies of undoped and variously doped PLD ZnO thin ﬁlms.

7.5 Demonstrator Devices with PLD ZnO Thin Films In this section, some selected recent application-oriented results of PLD ZnO ﬁlms and heterostructures will be presented in more detail. In particular, (a) large-area ZnO ﬁlms on sapphire optimized for high cathodoluminescence

7 PLD of ZnO Films

337

Fig. 7.24. AFM (left) and MFM (right) images of identical positions of 3d- or 4f-element doped PLD ZnO ﬁlms on sapphire (from top): ZnO:Mn(2.56 at. %) [77], ZnO:Co(0.2 at. %)Al(0.5 at. %) [84], and ZnO:Ti(9.9 at. %) [78] (see also [83]), and ZnO:Nd(0.1 at. %) [85]. The MFM images taken with a lift scan height of 50 nm of the low-coercivity tip clearly demonstrate the stripe-like or bubble-like magnetic domain formation. The z-scales of the AFM/MFM images are 50 nm/0.8◦ for ZnO:Mn, 50 nm/0.2◦ for ZnO:Co,Al, 100 nm/0.5◦ for ZnO:Ti, and 10 nm/0.8◦ for ZnO:Nd, respectively. Measured by H. Schmidt and M. Ungureanu

yield for scintillator applications, (b) Bragg mirrors consisting of oxide multilayers with maximum reﬂectivity at 3.3 eV together with quantum conﬁnement eﬀects in MgZnO/ZnO/MgZnO quantum well structures, and (c) Pd Schottky diodes to PLD ZnO thin ﬁlms with optimized rectifying behavior are reported. The quantum well structures and the Schottky contacts were already mentioned in Sect. 7.4.5. Finally, an overview on international highlights of PLD ZnO thin ﬁlms and heterostructures, e.g., the ﬁrst blue ZnO-based LED, will be given.

338

M. Lorenz

7.5.1 Large-area ZnO Scintillator Films We patent-pended a pulsed laser deposition (PLD) process [89,90] for n-type conducting ZnO thin ﬁlms, which show high CL intensities under optimized growth conditions. ZnO is described as a new scintillator material class because of its high luminescence intensity and the short decay time in the nanosecond range [91, 92]. Because of increased data rate and resolution in state of the art electron microscopic imaging there is a need for more homogeneous, faster, and if possible brighter scintillators based on epitaxial ZnO thin ﬁlms to replace the polycrystalline ZnO-based phosphors. We investigated ZnO thin ﬁlms grown from nominally undoped ZnO targets in either reducing or oxidizing environments, namely in N2 , N2 O, and O2 background gases, and in a wide 5 orders of magnitude range of background gas pressure from 3 × 10−4 to 10 mbar, as shown in Fig. 7.25 [89]. Because the luminescence spectra have to be detected both from the ﬁlm side (in reﬂection) and through the sapphire substrate (in transmission) [90], the ﬁlms were grown on double-side epi-polished a-plane sapphire substrates of the size 10 × 10 mm2 and 32.8 mm diameter. For the PLD growth of ZnO ﬁlms with area larger than 1 cm2 we employ the so called oﬀset-PLD with parallel target and substrate arrangement and a target to substrate distance of 100 mm [10]. Films grown at about 1 mbar background pressure show the highest normalized excitonic CL intensity, independently of the kind of gas, as shown in Fig. 7.25. This maximum of the near-band-edge luminescence in undoped ZnO ﬁlms is clearly correlated with maxima of both the free carrier concentration and the Hall mobility. For the high growth pressure around 1 mbar, the average surface roughness of the ﬁlms is considerably increased, as shown in Fig. 7.25. However, a rough surface probably promotes the outcoupling of the generated light from the ZnO ﬁlm into the environment. Figure 7.26 shows typical room-temperature CL spectra of three ZnO thin ﬁlms deposited at the optimum pressure of 0.9 mbar for the three background gases. All spectra show the dominating excitonic near-band-edge luminescence peak with a maximum at around 3.25 eV and the two-to-three orders of magnitude weaker defect-dominated luminescence band in the visible spectral range. The intensity of the defect-dominated CL signal increases in the order of used N2 – N2 O – O2 gas. To explain CL spectra taken in dependence on the excitation voltage and for detection of the luminescence light on the ﬁlm or the substrate side of the ﬁlm sample, a model based on self absorption of the generated photons in the ZnO ﬁlm was developed [90, 93]. This model was also used to explain the origin of the remarkable splitting and broadening of the excitonic CL peak of ﬁlms grown in oxygen (see Fig. 7.26). Both small crystallites with size in the 300 nm range and huge single crystalline regions with hexagonal shape in the 20 µm size range are able to emit maximum CL intensity, as shown in Fig. 7.27 [90]. These substantial diﬀerences of the surface morphology are due to the growth conditions, including substrate properties. However, up to now they cannot be related directly to

7 PLD of ZnO Films

339

Fig. 7.25. CL intensity, Hall mobility, and carrier concentration at 300 K of PLD ZnO thin ﬁlms on a-plane sapphire show maxima around 1 mbar background gas pressure of O2 , N2 O, and N2 during growth at 100 mm target substrate distance, indicating the growth condition for ZnO thin ﬁlm scintillators [89]. The average surface roughness is considerably increased at the high growth pressure of 1 mbar

Fig. 7.26. Typical room temperature CL spectra of ZnO thin ﬁlms grown at optimized O2 , N2 O, and N2 background gas pressure of 0.9 mbar [89]. The ﬁlm grown in O2 shows splitting and broadening of the excitonic peak. Only the ﬁlm grown in N2 shows nearly no green luminescence around 2.3 eV photon energy

340

M. Lorenz

3µm

1µm

Fig. 7.27. SEM images of ZnO thin ﬁlms on a-plane sapphire grown in 1 mbar oxygen. Both ﬁlms show very high UV CL intensity, but very diﬀerent ﬁlm morphology due to diﬀerent growth conditions [90]

adjustable PLD parameters. Probably, variations of the miscut of the a-plane sapphire substrates which is speciﬁed within a range of ±1◦ could be the origin for the diﬀerent ﬁlm morphologies. 7.5.2 Bragg Reﬂector Mirrors and ZnO Quantum Well Structures Dielectric Bragg reﬂector mirrors made from thin ﬁlm heterostructures are used for example to enhance the optical conﬁnement in resonators of semiconductor lasers. For the future application of ZnO as the active laser medium at room temperature, Bragg reﬂectors for the excitonic emission energy of ZnO, i.e., for 3.3 eV, are needed. To grow oxide based Bragg mirrors with a high reﬂectivity near 100%, the PLD technique had to be qualiﬁed concerning a precise control of the single layer thicknesses, the chemical composition, and so the refractive indices. We found that MgZnO could not be used as one of the layer materials because of the reasons mentioned in Sect. 7.4.5. We failed to obtain a stable MgZnO composition of consecutively deposited layers. Therefore, ﬁrst attempts of a 9.5 pair ZnO-MgO Bragg structure on sapphire showed a reﬂectivity of 85% at 2.3 eV phonon energy [94]. The isotope intensity depth proﬁle of this particular Bragg mirror is shown in Fig. 7.28. The thickness and the complex dielectric functions of the Bragg mirrors were derived from spectroscopic ellipsometry by ﬁtting the measured data with an appropriate layer model [95, 96]. The refractive index data were obtained from single layers of the used oxide materials. Further improvement of the structural and optical properties of the PLD Bragg mirrors was achieved by substituting ZnO by yttria stabilized zirconia (YSZ, with typically 9 at. % Y2 O3 ), as demonstrated in Fig. 7.29. A considerable increase of the maximum reﬂectivity of the Bragg structures from about 90–99% was realized by doubling the number of YSZ-MgO layer pairs from 5.5 to 10.5 as shown in Fig. 7.29 (top). The experimentally obtained single layer thicknesses of the 5.5 and 10.5 pair structure are given in the caption and show smaller variation compared to the MgO– ZnO structure of Fig. 7.28. Indeed, the SNMS isotope intensity depth proﬁle

7 PLD of ZnO Films

341

Fig. 7.28. SNMS isotope intensity depth proﬁle of a 9.5 pair ZnO-MgO Bragg mirror grown by PLD on c-plane sapphire. This particular Bragg structure had a maximum reﬂectivity of 85% at 2.3 eV photon energy [94]. The single layer thicknesses obtained from UV–vis ellipsometry varied from 80–96 nm (MgO) and 41–71 nm (ZnO)

in Fig. 7.29 (bottom) shows improved lateral thickness homogeneity of the 5.5 pair Bragg structure as expressed by the considerably improved isotope intensity oscillations. First potentially suitable laser structures consisting of YSZ/MgO Bragg–ZnO–YSZ/MgO Bragg are currently under test. Furthermore, ZrO2 –MgO Bragg mirrors were also deposited by PLD on ZnO nano- and micropillars with diameter in the 1 µm range and a considerable enhancement of the CL intensity from the pillar was detected [96]. First results on growth and optical investigation of ZnO quantum well layers embedded in MgZnO with higher band gap energy as future light emitters are shown in Fig. 7.30 [53]. Because the surface roughness Ra of the ﬁrst MgZnO layer was in the 3 nm range, the actual structure of the thinnest quantum wells with nominal thickness of 3 nm is not clear up to now. It could be more a spike-like structure of isolated ZnO dots rather than a closed 2D-like layer, as imaged by AFM in [53]. Nevertheless, Fig. 7.30 shows a remarkable blue-shift coupled with intensity enhancement of the excitonic ZnO peak by decreasing the nominal ZnO quantum well thickness from 25 nm down to 3 nm. The combination of blueshift and increasing intensity are a clear indication for optical conﬁnement of the excitons. 7.5.3 Schottky Diodes to ZnO Thin Films The application of capacitance spectroscopy as for example DLTS (see Sect. 7.4.3 and 7.4.5) requires high-quality Schottky contacts to ZnO single

342

M. Lorenz

Fig. 7.29. Top: Reﬂectivity at normal incidence of two PLD grown Bragg mirrors with 5.5 and 10.5 YSZ-MgO layer pairs obtained from the ellipsometry model analysis. By doubling the layer number, the reﬂectivity was increased from 90 to 99%. The UV–vis ellipsometry data were ﬁtted best with layer thicknesses of 38–46 nm YSZ/48–54 nm MgO for the 5.5 layer pair Bragg, and 46.4 ± 0.7 nm YSZ and 51.9±0.5 nm MgO for the 10.5 pair Bragg. Measured and calculated by R. SchmidtGrund. Bottom: SNMS isotope intensity depth proﬁle of this 5.5×YSZ/MgO Bragg structure

crystals and thin ﬁlms. Furthermore, because of the current diﬃculties to reproducibly obtain highly p-type conducting ZnO, Schottky contacts represent an alternative route to devices with rectifying electrical properties. Continuous improvement of the surface preparation technique prior to the evaporation of the Schottky contact metal, comparison of diﬀerent Schottky contact metals [60], the introduction of the innovative front-to-back Schottky

7 PLD of ZnO Films

343

Fig. 7.30. Photoluminescence spectra (2K) of PLD MgZnO-ZnO-MgZnO quantum well heterostructures on sapphire with nominal thickness of the ZnO quantum well of 25, 12, 6, and 3 nm [53]. The blueshift of the excitonic peak combined with the intensity enhancement is a clear indication of optical conﬁnement in the ZnO layer

diode conﬁguration with ohmic back contact ZnO:Al layer below the undoped or doped ZnO layer as already mentioned in Sect. 7.4.5 resulted in good Schottky contacts on PLD ZnO thin ﬁlms [55, 57]. On the insulating sapphire substrate, at ﬁrst an about 50 nm thick highly Al-doped ZnO ﬁlm was deposited that serves as the ohmic back contact. On top of this back contact layer, the nominally undoped or doped ZnO ﬁlms are grown. The ZnO:Al ﬁlms are contacted with sputtered Au ﬁlm. The Pd Schottky contacts were realized by thermal evaporation of Pd through shadow masks. The contact area is usually around 10−4 cm2 . The series resistance of such front-to-back Pd–ZnO–ZnO:Al–Au Schottky diodes is as low as about 50–200 Ω [55]. Figure 7.31 demonstrates the very good rectifying behavior of such a Pd Schottky diode on undoped ZnO thin ﬁlm. The current density ratio determined for bias voltages of +0.6 V and –3 V is about 104 as shown in the inset of Fig. 7.31. The ideality factor n is about 1.5. The temperaturedependent current–voltage (IV, see Fig. 7.31) and capacitance–voltage (CV) measurements from 210 to 300 K explain the reason for the slight deviation of the ideality factor from unity and the dependence of the reverse current on the reverse bias. The barrier heights of the diode of Fig. 7.31 ΦIV and ΦCV as determined from IV- and CV-measurements amount to 0.82 and 1.16 eV, respectively [97]. The diﬀerence of the two barrier height values is due to the diﬀerent eﬀect of lateral potential ﬂuctuations as explained in detail in [55, 57, 97].

344

M. Lorenz

10 0

300 K 290 K

10 -1 10

280 K 270 K

-2

260 K 240 K

-4

220 K

10

230 K 210 K

10 -5 10 -6 10 -7 10 -8 10 -9

10 0 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6 10 -7 10 -8 10 -9 10 -10

|jd|(A/cm2)

jd (A/cm2)

250 K

10 -3

-3

-2

-1

0

1

Vd (V)

0.0

0.5

1.0

Vd (V) Fig. 7.31. Superior rectifying behavior of a Pd Schottky contact on PLD ZnO ﬁlm with ohmic ZnO:Al back layer contact, with excellent stability at diﬀerent temperature. The inset shows the current density–voltage dependence for a larger voltage range at 290 K. Reprinted with permission from [55]

7.5.4 PLD of ZnO pn-Junctions, First LEDs, and Other Highlights Table 7.10 gives an overview on selected highlights of the PLD of ZnO thin ﬁlms, thereby demonstrating the state of the art. The table is divided into three sections, corresponding to (1) p-type ZnO ﬁlms and pn-junctions, (2) n-type ZnO and MgZnO ﬁlms on various substrate materials, and (3) ferromagnetic 3d-element doped ZnO ﬁlms. For the results of the Leipzig group see Table 7.9. The following results in Table 7.10 should be especially emphasized: (a) the p-type ZnO ﬁlms and the ZnO-LED structure grown by laser MBE of semiconductor-quality ZnO ﬁlms on lattice matched ScAlMgO4 (SCAM) single crystals by the group of M. Kawasaki at Tohuku University Sendai in Japan, (b) the systematic PLD of ZnO and MgZnO ﬁlms for solar blind photodetectors in the group of T. Venkatesan at University of Maryland in the U.S.A., and (c) the PLD-based systematics on ferromagnetic 3d-element doped ZnO by J. M. D. Coey at Trinity College, University of Dublin, Ireland. For a recent review on p-type doping of ZnO including a list of published ptype ﬁlms and their deposition technique and devices see also the review [43]. For more general reviews on ZnO covering results from all deposition techniques see [41, 42, 44, 45], and also [98].

7 PLD of ZnO Films

345

Table 7.10. Selected highlights of the PLD of (1) p-type ZnO ﬁlms, including pn-junctions, (2) n-type ZnO ﬁlms on various substrates, and (3) ferromagnetic ﬁlms Keyword

ZnO-based structure, methods, state of the art result

p-type ZnO:N ﬁlms

Temperature-modulation epitaxy of atomically smooth p-ZnO:N/i-ZnO/n-ZnO/buﬀer-ZnO on SCAM by laser MBE with p = 2 × 1016 cm−3 (cN = 2 × 1020 cm−3 ), µH = 8 cm2 (V s)−1 (300 K), and ND /NA = 0.8 (EA = 100 meV) ZnO LED ZnO-based LED with pn-junction and electroluminescence with maxima around 3.05 eV (405 nm) and 2.15 eV with 20 mA injection current ZnO:P ﬁlms doping behavior of PLD ZnO:P(1 − 5 at. %) ﬁlms, after annealing semi-insulating ﬁlms by deep level formation p-ZnO:As PLD of ZnO together with As from eﬀusion cell, p-type ﬁlms proved by Hall eﬀect: p = 8 × 1016 to 4 × 1017 cm−3 , and µ = 6–35 cm2 (V s)−1 , growth also on GaAs ZnO LED Au/p(i)-ZnO/n-ZnO single crystal/In structure grown by N2 O plasma enhanced PLD, rectifying behavior, bluish-white electroluminescence p-NiO/ ZnO/NiO/ITO on (111)YSZ grown by PLD, p-type n-ZnO NiO:Li(10 at. %) annealed, rectifying behavior in junction dependence on UV illumination, UV sensing properties PLD of ZnO and Zn3 P2 on sapphire(0001), laser Zn3 P2 / n-ZnO annealing of the Zn3 P2 , rectifying behavior diode n-ZnO on SCAM

n-ZnO on GaN n-ZnO on SrTiO3 n-ZnO on InP n-ZnO on Al2 O3

Laser-MBE of high-quality n-type ZnO and MgZnO ﬁlms on SCAM grown at 950◦ C and 10−7 mbar, µH = 440 cm2 (V s)−1 (300 K) and 5,000 cm2 (V s)−1 (100 K), eﬀect of Mgx Zn1−x O capping layers on the PL of ZnO/SCAM PLD of high-quality ZnO and GaN on c-plane sapphire, enhanced CL excitonic intensity compared to ZnO/sapphire PLD of undoped and Al-doped ZnO(11-20) on SrTiO3 (001) and (011), structure, ﬁeld eﬀect experiments (FET) Surface morphology evolution of PLD ZnO ﬁlms on InP(100) at 623 K growth temperature PLD/laser MBE of ZnO on c-plane sapphire, structure and electrical properties in dependence on oxygen pressure and growth temperature, optical properties, properties of A, B, C exciton, band gap and strain for ZnO on c- and r -plane sapphire, and fused silica

Refs. [99]

[99]

[100] [101]

[102]

[103]

[104]

[105, 106]

[107]

[108–110]

[111] [112–116]

346

M. Lorenz Table 7.10. (continued)

Keyword

ZnO-based structure, methods, state of the art result

Polarity controlled ZnO MgZnO

Laser MBE of Zn-face and O-face ZnO ﬁlms on ZnO(0001) single crystals, Zn-face shows Ra = 0.18 nm, Rrms = 0.21 nm and Rpeak−valley = 0.74 nm. PLD of Mgx Zn1−x O (0 ≤ x ≤ 1) in wurtzite and cubic phase, solar blind photodetectors

ZnO:3delements

Room temperature ferromagnetism found in PLD (110)ZnO:Sc, Ti, V, Fe, Co, Ni thin ﬁlms on r -plane sapphire, but not for ZnO:Cr, Mn, Cu (all 5 at. %). Large moments of 1,9 and 0.5 µB /atom for ZnO:Co and ZnO:Ti, respectively Origin of ferromagnetism in low-temperature processed ZnO:Mn is probably oxygen vacancy stabilized Mn2−x Znx O3−δ

Mn–Zn–O ﬁlms

Refs. [117]

[118, 119] [120, 121]

[122], see also [123]

Compare also the PLD review [22]. For results of the Leipzig group see Table 7.9

7.6 Advanced PLD Techniques In this section advanced developments of the PLD technique are shortly described, e.g., the combinatorial approach and laser MBE. Because of the increasing research interest in ZnO-based nanostructures, the established PLD (at background gas pressure 10−4 −3 mbar) was extended to much higher pressures of 50–200 mbar to grow arrays of free-standing ZnO nanowires. This unique high-pressure PLD process allows the growth of ZnO-based nanostructures with controlled shape and diameter and excellent optical properties. 7.6.1 Advances in PLD of Thin Films The ﬂexibility of the PLD caused by decoupling of laser and growth chamber promoted the development of a variety of advanced PLD techniques to open new growth possibilities and to overcome some limitations of the technique. Table 7.11 gives an overview on recent advances in PLD of thin ﬁlms. The PLD with ps- or ns-pulse length lasers and the pulsed electron beam deposition (PED) developed by Neocera refer to advanced pulsed energy sources for plasma generation. The PED process uses a pulsed electron beam gun instead of the expensive excimer laser and has therefore high potential to push the application of the technique towards commercial deposition. The other advanced PLD techniques modify the deposition conditions (laser MBE) or the geometric arrangement in multitarget deposition (combinatorial and gradient approaches) in order to ensure cleaner nucleation and deposition under UHV conditions or simultaneous deposition of composition libraries of new multielement compounds. For such libraries, appropriate characterization tools with high lateral resolution are necessary to beneﬁt from the fast

7 PLD of ZnO Films

347

Table 7.11. Advanced techniques for the PLD of thin ﬁlms PLD technique

Main innovation, advantages

References

PLD with sub-ns pulse length laser

Comparison fs-laser to ns-laser: higher kinetic energy of ablated ZnO species, higher mosaicity, smaller grain size, but also smaller residual stress of ZnO on c-sapphire. May be advantageous for special materials

[124, 125]

Continuous composition spread (CCS-PLD)

Continuous, controlled, vertical composition gradient by PLD, for example, applied for Ba1−x Srx TiO3 ﬁlms

[126–128]

Combinatorial PLD

Dramatic increase of the rate of discovery and improvement of new compounds, synthesis of up to thousands of diﬀerent compositions on one wafer in a single growth run

[129–131]

PLD in UHV (laser-MBE)

MBE-like background pressure and in situ RHEED to ensure clean and controlled deposition of high-quality nucleation layers and ﬁlms. For particular systems as SrTiO3 and BaTiO3 , atomically smooth surface and interface were obtained

[128, 132]

RF- and ion-beam PLD chamber equipped with RF or microwave assisted PLD plasma source or ion beam source to enhance the composition of particular ﬁlm components such as N or O to grow in-plane aligned ﬁlms on polycrystalline substrates, and to grow nanostructures

[128]

Oﬀ-axis and dual laser PLD for droplet reduction

Dynamic melt studies of the target surface, time synchronized irradiation of the target with CO2 and KrF lasers leads to particulate free ZnO ﬁlms

[133, 134]

PLD for series production

Pilot scale production PLD systems are in successful operation for X-ray mirrors and are commercially available with excellent thickness and composition homogeneity and reproducibility

[14, 128]

Pulsed electron beam deposition (PED)

Pulsed electron beam source emitting 100 ns long electron pulses with 10–20 keV and ∼kA intensity into the deposition chamber, no excimer laser is required, innovative complimentary technique to PLD, further extending the range of materials to be grown as thin ﬁlms by pulsed energy techniques

[128, 135]

348

M. Lorenz

and cost eﬀective deposition of complete mixing range materials by combinatorial PLD. Furthermore, hybrid techniques are used employing additional lasers, RF or microwave (ECR) plasma sources, or ion beam sources to modify the target ablation for droplet suppression or to improve ﬁlm composition and structure. Thus, Table 7.11 summarizes the innovative potential that is inherent to PLD. 7.6.2 High-Pressure PLD of ZnO-Based Nanostructures Only a few attempts to obtain nanostructured Si and ZnO using PLD have been published in the last years [136–139]. Figure 7.32 shows the scheme of the high-pressure PLD chamber specially designed for nano-heterostructures using a T-shaped quartz tube with an outer diameter of 30 mm in the Leipzig group [140]. A KrF excimer laser beam enters along the center of the T and is focused on the cylindrical surface of one of the two rotating laser targets, thus allowing in situ modulation of chemical composition of the grown nanostructures. The laser energy density on the target is about 2 J cm−2 , which is similar to conventional PLD ﬁlm growth conditions. An encapsulated heater with an arrangement of KANTHAL wire in ceramic tubes and FIBROTHAL isolation material is built around the T-shape quartz tube. The growth temperature is usually chosen between 500 and 950◦ C as measured by a thermocouple. A downstream gas ﬂow of argon of 0.05–0.2 l min−1 results in an Ar gas pressure from 25 to 200 mbar. The diameter of single ZnO wires could be varied between about 50 and 3,000 nm by control of the target to substrate distance from 5 to 35 mm [140]. The a-plane or c-plane sapphire KrF laser beam UV lens UV window

to vacuum pump

substrate heater with thermal isolation gas inlet

substrate plasma

viewport

quartz tube targets

linear-rotary feedthrough

Fig. 7.32. Schematic illustration of the high-pressure PLD chamber for ZnO-based nano-heterostructures consisting of a T-shaped quartz tube with 30 mm outer diameter. Reprinted with permission from [140]

7 PLD of ZnO Films

349

Fig. 7.33. Typical SEM images of PLD grown ZnO nanowires (100 mbar Ar, 840◦ C) on a-plane sapphire with Au colloides as nucleation sites [142]

substrates (size 1 × 1 cm2 ) were arranged oﬀ-axis, i.e., parallel to the expanding plasma plume. They were partially (using a hole mask) or fully covered before PLD with DC-sputtered gold ﬁlms of nominally 1 nm thickness. Alternatively, gold colloides of variable size and lateral density or laterally aligned gold seeds obtained by nanosphere lithography mask transfer technique [141] can be used as growth templates. Typical SEM images of ZnO nanowires grown by high-pressure PLD at 100 mbar Ar pressure and at 840◦ C on aplane sapphire covered with Au colloides are shown in Fig. 7.33 [142].

7.7 Summary This chapter about pulsed laser deposition demonstrated the extensive opportunities of PLD to grow nominally undoped and doped ZnO thin ﬁlms and heterostructures with well deﬁned structure, surface morphology, and electrical and optical properties. An outstanding advantage of PLD in comparison to other physical and chemical growth techniques is the ﬂexibility in doping and alloying ZnO thin ﬁlms with a variety of dopant elements and oxides, thus making PLD most suitable for exploratory research on new materials. Furthermore, because PLD is especially appropriate for all kind of oxides, multilayer heterostructures can be easily built from diﬀerent oxide materials by PLD. PLD ZnO thin ﬁlms grown at higher temperature of a few 100◦ C are usually crystalline and granular with minimum average and rms surface roughness down to 0.2 nm. For selected materials and growth conditions, atomically smooth ﬁlm surfaces were also obtained. The grains show a nearly perfect, single crystal-like vertical and lateral alignment on various single crystalline substrates as ScAlMgO4 (SCAM), c- and a-plane sapphire, and GaN. The combination of ZnO with dielectric and ferroelectric materials

350

M. Lorenz

(SrTiO3 , BaTiO3 ) with ﬁxed and switchable polarization, respectively, oﬀers new perspectives for innovative electronic and storage devices. The carrier concentration of n-type conducting PLD ZnO thin ﬁlms can be controlled from highly doped 1020 cm−3 down to the semi-insulating 1012 cm−3 range. The electron mobility of PLD ZnO ﬁlms on sapphire and SCAM exceeds 150 and 300 cm2 (V s)−1 at room temperature, respectively. The peak width of the donor bound exciton peak in low-temperature photoluminescence of high-quality PLD ZnO ﬁlms is around 1 meV, which is as narrow as measured for commercial ZnO bulk single crystals. First successful ZnO device demonstrations as for example stable homoand heteroepitaxial pn-junctions and LED structures, thin ﬁlm scintillators, and quantum well structures with optical conﬁnement, and oxide-based Bragg reﬂectors, and high-quality Schottky contacts are based on PLD grown thin ﬁlms. Several techniques as for example the PLD in UHV conditions (laser MBE), and gradient and combinatorial PLD, and high-pressure PLD for nano-heterostructures show the innovative potential of the advanced growth technique PLD. Acknowledgement. The author thanks all colleagues of the Semiconductor Physics Group of University Leipzig for their kind cooperation in the scientiﬁc and technical preparation of this manuscript, especially for providing graphs of published and unpublished results and for help on LaTeX. These PLD developments would not have been possible without the longstanding technical coworkers of the PLD lab Dieter Natusch, Holger Hochmuth and Gabriele Ramm. The PLD equipment and the initial development of the Leipzig PLD group was ﬁnanced mainly by the German Federal Ministry of Education and Research (BMBF) from 1990 to 2003 and by the Saxonian Ministry of Science and Art SMWK in Dresden, which is hereby highly acknowledged. Currently, the development of the high-pressure PLD process is supported within the DFG research group 522 “Architecture of nano- and microdimensional building blocks” and by the European Commission within the STREP collaboration project “Nanophotonic and Nanoelectronic Devices from Oxide Semiconductors” NANDOS. The growth of the magnetic ZnO-based ﬁlms and multilayers is supported by the BMBF Young Scientists group “Nanospintronics.”

References 1. J.T. Cheung, History and Fundamentals of Pulsed Laser Deposition. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 1–22 2. D. B¨ auerle, Laser Chemical Processing. In: Landolt-B¨ ornstein New Series, Group VIII Advanced Materials and Technologies, Vol. 1 Laser Physics and Applications, Subvolume C Laser Applications, ed. by R. Poprawe, H. Weber, G. Herziger (Springer, Berlin 2004) pp 311–354 3. G.W. Martin, L.A. Doyle, A. Al-Kateeb, I. Weaver, D. Riley, M.J. Lamb, T. Morrow, C.L.S. Lewis, Appl. Surf. Sci. 127–129, 710 (1998)

7 PLD of ZnO Films

351

4. A. Husmann, Pulsed Laser Deposition mittels g¨ utegeschalteter CO2 -Laser. PhD Thesis, RWTH Aachen, Aachen (1999) 5. T. Venkatesan, Pulsed Laser Deposition - Future Trends. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 313–326 6. J.S. Horwitz, Film Nucleation and Film Growth in Pulsed Laser Deposition of Ceramics. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 229–254 7. M. Lorenz, S. Becker, H.-J. Dietze, W. Schmitz, B. Brunner, K.F. Renk, Physica C 182, 114 (1991) 8. M. Lorenz, H. Hochmuth, D. Natusch, M. Kusunoki, V.L. Svetchnikov, V. Riede, I. Stanca, G. K¨ astner, D. Hesse, IEEE Trans. Appl. Supercond. 11, 3209 (2001) 9. J.A. Greer, Commercial Scale-Up of Pulsed Laser Deposition. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 293–312 10. M. Lorenz, H. Hochmuth, D. Natusch, H. B¨ orner, K. Kreher, W. Schmitz, Appl. Phys. Lett. 68, 3332 (1996) 11. J.S. Horwitz, D.B. Chrisey, R.M. Stroud, A.C. Carter, J. Kim, W. Chang, J.M. Pond, S.W. Kirchoefer, M.S. Osofsky, D. Koller, Appl. Surf. Sci. 127–129, 507 (1998) 12. D. B¨ auerle, Laser Processing and Chemistry, 3rd edn. (Springer, Berlin Heidelberg Newyork, 2000), pp 3–100 13. L.A. Doyle, G.W. Martin, A. Al-Kateeb, I. Weaver, D. Riley, M.J. Lamb, T. Morrow, C.L.S. Lewis, Appl. Surf. Sci. 127–129, 716 (1998) 14. Applied X-ray Optics AXO Dresden GmbH, Heidenau, Germany, http:// www.axo-dresden.de 15. AxynTeC D¨ unnschichttechnik GmbH, Augsburg, Germany, http:// www.axyntec.de 16. nanovation SARL, Orsay, France, http://www.nanovation.biz 17. B. Angstenberger, Fliehkraftunterst¨ utztes Laserbeschichten. PhD Thesis, Universit¨ at Stuttgart, Stuttgart (2000) 18. E.W. Kreutz, Appl. Surf. Sci. 127–129, 606 (1998) 19. K.L. Saenger, Angular Distribution of Ablated Material. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 199–228 20. L.-C. Chen, Particulates Generated by Pulsed Laser Ablation. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 167–198 21. A.E. Tselev, Cross-beam pulsed laser deposition as a method for preparation of thin ﬁlms of metastable solid solutions. PhD Thesis, TU Dresden, Dresden (2000) 22. A. Ohtomo, A. Tsukazaki, Semicond. Sci. Technol. 20, S1 (2005) 23. D.B. Geohegan, Diagnostics and Characteristics of Laser-Produced Plasmas. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 115–166 24. H. H¨ ugel, F. Dausinger, Fundamentals of Laser-Induced Processes. In: Landolt-B¨ ornstein New Series, Group VIII Advanced Materials and Technologies, Vol. 1 Laser Physics and Applications, Subvolume C Laser Applications,

352

25. 26.

27.

28. 29. 30.

31. 32. 33.

34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46.

47.

M. Lorenz ed by R. Poprawe, H. Weber, G. Herziger (Springer, Berlin Heidelberg New York 2004) pp 311–354 J. Gottmann, Dynamik der Schichtabscheidung von Keramiken mit KrFExcimer-Laserstrahlung. PhD Thesis, RWTH Aachen, Aachen (2001) R. Kelly, A. Miotello, Mechanisms of Pulsed Laser Sputtering. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 55–88 S.R. Foltyn, Surface Modiﬁcation of Materials by Cumulative Laser Irradiation. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp. 89–114 R.E. Leuchtner, Appl. Surf. Sci. 127–129, 626 (1998) Y. Kawaguchi, A. Narazaki, T. Sato, H. Niino, A. Yabe, Appl. Surf. Sci. 197–198, 268 (2002) S. Metev, Process Characteristics and Film Properties in Pulsed Laser Deposition. In: Pulsed Laser Deposition of Thin Films, ed. by D.B. Chrisey, G.H. Hubler (Wiley, New York Chichester Brisbane Toronto Singapore 1994) pp 229–254 S. Ohashi, M. Lippmaa, N. Nakagawa, H. Nagasawa, H. Koinuma, M. Kawasaki, Rev. Sci. Instrum. 70, 178 (1999) Information sheet LPX300i Series 1996 and 2004, Coherent Lambda Physik GmbH, G¨ ottingen, Germany, http://www.lambdaphysik.com Information sheet YG980 Pulsed Nd:YAG Laser Speciﬁcations, http://www. quantel.fr/uk/, for more details see http://www.bigskylaser.0com/pdf/ yg980web.pdf Neocera Inc., Beltsville, MD, U.S.A., http://www.neocera.com Twente Solid State Technology B.V., Hengelo, Netherlands, http:// www.tsst.nl Koinuma-Kawasaki Mobile Combinatorial PLD, Huntington Mechanical Laboratories Inc, Mountain View, CA, U.S.A., http://www.huntvac.com/PLD/ PVD Products Inc., Wilmington, MA, U.S.A., http://www.pvdproducts.com Surface, H¨ uckelhofen, Germany, http://www.surface-tec.com DCA instruments Oy, Turku, Finland, http://www.dca.ﬁ Pascal Technologies Inc., Fredericksburg, VA, U.S.A., http:// www.pascaltechnologies.com ¨ Ozg¨ ¨ ur, Ya. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, V. Avrutin, U. S.-J. Cho, H. Morko¸c, J. Appl. Phys. 98, 041301 (2005) R. Tribulet, J. Perri`ere, Prog. Cryst. Growth Charact. Mater. 47, 65 (2003) D.C. Look, B. Claﬂin, Phys. Stat. Sol. B 241, 624 (2004) C. Liu, F. Yun, H. Morko¸c, J. Mater. Sci. 16, 555 (2005) T. Yao, Zinc Oxide. In: Encyclopedia of Materials: Science and Technology, (Elsevier, Amsterdam London 2001) pp 9883–9888 W. Hirschwald, P. Bonasewicz, L. Ernst, M. Grade, D. Hoﬀmann, S. Krebs, R. Littbarski, G. Neumann, M. Grunze, D. Kolb, H.J. Schulz, Zinc Oxide. In: Current Topics in Materials Science, Vol. 7, ed. by E. Kaldis (North Holland, Amsterdam 1981) pp 143–482 A. Rahm, High-resolution X-ray diﬀraction of ZnO-based thin ﬁlms. Diploma thesis, Universit¨ at Leipzig, Leipzig (2003)

7 PLD of ZnO Films

353

48. M. Lorenz, H. Hochmuth, H. von Wenckstern, H. Schmid, W. Mader, M. Grundmann, Crystalline nanostructure of PLD ZnO and MgZnO thin ﬁlms on sapphire. In: Universit¨ at Leipzig, The Physics Institutes Report 2004, ed. by M. Grundmann (Leipzig 2005) p 156 49. S. Heitsch, C. Bundesmann, G. Wagner, G. Zimmermann, A. Rahm, H. Hochmuth, G. Benndorf, H. Schmidt, M. Schubert, M. Lorenz, H. Schmidt, M. Schubert, M. Grundmann, Thin Solid Films 496, 234 (2006) 50. M. Lorenz, H. Hochmuth, A. Jammoul, G. Ferro, C. F¨ orster, J. Pezoldt, J. Zuniga Perez, G. Benndorf, J. Lenzner, R. Schmidt-Grund, M. Grundmann, Wissenschaftlich-Technische Berichte of Research Center Rossendorf FZR433, 74 (2005) 51. E.M. Kaidashev, M. Lorenz, H. von Wenckstern, A. Rahm, H.-C. Semmelhack, K.-H. Han, G. Benndorf, C. Bundesmann, H. Hochmuth, M. Grundmann, Appl. Phys. Lett. 82, 3901 (2003) 52. Y.F. Chen, D.M. Bagnall, H. Koh, K. Park, K. Hiraga, Z. Zhu, T. Yao, J. Appl. Phys. 84, 3912 (1998) 53. S. Heitsch, G. Benndorf, G. Zimmermann, C. Schulz, D. Spemann, H. Hochmuth, H. Schmidt, Th. Nobis, M. Lorenz, M. Grundmann, Appl. Phys. A 88, 99 (2007) 54. M. Lorenz, H. Hochmuth, J. Lenzner, M. Brandt, H. von Wenckstern, G. Benndorf, M. Grundmann, Wissenschaftlich-Technische Berichte of Research Center Rossendorf FZR-433, 57 (2005) 55. M. Grundmann, H. von Wenckstern, R. Pickenhain, Th. Nobis, A. Rahm, M. Lorenz, Superlatt. Microstruct. 38, 317 (2005) 56. M. Lorenz, E.M. Kaidashev, H. von Wenckstern, V. Riede, C. Bundesmann, D. Spemann, G. Benndorf, H. Hochmuth, A. Rahm, H.-C. Semmelhack, M. Grund mann, Solid State Electron. 47, 2205 (2003) 57. H. von Wenckstern, S. Weinhold, G. Biehne, R. Pickenhain, H. Schmidt, H. Hochmuth, M. Grundmann, Donor Levels in ZnO. In: Advances in Solid State Physics, Vol. 45, ed. by B. Kramer (Springer, Berlin Heidelberg New York 2005) pp 263–275 58. H. von Wenckstern, G. Benndorf, S. Heitsch, J. Sann, M. Brandt, H. Schmidt, J. Lenzner, M. Lorenz, A.Y. Kuznetsov, B.K. Meyer, M. Grundmann, Appl. Phys. A 88, 125 (2007) 59. H. von Wenckstern, M. Brandt, H. Schmidt, G. Biehne, R. Pickenhain, H. Hochmuth, M. Lorenz, M. Grundmann, Appl. Phys. A 88, 135 (2007) 60. H. von Wenckstern, E.M. Kaidashev, M. Lorenz, H. Hochmuth, G. Biehne, J. Lenzner, V. Gottschalch, R. Pickenhain, M. Grundmann, Appl. Phys. Lett. 84, 79 (2004) 61. M. Grundmann, H. von Wenckstern, Proc. FVS-Workshop TCOs f¨ ur D¨ unnschichtsolarzellen und andere Anwendungen, Freyburg, Germany, 10–12. April 2005, Hahn–Meitner–Institut Berlin 62. B.K. Meyer, H. Alves, D.M. Hofmann, W. Kriegseis, D. Forster, F. Bertram, J. Christen, A. Hoﬀmann, M. Straburg, M. Dworzak, U. Haboeck, A.V. Rodina, Phys. Stat. Sol. B 241, 231 (2004) 63. W. Czakai, Photolumineszenz an ZnO. Diploma thesis, Universit¨at Leipzig, Leipzig (2004) 64. M. Strassburg, A. Rodina, M. Dworzak, U. Haboeck, I.L. Krestnikov, A. Hoﬀmann, O. Gelhausen, M.R. Phillips, H.R. Alves, A. Zeuner, D.M. Hofmann, B.K. Meyer, Phys. Stat. Sol. B 241, 607 (2004)

354

M. Lorenz

65. D. Spemann, E.M. Kaidashev, M. Lorenz, J. Vogt, T. Butz, Nucl. Instrum. Meth. B 219–220, 891 (2004) 66. C. Bundesmann, M. Schubert, D. Spemann, T. Butz, M. Lorenz, E.M. Kaidashev, M. Grundmann, N. Ashkenov, H. Neumann, G. Wagner, Appl. Phys. Lett. 81, 2376 (2002) 67. C. Bundesmann, M. Schubert, D. Spemann, A. Rahm, H. Hochmuth, M. Lorenz, M. Grundmann, Appl. Phys. Lett. 85, 905 (2004) 68. C. Bundesmann, A. Rahm, M. Lorenz, M. Grundmann, M. Schubert, J. Appl. Phys. 99, 113504 (2006) 69. R. Schmidt, B. Rheinl¨ ander, M. Schubert, D. Spemann, T. Butz, J. Lenzner, E.M. Kaidashev, M. Lorenz, M. Grundmann, Appl. Phys. Lett. 82, 2260 (2003) 70. R. Schmidt-Grund, M. Schubert, B. Rheinl¨ ander, D. Fritsch, H. Schmidt, E.M. Kaidashev, M. Lorenz, C.M. Herzinger, M. Grundmann, Thin Solid Films 455–456, 500 (2004) 71. R. Schmidt-Grund, A. Carstens, B. Rheinl¨ ander, D. Spemann, H. Hochmut, G. Zimmermann, M. Lorenz, M. Grundmann, C.M. Herzinger, M. Schubert, J. Appl. Phys. 99, 123701 (2006) 72. J. Zuniga-Perez, V. Munoz-Sanjose, M. Lorenz, G. Benndorf, S. Heitsch, D. Spemann, M. Grundmann, J. Appl. Phys. 99, 023514 (2006) 73. J. Zuniga Perez, V. Munoz-Sanjose, M. Lorenz, H. Hochmuth, G. Benndorf, S. Heitsch, D. Spemann, M. Grundmann, Third SOXESS workshop on ZnO, EC Contract G5RT-CT-2002-05075, 28 September – 1 October 2005, Gallipoli, Italy, Poster p.13 74. C. Bundesmann, N. Ashkenov, M. Schubert, D. Spemann, T. Butz, E.M. Kaidashev, M. Lorenz, M. Grundmann, Appl. Phys. Lett. 83, 1974 (2003) 75. S. Heitsch, Photoluminescence of p-doped ZnO thin ﬁlms. M.Sc. thesis, Universit¨ at Leipzig, Leipzig (2003) 76. H. von Wenckstern, S. Heitsch, G. Benndorf, D. Spemann, E.M. Kaidashev, M. Lorenz, M. Grundmann, AIP Conf. Proc. 772, 183 (2005) 77. M. Diaconu, H. Schmidt, H. Hochmuth, M. Lorenz, G. Benndorf, J. Lenzner, D. Spemann, A. Setzer, K.-W. Nielsen, P. Esquinazi, M. Grundmann, Thin Solid Films 486, 117 (2005) 78. H. Schmidt, M. Diaconu, H. Hochmuth, M. Lorenz, A. Setzer, P. Esquinazi, A. P¨ oppl, D. Spemann, K.W. Nielsen, R. Gross, G. Wagner, M. Grundmann, Superlatt. Microstruct. 39, 334 (2006) 79. M. Diaconu, H. Schmidt, A. P¨ oppl, R. B¨ ottcher, J. Hoentsch, A. Klunker, D. Spemann, H. Hochmuth, M. Lorenz, M. Grundmann, Phys. Rev. B 72, 085214 (2005) 80. M. Diaconu, H. Schmidt, A. P¨ oppl, R. B¨ ottcher, J. Hoentsch, A. Rahm, H. Hochmuth, M. Lorenz M. Grundmann, Superlatt. Microstruct. 38, 413 (2005) 81. H. Schmidt, M. Diaconu, H. Hochmuth, G. Benndorf, H. von Wenckstern, G. Biehne, M. Lorenz, M. Grundmann, Appl. Phys. A 88, 157 (2007) 82. M. Diaconu, H. Schmidt, H. Hochmuth, M. Lorenz, H. von Wenckstern, G. Biehne, M. Grundmann, Solid State Commun. 137, 417 (2006) 83. Qingyu Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. SchmidtGrund, D. Spemann, A. Rahm, M. Grundmann, Thin Solid Films 515, 2549 (2006)

7 PLD of ZnO Films

355

84. Qingyu Xu, L. Hartmann, H. Schmidt, H. Hochmuth, M. Lorenz, R. SchmidtGrund, C. Sturm, D. Spemann, M. Grundmann, Phys. Rev. B 73, 205342 (2006) 85. M. Ungureanu, H. Schmidt, Q.Y. Xu, H. von Wenckstern, D. Spemann, H. Hochmuth, M. Lorenz, M. Grundmann, E-MRS Spring Meeting, Nice, 29 May to 2 June 2006, Symposium K: ZnO and related materials, Poster K PII 02 86. M. Schubert, N. Ashkenov, T. Hofmann, M. Lorenz, H. Hochmuth, H. von Wenckstern, M. Grundmann, G. Wagner, Ann. Phys. (Leipzig) 13, 61 (2004) 87. N. Ashkenov, M. Schubert, E. Twerdowski, B.N. Mbenkum, H. Hochmuth, M. Lorenz, H.V. Wenckstern, W. Grill, M. Grundmann, Thin Solid Films 486, 153 (2005) 88. B.N. Mbenkum, N. Ashkenov, M. Schubert, M. Lorenz, H. Hochmuth, D. Michel, M. Grundmann, G. Wagner, Appl. Phys. Lett. 86, 091904 (2005) 89. M. Lorenz, H. Hochmuth, J. Lenzner, T. Nobis, G. Zimmermann, M. Diaconu, H. Schmidt, H. von Wenckstern, M. Grundmann, Thin Solid Films 486, 205 (2005) 90. R. Johne, M. Lorenz, H. Hochmuth, J. Lenzner, H. von Wenckstern, G. Zimmermann, H. Schmidt, R. Schmidt-Grund, M. Grundmann, Appl. Phys. A 88, 89 (2007) 91. W.W. Moses, Nucl. Instrum. Meth. A 487, 123 (2002) 92. S.E. Derenzo, M.J. Weber, E. Bourret-Courchesne, M.K. Klintenberg, Nucl. Instrum. Meth. A 505, 111 (2003) 93. R. Johne, Kathodolumineszenz-Untersuchung von ZnO-D¨ unnﬁlmen f¨ ur Szintillator-Anwendungen - Experiment und Modellierung. Diploma thesis, Universit¨ at Leipzig, Leipzig (2006) 94. M. Lorenz, H. Hochmuth, R. Schmidt-Grund, E.M. Kaidashev, M. Grundmann, Ann. Phys.(Leipzig) 13, 59 (2004) 95. R. Schmidt-Grund, T. Nobis, V. Gottschalch, B. Rheinl¨ander, H. Herrnberger, M. Grundmann, Thin Solid Films 483, 257 (2005) 96. R. Schmidt-Grund, T. G¨ uhne, H. Hochmuth, B. Rheinl¨ ander, A. Rahm, V. Gottschalch, J. Lenzner, M. Grundmann, SPIE 6038, 489 (2006) 97. H. von Wenckstern, G. Biehne, R. Abdel Rahman, H. Hochmuth, M. Lorenz, M. Grundmann, Appl. Phys. Lett. 88, 092102 (2006) 98. C. Klingshirn, M. Grundmann, A. Hoﬀmann, B. Meyer, A. Waag, Phys. J. 5, 33 (2006) 99. A. Tsukazaki, A. Ohtomo, T. Onuma, M. Ohtani, T. Makino, M. Sumiya, K. Ohtani, S.F. Chichibu, S. Fuke, Y. Segawa, H. Ohno, H. Koinuma, M. Kawasaki, Nat. Mater. 4, 2 (2005) 100. Y.-W. Heo, S.J. Park, K. Ip, S.J. Pearton, D.P. Norton, Appl. Phys. Lett. 83, 1128 (2003) 101. Y.R. Ryu, T.S. Lee, H.W. White, Appl. Phys. Lett. 83, 87 (2003) 102. X.-L Guo, J.-H. Choi, H. Tabata, T. Kawai, Jpn. J. Appl. Phys. 40, L177 (2001) 103. H. Ohta, M. Hirano, K. Nakahara, K. Nakahara, H. Maruta, T. Tanabe, M. Kamiya, T. Kamiya, H. Hosono, Appl. Phys. Lett. 83, 1029 (2003) 104. S.Y. Lee, E.S. Shim, H.S. Kang, S.S. Pang, J.S. Kang, Thin Solid Films 473, 31 (2005) 105. T. Makino, K. Tamura, C.H. Chia, Y. Segawa, M. Kawasaki, A. Ohtomo, H. Koinuma, Appl. Phys. Lett. 81, 2172 (2002)

356

M. Lorenz

106. A. Tsukazaki, A. Ohtomo, M. Kawasaki, Appl. Phys. Lett. 88, 152106 (2006) 107. R.D. Vispute, V. Talyansky, S. Choopun, R.P. Sharma, T. Venkatesan, M. He, X. Tang, J.B. Halpern, M.G. Spencer, Y.X. Li, L.G. Salamanca-Riba, A.A. Iliadis, K.A. Jones, Appl. Phys. Lett. 73, 348 (1998) 108. E. Bellingeri, D. Marre, I. Pallecchi, L. Pellegrino, A.S. Siri, Appl. Phys. Lett. 86, 012109 (2005) 109. E. Bellingeri, D. Marre, I. Pallecchi, L. Pellegrino, G. Canu, A.S. Siri, Thin Solid Films 486, 186 (2005) 110. M. Karger, M. Schilling, Phys. Rev. B 71, 075304 (2005) 111. E. Vasco, C. Zaldo, L. Vazquez, J. Phys.: Condens. Mat. 13, L663 (2001) 112. S. Choopun, R.D. Vispute, W. Noch, A. Balsamo, R.P. Sharma, T. Venkatesan, A. Iliadis, D.C. Look, Appl. Phys. Lett. 75, 3947 (1999) 113. X.W. Sun, H.S. Kwok, J. Appl. Phys. 86, 408 (1999) 114. A. Ohtomo, H. Kimura, K. Saito, T. Makino, Y. Segawa, H. Koinuma, M. Kawasaki, J. Cryst. Growth 214–215, 284 (2000) 115. J.F. Muth, R.M. Kolbas, A.K. Sharma, J. Appl. Phys. 85, 7884 (1999) 116. V. Srikant, D.R. Clarke, J. Appl. Phys. 81, 6357 (1997) 117. H. Matsui, H. Saeki, T. Kawai, A. Sasaki, M. Yoshimoto, M. Tsubaki, H. Tabata, J. Vac. Sci. Technol. B 22, 2454 (2004) 118. S. Choopun, R.D. Vispute, W. Yang, R.P. Sharma, T. Venkatesan, Appl. Phys. Lett. 80, 1529 (2002) 119. R.D. Vispute, S.S. Hullavarad, D.E. Pugel, V.N. Kulkarni, S. Dhar, I. Takeuchi, T. Venkatesan, Wide band gap ZnO and MgZnO heterostructures for future optoelectronic devices. In: Thin Films and Heterostructures for Oxide Electronics, ed. by S.B. Ogale (Springer, Berlin Heidelberg New York 2005) pp 301–330 120. M. Venkatesan, C.B. Fitzgerald, J.G. Lunney, J.M.D. Coey, Phys. Rev. Lett. 93, 177206 (2004) 121. A.K. Pradhan, K. Zhang, S. Mohanty, J.B. Dadson, D. Hunter, J. Zhang, D.J. Sellmyer, U.N. Roy, Y. Cui, A. Burger, S. Mathews, B. Joseph, B.R. Sekhar, B.K. Roul, Appl. Phys. Lett. 86, 152511 (2005) 122. D.C. Kundaliya, S.B. Ogale, S.E. Loﬂand, S. Dhar, C.J. Metting, S.R. Shinde, Z. Ma, B. Varughese, K.V. Ramanujachary, L. Salamanca-Riba, T. Venkatesan, Nat. Mater. 3, 709 (2004) 123. S.J. Pearton, W.H. Heo, M. Ivill, D.P. Norton, T. Steiner, Semicond. Sci. Technol. 19, R59 (2004) 124. J. Perriere, E. Millon, W. Seiler, C. Boulmer-Leborgne, V. Cracium, O. Albert, J.C. Loulergue, J. Etchepare, J. Appl. Phys. 91, 690 (2002) 125. M. Okoshi, K. Higashikawa, M. Hanabusa, Jpn. J. Appl. Phys. 40, 1287 (2001) 126. H.M. Christen, S.D. Silliman, K.S. Harshavardhan, Rev. Sci. Instrum. 72, 2673 (2001); & ORNL-CMSD progress report ORNL-6969, 76 (2002) 127. X. Zhu, H.L.-W. Chan, C.-L. Choy, K.H. Wong, D. Hesse, J. Appl. Phys. 97, 093503 (2005) 128. T. Venkatesan, K.S. Harshavardhan, M. Strikovski, J. Kim, Recent advances in the deposition of multi-component oxide ﬁlms by pulsed energy deposition. In: Thin Films and Heterostructures for Oxide Electronics, ed. by S.B. Ogale (Springer, Berlin Heidelberg New York 2005) pp 385–413 129. I. Takeuchi, Combinatorial synthesis of functional materials. In: Thin Films and Heterostructures for Oxide Electronics, ed. by S.B. Ogale (Springer, Berlin Heidelberg New York 2005) pp 333–352

7 PLD of ZnO Films

357

130. Z. Jin, T. Fukumura, M. Kawasaki, K. Ando, H. Saito, T. Sekiguchi, Y.Z. Yoo, M. Mukarami, Y. Matsumoto, T. Hasegawa, H. Koinuma, Appl. Phys. Lett. 78, 3824 (2001) 131. Z.W. Jin, Y.Z. Yoo, T. Sekiguchi, T. Chikyow, H. Ofuchi, H. Fujioka, M. Oshima, H. Koinuma, Appl. Phys. Lett. 83, 39 (2003) 132. A. Tsukazaki, A. Ohtomo, T. Kita, Y. Ohno, H. Ohno, M. Kawasaki, Science 315, 1388 (2007) & K. Ueda, H. Tabata, T. Kawai, Science 280, 1064 (1998) 133. M. Siegert, W. Zander, J. Lisoni, J. Schubert, C. Buchal, Appl. Phys. A 69, S 779 (1999) 134. P. Mukherjee, S. Chen, J.B. Cuﬀ, P. Sakthivel, S. Witanachchi, J. Appl. Phys. 91, 1828 (2002) 135. M. Strikovski, K.S. Harshavardhan, Appl. Phys. Lett. 82, 853 (2003) 136. A.M. Morales, C.M. Lieber, Science 279, 208 (1998) 137. Y.H. Tang, Y.F. Zhang, N. Wang, I. Bello, C.S. Lee, S.T. Lee, J. Appl. Phys. 85, 7981 (1999) 138. M. Kawakami, A.B. Hartando, Y. Nakata, T. Okada, Jpn. J. Appl. Phys. 42, L33 (2003) 139. M. Yan, H.T. Zhang, E.J. Widjaja, R.P.H. Chang, J. Appl. Phys. 94, 5240 (2003) 140. M. Lorenz, E.M. Kaidashev, A. Rahm, Th. Nobis, J. Lenzner, G. Wagner, D. Spemann, H. Hochmuth, M. Grundmann, Appl. Phys. Lett. 86, 143113 (2005) 141. A. Rahm, M. Lorenz, T. Nobis, G. Zimmermann, M. Grundmann, B. Fuhrmann, F. Syrowatka, Appl. Phys. A 88, 31 (2007) 142. A. Rahm, T. Nobis, M. Lorenz, G. Zimmermann, N. Boukos, A. Travlos, M. Grundmann, Adv. Solid State Phys. 46, in press (2007)

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells J. H¨ upkes, J. M¨ uller, and B. Rech

This chapter provides an overview of the physical principles and the application of zinc oxide in thin ﬁlm silicon solar cells. Focus will be on the method of magnetron sputtering followed by a wet-chemical etching step to achieve the required surface roughness of the ﬁlms. We will start with some basics of thin ﬁlm silicon solar cell design and operation, including the speciﬁc requirements that have to be met by the applied TCO front contact. Most emphasis will be given to the optical and light scattering properties. This is followed by a detailed discussion of the etching behavior of sputter deposited polycrystalline ZnO:Al ﬁlms. Finally, we will give an overview of state-of-the art thin ﬁlm silicon solar cells and modules using ZnO as transparent front electrode. The chapter ends with a summary, some concluding remarks, and an outlook on future developments.

8.1 Introduction For further development and world-wide market growth of photovoltaic (PV) power generation, a cost reduction of the PV systems is one of the major issues. Among the approaches followed in thin ﬁlm photovoltaics, solar cells based on hydrogenated amorphous (a-Si:H) or microcrystalline (µc-Si:H) silicon oﬀer speciﬁc advantages (see references in Zeman et al., Rech et al. and Rath [1–3]): Thin ﬁlm silicon solar cells are fabricated from extremely abundant raw materials and involve almost no ecological risk during manufacturing, operation, and disposal. Low process temperatures facilitate the use of a variety of low-cost substrate materials such as ﬂoat glass and metal or plastic foils [4–7]. The moderate energy demand for the whole module fabrication process leads to short energy payback times of the order of one year [8, 9]. Earlier and recent studies indicate a substantial cost reduction potential [10, 11] compared to the conventional crystalline silicon wafer technology. The low material costs and proven manufacturability of a-Si:H solar modules make them ideally suited for low-cost application on large scale. An integral part of these devices are the transparent conductive oxide (TCO) layers used as a front electrode and as part of the back side reﬂector. When applied at the front side, the TCO has to posses a high transparency in the spectral region where the solar cell is operating, a high electrical conductivity, and a surface structure which leads to suﬃcient scattering of the incoming

360

J. H¨ upkes et al.

light into the silicon absorber layer. The scattering at interfaces between neighboring layers with diﬀerent refractive indices and subsequent trapping of the incident light within the silicon absorber is crucial for all thin ﬁlm silicon solar cells to gain high eﬃciency. For amorphous or microcrystalline cells light scattering is usually achieved by nanotexturing the front TCO electrodes (with a typical root mean square surface roughness of 40–150 nm) and/or nanotextured back reﬂectors. In the ideal case, these rough layers can introduce nearly completely diﬀuse transmission or reﬂection of light, leading to very eﬀective light trapping. TCO is also used between silicon and the metallic contact as a part of the back reﬂector to improve its optical properties and to act as a diﬀusion barrier [12, 13]. Furthermore, applied in a-Si:H/µc-Si:H “micromorph” tandem solar cells, TCO can be used as an intermediate reﬂector between top and bottom cell to increase the current in the thin amorphous silicon top cell [14–17]. As a result the thickness of the a-Si:H top cell can be decreased to improve the stability of the cell against light induced degradation [18]. Finally, nano-rough TCO front contacts act as an eﬃcient anti-reﬂection coating due to the refractive index grading at the rough TCO/Si interface. In summary, TCOs play an important role in thin ﬁlm silicon solar cell structure and have a decisive inﬂuence on the eﬃciencies presently achievable in state-of-the art amorphous, microcrystalline or micromorph solar cells. Fluorine doped tin oxide (SnO2 :F) layers prepared by chemical vapor deposition (CVD) are commonly used as front contact. Tin oxide coated glass can be purchased on large area from glass coating companies and are usually prepared by a comparatively cheap in-line process. However, because of process restraints and low production volume, the quality of SnO2 :F ﬁlms available for industrial PV production is not ideal, leading to substantial eﬃciency losses. Therefore, zinc oxide (ZnO) has recently received growing attention, as this material can be produced with superior electrical conductivity and optical transparency [19]. Even more importantly, ZnO oﬀers ways to optimize surface texture and hence light scattering, which has led to substantially higher solar cell eﬃciencies on ZnO-coated glass than on commercially available SnO2 :F. Finally, ZnO shows a higher stability against hydrogencontaining plasmas [20], as commonly used for the deposition of Si-layers by plasma enhanced chemical vapor deposition (PECVD) [21, 22]. Among the diﬀerent preparation techniques for ZnO, the most widely used method is magnetron sputter deposition from ceramic ZnO or metallic Zn targets doped with Al (or other elements like Ga or In). The speciﬁc advantages and features of magnetron sputtering for ZnO deposition are explained in detail in Chap. 5 of this book. Other techniques include lowpressure chemical vapor deposition (LPCVD, see Chap. 6) or pulsed laser deposition (Chap. 7). Similarly to the sputtering process the LPCVD process is applicable for large areas and is used for preparation of boron doped zinc oxide (ZnO:B) for silicon thin ﬁlm solar module application.

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

361

8.2 Silicon Thin Film Solar Cells Although ZnO has also been applied in so-called amorphous/crystalline heterojunction solar cells consisting of a (doped) silicon wafer and thin doped a-Si:H layers to build the p–n junction, we will restrict ourselves here to solar cells and modules with amorphous and/or microcrystalline absorber layers, i.e., “real” thin ﬁlm silicon solar cells. For detailed information on the use of ZnO in crystalline silicon wafer based devices, the reader is referred to the literature (see e.g. [23, 24]). 8.2.1 Principle of Operation Silicon based thin ﬁlm solar cells have a p–i–n or n–i–p diode structure depending on the deposition sequence of doped and intrinsic layers (see also Sect. 8.2.3 and Fig. 8.4). For both structures the light enters through the p-layer, which eﬃciently supports hole collection in the device. The main reason for this is the smaller mobility of holes compared to electrons in amorphous silicon. A transparent conductive oxide (TCO) ﬁlm contacts the silicon diode from the front side and a metal ﬁlm or TCO/metal dual layer stack serves both as rear electrical contact and back reﬂector. This is illustrated in Fig. 8.1 together with a schematic band structure of the cell. As mentioned before, the interface between the individual layers are usually rough due to the necessity of light scattering. The very thin (10–30 nm) p- and ndoped layers build up an electric ﬁeld extending over the intrinsic (i) layer.

T C O potential (eV)

p

i

n

back reflector

CB electrons

EF VB holes

space coordinate

Fig. 8.1. Schematic energy band diagram of silicon thin ﬁlm solar cells. The symbols denote the Fermi level EF , valence band (VB), and conduction band (CB). Note that details of the band behavior at the various interfaces are not included. The light enters from the p-doped side

362

J. H¨ upkes et al.

Electrons and holes generated in the i-layer by incident light are driven to the n- and p-layer by the internal electric ﬁeld, respectively. The material quality of the intrinsic layer and the strength and distribution of the electric ﬁeld are responsible for the charge carrier collection and mainly determine the electrical solar cell performance. Defects aﬀect the charge carrier collection in two diﬀerent ways: On the one hand they act as recombination centers, and on the other hand their charge state modiﬁes the electric ﬁeld distribution in the i-layer. Figure 8.2 shows a J/V -curve of typical solar cells. The diode like characteristics are shifted towards negative current by the value of the photo current density JPH . In ﬁrst order approximation JPH is independent of the external voltage. The ﬁgure also includes the deﬁnitions of the so-called J/V parameters of the illuminated device: Short circuit current density JSC gives the current density at zero voltage and nearly equals the photo current density JPH . Open circuit voltage VOC describes the maximum voltage the device can theoretically operate at. The maximum power density that can be extracted from the solar cell is equal to the product of current density JMPP and voltage VMPP at the working point of the J/V -curve, the so-called maximum power point (MPP). From this, one deﬁnes a parameter called ﬁll factor (FF) as the ratio between maximum power density and the product of JSC and VOC FF =

JMPP × VMPP . JSC × VOC

(8.1)

J

VMPP

VOC

V

JMPP JSC

η =

FF × VOC × JSC Plight MPP

JPh

Fig. 8.2. J/V -curve with description of solar cell parameters: eﬃciency η, ﬁll factor F F , open circuit voltage VOC , and short circuit current density JSC . The photogenerated current density JPH is indicated by the dotted line. More details of these and all other parameters are given in the text

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

363

The more “rectangular” the J/V -curve appears, the higher the FF is and the better charge carriers are collected within the device. Finally, the conversion eﬃciency is deﬁned as the ratio between electrical power density delivered by the solar cell under standard illumination conditions and the power density of the incident light Plight η=

FF × VOC × JSC . Plight

(8.2)

The eﬃciency η is the most important parameter as it deﬁnes the overall quality of the device. 8.2.2 Amorphous and Microcrystalline Silicon Hydrogenated amorphous silicon (a-Si:H) diﬀers from crystalline silicon by the lack of long-range order and the high content of bonded hydrogen (typically around 10 at% in device-quality a-Si:H). Hydrogen atoms passivate most of the unsaturated silicon dangling bonds and make the material useful for (opto-)electronic devices. The disorder relaxes the momentum conservation rule, thus leading to high optical absorption coeﬃcients for photon energies exceeding the bandgap energy. However, disorder causes a high density of localized states (band-tail and defect states) within the energy gap impeding the carrier transport and enhancing recombination losses. The electrical properties of a-Si:H deteriorate under illumination (Staebler–Wronski eﬀect [18]), since during light soaking additional defects are created within the band gap acting as recombination centers. The photoconductivity of a-Si:H ﬁlms and the eﬃciency of solar cells decrease during light soaking until a certain saturation value is reached. A comprehensive review of the properties, preparation techniques, and applications of a-Si:H can be found for example in Street’s books [25,26], details on the transport properties in Overhof and Thomas [27]. The commonly used plasma-enhanced chemical vapor deposition (PECVD) is compatible with large area deposition and low process temperatures. This allows the use of a large variety of inexpensive substrate materials. By adjusting the deposition conditions in the right way, the material will start to develop crystalline regions within the amorphous matrix and can even be tuned to consist of almost 100 % crystalline volume fraction only separated by grain boundary regions. Once the ﬁlm contains such crystalline regions, it is usually called hydrogenated “microcrystalline” silicon (µc-Si:H), although nano- or polycrystalline silicon have also been used in literature to denote basically the same kind of material. In fact, it has turned out that for use as absorber material in photovoltaic applications, the most suited material lies close to the transition region from the amorphous to the microcrystalline phase [21,22,28]. One advantage of µc-Si:H compared to a-Si:H is the stability against light induced degradation [29]. However, newer results revealed light induced degradation of µc-Si:H solar cells, which can be suppressed to a

364

J. H¨ upkes et al.

large extent by high crystalline volume fractions [30, 31]. Another advantage of µc-Si:H is the extension of the wavelength range of absorption to the near infrared (NIR) region up to about 1 100 nm. The latter property is owed to the fact that the optical bandgap of µc-Si:H is very similar to crystalline silicon, although there are diﬀerences in the details of the absorption coeﬃcient (see Fig. 8.3). However, this also means that the bandgap is indirect and hence the absorption coeﬃcient for NIR light above 800 nm wavelength is rather low. Hence, an i-layer thickness of more than 1 µm and eﬃcient light trapping (see Sect. 8.2.4.1) are required for suﬃcient absorption. Another drawback of µc-Si:H PECVD is the high hydrogen dilution of the SiH4 process gas, which may cause chemical reduction of underlying layers and modify or damage the interface. Both a-Si:H and µc-Si:H can be doped by adding phosphorus or boron containing gases such as phosphine PH3 or trimethylboron B(CH3 )3 during the deposition process, so that p–i–n homojunctions and thus photovoltaic devices become possible. The optical absorption coeﬃcients α of typical a-Si:H, µc-Si:H, and crystalline silicon are shown in Fig. 8.3 as function of wavelength and photon energy. As a-Si:H ﬁlms possess a high optical absorption coeﬃcient over almost the whole visible sunlight spectrum, ﬁlms with a typical thickness of several 100 nm are suﬃcient as solar absorber layers. The optical bandgap of standard a-Si:H is typically around 1.7–1.8 eV, but can be tuned in a certain range by the deposition parameters and the hydrogen content in the material or, to a larger extend, by additions of carbon, oxygen, or germanium. Microcrystalline material, on the other hand, possesses a bandgap of

Fig. 8.3. Absorption coeﬃcient α of a-Si:H (dashed ), µc-Si:H (solid ), and crystalline silicon (dotted ) [32]. The x-axes represent the wavelength (bottom) and photon energy (top), respectively, of the incident light

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

365

around 1.1 eV. As seen in Fig. 8.3 this means that above about 800 nm a-Si:H is almost transparent, while light of a wavelength up to 1 100 nm can still be absorbed by µc-Si:H allowing a better utilization of the solar spectrum. 8.2.3 Design Aspects of Silicon Thin Film Solar Cells Here we describe the layer structure for single junction as well as for tandem solar cells consisting of a-Si:H and µc-Si:H. Further, this section will deal with the stability of silicon thin ﬁlm solar cells and the possibility to reduce degradation by special design. 8.2.3.1 Layer Structure and Deposition Sequence The doped and intrinsic silicon layers (p, i, n) are packed between a TCO front contact and a highly reﬂective back contact. The back contact is usually either a metal like silver (Ag) or aluminum (Al), or a TCO/metal double layer structure. The latter has been shown to reduce absorption losses due to a better grain growth of Ag layers onto ZnO. Additionally, absorption losses due to surface plasmons in the metal ﬁlm have to be considered [33]. Both eﬀects result in a higher reﬂectivity of the TCO/Ag back reﬂector. In module production, magnetron sputtered ZnO is usually applied as TCO-material for the back reﬂector in combination with either Ag (highest reﬂectivity) or Al (low cost). Depending on the deposition sequence of the doped and intrinsic silicon layers, one speaks of so-called superstrate (p–i–n) or substrate (n–i–p) cell structure (see Fig. 8.4). In the p–i–n (superstrate) conﬁguration, the light enters through the transparent, TCO coated superstrate. Glass is most commonly used as transparent front material. The p–i–n structure is fully compatible with laser scribing techniques to form a monolithically integrated series connection of

Incident light

Tandem cell

Incident light Superstrate TCO p i n TCO Metal

silicon

Substrate

deposition sequence

silicon

TCO p i n TCO Metal

TCO p i n p

a-Si:H

i

µc-Si:H

n TCO Metal

Fig. 8.4. Layer structure of single junction n–i–p (substrate) and p–i–n (superstrate) solar cells. Also included is an amorphous/microcrystalline tandem solar cell structure

366

J. H¨ upkes et al.

individual cells to an eﬃcient thin ﬁlm module (see Sect. 8.2.5). The use of glass substrates already provides an eﬀective encapsulation of the cell from the front side; the encapsulation materials “behind” the cell do not have to be transparent. As the front TCO is the ﬁrst deposited layer, one is almost free in the choice of TCO deposition parameters and even additional surface modiﬁcations (like wet-chemical etching, see Sect. 8.3) can be applied before starting with the silicon cell itself. This choice is limited in the n–i–p (substrate) cells, as the front TCO layer is deposited as the last process step and hence is subject to certain restrictions (like limited deposition temperature) to avoid any damage to the underlying cell structure. Manufacturing advantages of the n–i–p conﬁguration, on the other hand, are the possibility to use a variety of cheap transparent or nontransparent substrates (for example, stainless steel, plastic or metal foils, aluminum sheets, glass) including ﬂexible ones, and hence the application of roll-to-roll processing, which promises low production costs [5, 34]. Recently, a novel approach was presented combining some advantages of superstrate technology with roll-to-roll processing. Here, TCO and p–i–n cell processing are performed on a temporary ﬂexible superstrate. The complete solar cells can be laminated onto a deliberately chosen substrate, while the temporary superstrate is etched oﬀ [6, 35, 36]. In this chapter, we will focus on the application of sputter deposited ZnO:Al ﬁlms in silicon thin ﬁlm solar cells in superstrate conﬁguration on glass substrates. The stacked cell concept has emerged as a powerful tool to enhance the stability against light induced degradation of a-Si:H [37, 38]. Moreover, the use of component cells with diﬀerent optical band gaps reduces thermalization losses and provides a better utilization of the solar spectrum [39, 40]. While production facilities world-wide still mostly produce stacked cells consisting of amorphous silicon and its alloys, ﬁrst commercial production of tandem cells utilizing amorphous silicon for the top and microcrystalline silicon for the bottom cell absorber layers have been announced or already started [41–43]. The technology is being developed world-wide with high priority and includes work by equipment manufacturers to develop dedicated coating technologies and systems. Recent progress has been achieved in this ﬁeld by several companies [44–47]. There are several reasons why the a-Si:H/µc-Si:H stacked cell concept is considered very promising for the next generation of silicon thin ﬁlm solar cells: Stacked solar cells combine the high VOC and high absorption coeﬃcient in the visible region of a-Si:H with the utilization of the near infrared light (NIR) by µc-Si:H. Furthermore, µc-Si:H has been shown to be much more stable under light illumination [29–31], and hence only a small eﬃciency decrease is expected in the µc-Si:H cell. Figure 8.5 shows the quantum eﬃciency (QE) of an a-Si:H/µc-Si:H-tandem cell as a function of wavelength of the incident light. The QE gives the fraction of incident photons, which contribute to the photo current generation. The individual quantum eﬃciency curves of

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

367

Fig. 8.5. Quantum eﬃciency of an amorphous/microcrystalline silicon tandem junction solar cell. The individual quantum eﬃciency curves of the two component cells (a-Si:H top (dotted ) and µc-Si:H bottom (dashed )) are also included

the two component cells are also included. The ﬁgure nicely illustrates how a-Si:H-cells are only active in the visible range, while µc-Si:H can also absorb NIR radiation due to its lower band gap. 8.2.3.2 Long Term Stability of Silicon Thin Film Solar Cells In this paragraph we will brieﬂy discuss the stability behavior of silicon thin ﬁlm solar cells. Two aspects were considered: light induced degradation and stability in hot and humid environment. As mentioned in Sect. 8.2.2, a-Si:H degrades under light exposure due to the Staebler–Wronski eﬀect and solar cells reach stable operation after about 1 000 h of AM1.5 illumination. In Fig. 8.6 light soaking behavior of diﬀerent silicon thin ﬁlm modules with an aperture area of 8 × 8 cm2 is compared [45, 48]. An a-Si:H/a-Si:H tandem module was cut from an industrially produced large area (0.6 m2 ) module. Its stabilized aperture area eﬃciency is slightly above 6.5 %, which is the speciﬁed eﬃciency for this module type. The µcSi:H single junction solar module is nearly stable and its eﬃciency stabilizes at 8.2 %. The a-Si:H/µc-Si:H tandem module exhibits a degradation of less than 10 % relative. Each tandem cell design must be optimized with respect to the stabilized cell eﬃciency, i.e., layer thicknesses of top and bottom cell have to be adjusted accordingly. Their thickness values aﬀect for example the current matching between top and bottom cell. The inﬂuence of the current matching of a-Si:H/µc-Si:H tandem solar cells on the initial and stabilized cell performance was discussed by Repmann et al. [49]. The comparison demonstrates the potential of the a-Si:H/µc-Si:H tandem cell technology on textured ZnO coated glass substrates for highly eﬃcient large area thin ﬁlm solar modules. However, the transfer of the technology

368

J. H¨ upkes et al.

Fig. 8.6. Aperture area eﬃciency of diﬀerent silicon thin ﬁlm solar modules (aperture area 8 × 8 cm2 ) as function of light exposure time. The a-Si:H/µc-Si:H tandem and µc-Si:H single junction solar modules on texture etched ZnO:Al were prepared at the FZJ, the a-Si:H/a-Si:H tandem solar module was cut from an industrially produced (0.6 m2 ) module

to a cost-eﬀective module production still requires strong eﬀorts in research and technology development. Reliable stability data of the p–i–n solar cell itself are not easily obtained, especially for non-encapsulated cells or modules. One of these tests e.g. for EN/IEC 61646 certiﬁcation of modules is the so-called damp-heat test (85◦ C, 85 % humidity, up to 1 000 h). Recent studies were performed by Stiebig et al. [50, 51] exposing diﬀerent types of cells to harsh conditions. One of the most important results was the excellent stability of silicon thin ﬁlm solar cells. Remarkably, this is also valid for small area modules even without encapsulation [52]. This is of high interest because costs and eﬀorts for module encapsulation strongly depend on the inherent stability of the solar cells. As a more detailed treatment of this subject is beyond the scope of this chapter, the reader is referred to the original papers [50, 51]. 8.2.4 Requirements for TCO Contacts As already mentioned in the introduction, TCO ﬁlms are essential for most solar cell concepts. For thin ﬁlm silicon solar cells and modules, the TCO front contacts have to meet a number of requirements. The TCO must be highly transparent in the active range of the absorber layer (400–1100 nm in case of µc-Si:H) and also highly conductive to avoid ohmic losses. In the p–i–n structure, the front contact additionally has to possess a rough surface to provide eﬃcient light scattering. Furthermore, the TCO must feature favorable physico-chemical properties for the growth of silicon. For example, the TCO has to be inert to hydrogen-rich plasmas, which appear during deposition of silicon by CVD processes, and act as a good nucleation layer

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

369

for the growth of silicon. In general, the TCO/p-contact plays an important role for cell performance [53–56]. 8.2.4.1 Optical Properties and Light Trapping For all thin ﬁlm silicon solar cells, scattering at interfaces between neighboring layers with diﬀerent refractive indices and subsequent trapping of the incident light within the silicon absorber layers is crucial to gain high eﬃciency. The reason is the low absorption coeﬃcient α of amorphous and microcrystalline silicon, which is shown in Fig. 8.3 for typical device grade material. Because of the low band gap of µc-Si:H, light can be absorbed up to the near infrared spectral region allowing a better utilization of the solar spectrum. However, the band gap of crystalline silicon is indirect thus limiting the absorption coeﬃcient. As a result, in a µc-Si:H thin ﬁlm of no more than a few microns thickness, incoming light will not be completely absorbed during one single pass. On the other hand, to minimize process time and reduce lightinduced degradation in case of a-Si:H, the absorber layer should be as thin as possible. Quantitatively, this is shown by the total absorption A of an a-Si:H and µc-Si:H layer of typical device thickness (da−Si:H = 300 nm and dµc−Si:H = 1 µm, respectively) in Fig. 8.7. The absorption was calculated for one single light pass using the absorption coeﬃcients α of Fig. 8.3 according to A = 1 − e−αd .

(8.3)

Fig. 8.7. Absorption of 300 nm a-Si:H (dashed ) 1 µm µc-Si:H (solid ) and of two high quality ZnO:Al front contacts with diﬀerent aluminum doping levels. The ZnO:Al ﬁlms were prepared by RF sputtering from ceramic targets with a thickness of 700 nm. Aluminum concentration CAl in the targets was 0.8 at. % (dotted ) and 1.6 at. % (dash-dotted ), respectively. Sheet resistances of the ZnO:Al ﬁlms were 11 and 6 Ω, respectively

370

J. H¨ upkes et al.

The low absorption illustrates the necessity of light trapping for long wavelengths. The maximum current that can be gained in these regions from solar radiation if ideal light trapping is realized is 12 and 27 mA cm−2 for a-Si:H and µc-Si:H solar cells, respectively [17]. Hence, optical absorption inside the silicon layers has to be enhanced by increasing the optical path of solar radiation. For a-Si:H and µc-Si:H solar cells, light scattering is usually achieved by nanotexturing the front TCO electrodes to a typical root mean square surface roughness δrms of 40–150 nm and/or nanotextured back reﬂectors. In the ideal case, these rough layers can introduce nearly completely diﬀuse transmission or reﬂection of light. The absorption spectra of two ZnO:Al front contacts are included in Fig. 8.7, which were calculated from reﬂection and transmission measurements. The ZnO:Al ﬁlms were prepared by sputter deposition from targets with diﬀerent aluminum concentration (CAl ), leading to diﬀerent doping levels in the ﬁlms. Note, that CAl is calculated as quotient [Al]/([Zn] + [Al]), which might diﬀer from the notiﬁcation of other groups, who also take the oxygen atoms into account. The absorption of the ZnO:Al ﬁlm with higher doping concentration (CAl = 1.6 at%) is a few percent in the visible region but increases strongly towards the NIR range. This is caused by free carrier absorption, when the frequency of the incident light approaches the plasma frequency of the free electron gas in the material [57]. The plasma frequency of typical TCO materials is in the NIR and shifts with increasing carrier density towards the visible range. For lower doping level (CAl = 0.8 at%), the absorption is signiﬁcantly reduced over the whole spectrum. Even this highly transparent front TCO still absorbs more light of long wavelengths than the silicon layers. This illustrates the necessity of low TCO absorption to improve carrier generation in the solar cells. The eﬀect of light trapping on solar cell performance is demonstrated by ZnO:Al ﬁlms with diﬀerent surface textures. Figure 8.8 shows scanning electron microscope (SEM) images of these ZnO:Al ﬁlms. Sputtered ZnO:Al is initially smooth (Fig. 8.8a). By a simple wet-chemical etching step in diluted hydrochloric acid (typically 0.5 % HCl) one can roughen the ZnO:Al surface which is shown after a short dip (Fig. 8.8b) and after an optimized etch

Fig. 8.8. SEM micrographs of RF sputter deposited ZnO:Al ﬁlms directly after deposition (a), after a short dip in HCl (b), and after optimized etch duration (c). Some surface properties are given in Table 8.1

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

371

process (Fig. 8.8c). RMS roughness, haze at 700 nm of the ZnO:Al ﬁlms and short circuit current density of the respective µc-Si:H solar cells are listed in Table 8.1. The haze is a measure of milkiness of the ﬁlms and was calculated as the ratio of diﬀuse and total transmission. Upon etching, roughness δrms and haze increase from 7 to 162 nm and from 1 % to more than 50 %, respectively. The spectral response curves of µc-Si:H p–i–n cells prepared on these smooth and diﬀerently texture etched ZnO:Al substrates are compared in Fig. 8.9. The ﬁgure shows quantum eﬃciency and total cell absorption (1 − R, R = reﬂectivity) as a function of wavelength. The cell prepared on the optimized texture etched ZnO:Al substrate exhibits an increase in shortcircuit current density by more than 7 mA cm−2 , due to an enhanced long wavelength absorption (see Table 8.1). In addition, the interference fringes characteristic of a ﬂat ﬁlm disappear due to destruction of phase correlation between the diﬀerent reﬂected beams by the rough interfaces. However, a second look at Fig. 8.9 and the discussion of Fig. 8.7 above also demonstrates an intrinsic problem of light trapping: The TCO absorbs much more light of Table 8.1. Surface properties of the etched ZnO:Al ﬁlms shown in Fig. 8.8. Root mean square roughness (δrms ), haze at a wavelength of 700 nm, and resulting short circuit current density, when applied in 1.1 µm thick µc-Si:H solar cells Etching No Short Optimized

δrms (nm)

Haze@700 nm (%)

JSC (mA cm−2 )

7 72 162

1 15 54

16.3 19.8 23.4

Fig. 8.9. Measured quantum eﬃciency (QE) and cell absorption (1 − R) of µc-Si:H solar cells on the RF sputtered ZnO:Al front contacts shown in Fig. 8.8: smooth (dashed ), after a short etching step (solid ), and after optimized etch duration (dotted ). Resulting short circuit current densities are 16.3, 19.8, and 23.4 mA cm−2 , respectively

372

J. H¨ upkes et al.

longer wavelengths during a single pass than the respective silicon absorber. This single pass absorption within the TCO is strongly enhanced by internal light trapping, as a certain part of the light is reﬂected back and forth between bottom and top contact or the glass surface passes through the front TCO layer many times and experiences absorption during every pass. For this reason, the NIR transmission of front TCO is very important to minimize optical losses, which leads to the requirement of a low charge carrier density and hence a low doping level. The exact calculation of TCO absorption losses is rather diﬃcult, as neither the angular dependence nor the exact amount of light scattering at the individual interfaces is known or can be easily measured. A number of approaches attempting to predict the performance of thin ﬁlm silicon solar cells by theoretical modeling have been reported in literature (see e.g. [58–67]). As an example we present here the results on optical absorption losses for solar cells as obtained by Springer et al. [58]. The plot in Fig. 8.10 shows calculated losses in the diﬀerent layers of the solar cell. Apart from optical absorption losses in n- and p-layers, which may be reduced by further optimization, the strong inﬂuence of TCO absorption (single diagonally hatched ) in the light trapping region is emphasized. For details of the calculations and input data used, please refer to the original paper [58]. 8.2.4.2 Electrical Properties From an electrical point of view it is obvious that the TCO sheet resistance Rsq (resistance of a square of a homogeneous ﬁlm of a certain thickness) of

Fig. 8.10. Measured (dashed lines) and calculated (solid ) quantum eﬃciency (QE) and total cell absorption (1 − R) according to Springer et al. [58]. The area between QE and (1 − R) corresponds to the absorption losses in 330 nm thick front ZnO:Al (single diagonally hatched ), doped silicon (p layer double diagonally hatched, n layer white), silver (black ), and glass + back ZnO (horizontally hatched ). The area above the silver absorption (black ) and the total cell absorption (dashed ) corresponds to cell reﬂection R. Reprinted with permission from [58]

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

373

a TCO layer should be as small as possible to minimize Ohmic losses in the TCO when it transports the photocurrent. To keep the ﬁlm thickness small, this leads to the requirement of low TCO resistivity ρ or high conductivity σ. The relation between conductivity, mobility µ, and carrier density n is described by the formula σ=

1 = neµ , ρ

(8.4)

where e is the elementary charge. This means that either n or µ or both should be high to achieve high conductivity. However, an increase in n is undesirable, since this will increase free carrier absorption losses, as discussed earlier. Therefore, the perfect TCO material should have a high conductivity achieved by high mobility, while keeping carrier density low. In practice, this can possibly be achieved by using low doping levels in the TCO material. The relationships between n and µ of zinc oxide are reviewed in Chap. 2, dealing with diﬀerent theories and experimental data. As there are, of course, limitations to the achievable n- and µ-values, we have to deal with an optimization task, where good electrical and good optical properties are contradicting requirements and therefore the optimum for best overall performance has to be found. This will become more quantitative in the next section on module aspects. Furthermore, another important requirement, the light scattering ability, will be discussed in great detail below. 8.2.5 Solar Module Aspects A necessary prerequisite for high module eﬃciencies is the uniform distribution of material and cell properties on the substrate to avoid power losses due to averaging eﬀects. The total area of solar modules is divided into cell strips, which are interconnected by a sequence of deposition and cutting steps. Figure 8.11 illustrates the structure for series-connection in thin ﬁlm silicon wa

wd

Fig. 8.11. Monolithic series-connection of silicon thin ﬁlm solar modules by a sequence of deposition and laser scribing steps. The current ﬂow is indicated by the dotted arrow. For calculation of resistive and optical losses the active cell width wa and dead area width wd are needed

374

J. H¨ upkes et al.

solar modules with arrows indicating the current ﬂow. After TCO deposition onto glass substrate, the front contact is cut into strips of typically 1 cm width. The groove prepared into the silicon ﬁlms on top of the TCO directly next to the TCO line allows for a connection between front and back contact, which is later deposited through the silicon layer. The ﬁnal step is the separation of the individual cells by scribing of the back contacts. As a result, the back contact of one cell is connected through the grooves in the silicon layers to the front contact of the adjacent cell. Typically, the separations are done by laser ablation, which is a fast and dry process. Other series-connection approaches have been presented for example by Ichikawa et al. [34] and Basore et al. [68]. The aperture area of a solar module is deﬁned by the inner boundary of total laser scribed area. For every kind of series-connection, the active solar cell area is reduced by the area of the interconnection structure. Area losses increase if the cell width is reduced. On the other hand, electrical losses in contact layers become more severe with increasing cell width due to increasing cell current. Hence, optimization of the cell width must consider both area losses (dead area) due to patterning and series resistance losses due to TCO sheet resistance. These power losses can be described by the loss factors f , which is the sum of the loss factors fd (area losses) and fTCO (resistive losses in front TCO) [69]: f = fd + fTCO wd JMPP Rsq wa3 = + , wa + wd VMPP 3 wa + wd with active cell width wa , dead area width wd , TCO sheet resistance Rsq , and current density JMPP and voltage VMPP at the maximum power point (MPP). The expected active area module eﬃciency ηact and aperture area eﬃciency ηap for a given light intensity Plight are calculated by JMPP VMPP (1 − fTCO ) Plight JMPP VMPP (1 − f ) = , Plight

ηact = ηap

and

respectively. These equations are used to predict module eﬃciencies from small area cell results and to optimize cell width. Figure 8.12 shows calculated module eﬃciencies using typical stabilized J/V -parameters at MPP of 1 cm2 a-Si:H/µc-Si:H test cells on texture etched ZnO:Al (η = 11%, JMPP = 10 mA cm−2 , VMPP = 1.1 V) for an assumed dead area width wd due to patterning lines of 300 µm. The highest calculated aperture area eﬃciencies are 10.5, 10.4, and 10.2 % for the front TCO sheet resistances Rsq of 5, 10, and 20 Ω, respectively. The corresponding optimized cell widths are about 10, 8, and 6 mm.

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

375

Fig. 8.12. Calculated active area (gray) and aperture area eﬃciencies (black ) of a-Si:H/µc-Si:H tandem solar modules as function of cell width for diﬀerent TCO sheet resistance values RTCO : 5 Ω (dashed ), 10 Ω (solid ), and 20 Ω (dotted )

With idealized assumptions inherent losses due to interconnection are between about 5 and 8 % depending on the front TCO conductivity and the width of the interconnection structure. Additional losses to consider in module production compared to laboratory type small area test cells appear due to production type processes, nonuniformity, and peripheral ﬁlm removal [8, 70]. 8.2.6 Approaches for Light Trapping Optimization In superstrate technology, the TCO ﬁlms must exhibit high transparency, high conductivity, and an adapted surface roughness to introduce light trapping. In addition, the surface texture should not (or only slightly) reduce the open circuit voltage and must not increase the proneness to shunts. Fluorine doped tin oxide ﬁlms (SnO2 :F) prepared oﬀ-line by atmospheric pressure chemical vapor deposition (APCVD) fulﬁl these requirements to a large extent. State-of-the-art SnO2 :F coated glass substrates have been developed by Asahi Glass Co. (Asahi Type U) [71]. Unfortunately, this optimized TCO substrate material is not yet available at low costs. A similar CVD process is also applied for the pilot stage SnO2 :F deposition on aluminum foils [6, 36] mentioned in Sect. 8.2.3. For commercial applications, SnO2 :F ﬁlms are pyrolytically prepared by APCVD during ﬂoat glass production online. The installed production capacities are already capable of providing low-cost substrates for thin ﬁlm silicon module production in the giga watt range [72, 73]. Unfortunately, opto-electronical properties of production type SnO2 :F ﬁlms still need further improvement for silicon thin ﬁlm solar cell application [19]. Recently, promising progress has been achieved

376

J. H¨ upkes et al.

for developments that directly aim large area silicon thin ﬁlm solar module application (see e.g. [74]). While there have been eﬀorts to reduce the long wavelength absorption in the 3–4 mm thick glass substrate by using glass of low iron content [75–77], the SnO2 :F surface structure appears to be not ideally suited for long wavelength light scattering as needed in microcrystalline solar cells [56]. One of the reasons is the feature size of the SnO2 :F which is too small for the wavelength range above 800 nm. In addition, transmission of SnO2 :F-ﬁlms decreases due to chemical reduction of SnO2 :F to metallic Sn in hydrogen containing plasmas as present during µc-Si:H CVD. This eﬀect makes uncoated SnO2 :F unsuited as substrate for microcrystalline silicon single junction solar cells. A promising alternative is surface textured doped zinc oxide ﬁlms. ZnO ﬁlms can oﬀer excellent transparency and are highly resistant to hydrogen plasmas [78]. Textured ZnO ﬁlms have been prepared by several deposition techniques. Examples are boron doped zinc oxide (ZnO:B) prepared by lowpressure chemical vapor deposition (LPCVD) ([79, 80], see also Chap. 6) or ZnO ﬁlms deposited by expanding thermal plasma CVD [81]. Quite recently, ZnO ﬁlms for back contacts of solar modules have been developed using chemical bath deposition [82]. Other examples for approaches to achieve light trapping are periodic grating couplers prepared by photolithography and etching [83] or surface textured glass substrates. The latter can be realized by glass etching [84] or special coatings [85]. A very promising deposition method from a technological point of view is the use of magnetron sputtering: Low substrate temperatures can be applied which are usually compatible with all preceding process steps (especially important for n–i–p cells, see Sect. 8.2.3) and allow the use of a large variety of inexpensive substrate materials. Sputtering is an in-house technology for almost every producer of a-Si:H solar modules, since it is the standard technique for rear contact preparation or, in case of n–i–p (substrate) technology, also for front TCO contacts. In general, sputter deposited ZnO:Al ﬁlms are very smooth even though there have been approaches for surface textured ZnO:Al by adding water vapor to the sputtering atmosphere or otherwise modiﬁed growth conditions [86–89]. Sputter deposited but smooth ZnO:Al ﬁlms oﬀer an attractive feature to design optimized light trapping concepts: Depending on their structural properties, the surface morphology of sputtered ZnO:Al ﬁlms can be modiﬁed by chemical etching (for example in diluted hydrochloric acid) [90]. The initial ﬁlm properties and the duration of the etching process control feature size and shape of the resulting surface texture [91–93]. By this, the light scattering properties of ZnO:Al ﬁlms can be tuned over a wide range, while still excellent conductivity and transparency are maintained. The combination of sputtering and etching, which has already been introduced in Sect. 8.2.4.1, opens exciting possibilities to design tailored light trapping schemes for diﬀerent absorber materials

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

377

and device structures [94]. This approach, using sputtering and etching of ZnO:Al ﬁlms, will be described in more detail in Sect. 8.3.

8.3 Sputter Deposition and Etching of ZnO:Al Preparation of ZnO:Al ﬁlms by sputter deposition usually leads to smooth ﬁlms (see also Chap. 5). After deposition, the ZnO:Al surface can be roughened by wet-chemical etching (see Sect. 8.2.4.1 and [90]). This technique oﬀers the unique possibility to separate conductivity and transparency of the ZnO:Al ﬁlms from their light scattering behavior and allows to produce light scattering TCO ﬁlms for both p–i–n and n–i–p cells [95, 96]. Focus of this section will be on the diﬀerent etching behavior of zinc oxide and the possibilities and limits to control the surface texture after etching by the proper choice of deposition and etching conditions. It has been shown that it is possible to achieve high electron mobility for sputtered ZnO:Al on glass and thus reduce infrared absorption while maintaining good electrical conductivity [97]. Typical mobility values of sputter deposited ZnO:Al ﬁlms are between 25 and 50 cm2 (V s)−1 . Signiﬁcantly higher mobilities have been reported for other TCO materials like Cd2 SnO4 (80 cm2 (V s)−1 , Eg = 3.1 eV [98]). However, a textured TCO material with these properties, which is also feasible for the use in thin ﬁlm silicon solar cells, still has to be developed. Although excellent ZnO:Al ﬁlms are achieved with radio frequency (RF) sputtering, low deposition rate, high equipment costs, and the comparably high material costs for ceramic targets have led to the development of alternative processes. If ceramic targets are to be used, the high investment costs and low throughput can be overcome by direct current (DC) sputtering, leading to distinctly higher deposition rates. Additionally, the development of cost-eﬀective target manufacturing methods reduce running material costs. Another approach is reactive sputter deposition from metallic Zn:Al alloy targets also featuring high growth rates and low target costs. In this case, the compound is formed during deposition from the metallic target material and small amounts of oxygen added to the inert sputter gas. In addition to the important parameters substrate temperature and sputter pressure, the reactive process adds another inﬂuencing factor: the working point of the process. The working point is characterized by a certain discharge voltage, oxygen partial pressure, and plasma emission intensity of zinc at a certain reactive gas ﬂow. A too high oxygen (reactive gas) partial pressure means deposition in the so-called oxide or poisoned mode leading to highly transparent but nonconductive ﬁlms at low deposition rate. Too little oxygen fraction, on the other hand, leads to sputtering in metallic mode producing conductive but optically absorbing ﬁlms. It has turned out that ﬁlms with both high transparency and high conductivity have to be prepared in the so-called transition mode in between these two regimes. Special process control techniques have

378

J. H¨ upkes et al.

to be employed to stabilize the process in this transition region. The sputter technology and resulting ZnO:Al material properties are described in more detail in Chap. 5 of this book and with focus on application as front contact in silicon thin ﬁlm solar cells in H¨ upkes et al. [99]. 8.3.1 Properties of Sputter Deposited ZnO:Al Some important deposition parameters to control ZnO:Al ﬁlm properties are deposition pressure, substrate temperature, and amount of oxygen in the sputter gas mixture (see e.g. [93, 96, 100, 101]). Doping concentration in ZnO ﬁlms plays another important role for opto-electronic properties [97,100,102]. Figure 8.13 shows the spectral transmission of ﬁlms that were reactively sputtered from targets with diﬀerent aluminum concentration at optimized conditions [93]. Sputter conditions and ﬁlm properties are given in Table 8.2.

Fig. 8.13. Spectral transmission of diﬀerently doped ZnO:Al ﬁlms. The ﬁlms were prepared in an in-line system by reactive MF sputtering. The subscripts denote the aluminum concentration of the corresponding metallic targets in at. %. Deposition conditions and ﬁlm properties are given in Table 8.2

Table 8.2. Deposition pressure pdep , substrate temperature TS , ﬁlm thickness d, aluminum concentration in the ﬁlms CAl , resistivity ρ, carrier concentration n and mobility µ of the ZnO:Al ﬁlms given in Fig. 8.13. The subscripts denote the aluminum concentration of the corresponding metallic targets in at. % Film

pdep (Pa)

TS (◦ C)

d (nm)

CAl (at. %)

ρ (10−4 Ω cm)

n (1020 cm−3 )

µ (cm2 (V s)−1 )

Al0.5 Al1.2 Al2.4 Al4.7

0.6 0.6 1.2 0.6

360 320 260 240

620 870 896 866

2.3 2.7 3.8 6

3.3 2.8 2.6 5.2

4.4 5.3 8.0 9.0

42 41 30 13

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

379

The ﬁlms were prepared by reactive mid frequency (MF) sputtering on Corning 1737 glass [103] in an in-line system. Aluminum concentration CAl in the ﬁlms was determined by secondary ion mass spectroscopy (SIMS). All values for Al concentrations are given as ratio [Al]/([Zn] + [Al]) (in atomic % or at.%), where [Al] and [Zn] are the number of metal atoms in the target or ﬁlm per unit volume, i.e., the respective density. Note that this terminology diﬀers from the commonly used wt% of Al or Al2 O3 given by target manufacturers. Other groups, who also use atomic percent to characterize their material include the oxygen content for ceramic targets or oxide ﬁlms [101, 104]. Our terminology was chosen as we want to emphasize the inﬂuence of external doping which is determined by the Al-content relative to the number of Zn-atoms. Aluminum concentration of targets (at. %) is calculated from the nominal weight percent value CAl [wt] given by the target manufacturer using the formula CAl (at. %) = k × CAl (wt) . (8.5) The factor k is a function of molar masses of zinc, aluminum, and oxygen atoms and additionally slightly depends on CAl (wt). For our needs, k can be approximated by the constants kmetallic = 2.3–2.4 and kcompound = 1.5–1.6 for metallic and ceramic targets as well as for the compound ﬁlm, respectively. Carrier concentration and mobility were determined by van der Pauw Hall eﬀect measurements [105]. According to the Burstein–Moss eﬀect [106, 107], the optical band gap increases with carrier concentration n. This eﬀect can be observed at the short wavelength turn-on of transmission. Between 400 and 600 nm all ﬁlms show very high transmission of similar values above 82% in average. The diﬀerences above 600 nm can be attributed to free carrier absorption, resulting in lower transmission for highly doped ﬁlms [108]. Lower aluminum concentration in the targets and thus in corresponding ﬁlms signiﬁcantly improves the transmission in the red and near infrared (NIR) region. The very low resistivity of these ﬁlms even at the lower carrier concentration is maintained by an improved electron mobility (see Table 8.2). Note that the ﬁlm thickness of the ZnO:Al ﬁlm sputtered from the 0.5 at. % target is smaller than that for the ﬁlms with higher doping levels. While the improved NIR transmission for low doping also shows up in thicker ﬁlms (not included here), the smaller ﬁlm thickness was deliberately chosen, as we compare ZnO:Al ﬁlms relevant for solar cell front contacts. Because of the excellent electrical properties of the 0.5 at. % ﬁlm even at reduced thickness, such thin ﬁlms can be directly applied as high quality solar cell substrates making use of the additional optical gain due to the smaller thickness. 8.3.2 Etching Behavior of Zinc Oxide Both ZnO single crystals and polycrystalline ﬁlms are readily etched in many acidic [109–114] and alkaline solutions [93, 109, 115]. By chemical etching of

380

J. H¨ upkes et al.

ZnO:Al ﬁlms one can roughen the ﬁlm surface and thereby introduce a light scattering eﬀect which is of paramount importance for high eﬃciency silicon thin ﬁlm solar cells. In this section we describe the etching behavior of ZnO single crystals as well as of sputter deposited polycrystalline ZnO:Al ﬁlms and give some suggestions for an etching model.

8.3.2.1 Wet-Chemical Etching of ZnO Single Crystals First studies on the etching behavior of zinc oxide in diluted acids and lyes have been published several decades ago. Systematic experiments revealed a strong dependence of etching behavior on the crystal planes and etch solution [109, 110]. Figure 8.14 shows scanning electron microscopy (SEM) images of the zinc terminated (001) face (left) and oxygen terminated (001) face (right) of ZnO single crystals, which were etched in diluted hydrochloric acid (5% HCl at room temperature, top) or sodium hydroxide (10% NaOH at 50◦ C, bottom), respectively. During etching in acids the Zn-face develops hexagonal craters at positions of crystal defects, while the remaining surface stays smooth and probably unetched. The faces of these craters

Fig. 8.14. SEM micrographs of etched ZnO single crystals. The crystals were etched either in hydrochloric acid (top) or in sodium hydroxide solution (bottom). The graph shows the zinc terminated (001) (left) and the oxygen terminated (001) (right) surface after the respective etching step. The angles γ spanned by the two opposite edges and δ between the (101) faces of the etch crater are about 130◦ and 123◦ , respectively

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

381

belong to the (101) planes with an opening angle of δ = 123◦ as measured by atomic force microscopy. In contrast, the O-face is etched very rapidly in acids and becomes very rough due to the formation of hexagonal pyramids all over the surface. In alkaline solutions, craters are formed on the O-face similarly to the etching of Zn-faces in acids. On the Zn-terminated face no characteristic structure is developed in case of NaOH etching, although the model cited below [109] predicts that material may still be removed. During etching in acids the nonpolar (100) prism faces of ZnO exhibit triangular etch pits with the apex pointing towards the Zn face (not shown here) [109, 110]. The diﬀerent etching behavior of the polar faces was explained by Mariano and Hanneman [109] using a model for polar III-V semiconductors [116]. An illustration of this model is shown in Fig. 8.15. Both polar surfaces exhibit dangling bonds that are partially positively or negatively charged in case of Zn or O termination, respectively. These charged dangling bonds act as attractive centers for a fast reaction either with hydroxide (OH− ) or hydronium (H3 O+ ) ions. On the other hand, if the surface dangling bonds and the primary ions of the etching solution have the same charge polarity and hence repel each other, the etching mainly takes place at dislocations and is not masked by rapid etching of the whole surface. Similar observations have been made by Jo et al. [117] for sintered ZnO with large crystals of about 50 µm size after etching in an HCl/HF mixture. As three-dimensional shape of etch pits formed in ZnO crystals, they predict polyhedral cones composed of (101) and (001) planes. This model represents a fully three-dimensional extension of the earlier model by Mariano and Hanneman [109] for the special case of acidic etching.

OH–

δ+

001 (Zn-terminated)

surface bulk

Zn

bulk

O

surface

H3O+

δ–

001 (O-terminated)

Fig. 8.15. Model for etching mechanism of acidic and alkaline solutions on polar ZnO faces [109]. The dangling bonds at the surfaces exhibit partially positive (δ + for Zn) or partially negative (δ − for O) charge state. This leads to diﬀerent reactivity depending on the ion charge

382

J. H¨ upkes et al.

8.3.2.2 Wet-Chemical Etching of Sputter Deposited ZnO:Al Films There are many reports about diﬀerent applications for etching of ZnO ﬁlms like patterning during preparation of nanostructured devices [113, 114] and ﬂat panel displays (see for example Li et al. [112]). In this section, we focus on the surface texturing of magnetron sputtered ZnO:Al ﬁlms for the application as light scattering front contacts in thin ﬁlm silicon solar cells. The surface texturing of sputter deposited ZnO ﬁlms by post deposition etching has been presented by van de Pol et al. [111]. The application of such surface textured zinc oxide ﬁlms in silicon thin ﬁlm solar cells was ﬁrst experimentally realized and published by L¨ oﬄ et al. [90]. Depending on initial ﬁlm properties and etch parameters the ﬁlms develop distinctly diﬀerent surface morphologies during etching. Up to now, there is no theoretical model to predict the wet-chemical etching behavior of polycrystalline ZnO:Al ﬁlms as a function of deposition parameters or material properties. However, we can present empirical relationships resulting from many years of systematic studies, which allow surface morphology tuning of ZnO:Al ﬁlms for optimized light trapping in amorphous or microcrystalline silicon solar cells. First, we illustrate the evolution of typical surface structures during etching in a series with systematically varied etch time. Figure 8.16 shows SEM top views (i.e., perpendicular beam incidence) of ZnO:Al ﬁlms in the as deposited state (a) and after etching (b–f). Etching was done either in 0.5% HCl at room temperature (b–c) or in 33% potassium hydroxide (KOH) at 50◦ C (d–f) with increasing etch duration from left to right. The initial ﬁlms were nominally identical with similar thickness of 770 nm and were prepared on the same type of glass substrates under identical sputter conditions. The as deposited ﬁlm is rather smooth. Small roughness with lateral feature sizes

Fig. 8.16. SEM micrographs of ZnO:Al surfaces in the initial (as deposited) state (a), after etching in HCl for 5 s (b), and 10 s (c) or in potassium hydroxide solution KOH for 10 s (d), 30 s (e), and 100 s (f ), respectively

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

383

of less than 100 nm arises from the columnar structure of the ZnO:Al ﬁlms. Typical root mean square surface roughness δrms of as deposited ﬁlms is less than 15 nm as determined from AFM measurements. A short dip in hydrochloric acid leads to randomly distributed craters with diameters of about 500 nm, while – on a microscopic scale – parts of the ZnO:Al surface area are only slightly attacked and remain rather smooth. After prolonged etching, the crater formation proceeds until a textured surface with larger and smaller craters is obtained (Fig. 8.16c; see also Fig. 8.8). Although there is a certain distribution of crater depths and sizes, the overall surface appears homogeneously roughened on a larger scale. Upon HCl etching, the ﬁlm thickness of these ﬁlms was reduced at a rate of 5–6 nm s−1 . Typical etch rates are between 2 and 10 nm s−1 . Root mean square roughness increases approximately linearly with etch duration up to a certain limit, which is typically in the range of 100–150 nm. The ﬁnal roughness and distribution of craters strongly depends on initial ﬁlm properties. On the other hand, KOH solution attacks the surface mainly at a few, randomly distributed points, where deep and steep holes are formed, as illustrated in Figs. 8.16d–f. The surface between the etched holes appears to be hardly attacked by the KOH, although the overall ﬁlm thickness was reduced at a rate of about 1 nm s−1 . This thickness reduction also results in a slightly rougher surface structure after the longest etch time. The feature size of these morphologies is similar to the one of unetched ﬁlms and hence much smaller than the craters obtained upon HCl attack. In contrast to the HCl etching, the holes did not grow laterally with KOH as etchant, although this has been observed in another study [118]. From the results in Fig. 8.16 it is obvious that the attack of acidic solutions generally yields higher surface roughness than observed in case of alkaline etchants. This corresponds well to light scattering properties of the diﬀerent types of surface structures. Hence, acidic solutions are the preferred etchant to obtain light scattering surfaces and were used to obtain the highly eﬃcient solar cells and modules, which will be presented in Sect. 8.4. Figure 8.17 shows ZnO:Al cross section micrographs. The transmission electron microscopy (TEM) image in Fig. 8.17a represents a texture etched ZnO:Al ﬁlm (bottom) covered with µc-Si:H silicon. The SEM micrograph (Fig. 8.17b) shows cross-section and surface of a texture etched ZnO:Al ﬁlm. The columns of ZnO:Al grow with a diameter between 50 and 100 nm. They are clearly distinguishable from each other and extend from the substrate to the top of the ﬁlm. The c-axis is oriented perpendicular to the substrate. The craters cause a “wave-like” appearance of the interface region to silicon (left) and are also visible on the surface of the SEM image (right). The most important observation here is that the etched craters extend over a large number of columns. This means that, unlike the etching results obtained in alkaline solutions, the crater formation during acidic etching does not proceed along the grain boundaries of the polycrystalline material. The crater

384

J. H¨ upkes et al.

Fig. 8.17. (a) Cross-sectional TEM micrograph of an etched ZnO:Al ﬁlm covered with about 1 µm of a microcrystalline silicon p–i–n solar cell [119]; (b) SEM image of an etched ZnO:Al ﬁlm showing cross-section and etched surface. The crater edge on the ZnO surface is indicated by a black line

formation is hardly aﬀected by the grain boundaries between the columns. This eﬀect represents one of the diﬃculties in explaining the etching behavior by a microscopic model. However, based on the growth model for sputtered metals [120, 121], an empirical model was developed by Kluth et al. to describe systematic changes of ﬁlm properties as a function of the deposition conditions for ZnO:Al ﬁlms deposited by radio frequency sputtering from ceramic targets [96]. Figure 8.18 shows SEM micrographs of diﬀerent ZnO:Al ﬁlms, denoted type A, B, and C, before (left) and after etching (right) as well as a sketch of structural properties as a function of deposition pressure and substrate temperature. The zone 2 represents growth conditions with energetic particles leading to diﬀusion controlled growth and compact ﬁlms, while zone 1 ﬁlms are characterized by limited motion of ad-atoms at the surface. For type A ZnO:Al, the c-axis is predominantly oriented parallel to the substrate normal. Crystallites of types B and C material are also highly oriented, but their c-axis exhibits a small inclination with respect to the substrate normal. Crystalline columns can be easily distinguished in type A ZnO:Al by the SEM crosssections. This is more diﬃcult for type B and C ﬁlms, indicating an increasingly compact and denser structure for the transition from type A to type C material. After etching in diluted HCl structural diﬀerences between the three samples are more obvious and entirely diﬀerent surface morphologies can be observed (see Fig. 8.18 right side). Since type A material is homogeneously etched at high rate, chemical etching has no impact on the surface morphology but only reduces the ﬁlm thickness. Type B and C ZnO:Al are anisotropically etched leading to a crater-like surface structure. The opening angle of these craters is between 120◦ and 135◦ as measured by AFM [96]. The regular, rough morphology of the Type B ﬁlm consists of a homoge-

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

385

after etching

as deposited Sputtered RF-ZnO:Al on glass

C

C

µ

500 nm

B

B

B

increasing compactness

C

A 0.1

30 500 nm

A

100 TS (˚C)

500 nm

1.3 200

270

4.0

1µ

A

2.7 Sputter pressure (Pa) 1µ

Fig. 8.18. Modiﬁed Thornton-model for RF sputtered ZnO:Al ﬁlms according to Kluth et al. [96]: Structural properties and etching behavior as function of deposition pressure and substrate temperature TS . The SEM micrographs show cross-sections and the surface in the as-deposited state (left) and after etching in hydrochloric acid (right). The matrix of substrate temperature and deposition pressure contains structure zones according to the Thornton model [120,121]. Reprinted with permission from [96]

neous distribution of large craters on the complete surface. As will be seen later, this type of surface morphology exhibits the best light scattering properties in solar cells. Because of its extremely compact ﬁlm structure, Type C material is only partially etched leading to a few randomly spread big craters on the ﬁlm surface. This means that, for nonreactive sputtering from ceramic targets, etching behavior and thus surface structure after etching can be adjusted within certain limits by the choice of deposition pressure and substrate temperature. Similar correlations have been observed for reactively sputtered ZnO:Al ﬁlms [122]. Figure 8.19 shows SEM micrographs of texture etched ZnO:Al ﬁlms that were prepared by reactive mid-frequency sputtering from Zn:Al targets at diﬀerent argon pressures and optimized working points. In the low pressure regime (Fig. 8.19c), isolated large craters develop, while for high pressures (Fig. 8.19a) granular structures are observed. In the intermediate pressure range (Fig. 8.19b), regularly distributed, sharp, and deep craters are etched into the ZnO:Al surface.

386

J. H¨ upkes et al.

Fig. 8.19. Transition of etching behavior for reactively sputtered ZnO:Al ﬁlms prepared at diﬀerent deposition pressure: 7 Pa (a), 4 Pa (b), and 0.6 Pa (c) [122]

Fig. 8.20. SEM micrographs of etched ZnO:Al surfaces. The ﬁlms were prepared by reactive MF-sputtering at diﬀerent working points with decreasing oxygen partial pressure from left to right. The etching was done in KOH (top), HCl (middle), or by an ion beam of 3 keV (oxygen/argon gas mixture, gas ﬂows 50 sccm each) in a vacuum chamber (bottom). More details can be found in [124]

Berginski et al. extended the model to RF sputter deposited ZnO:Al ﬁlms with diﬀerent aluminum doping concentration [123]. They described diﬀerent types of surface structure depending on substrate temperature and target doping concentration. With reduced aluminum concentration in the employed targets, the transition temperature between diﬀerent surface structures after etching shifts to higher values. For reactive sputtering, the working point of the process is an additional important parameter to tune the ﬁlm properties. The working point basically describes the process conditions between deposition of pure metallic zinc without oxygen and oxygen surplus. As explained in detail in Chap. 5 of this book, the electrical and optical properties of zinc oxide change if the working point is varied. In this section, we compare the inﬂuence of the working point on surface morphology after etching either in 0.5% HCl (at room temperature) or 33% KOH (at 50◦ C) [93]. SEM images of selected

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

387

ﬁlms are shown in Fig. 8.20. This wet-chemical etching was also compared to a purely physical sputter process using a beam of argon and oxygen ions (gas ﬂows 50 sccm each) from a dedicated ion beam source at normal incidence. The total beam current was 330 mA (ion source length about 40 cm) at a maximum ion energy of 3 keV [124]. The working point during reactive mid frequency sputtering was stabilized by plasma emission monitoring [125] in transition mode and was varied from close to oxide mode (see Chap. 5, Fig. 5.5, high oxygen partial pressure p(O2 ), low voltage) over the transition mode towards metallic mode (low p(O2 ), high voltage). The ﬁlms etched with HCl (Fig. 8.20d–f) developed craters with sloped side walls similar to type B and C ﬁlms of the modiﬁed Thornton model (Fig. 8.18). KOH solution (Fig. 8.20a–c) etched small vertical holes into the ﬁlm, similar to what has been observed in Fig. 8.16 in case of nonreactively sputtered ﬁlms. Ion beam treatment of the ZnO:Al surface results in shallow indentations, which are shown in Fig. 8.20g–i. The structure formation is very diﬀerent for diﬀerent etchants. However, for each etchant, the density of points of attack decreases as the working point approaches the metallic mode. This is even valid for the treatment with the ion beam, which is pure physical sputtering rather than chemical etching. This suggests that the number of points of attack is given by the material itself, while the actual resulting surface morphology as well as shape and size of the surface features for a given material are determined mainly by the etching technique. Similar eﬀects on etching behavior have been observed by adding small amounts of oxygen during sputter deposition from ceramic targets [126]. The results presented above show that it is possible to obtain a certain surface morphology by an adjustment of deposition and etch parameters. One should keep in mind, however, that this is only possible if all other obvious or hidden parameters of the sputtering and etching process and the sputter system itself are known and controlled. These eﬀects are particularly important for up-scaling of processes from laboratory to industrial environment. Some but not all of these inﬂuencing factors have been identiﬁed (e.g. glass age and glass side, in-line or batch processing) and will be discussed

Fig. 8.21. SEM micrographs of ZnO:Al ﬁlm surface textures obtained after etching. The ﬁlms were prepared under similar conditions. The etching steps reveal diﬀerences in the ﬁlms. This means that there must be hidden parameters that have to be identiﬁed and controlled during deposition

388

J. H¨ upkes et al.

Table 8.3. Substrate temperature TS , sputter pressure pdep , and ﬁlm thickness in initial (as deposited) state (tinitial ), removed thickness (tremoved ) and thickness after etching in HCl (tafter ), etch time and rate of the ZnO:Al ﬁlms shown in Fig. 8.21 Figure

TS (◦ C)

pdep (Pa)

tinitial (nm)

Etch time (s)

Etch rate (nm s−1 )

tremoved (nm)

tafter (nm)

8.21a 8.21b 8.21c

160 200 200

0.65 0.97 1.1

845 980 710

40 45 30

3.9 3.9 4.0

155 175 120

690 805 590

in Sect. 8.3.3. To illustrate the large variety of achievable surface structures after etching without going into further details, Fig. 8.21 shows a few examples of nonoptimized surfaces of highly transparent and conductive sputtered ZnO:Al ﬁlms. Some deposition conditions as well as ﬁlm properties are given in Table 8.3. As seen from the table, all ﬁlms were obtained under similar deposition conditions, but one or more of these “hidden” parameters were obviously diﬀerent. Possible candidates to be one of these hidden parameters are the ion bombardment, particle energy, material properties of the target or the motion of the racetracks by moving magnets. The distinctly diﬀerent surface appearances illustrate once again the necessity to control all accessible parameters.

8.3.2.3 Dry Etching of ZnO For silicon thin ﬁlm solar cell manufacturing, not only the magnetronsputtered ZnO:Al layers, but also the subsequent silicon ﬁlms by plasma enhanced chemical vapor deposition (PECVD) are prepared in vacuum processes. Therefore, a replacement of the wet-chemical etching of the ZnO:Al in diluted acids by dry plasma etching methods might be attractive from a process point-of-view. In addition to a suﬃcient etching rate to make the process economically viable, the most important question is, with which techniques and gas chemistries the required surface roughness of the ZnO:Al ﬁlms can be achieved. A number of reports have been published on dry etching of ZnO. Most of the available studies aim at the application of ZnO for optoelectronic devices, which would require structuring and masking steps. Because of this targeted application, examined samples were usually either ZnO single crystals or ZnO ﬁlms grown epitaxially at high temperatures. Applied etching chemistries were manifold, but have been distinguished in a recent review article into Cl2 -, BCl3 -, and CH4 -based chemistries [127]. The reader is referred to this paper for an overview of recent results and a compilation of the related original papers. Often argon is also present in the plasma to assist the chemical etching process by physical removal of surface material. Usually the plasmas are inductively coupled yielding high plasma densities.

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

389

An exception is the work by Hiramatsu et al. [128], where capacitive coupling was used. The highest etch rates of up to 300 nm min−1 were achieved in CH4 /H2 /Ar [129] or BCl3 /CH4 /H2 [130] gas mixtures. For the purpose of ZnO texturing by dry etching for solar cell applications, they seem to be suﬃcient for an industrial application. However, it is questionable whether the required surface roughness can be achieved. Although the various reports are not easily comparable as the as-deposited samples had very diﬀerent surface roughness values, most of the studies especially with CH4 - and BCl3 based chemistries point to a smoothening rather than roughening eﬀect by plasma treatment of ZnO. Still, systematic studies on dry etching of solargrade magnetron-sputtered doped ZnO:Al would be needed to answer the question on the feasibility of such methods. Another approach that has been tried is the surface roughening by pure physical sputtering with argon or oxygen ions using either a magnetron sputter source at moderate discharge voltage levels (up to about 250 V) or alternatively a dedicated ion beam source capable of producing collimated ion beams with ion energies up to 3 keV. Results of the latter experiments have already been shown in Fig. 8.20. As evident from the SEM micrographs, the surfaces after Ar-ion bombardment remain rather smooth. Similar results were obtained in case of magnetron sputter discharge to etch ZnO:Al ﬁlms. Hence, purely physical sputtering seems to be no way to achieve the required surface roughness and lateral length scale of several 100 nm needed for good light scattering.

8.3.2.4 Discussion of ZnO Etching Behavior In the last sections we presented a number of examples for diﬀerent etch structures that were observed for sputter deposited ZnO:Al ﬁlms, depending on preparation and etching conditions. All ﬁlms described in this section have a columnar structure with pronounced orientation of the c-axis along the substrate normal. In addition, most ZnO:Al ﬁlms exhibit properties similar to the ﬁlms of zone 2 of the modiﬁed Thornton model for RF sputtered ZnO:Al ﬁlms, leading to a formation of craters with diﬀerent size, shape, and density during etching in HCl. Up to now, the diﬀerent etching behavior of these ZnO:Al ﬁlms could not be directly related to other ﬁlm properties determined prior to the etching step. This means that ﬁlms with nearly identical optical, electrical, and structural properties can be distinguished by diﬀerent etching behavior. While the microscopic etch mechanism of ZnO single crystals in alkaline and acid solutions is well understood [109, 117], a detailed understanding of the etching behavior of compact polycrystalline ﬁlms is still not available. In the following we will discuss the relation between etching behavior of

390

J. H¨ upkes et al.

polycrystalline sputtered ZnO:Al ﬁlms and ZnO single crystals. The following table compares the etch characteristics: ZnO single crystals

Sputtered polycrystalline ZnO:Al

– Surface termination well deﬁned –

– –

–

–

–

– Surface termination unknown and may vary for neighboring columns Hexagonal craters with sixfold – Often craters, but manifold etch symmetry or pyramids on etched structures possible c-faces Sixfold symmetry of craters – Craters etched in HCl without strong etched in HCl on (001) surface symmetry Fixed opening angle for craters – Opening angle of craters between of δ = 123◦ and γ = 130◦ , 120◦ and 135◦ as measured by respectively (see Fig. 8.14) AFM [96] Selective etching by acidic and – Selective etching of groups of crystal alkaline solutions depending on columns in alkaline solutions surface termination Crater formation at crystal – Criteria determining points of local defects attack unknown, grain boundaries are not suﬃcient Depending on surface termina- – Density of points of attack depends on ﬁlm properties and is similar for tion and etchant, either plane HCl, KOH, and even for sputter etchetching or attack only at defects ing (see Fig. 8.20)

The etch structures on polycrystalline ZnO:Al ﬁlms do not exhibit strong hexagonal symmetry, but sometimes etch pits with six edges are formed. Craters are common structures for acidic etching of compact, sputter deposited ZnO ﬁlm with columnar grains and a c-axis directed perpendicularly to the substrate. Crater formation seems to be inherent to chemical etching of ZnO c-planes. Even the opening angle of craters etched by HCl onto polycrystalline ZnO and the craters formed by the (101) planes on ZnO single crystals are very similar. We conclude that the microscopic etch mechanism must be the same for single crystals and sputter deposited, polycrystalline ZnO:Al. For the latter, the tendency for crater formation is masked by inhomogeneous chemical or physical properties like porosity, composition or, in case of dynamic deposition, multilayered ZnO:Al ﬁlms. This multilayer structure results from the fact that structural properties of ZnO:Al deposited by a sputter process varies depending on the position of the ﬁlm relative to the race track of the sputter target [131, 132]. This dependence is important for the etch rate of in-line sputter deposited ﬁlms [133]. If the sputter deposited ZnO:Al ﬁlm favors crater formation upon etching, the strong hexagonal symmetry of etched craters in single crystals is

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

391

relaxed by the missing long-range periodicity over many grain boundaries in polycrystalline ZnO:Al ﬁlms. The suggestion that the crater walls on etched ZnO:Al ﬁlms consist of facetted (101) faces of diﬀerent crystallites has not yet been conﬁrmed experimentally. Additionally, the local surface property which is responsible for the favored attack of etchants at speciﬁc locations (points of attack) on ZnO:Al ﬁlms is still not identiﬁed. The detailed role of the termination of crystal columns with either oxygen or zinc atoms for the etching behavior is also unclear, although the surface termination has been related to diﬀerent etching behavior [134]. In our experiments, the similar crater formation for single and polycrystalline ZnO:Al in HCl and the highly selective etching of KOH indicate a strong inﬂuence of termination on the etching behavior, following the model by Mariano and Hanneman [109]. However, our etching experiments of Zn-terminated ZnO single crystal surfaces in KOH show only very small etch rates, which contradicts an interpretation according to this model. A second argument against the inﬂuence of the termination is the similar density of points of attack observed for diﬀerent ﬁlms etched in HCl, KOH, or with an ion beam (see Fig. 8.20). This result suggests that the points of attack do not only appear with similar density but also are exactly the same for the diﬀerent etchants. This could mean that the points are not the result of a special surface chemistry, but of a structural defect that is less resistive against all kinds of etching. These defects must be much stronger than simple grain boundaries in compact ZnO:Al ﬁlms, because the craters extend over a large number of crystallites. Another possible explanation of the etching is based on the orientation of the polar faces of ZnO. The O-terminated crystallites are more sensitive to chemical etching as it was observed experimentally for single crystals with rapid overall etching and etching at defects in case of HCL and NaOH, respectively (see Fig. 8.14). The Zn-terminated faces were only attacked by the acid at defect positions, while no etching was observed in NaOH. That means that Zn-terminated columns are resistant against alkaline solutions but can be etched by an acid from sides leading to (101) faces of etched crystals. Under the assumption that the majority of ZnO columns is Zn-terminated, this attack from the side could only occur if either the ﬁlm is suﬃciently porous, e.g., for high depositions pressure, or one of the neighboring columns has already been etched. This model explains the formation of craters and deep holes at positions of O-terminated ZnO columns by termination-dependent sensitivity to acids and alkaline solutions, respectively. The similar density of points of attack for ion beam treatment might occur by chance or could be also related to termination of the crystals. As a general caveat, note that an interpretation of etch results in terms of surface termination is diﬃcult, as the microscopic structure of ZnO single crystal or ZnO:Al ﬁlm surfaces is unknown. Eﬀects like surface reconstruction or presence of adsorbates may constitute the main factor determining the etching behavior, thus masking all other eﬀects [135]. A direct microscopic

392

J. H¨ upkes et al.

measurement of the surface chemistry and termination of single columns and the microscopic structure of the points of attack is required to prove or extend these models. 8.3.3 Practical Aspects for Etching of Sputtered ZnO:Al One of the challenges of any R&D eﬀort is the transfer of laboratory results to a manufacturing environment with its restrictions in terms of cost, throughput, and availability of materials. In this section we will elaborate on some practical aspects speciﬁc to the etching process of magnetron sputtered ZnO:Al ﬁlms. An important factor whose inﬂuence is easily neglected in a laboratory environment are the properties of the glass sheet onto which the ZnO:Al front contact is deposited. For low-cost manufacturing, the preferred glass is standard so-called ﬂoat glass, which is produced in glass sheets of 6 × 3.2 m2 size (in European plants) with a throughput of millions of square meters per year. After manufacturing, the glass might be stored for extended periods of time in often uncontrolled environments before it is purchased. If the ambient conditions are humid, the Na-content of the glass may form NaOH at the glass surface, thereby leading to actual corrosion of the glass (see e.g. [136, 137]). Furthermore, standard ﬂoat glass is produced by letting the liquid glass “ﬂoat” onto a bath of molten tin. Because of this contact with the Sn-bath the so-called “bath-side” of the ﬁnished glass sheet will have diﬀerent surface properties [137–140]. Finally, there are diﬀerent methods of cleaning and drying a glass sheet prior to deposition [141]. Often, a speciﬁc deposition process or layer sequence requires an adopted cleaning process to obtain the best process and ﬁlm results. Many of these eﬀects are still not well understood and procedures often more rely on experience than on actual knowledge. As an example, the eﬀect of the glass surface is illustrated in Fig. 8.22, where SEM micrographs of two ZnO:Al ﬁlms are shown. The ﬁlms were prepared onto diﬀerent sides of a 3 mm thick ﬂoat glass in the same run by RF sputtering and subsequent texture etching in diluted HCl [19]. Both glass samples underwent the same cleaning procedure and were stored in a closed box ﬂooded with dry nitrogen prior to deposition. Initial thickness of both ZnO:Al ﬁlms was about 800 nm and their electrical and optical properties were identical. However, major diﬀerences show up in the etch rates, which are about 3 and 9 nm s−1 in 0.5% HCl for the bath- and ﬁre-side of the glass, respectively. The etch time was adjusted such as to yield approximately the same removed ﬁlm thickness for both samples. Totally diﬀerent surface structures develop during etching as shown in the Fig. 8.22. For the ﬁlm on the bath-side, a large number of holes is produced. One of them can be seen in the right part of the SEM image. Like in this example, such diﬀerences often become only obvious after the etching step, which is very sensitive to the structural ﬁlm properties

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

393

Fig. 8.22. SEM micrographs of two etched ZnO:Al ﬁlms deposited on the ﬁre (a) and bath side (b), respectively, of a 3 mm thick ﬂoat glass substrate. The etch times were adjusted to yield a removed ﬁlm thickness of about 180 nm in both cases

(see e.g. [19, 96]). Hence, for the ZnO:Al sputtering and etching process the properties of the glass surface have to be always considered and carefully monitored, as it is known for many other applications (see e.g. [137] and references therein). To obtain a well-deﬁned glass surface, even a deposition of thin interlayers like SiOx or SiOx Ny between glass and TCO might be recommended [142]. A second practical aspect is the up-scaling of the etching equipment, which is not available as an oﬀ-the-shelf product from equipment manufacturers. The task is to etch glass sheets of a size of about 1 m2 in diluted HCl with excellent homogeneity over the whole substrate area. This requires precise control of both etch time and etchant concentration, while additionally the requirements of a clean process have to be fulﬁlled to avoid layer imperfections or pinholes in the silicon layers. With a prototype in-line etcher, we have achieved excellent uniformity of the etching step on substrates up to 30 × 30 cm2 [48]. This process can be easily up-scaled to production size.

8.4 High Eﬃciency Silicon Thin Film Solar Cells In this section we present solar cells and modules on diﬀerent ZnO:Al front contacts. Having already established some general requirements for the use of TCO contacts in silicon thin ﬁlms solar cells (Sect. 8.2.4), we now discuss in detail the relationship between ZnO:Al properties and cell performance. Finally, we report the current status of high eﬃciency solar cells and modules on optimized sputtered and texture etched ZnO:Al ﬁlms. 8.4.1 Optimization of Solar Cells When we attempt to establish a relationship between cell performance and TCO, it is indispensable to break down the conversion eﬃciency of a solar cell

394

J. H¨ upkes et al.

or module into the parameters of the J/V -curve of the device, as introduced in Sect. 8.2.1 of this chapter. We will see that diﬀerent TCO features may or may not have a major inﬂuence on one or more of these parameters (open circuit voltage VOC , ﬁll factor FF, or short circuit current density JSC ). Some of these points have already been mentioned in Sect. 8.2.4, but will be treated in more detail below.

8.4.1.1 ZnO:Al/Si Interface The ﬁrst aspect to consider is the interface between ZnO:Al and the subsequently deposited layer, which will be p-doped amorphous or microcrystalline silicon if we restrict ourselves to the p–i–n or superstrate conﬁguration. Interface properties are mainly the chemistry of the interface, the band alignment, the nature and density of interface states, and the interface structure, which will aﬀect the light incoupling and scattering and can also change the eﬀective interface area. Some physical details about ZnO interfaces with other materials are discussed in Chap. 4 of this book. From an experimental point of view, the interface between ZnO:Al and the p-layer has to provide an electrical contact with a contact resistance as low as possible. While on tin oxide amorphous p-layers can be used to fulﬁll this requirement for a-Si:H solar cells, on ZnO:Al an adapted double p-layer design consisting of the combination of a microcrystalline and an amorphous p-layer is necessary to obtain low-ohmic contact required for a high FF without loss in the open circuit voltage VOC [56]. More details on the properties of the ZnO/p-interface can be found in literature (see e.g. [53–55, 94]). To illustrate the ZnO:Al/p-contact problem, Fig. 8.23 shows illuminated J/V -curves of two a-Si:H solar cells on texture-etched ZnO:Al, which only

Fig. 8.23. J/V curves of a-Si:H solar cells on texture etched ZnO:Al which diﬀer only by the p-layer design. The solid and dashed lines were measured for solar cells with a simple amorphous silicon and a µc-Si:H/a-Si:H double p-layer, respectively

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

395

diﬀer from each other by the p-layer (a-Si:H-p, vs. µc-Si:H-p/a-Si:H-p). The lower ﬁll factor in case of the single amorphous p-layer is caused by a high series resistance of the interface, which is strongly reduced by inserting a µc-Si:H p-layer between ZnO:Al and a-Si:H-p. On the other hand, the amorphous part of the p-layer is also important to maintain high VOC . Figure 8.24 compares the VOC values obtained for a-Si:H single junction p–i–n solar cells using either a microcrystalline p-layer only or a µc-Si:H/a-Si:H double p-layer design. The thickness of the a-Si:H p-layers was kept constant at about 8 nm. Data are plotted as function of µc-Si:H p-layer deposition time, since the exact layer thickness cannot be determined easily on rough ZnO:Al front contacts. For µc-Si:H-p deposition times of 1.5 min (for which we estimated a µc-Si:H p-layer thickness of 5–10 nm) a good ZnO:Al/p-contact, i.e., low series resistance and ﬁll factors above 70% are obtained. However, for a single µc-Si:H p-layer VOC saturates around 760 mV, while for a µc-Si:H/a-Si:H double p-layer the values are around 840 mV, independent of the µc-Si:H p-layer thickness as long as deposition time is above 30 s. To minimize absorption losses, one uses the smallest possible thickness, which still maintains both high VOC and high FF. Absorption in the double p-layer leads to slightly lower short circuit current density as can be seen in Fig. 8.23. Deposition times of 90 and 30 s for µc-Si:H and a-Si:H p-layer, respectively (corresponding to a thickness of approximately 5–10 nm in both cases) turned out as the optimal values. For µc-Si:H solar cells, on the other hand, a purely microcrystalline p-layer is suﬃcient to achieve a good TCO/p contact and a good p/i-interface, leading to VOC values typical for high quality µc-Si:H solar cells.

Fig. 8.24. Open circuit voltage VOC of a-Si:H solar cells prepared using a microcrystalline p-layer only (open squares) or a µc-Si:H/a-Si:H double p-layer design (solid triangles) as function of µc-Si:H-p deposition time. The thicknesses of the a-Si:H-p-layers were kept constant between 5 and 10 nm [143]

396

J. H¨ upkes et al.

8.4.1.2 Opto-Electrical ZnO:Al Properties Very obvious parameters of the TCO contact that directly inﬂuence the solar cell performance are electrical conductivity (or sheet resistance, Rsq ) and transparency. The relations have already been established in Sects. 8.2.4 and 8.2.5. An improved transparency of the TCO results in a higher short circuit current density, while an improved conductivity allows a higher ﬁll factor of an otherwise optimized solar cell or module. A low sheet resistance is most important for purely µc-Si:H solar cells due to the high short circuit currents. Figure 8.25, which shows spectral quantum eﬃciency QE and total cell absorbance (1 − R) [17], illustrates these relationships. Here, µc-Si:H cells were deposited on ZnO:Al ﬁlms of three diﬀerent initial thicknesses, yielding ZnO:Al with 330, 870, and 1 370 nm ﬁnal thickness and similar surface roughness after 30 s of etching in HCl. ZnO:Al thickness, sheet resistance Rsq , and solar cell parameters FF and JSC are listed in Table 8.4. The thick

Fig. 8.25. Quantum eﬃciency QE and cell absorption (1 − R) curves of µc-Si:H pin solar cells (i-layer thickness 1.2 µm, ZnO/Ag back reﬂector) for front ZnO:Al with diﬀerent thickness: 330 nm (dashed ), 870 nm (dotted ), and 1 370 nm (solid ). R is total cell reﬂectance. Sheet resistance of the front contacts as well as F F and JSC of the solar cells are given in Table 8.4 Table 8.4. Properties of 1.2 µm thick µc-Si:H solar cells on diﬀerent ZnO:Al front contacts: ZnO:Al thickness and sheet resistance Rsq as well as FF and JSC of the solar cells. QE curves of these solar cells are shown in Fig. 8.25 ZnO:Al thickness (nm)

Rsq ( Ω)

FF (%)

JSC (mA cm−2 )

330 870 1370

10.1 4.4 2.7

68 72 75

20.8 18.5 16.6

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

397

front contact has a lower Rsq , which increases FF due to reduced TCO series resistance losses. However, lower optical transparency leads to lower QE and thus JSC . Comparing the thin and the thick substrate, optical losses, given by the diﬀerence between cell absorbance and QE, appear mostly in the red and NIR region (>600 nm) due to free carrier absorption in the front contact. In this wavelength region the current density loss ∆JSC as compared to the thin ZnO:Al is approximately 3 mA cm−2 and can be attributed mainly to the reduced absorbance in thin ZnO:Al layer. The absorption in the TCO ﬁlms is further enhanced by multiple passes of scattered light within the solar cell structure. Small losses ∆JSC = 1 mA cm−2 ) are also observed in the short wavelength region (<600 nm). These eﬀects will be explained in detail in the next section. This experiment demonstrates the conﬂicting requirements of good electrical and optical TCO properties and the necessity of careful optimization with respect to each other. Recent progress has been made by optimizing the doping concentration to tailor the electrical and optical needs [123] (see also Fig. 8.28). 8.4.1.3 ZnO:Al Surface Structure and Light Trapping The probably most important inﬂuencing factor of the TCO on solar cell performance is the TCO surface structure, comprising individual surface features of diﬀerent shape, lateral and vertical sizes, and statistical distributions. As mentioned before, these relationships are highly complex and still not well understood. Some of the theoretical approaches to clarify these questions have been referenced in Sect. 8.2.4 already. Hence, our treatment of this subject will be mostly experimental without any attempt of detailed theoretical explanations. Let us ﬁrst have a second look at Fig. 8.9, which nicely illustrates the two major eﬀects of a rough ZnO:Al-surface. The ﬁrst one is eﬀective in the short wavelength region (400–600 nm). Here the enhanced QE for the cells on rough ZnO:Al is due to an improved light coupling into the device due to an antireﬂection eﬀect of the rough interface between front TCO and silicon. The roughness results in an eﬀectively graded index of refraction at the interface hence reducing the amount of reﬂected light, which is based on the eﬀective medium theory by Bruggeman [144]. The second eﬀect, in the longer wavelength region, is of course the aforementioned light trapping, which strongly increases the light path within the silicon absorber (more than a factor of 15) and hence leads to higher absorption and current generation. However, a look at the (1 − R) curves, which basically represents the amount of light trapped within the device, demonstrates also the downside of light trapping. During multiple reﬂections, light will always be absorbed in photovoltaically non-active layers (TCO, doped layers, and back reﬂector) at every pass. This eﬀect explains the increasing diﬀerence between light trapped in the device (1 − R) and QE for longer wavelengths and stresses once more the necessity of low TCO absorbance.

398

J. H¨ upkes et al.

Table 8.5. J/V parameters of identically prepared a-Si:H solar cells on diﬀerent ZnO:Al front contacts. The surface structures were already shown in previous sections ZnO:Al Fig. Fig. Fig. Fig.

8.19a 8.19b 8.19c 8.8c

η (%)

F F (%)

VOC (mV)

JSC ( mA/cm2 )

6.7 9.2 8.6 9.6

71.2 70.3 70.5 69.9

865 875 910 910

10.9 15.0 13.4 15.1

To investigate the inﬂuence of ZnO:Al surface details on solar cells, identical a-Si:H solar cells were prepared on diﬀerent ZnO:Al front contacts, whose surface structures are shown in Fig. 8.19. Table 8.5 lists J/V parameters of these solar cells. The ﬁne, granular surface (Fig. 8.19a) induces only poor light scattering, resulting in low short circuit current density. Regularly distributed craters of the ﬁlm in Figs. 8.19b and 8.8c perform very strong light scattering and good light trapping, which can be seen at the high JSC of 15 mA cm−2 . Surfaces with only a small number of large craters within an otherwise relatively smooth surface (Fig. 8.19c) improve current generation as compared to smooth TCO, but are clearly inferior to the previous surface morphology. There have been many attempts to quantify the surface appearance and light scattering behavior of textured front TCOs and correlate the results with solar cell performance using either optical measurements or measurements of the surface morphology by atomic force microscopy [58–60,62–64,66,96,145]. In the former case, a widely used parameter is the so-called haze value, which describes the ratio of scattered to total transmitted or reﬂected light intensity as a function of wavelength. This haze has been found to correlate well with surface morphology described by the root mean square roughness of etched ZnO:Al showing an increase in haze (at a given wavelength within the light scattering region) with increasing surface roughness. The results have been published by Lechner et al. [145] and are reproduced in Fig. 8.26. The ZnO:Al ﬁlms (series A, B, and M) were prepared using diﬀerent sputter techniques and subsequently etched in HCl. Asahi U is a laboratory type glass substrate with an SnO2 :F TCO layer deposited by CVD [71]. The correlation is largely lost if short circuit current JSC of identical µcSi:H solar cells prepared onto various TCOs is plotted as function of the haze of the corresponding TCO ﬁlm (Fig. 8.27). This indicates that the haze is not suﬃcient to characterize the light scattering behavior of a given TCO surface. Therefore, Lechner et al. [145] took additional parameters into account. One was the power spectral density (PSD) as obtained from AFM representing the surface proﬁle as it is composed of various sine waves with different amplitudes and periods. Another one was angular resolved scattering (ARS), which is governed by the size of the texture details and thus involves

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

399

Fig. 8.26. Haze as function of RMS roughness δrms for diﬀerent textured TCO front contacts according to Lechner et al. [145]. The TCO ﬁlms were prepared under diﬀerent conditions and consist of surface features with diﬀering distribution, size, and shape. The TCOs labeled series A, B, and M are texture etched ZnO:Al ﬁlms, while Asahi U is a high quality laboratory type SnO2 :F covered glass [71]. More details of the corresponding TCO ﬁlms can be found in the original papers

Fig. 8.27. JSC of µc-Si:H solar cells as function of haze for diﬀerent textured TCO front contacts, which have already been shown in Fig. 8.26. The µc-Si:H solar cells have identical i-layer thickness. For details regarding the diﬀerent TCO front contacts, the reader is referred to the original paper [145]

both “vertical” and “horizontal” dimensions of the surface morphology. The authors were able to ﬁnd some correlations and the reader is referred to the original papers for more details [145, 146]. Recently, Stiebig et al. continued this work [66] and developed a new approach to extract the light scattering behavior from AFM measurements using ray tracing by Snell’s diﬀraction law on nanoscopic scale. It should be mentioned that zinc oxide is also used between silicon and the metallic contact as a part of the back reﬂector to improve its optical properties and to act as a diﬀusion barrier [12, 13]. A layer thickness between about 60 and 100 nm is usually suﬃcient here, and there are no strict requirements for

400

J. H¨ upkes et al.

conductivity of the ZnO ﬁlm, as the lateral current is carried by the metal deposited onto the ZnO. Therefore, even undoped ZnO can be used here, which has the advantage of reduced free carrier absorption and therefore increased NIR transparency, which is desirable for light trapping especially in µc-Si:H solar cells. A more detailed discussion on the inﬂuence of the ZnO/metal back reﬂector on solar cell performance can be found in further literature [17, 58, 59, 147]. As already mentioned in Sects. 6.3.1.2 and 8.2.3, ZnO can be used as an intermediate reﬂector in a-Si:H/µc-Si:H tandem solar cells between top and bottom cell to increase the current in the thin amorphous silicon top cell [14, 15]. As a result, the thickness of the a-Si:H top cell can be reduced to improve the stability of the cell upon light induced degradation. Recent progress has even lead to improved light utilization and higher short circuit current density by an optimized intermediate reﬂector [16]. However, this intermediate reﬂector might consist of another material than ZnO. 8.4.1.4 Electrical Performance of Solar Cells We will now have a closer look at the ﬁll factor of solar cells on the different ZnO:Al front contacts. The inﬂuence of the electrical properties of the TCO and the TCO/p interface has already been discussed earlier. The requirements for TCO surface morphology are such that in general the surface features should not be too steep and resulting edges should not be too sharp. Such surface morphologies will increase the proneness to shunting, as sharp edges and steep inclinations might not be covered very well by subsequently deposited layers, e.g., simply through geometric shadowing and line-of-sight eﬀects. Hence, short circuit paths can more easily develop at such spots and are additionally promoted by the stronger electric ﬁeld at peaks or edges, which could lead to easier electrical breakdown. Experimentally, however, it is a hard task to separate these morphology eﬀects on ﬁll factor from contact resistance losses. Hence, these relations are more the result of experimental experience compiled over many years than of systematic experiments. Going back to Table 8.5, the last parameter we should look at is the open circuit voltage VOC . The highest VOC value is obtained for the ZnO:Al ﬁlm in Fig. 8.19c with the overall smoothest surface structure or in other words, the lowest total surface area. Note that this does not correlate necessarily with the lowest short circuit current value. VOC , like ﬁll factor, in general is strongly inﬂuenced by the details of the layer structure and in particular the TCO/p interface region. Hence, it is impossible to make general correlations between TCO surface and VOC . However, the observed inﬂuence on surface area can be understood in a simple picture using the simple diode model of a solar cell (see e.g. the book of Chopra and Das [148]): VOC =

JSC neﬀ kT ln(1 + ) e J0

(8.6)

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

401

with the parameters diode quality factor neﬀ , Boltzmann’s constant k, absolute temperature T , and dark reverse voltage saturation current density J0 . The J0 -value usually increases with the number of recombination events at the TCO/Si interface or at the interface between doped and undoped silicon, which in turn increases with the geometrical area of the interface region. As for rough TCO the interface area to the p-layer is much larger than that for smooth TCO, and this explains qualitatively the observed decrease in open circuit voltage for the rough TCO in Table 8.5. Note that the negative eﬀect of surface roughness on VOC is more than oﬀ-set by the increase in short circuit current due to the light trapping eﬀect. On the other hand, the high quality reference ZnO:Al ﬁlm can combine high VOC and high JSC . The inﬂuence of interface area on VOC is utilized in high eﬃciency crystalline silicon solar cells with the so-called point contact concept to increase the open circuit voltage by suppressing area related recombination [149]. Recently, the point contact concept has been applied also in µc-Si:H solar cells using photolithography or self-organized zinc oxide etch masks [150]. However, so far VOC improvement could not be demonstrated. For more details of preparation and application of ZnO etch masks the reader is referred to the original work [118].

8.4.2 Highly Transparent ZnO:Al Front Contacts Very good light trapping can be achieved for silicon thin ﬁlm solar cells by an adapted surface texture of the front TCO. However, current generation is still limited by signiﬁcant absorption in the rough ZnO/metal back reﬂector and by free carrier absorption in the front ZnO:Al ﬁlm for the near infrared (NIR) spectral range [58]. While the former eﬀect is beyond the scope of this article (see e.g. Vanecek et al. and Springer et al. [33, 147]), the potential of reduced red and NIR absorption of the ZnO:Al front contact by reduction of free carrier density was demonstrated by Rech et al. [45]. They succeeded in reducing the charge carrier density in ZnO:Al ﬁlms by utilizing a target with only 0.8 at. % alumina while still maintaining good electrical properties and a surface texture similar to Fig. 8.8c with excellent light scattering properties. The conductivity of the ﬁlms remained high despite the reduced carrier concentration by an increase in electron mobility, leading to sheet resistance values of the 700 nm thick ﬁlms after etching between 10 and 15 Ω. While these sheet resistance values probably need some further optimization for the use in large area solar modules (see Sect. 8.2.5), they are already good enough for highly eﬃcient µc-Si:H and particularly a-Si:H/µc-Si:H tandem cells. The quantum eﬃciency curves of two 1 µm thick µc-Si:H cells prepared on these optically improved substrates are shown in Fig. 8.28. A current gain of around 1 mA cm−2 can be achieved as compared to the standardly used reference substrate with a doping level of 1.6%. Absorption spectra of both ZnO:Al front contacts can be seen in Fig. 8.7. The short circuit current densities, calculated

402

J. H¨ upkes et al.

Fig. 8.28. Quantum eﬃciency and cell absorption (1-R) of µc-Si:H solar cells on ZnO:Al ﬁlms sputtered from targets with diﬀerent doping levels (0.8 at. % (solid ) and 1.6 at. % (dashed )). Optical data of these ZnO:Al ﬁlms are given in Fig. 8.7. Simulation results using improved optical components are also included (dotted ) [147]

from spectral response measurements, are as high as 23.0 and 24.3 mA cm−2 for 1 µm thick µc-Si:H cells on optimized ZnO:Al with 1.6 and 0.8 at. % doping level, respectively. Berginski et al. followed another approach to tailor the optical and electrical properties of texture etched ZnO:Al ﬁlms without aﬀecting the surface topography [151]. ZnO:Al ﬁlms were optimized with respect to surface texture after etching (see also Sect. 8.4.1.3) and high conductivity. By vacuum annealing after etching it is possible to reduce the carrier density and thus to improve NIR transparency. From experimental data of improved short circuit current density by reduced parasitic TCO absorption they estimated an optimized carrier density around 2 × 1020 cm−3 and a potential for solar module eﬃciency between 3 and 4.5% relative compared to untreated ZnO:Al ﬁlms for µc-Si:H single junction and a-Si:H/µc-Si:H/µc-Si:H triple junction solar modules, respectively. For this estimation, resistive losses in the TCO front contact are already included. For further details, the reader is referred to the original paper. From a theoretical point-of view, signiﬁcantly higher current densities are feasible, but require further improved front TCO ﬁlms and perfect mirrors as back reﬂectors. This is illustrated by the dotted curve in Fig. 8.28, which shows simulations of quantum eﬃciency for a 1 µm thick µc-Si:H solar cell. These simulations reveal a current potential of 29.2 mA cm−2 by improved optical components like reduced parasitic absorption in the front TCO, ideal Lambertian light scattering, dielectric back reﬂectors, and antireﬂection coatings on the front side [147]. However, this still has to be achieved experimentally.

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

403

8.4.3 Application of Texture Etched ZnO:Al Films in High Eﬃciency Solar Cells and Modules Having established the general relationships between front contact properties and solar cell performance, this section will summarize the present status of high eﬃciency solar cells and modules focusing on state-of-the-art texture etched ZnO:Al ﬁlms as substrates. The best stabilized cell eﬃciencies on ZnO:Al obtained at the FZJ in a PECVD reactor for 30×30 cm2 substrates were 8.0, 8.9, and 11.2% for a-Si:H p–i–n, µc-Si:H p–i–n, and a-Si:H/µc-Si:H tandem cells, respectively [152]. The best initial eﬃciency for µc-Si:H single junction solar cells was achieved by Mai et al., who reached 10.3% for an absorber layer thickness of 1.6 µm, which was deposited at a high deposition rate of 11 ˚ A s−1 [153]. Initial aperture area module eﬃciencies of 10.8 and 10.6% were achieved for a-Si:H/µc-Si:H tandem modules with an aperture area of 8 × 8 and 26 × 26 cm2 , respectively [154,155]. The eﬃciency of the small area module stabilized at an eﬃciency of 10.1% after 1 000 h of light soaking as conﬁrmed by the National Renewable Energy Laboratory (NREL, see Fig. 8.29). Although the excellent results cited above were achieved with RF-sputtered ZnO:Al, the low deposition rates and high equipment costs of RF sputtering have led to the development of alternative deposition processes for ZnO:Al. One example is high rate DC (or pulsed DC) sputtering from ceramic targets. For this process, ceramic targets with suﬃcient conductivity are needed. High material costs for these ceramic targets can be overcome by reactive mid-frequency sputter deposition from metallic Zn:Al alloy targets

Fig. 8.29. J/V curve of an a-Si:H/µc-Si:H tandem solar module after 1 000 h of light soaking. Measurement was conﬁrmed by NREL

404

J. H¨ upkes et al.

(see also Chap. 5). A direct comparison of high rate reactive and nonreactive sputter deposition of ZnO:Al ﬁlms for solar cell and module application is given elsewhere [99]. Reactively in-line deposited ZnO:Al ﬁlms were manufactured at dynamic deposition rates up to 115 nm m min−1 (from dual cathodes), which corresponds to a static rate of about 400 nm min−1 [93]. After etching the solar modules on these substrates achieved an initial aperture area eﬃciency of 10.4 and 9.8% on 8 × 8 and 26 × 26 cm2 aperture area, respectively [156]. For in-line pulsed DC sputter deposited ZnO:Al ﬁlms from ceramic targets (deposition rate of about 44 nm m min−1 from a single cathode) an initial aperture area eﬃciency of 10.7% was achieved for an 8 × 8 cm2 module [157]. Note that commercial ﬂoat glass served as substrate for these modules. Further improvement in solar module eﬃciency at the FZJ is expected by combining recent and future developments regarding the TCO and silicon preparation as well as new cell concepts. A totally diﬀerent approach to produce zinc oxide as transparent contacts with a rough surface utilizes low pressure chemical vapor deposition (LPCVD). Excellent results have been achieved with this type of substrate [158], which are described in detail in Sect. 6.3.2.2 of this book. The highest initial eﬃciencies of silicon thin ﬁlm solar cells in superstrate conﬁguration were achieved by Yamamoto et al. [159], who presented an amorphous silicon based solar cell in superstrate conﬁguration with an initial eﬃciency of 15.0%. Stabilized values are not yet published. Best overall silicon thin ﬁlm solar cell eﬃciency was achieved by Yan et al. [160]. They realized an a-Si:H/a-SiGe/µc-Si:H triple junction solar cell in substrate conﬁguration on stainless steel with Ag/ZnO back reﬂector and indium tin oxide front contacts leading to an initial eﬃciency of 15.1%, which stabilized at 13.3% after light soaking. The world best thin ﬁlm silicon module eﬃciencies have also been reported by Yamamoto et al. [159] with an initial aperture area eﬃciency of 13.5% (4 140 cm2 ) for an a-Si:H/µc-Si:H solar module with interlayer between top and bottom cell. Takatsuka et al. achieved an initial aperture area module eﬃciency of 12.8% (2 184 cm2 ) for an a-Si:H/µc-Si:H solar module [161].

8.5 Summary Solar modules based on amorphous and microcrystalline silicon and their alloys are one of the most promising options for future large-scale photovoltaic electricity generation. An integral part of these devices are the transparent conductive oxide layers used as a front electrode and as a part of the back reﬂector. These ﬁlms have to be highly conductive and optically transparent in the active region of the absorber material. In addition, silicon thin ﬁlm solar cells have to incorporate a mechanism to trap incoming light within the absorber layer to enhance the intrinsically low absorption of silicon and allow for a reduction of absorber layer thickness. The light trapping eﬀect is usually achieved by combination of a highly reﬂective back

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

405

contact with rough interfaces within the device to eﬀectively scatter incoming light. One approach to introduce rough interfaces is the surface texturing of the front TCO. While diﬀerent materials and preparation techniques are in use, in this review we focused on ZnO:Al ﬁlms, which are prepared by magnetron sputtering and subsequently textured by a simple wet-chemical etching step. One of the main advantages of this technique is that it allows to a certain degree to separate optical and electrical ﬁlm properties from the light scattering capability, which can then be tuned to achieve the best light trapping within the device. Next generation thin ﬁlm silicon solar modules will incorporate microcrystalline silicon as absorber material, which utilizes sunlight up to a wavelength of about 1 100 nm. Therefore, a high transparency in the visible and near infrared spectral region is required for an optimized TCO material. Such properties can be achieved for ZnO:Al by using sputter targets with low Al-dopant concentration and applying high substrate temperatures at the same time. This leads to higher NIR transmittance due to lower charge carrier concentration while maintaining excellent conductivity by an increased mobility of the ZnO:Al ﬁlms. Optimized ﬁlms were obtained for targets of only 0.5 at % Al and substrate temperature of 360◦ C, leading to highest transmittance and low electrical resistivity of 3.3 × 10−4 Ω cm. The etching behavior and resulting surface morphology of the ﬁlms after etching strongly depend on the initial ﬁlm properties and etching conditions. Although there is no microscopic model available to describe these relationships, some insight can be gained by comparing the etching properties of ZnO single crystals and the sputter deposited polycrystalline ﬁlms, which are all highly c-axis oriented. As in the case of single crystals, the etching behavior in acidic or alkaline solutions is distinctly diﬀerent. Replacing the wet-chemical etching step by dry etching (plasma or ion beam etching) seems to be currently no viable option. It turns out that surface morphologies favorable for light trapping in solar cells are achieved by etching in diluted acids. Such an etching step creates surfaces with regularly distributed large and small craters of lateral sizes of up to several microns, which lead to the highest cell eﬃciencies. A material structure leading to this kind of surface appearance can be achieved by a proper choice of sputter pressure and substrate temperature, which are the most important process parameters for this purpose. In case of reactive sputtering from metallic Zn:Al targets, the working point of the reactive process (oxygen fraction of sputter gas mixture) also plays a dominant role and has to be used to optimize structural ﬁlm properties. When applied in silicon thin ﬁlm solar modules, texture-etched ZnO:Al ﬁlms lead to excellent cell and module eﬃciencies. The best stabilized eﬃciencies achieved at the Institute of Energy Research – Photovoltaics (Forschungszentrum J¨ ulich GmbH) on RF sputtered ZnO:Al are 8.0, 8.9, and 11.2% for a-Si:H pin, µc-Si:H pin, and a-Si:H/µc-Si:H tandem cells, respectively. The best small area (64 cm2 aperture area) solar module has a stable aperture area eﬃciency of 10.1%. Initial aperture area eﬃciency of 10.4%

406

J. H¨ upkes et al.

has already been achieved on reactive mid-frequency sputtered ZnO:Al prepared at a high dynamic deposition rate. The high deposition rate makes the reactive process particularly interesting from an economical point of view. In summary, the process of magnetron sputtering and subsequent wet-chemical etching in diluted acid has proven its applicability in silicon thin ﬁlm solar cells and modules. The biggest advantages compared to commercially available substrates are achieved with microcrystalline silicon absorber layers. In parallel to working on the remaining scientiﬁc questions like an understanding of the ZnO etching behavior, the next step is to make the process viable for larger areas. If the excellent electrical, optical, and light scattering properties of magnetron sputtered and texture etched ZnO:Al substrates can be transferred to a large area, high-throughput industrial production, such ﬁlms could be the material of choice for future thin ﬁlm silicon solar module manufacturing. Acknowledgements. The authors would like to express their special thanks for their contributions to all former and current colleagues of the Institute of Energy Research – Photovoltaics in the Research Centre J¨ ulich (FZJ). Especially we appreciate numerous discussions with our experts for diﬀerent subjects: Anton L¨oﬄ, Oliver Kluth and Michael Berginski (TCO development), Aad Gordijn and Tobias Repmann (a-Si:H and µc-Si:H solar cells), and Helmut Stiebig (solar cell characterization and modelling). Financial support by the German Federal Ministry of Education and Research (BMBF) and the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU) is gratefully acknowledged.

References 1. M. Zeman, R.E.I. Schropp, Amorphous and Microcrystalline Silicon Solar Cells (Kluwer, Norwell, 1998) 2. B. Rech, H. Wagner, Appl. Phys. A 69, 155 (1999) 3. J.K. Rath, Sol. Energ. Mater. Sol. Cell. 76, 431 (2003) 4. J. M¨ uller, G. Sch¨ ope, O. Kluth, B. Rech, V. Sittinger, B. Szyszka, R. Geyer, P. Lechner, H. Schade, M. Ruske, G. Dittmar, H.P. Bochem, Thin Solid Films 442, 158 (2003) 5. J. Yang, B. Yan, S. Guha, Thin Solid Films 487, 162 (2005) 6. R. Schlatmann, B. Stannowski, E.A.G. Hamers, J.M. Lenssen, A.G. Talma, G.C. Dubbeldam, G.J. Jongerden, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 744 7. A. Madan, Surf. Coat. Technol. 200, 1907 (2005) 8. A. Shah, H. Schade, M. Vanecek, J. Meier, E. Vallat-Sauvain, N. Wyrsch, U. Kroll, C. Droz, J. Bailat, Prog. Photovolt. Res. Appl. 12, 113 (2004) 9. R.W. Miles, Vacuum 80, 1090 (2006) 10. J.M. Woodcock, H. Schade, H. Maurus, B. Dimmler, J. Springer, A. Ricaud, in Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 1997, p. 857

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

407

11. M.S. Keshner, R.R. Arya, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 149 12. J. Morris, R.R. Arya, J.G. ODowd, D. Wiedemann, J. Appl. Phys. 67, 1079 (1990) 13. C. Beneking, B. Rech, T. Eickhoﬀ, Y.G. Michael, N. Schultz, H. Wagner, in Proceedings of the 12th European Photovoltaic Solar Energy Conference, Amsterdam, The Netherlands, 1994, p. 683 14. D. Fischer, S. Dubail, J.A. Anna Selvan, N. Pellaton Vaucher, R. Platz, C. Hof, U. Kroll, J. Meier, P. Torres, H. Keppner, N. Wyrsch, M. Goetz, A. Shah, K.D. Ufert, in Proceedings of the 25th IEEE Photovoltaic Specialists Conference, Washington DC, USA, 1996, p. 1053 15. N. Pellaton Vaucher, J.L. Nagel, R. Platz, D. Fischer, A. Shah, in Proceedings of the 2nd World Conference on Photovoltaic Solar Energy Conversion, Vienna, Austria, 1998, p. 728 16. K. Yamamoto, A. Nakajima, M. Yoshimi, T. Sawada, S. Fukuda, T. Suezaki, M. Ichikawa, Y. Koi, M. Goto, T. Meguro, T. Matsuda, M. Kondo, T. Sasaki, Y. Tawada, Sol. Energy 77, 939 (2004) 17. J. M¨ uller, B. Rech, J. Springer, M. Vanecek, Sol. Energy 77, 917 (2004) 18. D.L. Staebler, C.R. Wronski, Appl. Phys. Lett. 31, 292 (1977) 19. J. M¨ uller, G. Sch¨ ope, B. Rech, H. Schade, P. Lechner, R. Geyer, H. Stiebig, W. Reetz, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion (WCPEC-3), Osaka, Japan, 2003, p. 1839 20. T. Minami, H. Sato, H. Nanto, S. Takata, Thin Solid Films 176, 277 (1989) 21. O. Vetterl, F. Finger, R. Carius, P. Hapke, L. Houben, O. Kluth, A. Lambertz, A. M¨ uck, B. Rech, H. Wagner, Sol. Energ. Mater. 62, 97 (2000) 22. J. Meier, E. Vallat-Sauvain, S. Dubail, U. Kroll, J. Dubail, S. Golay, L. Feitknecht, P. Torres, S. Fa¨ y, D. Fischer, A. Shah, Sol. Energ. Mater. 66, 73 (2001) 23. C. Beneking, W.A. Nositschka, O. Kluth, G. Sch¨ope, F. Birmans, H. Siekmann, B. Rech, in Proceedings of the 16th European Photovoltaic Solar Energy Conference, vol. II, Glasgow, UK, 2000, p. 1230 24. K.V. Maydell, H. Windgassen, W.A. Nositschka, U. Rau, P.J. Rostan, J. Henze, J. Schmidt, M. Scherﬀ, W. Fahrner, D. Borchert, S. Tordon, H. Stiebig, M. Schmidt, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 2005, p. 822 25. R.A. Street, Hydrogenated Amorphous Silicon, (Cambridge University Press, Cambridge, 1991) 26. R.A. Street, Technology and Applications of Amorphous Silicon (Springer, Berlin Heidelberg New York, 2000) 27. H. Overhof, P. Thomas, Electronic Transport in Hydrogenated Amorphous Semiconductors (Springer, Berlin Heidelberg New York, 1989) 28. B. Rech, T. Roschek, J. M¨ uller, S. Wieder, H. Wagner, Sol. Energ. Mater. Sol. Cell. 66, 267 (2001) 29. H. Keppner, J. Meier, P. Torres, D. Fischer, A. Shah, Appl. Phys. A 69, 169 (1999) 30. S. Klein, F. Finger, R. Carius, H. Stiebig, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, p. 1579 31. F. Meillaud, E. Vallat-Sauvain, X. Niquille, M. Dubey, J. Bailat, A. Shah, C. Ballif, in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, USA, 2005, p. 1412

408

J. H¨ upkes et al.

32. R. Carius, Photovoltaic and Photoactive Materials – Properties, Technology and Applications (Kluwer, Dordrecht, 2002), p. 93 33. J. Springer, A. Poruba, L. Mullerova, M. Vanecek, O. Kluth, B. Rech, J. Appl. Phys. 95, 1427 (2004) 34. Y. Ichikawa, T. Yoshida, T. Hama, H. Sakai, K. Harashima, Sol. Energ. Mater. Sol. Cell. 66, 107 (2001) 35. R.E.I. Schropp, E. Middelman, E. van Andel, H. Meiling, C.H.M. van der Werf, P.M.G.M. Peters, L.V. de Jonge-Meschaninova, J. Winkeler, R.J. Severens, M. Zeman, M.C.M. van de Sanden, A. Kuipers, C.I.M.A. Spee, G.J. Jongerden, in Proceedings of the 2nd World Conference on Photovoltaic Solar Energy Conversion, Wien, Austria, 1998, p. 820 36. M.N. van den Donker, T. Repmann, H. Stiebig, F. Finger, B. Rech, B. Stannowski, R. Schlatmann, G.J. Jongerden, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 857 37. Y. Uchida, in Proceedings of the 7th European Photovoltaic Solar Energy Conference, Seville, Spain, 1986, p. 395 38. H. Sakai, Y. Ichikawa, J. Non-Cryst. Solids 137–138, 1155 (1991) 39. J. Yang, A. Banerjee, S. Guha, Appl. Phys. Lett. 70, 2975 (1997) 40. G. Nakamura, K. Sato, Y. Yukimoto, in Proceedings of the 16th IEEE Photovoltaic Specialists Conference, New York, USA, 1982, p. 1331 41. Y. Tawada, H. Yamagishi, K. Yamamoto, Sol. Energ. Mater. Sol. Cell. 78, 647 (2003) 42. Sharp develops mass-production technology for triple-junction thin-ﬁlm solar cells, Press release, January 24, 2007, http://www.sharp-world.com/ corporate/news/070124.html. Last check 23 Apr 2007 43. Q-Cells AG, Brilliant den Grundstein gelegt, Press release, October, 11, 2006, http://www.q-cells.com/cmadmin 1 1055 0.html. Last check 23 Apr 2007 44. S. Wieder, M. Liehr, T. Repmann, B. Rech, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 145 45. B. Rech, T. Repmann, M.N. van den Donker, M. Berginski, T. Kilper, J. H¨ upkes, S. Calnan, H. Stiebig, S. Wieder, Thin Solid Films 511–512, 548 (2006) 46. T. Takagi, M. Ueda, N. Ito, Y. Watabe, H. Sato, K. Sawaya, Thin Solid Films 502, 50 (2006) 47. J. Meier, U. Kroll, T. Roschek, J. Spitznagel, S. Benagli, C. Ellert, G. Androutsopoulos, A. H¨ ugli, W. Stein, J. Springer, O. Kluth, D. Borello, M. Poppeller, G. B¨ uchel, A. B¨ uchel, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 238 48. T. Repmann, B. Sehrbrock, C. Zahren, H. Siekmann, J. H¨ upkes, B. Rech, W. Psyk, R. Geyer, P. Lechner, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 1574 49. T. Repmann, J. Kirchhoﬀ, W. Reetz, F. Biermans, J. M¨ uller, B. Rech, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 1843 50. H. Stiebig, W. Reetz, C. Haase, T. Repmann, B. Rech, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 561

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

409

51. H. Stiebig, W. Reetz, C. Zahren, T. Repmann, B. Rech, in Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, USA, 2006, p. 1521 52. W. Beyer, J. H¨ upkes, H. Stiebig, Transparent conducting oxide ﬁlms for thin ﬁlm silicon photovoltaics. to be published at ICMCTF, San Diego (2007) 53. M. Kubon, E. B¨ ohmer, F. Siebke, B. Rech, C. Beneking, H. Wagner, Sol. Energ. Mater. Sol. Cell. 41–42, 485 (1996) 54. E. B¨ ohmer, F. Siebke, B. Rech, C. Beneking, H. Wagner, Mater. Res. Soc. Symp. Proc. 426, 519 (1996) 55. H. Stiebig, F. Siebke, W. Beyer, C. Beneking, B. Rech, H. Wagner, Sol. Energ. Mater. Sol. Cell. 48, 351 (1997) 56. J. M¨ uller, O. Kluth, S. Wieder, H. Siekmann, G. Sch¨ope, W. Reetz, O. Vetterl, D. Lundszien, A. Lambertz, F. Finger, B. Rech, H. Wagner, Sol. Energ. Mater. Sol. Cell. 66, 275 (2001) 57. H. Ibach, H. L¨ uth, Solid State Physics (Springer, Berlin Heidelberg New York, 2003) 58. J. Springer, B. Rech, W. Reetz, J. M¨ uller, M. Vanecek, Sol. Energ. Mater. 85, 1 (2005) 59. H. Stiebig, A. Kresiel, K. Winz, N. Schultz, C. Beneking, T. Eickhoﬀ, H. Wagner, in Proceedings of the 1st World Conference on Photovoltaic Solar Energy Conversion, Waikaloa, USA, 1994, p. 603 60. F. Leblanc, J. Perrin, J. Schmitt, J. Appl. Phys. 75, 1074 (1994) 61. Y. Hishikawa, T. Kinoshita, M. Shima, M. Tanaka, S. Kiyama, S. Tsuda, S. Nakano, in Proceedings of the 26th IEEE Photovoltaic Specialists Conference, Anaheim, USA, 1997, p. 615 62. Y. Hishikawa, H. Tari, S. Kiyama, Technical Digest of the 11th International Photovoltaic Science and Engineering Conference, Sapporo, Japan, 1999, p. 219 63. M. Zeman, R.A.C.M.M. van Swaaij, J.W. Metselaar, R.E.I. Schropp, J. Appl. Phys. 88, 6436 (2000) 64. J. Krˇc, M. Zeman, F. Smole, M. Topiˇc, J. Appl. Phys. 92, 749 (2002) 65. J. Springer, B. Rech, T. Repmann, H. Siekmann, W. Reetz, J. M¨ uller, M. Ruske, M. Vanecek, in Proceedings of the 17th European Photovoltaic Solar Energy Conference, M¨ unchen, Germany, 2002, p. 2869 66. H. Stiebig, M. Schulte, C. Zahren, C.S. Haase, B. Rech, P. Lechner, in Proceedings of the Photonics Europe – Photonics for Solar Energy Systems, Vol. 6197, Strasbourg, France, 2006, p. 1 67. C. Haase, H. Stiebig, Prog. Photovolt. Res. Appl. 14, 629 (2006) 68. P.A. Basore, in Proceedings of the 29th IEEE Photovoltaic Specialists Conference, New Orleans, USA, 2002, p. 49 69. Y. Gupta, H. Liers, S. Woods, S. Young, R. DeBlasio, L. Mrig, in Proceedings of the 16th IEEE Photovoltaic Specialists Conference, San Diego, USA, 1982, p. 1092 70. C. Beneking, B. Rech, S. Wieder, O. Kluth, H. Wagner, W. Frammelsberger, R. Geyer, P. Lechner, H. R¨ ubel, H. Schade, Thin Solid Films 351, 243 (1999) 71. K. Sato, Y. Gotoh, Y. Wakayama, Y. Hayashi, K. Adachi, H. Nishimura, Reports Res. Lab., Asahi Glass Co. Ltd., 1992 72. P.F. Gerhardinger, R.J. Mccurdy, Mater. Res. Soc. Symp. Proc. 426, 399 (1996)

410

J. H¨ upkes et al.

73. M. Kambe, M. Fukawa, N. Taneda, Y. Yoshikawa, K. Sato, K. Ohki, S. Hiza, A. Yamada, M. Konagai, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 1812 74. B. Kuhn, D. Le Bellac, B. Rech, J. H¨ upkes, G. Sch¨ ope, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, p. 1733 75. T. Uchino, K. Nakaguchi, Y. Nagashima, T. Kondo, J. Non-Cryst. Solids 261, 72 (2000) 76. M. Hirata, A. Fujisawa, T. Otani, Y. Nagashima, K. Kiyohara, M. Hyodo, in Proceedings of the 3rd International Conference on Coatings on Glass, Maastricht, The Netherlands, 2001, p. 395 77. U. Blieske, T. Droege, P. Gayout, M. Neander, D. Neumann, A. Prat, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion (WCPEC-3), Osaka, Japan, 2003, p. 188 78. L. Yang, M. Bennett, L. Chen, K. Jansen, J. Kessler, Y. Li, J. Newton, K. Rajan, S. Willing, R. Arya, D. Carlson, Mat. Res. Soc. Symp. Proc. 420, 839 (1996) 79. R. van den Berg, H. Calwer, P. Marklstorfer, R. Meckes, F.W. Schulze, K.D. Ufert, H. Vogt, Sol. Energ. Mater. 31, 253 (1993) 80. S. Fa¨ y, S. Dubail, U. Kroll, J. Meier, Y. Ziegler, A. Shah, in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, 2000, p. 361 81. R. Groenen, J. Loeﬄer, P.M. Sommeling, J.L. Linden, E.A.G. Hamers, R.E.I. Schropp, M.C.M. van de Sanden, Thin Solid Films 392, 226 (2001) 82. N. Toyama, R. Hayashi, Y. Sonoda, M. Iwata, Y. Miyamoto, H. Otoshi, K. Saito, K. Ogawa, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 1601 83. N. Senoussaoui, M. Krause, J. M¨ uller, E. Bunte, T. Brammer, H. Stiebig, Thin Solid Films 451–452, 397 (2004) 84. N.P. Harder, D. Le Bellac, E. Royer, B. Rech, G. Sch¨ope, J. M¨ uller, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, p. 1355 85. J. Ji, P.A. Basore, R. Evans, International Patent WO 03/019676 A1, 2003 86. T. Nakada, Y. Ohkubo, A. Kunioka, Jpn. J. Appl. Phys. 30, 3344 (1991) 87. H. Sato, T. Minami, S. Takata, T. Mouri, N. Ogawa, Thin Solid Films 220, 327 (1992) 88. J.A. Anna Selvan, H. Keppner, A. Shah, Mater. Res. Soc. Symp. Proc. 426, J9.4 (1996) 89. O. Kluth, Dissertation, Rheinisch-Westf¨ ahlische Technische Hochschule (1996) 90. A. L¨ oﬄ, S. Wieder, B. Rech, O. Kluth, C. Beneking, H. Wagner, in Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 1997, p. 2089 91. O. Kluth, A. L¨ oﬄ, S. Wieder, C. Beneking, W. Appenzeller, L. Houben, B. Rech, H. Wagner, S. Hoﬀmann, R. Waser, J.A. Anna Selvan, H. Keppner, in Proceedings of the 26th IEEE Photovoltaic Specialists Conference, Anaheim, USA, 1997, p. 715 92. O. Kluth, B. Rech, L. Houben, S. Wieder, G. Sch¨ ope, C. Beneking, H. Wagner, Thin Solid Films 351, 247 (1999) 93. J. H¨ upkes, B. Rech, S. Calnan, O. Kluth, U. Zastrow, H. Siekmann, M. Wuttig, Thin Solid Films 502, 286 (2006)

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

411

94. B. Rech, S. Wieder, C. Beneking, A. L¨ oﬄ, O. Kluth, W. Reetz, H. Wagner, in Proceedings of the 26th IEEE Photovoltaic Specialists Conference, Anaheim, USA, 1997, p. 619 95. O. Kluth, O. Vetterl, R. Carius, F. Finger, S. Wieder, B. Rech, H. Wagner, Mater. Res. Soc. Symp. Proc. 557, 731 (1999) 96. O. Kluth, G. Sch¨ ope, J. H¨ upkes, C. Agashe, J. M¨ uller, B. Rech, Thin Solid Films 442, 80 (2003) 97. C. Agashe, O. Kluth, J. H¨ upkes, U. Zastrow, B. Rech, M. Wuttig, J. Appl. Phys. 95, 1911 (2004) 98. T.J. Coutts, D.L. Young, X. Li, W.P. Mulligan, X. Wu, J. Vac. Sci. Technol. A 18, 2646 (2000) 99. J. H¨ upkes, M. Berginski, V. Sittinger, F. Ruske, B. Rech, H. Siekmann, J. Kirchhoﬀ, B. Zwaygardt, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, p. 1779 100. T. Minami, H. Nanto, S. Takata, Jpn. J. Appl. Phys. 23, L280 (1984) 101. B. Szyszka, Thin Solid Films 351, 164 (1999) 102. K. Ellmer, J. Phys. D Appl. Phys. 34, 3097 (2001) 103. Abrisa Corporate Oﬃce, Data sheet and product information: Corning 1737, http://www.abrisa.com/downloads/corning 1737F.pdf. Last check 23 Apr 2007 104. J.M. Ting, B.S. Tsai, Mater. Chem. Phys. 72, 273 (2001) 105. L.J. van der Pauw, Research Report 1, Phillips, 1958 106. E. Burstein, Phys. Rev. 93, 632 (1954) 107. T.S. Moss, Proc. Phys. Soc. B 67, 775 (1954) 108. K.L. Chopra, S. Major, D.K. Pandya, Thin Solid Films 102, 1 (1983) 109. A.N. Mariano, R.E. Hanneman, J. Appl. Phys. 34, 384 (1963) 110. G. Heiland, P. Kunstmann, Surf. Sci. 13, 72 (1969) 111. F.C.M. van de Pol, F.R. Blom, T.J.A. Popma, Thin Solid Films 204, 349 (1991) 112. Y. Li, G.S. Tompa, S. Liang, C. Gorla, Y. Lu, J. Doyle, J. Vac. Sci. Technol. A 15, 1063 (1997) 113. H. Maki, T. Ikoma, I. Sakaguchi, N. Ohashi, H. Haneda, J. Tanaka, N. Ichinose, Thin Solid Films 411, 91 (2002) 114. J. Zhu, N.W. Emanetoglu, Y. Chen, B.V. Yakshinskiy, Y. Lu, J. Electron. Mater. 33, 556 (2004) 115. J.G.E. Gardeniers, Z.M. Rittersma, G.J. Burger, J. Appl. Phys. 83, 7844 (1998) 116. H.C. Gatos, J. Appl. Phys. 32, 1232 (1961) 117. W. Jo, S.J. Kim, D.Y. Kim, Acta Mater. 53, 4185 (2005) 118. J. H¨ upkes, Master thesis (2001) 119. Y. Mai, Ph.D. thesis, Nankai University (2006) 120. J.A. Thornton, J. Vac. Sci. Technol. 11, 666 (1974) 121. J.A. Thornton, J. Vac. Sci. Technol. A 4, 3059 (1986) 122. J. H¨ upkes, B. Rech, O. Kluth, T. Repmann, B. Sehrbrock, J. M¨ uller, R. Drese, M. Wuttig, Sol. Energ. Mater. Sol. Cell. 90(18–19), 3054 (2006) 123. M. Berginski, B. Rech, J. H¨ upkes, H. Stiebig, M. Wuttig, J. Appl. Phys. 101, 074903 (2007) 124. J. H¨ upkes, Ph.D. thesis, Rheinisch-Westf¨ahlische Technische Hochschule (2005)

412

J. H¨ upkes et al.

125. J. Str¨ umpfel, G. Beister, D. Schulze, M. Kammer, S. Rehn, in Proceedings of the 40th Annual Technical Conference of the Society of Vacuum Coaters, New Orleans, USA, 1997, p. 179 126. O. Kluth, G. Sch¨ ope, B. Rech, R. Menner, M. Oertel, K. Orgassa, H.W. Schock, Thin Solid Films 502, 311 (2006) 127. K.J. Nordheden, in Proceedings of the SPIE – Quantum Sensing and Nanophotonic Devices, Vol. 5359, San Jose, USA, 2004, ed. by M. Razeghi, G.J. Brown 128. M. Hiramatsu, K. Imaeda, N. Horio, M. Nawata, J. Vac. Sci. Technol. A 16, 669 (1998) 129. K. Ip, M.E. Overberg, K.W. Baik, R.G. Wilson, S.O. Kucheyev, J.S. Williams, C. Jagadish, F. Ren, Y.W. Heo, D.P.N.J.M. Zavada, S.J. Pearton, Solid State Electron. 47, 2289 (2003) 130. J.W. Bae, C.H. Jeong, H.K. Kim, K.K. Kim, N.G. Cho, T.Y. Seong, S.J. Park, I. Adesida, G.Y. Yeom, Jpn. J. Appl. Phys. 42, L553 (2003) 131. K. Tominaga, K. Kuroda, O. Toda, Jpn. J. Appl. Phys. 27, 1176 (1988) 132. T. Minami, H. Nanto, H. Sato, S. Takata, Thin Solid Films 164, 275 (1988) 133. J. H¨ upkes, B. Rech, O. Kluth, J. M¨ uller, H. Siekmann, C. Agashe, H.P. Bochem, M. Wuttig, Mat. Res. Soc. Symp. Proc. 762, A7.11 (2003) 134. T. Yamamoto, T. Shiosaki, A. Kawabata, J. Appl. Phys. 51, 3113 (1980) 135. A. Klein, Termination related features of ZnO etching behavior, private communication, 2005 136. H. Jebsen-Marwedel, R. Br¨ uckner, Glastechnische Fabrikationsfehler (Springer, Berlin Heidelberg New York, 1980) 137. H.J. Gl¨ aser, Large Area Glass Coating, vol. 1 (von Ardenne, Dresden, 2000) 138. W. Meisel, Glass Sci. Technol. 72, 291 (1999) 139. C. Wang, K. Appelfeller, T. Giese, G. Krausch, in Tagungsband der 78. Glastechnischen Tagung, N¨ urnberg, Germany, 2004, p. 176 140. S. K¨ urbitz, T. Rainer, K.J. Berg, G. Berg, in Tagungsband der 78. Glastechnischen Tagung, N¨ urnberg, Germany, 2004, p. 161 141. H.K. Pulker, Coatings on Glass (Elsevier, Amsterdam Oxford New York Tokyo, 1984) 142. V. Sittinger, F. Ruske, W. Werner, B. Szyszka, B. Rech, J. H¨ upkes, G. Sch¨ ope, H. Stiebig, Thin Solid Films 496, 16 (2006) 143. J. M¨ uller, G. Sch¨ ope, O. Kluth, B. Rech, M. Ruske, J. Trube, B. Szyszka, X. Jiang, G. Br¨ auer, Thin Solid Films 392, 327 (2001) 144. D.A.G. Bruggeman, Annalen der Physik 24, 636 (1935) 145. P. Lechner, R. Geyer, H. Schade, B. Rech, O. Kluth, H. Stiebig, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, p. 1591 146. O. Kluth, C. Zahren, H. Stiebig, B. Rech, H. Schade, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, p. 1587 147. M. Vanecek, J. Springer, A. Poruba, O. Kluth, T. Repmann, B. Rech, N. Wyrsch, J. Meier, A. Shah, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 1527 148. K.L. Chopra, S.R. Das, Thin Film Solar Cells (Plenum, New York, 1983) 149. M.A. Green, Silicon Solar Cells – Advanced Principles & Practice (Centre for Photovoltaic Devices and Systems, Sydney, 1995) 150. T. Brammer, J. H¨ upkes, M. Krause, O. Kluth, J. M¨ uller, B. Rech, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 176

8 Texture Etched ZnO:Al for Silicon Thin Film Solar Cells

413

151. M. Berginski, B. Rech, J. H¨ upkes, G. Sch¨ ope, M.N. van den Donker, W. Reetz, T. Kilper, M. Wuttig, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, p. 1539 152. B. Rech, G. Sch¨ ope, O. Kluth, T. Repmann, T. Roschek, J. M¨ uller, J. H¨ upkes, H. Stiebig, in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, 2003, p. 2783 153. Y. Mai, S. Klein, R. Carius, H. Stiebig, X. Geng, F. Finger, Appl. Phys. Lett. 87, 073503 (2005) 154. T. Repmann, T. Kilper, W. Appenzeller, C. Zahren, H. Stiebig, B. Rech, in Proceedings of the 31rd IEEE Photovoltaic Specialist Conference, Orlando, USA, 2005, p. 1383 155. B. Rech, T. Repmann, J. H¨ upkes, M. Berginski, H. Stiebig, W. Beyer, V. Sittinger, F. Ruske, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 2005, p. 1481 156. V. Sittinger, F. Ruske, W. Werner, B. Szyszka, B. Rech, G. Sch¨ope, J. H¨ upkes, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, 2005, p. 1667 157. T. Tohsophon, J. H¨ upkes, H. Siekmann, B. Rech, M. Schultheis, N. Sirikulrat, in Proceedings of the 6th International Conference on Coatings on Glass and Plastics, Dresden, Germany, 2006, p. 397 158. U. Kroll, J. Meier, S. Benagli, T. Roschek, J. Spitznagel, A. Huegli, D. Borello, M. Mohr, O. Kluth, D. Zimin, G. Monteduro, J. Springer, G. Androutsopoulos, C. Ellert, W. Stein, G. Buechel, A. Zindl, A. Buechel, D. Koch-Ospelt, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, p. 1546 159. K. Yamamoto, A. Nakajima, M. Yoshimi, T. Sawada, S. Fukuda, T. Suezaki, M. Ichikawa, Y. Koi, M. Goto, T. Meguro, T. Matsuda, M. Kondo, T. Sasaki, Y. Tawada, in Technical Digest of the 15th International Photovoltaic Science and Engineering Conference, Shanghai, China, 2005, p. 529 160. B. Yan, G. Yue, J.M. Owens, J. Yang, S. Guha, in Proceedings of the 4th World Conference on Photovoltaic Solar Energy Conversion, Waikaloa, USA, 2006, p. 1477 161. H. Takatsuka, Y. Yamauchi, M. Fukagawa, H. Mashima, K. Kawamura, K. Yamaguchi, T. Nishimiya, Y. Takeuchi, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, 2006, p. 1531

9 Chalcopyrite Solar Cells and Modules R. Klenk

Chalcopyrite-based solar modules are uniquely combining advantages of thinﬁlm technology with the eﬃciency and stability of conventional crystalline silicon cells. It is therefore believed that chalcopyrite based modules can take up a large part of the PV market growth once true mass production is started. The eﬃciency of lab-scale thin-ﬁlm devices is close to 20% [2], an eﬃciency comparable to the best multicrystalline silicon cells. Many scaling-up and manufacturing issues have been resolved. Pilot production lines are operational and modules are commercially available. As of 2006, the market share of chalcopyrite photovoltaic (PV) modules is not yet signiﬁcant but major problems that might prevent further commercialization have not been identiﬁed. The ﬁrst chalcopyrite-based solar cell has been published in 1974 [3]. The cell was prepared from a p-type CuInSe2 (CISe) single crystal onto which a CdS ﬁlm was evaporated in vacuum. This combination of a p-type chalcopyrite absorber and a wide-gap n-type window layer still is the basic concept upon which current cell designs are based. The typical design, ﬁrst described in 1985 [4] is shown in Figs. 9.1 and 9.2. The CuInSe2 crystal is replaced by a polycrystalline thin ﬁlm of the more general composition Cu(In,Ga)(S,Se)2 . The thick CdS ﬁlm is replaced by a very thin, typically 50 nm, CdS ﬁlm (buﬀer layer) and a stack of undoped (i-ZnO) and highly doped ZnO (TCO). Cells and modules are prepared in substrate conﬁguration, i.e., the nontransparent metal back contact is the ﬁrst ﬁlm to be deposited onto the substrate whereas the transparent ZnO front contact is deposited last. The requirements for ZnO in chalcopyrite-based photovoltaics are derived from its major functions within the device: – Formation of the heterojunction – Lateral current transport – Passing illumination from the cell surface to the absorber

9.1 Heterojunction Formation The CdS ﬁlm used in the ﬁrst chalcopyrite cell is in some aspects an ideal heterojunction partner for CISe: the conduction bands are reasonably well aligned and the lattice constants match. On the other hand, its relatively low

416

R. Klenk

Fig. 9.1. Schematic cross-section of a chalcopyrite-based thin-ﬁlm solar cell. Typical materials for the individual parts of the cell are given in square brackets

Fig. 9.2. Scanning electron micrograph of the cross-section of a typical chalcopyrite solar cell with Cu(In,Ga)Se2 (CIGSe) absorber (substrate now shown). Reprinted with permission from [1]

band gap (Eg ≈ 2.4 eV) causes signiﬁcant losses in photocurrent and its conductivity is limited. In an eﬀort to improve transparency and conductivity of the window layer without loosing the advantages of CdS concerning junction formation, a very thin CdS ﬁlm (buﬀer) was combined with a thick ZnO ﬁlm [4]. At the same time, the CdS preparation technique was changed from evaporation to chemical bath deposition (CBD). Because of the ideal step

9 Chalcopyrite Solar Cells and Modules

417

coverage of CBD, the rough absorber surface can be covered completely even by a very thin ﬁlm. A cross section of an actual device is shown in Fig. 9.2. It may appear from this description that the ZnO plays a role only in forming an nn-heterojunction, which is not critical in terms of cell performance. However, if the absorber/buﬀer interface and the buﬀer itself do not hold suﬃcient positive electrical charge to balance the space charge region of the absorber, the ZnO properties will inﬂuence the Fermi-level position at the pn-junction. It can be shown that this parameter is of crucial importance concerning device performance [5]. In general, the Fermi-level should be close to the conduction band of the absorber at the interface in order to minimise the density of holes available for recombination. Band line-up, buﬀer doping, and interface charge in chalcopyrite-based heterojunctions are under discussion and may vary depending on materials and preparation techniques. An example for a situation where the TCO does play a role in interface formation is shown in Fig. 9.3a. Here, a diﬀerent band line-up between buﬀer and TCO as well as diﬀerent doping of the TCO would inﬂuence the Fermi level position at the absorber/buﬀer junction and, hence, the performance of the device. There are indications that in most cases the situation is closer to the examples shown in Fig. 9.3c,b, i.e., that the absorber/buﬀer interface and/or the buﬀer bulk determine the interface Fermi level. In these cases, a ﬂuctuation in TCO properties will have only a small inﬂuence on the interface recombination rate. Further models for the chalcopyrite-based heterojunction are suggested in literature [6–8]. It is postulated that cadmium diﬀuses into the absorber surface, converts it to an n-type semiconductor, and causes a buried junction. It is also postulated that this eliminates interface recombination losses due to the spatial separation of the pn-junction and heterojunction. However, calculations reveal that two cases have to be distinguished, none of them necessarily resulting in improved device performance. In the ﬁrst case, the n-type layer is too shallow or the donor density is too small to equal the charge in the p-type absorber. The surface type conversion will have little inﬂuence and the situation is – in principle – still the same as described above and shown in Fig. 9.3a–c. In the second case the n-type layer can hold enough charge, the band diagram will resemble the situation shown in Fig. 9.3d, and the TCO will have little inﬂuence on device performance. But the holes photo-generated within the n-type part of the absorber will recombine at the interface to the buﬀer layer (where they are minority carriers). In analogy to a conventional homojunction solar cell with deep emitter and poorly passivated surface, this reduces the blue response of the cell. A high barrier in the conduction band due to unfavorable band line-up or interface charge at the buﬀer/TCO junction may impede the majority carrier transport at this nn-junction. There are no indications for a signiﬁcant barrier in standard cells but it has been made responsible for the poor performance of certain cells with alternative Cd-free buﬀer layers [9].

418

R. Klenk 1

1 EC

–1

EC 0

EF Chalcopyrite Absorber

–3.29

EV

–2

–3.3 2.1

Buffer

2.105

Energy (eV)

Energy (eV)

0

Chalcopyrite Absorber EV

–2

Buffer

–3

–3

(a) –4

–1

EF

1.7

(b)

TCO 1.8

1.9

2

2.1

–4

2.2

TCO

1.7

1.8 1.9 2 2.1 Distance from back contact (µm)

Distance from back contact (µm)

(a) Lightly doped buﬀer, no interface charge. The ﬁeld zone extends through the buﬀer and into the TCO

(b) Higher doping in the buﬀer layer. The ﬁeld zone is limited to absorber and buﬀer 1

1

EC

EC 0

Chalcopyrite Absorber EV

–2

Buffer

–1 Chalcopyrite Absorber EV

–2

Buffer –3

–3

(c) –4

EF

EF

Energy (eV)

Energy (eV)

0 –1

2.2

1.7

(d) TCO

TCO 1.8 1.9 2 2.1 Distance from back contact (µm)

2.2

(c) Positive interface charge causes inversion of the absorber surface

–4

1.5

1.6

1.7 1.8 1.9 2 2.1 Distance from back contact (µm)

2.2

(d) Buried junction. The ﬁeld zone is limited to the absorber

Fig. 9.3. Band diagrams of chalcopyrite/buﬀer/TCO heterojunctions illustrating the inﬂuence of buﬀer doping and interface charge

9.1.1 Why Use an Undoped ZnO Layer? In the standard conﬁguration, the ZnO ﬁlm next to the CdS is sputtered from an undoped target or prepared by MOCVD without a dopant gas. It may also be sputtered from the same doped ZnO target that is used for the deposition of the conductive layer but with additional oxygen in the working gas. This ﬁlm is commonly referred to as i-ZnO, which may be misleading because native defects cause a fairly high carrier density. Nevertheless, in view of the considerations made in the previous paragraph, depositing an i-ZnO ﬁlm onto the CdS appears to be counter-productive. In addition, using undoped as well as doped targets increases production cost, in particular because the undoped ceramic target requires RF-sputtering. Ruckh et al. [10] have varied the thickness of the sputtered i-ZnO layer between 0 and 250 nm and have found no inﬂuence on open-circuit voltage and ﬁllfactor of eﬃcient lab-scale cells. The inﬂuence of sputtering parameters such as total pressure, argon and oxygen ﬂow was also not signiﬁcant. The authors

9 Chalcopyrite Solar Cells and Modules

419

conclude that the i-ZnO does not at all participate in junction formation. The only eﬀect noted was a higher stability at 200◦ C in air, i.e., a certain encapsulation or diﬀusion-barrier function of the i-ZnO layer. Similar experiments carried out by Kessler et al. conﬁrm that high eﬃciencies can be achieved when omitting the undoped ZnO ﬁlm [11], but presumably only if the CdS buﬀer is suﬃciently thick [12]. Kessler et al. also report a better stability of cells with the undoped ZnO in accelerated ageing (damp heat test), which agrees with the encapsulation eﬀect found by Ruckh et al. Finally, Ramanathan et al. report identical parameters of high eﬃciency cells with and without i-ZnO, even for the case of thin CdS buﬀers [13]. The stability was not assessed. In spite of the described ﬁndings it is generally believed that – at least on module level – reproducibility and production-yield proﬁt from inclusion of i-ZnO. It is obvious that the inﬂuence of localized ﬂaws in the absorber ﬁlm, such as pin-holes, is more severe if those are directly in contact with the highly doped contact layer [14]. Similarly, if the device properties are slightly inhomogeneous on a microscopic scale, the i-ZnO may be beneﬁcial in terms of performance. Such inhomogeneities could be caused, e.g., by lateral bandgap ﬂuctuations in the absorber ﬁlm. In this case, the optimum resistivity of the i-ZnO will depend on the amount of ﬂuctuations present [15].

9.2 Transparent Front Contact While the ZnO double-layer may play a certain role in establishing the heterojunction, its major function is the lateral current transport to the contacts and the transmission of solar radiation into the absorber, i.e., acting as a transparent front contact. Unfortunately, high transparency and lowest sheet resistance are mutually exclusive ZnO properties. A given ZnO preparation process is therefore best described by plotting the achievable photocurrent vs. sheet resistance (Fig. 9.4). Transparency and resistance can be combined to give a ﬁgure of merit describing the performance of a ZnO ﬁlm [16]. Small and medium area chalcopyrite solar cells are prepared for lab-scale testing and some special applications. Here the current collection is typically assisted by a metal grid deposited on top of the ZnO ﬁlm, which relaxes the requirements for ZnO conductance. Depending on the spacing of the grid ﬁngers, the ZnO can be fairly thin (<0.5 µm), which is beneﬁcial in terms of optical losses (see below). General experience shows that nickel makes a good and stable ohmic contact to highly doped ZnO. Additional metals such as aluminum are often used on top of a thin nickel ﬁlm to reduce the grid resistance. However, large area modules are generally scribed into multiple cells which are monolithically connected in series and do not use a metal grid.

420

R. Klenk

Current loss (mA/cm2)

3

E G = 1.16 eV E G = 1.25 eV E G = 1.45 eV

2

1

0

10

15 20 25 30 Sheet resistance (Ohm/square)

35

Fig. 9.4. Typical loss of current density as a function of ZnO sheet resistance calculated for three values of the absorber bandgap (assuming air mass 1.5 illumination, redrawn with permission from [12])

9.2.1 Monolithic Integration The connection is made from the molybdenum back contact to the ZnO during ZnO deposition (Fig. 9.5). The common scheme requires three patterning steps: an isolation scribe in the molybdenum back contact (P1), an interconnect scribe of the absorber to create a gap, which is later ﬁlled by ZnO (P2), and an isolation scribe of the complete cell structure down to the molybdenum (P3). While the preferred tool for P1 patterning is a pulsed Nd-YAG laser, photo lithographic patterning is also possible. After laser patterning the substrate is subject to wet cleaning with rotating brushes to remove loose particles. P2 and P3 patterning are carried out by mechanical scribing. Appropriate tools must be used in order to avoid damaging the underlying molybdenum ﬁlm. P2 patterning can be carried out before or after deposition of the undoped ZnO layer, the latter method may give a better contact between ZnO and molybdenum in P2. Optimizing the ZnO requires a balance between conductance, transparency, and manufacturing cost. The series resistance of the device must be minimized to achieve a high ﬁll factor. It depends on carrier density and mobility in the ZnO, thickness of the ZnO ﬁlm, spacing of interconnects (grid ﬁnger spacing in case of cells), and the contact resistance between molybdenum and ZnO (nickel and ZnO). Higher doping may result in reduced mobility and higher optical losses, in particular in the long wavelength region (free carrier absorption/reﬂection). Increased thickness also aﬀects the transparency and increases deposition time and costs. On the other hand, higher ZnO sheet resistance requires reducing the interconnect or ﬁnger spacing which in turn results in losses of active area. A high mobility is desirable because it reduces the series resistance without adversely aﬀecting other parameters

9 Chalcopyrite Solar Cells and Modules

421

Fig. 9.5. Schematic cross-section of the cell interconnect in monolithic integration. This ﬁgure shows the variant where the P2 scribing is carried out after the deposition of the i-ZnO. Reprinted with permission from [1]

as long as the mobility can be achieved without costly modiﬁcations of the manufacturing process. Depending on the composition of the absorber, typical photocurrent densities under full illumination range from 20 (CuInS2 [19]) to 40 mA cm−2 (CISe [20]). Interconnect distance is 5–10 mm and causes a loss in active area of around 10%. The required ZnO thickness is around 1 µm. Nonideal ZnO transparency, series resistance, and loss in active area will cause an overall loss in module eﬃciency of about two percentage points compared to small area test cells prepared from the same materials. Results of a detailed calculation based on measured ZnO properties and carried out for diﬀerent photocurrent densities (chalcopyrite absorber bandgap) can be found in [12]. They show a clear advantage of wide gap absorbers in terms of module eﬃciency. A computer simulation has been made available to optimize the module patterning for a given set of ﬁlm properties [21]. Figure 9.6 shows data calculated by Wennerberg [17] using this program. Required input data are the ZnO sheet resistance and weighted optical transparency for varied thickness (in analogy to Fig. 9.4) as well as cell and interconnect properties. The data used here may be considered representative of the properties achievable in today’s pilot production of evaporated CIGSe absorbers and of ZnO ﬁlms sputtered from a ceramic target. Experimental veriﬁcation can be carried out using a straightforward method described by Klaer et al. [18]. Results for CuInS2 -based test structures are shown in Fig. 9.7. Given the current development status of ZnO thin ﬁlms, a metal grid will result in higher cell or module [22] eﬃciency because the grid shading losses can be over-compensated by improved ZnO transparency and lower series resistance. In spite of this, grids are not used in commercial production of modules, presumably due to prohibitive cost, added complexity, and aesthetical reasons. The ZnO ﬁlm plays an important role for module stability in accelerated lifetime testing under damp heat conditions, which forms a part of the

422

R. Klenk

Fig. 9.6. Calculated optimum interconnect distance and resulting module eﬃciency as a function of ZnO sheet resistance. Calculations are based on the properties of a small area CIGSe-based cell with 14.4% eﬃciency. Reprinted with permission from [17]

Fig. 9.7. CuInS2 -based module eﬃciency as a function of cell width for diﬀerent ZnO layer thicknesses. The data are based on measurements of single module cells and calculated for a module having interconnects of 0.6 mm width. Reprinted with permission from [18]

EN/IEC 61646 certiﬁcation. The lateral resistance tends to increase, giving rise to ﬁll factor losses. It is therefore mandatory to optimize ZnO preparation not only with respect to the as-grown properties but also by taking into account the degradation in damp heat (see Sect. 9.4).

9 Chalcopyrite Solar Cells and Modules

423

9.2.2 Optical Losses Chalcopyrite-based devices generally show rather high photocurrent densities even without antireﬂective coating. Hence, apart from the issues described above, not much consideration has been given to optical losses caused by the ZnO ﬁlms. The refractive indices within the device are not too far apart (ZnO ≈ 1.9, CdS ≈ 2.4, Absorber ≈ 2.9). Reﬂection at internal interfaces is therefore not severe. Interference fringes caused by the optical propagation in the window layer are sometimes noticeable in the spectral response of the device but only if the absorber is very smooth. Matching of refractive indices is also close to ideal for encapsulated devices. A simple single-layer antireﬂective coating deposited onto test structures without encapsulation will achieve maximum external quantum eﬃciencies of more than 95%. Light trapping does not play a signiﬁcant role. The high optical absorption in the chalcopyrites together with an absorber thickness in the range of 2–3 µm guarantees that the light is completely absorbed in a single passage through the absorber. This is likely to change in the future when absorber layers will be made thinner to save on raw materials and production cost. New optical concepts for the chalcopyrite device are therefore being investigated, including reﬂecting back contacts [23] and ZnO front contacts optimized for light-trapping. The quest for record eﬃciencies in the lab as well as applications where eﬃciency is of paramount importance, e.g., applications in space are other motivations to optimize the optical design [24–26].

9.3 Manufacturing Since in the chalcopyrite module the ZnO ﬁlms are the last to be deposited, the processing must be compatible with the remainder of the cell structure. This implies in particular that substrate temperatures must be limited to 200–250◦C [10] even though better ZnO properties could be achieved at higher deposition temperatures. Interdiﬀusion at the absorber/buﬀer interface has been made responsible for the instability [27] but it is believed that a detailed study using current state of the art material would be required to clarify this point. It is important to realize that the actual ﬁlm properties depend strongly on the substrate. A rough substrate (Fig. 9.8), such as absorber ﬁlms from two-stage preparation [28], leads to a signiﬁcant increase of the ZnO ﬁlm resistance. The use of witness samples based on ideally smooth glass substrates may be misleading. The throughput needs to be fairly high. Actual and projected module production capacities are ranging from 1 to 50 MWp/a per facility, which roughly corresponds to cycle times from 30 to below 1 min. Deposition of ZnO contributes signiﬁcantly to production costs [29]. In (pilot) production lines for chalcopyrite-based solar modules, ZnO is deposited by magnetron sputtering [29–31].

424

R. Klenk

Fig. 9.8. Scanning electron micrograph of a ZnO:Al/ZnO/CdS/CuInS2 solar cell. The larger structures correspond to the grains of the absorber layer

Fig. 9.9. Scanning electron micrograph of a ZnO:Al/ZnO/CdS/CuInS2 solar cell (cross section)

9 Chalcopyrite Solar Cells and Modules

425

9.3.1 Chemical Vapour Deposition1 The ﬁlm is prepared by reacting diethylzinc (DEZn) with water vapour at the surface of the substrate heated to moderate temperatures, typically in the range of 150–200◦C [32]. Inert gases are used to carry the reactants into the deposition chamber. Doping can be achieved by adding diborane to the atmosphere. Terzini et al. have compared the junction formation using sputtered and chemical vapour deposition (CVD) ZnO, in both cases without extrinsic doping, and have not identiﬁed signiﬁcant diﬀerences [33]. CVD of ZnO has been used in previous pilot production lines with good results [34].

9.3.2 Sputtering ZnO for cells and small modules can be prepared by RF-sputtering, which is known to yield good ﬁlm properties [35]. Results of an optimization study can be found in [11]. Fine-tuning must be carried out using the actual absorber, buﬀer layers, metal grids, and antireﬂective coating in order to optimize the complete system. The NREL group uses a combination of a 90 nm undoped and a 120 nm doped ﬁlm for their high eﬃciency devices [36]. The resistance is given as 60 Ω sq−1 but it is not clear whether this is measured in the actual device or on witness samples. For industrial production of large area modules, RF-sputtering of the doped ZnO is too slow and too expensive. RF-sputtering may still be used for deposition of the relatively thin undoped ZnO ﬁlm. The thicker, highly doped ﬁlm must be sputtered with DC or MF excitation of the plasma. DC sputtering from doped ceramic targets is feasible due to progress in target manufacturing and power supply technology (arc suppression). It is the preferred technology in pilot lines. Novel production methods have signiﬁcantly reduced the high costs of ceramic targets. Further progress in target utilization may be possible by using rotating targets. It is conceivable that DC sputtering from ceramic targets will remain the preferred method when transitioning from pilot production to commercial production with higher capacity. Otherwise, reactive sputtering from metal targets is an alternative process promising high deposition rates and low cost. Typical pilot line sputtering systems are in-line systems with dynamic deposition where the substrate passes one or more targets. The substrates are additionally heated in some but not in all systems. The systems are partly derived from those sold otherwise for ITO-deposition in ﬂat panel display 1

The CVD processing of ZnO for thin ﬁlm solar cells is extensively described in Chap. 6. Details on CVD ZnO in Cu(In,Ga)Se2 thin ﬁlm solar cells can be found in Sects. 6.3.1.1 and 6.3.2.1.

426

R. Klenk

manufacturing. There is no clear preference for vertical or horizontal conﬁguration. Reasonable ﬁlm properties, high throughput, and stable processes at low cost are obvious challenges to be met. In addition, ZnO deposited on chamber walls, shields, or even redeposited onto target areas outside the erosion zones exhibits particularly poor adhesion. ZnO ﬂakes, particles, and dust may cause pin-holes in the module. Frequent cleaning of the deposition system may be required and limits its useful uptime. 9.3.2.1 DC Sputtering from Ceramic Targets Results of a comprehensive study on DC sputtered ZnO for CIGSe-based modules can be found in [37]. Optical and electrical ﬁlm properties were determined for varied target doping (ZnO:Al, ZnO:Ga) and sputtering conditions, in particular the oxygen ﬂow. A normalized photocurrent was calculated from the measured transmission spectra using the spectral response of a highly eﬃcient CIGSe solar cell. A model curve describing the normalized photo current as a function of ZnO sheet resistance was extrapolated from the experimental data and used as input for numerical calculations of module eﬃciency. It was concluded that – given the properties of the DCsputtered ZnO – the optimal interconnect distance is 5 mm. An eﬃciency within 2% of the optimum was predicted for ZnO sheet resistance ranging from 7.5 to 20 Ω sq−1 . These values can be achieved with a target doping of 1.6–3.2% (metals ratio) and ﬁne-tuned with the oxygen ﬂow. Dynamic deposition rates of 80 nm m min−1 could be achieved. The authors point out the strong variations of the conductivity with minima directly opposite to the target erosion zones (race track) measured on test substrates deposited in static mode and conclude that this prevents achievement of optimum ﬁlm properties in dynamic mode. It was found later that the conductivity proﬁle under the target depends on the age of the target, i.e., the depth of the race track [38]. 9.3.2.2 Reactive Sputtering Reactive sputtering, i.e., sputtering from a metal rather than ceramic target is one of the options investigated for cost reductions. Already in 1996 it was reported that reactive DC sputtering can achieve the same cell eﬃciency as RF-sputtering from a ceramic target [10] even though the static deposition rate (300 nm min−1 ) was ten times higher using the former method. The authors point out the importance of tight process control, in this case plasma emission monitoring (PEM), to obtain the required ﬁlm properties at the low substrate temperature permitted by the chalcopyrite-based junction. Various problems concerning homogeneity, set point drift, and ﬂaking have been encountered when reactive sputtering was evaluated in the Wuerth Solar pilot line [39] using a PEM controlled DC/DC dual magnetron process.

9 Chalcopyrite Solar Cells and Modules

427

Nevertheless, eﬃciencies of more than 10% have been reported for full-size modules, only slightly lower than that of the standard modules prepared using a ceramic target. Sittinger et al. report on an MF-excited dual magnetron process using λ-sensors for oxygen partial pressure control. The process was originally developed for thin-ﬁlm silicon modules where higher substrate temperatures can be used [40]. High eﬃciency as well as good stability of CIGSebased mini-modules could be demonstrated after readjusting the deposition parameters [41].

9.4 Stability Outdoor testing of chalcopyrite-based modules has generally demonstrated excellent stability [42, 43]. Owing to the increasing production volume there is a growing number of installations where the actual performance [44] and long-term stability can be assessed. Chalcopyrites do not suﬀer from any form of light-induced degradation. They are also known for their extraordinary radiation hardness and for their capability to passivate defects at comparably low temperatures [45]. Accelerated lifetime testing, especially the damp heat testing procedure (85◦ C, 85% humidity, 1 000 h), which forms a part of the EN/IEC 61646 certiﬁcation, has, however, been cumbersome [46]. Partly, this is due to transient eﬀects which occur during stress tests [47]. These can lead to an apparent degradation but the eﬃciency recovers after several days of light-soaking. As we will describe in the following sections, the exact causes for degradation are still under investigation; nevertheless, empirical optimization has achieved modules that have been independently certiﬁed [43, 47, 48]. ZnO serves diﬀerent purposes in the chalcopyrite-based solar cells, hence, a degradation of ZnO properties in damp heat may aﬀect the device performance in several ways. If the stability is examined using small individual cells with metal grid, the observed device degradation is mainly due to some deterioration of the heterojunction itself [49]. The stability of an interconnected module is in addition also aﬀected by increasing ZnO sheet resistance and deteriorating interconnect properties and is therefore more severe [50]. Investigations aimed at identifying the basic degradation mechanisms are usually carried out by exposing non-encapsulated ﬁlms and devices to the conditions according to IEC/EN 61646. It is often silently assumed that the degradation observed under these extremely harsh conditions is governed by the same physical and chemical eﬀects as the small degradation, which is actually allowable for encapsulated modules passing the test (not more than 5% change in eﬃciency). Experimental results described by diﬀerent authors are sometimes contradictory, which indicates that the fundamental mechanisms are not well understood.

428

R. Klenk

9.4.1 Module Degradation Accelerated ageing according to IEC/EN 61646 implies elevated temperature as well as high humidity. It has been reported that there is a module power loss already due to storing the module at the elevated temperature (dry heat) [47]. This power loss could be reversed by prolonged illumination (lightsoaking), whereas the degradation caused by the humidity was irreversible. Other modules or test structures appear to be perfectly stable in dry heat and do not exhibit the reversible power loss [51]. A major degradation mechanism of modules is the decrease in ﬁll factor. This is caused by an increase in the diode quality factor of the cells making up the module and by an increase in series resistance. The former is related more to the absorber and heterojunction properties and less to the ZnO properties. The series resistance increases because the conductivity of the ZnO drops and because the interconnects are deteriorating. Wennerberg et al. have assessed the individual contributions to increased series resistance [50]. Klaer et al. [52, 53] have described a transmission-line test structure that allows to separate the contributions of contact and sheet resistance, respectively. The test structure is prepared by the same scribing techniques as those used in module manufacturing. 9.4.1.1 Interconnect Corrosion Increased series resistance due to interconnect corrosion has two possible origins: degradation of the Mo/ZnO contact (P2 interconnect via) and corrosion of the molybdenum which is exposed to the atmosphere in the P3 scribe-line. Bare molybdenum ﬁlms on glass corrode rapidly in damp heat and partially form a transparent oxide. Nevertheless, corrosion of the molybdenum in P3 does not seem to contribute signiﬁcantly to the module degradation until the molybdenum in the scribe has oxidized completely, the interconnect is broken, and the module eﬃciency is practically zero [54]. It has been reported that molybdenum corrosion occurs faster when the molybdenum ﬁlm is mechanically stressed [55]. Nevertheless, total breakdown of the interconnect should happen only under unrealistically extreme conditions. Wennerberg et al. argue that degradation of the Mo/ZnO contact resistance (P2 interconnect via) is to blame for most of the increased resistance [50]. They ﬁnd that the contact resistance increases by almost two orders of magnitude, whereas the ZnO sheet resistance increases by only 50% after 500 h. Using locally resolved X-ray emission spectroscopy, Fischer et al. have found indications that, in P2, the ZnO reacts with molybdenum-chalcogenides during damp heat. This may explain contact degradation by formation of compounds with low conductivity [56] (molybdenum-chalcogenides can form during absorber preparation). Klaer et al. conﬁrm a certain degradation of the interconnect but conclude that the increase in ZnO sheet resistance is more signiﬁcant [52,53]. They ﬁnd a better stability of the interconnect when

9 Chalcopyrite Solar Cells and Modules

429

the P2 scribe is carried out after the deposition of i-ZnO and before deposition of n+ -ZnO. This process sequence results in an Mo/n+ -ZnO via rather than an Mo/i-ZnO/n+-ZnO interconnect. The disadvantage of this sequence lies in higher manufacturing cost due to the necessity to interrupt the ZnO deposition and break the vacuum for the patterning step. 9.4.1.2 ZnO Degradation As described above, there is no general consensus in literature concerning the relative contributions of contact corrosion and increasing ZnO sheet resistance to increased module series resistance. However, experiments carried out within a recent joint research project in Germany suggest that degradation of the ZnO sheet resistance may indeed be the principle cause of damp heat instability. Large area absorbers provided by Wuerth Solar were diced into many nominally identical samples and coated with likewise nominally identical CdS buﬀer layers. ZnO ﬁlms were then deposited onto these samples by several project participants using a range of diﬀerent sputtering systems and parameters. The ZnO sheet resistance was measured directly on these substrates before and after diﬀerent periods of damp heat ageing (without any encapsulation). All samples showed a sheet resistance increasing systematically with ageing. There was also a clear correlation between the increased ZnO sheet resistance and degradation of module test structures prepared under the same conditions. However, the described eﬀects varied strongly among the diﬀerent ZnO layers [38, 41]. In general, conditions which favor the growth of dense ﬁlms and good step coverage (RF excitation, low pressure, high temperature, moving substrate) tend to yield better stability. In addition to the deposition conditions, the ZnO microstructure is also heavily inﬂuenced by the substrate. In view of the ﬁndings described above it is not surprising that ZnO ﬁlms deposited onto smooth substrates are much more stable than those deposited onto rough substrates. This is best illustrated by experiments [57] where polished and texture-etched silicon wafers were used as model substrates (Figs. 9.10 and 9.11). The free carrier optical reﬂection of test modules before and after damp heat indicates that the eﬀective carrier density is not much aﬀected [58]. Hence, the degradation of the ZnO sheet resistance is probably more of a carrier transport problem. It is, at present, unclear where electron barriers are located. They may be present at the grain boundaries in general [59]. In this case, the disturbances of the ZnO microstructure (induced by the substrate but also depending on preparation parameters) are only harmful because they allow a faster penetration of the humidity into the ﬁlm. On the other hand, the disturbed regions may themselves be highly resistive after damp heat exposure, which forces the current to percolate around these areas.

430

R. Klenk

Fig. 9.10. Scanning electron micrograph of a ZnO thin ﬁlm on a texture-etched silicon substrate. The ZnO microstructure is disturbed where one pyramid of the substrate borders the next pyramid

Resistivity (mΩ cm)

8

ZnO on polished wafer ZnO on texture-etched wafer

6

4

2

0 0

10

50 100 500 Damp Heat Exposure (h)

1000

Fig. 9.11. Resistivity of ZnO thin ﬁlms as a function of damp heat exposure (no encapsulation)

9 Chalcopyrite Solar Cells and Modules

431

9.5 Nonconventional and Novel Applications 9.5.1 Direct ZnO/Chalcopyrite Junctions The standard device comprises a thin CdS buﬀer layer as described above. It is believed that market acceptance of chalcopyrite-based photovoltaics could be improved by introducing a Cd-free buﬀer layer. There may also be cost beneﬁts in view of the cost associated with (occupational) safety, handling of toxic waste in production, and recycling of modules at their end of life. Research has identiﬁed Cd-free materials well suited for alternative buﬀer layers. They can be deposited by CBD in analogy to the standard CdS buﬀer layers or by other processes. In particular, dry processes are attractive because they oﬀer a better compatibility with the other process steps used for the remainder of the module. Ultimately, the best solution would be to omit the buﬀer layer altogether in favor of a direct chalcopyrite-sputtered/MOCVD ZnO junction. Here we will limit the discussion to the state of these latter direct junctions and to ZnO-based buﬀer layers (Table 9.1). Results achieved with other materials can be found in the literature [67, 68]. It had been assumed originally that CdS is required for lattice matching and conduction band alignment. It is therefore somewhat surprising that it is in eﬀect possible to use a ZnO buﬀer layer. As we have shown elsewhere, however, interface recombination can be suppressed to an insigniﬁcant level not only by reducing the interface recombination velocity (which would require lattice matching) but also by reducing the number of carriers available for recombination [5]. In consequence, an eﬃcient device can be expected whenever we succeed in positioning the Fermi-level close to the conduction band at the hetero-interface. In general this can be achieved by a suitable defect distribution (shallow donors) and band alignment at the interface. Judging from the experimental approaches that have been successful in terms of device performance, wet chemical surface conditioning, and mixing ZnO with other compounds seem to play an important role in establishing Table 9.1. Performance of ZnO/chalcopyrite solar cells without buﬀer layer Reference Negami et al. [60] Ramanathan et al. [61] B¨ ar et al. [62] Glatzel et al. [63] Olsen et al. [64] Strohm et al. [65] Lincot et al. [66]

Technology Evaporation of Zn, sputtering of (Zn,Mg)O Partial electrolyte, sputtering Partial electrolyte, ion layer gas reaction Sputtering of (Zn,Mg)O Chemical vapour deposition Sputtering of (Zn,Mg)O Electrodeposition (2 mm2 )

Eﬃciency(%) 16.2 15.3 15 12.5 12 11.5 9.2

432

R. Klenk Table 9.2. Superstrate cells

Reference

TCO

Buﬀer

Absorber

Klenk et al. [69] Yoshida et al. [70] Negami et al. [71] Haug et al. [72] Nakada et al. [73]

ZnO ITO ZnO ZnO ZnO

None CdS CdS None None

CuGaSe2 CuInSe2 CuInSe2 Cu(In,Ga)Se2 Cu(In,Ga)Se2

Eﬃciency Voc > 800 mV 6.6% 6.7% 11.2% 12.8%

the described criteria. It is interesting that in the reversed superstrate structure (see below) direct junctions perform better than those with a buﬀer layer. 9.5.2 Superstrates The conventional cell structure is supported by a substrate onto which the layers are deposited with the doped ZnO being the ﬁnal layer. Reversing the whole structure by depositing ﬁrst the ZnO followed by the remaining layers results in a structure where the light enters the cell through the superstrate (Fig. 9.12). This concept has some potential advantages: – The ZnO can be deposited at high temperature, resulting in improved optical and electrical properties and stability. Presumably its thickness could be reduced resulting in lower cost. – Since the light enters through the superstrate the encapsulation does not have to be transparent. Nontransparent foils could be more cost eﬀective than glass and reduce the module weight. – Mechanically stacked tandem structures could be realized in a straightforward manner by fabricating one part in substrate and the other in superstrate conﬁguration and then laminating them together. However, preparing a blocking contact in superstrate structure has been diﬃcult. Only small area cells have been demonstrated so far and even those show limited performance (Table 9.2). To achieve optimum properties, the absorber layer must be prepared at temperatures of about 500◦ C, which implies that interdiﬀusion is signiﬁcant and may lead to a deterioration of junction and absorber bulk properties. It is interesting to note that approaches not using buﬀer layers have resulted in higher eﬃciency than those using CdS buﬀers prepared by various methods. 9.5.3 Transparent Back Contact All ZnO/(buﬀer)/chalcopyrite junctions described so far in this contribution are blocking junctions as may be expected from the conductivity type

9 Chalcopyrite Solar Cells and Modules

433

Fig. 9.12. Schematic cross-sectional views of chalcopyrite solar cell conﬁgurations. Reprinted with permission from [1]

of ZnO (n-type) and chalcopyrite (p-type), respectively. However, in the absence of technically viable p-type TCOs there is a certain interest also in making ohmic (nonblocking) contacts to n-type TCO. Experimental results demonstrate that this is feasible. Replacing the molybdenum ﬁlm of the conventional structure by a TCO ﬁlm then results in a cell with a transparent back contact. Such a structure may be useful in implementing bifacial or tandem cells, or semi-transparent modules [73] (Fig. 9.12). Again, the high substrate temperature necessary for absorber preparation is problematic and may lead to a deterioration of transparency and conductivity of the TCO ﬁlm. In addition the absorber may be poisoned by elements diﬀusing from the TCO ﬁlm. In many studies ZnO was found to be less suitable than other materials (such as SnO2 or ITO) for this particular application [73, 74]. It has been demonstrated, however, that a very thin molybdenum ﬁlm deposited onto the ZnO signiﬁcantly improves the performance of devices with ZnO back contact. It is believed that the molybdenum is converted to MoSe2 (a semiconductor with Eg = 1.2 eV) during deposition of the absorber [75]. The eﬃciency for backside illumination (bi-facial cells) is limited because the carriers are generated outside the ﬁeld zone in proximity to the poorly passivated contact (poor blue response). Further optimization of absorber thickness, diﬀusion length, and contact passivation appears to be feasible. The eﬃciency of wide gap chalcopyrite cells, even on standard nontransparent substrate is too low for the realization of tandem cell. Using a TCO back contact reduces the eﬃciency even further. Establishing a high transparency is also not straightforward. Nevertheless, the potential of tandem

434

R. Klenk

cells is reﬂected in wide-spread research activities. Preliminary prototypes are described in literature [76, 77]. The standard molybdenum contact is characterized by low optical reﬂection, which becomes relevant in eﬀorts to reduce the absorber thickness (light trapping). A study of other metals has not identiﬁed clearly promising alternatives [78]. Hence, an ohmic contact between TCO and chalcopyrite with good electrical and optical properties could also be useful in developing a cell with a high reﬂectivity metal/TCO back contact.

References 1. J. Poortmans, V. Archipov, “Thin Film Solar Cells”, Wiley & Sons, 2006 2. K. Ramanathan, M.A. Contreras, C.L. Perkins, S. Asher, F.S. Hasoon, J. Keane, D. Young, M. Romero, W. Metzger, R. Nouﬁ, J. Ward, A. Duda, Prog. Photovolt: Res. Appl. 11, 225 (2003) 3. S. Wagner, J.L. Shay, P. Migliorato, H.M. Kasper, Appl. Phys. Lett. 25, 434 (1974) 4. R.R. Potter, C. Eberspacher, L.B. Fabick, in Proceedings of the 18th IEEE Photovoltaic Specialists Conference, 1985, p. 1659 5. R. Klenk, Thin Solid Films 387, 135 (2001) 6. D. Schmid, M. Ruckh, F. Grunwald, H.W. Schock, J. Appl. Phys. 73, 2902 (1993) 7. T. Wada, S. Hayashi, Y. Hashimoto, S. Nishiwaki, T. Sato, T. Negami, M. Nishitani, in Proceedings of the 2nd World Conference Photovoltaic Solar Energy Conversion, 1998, p. 403 8. K. Ramanathan, R. Nouﬁ, J. Granata, J. Webb, J. Keane, Solar Energy Mater. & Sol Cells 55, 330 (1998) 9. Q. Nguyen, K. Orgassa, I. Koetschau, U. Rau, H.W. Schock, Thin Solid Films 431–432, 330 (2003) 10. M. Ruckh, D. Hariskos, U. R¨ uhle, H.W. Schock, R. Menner, B. Dimmler, in Proceedings of the 25th IEEE Photovoltaic Specialists Conference, 1996, p. 825 11. J. Kessler, J. Norling, O. Lundberg, J. Wennerberg, L. Stolt, in Proceedings of the 16th European Photovoltaic Solar Energy Conference, 2000, p. 775 12. J. Kessler, S. Wiedeman, L. Russell, J. Fogleboch, S. Skibo, R. Arya, D. Carlson, in Proceedings of the 25th IEEE Photovoltaic Specialists Conference, 1996, p. 885 13. K. Ramanathan, J. Keane, R. Nouﬁ, in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, 2005, p. 195 14. U. Rau, M. Schmidt, Thin Solid Films 387, 141 (2001) 15. U. Rau, P.O. Grabitz, J.H. Werner, Appl. Phys. Lett. 85, 6010 (2004) 16. G. Haacke, Appl. Phys. Lett. 47, 4086 (1976) 17. J. Wennerberg, Design and Stability of Cu(In,Ga)Se2 -Based Solar Cell Modules, Ph.D. Thesis, Uppsala University, 2002 18. J. Klaer, I. Luck, A. Boden, R. Klenk, I. Gavilanes-Perez, R. Scheer, Thin Solid Films 431–432, 534 (2003) 19. R. Klenk, J. Klaer, R. Scheer, M.C. Lux-Steiner, I. Luck, N. Meyer, U. R¨ uhle, Thin Solid Films 480–481, 509 (2005)

9 Chalcopyrite Solar Cells and Modules

435

20. J. AbuShama, R. Nouﬁ, S. Johnston, S. Ward, X. Wu, in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, 2005, p. 299 21. M. Burgelman, A. Niemegeers, Solar Energy Mater. & Sol Cells 51, 129 (1998) 22. J. Kessler, J. Wennerberg, M. Bodeg˚ ard, L. Stolt, Solar Energy Mater. & Sol Cells 75, 35 (75) 23. K. Orgassa, H.W. Schock, J.H. Werner, Thin Solid Films 431–432, 387 (2003) 24. K. Orgassa, Q. Nguyen, I.M. K¨ otschau, U. Rau, H.W. Schock, J.H. Werner, in Proceedings of the 17th European Photovoltaic Solar Energy Conference, 2001, p. 391 25. K. Orgassa, U. Rau, Q. Nguyen, H.W. Schock, J.H. Werner, Prog. Photovolt: Res. Appl. 10, 457 (2002) 26. C.A. Kaufmann, A. Neisser, R. Klenk, R. Scheer, H.W. Schock, Mater. Res. Soc. Symp. Proc. 865, F7.5.1 (2005) 27. L.L. Kazmerski, O. Jamjoum, P.J. Ireland, R.A. Mickelsen, W.S. Chen, J. Vac. Sci. Technol. 21, 486 (1982) 28. B.M. Basol, V.J. Kapur, in Proceedings of the 21st IEEE Photovoltaic Specialists Conference, 1990, p. 546 29. N. Meyer, I. Luck, U. R¨ uhle, J. Klaer, R. Klenk, M.C. Lux-Steiner, R. Scheer, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004, p. 1698 30. B. Dimmler, M. Powalla, R. Schaeﬄer, in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, 2005, p. 189 31. M. Oertel, B. Dimmler, in Proceedings of the Workshop TCO f¨ ur D¨ unnschichtsolarzellen und andere Anwendungen III, 2005, p. 61 32. P.S. Vijayakumar, K.A. Blaker, R.D. Wieting, B. Wong, A.T. Halani, U.S. Patent 4,751,149, 1988 33. E. Terzini, A. Antonaia, P. Thilakan, S. Aprea, I. Luck, in Proceedings of the 16th European Photovoltaic Solar Energy Conference, 2000, p. 706 34. R.D. Wieting, in Proceedings of the 29th IEEE Photovoltaic Specialists Conference, 2002, p. 478 35. R.H. Mauch, J. Hedstr¨ om, D. Lincot, M. Ruckh, J. Kessler, R. Klinger, L. Stolt, J. Vedel, H.W. Schock, in Proceedings of the 22nd IEEE Photovoltaic Specialists Conference, 1991, p. 898 36. K. Ramanathan, J.C. Keane, B. To, R.G. Dhere, R. Nouﬁ, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, 2005, p. 1695 37. R. Menner, R. Sch¨ aﬄer, B. Sprecher, B. Dimmler, in Proceedings of the 2nd World Conference on Photovoltaic Solar Energy Conversion, 1998, p. 1161 38. R. Menner, M. Powalla, in Proceedings of the Workshop TCO f¨ ur D¨ unnschichtsolarzellen und andere Anwendungen III, 2005, p. 71 39. R. Menner, M. Oertel, M. Powalla, M. Dimer, S. Rehn, J. Str¨ umpfel, B. Dimmler, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004, p. 1694 40. V. Sittinger, F. Ruske, W. Werner, S. B, B. Rech, J. H¨ upkes, G. Sch¨ ope, H. Stiebig, Thin Solid Films 496, 16 (2006) 41. V. Sittinger, F. Ruske, W. Werner, B. Szyszka, R. Menner, M. Powalla, B. Dimmler, in Proceedings of the 21st European Photovoltaic Solar Energy Conference, 2006 42. L. Mrig, S. Rummel, in Proceedings of the 21st IEEE Photovoltaic Specialists Conference, 1990, p. 1038

436

R. Klenk

43. M. Powalla, B. Dimmler, R. Sch¨aﬄer, G. Voorwinden, U. Stein, H.D. Mohring, F. Kessler, D. Hariskos, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 1663, p. 1663 44. H.D. Mohring, D. Stellbogen, R. Sch¨ aﬄer, S. Oelting, R. Gegenwart, P. Konttinen, T. Carlsson, M. Cendagorta, W. Herrmann, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004, p. 2098 45. A. Jasenek, A. Boden, K. Weinert, M.R. Balboul, H.W. Schock, U. Rau, Mater. Res. Soc. Symp. Proc. 668, H3.2.1 (2001) 46. J. Malmstr¨ om, J. Wennerberg, L. Stolt, Thin Solid Films 431–432, 436 (2003) 47. F. Karg, H. Calwer, J. Rimmasch, V. Probst, W. Riedl, W. Stetter, H. Vogt, M. Lampert, in Proceedings of the 11th International Conference on Ternary and Multinary Compounds, 1998, p. 909 48. J. Palm, V. Probst, W. Stetter, R. Toelle, S. Visbeck, H. Calwer, T. Niesen, H. Vogt, O. Hernandez, M. Wendl, F.H. Karg, Thin Solid Films 451–452, 544 (2004) 49. J. Wennerberg, J. Kessler, M. Bodeg˚ ard, L. Stolt, in Proceedings of the 2nd World Conference on Photovoltaic Solar Energy Conversion, 1998, p. 1161 50. J. Wennerberg, J. Kessler, L. Stolt, in Proceedings of the 16th European Photovoltaic Solar Energy Conference, 2000, p. 309 51. R. Scheer, R. Klenk, J. Klaer, I. Luck, Solar Energy 77, 777 (2004) 52. J. Klaer, R. Scheer, R. Klenk, A. Boden, C. K¨oble, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004, p. 1847 53. J. Klaer, R. Klenk, A. Boden, A. Neisser, C. Kaufmann, R. Scheer, H.W. Schock, in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, 2005, p. 336 54. M. Powalla, B. Dimmler, Solar Energy Mater. & Sol Cells 67, 337 (2001) 55. J. Klaer, R. Klenk, C. K¨ oble, A. Boden, R. Scheer, H.W. Schock, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, 2005, p. 1914 56. C.H. Fischer, M. B¨ ar, T. Glatzel, I. Lauermann, M.C. Lux-Steiner, Solar Energy Mater. & Sol Cells 90, 1471 (2006) 57. R. Klenk, M. Linke, H. Angermann, C. Kelch, M. Kirsch, J. Klaer, C. K¨oble, in Proceedings of the Workshop TCO f¨ ur D¨ unnschichtsolarzellen und andere Anwendungen III, 2005, p. 79 58. C. K¨ oble, R. Klenk (unpublished results) 59. J.Y.W. Seto, J. Appl. Phys. 46, 5427 (1975) 60. T. Negami, T. Aoyagi, T. Satoh, S. Shimakawa, S. Hayashi, Y. Hashimoto, in Proceedings of the 29th IEEE Photovoltaic Specialists Conference, 2002, p. 656 61. K. Ramanathan, F.S. Hasoon, S. Smith, D.L. Young, M.A. Contreras, P.K. Johnson, A.O. Pudov, J.R. Sites, J. Phys. Chem. Solids 64, 1495 (2003) 62. M. B¨ ar, C.H. Fischer, H.J. Muﬄer, B. Leupolt, T.P. Niesen, F. Karg, M.C. LuxSteiner, in Proceedings of the 29th IEEE Photovoltaic Specialists Conference, 2002, p. 636 63. T. Glatzel, H. Steigert, R. Klenk, M.C. Lux-Steiner, T.P. Niesen, S. Visbeck, in Technical Digest of the 14th International PVSEC, 2004, p. 27–5 64. L.C. Olsen, H. Aguilar, F.W. Addis, W. Lei, J. Li, in Proceedings of the 25th IEEE Photovoltaic Specialists Conference, 1996, p. 997 65. A. Strohm, T. Schl¨ otzer, Q. Nguyen, K. Orgassa, H. Wiesner, H.W. Schock, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004, p. 1741

9 Chalcopyrite Solar Cells and Modules

437

66. D. Lincot, B. Canava, S. Quenet, S. Peulon, in Proceedings of the 14th European Photovoltaic Solar Energy Conference, 1997, p. 2168 67. S. Siebentritt, Solar Energy 77, 767 (2004) 68. N.A. Allsop, A. Sch¨ onmann, H.J. Muﬄer, M. B¨ ar, M.C. Lux-Steiner, C.H. Fischer, Prog. Photovolt: Res. Appl. 13, 607 (2005) 69. R. Klenk, R. Mauch, R. Sch¨ aﬄer, D. Schmid, in Proceedings of the 22nd IEEE Photovoltaic Specialists Conference, 1991 70. T. Yoshida, R.W. Birkmire, in Proceedings of the 11th E.C. Photovoltaic Solar Energy Conference, 1992, p. 811 71. T. Negami, M. Nishitani, M. Ikeda, T. Wada, Solar Energy Mater. & Sol Cells 35, 215 (1994) 72. F.J. Haug, D. Rudmann, H. Zogg, A.N. Tiwari, Thin Solid Films 431–432, 431 (2003) 73. T. Nakada, Y. Hirabayashi, T. Tokado, D. Ohmori, T. Mise, Solar Energy 77, 739 (2004) 74. S. Bakehe, R. Klenk, M.C. Lux-Steiner, in Proceedings of the 19th European Photovoltaic Solar Energy Conference, 2004 75. P.J. Rostan, J. Mattheis, G. Bilger, U. Rau, J.H. Werner, Thin Solid Films 480–481, 67 (2005) 76. S. Nishiwaki, S. Siebentritt, P. Walk, M.C. Lux-Steiner, Prog. Photovolt: Res. Appl. 11, 243 (2003) 77. M. Symko-Davies, R. Nouﬁ, in Proceedings of the 20th European Photovoltaic Solar Energy Conference, 2005, p. 26 78. K. Orgassa, H.W. Schock, J.H. Werner, Thin Solid Films 431–432, 387 (2003)

Index

(Zn,Cd)O, 13 (Zn,Mg)O, 13 absorption edge, fundamental, 111 MgZnO, 117, 118 temperature, 112–114, 116 air mass (AM), 250 Aluminium concentration, 135 angular distribution function, 287 anisotropy of conductivity, 48 atomic layer deposition, 293 band alignment amorphous interface, 164 II–VI semiconductors, 13 ZnO/(Zn,Mg)O, 13, 14 band gap, 274–276 shrinkage at high doping level, 276 band gap engineering, 13 band gap, fundamental, 109, 111 MgZnO, 117, 118 temperature, 112–114, 116 band-to-band transitions, 108, 109 Bose–Einstein model, 96, 112 parameters, 97, 113 bowing, 117 brass, 21 Burstein–Moss eﬀect, 274–276 cathodoluminescence, 26 Cauchy formula, 105 parameters, 107, 108 CdO, 5 CIGS solar cells CVD ZnO buﬀer layers, 289 collision cascade, 192 combinatorial PLD, 347 conduction band oﬀset, 126

conductivity of ZnO single crystals, 36 critical points, 108 parameters, 109 model dielectric function, 87 crystal structure, 81 rocksalt, 83 wurtzite, 82 crystal-ﬁeld splitting, 110, 112 crystallographic structure, 4 Cu(In,Ga)Se2 Cu depletion, 168 inﬂuence of Na, 169 Se cap layer, 164 solar cell, 128 cubic ZnO, 4 CVD atmospheric pressure (APCVD), 235, 238, 241 deposition temperature, 261 grain size, 245, 247 growth kinetics, 238, 239 low pressure (LPCVD), 239 precursors, 236 damp-heat test, 368 Debye-temperature, 112–114 defect annealing, 21 defects antisites, 14 concentration, 15 concentration in ZnO, 19 formation enthalpy, 15 formation enthalpy for ZnO, 16 Frenkel, 14, 17 in ZnO, 14–23 oxygen dumbbell interstitial, 17, 20, 21

440

Index

oxygen interstitial, 14 oxygen split interstitial, 17 oxygen vacancy, 14, 17, 19, 35, 38 oxygenvacancies, 38, 39 Schottky, 14 self compensation, 18 transition energies, 18 zinc interstitial, 14, 19, 35, 38, 39 zinc vacancy, 14, 17, 18 demonstrator devices, 336 density, 6 dielectric constants, 85, 91, 105 MgZnO, 91 ZnO, 91 dielectric function, 89–90 ZnO - temperature, 113 ZnO - VIS-VUV, 108, 113 diﬀusion in ZnO, 19–23 migration enthalpy, 19, 21 of oxygen in ZnO, 20–22 of zinc in ZnO, 21, 23 doping of CVD ZnO ﬁlms, 266 doping eﬃciency, 18, 203, 249, 257, 272, 273 doping limits, 18 for ZnO and In2 O3 , 19 doping uniformity of CVD ZnO ﬁlms, 257 droplets, 304 eﬀective electron mass, 103 electroluminescence, 26 electron mobility of AP-CVD ZnO ﬁlms, 257 of CVD ZnO ﬁlms, 258 electron-phonon interaction, 112 electronic structure, 12 ellipsometry, 81, 88 data analysis, 89 generalized, 89 standard, 88 energy band diagram, 126 epitaxial ZnO thin ﬁlms, 313 n-type doped, 322 pn-junctions, 344 p-type doped, 322, 344 AFM, 319

applications, 336 Bragg reﬂectors, 340 carrier concentration, 323 cathodoluminescence, 327, 338 chemical composition, 331 composition transfer, 334 deep level transient spectroscopy, 325 demonstrator devices, 336 diﬀusion barrier layer, 322 doping, 331 electrical properties, 322 excitons, 329 Hall mobility, 323 highlights, 344 magnetic domain formation, 337 phonons, 329 photoluminescence, 327 quantum well structures, 340 resistivity, 323 results of the Leipzig group, 335 RHEED, 314 Schottky contact, 327, 341 scintillators, 338 semiinsulating, 323 structure, 314 surface morphology, 319 TEM, 314 temperature-dependent Hall eﬀect, 325 trace element concentrations, 334 XRD, 314 etching behavior acidic solution, 370, 380, 382, 384, 386 alkaline solution, 380, 382, 386 discussion, 389 inﬂuence of aluminum doping, 385 inﬂuence of deposition pressure, 384 inﬂuence of glass substrate, 392 inﬂuence of hidden parameters, 387 inﬂuence of substrate temperature, 384 inﬂuence of working point, 386 ion beam treatment, 387 modiﬁed Thornton model, 384 plasma etching, 388 polycrystalline ﬁlms, 382 single crystal, 380

Index etching of sputtered ZnO layers, 213 excimer laser, 309 excitons, 108, 109, 111, 113 Fermi level pinning, 127 at CdS/ZnO interface, 160, 164 at In2 S3 /ZnO interface, 177 ﬁeld eﬀect transistor, 26 ﬁgure of merit, 287 formation enthalpy, 6, 166 free-charge-carriers, 80, 86, 102–105 fundamental absorption edge, 86 fundamental PLD processes ablation, 306 ablation threshold, 308 absorption, 306 condensation, 308 ﬁlm growth, 308 nucleation, 308 plasma expansion, 307 Ga2 O3 , 166 Gr¨ uneisen parameters, 97 grain boundaries, 59 grain boundary scattering, 257, 259, 278 grain size of CVD ﬁlms, 247 of LPCVD ﬁlms, 245, 247 growth of ZnO single crystals, 9 Hall coeﬃcient, 44, 46 Hall eﬀect, 37 hidden eﬀects, 310 high pressure phase, 5 high-Tc superconductor, 303 high-pressure PLD, 348 hydrogen, 19 In2 O3 , 5, 6, 18, 166 index of refraction, 105 birefringence, 106 MgZnO, 106, 108 temperature, 107 ZnO, 106, 108 MgZnO, 107 temperature, 106 ZnO, 107 interface dipole

441

at In2 S3 /ZnO interface, 177 ionization potential inﬂuence on band alignment, 160 of CdS, 144 of CdTe, 144 of ZnO, 143 ionized impurity scattering, 257 ITO, 18, 19, 25, 227 laser pulse energy, 310 laser-MBE, 347 lattice mismatch, 128 LCAO theory, 12 LCD display, 25 light scattering, 369 Lambertian scattering, 402 of APCVD ZnO, 259 of LPCVD ZnO, 259, 260 light trapping, 369, 370, 375 periodic gratings, 376 textured glass, 376 tin oxide, 375 zinc oxide (CVD), 376 zinc oxide (texture etched), 376 low energy ion scattering (LEIS), 133 Lydanne–Sachs–Teller relation, 85 magnetic semiconductors, 27 material properties, 6 micromorph Si solar cell, 284, 285 mobility, 38, 42–53, 377 as a function of carrier concentration, 49 model dielectric function, 85–88, 109 critical points, 86, 87 excitons, 87 free charge carriers, 86 phonons, 85 plasmons, 86 module encapsulation, 368 multi-element compounds, 304 Nd-YAG laser, 309 nucleation amorphous layer, 155, 164 oxygen dissociation, 138 nucleation layer, 268

442

Index

organic LED, 25, 227 organic solar cells, 227

rocksalt structure, 5 roll-to-roll process, 366

p–d repulsion, 12 peroxide, 7, 17, 138, 153, 154 phase diagram, 7 phonons, 83–84, 92–102, 112 broadening parameters, 80, 100–102 MgZnO, 99–100 rocksalt-structure, 84 wurtzite-structure, 83 ZnO, doped, 98–99 ZnO, undoped, 92–96 pressure, 96 temperature, 95–96 photoelectron spectroscopy, 128 interface studies, 129 of ZnO ﬁlms, 133 plasmon excitation, 133, 148, 149 piezoelectric coeﬃcient, 6 piezoelectric properties, 1, 26, 213 plasmons, 80, 86, 102–105 PLD chamber, 311 PLD parameters, 312 pulsed electron beam deposition, 347 pulsed laser deposition, 303 advanced PLD techniques, 346 basics, 303 demonstrator devices, 336 epitaxial ZnO thin ﬁlms, 313 ﬂexibility, 313 fundamental processes, 305 history, 303 instrumentation, 309 modelling, 305 nanostructures, 348 plasma diagnostics, 305 ZnO growth parameters, 309

sapphire substrate, 313 SAW devices, 213 ScAlMgO4 substrate, 313 scattering processes acoustical mode scattering, 44 dislocation scattering, 57 grain boundaries, 59 ionized impurity scattering, 45 neutral impurity scattering, 48 optical mode scattering, 43 piezoelectric mode scattering, 44 Schottky contact, 126 of ZnO, 127 Seebeck coeﬃcient, 38 segregation, 26 self compensation, 18 sheet resistance, 372 Silicon, 6 silicon absorption, 369 absorption coeﬃcient, 364 amorphous silicon, 363 band structure, 361 bandgap, 364 crystalline silicon, 361 defects, 362 microcrystalline silicon, 363 nanocrystalline silicon, 363 polycrystalline silicon, 363 SnO2 , 5, 6 solar cells J/V -curve, 362 back reﬂector, 361, 365, 376, 399 eﬃciency, 363, 403, 404 ﬁll factor, 362 heterojunction, 361 internal electric ﬁeld, 362 light induced degradation, see Staebler–Wronski eﬀect light trapping, 369, 370, 397 module, see solar modules n–i–p structure, 361 open circuit voltage, 362 optical losses, 372 optical simulation, 372 p–i–n structure, 361

quantum well structures, 13 radiation resistance, 21 Raman scattering, 81, 84 scattering conﬁgurations, 84 selection rules, 85 residual conductivity, 18, 19

Index photo current, 362 quantum eﬃciency, 366, 371, 396 short circuit current, 362 stability, 367 structure, 365 substrate conﬁguration, 365 superstrate conﬁguration, 365 tandem structure, 366 TCO-p contact, 394 solar modules, 373 active area, 374 aperture area, 374 cell width, 374 dead area, 374 eﬃciency, 403 encapsulation, 366 interconnection area, 374 series-connection, 373 space group, 6 spin-orbit coupling, 110, 112 spintronics, 27 sputter damage, 128 sputter yield, 190 Staebler–Wronski eﬀect, 363, 367, 400 STM of ZnO surfaces, 132, 133 stoichiometry, 7, 8 of sputtered ZnO, 134 strain, 80, 112, 114 sublimation, 7 surface acoustic wave devices, 26, 27 surface roughness, 247, 248 surface states of ZnO, 140 surface structure stability of polar surfaces, 132 wurtzite lattice, 132 wurtzite(10¯ 10), 132 wurtzite(11¯ 20), 132 zincblende lattice, 132 zincblende(110), 132 zincblende(111), 132 zincblende(¯ 1¯ 1¯ 1), 133 ZnO(0001), -(000¯ 1), 132, 133 surface termination, 132, 213 of ZnO, 144

443

texture of sputtered ZnO ﬁlms, 132, 144, 145, 157 thermal conductivity, 6 thermal expansion, 8 thermal expansion coeﬃcient, 6 thermovoltage, 37 thin ﬁlm solar cells, 24 tin oxide, 375 transparent electrodes, 24 two-oscillator model, 112, 116 parameters, 114 valence band oﬀset, 126 valence-band ordering, 110, 112 vapor pressure, 7 varistor, 53 varistors, 25 Varshni model, 112 parameters, 113 volume deposition rate, 304 work function of ZnO, 142, 143 YBa2 Cu3 O7−δ , 303 zinc oxide CAl , 370 doping level, 370 electrical properties, 372, 378 etching behavior, see etching behavior haze, 371 optical properties, 369, 370, 378 parasitic absorption, 401, 402 roughness, 371 sputter deposition, 377, 403, 404 surface texture, 371, 397 Zn resources, 3 Zn2 SiO4 , 26 ZnGa2 O4 , 26 ZnO ceramics, 26 ZnO nanostructures, 11 ZnO single crystals growth of, 9, 10 ZnO surfaces, 131–149 ZnO:Al, see zinc oxide

Springer Series in

materials science Editors: R. Hull

R. M. Osgood, Jr.

40 Reference Materials in Analytical Chemistry A Guide for Selection and Use Editor: A. Zschunke 41 Organic Electronic Materials Conjugated Polymers and Low Molecular Weight Organic Solids Editors: R. Farchioni and G. Grosso 42 Raman Scattering in Materials Science Editors: W. H. Weber and R. Merlin 43 The Atomistic Nature of Crystal Growth By B. Mutaftschiev 44 Thermodynamic Basis of Crystal Growth P–T–X Phase Equilibrium and Non-Stoichiometry By J. Greenberg 45 Thermoelectrics Basic Principles and New Materials Developments By G. S. Nolas, J. Sharp, and H. J. Goldsmid 46 Fundamental Aspects of Silicon Oxidation Editor: Y. J. Chabal 47 Disorder and Order in Strongly Nonstoichiometric Compounds Transition Metal Carbides, Nitrides and Oxides By A. I. Gusev, A. A. Rempel, and A. J. Magerl 48 The Glass Transition Relaxation Dynamics in Liquids and Disordered Materials By E. Donth

J. Parisi

H. Warlimont

51 Point Defects in Semiconductors and Insulators Determination of Atomic and Electronic Structure from Paramagnetic Hyperﬁne Interactions By J.-M. Spaeth and H. Overhof 52 Polymer Films with Embedded Metal Nanoparticles By A. Heilmann 53 Nanocrystalline Ceramics Synthesis and Structure By M. Winterer 54 Electronic Structure and Magnetism of Complex Materials Editors: D.J. Singh and D. A. Papaconstantopoulos 55 Quasicrystals An Introduction to Structure, Physical Properties and Applications Editors: J.-B. Suck, M. Schreiber, and P. H¨aussler 56 SiO2 in Si Microdevices By M. Itsumi 57 Radiation Effects in Advanced Semiconductor Materials and Devices By C. Claeys and E. Simoen 58 Functional Thin Films and Functional Materials New Concepts and Technologies Editor: D. Shi 59 Dielectric Properties of Porous Media By S.O. Gladkov 60 Organic Photovoltaics Concepts and Realization Editors: C. Brabec, V. Dyakonov, J. Parisi and N. Sariciftci

49 Alkali Halides A Handbook of Physical Properties By D. B. Sirdeshmukh, L. Sirdeshmukh, and K. G. Subhadra

61 Fatigue in Ferroelectric Ceramics and Related Issues By D.C. Lupascu

50 High-Resolution Imaging and Spectrometry of Materials Editors: F. Ernst and M. R¨uhle

62 Epitaxy Physical Principles and Technical Implementation By M.A. Herman, W. Richter, and H. Sitter

Springer Series in

materials science Editors: R. Hull

R. M. Osgood, Jr.

J. Parisi

H. Warlimont

63 Fundamentals of Ion-Irradiated Polymers By D. Fink

75 Wafer Bonding Applications and Technology Editors: M. Alexe and U. G¨osele

64 Morphology Control of Materials and Nanoparticles Advanced Materials Processing and Characterization Editors: Y. Waseda and A. Muramatsu

76 Spirally Anisotropic Composites By G.E. Freger, V.N. Kestelman, and D.G. Freger

65 Transport Processes in Ion-Irradiated Polymers By D. Fink 66 Multiphased Ceramic Materials Processing and Potential Editors: W.-H. Tuan and J.-K. Guo 67 Nondestructive Materials Characterization With Applications to Aerospace Materials Editors: N.G.H. Meyendorf, P.B. Nagy, and S.I. Rokhlin 68 Diffraction Analysis of the Microstructure of Materials Editors: E.J. Mittemeijer and P. Scardi 69 Chemical–Mechanical Planarization of Semiconductor Materials Editor: M.R. Oliver 70 Applications of the Isotopic Effect in Solids By V.G. Plekhanov 71 Dissipative Phenomena in Condensed Matter Some Applications By S. Dattagupta and S. Puri 72 Predictive Simulation of Semiconductor Processing Status and Challenges Editors: J. Dabrowski and E.R. Weber 73 SiC Power Materials Devices and Applications Editor: Z.C. Feng 74 Plastic Deformation in Nanocrystalline Materials By M.Yu. Gutkin and I.A. Ovid’ko

77 Impurities Conﬁned in Quantum Structures By P.O. Holtz and Q.X. Zhao 78 Macromolecular Nanostructured Materials Editors: N. Ueyama and A. Harada 79 Magnetism and Structure in Functional Materials Editors: A. Planes, L. Ma˜nosa, and A. Saxena 80 Micro- and Macro-Properties of Solids Thermal, Mechanical and Dielectric Properties By D.B. Sirdeshmukh, L. Sirdeshmukh, and K.G. Subhadra 81 Metallopolymer Nanocomposites By A.D. Pomogailo and V.N. Kestelman 82 Plastics for Corrosion Inhibition By V.A. Goldade, L.S. Pinchuk, A.V. Makarevich and V.N. Kestelman 83 Spectroscopic Properties of Rare Earths in Optical Materials Editors: G. Liu and B. Jacquier 84 Hartree–Fock–Slater Method for Materials Science The DV–X Alpha Method for Design and Characterization of Materials Editors: H. Adachi, T. Mukoyama, and J. Kawai 85 Lifetime Spectroscopy A Method of Defect Characterization in Silicon for Photovoltaic Applications By S. Rein 86 Wide-Gap Chalcopyrites Editors: S. Siebentritt and U. Rau 87 Micro- and Nanostructured Glasses By D. H¨ulsenberg and A. Harnisch