Mixed Ionic Electronic Conductors for Solid Oxide Fuel Cells

Diss. ETH No. 11348

Mixed Ionic Electronic Conductors for Solid Oxide Fuel Cells

A dissertation submitted to the SWISS FEDERAL INSTITUTE OF TECHNOLOGY

ZÜRICH for the degree of

Doctor of Technical Science

presented by

Martin Gödickemeier Dipl. Werkstoffing. ETH

born on March 5, 1967 citizen of Rickenbach ZH

ACCEPTED ON THE RECOMMENDATION OF

Prof. Dr. L.J. Gauckler, examiner Prof. Dr. H. Böhni, co-examiner

Prof. Dr. I. Riess, co-examiner Dr. K. Honegger, co-examiner

Zürich 1996

Acknowledgements

I wish to thank Prof. Dr. L.J. Gauckler for the fascinating subject, his encouragement and many hours of helpful discussions throughout the progress of my thesis. In particular I wish to thank him for the gratitude of giving me the opportunity to present my results to the public at international symposia and conferences.

I would like to thank Prof. Dr. Ilan Riess from Technion, Haifa (Israel) who introduced me to the field of mixed ionic electronic conductors, for many hours of fruitful but demanding discussions, and to Prof. Dr. H. Böhni for useful suggestions on electrochemical terminology and for the time he spent improving my thesis.

I wish to thank to Dr. Kaspar Honegger from Sulzer Innotec for his excellent support as an industrial partner and friend and for giving me experimental assistance.

I am in debt to many colleagues at the chair of Nonmetallic materials, especially to

Dr. Kazunari Sasaki, who gave me a lot of help throughout this work as a good colleague.

my colleagues Felix Baader and Martin Woodtli for help in experimental setup and for interesting discussions on scientific and philosophical topics,

Irene Urbanek for support in bureaucracy and Peter Kocher for technical help,

Prof. G. Bayer and Dr. P. Bohac for helpful discussions on scientific questions,

my students Piyasiri Ekanayake, Bruno Schelling, Daniel Schneider, Simon Sutter, Marco Wieland and Nicole Zech for their significant contribution to many aspects of this work.

I wish to thank to all members of the Swiss SOFC society who supported me during my work

Andreas Mitterdorfer and Julia Will, ETH Zürich

Dr. Augustin McEvoy, Dr. Claude Revilliod and Dr. Jan Van herle, EPFL Lausanne

Roland Diethelm, Dr. Emad Batawi and T. Gamper, Sulzer Innotec, Winterthur.

I wish to thank my parents, my wife Dorothée, and all my friends for their support and encouragement during my education.

Financial support from the Swiss Priority Program on Materials (PPM) of the board of the Swiss Federal Institutes of Technology is gratefully acknowledged.

Table of Contents

1. Mixed Conductors for Solid Oxide Fuel Cells

Summary / Zusammenfassung...................................................................................1

2. Introduction to Mixed Ionic Electronic Conductors in Solid Oxide Fuel Cells

1. Solid oxide fuel cells.............................................................................................4 2. Solid electrolytes..................................................................................................5 3. Mixed ionic electronic conductors.........................................................................6 4. Cathode...............................................................................................................7 5. Anode .................................................................................................................8 6. Interconnector......................................................................................................8 7. Compatibility of the materials system.....................................................................8 8. System efficiency and fuel cell design ....................................................................9 9. Terminology.........................................................................................................9 10. Structure of the thesis .......................................................................................10

3. Aim of the Study........................................................................................................13

4. Processing of Ceria Powders and Membranes

1. Introduction.......................................................................................................14 2. Powder preparation ...........................................................................................15 3. Tapecasting........................................................................................................21 4. Sintering of tapes................................................................................................24 5. Summary...........................................................................................................28 6. References.........................................................................................................29

5. Mixed Ionic Electronic Conductivity in Ceria-based Ceramic Membranes

1. Introduction.......................................................................................................30 2. Ionic and electronic conductivity.........................................................................32 3. Experimental......................................................................................................34 4. Results and discussion........................................................................................41 5. Four point measurements....................................................................................45 6. Summary & conclusions .....................................................................................54 7. References.........................................................................................................55

6. Nonstoichiometry and Defect Chemistry of Ceria Solid Solutions

1. Introduction.......................................................................................................57 2. Experimental......................................................................................................58 3. Results and discussion........................................................................................63 4. Defect models....................................................................................................67 5. Summary...........................................................................................................73 6. References.........................................................................................................74

7. Thermal and Isothermal Expansion of Ceria-based Ceramic Membranes

1. Introduction.......................................................................................................75 2. Experimental......................................................................................................76 3. Results and discussion........................................................................................80 4. Summary & conclusions .....................................................................................85 5. References.........................................................................................................86

8. Characterization of Solid Oxide Fuel Cells Based on MIECs

1. Introduction.......................................................................................................87 2. I-V relations.......................................................................................................88 3. Interpretation of four probe measurements on SOFCs.........................................98 4. Limitation on decreasing the MIEC resistance...................................................101 5. Comparison with experiments...........................................................................102 6. Chemical potential profiles................................................................................110 7. Summary.........................................................................................................112 8. References.......................................................................................................113 List of abbreviations and symbols .........................................................................115

9. Electrochemical Characteristics of Cathodes on Ceria based Electrolytes

1. Introduction.....................................................................................................117 2. Ionic and electronic currents in MIEC-fuel cells with MIECs.............................119 3. Experimental....................................................................................................126 4. Results and discussion......................................................................................133

4.1 Fitting steps in the evaluation of overpotentials.....................................133 4.2 La0.84Sr0.16CoO3 cathodes................................................................135 4.3 Platinum cathode................................................................................143 4.4 Silver cathode....................................................................................146 4.5 Gold cathode .....................................................................................150 4.6 La0.8Sr0.2MnO3 cathodes ..................................................................153

5. Summary & Conclusions ..................................................................................160 6. References.......................................................................................................162

10. Electrochemical Characteristics of Mixed Ionic Electronic Conducting Ni-Ce0.9Ca0.1O1.9 Cermet Anodes for Ceria

1. Introduction.....................................................................................................165 2. Anodes for SOFC, materials and reaction mechanisms......................................167 3. Interpretation of current interruption measurements on ceria-based ........electrolytes 170 4. Experimental....................................................................................................173 5. Screening test...................................................................................................177 6. Electrochemical characterization by current interruption measurements:Results and Discussion 183 7. Summary.........................................................................................................187 8. References.......................................................................................................188

11. Engineering of Solid Oxide Fuel Cells with Mixed Ionic Electronic Conductors

1. Introduction.....................................................................................................190 2. Ionic and electronic currents in fuel cells with mixed conducting electrolytes ......................................................................................................................191 3. Electrical and electrochemical cell parameters ...................................................195 4. Upscaling to prototype size...............................................................................199 5. Efficiency of fuel cells with mixed conductors ....................................................202 6. Oxygen chemical potential across the MIEC .....................................................206 7. Cell design and operating conditions .................................................................209 8. Summary and conclusions.................................................................................215 9. References.......................................................................................................215

Further Work ................................................................................................................217

Curriculum Vitae ...........................................................................................................220

1. Mixed Conductors for Solid Oxide Fuel Cells

Summary

Current solid oxide fuel cells (SOFC) employ stabilized zirconia as electrolyte. However, due to its high resistivity to ionic current, fuel cells with this material have to be operated at temperatures of 900 °C and higher. Decreasing the operation temperature to 700 °C would reduce system cost and enhance the reliability. To achieve this, the ionic resistance of the fuel cell has to be lowered e.g. by application of an electrolyte material with a lower resistivity.

Ceria based ionic conductors are promising candidates for such alternative electrolytes. They exhibit an ionic conductivity which is 3 to 5 times higher than that of zirconia electrolytes. However, at low oxygen partial pressures, such as they prevail at the anode side of SOFC, ceria is slightly reduced and develops additional n-type electronic conductivity. Thus, under SOFC operating conditions it is a mixed ionic electronic conductor and was not considered as fuel cell electrolyte up to now.

The scope of the present study is to show the possibility to apply mixed ionic electronic conducting ceria based materials as electrolyte to SOFC and to investigate the physical and electrochemical properties of such systems. Electrical conductivity, oxygen nonstoichiometry and thermal expansion are investigated under varying oxygen chemical potentials. Electrochemical measurements of fuel cells with ceria based membranes as electrolyte and different cathode and anode materials are interpreted by a novel model for mixed ionic electronic conductors. It is thereby important to consider the cathode/electrolyte/anode compound as a materials system to model its electrical and electrochemical properties.

The model gives a consistent description with explicit expressions for the oxygen transport properties of SOFC systems with mixed ionic electronic conductors as electrolyte. Partial electronic and ionic currents within this SOFC system are described as a function of the operating conditions under consideration of the defect chemistry of the electrolyte and the electrochemical performance of the electrodes. This information provides the base for design and construction of fuel cells operating with mixed ionic electronic conducting electrolytes. For the electrochemical reactions the model employs a straight-forward technical interpretation considering a basic concept of a superposition of diffusion and charge transfer overpotential.

Fuel cells with ceria based electrolytes exhibit a power output at 700 °C which is comparable to state of the art of zirconia based systems at 900 °C. However, due to its mixed ionic electronic conductivity the electrical conversion efficiency is slightly lower compared to pure ionic conductors. The lower operating temperature and the lower electrical conversion efficiency favor ceria based SOFC systems after all for small combined heat/power plants.

Summary/Zusammenfassung 2

From a systems engineering point of view, mixed conductors are more difficult to handle. They exhibit a small operating range between a maximum in efficiency at relatively high current densities and the maximum power output. Additionally, the thermal expansion coefficient of ceria based materials is dependent on oxygen partial pressure. Ceria based materials are therefore preferably used at temperatures below 800 °C.

Zusammenfassung

In heutigen Festelektrolyt-Brennstoffzellen findet stabilisiertes Zirkonoxid als Elektrolyt Verwendung. Wegen des hohen ionischen Widerstands dieses Materials müssen sie allerdings bei Temperaturen von über 900 °C betrieben werden, was ihre Lebensdauer beeinträchtigt und die Systemkosten erhöht. Diese Nachteile könnten durch eine Senkung der Betriebstemperatur behoben werden. Dafür muss allerdings der Widerstand der Zellen gesenkt werden, was durch die Verwendung von alternativen Elektrolytmaterialien mit grösserer Leitfähigkeit erreicht werden kann.

Ionische Leiter auf der Basis von Ceroxid sind vielversprechende Kandidaten für den Ersatz von Zirkonoxid, da sie eine ionische Leitfähigkeit besitzen, die drei bis fünf mal höher liegt. Diese Materialien werden jedoch bei tiefen Sauerstoffpartialdrucken, wie sie auf der Anodenseite von Brennstoffzellen herrschen, leicht reduziert und entwickeln zusätzlich elektronische Leitfähigkeit. Unter Betriebsbedingungen verhalten sie sich als ionisch/elektronische Mischleiter und wurden bisher deshalb nicht in Betracht gezogen.

Das Ziel dieser Studie ist es, die Verwendungsmöglichkeit von Ceroxid als mischleitendem Elektrolyt in Brennstoffzellen zu zeigen und die physikalischen und elektrochemischen Eigenschaften solcher Systeme zu bestimmen. Die elektrische Leitfähigkeit, Sauerstoffnichtstöchiometrie und die thermische Ausdehnung wird als Funktion des chemischen Potentials des Sauerstoffs untersucht. Die elektrochemischen Eigenschaften von Brennstoffzellen mit Ceroxid-Elektrolyten mit verschiedenen Kathoden und Anoden werden mit Hilfe eines neuen Modells interpretiert. Für das Verständnis dieses Modells ist es wichtig, den Verbund von Kathode/Elektrolyt/Anode als Materialsystem aufzufassen.

Dieses Modell erlaubt eine konsistente Beschreibung der Sauerstofftransport-eigenschaften von Festelektrolyt-Brennstoffzellen mit ionisch/elektronischen Misch-leitern mittels expliziter analytischer Gleichungen. Partielle elektronische und ionische Ströme innerhalb dieses Brennstoffzellen-Systems können mit Hilfe dieser Gleichungen als Funktion der Arbeitsbedingungen ermittelt werden. Hierfür müssen allerdings die elektrochemischen Eigenschaften der Elektroden bekannt sein, die in dieser Arbeit mit einem Konzept der Überlagerung von einer Ladungstransfer- und einer Diffusions-überspannung behandelt werden.

3 Summary / Zusammenfassung

Brennstoffzellen mit Ceroxid-Elektrolyten weisen bei 700 °C eine vergleichbare Leistungsdichte auf wie Brennstoffzellen mit Zirkonoxid-Elektrolyten bei 900 °C. Wegen der gemischten ionisch/elektronischen Leitfähigkeit ist jedoch bei Ceroxid der Wirkungsgrad der Umwandlung von chemischer in elektrische Energie etwas geringer. Die niedrige Betriebstemperatur auf der einen und der etwas geringere Wirkungsgrad auf der anderen Seite machen Ceroxid vor allem für kleinere Anlagen mit Kraft-Wärme Kopplung interessant.

Brennstoffzellen mit Ceroxidelektrolyten stellen ausserdem höhere Anforderungen an das System-Engineering. Aufgrund der Mischleitung besitzen sie nur einen kleinen Stromdichtebereich, in dem sie mit hoher Leistung und gutem Wirkungsgrad betrieben werden können. Wegen der zunehmenden Reduktion des Ceroxids bei Temperaturen von über 800 °C ist ein Betrieb bei tieferen Temperaturen von Vorteil.

2. Introduction to Mixed Ionic Electronic Conductors in Solid Oxide Fuel Cells

1. Solid oxide fuel cells

Fuel cells are energy conversion devices which convert chemical energy directly to electrical energy and heat by electrochemical combination of a fuel with an oxidant. Fuel cells offer substantially higher conversion efficiencies and lower emission of pollutants [1, 2] than other energy conversions systems. There exist a number fuel cell types which differ in electrode and electrolyte materials used, in the requirements on fuel and oxidant purity, and in operating temperature [2]. Solid oxide fuel cells (SOFC) have several advantages over other fuel cell types. Due to their operation temperature of more than 900 °C they are flexible in the use of fuel and offer the possibility of cogeneration of electrical power and heat. The high operating temperature enables to operate the fuel cell without precious electrocatalysts at the electrodes. The use of solid oxygen conductors as electrolyte eliminates corrosion and electrolyte management problems as they are present in other fuel cell types [1]. Solid oxide fuel cells consist of two electrodes and an electrolyte. To the anode fuel is (e.g. hydrogen or hydrocarbons) supplied continuously. Oxidant (air or pure oxygen) is fed to the cathode. The electrolyte of SOFC is an oxygen ion (O=) conductor. The working principle of this type of fuel cell is described as follows: Oxygen is fed to the cathode where it adsorbs and is reduced according to

O e O2 4 2+ → = (2-1)

These oxygen ions enter the electrolyte and react at the anode with the fuel gas, e.g. hydrogen:

2 2 2 42 2H O H O e+ → += (2-2)

The overall fuel cell reaction is given as

2 22 2 2H O H O+ → (2-3)

The overall fuel cell reaction can therefore be regarded as a spatially separated combustion reaction of hydrogen with oxygen, where the electrons which are transferred in the reaction are used as electrical energy in an external circuit. The driving force for the oxygen ions is the chemical potential difference between anode (low oxygen chemical potential) and cathode (high oxygen chemical potential). These electrode/electrolyte compounds are assembled to fuel cell stacks, i.e. they are connected in series with an interconnector between the single cells. This interconnector separates the gas room of fuel and oxidant and must be electrically conducting. It is important to treat this interconnector - single cell compound as one functional materials system, since all materials have to be chemically and mechanically compatible and the

5 Introduction

weakest link of the chain determines the performance of the whole system. Traditionally, this systems engineering point of view has not always been considered with the result, that until today no really competitive SOFC system is available. State of the art SOFC are based on the standard materials yttria stabilized zirconia for the electrolyte, La(1-x)SrxMnO3 for the cathode, La-chromite for the interconnect and Ni-cermets for the anode [3, 4]. These fuel cells operate at temperatures from 900 - 1000 °C, due to the thermally activated ionic conductivity of zirconia and the thermally activated electrochemical reactions at the electrodes. For smaller combined heat power systems lower operation temperatures are of advantage as some cell components could be fabricated from less expensive materials (e.g. metallic interconnects). For this purpose, the operation temperature has to be lowered to temperatures of 700 - 800 °C [5]. In the following sections an overview over electrolyte, cathode, anode and interconnector materials is given, and the requirements to the single materials to act as one materials system are formulated.

2. Solid electrolytes

The electrolyte for SOFC is an oxygen ionic conductor. Basically, its thickness and ionic conductivity are a necessary but not sufficient criterion to define the operating temperature of the fuel cell. The electrolyte has to be stable under the oxidizing atmosphere prevailing at the cathode and under the reducing atmosphere at the anode. High oxygen ionic conductivity has been found in oxides which possess an open crystal structure, such as fluorites and related structures and perovskites [6, 7]. Most of the good ionic conductors exhibit fluorite related structures. This structure is relatively open and has an exceptional tolerance for high levels of atomic disorder introduced either by doping, or nonstoichiometry due to oxidation or reduction [7 - 89]. By doping with divalent or trivalent oxides oxygen vacancies are created which give rise to oxygen ionic conductivity. As the atomic disorder, however, extents on the anion lattice only, the mobility of the cations in the fluorite structure is orders of magnitude lower. Examples of oxides with the fluorite structure and high ionic conductivity are doped CeO2 [

10-1112], doped ZrO2 [7, 13] or PuO2

[14]. Additionally, the δ-Bi2O3 phase possesses an anion deficient fluorite structure for which the highest oxygen ionic conductivities have been reported so far [6]. In perovskite related structures comparable ionic conductivities have been reported only recently for doped LaGaO3 [15]. High ionic conductivities have also been found for the perovskite related brownmillerite structure in which high oxygen ionic conductivity is introduced by an order-disorder transition, an example for this is doped Ba2In2O5 [

16].

Introduction 6

Most of these compounds have been found to be unstable under the reducing atmosphere prevailing at the anode side of SOFC, e.g. Bi2O3 is reduced completely and forms a metallic melt when exposed to hydrogen at high temperatures. Also PuO2 is disqualified for obvious reasons. The only compounds with sufficient stability under oxidizing as well as reducing conditions are doped ZrO2 and doped CeO2. Also doped LaGaO3 seems to be stable, however, only very little is known up to now about this new compound. Compositions with scandia or yttria exhibit the best ionic conductivities among the zirconia systems [13, 17, 18]. For its high price scandia is also disqualified and most SOFC systems rely on zirconia doped with 8 to 10 mole% of Y2O3. This fact is interesting if one considers the ionic conductivity of doped ceria for comparison. Ceria doped with 20 mol% of GdO1.5 or SmO1.5 is reported to have an oxygen ionic conductivity which is 3 to 5 times higher than the conductivity of yttria doped zirconia and has been suggested as an alternative to the latter [10, 19 - 202122].

3. Mixed ionic electronic conductors

Under reducing conditions, as on the anode side of solid oxide fuel cells, ceria is partially reduced giving rise to mixed ionic electronic conductivity [9, 10]. This leads to a considerable ionic conductivity of the electrolyte material under fuel cell operating conditions. It is a common believe that electrolytes for SOFC have to be pure ionic conductors with an electronic conductivity level which has to be below 1 % of the total electrical conductivity. Therefore, ceria has normally not been considered as SOFC electrolyte. Despite this believe, ceria is an interesting candidate to replace zirconia in attempting to reduce the operating temperature of SOFCs due to its high ionic conductivity. The main emphasis of this dissertation is to identify the advantages and limitations of the use of ceria which is a mixed ionic electronic conductor as electrolyte in a SOFC system.

The idea of applying mixed conductors to electrochemical systems is of course not new. Relying on Wagner's theory for mixed ionic electronic conductors [23, 24], the possibility to use mixed ionic electronic conductors as SOFC electrolytes was pointed out by several authors [25-262728]. However, these considerations were of pure theoretical nature, assuming an ideal system with no electrolyte/electrode interactions. Despite the high operation temperature of solid oxide fuel cells, their electrodes are not reversible, i.e. massive changes in oxygen chemical potential occur at the electrodes. This change in oxygen chemical potential affects the conductivity of mixed ionic conductors (MIEC) and influences the I-V relations and the efficiency of the system. It is the aim of this thesis to quantify the changes in oxygen chemical potential at the electrode/MIEC interfaces and to determine the resulting partial ionic and electronic currents

7 Introduction

within the electrochemical system. From these considerations optimum system design and operating conditions are derived to yield a high electrochemical conversion efficiency and a high electrical power output. These electrochemical and system engineering considerations are presented in chapters 8 to 11.

4. Cathode

The electrochemical performance of SOFCs is controlled to a great extent by electrode processes. It is therefore important for the cathode to find materials with a high electrochemical activity for the oxygen reduction reaction. To prevent limitations due to supply of oxygen to the electrolyte, it is necessary to tailor the microstructure of the electrodes in terms of thickness and porosity with respect to their electrical and electrochemical properties. State of the art zirconia based SOFC systems employ La(1-x)SrxMnO3 (LSM) as cathode material [3, 4]. This material is an electronic conductor and is chemically compatible with zirconia electrolytes. It has a high activity for the dissociative adsorption of oxygen on the surface, however it possesses only very limited oxygen ionic conductivity [29]. For this reason, the transfer of oxygen from the electrode into the electrolyte is restricted to the contact area of the gas phase with the electrolyte and the cathode, the so called triple phase boundary. There exist also potential cathode materials with mixed ionic electronic conductivity allowing the oxygen ions to pass across the whole electrode/electrolyte cross section. An example for such a material is La(1-x)SrxCoO3 (LSC) and related perovskite structures doped with Fe or Ni [30 - 3132]. However, these materials react with zirconia based electrolytes and form insulating phases at the cathode/electrolyte interface [33]. Ceria based electrolytes on the other hand, are chemically compatible with LSC. This allows the construction of a mixed ionic electronic conducting cathode/electrolyte compound with a gradually changing ratio of ionic to electronic conductivity. The effect of the electrical transport properties of the cathode on cell performance is discussed for several cathode materials in chapter 9.

5. Anode

SOFC anodes should exhibit a good electrical conductivity and a high activity for the oxidation of the fuel gas. For state of the art SOFC systems a Ni-cermet, i.e. a mixture of metallic nickel and a ceramic component are used. The ceramic component normally is yttria-doped zirconia [2, 3]. The ceramic compound is added to the cermet to adjust the thermal expansion coefficient of the anode to the electrolyte and to prevent the nickel particles in the porous structure from coalescence. Since nickel is a metallic conductor and zirconia is an oxygen ionic

Introduction 8

conductor, this anode can also be regarded as a mixed ionic electronic conductor. An obvious alternative to Ni-cermet anodes based on zirconia are Ni-ceria cermet anodes in case ceria is used as electrolyte. The use of Ni-ceria cermet, LSC as cathode and ceria as electrolyte allows now to construct an electrode/electrolyte systems consisting only of mixed conductors.

6. Interconnector

For the interconnector basically the same properties are required as for the electrolyte. Except, its conductivity should be of pure electronic nature. Today's SOFC employ either ceramic interconnectors such as La(1-x)SrxCrO3 or Cr-based oxide dispersion strengthened alloys [2, 3, 34]. These materials are extremely costly for machining and for their high chromium content. An inexpensive alternative to these materials would be the use of conventional ferritic chromium steels. These steels can withstand SOFC conditions only up to temperatures of about 750 °C, therefore, a basic need for fuel cells operating only up to this temperature is obvious. Most electrochemical measurements of the present study were carried out on single cells avoiding the use of an interconnector. However, the properties of ferritic steels as interconnector were taken as boundary conditions for the development of a fuel cell system operating in the intermediate temperature range (600 to 800 °C).

7. Compatibility of the materials system

In today's SOFCs the electrolyte is the supporting and load bearing structural element in the single cell with a thickness of 150 to 250 µm. It supports the thin (10 to 30 µm) electrodes. The interconnector used for stacking the single cells is also a thick, stiff structure with a thickness of about 2 to 5 mm. The compatibility requirements for the materials system can be summarized as follows: • The thermal expansions of the single components of the cell have to match. This is after all

important for the thermal expansion of the load bearing structural elements. Except the thermal expansion of the porous thin electrodes might be somewhat different from the electrolyte without causing problems.

• Chemical compatibility is necessary. No intermediate phases should be formed at the interfaces during manufacturing and operation.

• The electrode performance should allow intermediate temperature operation.

9 Introduction

• High ionic conductivity of the electrolyte is necessary. The electrode performance alone is not sufficient for good performance of the whole system.

8. System efficiency and fuel cell design

In fuel cells with mixed conducting electrolytes a part of the ionic current through the electrolyte is compensated by an internal electronic short circuit current. For this reason not the entire chemical energy fed to the system as fuel is converted to useful electrical energy, in addition to the already existing losses by electrical resistance of electrolyte and electrodes. Under open circuit conditions, i.e. zero external load current, ionic and electronic current through the electrolyte are equal. The ratio of ionic to electronic current through the electrolyte, however, increases with decreasing cell voltage. This leads at fairly high ionic current densities to a maximum in conversion efficiency. Fuel cell systems with mixed conducting electrolytes have therefore to be operated under somewhat different conditions compared to solid oxide fuel cells with pure ionic conductors. The amount of ionic and electronic currents depends on the gradient of the electrochemical potential across the electrolyte. For very thin electrolytes it is obvious, that these partial currents increase. This leads to additional losses due to higher electrode polarization and to a lower fuel cell conversion energy. In chapter 11 the effects of mixed ionic electronic conductivity on the efficiency and its influence on fuel cell operation and design are discussed.

9. Terminology

In electrochemistry several terms are defined for materials with special properties and well defined abbreviations are used in general. However, most of the present work deals with materials which are difficult to describe in traditional electrochemical terms. To prevent confusion, some of the terms used throughout this work will now be defined. Traditionally the term "electrolyte" is used for a material or a solution with mobile ions which exhibits only ionic conductivity or at least an ionic conductivity which is orders of magnitude higher than its electronic conductivity. In this work the term "electrolyte" is also used for mixed ionic electronic conductors (MIECs), if it describes the structural element of a SOFC. The "electromotive force" of an electrochemical system is normally defined as the electrical voltage measured between cathode and anode of an electrochemical system. If the electrolyte is a pure ionic conductor this voltage E corresponds to the difference in chemical potential between cathode and anode and is given by the Nernst equation

Introduction 10

EG

zF= −

∆

(2-4)

where ∆G is the difference in chemical potential, z is the amount of electrons transferred per formula unit and F is the Faraday constant. For a MIEC as "electrolyte" this voltage is lowered by the internal electronic short circuit within the MIEC. To avoid confusion the voltage on the MIEC is defined further as V(MC) and the (hypothetical) voltage given by the (applied) chemical potential difference between cathode and anode is defined as Vth,app . An additional problem arises when measuring the electrode overpotential of electrodes on MIECs directly. As will be described in chapter 8, the measured overpotential is lower than the effective one. Traditionally the measured electrode overpotential is abbreviated as η or π. In the further descriptions η is used for the measured overpotential, the real effective overpotential is however different from the measured one and is abbreviated as δVth , i.e. as the deviation from the voltage Vth,app given by the applied chemical potential.

10. Structure of the thesis

This dissertation deals with mixed ionic electronic conductors in SOFC systems. The main emphasis is put on the understanding and description of the nature of the electrode/electrolyte system in terms of electrochemical behavior using MIECs. Processing and physical properties of these materials are discussed in chapters 4 to 7. The electrochemical properties of fuel cells with MIECs are discussed in chapters 8 to 10. In chapter 11 principles for fuel cell design and system engineering are derived. All chapters of this thesis are stand alone pieces of information. For each chapter the necessary experimental and theoretical background as well as the literature is included separately as some chapters are already submitted for publications. Due to this structure of the work repetitions and redundancy can not be avoided.

11. References

1. N.Q. Minh, J. Am. Ceram. Soc., 76, 563 (1993).

2. "Fuel Cell Systems", L.J.M.J. Blomen and M.N. Mugerwa, Eds., Plenum Press NY, USA

(1993).

3. M.C. Williams, in Solid Oxide Fuel Cells IV, M. Dokiya, O. Yamamoto, H. Tagawa and

S.C. Singhal, Eds., PV 95-1, p. 10, The Electrochem. Soc. Proceedings Series, Pennington NJ, USA (1995).

4. S.C. Singhal, in Proc. 3rd Internat. Symp. SOFC, S.C. Singhal and H. Iwahara, Eds.,

PV 93-4, p. 665, The Electrochem. Soc. Proceedings Series, Pennington NJ, USA (1993).

11 Introduction

5. H. Altdorfer, K. Honegger, in Proc. 1st Swiss Conference on Materials Research for

Engineering Systems, B. Ilschner, M. Hofmann and F. Meyer-Olbersleben, Eds. p.56, Technische

Rundschau, Sion, Switzerland (1994).

6. B.C.H. Steele, in "High Conductivity Solid Ionic Conductors", T. Takahashi, Editor,

World Scientific, Singapore (1989).

7. T.H. Etsell and S.N. Flengas, Chem. Rev., 70, 339 (1970).

8. J.A. Kilner and B.C.H. Steele, in "Nonstoichiometric Oxides", O.T. Sorensen, Editor,

p. 233, Academic Press New York, USA (1981).

9. H.L. Tuller, in "Nonstoichiometric Oxides", O.T. Sorensen, Editor, p. 271, Academic

Press New York, USA (1981).

10. H.L. Tuller, A.S. Nowick, J. Electrochem. Soc. 122, 255 (1975).

11. R.N. Blumenthal, F.S. Burgner, and J.E. Garnier, J. Electrochem. Soc., 120, 1230 (1973).

12. T. Kudo and H. Obayashi, J. Electrochem. Soc., 122, 142 (1975).

13. S.P.S. Badwal, J. Mat. Sci., 22, 4125 (1987).

14. Hj. Matzke, in [8], p. 155.

15. T. Ishihara, H. Minami, H. Matsuda, and Y. Takita, in [2], p. 344.

16. J.B. Goodenough, J.E. Ruiz-Diaz and Y.S. Zhen, Solid State Ionics, 44, 21 (1990).

17. F.T. Ciacchi and S.P.S. Badwal, J. Europ. Ceram. Soc., 7, 197 (1991).

18. T. Ishii and Y. Tajima, J. Electrochem. Soc., 141, 3450 (1994).

19. C. Milliken, S. Elangovan, and A.C. Khandkar, in "Ionic and Mixed Conductors", T.A.

Ramanarayanan, W.L. Worrell and H.L. Tuller, Eds., PV 94-12, p. 466, The Electrochem. Soc. Proceedings

Series, Pennington NJ, USA (1994).

20. M. Gödickemeier, K. Sasaki, P. Bohac, and L.J. Gauckler, in Proc. 6th IEA SOFC

Workshop, ENEA, Rome, Italy (1994).

21. C. Milliken, S. Elangovan, and A.C. Khandkar, in [2], p. 1049.

22. M. Gödickemeier, K. Sasaki, and L.J. Gauckler, in [2], p. 1072.

23. C. Wagner, Z. Phys. Chem., B 21, 25 (1993).

24. C. Wagner, Z. Elektrochemie, 60, 4 (1956).

25. N.S. Choudhury and J.W. Patterson, J. Electrochem. Soc., 118, 1398 (1971).

26. P.N. Ross, Jr. and T.G. Benjamin, Journal of Power Sources, 1, 311 (1976/77).

27. D.S. Tannhauser, J. Electrochem. Soc., 128, 1277 (1978).

28. I. Riess, J. Electrochem. Soc., 128, 2077 (1981).

29. B.C.H. Steele, S. Carter, J. Kajada, I. Kontoulis and J.A. Kilner, in Proc. 2nd Internat.

Symp. SOFC, p. 517, F. Grosz, S.C. Singhal and O. Yamamoto, Eds., Commission of the Europ. Communities,

Athens Greece (1991).

30. Y. Teraoka, H.M. Zhang, K. Okamoto and N. Yamazoe, Mat. Res. Bull., 23, 51 (1988).

31. Ch. Ftikos, S. Carter, and B.C.H. Steele, J. Europ. Ceram. Soc., 12, 79 (1993).

Introduction 12

32. L.-W. Tai, M.M. Nasrallah, H.U. Anderson, D.M. Sparlin and S.R. Sehlin, Solid State

Ionics, 76, 259 (1995).

33. B.A. van Hassel, T. Kawada, N. Sakai, H. Yokokawa, M. Dokiya and H.J.M.

Bouwmeester, Solid State Ionics, 66, 295 (1993).

34. W. Köck, H.P. Martinz, H. Greiner and M. Janousek, in [2], p. 841.

3. Aim of the Study The aim of this dissertation is to demonstrate the feasibility of mixed ionic electronic conductors (MIECs) with a high ionic and a moderate electronic conductivity as electrolytes for solid oxide fuel cells operating at intermediate temperatures. As model substance for such a material ceria solid solutions are selected. The electrical and electrochemical properties of solid oxide fuel cells (SOFCs) with ceria based electrolytes should be characterized. For this characterization the physical properties of the electrolyte and the electrochemical behavior of the electrode/electrolyte system should be established by appropriate means. To describe the cell current-voltage characteristics and electrode overpotentials a model should be proposed which correlates electrochemistry and defect chemistry under all operating conditions considering non-reversible electrodes. With this model it should be possible to determine the partial electronic and ionic currents in order to find the electrode overpotential - ionic current relations for cathode and anode. Guidelines for design of the subsystem cathode/electrolyte/anodes should be established and the optimum operating conditions are expected to be derived by this model.

4. Processing of Ceria Powders and Membranes

Abstract

Ceria solid solutions with 20% Sm3+ and 10% Ca2+ were prepared by wet chemical coprecipitation of the corresponding nitrates or chlorides with ammonium hydroxide and ammonium oxalate. After calcination in air the solid solutions exhibited the cubic fluorite structure which was verified by X-ray diffraction analysis. Precipitation with ammonium hydroxide resulted in agglomerated powders which were not sinterable, whereas precipitation with ammonium oxalate produced homogeneous powders with good sinterability. From these powders ceramic membranes with a thickness of 220±40 µm were prepared by tapecasting and sintering in air at 1600 to 1700 °C. The density of sintered ceria membranes was larger than 95% of the theoretical density (TD).

1. Introduction

Ceria solid solutions with Sm, Gd, Y and Ca belong to the best solid conductors for oxygen ions. They are relatively stable under oxidizing as well as under reducing atmospheres and are therefore considered as electrolytes for solide oxide fuel cells (SOFC) [1-23]. Ceria solid solutions can be prepared by mixing the powders of the corresponding oxides in a ball mill, pressing, and calcining at temperatures between 1700 and 1800 °C. This procedure yields a powder with low sinterability and usually results in sintered compacts with densities around 90% of TD [4, 5]. This density is too low for SOFC electrolyte applications, since the electrical conductivity is low and gas tightness can usually not be guaranteed for samples with less than 95% of TD. Moreover, the mechanical stability is reduced in porous materials. An alternative approach described in the present chapter is the one of chemically coprecipitated powders of ceria solid solutions starting with ceria and rare earth salts by adding e.g. ammonium hydroxide or ammonium oxalate. This procedure results in homogeneous fine powders. After calcination at temperatures between 400 and 800 °C for 2 h these powders decompose into high purity CeO2 - solid solutions with a high surface area and good sinterability. The preparation of powders and their characterization by surface area analysis (BET), X-ray diffraction and particle size distribution is subject of section 2. Most solid oxide fuel cell designs rely on the planar concept with the electrolyte as load bearing element in the cathode/electrolyte/anode compound [6, 7]. The most efficient method for the preparation for these relatively thick (~200 µm) membranes is tape-casting [8, 9]. This

15 Processing

process starts with a slurry of inorganic ceramic materials dispersed in a liquid containing dissolved organic binders and plasticizing modifiers and dispersants. The slip is spread on a surface by means of a doctor blade and the solvent is allowed to evaporate. After drying the tape is removed from the surface and is sintered. This tape-casting process is well established and used in the microelectronics industry, however, a crucial point for the application of this process using a new powder is the composition of the organic slurry. After all the choice of dispersants – "wetting agents" for the dispersion of ceramic powders in a liquid – is important to achieve a slurry with a low viscosity and a high solids loading. In section 3 the tapecasting process for ceria solid solutions is illustrated and recipes for the preparation of ceramic tapes with a green thickness of 300 to 400 µm are given. In section 4 the sintering of ceria-based membranes is described. For the sintering of ceramic tapes commonly alumina substrates are employed. Ceria solid solutions, however, tend to react with alumina at temperatures above 1450 °C to form cerium aluminate [10]. Therefore, magnesia or zirconia substrates were employed for the sintering of ceria membranes in this work. Ceria tapes with a density of >95% TD were obtained from sintering at temperatures above 1600 °C. The intention of this chapter is to describe the processing route from the nitrate precursors over the ceramic powder to the sintered ceramic membrane. It is thought as a collection of recipes for powder preparation and tapecasting.

2. Powder preparation

2.1 Hydroxide coprecipitation

Ceria solid solutions with 20% Sm3+ were prepared by coprecipitation with ammonium hydroxide. A solution of CeIII(NO3)3

.6H2O (>99%, Fluka, Buchs, CH) and Sm2O3 (>99%, Rhône-Poulenc, La Rochelle, F) dissolved in HCl was added dropwise to a ammonium hydroxide solution at 25 °C. The concentration of the of the ammonium hydroxide solution was 1 M and the solution of cerium nitrate and SmCl3 was 1 M in Ce3+ and 0.25 M in Sm3+. After precipitation the solution was stirred for 15 min and subsequently filtered through filter-paper (5893 Schleicher & Schüll, Feldbach, CH). The filtrate was washed 3 times with distilled water. For coprecipitation and washing the pH of the solution was controlled and adjusted to a pH-value of 8. The precipitate had a pink color which turned brownish when exposed to air. This change in color is attributed to an oxidation from Ce3+ to Ce4+ [11]. The gel-like precipitate was difficult to filter due to its high viscosity. After filtration the filtrate was dried overnight at 200 °C and calcined at temperatures of 500 to 1200 °C.

Processing 16

The calcined powders were die-pressed into pellets with a diameter of 30 mm and 10 mm in height at an axial pressure of 200 MPa. The pellets were sintered at 1600 °C/2 hrs with a heating rate of 2 °C/min. It was impossible to obtain dense crack-free pellets from this powders when sintered in air. Only sintering in argon (99.95% pure) resulted in crack-free samples. After cooling to ambient temperature these samples were dark blue and had a density of <80% of TD. The poor sinterability was attributed to strong agglomeration of the very reactive ceric oxide hydrate [9] which is formed by precipitation with ammonium hydroxide.

2.2 Ammonium oxalate coprecipitation

Upon thermolysis oxalates Men+(C2O4)n/2 decompose into MeOn/2 and 2CO2. Thereby a large quantity of carbon dioxide is liberated, preventing the particles from agglomeration. It is expected that those powders will sinter to dense materials. The starting materials used were CeIII(NO3)3

.6H2O (>99%, Fluka, Buchs, CH), Sm2O3 (>99%, Rhône-Poulenc, La Rochelle, F) dissolved in HCl, and Ca(NO3)2

.4H2O (>99%, Merck, Darmstadt, FRG). Ammonium oxalate (NH4)2(C2O4) (>98% Fluka, Buchs, CH) was used for precipitation of the metal cations. Ammonium oxalate was dissolved in 3 l of distilled water to obtain a solution with a concentration of 0.52 M. Due to the low solubility of ammonium oxalate at room temperature this solution was heated to 43 °C. The pH of the solution was adjusted to 7.5 by addition of small amounts of NH3. The nitrate precursor solution was prepared by dissolving 0.6 mol of CeIII(NO3)3

.6H2O plus 0.067 mol of Ca(NO3)2.4H2O in 0.6 l of distilled water. Samarium

chloride was used instead of calcium nitrate for the preparation of Sm-doped ceria. The salt solution was added dropwise (2 - 3 drops/s) to the ammonium oxalate solution while the precipitation bath was heavily stirred. Immediately a white precipitate was formed upon dropping into the ammonium oxalate solution. Subsequently, the suspension was stirred for another 15 min to assure complete precipitation and was then cooled down to room temperature. The precipitate was filtered with filter paper (5893 Schleicher & Schüll, Feldbach, CH) and washed twice with distilled water at a pH = 7.5. The oxalate powders were dried evaporating H2O at a pressure of 104 Pa and a temperature of 80 °C (Rotavap RE 111, Büchi, Flawil, CH). After drying the powders were calcined at temperatures of 400 to 1200 °C in air for 2 h to form the mixed oxides from the oxalate precipitates. During calcination traces of cyanides were formed due to reaction of oxalate and traces of NH3. The precipitation conditions for hydroxide and oxalate coprecipitation are listed in Tab. 4-1.

17 Processing

90% CeO2

10% CaO

80% CeO2

20% SmO1.5

80% CeO2

20% SmO1.5

Volume 600 ml 600 ml 600 ml

Precursor Ce(NO3)3.6H2O 1 M 1 M 1 M

Ca(NO3)2.4H2O 0.11 M

SmCl3 0.25 M 0.25 M

Temperature 24 °C 23 °C 23 °C

pH 2.8 - -

Volume 3000 ml 3000 ml 3000 ml

Precipitation (NH4)2C2O4.H2O 0.47 M 0.52 M

bath NH4OH 1 M

Temperature 45 °C 43 °C 25 °C

pH 7.7 8 9

Tab. 4-1 Precipitation parameters for Sm and Ca doped ceria.

Ceramic powders with small crystallite sizes (~20 - 30 nm) and high surface area exhibit the highest sintering activity. For ceramic forming on the other hand large amounts of organic additives are needed to handle very fine powders. Therefore, a compromise between sintering activity and the amount of organic additives is aimed for. For tapecasting of ceramic powders with organic slurries a surface area of the ceramic powders of ~10 m2/g was found to be desirable. The dependence of the specific surface area of calcined powders on the calcination temperature was followed by BET surface analysis (Area-Meter II, Ströhlein, Leonberg, FRG) for 20% samaria doped ceria (CSO), 10% calcia doped ceria (CCO) and a commercial 20% gadolinia doped ceria powder (CGO, Rhône-Poulenc, La Rochelle, F). The BET-surface area of CCO depending on the calcination temperature is presented in Fig. 4-1.

Processing 18

400 600 800 1000 12000

20

40

60

80d [nm

]S B

ET [m

2 /g]

T [°C]

0

200

400

600

800

1000

1200

Fig. 4-1 Bet surface area SBET versus calcination temperature (left scale) and corresponding

crystallite size d assuming spherical particles (right scale) for CCO.

Upon calcination at 400 °C the specific surface area of the samples increases from 1.77 m2/g (specific surface of the oxalate) to 82.8 m2/g. It decreases continuously to 0.74 m2/g at a calcination temperature of 1200 °C. After calcination at 750 °C the powder exhibits a specific surface area of 10.50 m2/g with a corresponding crystallite diameter of 85.2 nm. The decomposition of the oxalate was monitored by differential thermal analysis (DTA) in air (STA 501, Bähr, Hüllhorst, FRG) at a heating rate of 5 °C/min. In Fig. 4-2 the DTA signal versus temperature is shown for CCO. In the temperature range between 100 and 200°C the adsorbed water is removed (endothermic process) and the oxalate is completely decomposed to the corresponding mixed oxide (exothermic process) between 300 and 350 °C. The specific surface area of CSO prepared by coprecipitation of samarium chloride and cerium nitrate is presented in Fig. 4-3. The CSO powder shows a similar dependence of the surface area and crystallite size on calcination temperature as observed for CCO (Fig. 4-1). For hydroxide precipitated CSO the surface area decreases only slowly to 40 m2/g by calcination at 1000 °C, followed by a rapid decrease in specific surface at higher calcination temperatures. After calcination the powders were white to yellowish (CSO) or had a slightly brownish color (CCO).

19 Processing

T [°C]

DTA

[

V]

-20-10

01020304050

0 500 1000 1500

exot

herm

ic

µ

Fig. 4-2 Differential thermal analysis of the decomposition of CCO. The removal of adsorbed

water is observed in the temperature range of 100 to 200 °C. Between 300 and 350 °C

the oxalate is decomposed to the oxide solid solution .

400 600 800 1000 12000

20

40

60

80

d [nm]

S BET

[m2 /g

]

T [°C]

0

50

100

150

200

Fig. 4-3 BET surface area SBET versus calcination temperature (left scale) and corresponding

crystallite size d assuming spherical particles (right scale) for CSO. Circles: oxalate

precipitation; triangles: hydroxide precipitation.

Processing 20

As verified by X-ray diffraction measurements all powders exhibited the cubic fluorite structure. The cubic lattice constants of the different ceramic powders were measured with an X-ray diffractometer (D5000, Siemens, FRG) with CuKα radiation using a step width of 0.003 ° and a measuring time of 5 s per step between 2θ = 25 to 60°. Silver was used as an internal standard for calibration. As an example the XRD spectrum of CGO is given in Fig. 4-4. The composition of the samples was measured with microprobe analysis or EDX. Cubic lattice constants and sample composition with respect to Ce and Sm, Ca or Gd are summarized in Tab. 4-2.

25 30 35 40 45 50 55 60

(Ag 111)

(222)

(113)

(022)

(002)

(111)

Inte

nsity

[arb

. sca

le]

2θ / °

Fig. 4-4 XRD spectrum of CGO at room temperature.

Sample Composition [atom %]

Cubic lattice constant [Å]

Ionic radius [Å] of dopant [12]

Ce0.8Sm0.2O1.9 18.82X 5.4341±0.0008 1.079

81.18X

Ce0.8Gd0.2O1.9 19.96X 5.4312±0.0034 1.053

80.04X

Ce0.9Ca0.1O1.9 10.02E 5.4177±0.0014 1.12

89.98E X by microprobe (Sulzer Innotec, Winterthur, CH), E by EDX (Tracor Northern Z-max 30)

21 Processing

Tab. 4-2 Composition and cubic lattice constant of sintered ceria-based membranes. The ionic

radius of the host cation Ce4+ is 0.97 Å [12].

3. Tapecasting

3.1 Slurry preparation

A specific powder surface area of ~10m2/g was regarded as a good compromise between sinterability and amount of organic additives needed for the slurry. Therefore, the precipitated powders were calcined at 750 °C for 2 h. The particle size distributions of the powders were measured by sedigraphy (Sedigraph, Micromeritics®, Norcross GE, USA). The results are shown in Fig. 4-5. The specific surfaces of the powders are listed in Tab. 4-3.

Material SBET [m2/g] d50 [µm]

CSO 11.94 1.2

CGO 8.10 1.1

CCO 10.50 2.5

Tab. 4-3 Specific surface area SBET from BET and mean particle size d50 from Sedigraph.

For a similar specific surface area CCO shows a slightly higher d50 compared to CSO and CGO. A possible explanation of this finding is the formation of larger agglomerates in the CCO powder compared to CSO. The commercial CGO powder contained the largest fraction of sub-micron particles.

Processing 22

100 10 1 0.10

20

40

60

80

100

cum

ulat

ive

mas

s pe

rcen

t

CGO

CSO

CCO

equivalent spherical diameter [µm]

Fig. 4-5 Particle size distribution of the powders used for tape-casting.

Tapecasting slurries normally consist of a liquid, several plasticisers, a dispersant and a polymeric binder. In the present study ethanol was used as solvent. As plasticisers polyethylene glycol (PEG 400, Fluka AG, Buchs, CH) and dioctyl phthalate (C24H38O4, >97%, Fluka) were used. As binder in most commercial tapes PVB is employed [13], in the present study PVB (B 20H, Hoechst, Frankfurt, FRG) was used. The proper choice of dispersants turned out to be the crucial point for the preparation of tapecasting slurries. Two different organic additives were used as dispersants, glycerol trioleate ((C17H33COOCH2)2CHOCO(CH2)5, >99%, Fluka) and triethanol amine ((HOCH2CH2)3N, >99%, Fluka). Whereas both dispersants worked satisfactorily with the ceramic powders prepared in our laboratory (CSO and CCO), the dispersants were not optimal for the commercial powder (CGO). Tapes prepared from this powder always were brittle and cracked during drying into pieces of approx. 5 x 5 cm. This brittleness could also be traceable to the high sub-micron particle content of this powder consuming a large amount of the dispersants and binders. The composition of the tapecasting slurry was related to the surface area of the ceramic powder (SBET x weight of powder). It is given in Tab. 4-4.

23 Processing

Additives Additives per surface area of the ceramic powder [10-3 g/m2]

Glycerol trioleate or triethanol amine

1.44

PEG 400 4.9

Dioctyl phthalate 4.23

PVB 6.83

Ethanol 28 - 42

Tab. 4-4 Organic additives for the tapecasting slurry per surface area of the ceramic powder.

For the slurry 100 g ceramic powder were added into a mixture of ethanol and the dispersant. If the viscosity of this slurry was too high, i.e. homogeneous mixing could not be obtained, small amounts of ethanol were added until the slurry had a creamy consistence. The slurry was transferred into a zirconia beaker of 300 ml internal volume and 10 zirconia balls (15 mm in diam.) were added. The beaker was closed with a zirconia lid and placed in a planetary mill (Retsch, Haan, FRG). The slurry was homogenized during 4 h at 200 rpm. Subsequently the plasticisers and the binder were added to the slurry and the mixture was homogenized for additional 2 h. The treatment in the planetary mill led to a heating of the slurry to approx. 35 °C. The slurry was cooled to room temperature and degassed at 104 Pa for 15 min to remove small gas bubbles. The optimal viscosity of the slurry for tapecasting was found to be between 3.5 and 4 poise.

3.2 Tape-casting

A tape-casting equipment of EPH Associates Inc. (Orem, Utah, USA) was used for casting. The slurry was poured on a dry float glass plate and spread with a doctor blade with an opening of 350 to 700 µm between blade and glass plate at a speed of 30 cm/min. The slurry was dried at ambient temperature for 2 h and the resulting green tape was removed from the glass plate. The consistency of the tape was leathery in case of CCO and CSO but very brittle in case of CGO powders. The upper surface of the tape was smooth (CCO and CSO) whereas in case of CGO the tape exhibited microcracks parallel and perpendicular to the pulling direction. The bottom surface of the tape was very smooth and reflecting mirror-like. After drying the tapes had a green density of 30 to 40% in relation to the ceramic powder (i.e. percent of TD of the ceramic) and a thickness of 300 to 500 µm. The green densities of the tapes determined after burnout of the binder at 600 °C for 2 h are summarized in Tab. 4-5 together with the green densities, blade openings and the thicknesses of the sintered tape.

Processing 24

Sample green density [% of TD]

blade opening

[µm]

thickness after

drying [µm]

thickness after

sintering [µm]

lateral shrinkage

[%]

CCO 29 700 365 260 25

CSO 1 30 700 345 239 25

CSO 2 38 500 320

CSO 3 37 700 295

CGO 1 39 600 300 220 25

CGO 2 26 700 365 240 28

Tab. 4-5 Green density, blade opening, thickness after drying and sintering, and lateral

shrinkage of CCO, CGO and CSO tapes.

Although it seems to be favorable to have a high green density in the dried tape to minimize shrinkage, it turned out that tapes with green densities between 26 and 39% exhibited almost the same lateral shrinkage. Moreover tapes with a higher green density were brittle and therefore difficult to handle.

4. Sintering of tapes

4.1 Substrates

Ceria needs sintering temperatures higher than 1500 °C. The most commonly used substrate for sintering is alumina. Ceria, however is known to react with alumina to form CeAlO3 at temperatures above 1450 °C [10]. In order to check the compatibility with different polished substrates ceria tapes were sintered on MgO plates (Metoxit, Thayngen, CH), Al2O3 plates, 3 mol% yttria stabilized zirconia (TZP, Tosoh) and fully stabilized zirconia (Tosoh). Upon sintering at 1650 °C the tapes reacted with alumina substrates forming a yellow melt (CGO) or a dark green melt (CSO). CCO was laminated onto these substrates. Also on MgO and TZP the tapes could not be removed without destruction. The only feasible substrate was fully stabilized zirconia (YSZ). To ensure flatness of the tapes after sintering the tapes were loaded uniformly with 1.5 - 2 g powder per cm2. For the powder loading CGO powder was sintered at 1700 °C for 6 h to obtain a coarse inactive powder of the same composition. Care was taken to avoid contamination with alumina powders, since even smallest amounts of alumina formed highly reactive melts, which destroyed the ceria tapes.

25 Processing

4.2 Sintering and microstructure

The tapes were sintered at temperatures between 1200 and 1700 °C. The samples were heated with 1 °C/min to 600 °C to burn out the organic additives followed by heating with 3 °C/min to temperatures of 1600 to 1700 °C with a dwell time of 2 h at the maximum temperature. The cooling rate to room temperature was 5 °C/min. After sintering the tapes were light orange to brownish and were translucent. For CCO a sintering temperature of 1600 °C was high enough to form dense tapes. CSO and CGO were found to be less active. For these powders a sintering temperature of at least 1650 °C during 2 h was necessary to form tapes with a density of >95% TD. Moreover, sintered samples of CGO powder always exhibited spots of the size of one to several millimeters which were not translucent. SEM investigations revealed a slightly higher porosity in these spots (6 to 7% of porosity instead of 5%). In Fig. 4-6 to 4-8 SEM micrographs of the development of the microstructure as a function of maximum sintering temperature (without dwell time) is presented for CSO. In Fig. 4-9 the surface of a CSO tape sintered at 1650 °C with a dwell time of 2 h is shown. At 1200 °C the microstructure consists of grains with an average crystallite size of 0.25 µm agglomerated to rectangular platelets of ~1.5 to 2 µm with a thickness of 0.25 to 0.5 µm. These platelets form spherical grains with a diameter of 0.5 µm during sintering at 1400 °C to 1600 °C. The grain size for all three compositions as a function of maximum sintering temperature is presented in Fig. 4-10. CCO exhibited the rapidest grain growth of all compositions. The average grain size of the sintered tapes and the (closed) porosity determined by line intersection method from fracture surfaces of the tapes are summarized in Tab. 4-6.

Processing 26

Fig. 4-6 Microstructure of a CSO tape sintered at 1200 °C. The crystallite size is 0.25 µm. The

particles are agglomerated to platelets of ~1.5 x 2 x 0.5 µm.

Fig. 4-7 Microstructure of a CSO tape sintered at 1400 °C. The average grain size is 0.5 µm.

27 Processing

Fig. 4-8 Microstructure of a CSO tape sintered at 1600 °C. The average grain size is 2.1 µm.

Fig. 4-9 Surface of a CSO tape sintered at 1650 °C for 2 h.

Processing 28

1200 1300 1400 1500 16000

2

4

6

8

aver

age

grai

n si

ze [µ

m]

CGO

CSO

CCO

TS [°C]

Fig. 4-10 Average grain size of CSO, CCO and CGO as a function of sintering temperature.

Material average grain size closed porosity

CSO 7.7 µm (1650 °C/2 h) 3.1%

CGO 6.7 µm (1650 °C/2 h) 5%

CCO 7.5 µm (1600 °C/2 h) 1.3%

Tab. 4-6 Average grain size and closed porosity of sintered ceria tapes.

5. Summary

Ceria solid solutions with 20% Sm3+ or Gd3+ and compositions with 10% Ca2+ were prepared by coprecipitation. Precipitation with ammonium hydroxide led to very reactive cerium oxide hydrate which cracked during sintering. Precipitation with ammonium oxalate on the other hand produced fine, sinterable ceramic powders. By adequate choice of the calcination conditions the surface area of these powders was tailored to obtain a specific surface area of about 10 m2/g which was found suitable for tape-casting.

29 Processing

Tape-casting was carried out using organic slurries with homogeneously dispersed ceramic powders. The proper choice of the dispersant proved to be the crucial point for this process. Whereas the powders prepared in our institute (CSO and CCO) did not cause difficulties, the commercial CGO powder (Rhône-Poulenc, La Rochelle, F) was not fully compatible to the employed dispersants. These tapes were always very brittle and difficult to handle. Sintering of the tapes at temperatures of 1600 to 1650 °C led to tapes with an average grain size of ~7 µm and a closed porosity of <5%. To ensure flatness of the tapes they were sintered under a powder load of 1.5 to 2 g/cm2 of coarse ceria powder. The lateral shrinkage during sintering was around 25% and was found to be nearly independent of the green density of the tapes in the range of 26 to 39% of the theoretical density. 6. References

1. T. Kudo, Y. Obayashi, J. Electrochem. Soc. 123, 415 (1976).


3. H. Yahiro, Y. Eguchi, K. Eguchi, H. Arai, J. Appl. Electrochem., 18, 527 (1988).

4. R.T. Dirstine, R.N. Blumenthal, and T.F. Kuech, J. Electrochem. Soc., 126, 264 (1979).

5. M. Mogensen, T. Lindegaard, U.R. Hansen, and G. Mogensen, J. Electrochem. Soc., 141,

2122 (1994).

6. R. Diethelm and K. Honegger, in Proc. 3rd Internat. Symp. SOFC, S.C. Singhal and H.

Iwahara, Eds., PV 93-4, p. 822, The Electrochem. Soc., Pennington NJ, USA (1993).

7. N.Q. Minh, J. Am. Ceram. Soc., 76, 563 (1993).

8. E.P. Hyatt, Ceram. Bull. Am. Ceram. Soc., 65, 637 (1986).

9. H. Raeder, C. Simon, T. Chartire and H.L. Toftegaard, J. Europ. Ceram. Soc., 13, 485

(1994).

10. A. Cuneyt Tas and M. Akinc, J. Am. Ceram. Soc., 77, 2961 (1994).

11. B.T. Kilbourn, Cerium - a Guide to its Role in Chemical Technology, Molycorp, Inc.,

White Plains, NY, USA (1992).

12. R.D. Shannon, Acta Cryst. A, 32, 751 (1976).

13. J.S. Reed, Principles of Ceramic Processing, p. 395, Wiley & Sons, New York, USA

(1988).

5. Mixed Ionic Electronic Conductivity in Ceria-based Ceramic Membranes

Abstract

The electrical conductivity of tapecast membranes of ceria solid solutions with calcia (10 at. %), samaria and gadolinia (20 at. %) was measured in the temperature range from 400 to 850 K by impedance spectroscopy in air and by DC 4-pt measurements in an oxygen partial pressure range from 0.21 to 10-25 atm over a temperature range from 873 K to 1073 K. These measurements were performed in a Ca-stabilized zirconia cell in which the oxygen partial pressure was adjusted by pumping out oxygen electrochemically. At high oxygen partial pressure the conductivity is mainly ionic with an activation energy from 0.98 eV (Ca-doped) to 0.93 eV (Sm-doped) in the temperature region up to 670 K and an activation energy of 0.83 eV (Ca-doped) to 0.70 eV (Sm-doped) in the temperature region of 670 to 1073 K. All samples exhibited mixed conduction at low oxygen partial pressures. The electrical domain boundary (σe = σi ) at 973 K is at 1.5.10-20 atm for Sm-doped ceria, 10-20 atm for Gd-doped ceria and 6.10-18 atm for Ca-doped ceria. Combination with nonstoichiometry data (chapter 6) electronic mobilities in Sm, Gd and Ca doped ceria were calculated. The mobility was 1.76.10-3 for Sm-doped, 3.10-3 for Gd-doped and 4.22.10-3 for Ca-doped ceria at 973 K.

1. Introduction

Oxygen ion conducting solid electrolytes have been studied extensively for technological applications which include oxygen sensors, oxygen pumps and solid oxide fuel cells (SOFC). To minimize the electrolyte resistivity losses in devices such as oxygen pumps and fuel cells there is a need for solid electrolytes having a high ionic conductivity. A beneficial effect of a lower resistivity is also the possibility to decrease the operation temperature from today's 1173 to 1273 K (mainly using yttria-stabilized zirconia) to the intermediate temperature range of 873 to 1073 K. Doped CeO2, δ-Bi2O3 and certain perovskites such as the only recently discovered LaGaO3 [1] have been proposed as electrolytes in intermediate temperature SOFCs. The higher ionic conductivity of these materials compared to yttria-stabilized zirconia enables to operate fuel cells in the intermediate temperature range (873 - 1073 K). Among these alternative materials ceria solid solutions are of special interest due to their mechanical and thermodynamic stability.

31 Electrical Conductivity

Several authors [2-3456789] pointed out the possible usefulness of ceria as fuel cell electrolyte. Ceria possesses the cubic fluorite structure. It can dissolve up to 15 mole% of CaO and more than 50 mole% of different rare earth oxides such as SmO1.5, GdO1.5 or YO1.5 [10]. Ceria solid solutions are of special scientific interest, since they offer the possibility to investigate ionic conductivity in a solid oxygen ion conductor over a wide range of dopant concentrations without changing the lattice crystal structure. The ionic conductivity of pure ceria and ceria solid solutions has been investigated extensively in the last 20 years. The ionic conductivity of calcia doped ceria has been studied by several authors [11-12131415]. The conductivity of rare earth doped ceria was investigated by Kudo and Obayashi [4, 16] for ceria solid solutions with up to 50 mole% of LnO1.5 (Ln = lanthanide cation). It is generally agreed upon, that doping with divalent or trivalent cations results in substitution of Ce4+ in the lattice compensated by oxygen vacancies. The highest conductivities were found for a dopant level introducing 5% of vacant oxygen sites in the lattice, e.g. for a dopant level of 20 mole% of SmO1.5 and GdO1.5 [1718 - 19]. Whereas at high oxygen partial pressures doped ceria is an ionic conductor, at low oxygen partial pressures it is partially reduced and develops electronic conductivity. It is therefore denoted as a mixed ionic electronic conductor where unlike e.g. in yttria-stabilized zirconia only n-type electronic conductivity has been reported to date. Electronic conductivity studies of ceria solid solutions with YO1.5 [5] and CaO [11] at low oxygen partial pressures revealed n-type electronic conductivity, where the electronic conductivity was confirmed by Naik and Tien [20] and Tuller and Nowick [21] to proceed by a small polaron transport mechanism. Whereas for calcia and yttria doped ceria there exist reliable studies on the mixed ionic and electronic conductivity, for samaria and gadolinia doped ceria only preliminary and contradictory data are available. Kudo and Obayashi [4] studied the ionic transference number of oxygen by EMF studies. They found the electrical domain boundary (EDB), where ionic and electronic conductivity are equal at 1.24.10-19 atm at 700 °C. This method, however, is problematic since it generates values which are only system parameters, because the influence of the electrodes on the ionic transport number cannot be separated (see e.g. chapter 8&9). Maricle et al. [6] studied the ionic and electronic conductivity of Gd-doped ceria as a function of oxygen partial pressure and found the EDB at 3.5.10-19 atm. For small additions of Sm or Pr (1 to 3%) to gadolinia doped ceria they reported a lowering of the EDB of two orders of magnitude. The aim of the present study was to investigate the ionic and electronic conductivity of calcia, samaria and gadolinia doped ceria. The measurements were performed on tapecast membranes of solid solutions with 10 mole% CaO and 20 mole% of SmO1.5 or GdO1.5. Microstructural aspects such as grain size and porosity greatly influence the electric conductivity of polycrystalline materials. Impedance spectroscopy measurements were carried

Electrical Conductivity 32

out in the temperature range from 400 to 850 K and in the frequency range from 40 Hz to 1 MHz in order to evaluate grain boundary and intragrain conductivities. The specific grain boundary resistivity was evaluated by using the brick-layer model. The electrical conductivity as a function of oxygen partial pressure was measured by 4-pt measurements in the temperature range from 873 to 1073 K and at oxygen partial pressures ranging from 0.21 to 10-25 atm. In section 2 of this chapter the theory that describes the ionic and mixed ionic electronic conductivity is summarized. The sample preparation and the experimental setup are presented in section 3. Section 4 presents the results of the impedance spectroscopy study in air. The oxygen partial pressure dependent measurements are presented in section 5. Summary and conclusions are given in section 6.

2. Ionic and electronic conductivity

2.1 Ionic conductivity

The ionic conductivity is expressed as

σ νi iN q= ( )2 (5-1)

where σi denotes the ionic conductivity, (2q) the charge of an oxygen vacancy, N their concentration and νi the ionic mobility. According to [22] the ionic mobility can be expressed as

ν imB

THkT

= −exp( )∆

(5-2)

where k the Boltzmann constant and ∆Hm is the enthalpy of migration of an oxygen vacancy, the pre-exponential factor B contains geometrical terms and the jump attempt frequency. From Eqs. (5-1) and (5-2) the ionic conductivity is expressed as

σ iAq

BNT

EkT

= −2 exp( ) (5-3)

Where EA, the activation energy of the ionic conductivity contains the enthalpy of migration ∆Hm and additional terms due to association of oxygen vacancies with dopant cations. This association is mainly due to coulombic attraction of dopant cations and oxygen vacancies. For association of oxygen vacancies with divalent cations such as Ca2+ the activation energy is given as EA = ∆Hm+ ∆HA1/2 . For associates with trivalent dopants the association energy can be expressed as EA = ∆Hm+ ∆HA2 where ∆HA1 and ∆HA2 are the association enthalpies for divalent and trivalent associates, respectively [24]. Usually two different ranges of the ionic conductivity are distinguished. In the low temperature range (T < 1200 K) almost all oxygen vacancies are associated and the activation energy for the ionic conductivity is of the order of


0.8 to 0.9 eV. At temperatures above 1200 K the charge carriers are almost unassociated and the activation energy is of the order of 0.6 to 0.7 eV [22].

2.2 Electronic conductivity

At low oxygen partial pressures ceria is partially reduced and develops n-type electronic conductivity according to

O V e OOx

O↔ + +•• 2 12 2 (5-4)

With the law of mass action it follows, that

n K N p(Oe=− −

intr1

21

42) (5-5)

where n denotes electron concentration and Ke is a constant. Nintr is the number of additional oxygen vacancies introduced by partial reduction. Similar to the ionic conductivity, the electronic conductivity is expressed as

σ νi enq= (5-6)

and the dependence of the electronic conductivity on the oxygen partial pressure is given as (Eqs. (5-5) and (5-6))

σ e p(O m∝ −2

1) (5-7)

The exponent 1/m in Eq. (5-7) is equal to 0.25, if the number of oxygen vacancies N can be assumed as constant. If the number of oxygen vacancies introduced by reduction increases at very low oxygen partial pressures, 1/m approaches 1/6. For association of quasifree electrons with oxygen vacancies, the exponent 1/m can increase up to 1/2 (chapter 6).

3. Experimental

3.1 Sample preparation

Ce0.9Ca0.1O1.9-x (CCO) was prepared by wet-chemical coprecipitation from nitrate precursors. An aqueous solution ([Ce3+] = 1 M, [Ca2+] = 0.11 M) of CeIII(NO3)3

.6H2O (>99%, Fluka, Buchs, CH) and Ca(NO3)2

.4H2O (>99%, Merck, Darmstadt, FRG) was precipitated with a 0.47 M solution of ammonium oxalate (>99.5%, Fluka) at 45 °C. This


solution was washed three times with water at a pH of 8 and calcined at 750 °C for 2 hrs. Ce0.8Sm0.2O1.9-x (CSO) was prepared under the same conditions with a Sm-nitrate solution prepared from Sm2O3 (>99%, Rhône-Poulenc, La Rochelle, F) dissolved in nitric acid. Ce0.8Gd0.2O1.9-x (CGO) powder was obtained from Rhône-Poulenc (Lot. NR. 94007/99, La Rochelle, F). The exact procedure for the powder preparation is given in chapter 4. All samples had the cubic fluorite structure as was verified by XRD. The composition was analyzed by microprobe analysis and by EDX. Composition with respect to Ce4+ and dopant as well as lattice constants and ionic radii are given in Tab. 5-1. As minor impurities SiO2 and ZrO2 were found (below 0.1% wt.).



Ce0.8Sm0.2O1.9 18.82X 5.4341

81.18X

Ce0.8Gd0.2O1.9 19.96X 5.4312

80.04X

Ce0.9Ca0.1O1.9 10.02E 5.4177




3.2 Tapecasting and sintering

100 g of ceria powder was mixed into a organic slurry using ethanol as solvent (30 g), polyethylene glycol 400 (2.95 g) (pract., Fluka) and dioctyl-phtalate (2.8 g) (>97%, Fluka) as plasticisers, triethanolamine (>99% Fluka) as dispersant (0.38 g), and PVB (Mowital B 20H, Hoechst, Frankfurt, FRG) as binder (5.8 g). The slurry was homogenized in a planetary mill, degassed at a pressure of 100 mbar for 15 min and cast into green ceramic tapes by use of a doctor blade technique. Samples were cut from the green tapes and sintered at 1650 °C. To ensure flatness of the samples they were sintered on polished yttria-stabilized zirconia plates and covered with ~1.5 to 2 gcm-2 of a coarse powder with the same composition. For the impedance spectroscopy study three layers of tape (1 x 1 cm2) were laminated with ethanol and sintered together. For the 4-pt conductivity study strips of 0.5 to 3 cm were sintered. The sample had a density of more than 95% of the theoretical density and an average grain size of 3.6 to 6.1 µm. Average


grain size and density were determined from scanning electron micrographs (Jeol, JEM-6400). Density and average grain size of the samples used in this study are summarized in Tab. 2.

Sample Theor. density [g/cm3]

% of TD measured

Av. grain size [µm]

Ce0.8Sm0.2O1.9 7.14 98 ± 1 6.1 impedance study

97 ± 2 7.7 4-pt measurement

Ce0.8Gd0.2O1.9 7.21 99 ± 1 5.2 impedance study


Ce0.9Ca0.1O1.9 6.73 96 ± 2 3.6 impedance study


Tab. 5-2 Theoretical density, measured density and average grain size of the samples used for

the impedance and 4-pt conductivity studies.

Typical microstructures of sintered tapes are shown in Figs. 5-1 (as sintered surface of a CSO tape) and 5-2 (fracture surface of CSO).

3.3 Impedance spectroscopy

Complex impedance spectroscopy in the frequency range from 40 Hz to 1 MHz was used to distinguish between the intragrain and the grain boundary contribution to the charge transport. The total impedance of polycrystalline ceramic specimens originates from the impedance contribution of the grains and from interfacial effects on grain boundaries and electrodes.


Fig. 5-1 SEM micrograph of the surface of a sintered CSO tape (1650 °C, 2 h).

Fig. 5-2 SEM micrograph of the fracture surface of a sintered CSO tape (1650 °C, 2 h).


Complex impedance measurements using a PC-controlled HP Precision LCR meter 4284A were carried out between 400 and 1100 K in air on rectangular samples with a thickness of 0.8 to 1 mm and an area of ~1 cm2 using a two-probe method with separate wires for current and voltage. Platinum electrodes were applied by sputtering (Union SCD-040, Balzers, FL) a layer of 0.5 µm of platinum onto the polished surface of the specimens. Pt-mesh (Aldrich 99.9%, 52 mesh, Aldrich, Milwaukee, USA) was attached by platinum paste (Heraeus C3605 P, Heraeus, Hanau, FRG) on these sputtered electrodes. The temperature was controlled to ± 2 °C close to the specimen. The data were analyzed using the NLLS-fit program of Boukamp [24]. The specific resistivity of the grain boundaries was evaluated by use of the brick-layer model [14, 25]. According to this model, it is assumed that the bulk material consists of conducting grains, approximated by cubes with an edge length a, separated by a thin homogeneous grain boundary layer of thickness δGB. For δGB << a and σGB << σG , the parallel conduction along the grain boundaries is negligible. The equivalent circuit model of the sample then consists of a simple series combination of three RC-elements (grain, grain boundary and electrodes). An idealized schematic picture of the observed impedance plot in the complex plane is given in Fig. 5-3. The spectrum in Fig. 5-3 consists of three semicircles which may be tilted below the real axis Z' , since the sample reacts not as an ideal RC element. The frequency ω increases with decreasing resistance. Below the schematic impedance plot, the equivalent circuit is given with the corresponding resistors and capacitors for grain (RG, CG), grain boundary (RGB, CGB) and electrode dispersions (RE, CE). The resistance of the electrode in this case corresponds to the charge transfer resistance and the capacitance to the electrochemical double layer capacity. The relaxation time τ of the single processes can be determined from the crest of the semicircles were ωτ = 1 . The total (DC) resistance of the ceramic material RT = RG + RGB, is the sum of the resistance of grains RG and grain boundaries RGB. Usually macroscopic specific conductivities σT', σG' and σGB' are calculated from resistivity values using the relation

( )σ ρi i iT

TR

LA

' '= =

−−

11

(5-8)

where LT = LG + LGB = n(a+δGB) is the total length of the sample, n = LT/a is the number of grain boundaries perpendicular to the current direction, and AT = AG + AGB is the total cross-sectional area of the sample. Since LGB/LT = δGB / (a + dGB), it can be shown that the length fraction of grain boundaries LGB amounts to L δGB / (a + δGB) ≈ LT δGB/a and the length fraction of grains LG to LT a / (a + δGB) ≈ LT. For the cross section of the current path through the area fractions of grains AG we obtain:


AA a

aAG

T

GBT=

+≈

2

2( )δ (5-9)

If the grain boundary conduction path along the length of the sample can be omitted, the microscopic specific conductivity of grain and grain boundary is given by

σ σG G= ' and σ σδ

GB GBGBa

= ' (5-10)

In order to eliminate the size effect of ceramic grains, the macroscopic specific grain boundary resistivities, ρGB' = RGB AT / LT, have to be converted to the normalized macroscopic resistivities rGB per unit area of the grain boundary surface, according to the relation

r aGB GB GB GB= =ρ ρ δ' (5-11)

where a = 1.5l is the corresponding cube edge length derived from the linear average grain size l of the ceramic sample [26].

Fig. 5-3 Idealized complex plane impedance plot schematically showing the observed

semicircles for grain, grain boundary and electrodes. Below the impedance plot the

equivalent electronic circuit is given.

3.4 Four-point electrical conductivity measurements

4-pt conductivity measurements were carried out on sintered strips of 0.5 x 3 cm with a thickness of 250 µm. On these strips Pt electrodes were painted and sintered at a temperature


of 800 °C for 2 hrs. The length of the sample between the inner electrodes (voltage probes) was 1 cm. The specific resistance of the samples was measured as a function of oxygen partial pressure in the p(O2) range from 10-25 to 0.21 atm at 600, 650 700, 750 and 800 °C. The measurement setup is shown schematically in Fig. 5-4. The samples were placed in a Ca-stabilized zirconia tube which was closed at one end by a brass flange and a polymer (PMMA) lid. Between the flange and the lid vacuum grease was applied to ensure gas-tightness. Within the zirconia tube the oxygen partial pressure was adjusted by pumping out oxygen electrochemically by application of a voltage between the lower electrodes with a laboratory power supply (GPR 3030D, G.W. Instruments, Taiwan). The pumping current was adjusted to 6 - 10 mA and the oxygen partial pressure was calculated from the EMF of the upper electrodes by the Nernst equation measured with an electrometer (Keithley 517 programmable electrometer). For the conductivity measurements oxygen was removed from the cell until a p(O2) of about 10-25 atm was reached. Subsequently, the current was switched off and the resistivity of the sample was measured with a micro-ohmmeter (Keithley 580). Leakage through the seals on top of the cell led to a slow increase of p(O2) corresponding to a increase in EMF of about 0.025 V/hr. The resistivity of the sample, the EMF of the upper electrodes and the temperature close to the sample were recorded with a pen-chart recorder (SE120, ABB Metrawatt) connected to the analog output of the respective devices. The EMF varies very quickly with time in an oxygen partial pressure range from 10-15 to 10-5 atm for a constant oxygen leakage current. Therefore, in this oxygen partial pressure range the samples were not in equilibrium with the surrounding atmosphere and were not included. Only for CSO at 873 K special care was taken to determine the conductivity also in this range. By 4-pt measurements the total conductivity of the sample, σt, is measured. It is the sum of electronic, σe, and ionic, σi, conductivities. The total conductivity in air was assumed as the ionic conductivity and the electronic conductivity of the samples was calculated as

σe = σi - σtot(air) (5-12)

since for high level doping the ionic conductivity of the sample can be assumed to be constant over a wide range of oxygen partial pressures.


1 PMMA lid 2 brass flanges 3 upper electrodes for EMF sensing (Pt) 4 lower electrodes for p(O2) adjustment (Pt) 5 thermocouple (type K, Interstar, Cham, CH)) 6 sample in 4-pt setup (Pt electrodes) 7 Ca-stabilized zirconia tube (Zr 23, Degussa, Baar, CH) A Micro-Ohmmeter (Keithley 580) B Furnace controller (CRL455, Thermotronic, Bern, CH) C Digital multimeter (Keithley 517 Programmable Electrometer) D Digital multimeter (Keithley 179A) E Power supply (GPR-3030D, G.W. Instruments Co., Taiwan) F Electrical furnace

Fig. 5-4 Schematic setup for the 4-pt electrical conductivity measurements as a function of

temperature and oxygen partial pressure. The cell volume was 13.9 cm3.


4. Results and discussion

4.1 Impedance spectroscopy

Impedance spectroscopy measurements were carried out in the frequency range of 40 Hz to 1 MHz. For this reason only a small window of the grain, grain boundary and electrode resistances as depicted schematically in Fig. 5-3 is visible at a given temperature. The intragrain conductivity of CSO, CGO and CCO can be measured up to a temperature of 850 K. Total conductivity of the samples and grain boundary resistance can be measured accurately up to temperatures of 900 K. In Fig. 5-5 a plot of log(frequency) versus real part of the resistivity is shown for CGO in the temperature range from 559 to 1122 K. In this plot the dispersions on grain, grain boundary and electrodes are visible for different temperature regimes. Small numbers indicate the temperature in Kelvin.

1 2 3 4 5 610-1

100

101

102

103

104

Z' [ Ω

m]

CGO

electrodes

grain boundary

grain

559

627

711

749

861

942

1032

1122

1070

log (frequency / Hz)

Fig. 5-5 Grain, grain boundary, and electrode dispersions of the real part of complex specific

impedance for CGO in air.

A typical impedance plot with real and imaginary part of the specific resistivity is given in Fig. 5-6 for a CSO sample consisting of three laminated tapes at 481 K. Small numbers indicate the corresponding frequency in Hertz.


0 7500 15000 22500 300000

5000

10000

15000

20000Z"

[Ωm

]

100K

1M

10K1K

100

481 K

Z' [Ωm]

Fig. 5-6 Complex impedance diagram at 481 K (CSO, three laminated tapes) in air.

By using a NLLS-fit program ("Equivalent circuit", Boukamp [24]) the specific intragrain conductivity was determined in the temperature range from 400 to 850 K in air. The activation energy of the ionic conductivity of CSO, CGO and CCO in this temperature range was found to vary from 0.93 - 0.98 eV in the temperature range of 400 to 673 K to 0.77 - 0.83 eV in the temperature range from 673 to 850 K. The ionic conductivity of the three compositions is shown in Fig. 5-7 as σiT versus 103/T. The corresponding activation energy and the ionic conductivity are summarized in Tab. 5-3. These measurements were carried out on laminated tapes and are slightly lower than the conductivities measured for single tapes as well as for the 4-pt conductivity measurements. It was found for these samples, that not the entire cross-section of the tapes were well laminated and therefore the effective microscopic cross-section was slightly smaller than the macroscopic one. However, this does not influence the activation energy. The activation energy is found to decrease with increasing temperature. Assuming a concept of association of oxygen vacancies with dopant cations, this can be explained with a decreasing degree of association at higher temperature leading to a higher mobility of oxygen vacancies.


1.2 1.4 1.6 1.8 2.0 2.2 2.410-3

10-2

10-1

100

101

102

103

specific intragrain conductivity

σ iT

(Gra

in) [

K S

/m] CSO

CGO CCO

103/T [1/K]

Fig. 5-7 Specific intragrain conductivity measured by impedance spectroscopy of CSO, CGO

and CCO depicted as σiT vs. 103/T.

Composition Intragrain conductivity at 800 K [S/m]

Activation energy 400..........673 K

Activation energy 673..........850 K

Ce0.8Sm0.2O1.9 0.32 0.93 eV 0.78 eV

Ce0.8Gd0.2O1.9 0.28 0.97 eV 0.77 eV

Ce0.9Ca0.1O1.9 0.12 0.98 eV 0.83 eV

Tab. 5-3 Intragrain ionic conductivity at 800 K and activation energies for the intragrain

conductivity estimated following Eq. (5-3) for CSO, CGO and CCO.

The activation energies compare well to values given by Arai et al. [15], who found an average activation energy of 0.89 eV (660 to 1100 K) for Ce0.9Ca0.1O1.9 . Kudo and Obayashi [16] found activation energies of 0.98 eV for Ce0.7Sm0.3O1.85 in the temperature range from 700 to 900 K and 0.78 eV for 900 to 1100 K. For Ce0.7Gd0.3O1.85 they found activation energies of 1.05 and 0.84 eV for the low and high temperature range, respectively. The normalized macroscopic grain boundary resistance rGB of the samples was determined from the macroscopic specific grain boundary resistivity according to the brick-layer model (Eqs.(5-9) to (5-11)) to eliminate the size effect of the grain size of the samples. It is equal to


the product of specific grain boundary resistance ρGB and the thickness of the grain boundaries δGB. In Fig. 5-8 ρGBδGBT is shown versus 103/T for the three investigated materials CSO, CGO, and CCO. The activation energy of the grain boundary conductivity was determined from this plot as 1.18 eV for CSO, 1.11 eV for CGO and 1.02 eV for CCO. Unlike for the intragrain conductivity no different activation energies were found for the low and high temperature range.

1.0 1.2 1.4 1.6 1.8 2.0 2.210-9

10-8

10-7

10-6

10-5

10-4

10-3

ρ GB

δ GBT

-1 [

Ωm

2 K-1]

CCO 1.18 eVCGO 1.11 eVCSO 1.02 eV

103/T [1/K]

Fig. 5-8 Macroscopic specific grain boundary resistivity of CSO, CGO and CCO as ρGBδGBT

vs. 103/T.

For CSO the highest macroscopic specific grain boundary resistivity was found followed by CGO and CCO which exhibited only small differences. The values of the activation energy of ρGBδGBT versus 103/T compare well to values found by El Adham and Hammou [14] who found an activation energy for the grain boundary resistance of 0.96 to 1.2 eV for polycrystalline samples of calcia doped ceria. Maricle et al. reported an activation energy for the grain boundary conductivity of 1.16 eV for Ce0.8Gd0.2O1.9 . For the electrochemical characterization of various electrodes on ceria-based electrolytes by current interruption measurements (chapters 8 to 10) it is important to know in which time scale the polarization of the electrolyte under load relaxes. Therefore, the relaxation time of the slowest process taking place in the bulk, i.e. the relaxation time of the grain boundaries, was estimated from the frequency on the crest of the semicircles observed in the impedance measurements (see Fig. 5-3). For the small frequency window, the relaxation time was only


observed in the temperature range from 400 to 670 K. The relaxation time of the grain boundaries for CSO, CGO and CCO is plotted versus 103/T in Fig. 5-9 and extrapolated to values up to 1000 K. In the intermediate operating temperature range the present ceria-based SOFC are designed for (873 to 1000 K), the grain and grain boundary polarization exhibits relaxation times below 0.1 µs for all investigated samples.

1.0 1.5 2.0 2.510-9

10-8

10-7

10-6

10-5

10-4

τre

laxa

tion

time

[s

]

CSOCGOCCO

103/T [1/K]

Fig. 5-9 Grain boundary relaxation times versus inverse temperature estimated from the

frequency on top of the observed semicircles in impedance spectroscopy.

5. Four point measurements

5.2 Total and ionic conductivity

4-pt resistivity measurements were carried out in the temperature range from 873 to 1173 K in steps of 50 K as a function of oxygen partial pressure. Fig. 5-10 shows the total electrical conductivity σtot of CSO, CGO and CCO as a function 103/T at an oxygen partial pressure of 0.21 atm. The corresponding activation energies calculated from σtotT vs. 1/T are 0.70 eV for CSO, 0.73 eV for CGO and 0.83 eV for CCO. Fig. 5-11 shows a comparison of the total electrical conductivities in air determined by impedance spectroscopy on single tapes of 250 µm thickness and 4-pt measurements on tape strips of 250 µm thickness as a function of inverse temperature. Open symbols represent 4-pt measurements, closed symbols impedance spectroscopy measurements. The measurements


are in good agreement, only for CSO slightly higher values were gained from 4-pt measurements.

0.9 1.0 1.1 1.2

1

3

10σ t

ot [S

/m]

CSO 0.70 eVCGO 0.73 eVCCO 0.83 eV

103/T [1/K]

Fig. 5-10 Total electrical conductivity of CSO, CGO and CCO in air determined by 4-pt

measurements on sintered polycrystalline tape strips.

0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2

10-4

10-3

10-2

10-1

100

101

σ tot [S

/m]

CSO

CGO

CCO

103/T [1/K]


Fig. 5-11 Arrhenius plot of the total electrical conductivity determined from 4-pt measurements

(open symbols) and impedance spectroscopy (closed symbols).

The total electrical conductivity of ceria solid solutions is the sum of the ionic and n-type electronic conductivity. At low oxygen partial pressures the electronic conductivity increases due to partial reduction of Ce4+ to Ce3+. The total electrical conductivity of Ce0.8Sm0.2O1.9-x (CSO) is plotted in Fig. 5-12 as a function of oxygen partial pressure for 873, 973 and 1073 K. In Fig. 5-13 and 5-14, the total electrical conductivity as a function of p(O2) is shown for Ce0.8Gd0.2O1.9-x and Ce0.9Ca0.1O1.9-x, where x indicates the deviation from the stoichiometric oxygen content due to partial reduction according to Eq. (5-4). The solid lines are fits to Eq. (5-7) where 1/m was fixed to 0.25. From these fit curves the electrical domain boundaries (EDB) for CSO, CCO and CGO, i.e. the oxygen partial pressure where the ionic conductivity equals the electronic conductivity, were determined. These EDB values are given in Fig. 5-15 as a function of inverse temperature. Arrhenius plots of the EDB versus 1/T revealed activation energies of 6.38 eV for CSO, 5.78 eV for CGO and 6.58 eV for CCO.

30 25 20 15 10 5 01

10

40

-log(pO2 / atm)

σ tot

[S/m

]

Ce0.8Sm0.2O1.9-x

1073 K

973 K

873 K

Fig. 5-12 Total electrical conductivity of CSO as a function of oxygen partial pressure.


30 25 20 15 10 5 01

10

100

-log(pO2 / atm)

σ tot

[S/m

]Ce0.8Gd0.2O1.9-x

1073 K

973 K

873 K

Fig. 5-13 Total electrical conductivity of CGO as a function of oxygen partial pressure.

25 20 15 10 5 0

1

10

60

-log(pO2 / atm)

σ tot

[S/m

]

Ce0.9Ca0.1O1.9-x

1073 k

973 K

873 K

Fig. 5-14 Total electrical conductivity of CCO as a function of oxygen partial pressure.


0.90 0.95 1.00 1.05 1.10 1.15 1.2010-2410-2310-2210-2110-2010-1910-1810-1710-1610-15

EDB

[atm

]

CSOCGOCCO

( )

103/T [1/K]

Fig. 5-15 Arrhenius plot of the electrolytic domain boundary (EDB) of CSO, CGO and CCO.

The electrical domain boundary and the ionic conductivity at 973 K, the activation energy of the EDB and the ionic conductivity are summarized in Tab. 5-4 and compared to literature values. The activation energy of the ionic conductivity found by 4-pt measurements in this work compares well to literature values of Kudo and Obayashi [4] and Maricle et al. [6]. The electrolytic domain boundary for CSO and CGO was one order of magnitude lower than that found in the literature, but the activation energy of the EDB compares well with values of [4]. The activation energy of the EDB includes the activation energy of the ionic mobility, the activation energy of the electronic mobility and the partial enthalpy of reduction for CSO, CGO and CCO, respectively. For CCO the lowest EDB-value of 5.99.10-18 was found at 973 K in this study. Maricle et al. found an increase of the electrolytic domain boundary by two orders of magnitude for small additions of samaria or praseodymia (1 to 3 atomic percent) compared to their Ce0.8Gd0.2O1.9. In this work the mechanism for the lowering of the EDB by small additions of Sm or Pr is however unclear. A possible explanation is a trapping of the electronic charge carriers introduced by partial reduction of the Sm or Pr cations.


Material σ i [S/m] 973 K

EA of σ i [eV]

EDB [atm] at 973 K

EA of EDB [eV]

Ce0.8Sm0.2O1.9-x 3.45 0.70 1.48.10-20 6.38

Ce0.8Gd0.2O1.9-x 3.62 0.73 1.14.10-20 5.78

Ce0.9Ca0.1O1.9-x 1.99 0.83 5.99.10-18 5.52

Ce0.8Gd0.2O1.9-x [4] 4.7 0.75 1.24.10-19 5.99

Ce0.8Gd0.2O1.9-x [6] 2.8 3.5.10-19

Ce0.8Gd0.17Pr0.03O1.9-x [6] 4.3 0.73 1.1.10-21

Tab. 5-4 Ionic conductivity activation energy for ionic conductivity, electrolytic domain boundary

(EDB) and activation energy of the EDB for CSO, CGO and CCO determined from 4-pt

measurements. For comparison literature data of Kudo and Obayashi [4] and Maricle

et al. [6] are given.

5.2 Electronic conductivity and mobility

The dependence of the electronic conductivity of ceria solid solutions on p(O2) should follow a power law with an exponent 1/m (Eq. (5-7)) which is ~1/4 for small deviations from stoichiometry and ~1/6 for large deviations from stoichiometry. However, if oxygen vacancies associate with Ce3+, the slope of σe vs. p(O2) can increase to values of ~1/3 to 1/2 (chapter 6). The electronic conductivity was calculated according to Eq. (5-14) from the total conductivity vs. p(O2). In Fig. 5-16 the dependence of the electronic conductivity of CSO on the oxygen partial pressure is given for 873, 973 and 1073 K. The slopes indicate the possible σe vs. p(O2) dependencies as explained in chapter 6. The location of the EDB is also shown.


12 14 16 18 20 22 24 26 28

1

10

50

σ e [S

/m]

EDB

EDB

EDB 1/41/6

1/3

1/4

CSO

1/3

1/4

1/6

1073 K973 K 873 K

-log(p(O2) / atm)

Fig. 5-16 Dependence of the electronic conductivity on the oxygen partial pressure for CSO at

873, 973 and 1073 K. Solid lines indicate the slope 1/m (Eq. (5-7).

From Fig. 5-16 it can be seen, that the electronic conductivity follows a p(O2)-1/4 law over wide ranges of the oxygen partial pressure. A fit of σe vs. p(O2)-1/4 can therefore be regarded as a reasonable approximation. In Fig. 5-17 the electronic conductivity of CSO, CGO and CCO is given as a function of oxygen partial pressure at 973 K. The values for CSO and CGO are more or less identical whereas CCO exhibits a significantly higher electronic conductivity. Therefore, CSO and CGO are regarded as suitable electrolyte materials for SOFC. CCO, due to its sufficiently high ionic conductivity and its higher electronic conductivity compared to CSO and CGO might be considered as an interesting material for SOFC anodes, since it is operated thereby at low p(O2) and can profit from its high ionic and electronic conductivity. By combination of these electronic conductivity values vs. oxygen partial pressure, with nonstoichiometry data, the electronic mobilities in ceria solid solutions with CSO, CGO and CCO can be determined. Fig. 5-18 shows the oxygen nonstoichiometry x vs. p(O2) at 973 K. The data are taken from chapter 6 and were obtained by coulometric titration.


16 18 20 22 24 26 28

1

10

80

σ e [S

/m]

973 K

CSOCCOCGO

1/3

EDBEDB

1/3

1/4

1/6

-log(p(O2) / atm)

Fig. 5-17 Dependence of the ionic conductivity on -log(p(O2)) for CSO, CGO and CCO at 973 K.

16 18 20 22 24 26 2810-3

10-2

10-1

x

1/4

1/3

- log (p(O2)/atm)

973 K

CGO

CSO

CCO

Fig. 5-18 Oxygen nonstoichiometry x of Ce0.8Sm0.2O1.9-x, Ce0.8Gd0.2O1.9-x and

Ce0.9Ca0.1O1.9-x as a function of -log(p(O2).


Fig. 5-17 and Fig. 5-18 exhibit approximately the same oxygen partial pressure dependence. From the σe vs. p(O2) and x vs. p(O2) measurements electronic mobility νe data were calculated according to ( with Eq. (5-6))

νσ

eea

q xcmVs

= ⋅

3 2

4 (5-13)

In Eq. (5-13) a signifies the cubic lattice constant (Tab. 5-1), q is the elementary charge and x is the nonstoichiometry. The factor 4 comes from the fact that an elementary cell of the fluorite lattice contains two formula units of ceria solid solution and that the electron concentration [e] = 2x. Similarly the ionic mobility νi is calculated from the ionic conductivity as

νσ

iia

q ycmVs

= ⋅

3 2

2 (5-14)

where y is the number of oxygen vacancies per formula unit. For the given compositions Ce0.8Sm0.2O1.9-x , Ce0.8Gd0.2O1.9-x , and Ce0.9Ca0.1O1.9-x y = 0.1. Since y >> x the influence of nonstoichiometry was neglected for the calculation of νi. The ionic mobility values are not discussed in further detail here, since they are proportional to the ionic conductivity values already discussed in section 5.1. Ionic mobility values and electronic mobilities at an oxygen nonstoichiometry x = 0.01 are summarized in Tab. 5-5. It should be noted that these values exhibit a large uncertainty since they were determined by two different measurements. However, they compare well with electronic mobility values for pure ceria given by Blumenthal and Panlener [27] who found electronic mobilities of 5.6.10-3 to 7.28.10-3 cm2V-1s-1 at 1073 K. Considering an activation energy of 0.1 to 0.5 eV for the electronic mobility as it is typical for a small polaron hopping mechanism which is reported for the electronic conductivity of ceria, these values compare also well to electronic mobilities at 1273 K reported by Tuller and Nowick [21] of 8.1 10-3 cm2V-1s-1 at x = 0.008 on ceria single crystals. Naik and Tien [20] found slightly higher mobility values of 2.10-2 cm2V-1s-1 at 1273 K and x = 0.025. From Tab. 5-5 it can be seen that the electronic mobilities are one order of magnitude higher than the ionic mobilities. Assuming an activation energy for the electronic mobility of ~0.4 eV (small polaron hopping) and an activation energy for the ionic mobility of ~0.8 eV, it is obvious that the ratio of νi/νe decreases with rising temperature. Moreover, the reducibility of ceria increases with temperature. Therefore, ceria-based materials can be regarded as suitable SOFC electrolytes in a temperature range below 1073 K only. Preferably they are operated at 800 to 1000 K.

Material T [K] ν e [cm2V-1s-1] ν i [cm2V-1s-1]


Ce0.8Sm0.2O1.9-x 973 1.76.10-3 1.73.10-4

1073 5.01.10-3 3.61.10-4

Ce0.8Gd0.2O1.9-x 973 3.00.10-3 1.81.10-4

1073 3.00.10-3 3.84.10-4

Ce0.9Ca0.1O1.9-x 973 4.22.10-3 9.89.10-5

1073 5.22.10-3 1.09.10-4

CeO2-x [21] 1273 8.1.10-3 x = 0.008

2.2.10-3 x = 0.025

CeO2-x [20] 1200 2.10-2 x = 0.025

CeO2-x [27] 1073 7.28.10-3

1273 9.19.10-3

Tab. 5-5 Ionic mobilities and electronic mobilities of ceria solid solutions (this work) at x = 0.01

compared to electronic mobilities reported in literature.

The highest electronic mobilities were found for CCO compared to CSO and CGO, however, even these are slightly lower than any data reported in literature. Possibly the electron mobility in ceria solid solution is diminished by association of electrons and oxygen vacancies. This is also seen from results of Tuller and Nowick [21] who found a decreasing electronic mobility with increasing defect concentration in pure ceria at 1273 K. Since the number of oxygen vacancies in ceria solid solutions is considerably higher than in pure ceria (due to nonstoichiometry and doping) an association would lower the electronic mobility compared to pure ceria.

6. Summary and conclusions

The electrical conductivity of tapecast membranes of ceria solid solutions with calcia (10 at. %), samaria and gadolinia (20 at. %) was measured in the temperature range from 400 to 850 K by impedance spectroscopy in air. From these measurements activation energies for the ionic conductivity in the bulk of 0.98 eV (Ca-doped) to 0.93 eV (Sm-doped) in the temperature region up to 670 K and an activation energy from 0.83 eV (Ca-doped) to 0.70 eV (Sm-doped) in a temperature region of 670 to 1073 K were determined. The activation energy for the grain boundary resistance was 1.02 eV (Sm-doped) to 1.18 eV (Ca-doped) over a temperature range from 450 to 900 K.


By DC 4-pt measurements in an oxygen partial pressure range from 0.21 to 10-25 atm over a temperature range from 873 K to 1073 K, electronic and ionic conductivities were determined. These measurements were performed in a Ca-stabilized zirconia cell in which the oxygen partial pressure was adjusted by pumping out oxygen electrochemically. At high oxygen partial pressure, the conductivity is mainly ionic, but all samples exhibited mixed conduction at low oxygen partial pressures. The electrical domain boundary (σe = σi ) at 973 K is at 1.5.10-20 atm for Sm-doped ceria, 10-20 atm for Gd-doped ceria and 6.10-18 atm for Ca-doped ceria. Combination with nonstoichiometry data (chapter 6) revealed electronic mobilities in Sm, Gd and Ca of 1.76.10-3 for Sm-doped, 3.10-3 for Gd-doped, and 4.22.10-3 for Ca-doped ceria at 973 K. The ratio of ionic to electronic mobility decreases with increasing temperature in favor to higher electronic conductivity. Therefore, it is concluded, that ceria solid solutions are suitable SOFC electrolytes in the temperature range below 1000 K only.

7. References

1. T. Ishihara, H. Minami, H. Matsuda, and Y. Takita, in Solid Oxide Fuel Cells IV, M.

Dokiya, O. Yamamoto, H. Tagawa and S.C. Singhal, Eds., PV 95-1, p. 344, The Electrochemical Soc.

Proceedings Series, Pennington NJ, USA (1995).

2. S. Singman, J. Electrochem. Soc., 113, 502 (1966).

3. H. Yahiro, Y. Eguchi, K. Eguchi, H. Arai, J. Appl. Electrochem., 18, 527 (1988).

4. T. Kudo, Y. Obayashi, J. Electrochem. Soc. 123, 415 (1976).


6. D.L. Maricle, T.E. Swarr and S. Karavolis, Solid State Ionics, 52, 173 (1992).

7. C. Milliken, S. Elangovan, and A.C. Khandkar, in Ref. [1], p. 1049.

8. F.P.F. van Berkel, G.M. Christie, F.H. van Heuveln and J.P.P. Huijsmans, in Ref. [1], p.

1062.

9. M. Gödickemeier, K. Sasaki and L.J. Gauckler, in Ref. [1], p. 1072.

10. T.H. Etsell and S.N. Flengas, Chem. Rev., 70, 339 (1970).

11. R.N. Blumenthal, F.S. Burgner, and J.E. Garnier, J. Electrochem. Soc., 120, 1230 (1973).

12. R.T. Dirstine, R.N. Blumenthal, and T.F. Kuech, J. Electrochem. Soc., 126, 264 (1979).

13. M.D. Hurley and D.K. Hohnke, J. Phys. Chem. Solids, 41, 1349 (1980).

14. K. El Adham and A. Hammou, Solid State Ionics, 9&10, 905 (1983).

15. H. Arai, T. Kunisaki, Y. Shimizu and T. Seiyama, Solid State Ionics, 20, 241 (1986).

16. T. Kudo and H. Obayashi, J. Electrochem. Soc., 122, 142 (1975). 17. I. Riess, D. Braunsthein, and D.S. Tannhauser, J. Am. Ceram. Soc., 64, 479 (1981).

18. T. Inoue, T. Setoguchi, K. Eguchi and H. Arai, Solid State Ionics, 35, 285 (1989).


19. K. Eguchi, T. Setoguchi, T. Inoue and H. Arai, Solid State Ionics, 52, 165 (1992).

20. I.K. Naik and T.Y. Tien, J. Phys. Chem. Solids, 39, 311 (1978).

21. H.L. Tuller and A.S. Nowick, J. Phys. Chem. Solids, 38, 859 (1977).

22. J.A. Kilner and B.C.H. Steele, in Nonstoichiometric Oxides, O.T. Sorensen, Editor, p.233,

Acad. Press, New York , USA (1981).

23. R.D. Shannon, Acta Cryst. A, 32,751 (1976).

24. B.A. Boukamp, Equivalent Circuit, Report CT88/265/128, University of Twente,

Enschede, NL (1989).

25. M. Gödickemeier, B. Michel, A. Orliukas, P. Bohac, K. Sasaki, L. Gauckler, H. Heinrich, P.

Schwander, G. Kostorz, H. Hofmann and O. Frei, J. Mat. Res., 9, 1228 (1994).

26. H.E. Exner und H.P. Hougardy, Einführung in die quantitative Gefügeanalyse, DGM

Informationsgesellschaft Verlag, p. 30, FRG (1986).

27. R.N. Blumenthal and R.J. Panlener, Phys. Chem. Solids, 31, 1190 (1969).

6. Nonstoichiometry and Defect Chemistry of Ceria Solid Solutions

Abstract

The nonstoichiometry x of CeO2-x solid solutions with 20% of SmO1.5 or GdO1.5 and 10% of CaO was investigated by isothermal coulometric titration using a solid electrolyte zirconia cell in the temperature range from 700 to 900 °C and in the oxygen partial pressure range from 10-12 to 10-25 atm. The nonstoichiometry of the 3 different compositions at 900 °C is comparable to pure ceria. At lower temperatures, however, pure ceria is more stable against reduction than ceria solid solutions. The dependence of nonstoichiometry on oxygen partial pressure suggests a defect model with oxygen vacancies and their associates with reduced cerium cations and dopant cations. At conditions prevailing on the anode side of solid oxide fuel cells the nonstoichiometry of ceria solid solutions is x = 0.005 to 0.03 at temperatures of 700 °C to 800 °C.

1. Introduction

Ceria and ceria solid solutions are reduced and become nonstoichiometric under the conditions prevailing at the anode side of solid oxide fuel cells (SOFC). The reduction of ceria introduces free electrons responsible for mixed ionic electronic conductivity. Furthermore, it leads to an isothermal expansion of the material which can cause mechanical failure of ceria-based SOFC membranes and delamination of the electrodes. The nonstoichiometry of pure undoped ceria was investigated intensively [1 - 23456789]. Most of these studies were carried out using thermogravimetrical methods [1-5] or solid state coulometric titration [8, 9] to determine the extent of nonstoichiometry. For ceria solid solutions on the other hand only limited nonstoichiometry studies exist. Garnier et al. [10] and Park et al. [11] measured the reducibility of Ca substituted ceria at 800 - 1500 °C by thermogravimetry and at temperatures between 850 and 920 °C by coulometric titration, respectively. For Gd substituted ceria, the nonstoichiometry at 1000 °C was investigated by Zachau-Christiansen et al. [12]. It was shown by Bevan and Kordis [1], that pure ceria retains its cubic fluorite structure up to a nonstoichiometry of x = 0.22 at temperatures above 685 °C. The partial enthalpy of oxygen, ∆H(O2), was calculated as ~ -9.8 eV for small nonstoichiometries (log x = -2.6) of pure ceria by Sorensen [4]. This is in good agreement with values found by Panlener et al. [2], who

Nonstoichiometry 58

found values around -10 eV for small deviations from stoichiometry (log x = -2.8). For 10% calcia substituted ceria on the other hand a partial enthalpy of oxygen of -8.64 eV at log x = -2.59 was found by Park et al. [10]. Different defect chemical models were suggested to describe the dependence of the nonstoichiometry on the oxygen partial pressure [2, 8, 13]. Panlener et al. [2] and Panhans et al. [8] suggested a model considering only doubly ionized oxygen vacancies without consideration of defect interactions. Tuller and Nowick [13] used a model including interactions between oxygen vacancies and quasi-free electrons introduced by deviation from nonstoichiometry also. The purpose of this study is to determine the nonstoichiometry of ceria solid solutions under SOFC operating conditions and to elucidate the dependence of nonstoichiometry on the oxygen partial pressure by a defect chemical model. The extent of reduction is also needed for the proper description of solid oxide fuel cells based on an electrolyte with mixed ionic electronic conductors over a wide range of operating conditions as outlined in chapter 8. Nonstoichiometry data of ceria solid solutions are obtained by solid state coulometric titration with a glass-sealed solid electrolyte coulometric titration cell. The experimental setup is described in section 2. In section 3 and 4 the results are presented and interpreted by a defect chemical model considering also defect interactions. A summary is given in section 5.

2. Experimental

2.1 Powder Preparation

Ce0.9Ca0.1O1.9-x was prepared by wet-chemical coprecipitation from nitrate precursors. An aqueous solution ([Ce3+] = 1 M, [Ca2+] = 0.11 M) of CeIII(NO3)3


.4H2O (>99%, Merck, Darmstadt, FRG) was precipitated with a 0.47 M solution of ammonium oxalate (>99.5%, Fluka) at 45 °C. This solution was washed three times with water at a pH of 8 and calcined at 750 °C for 2 hrs. Ce0.8Sm0.2O1.9-x was prepared under the same conditions with a Sm-nitrate solution prepared from Sm2O3 (>99%, Rhône-Poulenc, La Rochelle, F) dissolved in hydrochloric acid. Ce0.8Gd0.2O1.9-x powder was obtained from Rhône-Poulenc (Lot. No. 94007/99, La Rochelle). The exact procedure for the powder preparation is given in chapter 4. The composition was verified by microprobe analysis and by EDX. Composition of the samples is given in Tab. 6-1. Pure ceria powder was obtained from a commercial source (>99.9%, Aldrich, Milwaukee, WI, USA).

Sample Composition [atom %] Specific surface BET [m2/g]

Ce0.8Sm0.2O1.9 18.8 Sm 11.9

81.2 Ce

59 Nonstoichiometry

Ce0.8Gd0.2O1.9 19.6 Gd 16.6

80.2 Ce

0.2 La

Ce0.9Ca0.1O1.9 9.9 Ca 10.5

88.9 Ce

1.2 Zr

Tab. 6-1 Composition and specific surface area of the powder samples used for coulometric

titration.

2.2 Solid-state coulometric titration

2.2.1 Experimental setup. The experimental setup with the coulometric titration cell is shown in Fig. 1. The cell consists of an alumina tube 14.5 mm in diameter 20.8 mm high, which is closed at one end. The top of the cell is an 8 mole% yttria stabilized zirconia (YSZ) solid electrolyte disk with a thickness of 1 mm. On both sides of the electrolyte platinum paste (Heraeus C 3605 S) was applied as electrodes. The inner electrode was connected to the measurement circuit through the seal. Ceria solid solutions were filled into the cavity. It is well known, that Ce3+ tends to react with alumina to form cerium aluminate CeAlO3 [14], however this reaction is only observed at temperatures above 1450 °C. The measurement setup consists of a programmable power supply (GPM 6030, Good Will Instrument CO., Taiwan) and a scanning digital multimeter (Keithley, DMM 2001) equipped with a 10 channel scanning card for the detection of electrical current and EMF. The measurement is controlled by an IBM compatible personal computer connected to the measurement devices by an IEEE-bus. The measurement of the EMF and the integration of current after time was accomplished by the measurement program written in Basic™. The sample was placed in an electrically heated furnace and the temperature was controlled by a thermocouple placed on top of the solid electrolyte. The solid electrolyte with the platinum electrodes (Fig. 1) served as both oxygen pump and EMF sensor.

Nonstoichiometry 60

Fig. 6-1 Solid electrolyte coulometric titration cell for the nonstoichiometry measurements.

2.2.2 Measurement. Approximately 4 g of sample powder was weighted into the cell. To seal the cell an organic slurry containing a glass (Cerdec 90016, Degussa, Frankfurt, FRG) with a good wettability [15] for zirconia as well as for alumina was applied between the alumina crucible and the electrolyte. The slurry consisted of a mixture of ethanol (23 wt.%) PVB (2 wt.% Mowital B20H, Hoechst, Frankfurt, FRG) and the glass powder (75 wt.%). The cell was heated to 920 °C and then slowly cooled down to 900 °C. In a first coulometric titration run oxygen was removed from the cell. The gas-tightness of the cell was verified by following the EMF vs. time. When a steady state was reached, i.e. the EMF decreased by less than 0.3 mV/min the cell was judged to be sufficiently gas-tight. If the EMF change was higher the cell was again cooled down to room-temperature and another layer of glass paste was applied. This procedure was repeated until a gas-tight cell was obtained. For this easy cell assembly only about 20% of the cells were gas-tight for the first time and the sealing procedure had to be performed up to 5 times to obtain gas-tight cells. The oxygen content in nonstoichiometric ceria is changed by the electrochemical reaction

CeO x O CeO ex2 2 2− ⋅ ↔ +=− (6-1)

61 Nonstoichiometry

where the oxygen content of the sample can be changed by electrochemically pumping out oxygen from or to the solid electrolyte cell (Coulometric titration). The electrical charge Q which is exchanged during the titration is given as

Q I t= ⋅ (6-2)

where I is the electrical current across the solid electrolyte. By integration of this current over time the total electrical charge is obtained. However, in this calculation we neglect the charge which might be exchanged due to the minor n and p type electronic conductivity of the YSZ solid electrolyte. From the exchanged electrical charge the exchanged amount of oxygen is calculated as

n OI t

Ftot( )= =⋅⋅2 (6-3)

where F is the Faraday constant. From this value n(O2), the amount of gaseous oxygen which is contained in the titration cell before measurement has to be subtracted and the amount of oxygen removed from the sample is then given as

n O n O n Otot( ) ( ) ( )= == − ⋅2 2 (6-4)

where oxygen is treated as an ideal gas. The nonstoichiometry of the material is expressed as

xn O

mM

=⋅

=( )δ (6-5)

In Eq. (5) δ denotes the stoichiometric composition of ceria solid solutions , i.e. δ = 2-y/2 for trivalent dopants (Sm, Gd) and δ = 2-y for divalent dopants (Ca). M is the weight of the sample material per formula unit (g/mol) and m is the sample mass (g). By measuring the electromotive force of the titration cell the corresponding equilibrium oxygen partial pressure is determined by the Nernst equation. The cell was cooled down to measurement temperature at 0.5 °C/min. Cooling rates higher than this value led to cracking of the glass seal, especially at temperatures below 800 °C. After reaching the measuring temperature of 700, 800 or 900 °C, the cells were held at this temperature for at least 30 min to allow for temperature stabilization. The cells were short circuited until the EMF was zero, i.e. the oxygen partial pressure at the inner electrode corresponded to the one at the outer electrode. The composition of the samples was assumed to be stoichiometric under these conditions. For the nonstoichiometry measurements oxygen was removed from the cell at a current of 5 to 15 mA, where the exact amount of oxygen was determined by integration of the electrical current over time. After a preset time the current was switched off and the stabilization of the EMF over time was observed until equilibrium was reached. An equilibrium criterion was introduced, where a change of EMF of less than 0.2 - 0.3 mV per minute was taken as

Nonstoichiometry 62

equilibrium criterion. Normally the equilibrium criterion was fulfilled after 60 - 200 min at 700 °C and 40 - 80 min at 900 °C. Measurements were carried out with Ce0.8Sm0.2O1.9, Ce0.8Gd0.2O1.9 and Ce0.9Ca0.1O1.9 at 700, 800 and 900 °C. A typical example of the time needed for equilibration is given in Fig. 6-2 for a Ce0.8Gd0.2O1.9-x sample at 700 °C. A measurement run with the titration period (a) and the equilibration period (b) is given in Fig. 6-3 for Ce0.9Ca0.1O1.9-x at 700 °C.

0 50 100 150 2000.95

1.00

1.05

1.10

1.15

1.20

1.25

t [min]

EM

F vs

. air

[V]

700 °C

Fig. 6-2 EMF measured over the YSZ-electrolyte after the titration current was switched off

(t =0) for different titration runs (Ce0.8Gd0.2O1.9-x at 700 °C).

63 Nonstoichiometry

0 20 40 60 80 100 120-40

-30

-20

-10

0

t [min]

log(p(O2 )/atm

)

b)a)

equilibrium p(O2)

700 °C

9.0

9.3

9.6

9.9

10.2

10.5tit

ratio

n cu

rren

t [m

A]

Fig. 6-3 Coulometric titration run with the titration period (a) and the equilibration period (b).


The dependence of the nonstoichiometry on the oxygen partial pressure was calculated according to Eqs. (6-3 - 6-5). The nonstoichiometry, x, versus p(O2) is shown in Figs. 6-4 a to c for 900, 800, and 700 °C. For comparison literature data [2] for pure CeO2-x are given also. Among the different ceria solid solutions only small differences in nonstoichiometry are found. The difference between the ceria solid solutions and pure ceria, however, increases with decreasing temperature. At 900 °C pure ceria and ceria solid solutions show the same stability against reduction. At lower temperatures pure ceria is more stable than ceria solid solutions, i.e. the partial enthalpy of oxygen is higher for pure ceria than for its solid solutions. This finding is in accordance with findings of Sorensen [4], and Park et al. [2], who found a partial enthalpy of oxygen of -9.8 to -10 eV at log(x) = 2.6 for pure ceria, and Garnier [10] or Park [11], who found a partial enthalpy of oxygen of -8.64 eV for 10% CaO doped ceria (log(x) = 2.6). Assuming a simple defect model as presented in section 4 of this chapter (see Eq. (6-12) to (6-15)) a partial enthalpy of oxygen ∆H(O2) of -7.45 eV is found for Ce0.8Gd0.2O1.9-x, -7.26 eV for Ce0.9Ca0.1O1.9-x, -7.24 eV for Ce0.8Sm0.2O1.9-x at log x = -2.6.

Nonstoichiometry 64

For comparison coulometric titration measurements with pure CeO2-x were carried out at 900 °C and compared to literature data obtained by thermogravimetric measurements. Fig. 6-5 shows a plot of nonstoichiometry against oxygen partial pressure for pure CeO2-x. The data obtained for pure ceria in this work are in good agreement with the literature data [2, 4]. The slope of the curves in fig. 6-5 was found to increase from ~1/5 at higher oxygen partial pressures to ~1/2 at p(O2) around 10-16 atm. At oxygen partial pressures lower than 10-18 atm the slope decreases again. In Fig. 6-6 the reciprocal of the slope of log(x) vs. -log(p(O2)) is plotted at 900 °C. It is given as m(x) = dlog p(O2) / dlog(x). For pure ceria (Ref. [2]) and Ce0.9Ca0.1O1.9-x a minimum of m(x) is observed at x 0.02 and x 0.05, respectively. For Ce0.8Gd0.2O1.9-x m(x) increases from 4 at lower nonstoichiometries to values comparable to pure ceria and Ce0.9Ca0.1O1.9-x. In the next section defect chemical modes are presented to explain this behavior. Values for Ce0.8Sm0.2O1.9-x are not included, since due to the lack of enough data points, they show considerable scattering.

9 12 15 18 21

-3.2

-2.8

-2.4

-2.0

-1.6

-1.2

-0.8

-0.4

log(

x)

- log (p(O2)/atm)

900 °C

Ce0.8Sm0.2O1.9-x

Ce0.8Gd0.2O1.9-x

Ce0.9Ca0.1O1.9-x

CeO2-x

CeO2-x Panlener [2]

Fig. 6-4a Dependence of nonstoichiometry of ceria solid solutions and pure ceria at 900 °C.

65 Nonstoichiometry

9 12 15 18 21

-3.6

-3.2

-2.8

-2.4

-2.0

-1.6

-1.2

-0.8

- log (p(O2)/atm)

log

(x)

800 °C

Ce0.8Sm0.2O1.9-x

Ce0.8Gd0.2O1.9-x

Ce0.9Ca0.1O1.9-x

CeO2-xPanlener [2]

Fig. 6-4b Dependence of nonstoichiometry of ceria solid solutions and pure ceria at 800 °C.

15 18 21 24 27

-3.6

-3.2

-2.8

-2.4

-2.0

-1.6

-1.2

-0.8

- log (p(O2)/atm)

log

(x)

700 °C

Ce0.8Sm0.2O1.9-x

Ce0.8Gd0.2O1.9-x

Ce0.9Ca0.1O1.9-x

CeO2-x Panlener [2]

Fig. 6-4c Dependence of nonstoichiometry of ceria solid solutions and pure ceria at 700 °C.

Nonstoichiometry 66

10 12 14 16 18 20 22-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

1/5

1/2

1/4

1/6

this workPanlener [2]Sorensen [4]

log(

x)

-log(p(O2) / atm)

Fig. 6-5 Dependence of nonstoichiometry on oxygen partial pressure compared to literature

data from thermogravimetric measurements at 900 °C.

-3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8

3

4

5

6

7

8

log (x)

CeO2-x

CCO

CGO

m(x

)

900 °C

Fig. 6-6 Slope of m(x) = dlog p(O2) / dlog(x) for pure CeO2-x [2], Ce0.8Gd0.2O1.9-x (CGO) and

Ce0.9Ca0.1O1.9-x (CCO) at 900 °C.

67 Nonstoichiometry

4. Defect models

In this section defect models are derived for pure ceria and ceria solid solutions with divalent or trivalent dopants. First a simple defect model is presented which, however, cannot explain the different slopes of the nonstoichiometry curves presented in Fig. 6-5 in section 3. Therefore, in a second step the simple model is extended to a more complex model allowing also defect interactions.

4.1 Simple Defect Model

The overall reduction reactions for ceria solid solutions are for divalent dopants (e.g. Ca2+)

Ce Ca O Ce Ca Ox

O gy y y y y y x1 2 1 2 22− − − − −↔ + ( ) (6-6)

For trivalent dopants (e.g. Sm3+) this reaction is

Ce Sm O Ce Sm Ox

O gy y y y y y x1 2 21 2 2

22− − − − −↔ + ( ) (6-7)

The site fractions of the different species can be expressed as follows:

[ ]Ce y xCex = − −1 2

(6-8)

[ ]Ce xCe' = 2

(6-9)

[ ] ( )O x yOx = − −1

2 2 and [ ] ( )V x yO•• = +1

2 for di-valent dopants (6-10a)

[ ] ( )O xOx y= − −1

2 22 and [ ] ( )V xOy•• = +1

2 2 for trivalent dopants (6-10b)

The reduction reaction in Kröger-Vink notation is

2 212

12Ce O Ce V O gCe

xOx K

Ce O+ ← → + +••' ( ) (6-11)

The mass action constant associated with the above reaction for di-valent dopants is given as

K Cax x y

y x y xp(O1

2

2 21

241 2 2

( )( )

( ) ( ))=

⋅ +− − ⋅ − −

⋅ (6-12)

For trivalent dopants the mass action constant is given as

K Smx x y

y x y xp(O1

2

2 21

24 21 2 2 2

( )( )

( ) ( ))=

⋅ +

− − ⋅ − −⋅ (6-13)

The corresponding relative Gibb's free energy of oxygen ∆G(O2) can be given as

Nonstoichiometry 68

∆ G O kT K( ) ln( )2 12= (6-14)

From a fit of ln(Κ1) against 1/T one can obtain the corresponding relative enthalpy and entropy of oxygen according to

∆ ∆ ∆G O H O T S O( ) ( ) ( )2 2 2= − (6-15)

Values for the relative enthalpy of oxygen are given already in section 3, but since only three temperatures were measured they are not very accurate. Simplifying the relation between p(O2) and x (Eq. (6-13) under the assumption that doubly ionized vacancies are dominating and defect interaction can be neglected we get

x p(O m x∝−

21

) ( ) (6-16)

where m(x) is the reciprocal slope in the plot log(x) vs. -log(p(O2) for di-valent dopants

m xd p(O

d xx

x yx

y xxy x

( )log( ))

log( )= = +

++

− −+

− −2 4

2 81 2

22

(6-17)

and for trivalent dopants

m xd p(O

d xx

x yx

y xx

y x( )

log( ))log( )

= = ++

+− −

+− −

2 42

2

81 2

2

2 2 (6-18)

From Eqs. (6-17) and (6-18) it follows, that for small deviations from stoichiometry the slope log(x) vs. log(p(O2)) is 1/m(x)= 1/6 for pure CeO2-x and 1/4 for ceria solid solutions with divalent and trivalent dopants (y >> x). In Fig. 6-7 the calculated reciprocal slopes m(x) for pure and doped ceria are shown. Using the simple defect model an increasing reciprocal slope m(x) is expected for increasing nonstoichiometry. This is in contradiction to the experimental findings for pure ceria and calcia doped ceria, and obviously the applied simple defect model has to be modified including defect interactions such as vacancy clustering and defect association.

69 Nonstoichiometry

10-4 10-3 10-2 10-13

4

5

6

7

8

9

10

m(x)

nonstoichiometry x

CeO2-x

Ce0.8Sm0.2O1.9-x

Ce0.9Ca0.1O1.9-x

Fig. 6-7 Calculated reciprocal slope m(x) depending on the oxygen nonstoichiometry for pure

ceria and ceria solid solutions (Eqs. (6-17) and (6-18)).

4.2 Associated Defects

Defects of opposite charge may form defect complexes in the ceria lattice due to their electrostatic interactions. Oxygen vacancies can form associates with one of two trivalent cations. These cations can either be reduced host cations or trivalent dopants such as e.g. Sm3+ or Gd3+(Eqs. (6-19a) - (6-19d).

( )Ce V Ce VCe OK

Ce O' '+ ← → +•• ••

•2

(6-19 a)

( ) ( )Ce Ce V Ce V CeCe Ce OK

Ce O Ce' ' ' '+ + ← →••

•••3

(6-19 b)

( )Sm V Sm VCe OK

Ce O' '+ ← →•• ••

•4

(6-19 c)

( ) ( )Sm Sm V Sm V SmCe Ce OK

Ce O Ce' ' ' '+ ← →••

•••5

(6-19 d)

Nonstoichiometry 70

With divalent dopants, such as Ca2+ neutral defect complexes are formed according to Eq. (6-19e).

( )Ca V Ca VCe OK

Ce O" "+ ← →•• ••6

(6-19 e)

K2 to K6 are the appropriate equilibrium constants for Eqs. (6-19a - 6-19e). The electroneutrality condition (Eq. (6-20)), the dopant concentration (Eq. (6-21)) and the oxygen nonstoichiometry (Eq. (6-22)) are expressed by the following equations for the example of di-valent doping

[ ] [ ] ( ) [ ]2 2⋅ + =

+ ⋅•••

••Ca Ce Ce V VCe Ce Ce O O" ' '

(6-20)

[ ] ( )Ca Ca V yCe Ce O" "+

=•• (6-21)

[ ] ( ) ( )Ce Ce V Ce V Ce xCe Ce O Ce O Ce' ' ' '+

+ ⋅

=•••

••2 2 (6-22)

From Eqs. (6-19a) to (6-22) a system of 3 equations is obtained to express the oxygen nonstoichiometry including defect interactions:

[ ] [ ] [ ] [ ] [ ]2 22⋅ + = ⋅ ⋅ + ⋅•• ••Ca Ce K Ce V VCe Ce Ce O O" ' '

(6-23)

[ ] [ ] [ ]Ca K Ca V yCe Ce O" "+ ⋅ ⋅ =••

6 (6-24)

[ ] [ ] [ ] [ ] [ ]Ce K Ce V K K Ce V xCe Ce O Ce O' ' '+ ⋅ ⋅ + ⋅ ⋅ =•• ••

2 2 32

2 2 (6-25)

Analogous equations can be formulated for trivalent dopants. For given parameters K1 - K6, y, and x this system can be solved and with Eq. (6-11) and Eq. (6-19) to (6-22) the dependence of log(p(O2)) on log(x) is expressed as follows:

[ ] [ ]log( )) log log

'

( ) ( )p(O K

Ce V

y x y x

Ce O2 1

2

22 21 2 2

− = −⋅

− − ⋅ − −

••

(6-26)

Equivalent equations for pure ceria as well as for ceria solid solutions with trivalent oxides can be calculated analogously. For these solid solutions one obtains

71 Nonstoichiometry

[ ] [ ]log( )) log log

'

( ) ( )p(O K

Ce V

y x x

Ce Oy2 1

2

22

2 21 2 2

− = −⋅

− − ⋅ − −

••

(6-27)

Calculations for pure ceria and ceria solid solutions with divalent and trivalent oxides with different K1 to K6 are shown in Fig. 6-8. For these model calculations the parameters K4, K5 and K6 were assumed equal to K2 for trivalent and to K2

2 for divalent dopants. The oxygen partial pressure was normalized by K1. In this figure it is shown, that an association of Ce3+ with oxygen vacancies leads to a decrease in oxygen vacancy activity and therefore to a steeper increase of the nonstoichiometry with decreasing oxygen partial pressure in an intermediate nonstoichiometry region (0.01 < x < 0.06) than for the simple defect model. At higher nonstoichiometries (x > 0.06) the slope of log(x) vs. log (p(O2)) is flattening again. Higher assumed defect association constants lead to steeper slopes of the nonstoichiometry curves. At very low nonstoichiometries, the defect interactions are small and the slopes of the nonstoichiometry curves approaches the one expected from Eq. (6-18), i.e. -1/4 for doped ceria and -1/6 for pure ceria. As shown in Fig. 6-9 the defect interaction increases with decreasing temperature. This results in a steeper slope at high oxygen nonstoichiometries, specifically for Ce0.9Ca0.1O1.9-x. Therefore, this composition exhibits the highest oxygen nonstoichiometry at low oxygen partial pressures for all three solid solutions examined in this study.

Nonstoichiometry 72

CeO2-x

pure ceria p K2 = K3 = 100 l K2 = K3 = 20 + K2 = K3 = 5 Ce0.9Ca0.1O1.9-x

divalent doping K6 = K2

2 p K2 = K3 = 100 l K2 = K3 = 20 + K2 = K3 = 10 Ce0.8Sm0.2O1.9

trivalent doping K4 = K5 = K2 p K2 = K3 = 100 l K2 = K3 = 20 + K2 = K3 = 5

Fig.6-8 Normalized calculated nonstoichiometry based on a defect model with associated

defects. The dotted lines show the -1/6 (pure ceria) and the -1/4 (solid solutions)

dependence of log(x) on log(p(O2)).

-15 -10 -510-3

10-2

10-1

1/6

-15 -10 -510-3

10-2

10-1

1/4

nons

toic

hiom

etry

x

-15 -10 -510-3

10-2

10-1

1/4

2 log(K1) - log(p(O2))

73 Nonstoichiometry

12 15 18 21 24-3.2

-2.8

-2.4

-2.0

-1.6

-1.2

-0.8

log(

x)

- log (p(O2)/atm)

700 °C

900 °C

800 °C

1/4

1/3

Ce0.8Sm0.2O1.9-x

Ce0.8Gd0.2O1.9-x

Ce0.9Ca0.1O1.9-x

Fig. 6-9 Oxygen nonstoichiometry vs. log(p(O2)) at 900, 800 and 700 °C for ceria solid

solutions with Ca2+, Sm3+ and Gd3+. Note defect interactions at higher log(x).

5. Summary

The nonstoichiometry of ceria solid solutions with 10% CaO, 20% SmO1.5 and 20% GdO1.5 was investigated by isothermal coulometric titration at 700, 800 and 900 °C in an oxygen partial pressure range from 10-12 to 10-25 atm. The ceria solid solutions show a slightly lower relative enthalpy of oxygen compared to pure ceria. This leads to a higher nonstoichiometry at 700 and 800 °C while the nonstoichiometry was found to be equal to pure ceria at 900 °C. To obtain quantitative data for the relative enthalpy and entropy of oxygen, more measurements at different temperatures would be necessary. Comparison of nonstoichiometry data obtained in this work by the use coulometric titration agree very well with literature data obtained by thermogravimetry. The dependence of nonstoichiometry of ceria solid solutions on oxygen partial pressure deviates slightly from -1/4, a value which would be expected for the simple defect model without defect interactions. The deviation is more pronounced for ceria calcia solid solutions than for ceria gadolinia solid solutions. For ceria samaria solid solution the amount of data points was not sufficient to conclude for defect interactions. The steeper slope 1/m(x) at nonstoichiometries between 0.01 < x < 0.06 at temperatures between 700 - 900 °C can be explained by association of oxygen vacancies with Ce3+

Nonstoichiometry 74

cations. However, similar results might be obtained assuming clusters of oxygen vacancies which might be formed at higher oxygen vacancy concentrations [16]. For ceria calcia solid solutions a stronger defect association was found than for ceria gadolinia solid solutions.

6. References

1. D.J.M. Bevan and J. Kordis, J. Inorg. Nucl. Chem., 26, 1509 (1964).

2. R.J. Panlener, R.N. Blumenthal and J.E. Garnier, J. Phys. Chem. Solids, 36, 1213 (1975).

3. B. Iwasaki and T. Katsura, Bull. Chem. Soc. Japan, 44, 1297 (1971).

4. O.T. Sorensen, J. Solid State Chem., 18, 217 (1976).

5. I.K. Naik and T.Y. Tien, J. Phys. Chem. Solids, 39, 311 (1978).

6. M. Ricken, J. Nölting, and I. Riess, J. Solid State Chem., 54, 89 (1984).

7. J. W. Dawicke and R.N. Blumenthal, J. Electrochem. Soc., 133, 904 (1986).

8. M.A. Panhans and R.N. Blumenthal, Solid State Ionics, 60, 279 (1993).

9. H. Janczikowski, Thermodynamische Untersuchungen an CeO2-x mittels einer

neuartigen elektrochemischen Festkörperzelle, Ph.D. Thesis, Georg-August-Universität Göttingen,

Germany (1985).

10. J.E. Garnier, R.N. Blumenthal, R.J. Panlener and K.R. Sharma, J. Phys. Chem. Solids, 37,

369 (1976).

11. J.-H. Park, R.N. Blumenthal, and M.A. Panhans, J. Electrochem. Soc., 135, 855 (1988).

12. B. Zachau-Christiansen, T. Jacobsen, K. West and S. Skaarup, Proc. 3rd Intl. Symp.

SOFC, S.C. Singhal and H. Iwahara, Editors, PV 93-4, p. 104, The Electrochem. Soc. Proceedings Series,

Pennington, NJ (1993).

13. H.L. Tuller and A.S. Nowick, J. Electrochem. Soc., 126, 209 (1979).

14. A. Cuneyt Tas and M. Akinc, J. Am. Ceram. Soc., 77, 2961 (1994).

15. H. Vogt, Evaluation von Glasloten für Brennstoffzellendichtungen, Diplomarbeit, ETH-

Zürich, Nichtmet. Werkstoffe, Report 94/01, Zürich, CH (1994).

16. O.T. Sorensen, in Nonstoichiometric Oxides, O.T. Sorensen, Editor, p. 66, Academic

Press, New York (1981).

7. Thermal and Isothermal Expansion of Ceria-based Ceramic Membranes

Abstract

Thermal and isothermal expansion of tapecast ceria membranes was investigated by differential dilatometry as a function of oxygen partial pressure in the temperature range from 50 to 900 °C. The compositions investigated were Ce0.8Sm0.2O1.9, Ce0.8Gd0.2O1.9 and Ce0.9Ca0.1O1.9. A unique form of tapecast specimens was used to allow a rapid equilibration in different oxygen atmospheres. The samples revealed an isothermal reduction of 0.35 to 0.72% at 900 °C when the atmosphere was changed from 0.21 to 5.10-18 atm. At temperatures up to 700 °C no isothermal expansion was detectable within experimental error. The linear thermal expansion coefficients of these membranes in air were 12.9.10-6/K (50 to 900 °C) for Ce0.8Gd0.2O1.9, 13.3.10-6/K for Ce0.8Sm0.2O1.9 and 14.0.10-6/K for Ce0.9Ca0.1O1.9.

1. Introduction

Solid oxide fuel cells (SOFC) consist of at least four different materials in intimate contact with each other. The match of thermal expansion of the cathode, the electrolyte, the anode and the interconnector is one of the most important criteria for the mechanical stability of such a multicomponent system. Especially, the expansion behavior of the thickest load bearing components has to be adjusted carefully, i.e. in SOFC with self-sustaining electrolyte membranes this is the expansion behavior of the electrolyte and the current-collector. Nonstoichiometric oxides such as doped ceria or La(1-x)SrxCrO3-δ have been reported to exhibit an expansion behavior depending not only on temperature but also on oxygen stoichiometry [1-23456789]. In SOFC ceria-based electrolyte membranes are exposed to different oxygen chemical potentials at the anode and at the cathode side respectively. An expansion behavior depending on oxygen nonstoichiometry can therefore lead to different expansion. This can result in case of membranes which are reduced at one surface only to a bending and to mechanical failure a was observed for plates made from La(1-x)SrxCrO3-δ [8, 9]. For pure CeO2-x extensive investigations on the thermal expansion behavior as function of the oxygen nonstoichiometry were carried out in order to investigate phase transitions in the system CeO2 to CeO1.8 by Körner et al. [1]. H.-W. Chiang et al. [6] determined the defect structure of CeO2 by combining the results of dilatometry and X-ray diffraction. Thermal

Thermal expansion 76

expansion behavior of ceria solid solutions with 10 mole% calcia was investigated by Chiang [10] in the temperature range from 500 to 900 °C as a function of oxygen nonstoichiometry. Mogensen et al. [4, 5] investigated the thermal expansion behavior as a function of oxygen partial pressure of gadolinia and calcia doped ceria at 1000 °C by dilatometry. The lattice constant of pure nonstoichiometric CeO2-x increases isothermally with rising oxygen nonstoichiometry x [6, 7, 9]. By comparison of XRD and dilatometry data [6, 10] and by comparison of XRD and pycnometric density data it was found, that the predominant defect in reduced ceria are oxygen vacancies due to the reduction of Ce4+ to Ce3+ according to Eq. (7-1)

O Ce V Ce O gOx

Cex

O Ce+ ↔ + +••2 2 12 2

' ( ) (7-1)

The ionic radius of Ce3+ (1.143 Å [11]) is larger compared to Ce4+ (0.97 Å [11]), therefore the lattice constant increases with increasing Ce3+ concentration. The increase in lattice constant is comparable to the increase found with increasing concentration of trivalent dopants with similar ionic radius such as La3+ (1.16 Å [11], [12]). The objective of the present study is to characterize the thermal expansion behavior of tapecast ceria-based membranes in oxidizing and reducing atmospheres. These membranes are potential candidates for the use in intermediate temperature SOFC (600 °C to 800 °C) due to the higher ionic conductivity of ceria-based materials compared to the presently used stabilized zirconia [13-1415161718]. Thermal and isothermal expansion measurements were performed on ceramic membranes with the composition Ce0.8Sm0.2O1.9-x, Ce0.8Gd0.2O1.9-x and Ce0.9Ca0.1O1.9-x. The measurements were accomplished with a differential dilatometer at temperatures between 600 and 900 °C and oxygen partial pressures between 1 to 10-25 atm.

2. Experimental

Ce0.9Ca0.1O1.9-x was prepared by wet-chemical coprecipitation from nitrate precursors. An aqueous solution ([Ce3+] = 1 M, [Ca2+] = 0.11 M) of CeIII(NO3)3


.4H2O (>99%, Merck, Darmstadt, FRG) was precipitated with a 0.47 M solution of ammonium oxalate (>99.5%, Fluka, CH) at 45 °C. This solution was washed three times with water at a pH of 8 and calcined at 750 °C for 2 hrs. Ce0.8Sm0.2O1.9-

x was prepared under the same conditions with a Sm-nitrate solution prepared from Sm2O3 (>99%, Rhône-Poulenc, La Rochelle, F) dissolved in nitric acid. Ce0.8Gd0.2O1.9-x powder was obtained from Rhône-Poulenc (Lot. No. 94007/99, La Rochelle). The exact procedure for the powder preparation is given in chapter 4. All samples had the cubic fluorite structure as was verified by XRD. The composition was checked by microprobe analysis and by EDX. Composition with respect to Ce4+ and dopant as well as lattice constants and ionic radii are

77 Thermal Expansion

given in Tab. 7-1. Minor impurities of SiO2 and ZrO2 were found, these impurities were below 0.1% wt. . The powders were mixed into an organic solution using ethanol as solvent, polyethylene glycol 400 (pract., Fluka) and dioctyl-phtalate (>97%, Fluka) as plasticisers, triethanolamine (>99% Fluka) as dispersant, and PVB (Mowital B 20H, Hoechst, Frankfurt, FRG) as binder. The slurry was homogenized in a planetary mill, degassed and cast into green ceramic tapes by use of a doctor blade technique. Further details on the tapecasting procedure are found in chapter 4.



Ionic radius [Å] of dopant [11]

Ce0.8Sm0.2O1.9 18.82X 5.4341 1.079

81.18X

Ce0.8Gd0.2O1.9 19.96X 5.4312 1.053

80.04X

Ce0.9Ca0.1O1.9 10.02E 5.4177 1.12




To assure a rapid equilibration of the samples under different oxygen partial pressures hollow samples with a wall thickness of 300 µm were prepared from green tapecast sheets. From these sheets pieces of approx. 1.5 to 1.5 cm were cut and joined together at the edges with ethanol to form a triangular hollow body (Fig. 7-1). The as formed specimens were sintered at 1650 °C for 2 hrs in a powder bed. The samples exhibited a yellow to orange color and had a density of >95% of the theoretical density after sintering. Expansion measurements were carried out on a Theta Differential Dilatometer (Dilatronic™ II). The length change of the three ceria samples was measured in reference to a sapphire single crystal with a length of 10.04 mm. Length changes of sample and reference were transduced to the measuring head (Dilaflex, 1/2%) by alumina pushrods. The temperature of the measuring head was kept at a constant temperature of 20±0.5 °C by circulating water through passages in the head with a Haake FE constant temperature circulator.


Fig. 7-1 Hollow dilatometer specimen formed from tapecast sheets. The wall thickness is 200

to 250 µm.

The oxygen partial pressure in the sample chamber was adjusted by the use of 4 different gas mixtures, 1 atm O2, dry air, pure Ar (99.999%), and 90% Ar-10% H2 humidified at ambient temperature. The gas flow was controlled by a flow meter (Vögtlin, V-100, Aesch BL, CH). The flow rate was 60 ml/min. An yttria stabilized zirconia (YSZ) tube closed at one end was fixed around the measuring head. To control the oxygen partial pressure close to the sample, Pt-electrodes were applied on this tube and the electromotive force (EMF) was measured versus air by an electrometer (Keithley 617 programmable electrometer). From EMF measurements the oxygen partial pressure in the sample chamber was calculated by the Nernst equation. A schematic picture of the dilatometer setup is given in Fig. 7-2.

Fig. 7-2 Schematic setup for the dilatometer measurements.

The expansion of the samples was measured for every gas composition according to the temperature program given in Fig. 6-3. At 600, 700, 800 and 900 °C, the samples were held


for 60 to 140 min to allow equilibration with the ambient atmosphere. The dilatometer exhibited a significant drift during the dwell times. These drifts were found to be a function of different thermal fluxes in the dilatometer caused be the different gas compositions. Therefore correction measurements were carried out to compensate for the expansion of sample holder and pushrods. For these measurements the sample was replaced by a sapphire single crystal with the same length and the thermal expansion was measured according to the same program and with the same gas compositions. The thermal and isothermal expansion was determined as the difference between the expansion and the correction measurement. By this procedure also a possible change of the transducer output due to different gas atmospheres as reported by Chiang et al. [6] could be corrected. The temperature program for the expansion and correction measurements is given in Fig. 7-3. The accuracy and reproducibility of the thermal and isothermal expansion an of the linear thermal expansion coefficient was estimated to ∆L/L0 = ±0.025%.

0 200 400 600 8000

300

600

900

T [°C]

t [min]

Fig. 7-3 Temperature program for the dilatometer measurements for every gas composition.

Additionally, oxidation - reduction cycles were performed at 700, 800 and 900 °C on some specimens to verify whether cyclic contraction and expansion lead to mechanical damage of the samples.



In Fig. 7-4 the differential length change ∆L/L0 of Ce0.8Sm0.2O1.9-x (CSO), Ce0.8Gd0.2O1.9-x (CGO), and Ce0.9Ca0.1O1.9-x (CCO) in air is given as function of the temperature. The samples show an increasing slope of ∆L/L0 vs. temperature. This increase is only small for CSO and CGO, but distinct for CCO. It is attributed to a slight reduction of Ce4+ to Ce3+ [19] leading to an increase in the lattice constant [6], since CCO shows a slightly lower stability than CSO and CGO (chapter 6) and is easier reduced.

200 400 600 8000.0

0.2

0.4

0.6

0.8

1.0

1.2

∆L/L

0 [%

]

CSO CCO CGO

air

T [°C]

Fig. 7-4 Differential length change of CCO, CSO and CGO in air.

L0 denotes the length of the sample at a reference temperature (TRef) of 50 °C. The linear thermal expansion coefficient α is given by

α =−

∆LL

T T0

Ref (7-2)

where ∆l is the length change of the sample. In Eq. (7-2) the temperature has to be given in units of °C, the thermal expansion coefficient is usually expressed in units of 10-6/K. The linear expansion coefficient according to Eq. (7-2) is given in Figure 7-5. For CGO, the smallest linear thermal expansion coefficient was found (12.7 to 12.9.10-6/K) followed by CSO (13 - 13.3.10-6/K) and CCO (13.5 to 14.10-6/K). These results are in accordance with Mogensen et al. [5] who found linear thermal expansion coefficients (50 to 1000 °C) of 12.5.10-6/K for


Ce0.8Gd0.2O1.9-x and 12.8.10-6/K for Ce0.9Ca0.1O1.9-x. Körner et al. [1] found for pure ceria in air values of 12.7 to 13.5.10-6/K at temperatures of 800 to 1100 K.

500 600 700 800 90012

13

14

CSO CGO CCO

α [1

0-6/K

]air

T [°C]

Fig. 7-5 Linear thermal expansion (50 °C to T)of CCO, CSO and CGO in air.

The thermal expansion coefficient of ceria solid solutions in air matches very well with that of high chromium content ferritic alloys, which are potential interconnectors in intermediate temperature SOFCs. The linear thermal expansion coefficient of these alloys is of the order of 12 to 13.10-6/K [20]. In reducing atmosphere, the expansion of the three investigated compositions remains approximately constant up to temperatures of about 700 °C. At temperatures above 700 °C the samples show an increase in thermal expansion at oxygen partial pressures below 10-5 atm. Fig. 7-6a to 7-6c show the isothermal differential length change of CCO, CSO and CGO in relation to the reference length L0 of the specimen in air at 50 °C. Again for Ce0.9Ca0.1O1.9-x, the increase in differential length change is more pronounced than for Ce0.8Sm0.2O1.9-x and Ce0.8Gd0.2O1.9-x. Similar observations were made by Mogensen et al. [5]. However, they found no difference in isothermal expansion between Ce0.9Ca0.1O1.9 and Ce0.8Gd0.2O1.9 at 1000 °C. Chiang et al. [6] reported similar ∆L/L0 values for pure nonstoichiometric ceria under reducing atmospheres. They reported a significant increase in thermal expansion only at values above 10-20 atm at 800 °C and 10-17 atm at 900 °C.


Measurements of the linear thermal expansion coefficient in humidified Ar-H2 atmosphere (50 °C to measurement temperature, see Eq. (7-1)) are shown in Fig. 7-7. These measurements reveal an increase in the linear thermal expansion coefficient from 12.7 to 13.3.10-6/K at 700 °C to 15.5 to 18.7.10-6/K at 870 °C. It was noted, that after cooling to room temperature under reducing atmospheres the samples were black. This blackening is an indication of quasi-free electrons due to oxygen nonstoichiometry and was reversible. Even at ambient temperatures in air the samples turned partly orange again after several weeks. Up to a temperature of 700 °C the isothermal expansion of ceria solid solutions at lower oxygen partial pressures compared to air or pure oxygen is within the measurement uncertainty and can therefore be neglected. From this point of view there is no objection against ceria solid solutions as electrolytes in intermediate temperature SOFC at operation temperatures up to 700 °C.

0 -5 -10 -15 -20 -250.6

0.8

1.0

1.2

1.4

1.6

1.8

∆L/L

0 [%

]

CCO

600 °C

700 °C

800 °C

900°C

log(p(O2)/atm)

Fig. 7-6a Differential length change of Ce0.9Ca0.1O1.9-x as a function of temperature and p(O2). L0

is the length of the specimen at 50 °C in air.


0 -5 -10 -15 -20 -250.6

0.8

1.0

1.2

1.4

∆L/L

0 [%

]

CSO

600 °C

700 °C

800 °C

900°C

log(p(O2)/atm)

Fig. 7-6b Differential length change of Ce0.8Sm0.2O1.9-x as a function of temperature and p(O2).

L0 is the length of the specimen at 50 °C in air.

0 -5 -10 -15 -20 -250.60.70.80.91.01.11.21.31.4

∆L/L

0 [%

]

CGO

600 °C

700 °C

800 °C

900°C

log(p(O2)/atm)

Fig. 7-6c Differential length change of Ce0.8Gd0.2O1.9-x as a function of temperature and p(O2).

L0 is the length of the specimen at 50 °C in air.


500 600 700 800 900

10

12

14

16

18

20

CSO CGO CCO

9.10-25 10-222.5.10-20

5.10-18

(90 % Ar - 10 % H2)- 3 % H2Oα

[10-6

/K]

T[°C]

Fig. 7-7 Linear thermal expansion (50 °C to T) of CCO, CSO and CGO in Ar-H2-H2O. Numbers

indicate the oxygen partial pressures at the different temperatures.

Under SOFC operating conditions the oxygen partial pressure at the electrode/ceria interface is depending on the electrode overpotential [21]. By changing the load current of fuel cells, the p(O2) at this interface can change quite fast. Changes in thermal expansion due to different oxygen partial pressures at these interfaces can lead to cracks in the ceria membranes and to possible delamination of the electrodes. To investigate the stability of the sample against such reduction/oxidation cycles, a CSO specimen was subjected to several cycles at 700 and 800 °C. The sample was stable and no mechanical damage was observed after cooling to room temperature. A typical example of such a reduction/oxidation cycle is shown in Fig. 7-8. On the vertical axis the absolute measured length change of the sample is shown, without correcting expansion of the sapphire reference and the expansion of pushrods and sample holder. It was noted, that the reduction reaction of the sample was much slower than the oxidation. Since the mobility of oxygen vacancies in ceria is very fast, this difference was attributed to a surface effect. Obviously, the adsorption of oxygen at high oxygen partial pressures on the ceria surface is much faster than the desorption at low p(O2).


0 200 400 600 800 1000

30

32

34

36

38

40

42

700 °C

oxidation

reduction

air air90 % Ar - 10 % H2 + 3 % H2O

t [min]

∆l/µ

m (n

ot c

orre

cted

)

Fig. 7-8 Typical oxidation/reduction cycle of a CSO sample at 700 °C. The length change ∆L

was not corrected for thermal expansion of reference, sample holder and pushrods.


Dilatometer measurements of ceria solid solutions with samaria, gadolinia and calcia revealed a linear thermal expansion coefficient of 12.9.10-6/K (50 to 900 °C) for Ce0.8Gd0.2O1.9 (CGO), 13.3.10-6/K for Ce0.8Sm0.2O1.9 (CSO) and 14.10-6/K for Ce0.9Ca0.1O1.9 (CCO) in air. These values increased to 15.5.10-6/K for CGO, 15.7.10-6/K for CSO and up to 18.7.10-6/K for CCO. In the temperature range up to 700 °C and the oxygen partial pressure range from 1 to 10-25 atm the isothermal relative length change was not detectable within experimental error. It is therefore be concluded, that ceria-based materials are suitable electrolytes for intermediate temperature SOFC from this point of view. Oxidation/reduction cycles at 700 and 800 °C did not affect the mechanical integrity of the samples. It was noted during these measurements, that the oxygen uptake of the samples was on order of magnitude faster than the reduction. Since ceria exhibits a very high oxygen vacancy mobility this effect can not be caused by bulk properties and was therefore attributed to surface effects such as adsorption/desorption of oxygen.


5. References

1. R. Körner, M. Ricken, J. Nölting, and I. Riess, J. Solid State Chem., 78, 136 (1989).

2. P. Gode, Journal Less-common Metals, 160, 53 (1990).

3. M.A. Panhans and R.N. Blumenthal, Solid State Ionics, 60, 279 (1993).

4. G. Mogensen, M. Mogensen, Thermochim. Acta, 214, 47 (1993).


2122 (1994).

6. H.-W. Chiang, R.N. Blumenthal and R.A. Fournelle, Solid State Ionics, 66, 85 (1993).

7. R.G. Schwab, R.A. Steiner, G. Mages and H.-J. Beie, Thin Solid Films, 207, 288 (1992).

8. P.V. Hendriksen, J.D. Carter and M. Mogensen, in Solid Oxide Fuel Cells IV, M. Dokiya,

O. Yamamoto, H. Tagawa and S.C. Singhal, Editors, PV 95-1, p. 934, The Electrochem. Soc., Pennington

(NJ) (1995).

9. T.R. Armstrong, J.W. Stephenson, L.R. Pederson, and P.E. Raney, in Solid Oxide Fuel

Cells IV, M. Dokiya, O. Yamamoto, H. Tagawa and S.C. Singhal, Editors, PV 95-1, p. 944, The Electrochem.

Soc., Pennington (NJ) (1995).

10. H.-W. Chiang, A High Temperature Lattice Parameter and Dilatometric Study of

Nonstoichiometric Ceria and Calcia-doped ceria, Marquette University, Milwaukee (WI), USA (1992).

11. R.D. Shannon, Acta Cryst. A, 32, 751 (1976).

12. H.-H. Möbius, Z. Chem. 4, 81 (1964).

13. T. Kudo and Y. Obayashi, J. Electrochem. Soc., 123, 415 (1976).


15. M. Gödickemeier, K. Sasaki and L.J. Gauckler, in Solid Oxide Fuel Cells IV, M. Dokiya, O.

Yamamoto, H. Tagawa and S.C. Singhal, Editors, PV 95-1, p. 1072, The Electrochem. Soc. Proceedings

Series, Pennington, NJ (1995).

16. F.P.F. van Berkel, G.M. Christie, F.H. van Heuveln and J.P.P. Huijsmans, in Ref. 15,

p. 1062

17. C. Milliken, S. Elangovan, and A.C. Khandkar, in Ref. 15, p. 1049.

18. B.C.H. Steele, K. Zheng, R.A. Rudkin, N. Kiratzis and M. Christie, in Ref. 15, p. 1028.

19. Ch. Ftikos, M. Nauer and B.C.H. Steele, J. Europ. Ceram. Soc., 12, 267 (1993).

20. K. Honegger, Final Report, PPM, Project 1.C.2, Switzerland (1995).

21. I. Riess, M. Gödickemeier, and L.J. Gauckler, Solid State Ionics, submitted.

8. Characterization of Solid Oxide Fuel Cells Based on Mixed Ionic Electronic Conductors

Abstract

The relation between cell voltage (VCell), applied chemical potential difference (∆µ(O2)) and cell current (It) for solid oxide fuel cells (SOFC) based on mixed ionic electronic conductors is derived by considering also the effect of electrode impedance. Four-probe measurements, combined with current interruption analysis are considered to yield the relation between ionic current (Ii) and overpotential (η). These theoretical relations are used to analyze experiments on fuel cells with Ce0.8Sm0.2O1.9 and Ce0.8Gd0.2O1.9 electrolytes with La0.84Sr0.16CoO3 or Pt as cathode and Ni/Ce0.9Ca0.1O1.9 or Pt as anode. The electrode overpotentials of these cells determined by current interruption measurements are discussed assuming different models including impeded mass transport in the gas phase for molecular and monoatomic oxygen and Butler-Volmer type charge transfer overpotential.

1. Introduction

The I-V relations that characterize solid oxide fuel cells (SOFC) are determined by both the impedance of the solid electrolyte (SE) or mixed ionic electronic conductor (MIEC) and the electrodes of the cell. The I-V relations for solid oxide fuel cell based on MIECs can be derived from the general transport equations, considering defect concentrations in relation to the chemical potential of oxygen. In the literature [1-5], these transport equations are reported for cells with reversible electrodes only. In the present chapter it will be shown, that electrode reactions can be included. The potential drops on the electrodes have been considered before [6-11] but not for SOFCs based on MIECs under fuel cell operating conditions. It is clear, that as in classical electrochemistry one can identify three electrode polarization processes that lead to potential drops: a) mass transport b) charge transfer c) chemical reaction [12] In section 2 of this chapter, the equations that describe the I-V relations for SOFCs based on MIECs are derived. In section 3 a four probe fuel cell arrangement is analyzed, and it is shown how to derive the chemical potential drops from the measurements. Section 4 presents considerations of the optimum thickness for MIEC, which differs from the case of purely ionic

MIEC 88

conducting solid electrolytes. In section 5, experimental data on SOFCs based on the MIECs Gd2O3 and Sm2O3 doped ceria is presented and analyzed using the theory. Measurements with different types of electrodes, Pt, La0.84Sr0.16CoO3, Ni/Ce0.9Ca0.1O1.9-x and different oxygen partial pressures at the cathode are presented. For these experimental results, oxygen chemical potential profiles across the electrolyte are calculated in section 6. A summary is given in section 7. At the end of this chapter a list of the abbreviations and symbols used throughout the electrochemical part of this dissertation is given.

2. I-V relations

2.1 General equations

For the following derivations we consider doped CeO2, a MIEC in which electronic conductivity is introduced by deviation from stoichiometry at low oxygen partial pressures. This electrolyte is a predominantly O= - ionic conductor in oxidizing atmospheres due to oxygen vacancies introduced by doping with three-valent rare earth cations. However, by low oxygen partial pressures it is reduced and shows significant electronic conductivity (n-type). The introduction of extrinsic charge carriers can be described as

Me O Me O VCeOCe O

xO2 3

2 2 3 → + + ⋅⋅' (8-1)

the extrinsic charge carrier concentration is given by the dopant concentration

[ ] [ ]V MeO Ce⋅⋅ = 1

2'

(8-2)

and the ionic conductivity (σi) is described by

σ νi iqN= 2 (8-3)

where q is the elementary charge, νi is the mobility of oxygen ions and N their concentration. At low oxygen partial pressures, the electrolyte is partially reduced and develops n-type electronic conductivity according to:

O V e OOx

O↔ + +⋅⋅ 2 12 2 (8-4)

With the law of mass action follows:

n K N p(Oe= − −12

142) (8-5)

where n denotes the concentration of electrons and Ke is a constant. Since in the temperature range of 600 to 800 °C, the concentration of additional oxygen vacancies by partial reduction is at least one order of magnitude smaller than the one introduced by doping, it is justified to assume, that N remains constant (chapter 5) in this regime.

89 MIEC

Therefore:

N N Nintr extr extr+ ≈ (8-6)

The dependence of the electronic conductivity (σe = nqνe) on oxygen partial pressure therefore is given by

σ e p(O∝ −2

14) (8-7)

Although n<<N, the conductivity is found to be mainly electronic in reducing atmospheres since νe>>νi (chapter 5). The transport equations for the considered mixed ionic and n-type electronic conductor are given by

jq

Oii= − ∇ =σ

µ2

~( ) (8-8)

and

jqee

e= ∇σ

µ~

(8-9)

where j is the current density and ∇ =~( )µ O and ∇~µ e are the gradients in electrochemical

potential of oxygen ions and electrons, respectively. These electrochemical potentials are defined as the sum of the chemical potential and the electrical potential times the effective charge of the charge carrier

~( ) ( )µ µ ϕO O q= == − 2 (8-10)

and ~µ µ ϕe e q= − (8-11)

where µ(O=) and µe are the chemical potentials of oxygen ions and electrons and ϕ is the electrical potential within the electrolyte.

µ µ( ) ( , ) ln( )O O T kT N= = °= + (8-12)

µ µe e T kT n= +°, ln( ) (8-13)

Based on Eq. (8-8) to (8-11), the transport equations are formulated as

j kT q Nxi i

Nx i= − +ν ν

∂ϕ∂

∂∂ 2

(8-14)

j kTnx

q nxe e e= −ν

∂∂

ν∂ϕ∂ (8-15)

Here, k is the Boltzmann constant . Using Eq. (8-3) and (8-6), the first term of Eq. (14) vanishes, since ∇ ==µ( )O 0 for N ≈ constant. Eq (8-14) can therefore be rewritten as:

MIEC 90

jxi i= σ

∂ϕ∂ (8-16)

Since for N ≈ constant ∇ ==µ( )O 0 , and ∂∂jxi = 0 , it follows from Eq. (8-16) that

∂ ϕ∂

2

2 0x

= . The electric field within the electrolyte is therefore given as,

∂ϕ∂x

V MC V MCL

th=−( ) ( )

(8-17)

where V(MC) is the voltage drop across the mixed conductor and Vth(MC) is the Nernst potential across the MIEC, given by

V MCO O

qthL

( )( ) ( )

=−µ µ2 2

0

4 (8-18)

µ(O2)L and µ(O2)0 denote the oxygen chemical potential just inside the MIEC, near the surface but beyond any possible space charge at the interface MIEC/electrodes. The considered linear configuration is shown in Fig. 8-1. The Nernst potential across the MIEC can differ from the applied Nernst potential given by the atmospheres at the cathode and anode side, respectively.

V MC O Oqth app

high low

, ( ) ( ) ( )= −µ µ2 24 (8-19)

For an ideal gas, chemical potential and oxygen partial pressure are related by:

µ µ( ) ( , ) , ln( )O O T T kT p(O2 2 2= +° (8-20)

The I-V relations for the MIEC proper (i.e. once the oxygen ion is within the solid) can be expressed by

IV MC

Ree

= −( )

(8-21)

and

IV MC V MC

Rith

i=

−( ) ( )

(8-22)

91 MIEC

Fig. 8-1 Schematic representation of the considered MIEC. p(O2)high at the air side and p(O2)low

at the fuel side. p(O2)A and p(O2)C represent the oxygen partial pressures just after the

drop in oxygen chemical potential at the anode and cathode side, respectively. The

charge transfer region is shown schematically. It can be located at the MIEC/electrode

interface or entirely in the electrode.

The cell voltage under operating conditions as measured on the electrodes is denoted as VCell. This cell voltage is not necessarily equal to V(MC), due to electrode (ohmic) resistance. Ri is a constant, therefore, Eq. (8-22) can be used immediately. Since Re is not a constant, Eq. (8-21) is not sufficient for the purpose of the analysis here. A more explicit expression is needed which relates Ie , V(MC) and the local electronic conductivities, σe

0 and σeL, beyond the

surfaces of the MIEC at x=0 and x=L, respectively. In order to derive this relation, we have to introduce some boundary conditions. First, it is assumed that the electronic current is dominated by only one type of electronic charge carriers (by electrons) and that they are generated by deviation of the CeO2 from stoichiometry. An explicit expression for the electronic current is derived by combining Eqs. (8-10 - 12), (8-15), (8-17), (8-18), and (8-20). With respect to the electronic defect distribution (Eq. (8-5) and (8-7)) and Eqs. (8-12) and (8-18), the ratio n (L)/n (0) is expressed as

n Ln

e q V MCth( )( )

( )

0= − β

(8-23)

MIEC 92

Where β=1/kT. With the boundary conditions T, ∂ϕ ∂/ x (Eq. (8-17)), n(L) and n(0), the

transport equation for electrons (Eq. 8-15) is expressed as

( )I SV MC V MC

Le

ee e

L thqV MC

q V MC V MCth= −

−⋅

−

− − −σβ

β( ) ( ) ( )

( ) ( )1

1 (8-24)

2.2 Overpotential

Electrode processes are generally coupled with a voltage drop due to several processes, such as charge transfer, diffusion processes and chemical reactions. These processes are lowering the electromotive force Vth(MC), which is effective on the MIEC. Therefore Vth(MC) < Vth,app. We begin now to evaluate the differences δVth = Vth,app - Vth(MC) and δV = VCell - V(MC). These differences can first be separated to those at the cathode and those at the anode:

δ δ δV V Vth th C th A= +, , (8-25)

δ δ δV V VC A= + (8-26)

The voltage drop δV is only due to the resistance of the two electrodes, resistances RA and RC

, electrode processes do not contribute to δV. Therefore, δV arises from the resistance of the electrodes to the cell current It and if It = O , V(MC) = VCell:

V V MC I R RCell t A C= − +( ) ( ) (8-27)

where

I I It e i= + (8-28)

is the total current through the cell. For the drop in Vth we assume that the three possible processes at each electrode operate in series and therefore, that their contributions to δVth can be added.

δ δ δ δV V V Vth C th CD

th CCT

th CR

, , , ,= + + (8-29)

δ δ δ δV V V Vth A th AD

th ACT

th AR

, , , ,= + + (8-30)

where δVth CD

, and δVth AD

, ,δVth CCT

, and δVth ACT

, , δVth CR, and δVth A

R, , are the contributions

due to mass diffusion , charge transfer and reaction at both electrodes respectively. To summarize: one uses in principle Eqs. (8-18) and (8-24) now with Vth(MC) = Vth,app - δVth and V(MC) = VCell-δV and one can therefore relate the partial currents Ie and Ii as well as the total measured current It = Ie + Ii to the measurable

93 MIEC

parameters Vth,app and VCell. It is however important to find a way to measure also δVth,C and δVth,A. This can be done using a four probe method as analyzed in section 3.

2.3 Overpotential due to mass diffusion

In order to be able to calculate the SOFC I-V relations one has to assume specific polarization processes, i.e. specific relations between the potential losses and the ionic current. For mass diffusion polarization losses we consider the following processes: A gradient in p(O2) might occur in the porous electrodes due to slow gas diffusion. The p(O2) at the end of the diffusion path in the gas phase (near the electrode/MIEC interface) for the cathode, is

p(O p(O C IC highi2 2 1) )= − (8-31)

where C1 is a positive constant decreasing inversely with the diffusion rate in the gas phase. In view of Eqs. (8-19), (8-20) and (8-31)

δVkT

qp(O

p(O C Ith CD

high

highi

, ln)

)=

−

42

2 1 (8-32)

For the anode one has to notice that the diffusing gases are the fuel (e.g. H2) and the reaction products (e.g. H2O). Assuming local equilibrium for the reaction H2 + 1/2 O2 -> H2O at the electrode/MIEC interface, it follows

p(O Kp(H Op(H2 1

2

2

12)

))

= (8-33)

Assuming that H2 diffuses fast and diffusion limitations occur by H2O only:

p(H O p(H O A IA exti2 2 1) )= + (8-34)

where p(H2O)ext is the value in the gas phase outside the anode and A1 is a positive constant decreasing inversely with the diffusion of H2O in the gas phase in the pores. Then

δVkT

qp(H O A I

p(H Oth AD

exti

ext, ln)

)=

+

22 1

2 (8-35)

If these were the only polarization losses we obtain from Eq. (8-19), (8-20), (8-32)-(8-34):

( ) ( )V MCkT

qp(O C I

K Tp(H

p(H O AIthhigh

iext

i( ) ln ) ln( )

)ln )= − −

− +

4

2 22 11

22

(8-36)

In case the diffusion is limited by diffusion of monoatomic oxygen to the reaction sites, the oxygen partial pressure at the interface cathode/MIEC is given by

MIEC 94

( ) ( )p(O p(O C IC highi2 2 2

12

12) )= −

(8-37)

and

( )∂V

kTq

p(O

p(O C Ith CD

high

highi

, ln)

)

=

−

42

2 2

212

(8-38)

Analogous diffusion processes also can be formulated for diffusion limitation due to monoatomic or molecular diffusion at the anode with the parameters A2 for monoatomic and A3 for molecular diffusion of hydrogen.

2.4 Charge transfer processes

Charge transfer processes may take place inside the MIEC near surface layers, inside the electrode near surface layers or at the MIEC/electrode interface. In all cases a Butler-Volmer type I-V polarization loss is anticipated. We use the third possibility to demonstrate this. Let us assume that an O2 molecule decomposes into O+O which by contact with the electrode turns finally into O= ions at the interface. An O= ion upon entering the MIEC has to overcome a potential barrier associated with a space charge near the MIEC boundary. Because of the high concentration of (ionic) charge carriers in the MIEC, the space charge region is rather narrow of the order of a few atomic layers (as is also the case in liquid electrolytes). We assume that the change of the electrochemical potential of the electrons across the space charge regions (L-C and A-0 in Fig. 1) can be neglected (∆~µ e = 0 there). This may be justified in view of the high bulk resistance, Re, so that most of the drop ∆~µe (A-C) appears across the MIEC, i.e. between 0-L in Fig. 1. Then the change in the electrochemical potential of the oxygen ions across the space charge regions, ∆~( )µ O= (A-0) and ∆~( )µ O= (L-C) reduce to the corresponding changes of ∆µ( )O2 /2 and the contribution to the overpotential is attributed to δVth. If ∆~µe at A-0 and L-C cannot be neglected one can show that similar ionic current -overpotential (Ii-η) relations exist except that they are not attributed solely to δVth. The electronic current-overpotential (Ie-η) relations are slightly modified but the difference is assumed to be negligible. We therefore assume from here on that ∆~µe (A-0) and ∆~µe (L-C) vanish. The O= ions on the other hand are classical particles and must cross over the potential barrier. They are driven by ∇ =~( )µ O where ~( )µ O= is the electrochemical potential of the ions. µ(O=) is their chemical potential, uniform in the space charge region). The applied driving force is expected to modify the electric potential barrier. This is true, if ∆~( )µ O= does not change the equilibrium concentration of the ions and is analogous to the case described by the Butler-Volmer equation. Then the I-V relation can be written as [12]

95 MIEC

I SJ e eiO O= −

= =− −0

1αβ∆µ α β∆µ~( ) ( ) ~( )

(8-39)

where S is the electrode area J0 the exchange current per unit area and α the transfer coefficient. For ½ O2 + 2e- ⇔ 2 O=

∆µ ∆ ∆( ) ~( ) ~O O e2 2 4= −=µ µ (8-40)

Inserting Eq. (40) with ∆~µe = 0 (for the space charge region) into Eq. (8-39) yields

( )I I e eiO O= − − −

02 1 22 2αβ∆µ α β∆µ( ) / ( ) ( ) /

(8-41)

where I0 = SJ0. For the cathode the relevant ∆µ(O2) equals 4qδVth,CCT. Under fuel cell

conditions the ionic current flows only in one direction (positive x direction in Fig. 8-1). Eq. (8-41) can then be approximated by

I I q Vi C C th CCT= 2 20, ,sinh( )α βδ (8-42)

This approximation is valid for large overpotentials (positive) and for zero overpotential. Therefore it is a good approximation for positive values of δVth,C

CT as can also be seen in Fig. 8-2. Similarly for the anode,

I I q Vi A A th ACT= 2 20, ,sinh( )α βδ (8-43)

With this approximation one can invert the I-V relations and write

δα

VkTq

IIth

CT

C

i

C=

2 2 0arsinh

, (8-44)

and similar for the anode side.

MIEC 96

-0.06 0.00 0.06 0.12 0.18-3

0

3

6 Butler-Volmer-Equation Tafel-Equation sinh-Approximation

α=0.4

δVth,CCT or δVth,A

CT [V]

I i / I

0C o

r I i

/ I 0

A

Fig. 8-2 Comparison between calculated approximations for the Butler-Volmer equation with a

charge transfer coefficient α=0.4.

2.5 Transport equations considering electrode processes

The O2 chemical potential loss due to reactions at the electrodes will not be discussed in details in this chapter, this will be given in chapters 9 (cathode) and 10 (anode). For the interpretation of the data presented in this chapter, it is sufficient to realize that electrode overpotentials can be understood as a drop in oxygen chemical potential. This overpotential consists of a gas polarization (concentration polarization) and a charge transfer term. Since reaction overpotential also can be understood as a concentration overpotential, we neglect δVth,C

R and δVth,AR for the interpretation used here. In summary, taking both O2 gas

polarization and charge transfer polarization into consideration Vth(MC) is (from Eqs. (8-20), (8-21), (8-37), (8-42)-(8-44))

( ) ( )( )

V MCkT

qp(O C I K T p(H

p(H O A IkTq

II

kTq

II

thhigh

i

exti

C

i

C A

i

A

( ) ln ) ln ( ) / )

ln ) arsinh arsinh

= − −

− + −

−

42

22 2 2 2

2 1 1 2

2 10 0α α

(8-45)

where K1 is the constant for the equilibrium between H2, H2O and O2. In addition for Vth(MC) which is given e.g. by Eq. (8-27) we have always to use also Eq. (8-22)

97 MIEC

IV MC V MC

Rith

i=

−( ) ( )

(8-22)

Eq (8-9)

V V MC I R RCell t A C= − +( ) ( ) (8-27)

and Eq. (8-24)

( )I SV MC V MC

Le

ee e

L thqV MC

q V MC V MCth= −

−⋅

−

− − −σβ

β( ) ( ) ( )

( ) ( )1

1 (8-24)

where σeL has to be related to the value under p(O2)high by

σ σ σ β δe

Le

highL

high ehigh q Vp(O

p(Op(O

p(O e th C=

=

−

( ) ))

)( ) ) ,

22

22

14

(8-46)

in which δVth,C = δVth,CD + δVth,C

CT (Eq. (11) with δVth,CR neglected).

For cathode polarization determined by O2 gas diffusion and charge transfer,

σ σ αe

Le

highhigh

highi

IIp(O

p(Op(O C I

e C

i

C=−

( ) ))

)

arsinh

22

2 1

12q 2 0

(8-47)

3. Interpretation of four probe measurements on SOFCs

The four point arrangement is shown in Fig. 8-3. It is assumed that a) the distance ∆y between the main electrodes and the reference electrodes is large compared to the width L of the MIEC itself

∆y>>L (8-48)

and b) that the area of the reference electrodes is much smaller than that of the working electrodes. One reference electrode (Ref C) is exposed to the same (ambient) gas as the cathode. The other reference electrode (Ref A) is exposed to the same fuel atmosphere as the anode. The open circuit voltage VOC on the reference electrodes as well as on the working electrodes is smaller than ti Vth,app where ti is the average ionic transference number for the MIEC

proper given by [13]

tkT

qV MCith

eL

i

e i= +

++

1 0( )

lnσ σσ σ (8-49)

The reason is that for MIECs even under open circuit Ii ≠ 0 only Ii + Ie = It = 0 thus Ii = -Ie. Due to electrode polarizations Vth(MC) < Vth,app and the effective driving force on the MIEC

MIEC 98

for the ionic current is smaller (there is also a small modification in ti through Vth(MC), σeL

and σe0).

The open circuit voltage on the reference electrode and working electrode need not coincide as the electrode polarization may be different. However, if the type of material and microstructure are similar one expects only small differences on the open circuit voltage, otherwise large differences may occur (for instance for similar La0.84Sr0.16CoO3 electrodes on Sm2O3 doped CeO2 we were able to reduce the difference to 10-30 mV at 800 °C). Let us consider the cathode side. The measured voltage ηC between the cathode and Ref C given by the corresponding difference in ~µ e

− = −q ext CC e eη µ µ~ (Ref C) ~ ( , ) (8-50)

The reaction 1/2O2 + 2e ⇔ O= can be considered to occur at the interface where we assume it to be in equilibrium. There

∆~ ~ (Ref C) ~ ( ) ~ (Ref C) ~ ( )µ µ µ µ µe e e e e t CC ext C qI R= − = − + (8-51)

Using Eq. (8-40) for ∆~µe

( ) ( )− = − − + − −= =q O O O O qI RCC C

t Cη µ µ µ µ14 2 2

12( ) ( ) ~( ) ~( )Ref C Ref C

(8-52)

The first term can be rewritten as

( )µ µ µ µ µ µ

δ

( ) ( ) ( ) ( ) ( ) ( )Ref C Ref C

,

O O O O O O

qC q V

C high C high

th CD

2 2 2 2 2 2

34 4

− = − − −

= − + (8-53)

where ( )− −µ µ( ) ( )O O high2 2

Ref Cis replaced by a constant C3 (C3 >0) because it can be

considered quite independent of the current through the remote working electrodes and to be determined solely by the open circuit conditions on the reference electrodes. The second term can be written as,

( )~( ) ~( ) ~( ) ~( ) ~( ) ~( )Ref Cµ µ µ µ µ µO O O O O OC Ref C L C L= = = = = =− = − − − (8-54)

~( ) ~( )µ µ γO O qI RRef C LC i i

= =− = 2 (8-55) where γCIiRi is a fraction (0<γC<1) of the IiRi drop between the two working electrodes (γCIiRi is affected also by the reference electrodes, in particular for small currents).

99 MIEC

Fig. 8-3 Schematics of the measurement setup for MIEC characterization under fuel cell

operating conditions. ηC is the voltage measured between cathode and the reference

electrode at the air side (Ref C), ηA is the voltage between anode and Ref A.

~( ) ~( )µ µO OC L= =− is the difference in ~( )µ O= required to drive the ionic current through the

space charge region.

~( ) ~( ) ,µ µ δO O q VC Lth CCT= =− = 2 (8-56)

Combining Eqs. (8-52 - 8-56) yields

η δ γC th C C i i t CV C I R I R= − + +, 3 (8-57)

where we have used δVth,C = δVth,CD + δVth,C

CT. The measurement is being done after current interrupt. The value of ηC immediately after current interrupt is denoted as ηC

+. We use a fast oscilloscope (LeCroy 9450 A) and follow the relaxation over time scale of µs. We shall show that within this time not only the gas polarization but also the charge transfer polarization cannot relax. Therefore

η δ γC th C C i iV C I R+ − += − + ⋅ ⋅, 3 (8-58)

where δVth,C- is the polarization loss just before current interrupt and Ii

+ is determined by the open circuit condition on both the reference electrode and the working electrode. It may

MIEC 100

change a little with Ii- as the net driving force at the working electrode is affected by the

polarization there which depends on Ii-. However, this is a small correction and we

approximate γCIi+Ri by a constant. Then Eq. (8-58) becomes

η δC th CV C+ −= −, 4 (8-59)

In analogy for the anode

η δA th AV A+ −= −, 4 (8-60)

where A4 is a constant. δVth,C and δVth,A were discussed in section 2. To obtain δVth,C- and

δVth,A- one has to consider these expressions for Ii = Ii

-. The cell voltage after interrupt will relax and increase towards a steady state open circuit value. Before current interrupt Eqs. (8-18) and (8-27) yield

V V V V I R I R RCell th app th C th A i i t A C= − − − − +, , , ( )δ δ (8-61)

After interrupt (within about 1 µs)

V V V V I RCell th app th C th A t i+ − − += − − −, , , ( )δ δ

(8-62)

In view of Eqs. (59) and (60),

V V DCell th app C A+ + +≈ − − −, η η

(8-63)

where D is a constant (D = A4+C4). The effective overpotentials and the characteristic constants (A1...4, C1...4, D and S) can be obtained by from the measured It and VCell using Eqs. (8-24), (8-28), (8-46) and (8-71). This will be done in section 5.

4. Limitation on decreasing the MIEC resistance

One expects intuitively that reducing the MIEC resistance need improve the SOFC performance. However this is not true when the electrode impedance stays constant. Then, decreasing the MIEC resistance too much will decrease the power output and efficiency. Thus reducing the MIEC thickness should be considered carefully. A reduction in thickness decreases both Re and Ri within the MIEC. For reversible electrodes VOC for example does not change [13]

V VkTqOC th

eL

i

e i= +

++

ln

σ σσ σ0

(8-64)

101 MIEC

However, when the electrodes are not reversible the driving force Vth(MC) on the MIEC decreases as Re and Ri decrease. Let us consider the open circuit voltage on the MIEC

V MC V MCOC th( ) ( )≤ (8-65)

Let us assume only polarization in the gas phase at the cathode

( ) ( ) V MCkT

qp(O C I p(Oth

highi

low( ) ln ) ln )= − −4 2 1 2

(8-66)

Under open circuit condition Ii = -Ie. Using Eq. (1) for Ie

( )04 2

12≤ ≤ −

−

V MCkT

qp(O

C V MCR

p(OOChigh OC

e

low( ) ln )( )

ln ) (8-67)

When the thickness (d) of the MIEC is reduced d→0 then Ri→0 and Re→0. This results in VOC→0. We arrive at the same conclusion when the electrode polarization is due to charge transfer limitation at the cathode

V MC VkTq

IIth th app

C

i

C( ) arsinh,

,= −

2 2 0α (8-68)

and

02 2 0

≤ ≤ −

V MC V

kTq

VI ROC th app

C

OC

C e( ) arsinh,

,α (8-69)

As Re goes to zero, VOC→0. Then VCell and power (VCell.It) output vanish as well, since

0 ≤ ≤V VCell OC .

As a consequence, there exists a optimum electrolyte thickness for every set of temperature, applied oxygen partial pressures and (non-ideal) electrode characteristics. This will be discussed in chapter 11.

5. Comparison with experiments

5.1 General

The theory developed in the previous sections will now be applied to measurements carried out on SOFC. We consider the following measurements: a) Cell voltage, VCell, as a function of cell current, It (Fig. 8-4) b) Overpotential just after current interrupt both at the cathode, ηC

+, and the anode, ηA+, as a

function of cell current, It (Figs. 8-5a and 8-5b).

MIEC 102

c) Cell open circuit voltage as a function of p(O2)high and therefore as a function of different MIEC to electrode impedance ratios (Fig. 8-11) as well as cell open circuit voltage VOC as a function of the thickness of the MIEC (Fig. 8-9). The fitting of VCell-It relations to experimental data can be done in two ways. First, one can assume certain electrode polarization mechanisms, derive σe(p(O2)high) of the MIEC from σe vs. p(O2) measurements, use the size of the MIEC, and the applied p(O2)high, p(O2)low and solve the coupled Equations (8-18), (8-24), (8-27), (8-46), and (8-45). Fitting the solution to the data should yield the parameters: RA, RC, Ri, αC, αA, I0,A, I0,C and possibly C1 and A1 or C2 and A2. Ri, the resistance to ionic current needs not equal L/(Sσi) due to constriction resistance frequently observed [6, 9, 14]. This procedure turn out to be complicated and in view of the many parameters to be fitted, possibly also inaccurate. An alternative fitting procedure that we have adopted uses the experimental values of ηC

+ and ηA

+ to determine δVth,C and δVth,A up to a constant (see Eqs. (8-59) and (8-60)). We then determine Vth(MC) experimentally up to a constant (C4 + A4)

V MC V C Ath th app C A( ) ( ),= − − − ++ +η η 4 4 (8-70)

where Vth,app is also known ( kTq

p O p Ohigh low4 2 2ln ( ) ln ( )− ). Eq. (8-18) can be expressed in

terms of measured parameters and the unknowns RA, RC, (C4 + A4) and S in Ri = L/(Sσi):

IL

SV C A V I R Ri

ith app C A Cell t A Cσ

η η= − − − + − + ++ +, ( ) (4 4

(8-71)

For the MIEC Ri ~ 1 Ω. From the electronic conductivity of LSM and LSC [15, 16] electrodes one can estimate for the usual electrode geometries RC, RA~10-3 Ω and even less for metallic electrodes. Since I It i≤ (under fuel cell operating conditions) the term

It(RA + RC) can be neglected in Eq. (8-71) as well as in Eq. (8-27) i.e. VCell = V(MC). The electronic current can also be expressed in terms of experimentally known parameters and the constant (C4+A4). One uses Eq. (8-70), (8-27) (with RC = RA ≈ 0) and Eq. (8-46) with δVth,C = ηC

+ + C4 in Eq. (8-6). This leaves the Ie equation with the unknowns S, C4 and (C4 + A4). Fitting the theory (Eqs. (8-24) and (8-71)) to the measured VCell - It relation yields those three parameters. Using S, C4, A4 one can determine Ii using Eq. (8-71). Now the ηC

+ - Ii (and ηA+ - Ii)

relations as measured can be examined, electrode impedance models can be assumed. The corresponding theory of δVth,C (and δVth,A) can be fitted to the measured values (Fig. 8-8a, b).

5.2 Cell voltage - It and overpotential -Ii relations

103 MIEC

Here, measurements on Ce0.8Sm0.2O1.9 with La0.84Sr0.16CoO3 as cathode and Ni/Ce0.9Ca0.1O1.9-x cermet as anode are presented as an example. The measured It - VCell relations at 700 °C are presented in Fig. 8-4. The corresponding ηC

+ and ηA+ measurements

are presented in Figs. 8-5a and 8-5b, Fig. 8-6 shows the time dependence of ηC in the current interrupt measurements (It = 0.401 A before current interrupt) and the definition of ηC

+. The drop is interpreted as due to the RiIi voltage change. This is valid, if the IiRi changes are faster than ~1 µs at 700 °C and if the electrode processes including gas polarization and charge transfer are slower than 1 µs [14]. This assumption is justified, since the relaxation frequency for the ionic and electronic charge carriers at operating temperatures is in the MHz range (chapter 5).

0.0 0.2 0.4 0.6 0.8 1.00.0

0.2

0.4

0.6

0.8

1.0

La0.84Sr0.16CoO3 / Ce0.8Sm0.2O1.9 / Ni-Ce0.9Ca0.1O1.9-x

700 °C

VC

ell [

V]

It [A]

Fig. 8-4 I-V relation of a fuel cell air / LSC / MIEC / Ni-Cermet / H2O-H2 at 700 °C. The solid line

represents a fit to Eq. 8-53 with ηC+ and ηA

+ obtained by current interruption

measurements, Vth,app = 1.029 V and RC and RA ≈ 0.

It is further assumed, that Ii

+ can be considered as constant, roughly equal to the open circuit steady state value. This is certainly true for small It where the electrode polarization is relatively small. To check our interpretation and approximation of Ii

+ ~ const., we present in Fig. 8-7 the measured voltage drop ηC - ηC

+ = γCIiRi - const. as a function of the Ii values determined by our procedure. A linear relation as expected shows that our interpretation is valid. ηC

+ and ηA+ can now be presented as a function of Ii rather than It. This is shown in Fig.

8-8a and 8-9b. Fig. 8-8a indicates no limiting current behavior and the curve is fitted assuming

MIEC 104

only a Butler-Volmer type overpotential as given in Eq. (8-44). This yields αC = 0.45 and I0 = 0.18 A. For this type of cathode the effective electrode area is close to the nominal one [14]. For this cell S ≈ 1 cm2, hence J0 = 0.18 A/cm2 (The effective electrode area for electrode processes may be smaller than for the bulk current consideration. However, if it is at its maximum for the electrodes then it must be equal in both cases. Indeed the fitting yields S ≈ 1 cm2 for Ri = L/(Sσi)). This is also obvious in Fig. 8-9, where the ionic conductivity obtained from 4-pt measurements is compared to the conductivity determined by current interruption, without correction for S < 1 cm2.

0.0 0.2 0.4 0.6 0.8 1.00.00

0.06

0.12

0.18

0.24

air / LSC / Ce0.8Sm0.2O1.9

It [A]

700 °C

η C+ [V

]

0.0 0.2 0.4 0.6 0.8 1.00.00

0.06

0.12

0.18

0.24

b)

η A+ [V

]

It [A]

700 °C

Ce

0.8 Sm

0.2 O1.9 / N

i-Ce

0.9 Ca

0.1 O2-x / H

2 O-H

2

Fig. 8-5 Measured overpotential of the cathode ηC+ just after current interrupt versus cell current

It (a) and measured overpotential of the anode ηA+ just after current interrupt versus cell

current It (b).

The overpotential of the anode exhibits an increase at high currents indicative of limiting current polarization in addition to a possible charge transfer overpotential dominant at low ionic currents. The solid line in Fig. 8b exhibits a best fit to the experimental data. It assumes that in addition to charge transfer overpotential (according to Eq. (44) there is a diffusion limitation on the

105 MIEC

transport of H2O molecules (according to Eq. (35)) and of H atoms according to an equation analogous to Eq. (38) (with p(O2) replaced by p(H2) for the anode side). The fit yields αA = 0.87, J0 = 0.28 A/cm2. The fit is not unique and other parameters may yield rather similar fitting. All fits deviate from the experimental curve by 10 - 20 mV the way it is shown in Fig. 8b. This may be due to small measurement inaccuracy, approximation of the calculation or that the electrode overpotential is governed by a reaction of the gases with the electrode material e.g. with Ni for which I-V relations were not considered in this work.

0 100 200 300 400 5000

40

80

120

160

200

240

ηC+

700 °C

γC(Ii-Ii+)Ri

η C(t)

[mV

]

t [µs]

Fig. 8-6 Current interruption measurement of a LSC cathode on Ce0.8Sm0.2O1.9 at It = 0.401 A,

S ≈ 1 cm2 . The (fast) ohmic drop is represented as γC(Ii-Ii+)Ri, the measured value just

after the interrupt is ηC+. An exponential fit to the slow decay (solid line) gives a time

constant τ = 68 µs.

MIEC 106

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

0.3

0.4

0.5

Ii+

γCIi+.Ri

η C-η

C+ +

γ CI i+.

R i

700 °C

Ii [A]

Fig. 8-7 Measured voltage drop ηC - ηC+ + const (const = γCIi

+Ri) as a function of the ionic

current, γC = 0.666 (electrode area 1 cm2).

0.0 0.2 0.4 0.6 0.8 1.00.00

0.04

0.08

0.12

0.16

Ii+

C4

air / LSC / Ce0.8Sm0.2O1.9

700 °C

η C+ +

C4

Ii [A]

Fig. 8-8a Fit for the theory of δVth,C to the measured ηC+. The parameter C4 is obtained from a fit

of the measured VCell and It to Eq. (8-71), using also Eqs.(8-21), (8-24), and (8-48). The

solid line is obtained through interpretation of ηC+ + C4 as a pure charge transfer

process (Eq. (44), I0,C = 0.18 A, αC = 0.45).

107 MIEC

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

0.3

η A +

A4

[V]

Ii+

A4

Ce0.8Sm0.2O1.9 / Ni-Ce0.9Ca0.1O1.9-x / H2O-H2

700 °C

Ii [A]

Fig. 8-8b Fit for the theory of δVth,A to the measured ηA+. The parameter A4 is obtained from a

best fit of the measured VCell and It to Eq. (8-71), using also Eqs.(8-21), (8-24), and

(8-48). The solid line is a fit to a model assuming charge transfer dominating at small

current densities (αA = 1, J0 = 0.285 A/cm2) and diffusion limitation due to slow

transport of H (or H+) to the reaction sites (A2 = 0.319 atm½/A).

0.90 0.95 1.00 1.05 1.10 1.151

10

current interruption4-pt measurement

σ i [S

/m]

103/T [K-1]

Fig. 8-9 Comparison of the ionic conductivity of the MIEC, obtained from 4 pt measurements

(chapter 5) with the ohmic drop determined by current interruption (σi = L/(SRi)) for

MIEC 108

several Ce0.8Sm0.2O1.9 fuel cells with La0.84Sr0.16CoO3 cathodes and Ni-

Ce0.9Ca0.1O1.9-x anodes.

5.3 Reduction of the MIEC/electrode ratio It was argued in section 4 that reducing the impedance of the MIEC too much with respect to the electrode impedance results in deteriorating the SOFC performance. The open circuit voltage can serve as a measure to demonstrate this effect. Fig. 10 compares the open circuit voltage of different SOFCs based on Ce0.8Sm0.2O1.9 with La0.84Sr0.16CoO3 cathode and Ni/Ce0.9Ca0.1O1.9-x cermet anode as well as Ce0.8Gd0.2O1.9 with Pt electrodes. The measured open circuit voltage decreases as the MIEC thickness is decreased thereby increasing the permeating current and therefore electrode overpotential loss. The low values measured by us are consistent with values measured by Milliken et al. [17] and by Eguchi et al. [18]. Instead of changing the MIEC/electrode impedance ratio by changing the MIEC resistance one can change the electrode impedance. By decreasing p(O2)high on the cathode the overpotential due to diffusion for each current Ii increases. With a constant MIEC thickness the impedance ratio between MIEC/electrode decreases as p(O2)high decreases. This is shown in Fig. 8-11. The effect is more distinct at elevated temperatures since the MIEC resistance decreases exponentially with T while δVth,C

D depends weakly on temperature.

0 500 1000 1500 20000.70

0.75

0.80

0.85

0.90

VO

C /

Vth

,app

VOC / Vth,app for reversible electrodes

800 °C

700 °C

800 °C

700 °C

800 °C

700 °C

MIEC thickness [µm]

Fig. 8-10 Dependence of the average ionic transference number ( t V Vi OC th app= / , ) of

electrode/electrolyte systems on the MIEC thickness. (∆) LSC / Ce0.8Sm0.2O1.9 / Ni-

Ce0.9Ca0.1O1.9, (•) Pt / Ce0.8Gd0.2O1.9 / Pt , (+) LSM / Ce0.8Sm0.2O1.9/ Ni-ZrO2 (800 °C)

109 MIEC

[18], (*) LSC / Ce0.8Gd0.19Pr0.01O1.9 / Ni-CeO2 at 700 °C and 800 °C [17]. All

measurements with air at the cathode side and 90% Ar - 10% H2 saturated with H2O as

fuel.

0.001 0.01 0.1 10.0

0.2

0.4

0.6

0.8

1.0V O

C /

Vth

,app

800 °C700 °C600 °C

p(O2)high [atm]

Fig. 8-11 Dependence of the average ionic transference number (VOC) of a fuel cell

p(O2)high / LSC / Ce0.8Sm0.2O1.9 / Ni-Ce0.9Ca0.1O1.9-x / H2-H2O on the applied

oxygen partial pressure (p(O2)high).

6. Chemical potential profiles

In the following section the connection between electronic defect distribution and oxygen chemical potential is derived. The difference in oxygen chemical potential of a location x within the MIEC and the atmosphere at the anode side (µ(O2)low) is given by (Eq. (8-10), (8-11), (8-13), (8-20) and (8-40))

12 2 2 2∆µ ∆µ( ) ln

( )( )

O kTn x

n ext Ae= − = −

(8-72)

where from Eqs. (8-10), (8-11), (8-15), (8-17), (8-18), and (8-20),

( )n xn ext A

ee

eeq V MC

q V MC V MC

q V MC V MCq Vth

thxL

th

th A( )( )

( )( ( ) ( ))

( ( ) ( )),= − −

−

−

−− −

− −−1 1

1

1β

β

ββδ

(8-73)

The use of Eq. (8-5) to (8-7) yields

∆µ( ) ln)

)O kT

p(Op(O

x

low22

2=

(8-74)

MIEC 110

From Eq. (8-5) and (8-20), the variation in oxygen chemical potential across the electrolyte is expressed in terms of p(O2).

p(O p(On x

n ext Ax low

2 2

4) )

( )( )

=

−

(8-75)

In Fig. 12 oxygen chemical potential profiles for a Ce0.8Sm0.2O1.9 electrolyte at 700 °C are shown for the electrodes of Fig. 8-8 a-d (for comparison, dashed lines indicate the profile for reversible electrodes). The electrode overpotential induces a potential step at the anode/electrolyte interface (x = 0) and at the cathode/electrolyte interface (x = L), due to electrode overpotentials (δVth,C and δVth,A). The electrode overpotential shifts the oxygen chemical potential to higher values at the anode and to lower values at the cathode. For low cell voltages, the oxygen chemical potential is shifted to higher values, leading to a lower electronic conductivity of the MIEC and therefore to a much smaller electronic current (since both, V(MC), the driving force for electrons, and σe are decreased).

0.0 0.2 0.4 0.6 0.8 1.0-25

-20

-15

-10

-5

0

p(O 2)lo

w

p(O

2)high

0.25 V

0.57 V

VOC

VOC

0 V

0 V

log

(p(O

2)/a

tm)

x/L

Fig. 8-12 Oxygen partial pressure profiles across the MIEC. Solid lines are for the cell presented

in Fig. 8-4 to Fig. 8-8. Dotted lines: assumption of no electrode overpotential (CSO

700 °C).

111 MIEC

0.0 0.2 0.4 0.6 0.8 1.00.0

0.2

0.4

0.6

0.8

1.0

EMFVOCVOC

0 V

0 Vn(x)

/ n 0

(0)

x/L

Fig. 8-13 Electronic defect distribution within the electrolyte with respect to n0(0) for the same cell

as shown in Fig. 8-12 (solid lines). Dotted lines: assumption of no electrode

overpotential (n0(0), see text) (CSO 700 °C).

In Fig. 8-13, the corresponding electronic defect distribution is depicted as the ratio n(x)/n0(0), the parameters are the same as in Fig. 8-12. Here, n0(0) denotes the electronic defect concentration, one would find for the electrolyte material in equilibrium with the oxygen partial pressure at the anode side (x = 0), without consideration of electrode overpotentials and chemical potential gradients. Overpotentials at the anode and low cell voltages lower the electronic defect concentration within the MIEC and therefore, lead to an increase of the electronic resistivity.

7. Summary

The relations between cell voltage, VCell, applied chemical potential difference ∆µ(O2)app and cell current It for solid oxide fuel cells (SOFC), based on mixed ionic electronic conductors (MIECs) have been evaluated, taking also electrode impedance into consideration. The analysis yields also the relation of VCell and ∆µ(O2)app (via a theoretical Nernst voltage, Vth,app) to the partial ionic current, Ii, and partial electronic current through the cell. The analysis assumes that the MIEC conducts one type of electronic charge carriers (electrons or holes) and that the concentration of them is determined by deviation from stoichiometry. This is specifically used in Eq. (8-6) and (8-28). One can however consider also the case

MIEC 112

where the MIEC conducts both electrons and holes. The Ie - V(MC), Vth(MC) implicit relations can be found in Ref. [19]). This type of MIECs is however not common. We have also analyzed the four probe method for characterizing SOFCs and have evaluated the cathode overpotential, ηC

+, and anode overpotential, ηA+, current relations.

The equations derived depend on chemical potential drops, δVth,C and δVth,A, reflecting differences between µ(O2) as applied and on the MIEC. These differences have to be expressed as I-V relations for specific electrode processes. We have considered three types of chemical potential drops, one due to diffusion of molecules and diffusion of atoms and a third one due to charge transfer. Other processes may also be relevant in certain SOFC systems, for instance overpotential arising from reactions at the electrodes. The corresponding I-V relations have to be developed and substituted for δVth,C

R or δVth,AR.

The fitting of the theory to the data can be simplified and is more certain, if the various electrode processes are enhanced or reduced by changing the p(O2) conditions at both electrodes. Thus by reducing p(O2) at the cathode one can introduce diffusion limitation and by raising p(O2) on can remove this limitation. The p(O2) dependence of the limiting current teaches us which of the overpotential models is the relevant one in this p(O2) range. In addition, by increasing the current through the cell using an applied voltage, beyond the fuel cell operation condition (Vapp < 0) first Ie << Ii ~ It [13] and secondly one can enhance gas polarization impedance. This then shows experimentally at what current density gas polarization is important and whether under fuel cell operating conditions it is small or can even be neglected as seems to be the case presented in Fig. 8-8a. It is shown that in SOFCs based on MIECs it is not advantageous to reduce the MIEC thickness below a certain limit. This limit is reached when roughly the resistance to electronic current through the MIEC is not much larger than the electrode impedance. The reason is that then the internal electronic leak short circuits the cell, even if Re is relatively large with respect to the resistance of the MIEC to ionic current, Ri. This is different from the case of SOFCs based on ideal solid electrolytes where a decrease in their thickness is an advantage (being limited by mechanical considerations only). The effect of V(MC) on the electronic current is visualized by oxygen chemical potential profiles, from which one can see, that not only the electronic defect concentration in the electrolyte is diminished, but also the driving force for the electrons (V(MC)). This results in an exponential decrease of the electronic current towards lower cell voltages.

113 MIEC

8. References


2. P.N. Ross, Jr. and T.G. Benjamin, Journal of Power Sources, 1, 311 (1976/77).

3. D.S. Tannhauser, J. Electrochem. Soc., 125, 1277 (1978).


5. I. Riess, Solid State Ionics, 52, 127 (1992).

6. C.S. Tedmon, Jr., H.S. Spacil, and S.P. Mitoff, J. Electrochem. Soc., 116, 1170 (1969).

7. D.Y. Wang and A.S. Nowick, J. Electrochem. Soc., 126, 1155 (1979).

8. D.R. Franceschetti and J.R. Macdonald, J. Electoral. Chem., 82, 271 (1977)

9. D. Braunsthein, D.S. Tannhauser, and I. Riess, J. Electrochem. Soc., 128, 82, (1981).

10. M.J. Verkerk, M.W.J. Hammink, and A.J. Burggraaf, J. Electrochem. Soc., 130, 70 (1983).

11. B.A. van Hassel, B.A. Boukamp and A.J. Burggraaf, Solid State Ionics, 48, 139 (1991)

12. K.J. Vetter, Electrochemical Kinetics, Acad. Press, New York (1967), p. 117.

13. I. Riess, J. Phys. Chem. Solids, 47, 129 (1986)

14. K. Sasaki, J.P. Wurth, R. Gschwend, M. Gödickemeier, and L.J. Gauckler, J. Electrochem.

Soc., to be published.

15. M. Kertesz, I. Riess, D.S. Tannhauser, R. Langpape, and F.J. Rohr, J. Solid State Chem.,

42, 125 (1982).

16. A.N. Petrov and P. Kofstad, in Proc. Third Intl. Symp. SOFC, S.C. Singhal and H.

Iwahara, Editors, PV 93-4, p. 220, The Electrochem. Soc. Proceedings Series, Pennington, NJ (1993).

17. C. Milliken, S. Elangovan and A.C. Khandkar, in Ionic and Mixed Conducting Ceramics,

T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller, Editors, PV 94-12, p. 466, The Electrochem. Soc.

Proceedings Series, Pennington, NJ (1994).


19. I. Riess, Phys. Rev. B 35, 5740 (1987).

MIEC 114

List of abbreviations and symbols

Abbreviations SOFC Solid Oxide Fuel Cell SE Solid Electrolyte MIEC, MC Mixed Ionic Electronic Conductor LSC Strontium doped Lanthanum Cobaltite LSM Strontium doped Lanthanum Manganite Symbols N (n) concentration of ionic (electronic) carriers Nintr, Nextr intrinsic and extrinsic ionic charge carrier concentration q elementary charge k Boltzmann constant β 1/kT σe electronic conductivity νe electronic mobility σi ionic conductivity νi ionic mobility p(O2)high applied oxygen partial pressure at the cathode side p(O2)low applied oxygen partial pressure at the fuel (anode) side Ie, (je) electronic current (density) Ii, (je) ionic current (density) Ii

+ , Ii- ionic current just after (+), and before (-) current interrupt

It cell current It = Ii + Ie Re resistance of the cell to electronic current f(VCell, p(O2)high, p(O2)low) Ri resistance of the cell to ionic current (approx. constant) RC ohmic sheet resistance of the cathode RA ohmic sheet resistance of the anode VCell measurable cell voltage V(MC) voltage drop across the MC Vth(MC) Nernst voltage across the MC µ(O2) oxygen chemical potential q elementary charge µ(O2)high applied oxygen chemical potential at the cathode µ(O2)low applied oxygen chemical potential at the anode k Boltzmann constant

115 MIEC

Vth,app Nernst voltage given by the externally applied oxygen chemical potential S electrode cross section, current constriction occurs, if S is smaller than the apparent electrode area σe

L electronic conductivity at x = L σe

0 electronic conductivity at x = 0 δVth, δVth,C, δVth,A deviation from Vth through electrode overpotential δVth,C

D, δVth,AD

deviation part of δVth due to diffusion δVth,C

CT, δVth,ACT

deviation part of δVth due to charge transfer δVth,C

R, δVth,AR

deviation part of δVth due to a chemical reaction δV, δVC, δVA voltage drop through electrode (ohmic sheet) resistance A1, C1, C2 constant connected with diffusion at cathode and anode O= oxygen ion ~µ e electrochemical potential of electrons ~( )µ O=

electrochemical potential of oxygen ions ∇ =~(µ O gradient of electrochemical potential of oxygen ions J0 exchange current per unit area in the Butler-Volmer equation α, αC, αA charge transfer coefficient in the B-V equation I0,C, I0,A exchange current for the cathode and the anode, resp. K1 constant for the H2, H2O, O2 equilibrium ∆y distance between working and reference electrodes ηC measured voltage between cathode and Ref C ηC

+ measured ηC immediately after current interrupt ηA measured voltage between cathode and Ref A ηA

+ measured ηA immediately after current interrupt C3, A3,C4,A4, D constants δVth,C

-, δVth,A- deviation of δVth just before current interrupt

δVth,C+, δVth,A

+ deviation of δVth just after current interrupt

γC, γA fraction of IiRi measured between cathode and Ref C or anode and Ref A n0(0) electronic defect distribution of the MIEC in equilibrium with the ambient atmosphere at the anode without consideration of overpotentials and chemical potential gradients

9. Electrochemical Characteristics of Cathodes on Ceria based Electrolytes

Abstract

The ionic current - overpotential characteristics of cathodes on ceria-based electrolytes have been evaluated by galvanostatic current-interrupt measurements under fuel cell operating conditions. The effect on the oxygen reduction kinetics of the simultaneous transport of electrons and oxygen (either in the form of ions or neutral species) was investigated. As model substance for cathodes served La0.8Sr0.2MnO3 , Ag, La0.84Sr0.16CoO3 , Pt and Au. All five materials exhibit good electronic conductivity. Pt and La0.8Sr0.2MnO3 have a low ionic conductivity. Ag and La0.84Sr0.16CoO3 allow transport of oxygen or electrons/holes. These electrodes are compared with Au electrodes which do not allow oxygen transport and which are known to exhibit a a high overpotential for oxygen reduction. Electrochemical parameters for diffusion and charge-transfer governed electrode kinetics were obtained by applying the theory derived in the previous chapter. The best electrochemical performance was found for La0.84Sr0.16CoO3, a cathode material showing mixed ionic and electronic conductivity.

1. Introduction

The electrochemical performance of solid oxide fuel cells (SOFCs) is controlled mainly by electrode processes. Since these processes are thermally activated, higher electrode losses are anticipated for lower fuel cell operating temperatures. The I-V relations of mixed ionic electronic conductors (MIECs) are even more affected by electrode overpotentials compared to ZrO2-based solid electrolytes. With MIECs as electrolytes not only short circuit current and maximum power output, but also the open circuit voltage and therefore the efficiency depend on electrode performance. This was demonstrated in chapter 8. For MIEC under fuel cell operating conditions hardly any electrode overpotential measurements were done, moreover, the existing measurements show only qualitative differences between several electrode materials and no attempt was made so far to describe the electrode overpotential characteristics in a more detailed manner. On the other hand, many investigations on the overpotential-current relations were performed in electrochemical half cells with ceria electrolytes, where the ionic current through the cell is driven by an applied electrical potential, in the "oxygen pump" mode [1-2345]. Interfacial polarization has been investigated by impedance spectroscopy for Fe doped La(1-x)SrxCoO3 and for Pt electrodes [6-78]. However, these investigations do not describe the system under

Cathode 118

real conditions, i.e. under conditions where MIECs exhibit both large ionic and electronic currents. In this chapter, cathode reactions are investigated under operating conditions, in order to find the limiting steps in the overall cathode reaction. The method to separate ohmic losses and nonohmic contributions of the cathodes are separated by the well-established current-interrupt technique under fuel cell operating conditions [4, 9]. The investigated electrode materials are La(1-x)SrxMnO3 (LSM), which is the typical cathode material for zirconia based solid oxide fuel cells (SOFC), La(1-x)SrxCoO3 (LSC) which has a low impedance for oxygen reduction and exhibits mixed ionic electronic conductivity, Pt and Ag which are known to have a low impedance for the reduction of oxygen and exhibit surface diffusion of oxygen (Pt) or bulk diffusion (Ag), and Au which is known to have a high impedance for oxygen reduction. Previously, it was argued, that the mixed conductivity of ceria electrolytes might have a positive effect on the cathode reaction, e.g. by broadening the triple phase boundary electrode/gas-phase/electrolyte [5] However, results which indicate little influence of the mixed conductivity enhancing electrode kinetics are found also [2]. Therefore, LSM electrodes were investigated on ceria electrolytes as well as on zirconia electrolytes to evaluate the influence of the electrolyte material. With Au electrodes having a poor activity for oxygen reduction, the influence of "blocking" electrodes on the fuel cell characteristics was investigated. Quantitative parameters, which describe the overpotential-current relations are obtained by fitting the measured current - overpotential characteristics to the relations derived in chapter 8. Additionally, for the most promising cathode material, LSC, measurements under varying oxygen partial pressures at the cathode are performed to obtain more detailed information on the oxygen reduction mechanisms. In section 2, the transport equations presented in chapter 8 are summarized and possible limiting steps in the cathode reaction are derived. The experimental procedure for the cathode preparation and for the electrochemical measurements is given in section 3. In section 4 results of the electrochemical measurements of the different cathode materials are presented and discussed, where the different cathodes are compared in section 4.5 . Conclusions are given in section 5.

2. Ionic and electronic currents in fuel cells with mixed conducting electrolytes

2.1 Partial ionic and electronic currents

119 Cathode

The current-voltage characteristics for SOFCs based on mixed ionic electronic conductors with non reversible electrodes have been derived in chapter 8. The equations are valid for a mixed conductor, which conducts only one type of electronic charge carries (e.g. electrons) and whose electronic conductivity is induced by deviation from stoichiometry. For the partial ionic and electronic currents we can write [10, 11]

IV MC V MC

Rith

i=

−( ) ( )

(9-1)

and

IV MC

Ree

= −( )

(9-2)

where Ii, the ionic current is expressed as a function of Vth(MC), V(MC) and Ri. Ri is the resistance to ionic current and is treated here as a constant due to doping of the electrolyte with lower valent cations (Sm or Gd). Vth(MC) is a virtual electromotive force given by the oxygen chemical potentials just inside the MIEC, just beyond a possible space charge region. V(MC) is the corresponding voltage drop across the MIEC, it is related to the measurable cell voltage by

V MC V I R RCell t C A( ) ( )= + + (9-3)

where RC and RA are the electrode resistances (in plane and cross plane) and It is the cell current It=Ii+Ie. In most cases, after all by the use of metallic electrodes or electrodes with a high electrical conductivity (LSM, LSC) the electrode resistances can be neglected and V(MC)=VCell. The resistance of the MIEC to electronic current Re (Eq. 9-2) is not a constant. Instead, It is a function of cell geometry, the voltage V(MC) on the MIEC, the applied oxygen partial pressures at the cathode and at the anode, and of the cathode and anode overpotentials. It is given by (chapter 8)

( )

( )

ISL

V V V V

p(Op(O

p(Oe

e

e th app Cell th C th A

ehigh

L

high

qV

q V V V V

Cell

th app Cell th C th A

= − − − −

−

−

−

− − − −

, , ,

( ) ))

) , , ,

δ δ

σβ

β δ δ22

2

14 1

1 (9-4a)

( )ISL

V V V Vii

th app Cell th C th A= − − −σ

δ δ, , , (9-4b)

Here, Vth,app denotes the Nernst potential given by the applied outer atmospheres p(O2)high at the cathode and p(O2)low at the anode. S is the effective cross-sectional area of the electrodes and L is the thickness of the electrolyte, β = 1/kT. The oxygen partial pressure at the air side just inside the MIEC p(O2)L is related to the overpotential by

Cathode 120

p(O p(O eL high VqkT th C

2 24

) ) ,= ⋅ − δ (9-4c)

Vth,app is related to Vth(MC) by

V MC V V Vth th app th C th A( ) , , ,= − −δ δ (9-5)

Where δVth,C and δVth,A are the overpotentials at cathode and anode, respectively, which have been shown to be equivalent to a drop in oxygen chemical potential (chapter 8).

2.2 The oxygen reduction reaction

The overall cathode reaction is

O g V OO Ox

2 2( ) + =•• (9-6)

where an oxygen molecule from the gas phase is reduced by four electrons and is incorporated into the electrolyte. This overall reaction can be separated into several consecutive steps, which all can limit the cathode reaction rate. Depending on the transport and oxygen reduction properties of the cathode material the following steps can take place: a) adsorption of oxygen and dissociation of O2 into 2 O b) diffusion of adsorbed oxygen to the reaction site, where the charge transfer reaction takes

place. This process can proceed either by surface diffusion (b) or by bulk diffusion (b"). Also surface diffusion of oxygen ions can be considered (b').

c) charge transfer reaction. This reaction is considered to take place at the triple phase boundary for a material with low activity for oxygen reduction (c1). For considerable bulk diffusion it can take place at the whole interface between electrode and electrolyte (c2) or directly on the surface for a cathode material with high activity for oxygen reduction (c3). Alternatively this process can consist of two steps, one (c'3) on the cathode material and one at the triple phase boundary (c4).

d) additionally, the oxygen supply to the active surface can be limited by gas phase diffusion in the electrode The possible reaction steps are depicted schematically in Fig. 9-1. A material with low oxygen reduction activity is depicted on the left hand side. The charge transfer reaction takes place predominantly at the triple phase boundary of cathode/gas phase/electrolyte. In the middle of Fig. 9-1 a cathode material with bulk diffusion of oxygen and at the right hand side a material with oxygen ionic conductivity is shown.

121 Cathode

Fig.-9-1 Schematic picture of the possible reaction steps at an SOFC cathode.

a dissociative adsorption b, b', b" diffusion of atomic oxygen or oxygen ions (O-) c1, c2, c'3, c3, c4 charge transfer reaction (electronation of oxygen) d gas diffusion of oxygen

a) The adsorption reaction of oxygen on the electrode surface is generally treated as a dissociative adsorption mechanism described by [1 - 5]

O g O O O V Oad ad2( ) → + + → (9-7)

where Vad and Oad denote a vacant and occupied adsorption site on the electrode surface, respectively. By the use of the Langmuir equilibrium adsorption equation one can express this equation as

θθ1 2

12

−= K p(Oad )

(9-8)

where θ is the surface coverage of absorbed oxygen atoms and Kad is the equilibrium constant of this reaction. p(O2) is the local value near the covered surface. b) The following step in the cathode reaction is the diffusion of adsorbed oxygen atoms to the reaction sites, described by

2 2( ) ( )O V O Vad rs rs ad+ → + (9-9)

where Vrs and Ors denote vacant and occupied reaction sites. A diffusion overpotential, due to limited supply of oxygen atoms is

Cathode 122

( )δV

kTq

p(O

p(O C Ith CD

high

highi

, ln)

)

=

−

42

2 2

212

(9-10)

where δVth,CD is the part of the cathode overpotential δVth,C which is caused by the diffusion

of atomic oxygen. Of course the same mechanism holds not only for surface diffusion of oxygen atoms, but also for diffusion of atomic oxygen in the bulk of the cathode (reaction b') or for the diffusion of oxygen ions (with single or double charge) on the surface. For this reaction a limiting ionic current (Ii,Lim) is expressed as

( )I

p(O

Ci Lim

high

,)

=2

2

12

(9-11)

Therefore, if this step determines the cathode reaction, the limiting current follows a p(O2)1/2 dependence. The corresponding δVth,C versus Ii dependence is shown in Fig. 9-2a. In Fig. 9-2a to 9-2c possible overpotential - current relations are given. Fig. 9-2a describes this relation in case a pure diffusion process is rate limiting, in Fig. 9-2b a charge transfer is rate limiting. Fig. 9-2c is the combination of the single processes, i.e. at low ionic current densities charge transfer is dominant and at high ionic current densities diffusion dominates. Diffusion of atomic oxygen can be of importance, if oxygen is supplied to the electrode by bulk diffusion of oxygen or diffusion of ions in a ionically conducting electrode. , Diffusion limitation might also occur under low oxygen partial pressures at the cathode side, where the density of adsorbed oxygen is small. In this case the mobility of oxygen decreases at low surface coverage. This has been reported by Gland et. al. for the adsorption of oxygen on Pt, where a decreasing enthalpy for adsorption was found for an increasing surface coverage [12].

123 Cathode

0.0 0.1 0.2 0.3 0.4 0.50.0

0.1

0.2

0.3a

molecular oxygen

atomic oxygen

δVth

,CD [V

]

Ji [A/cm2]

0.0 0.1 0.2 0.3 0.4 0.50.0

0.1

0.2

0.3

b

δVth

,CC

T [V

]

Ji [A/cm2]

0.0 0.1 0.2 0.3 0.4 0.50.0

0.1

0.2

0.3

0.4

CT + molecular diffusion

CT + atomic diffusion

c

δVth

,C [V

]

diffusioncontrol

charge transfercontrol

Ji [A/cm2]

Fig. 9-2 Calculated overpotential - ionic current characteristics, for a pure diffusion limited process (Fig. 9-1a), a pure charge transfer controlled process (Fig. 9-1b), and a process with charge transfer control (CT) at low and diffusion control at high current densities (Fig. 9-1c). Parameters: C1 = 0.42 cm2/A.atm, C2 = 0.92 cm2A.atm1/2, α = 0.5, I0 = 0.4 A/cm2.

c) The charge transfer reaction is described by a Butler-Volmer type equation:

I I e ei CV VC

qkT th C

CTC

qkT th C

CT= −

− −0

2 1 2,

( ), ,α δ α δ

(9-12)

For positive currents, this equation can be approximated by

I Iq V

kTith

CT

CC C= 2

20,

, )sinh(α δ

(9-13)

Cathode 124

Here δVth,CCT is the part of the cathode overpotential δVth,C due to charge transfer, αC is the

charge transfer coefficient and I0,C is the exchange current density of the cathode reaction (Fig. 9-2b). The exchange current density of the Butler-Volmer equation (Eq. (9-12)) depends on the oxygen concentration at the electrode and/or the electrolyte surface [6, 13]. If one assumes a dissociative adsorption step in the oxygen reduction reaction, the concen-tration of adsorbed oxygen on the surface in equilibrium with the surrounding atmosphere can be expressed as

( )I C0 11

2, ( )∝ −θ θ (9-14a)

Inserting Eq. (9-8) into Eq. (9-14a) yields

I K p(OC ad0 21

21

4, )∝ ⋅− −θ or I K p(OC ad0 21

12

14

, ( ) )∝ − ⋅θ (9-14b)

At high oxygen partial pressures the surface is assumed to be covered with oxygen, θ~1 and (1-θ)~p(O2)-1/2 (Eq. (9-8)). Therefore, I0,C is proportional to p(O2)-1/4 (Eq. (9-14)). At oxygen partial pressures where (1-θ)~1, θ ∼ p(O2)1/2 and I0,C ~ p(O2)1/4. This was also found experimentally [1, 2, 4]. In the form the Butler-Volmer equation is formulated in Eq. (9-12) it is in principle only valid for a one step one electron transfer reaction. However, by introducing the factor of 2 in both the exponential terms for the cathodic and for the anodic reaction, the equation can be applied to a charge transfer reaction where two electrons are transferred in one step. The transfer coefficient αC is then a measure for the symmetry of the potential barrier an oxygen ion has to overcome between the electrode and the electrolyte. In Fig. 9-3, a schematic cross section through the electrochemical double layer is shown. The picture contains three curves. Curve 1 for the chemical potential, without any double layer considered, curve 2 for the electrical potential, and curve 3, the sum of curves 1 and 2. The rate constants for the electrochemical reaction, without considering electrode overpotential would be proportional to exp(-∆µ(O2)k

‡/kT) for the cathodic and exp(-∆µ(O2)a‡/kT) for the anodic reaction where ∆µ(O2)‡

is the chemical potential maximum. The electrical potential (curve 2) is approximated by a Helmholtz-model, i.e. the electrical potential varies linearly between x = 0 and x = d. The transfer of oxygen from the cathode (µ(O2) = µ(O2)0) into the electrolyte (µ(O2) = µ(O2)d) is an activated process. The oxygen ions have to overcome a potential barrier (µ(O2) = µ(O2)‡). This potential barrier (curve 3), however, is depending on the electrical potential difference between cathode and electrolyte (i.e. the overpotential) and is the sum of curves 1 and 2. By applying a potential difference ∆ϕ between cathode and electrolyte, the electrochemical process for the cathodic reaction can be promoted and the anodic reaction is hindered. ∆ϕ is given as ∆ϕ = δVth,C - ∆ϕrev, i.e. as the difference of the cathode overpotential and the

125 Cathode

reversible potential difference between cathode and electrolyte at electrochemical equilibrium (VCell = Vth,app). In the further considerations, however, ∆ϕrev is neglected and ∆ϕ = δVth,C.

Fig. 9-3 Schematic cross-section through the electrochemical double-layer at the cathode.

1 chemical potential 2 electrical potential 3 electrochemical potential

d) If the oxygen supply to the electrode surface, where adsorption and charge transfer take place, is limited by diffusion of oxygen molecules, e.g. through the porous microstructure (Fig. 9-2a), then the overpotential (δVth,C

D) is related to p(O2)high as follows:

δVkT

qp(O

p(O C Ith CD

high

highi

, ln)

)=

−

42

2 1 (9-15)

where C1 is a constant. The limiting current depends on the oxygen partial pressure by

Ip(O

Ci Lim

high

,)

= 2

1 (9-16)

Cathode 126

Generally, the overpotential of the overall cathode reaction is limited by one of the proposed steps. If the reaction steps are consecutive, the overall reaction is the sum of the single steps. However, the overall reaction can be dominated by one of the mechanisms in different overpotential regimes. It is for example possible, that the charge transfer reaction determines the reaction rate at low current densities and that at increasing current densities diffusion is predominant (Fig. 9-2c). The dependence of the different reaction determining steps on the oxygen partial pressure is given in Table 9-1.

linear dependence on p(O2)

oxygen species examples

log(Ii,Lim) vs. p(O2) molecular oxygen gas phase diffusion

log(Ii,Lim) vs. p(O2)½ atomic oxygen surface diffusion of Oad bulk diffusion of O or O=

log(I0,C) vs. p(O2)±¼ atomic oxygen dissociative adsorption of oxygen

Table 9-1 Dependence of limiting currents and exchange current on p(O2).

In this analysis of the overpotential and the partial ionic and electronic currents it has been assumed, that only the oxygen chemical potential is affected by the overpotentials and that electrons due to their nature as quantum particles have the possibility to tunnel across the (narrow) space charge region at the electrode/MIEC interface.

3. Experimental

All experiments were carried out on ceria based electrolyte membranes with the composition Ce0.8Sm0.2O1.9 (CSO) or Ce0.8Gd0.2O1.9 (CGO). The preparation of these electrolyte membranes was done by tape casting as described in chapter 4.

3.1 Electrode preparation

The perovskite electrodes (LSC and LSM) were prepared by screen printing. Starting powders for the electrode preparation were La0.84Sr0.16CoO3 and La0.8Sr0.2MnO3 (SSC Inc., Salt Lake City USA). The powders were prepared by a spray pyrolysis process and were calcined at 900 °C, followed by ball milling.

127 Cathode

To obtain a suitable paste, the powders were mixed in an agate mortar with a dispersant (Beycostat C213, Ceca SA, Paris, F) a binder (Ethyl cellulose, Ethocel 9004-57-3, Aldrich, Milwaukee, WI, USA) and a solvent with a high boiling point (Diethylene glycol monobutyl ether acetate, Fluka AG, Buchs, CH). The recipe is given in Table 9-2. Subsequently the cathodes were printed on dense Ce0.8Sm0.2O1.9 (CSO) or Ce0.8Gd0.2O1.9 (CGO) electrolytes, using a screen printing mask with an area of 1 by 1 cm for the working electrode and 0.1 by 0.3 cm for the reference electrode. The thickness of the sintered cathode layers was about 15 µm. The cathodes were sintered at 1050 °C (LSC) and 1100 °C (LSM) for 2 hrs. The heating rate was 2 °C/min from room temperature to 600 °C with a dwell time of 1 hr at this temperature followed by heating at a rate of 3 °C/min to peak temperature. The samples were cooled down to room temperature with 5 °C/min in air.

weight % of component

Paste component

71 La0.84Sr0.16CoO3 or

La0.8Sr0.2MnO3

2 dispersant (Beycostat C 213)

17 solvent (Diethylene glycol monobutyl ether acetate)

10 binder (Ethyl cellulose)

Table 9-2 Composition of a typical screen printing slurry

Pt mesh was used as a current collector. To ensure an intimate contact between the current collector and the electrodes, the Pt mesh (Engelhard, 52 mesh, wire diameter 100 µm) was pressed onto the electrode before sintering. At the sintering temperature the current collector became intimately bonded to the cathode. A schematic picture of the measurement setup is shown in Fig. 9-4. Ag, Pt and Au electrodes were prepared by sputtering from metallic substrates in Ar atmosphere. The electrode area was masked with polymer tape to obtain the same electrode area as for the perovskite electrodes. The thickness of all sputtered electrodes was fixed to 0.5 µm. Current collectors (Pt mesh) were applied with the corresponding metal paste for the silver electrodes (Ag, Heraeus C 1075) and for Pt electrodes (Pt, C 3650). For the gold electrodes, the current collector was pressed onto the electrode and fixed along the edge of the cathode by a ceramic binder. Before the measurement, the electrode was held at the highest operating temperature for several hours to introduce sufficient porosity and to obtain a stable microstructure.

Cathode 128

A schematic picture of a sample arrangement with Pt current collectors and cathode on an electrolyte sample is given in Fig. 9-5. The current collector is connected to two wires, one for the current and one for the voltage measurement.

Fig. 9-5 Schematic view of the quartz-glass sample holder (upper image) and the sample for the electrochemical measurements (lower image).

Anodes and current collectors were attached on the other side of the electrolytes in a similar manner. The anodes were either made from Pt, Au or from a nickel cermet, a mixture which contains Ni and Ce0.9Ca0.1O1.9-x, denoted further as Ni-CCO. A detailed description of the anode preparation is given in chapter 10.

129 Cathode

3.2 Overpotential measurements

To determine the overpotential at the electrodes, a four point arrangement as shown in Fig. 9-4 was used. As we have shown in chapter 8, the ohmic and nonohmic contributions of the measured voltage, ηC, between cathode and reference electrode Ref C can be determined separately by current interrupt measurements, since the ohmic drop (relaxation of ionic and electronic charge carriers) is very fast (< 0.1 µs, chapter 5), and the nonohmic processes (discharge of the double layer at the cathode/MIEC interface and diffusion processes) are some orders of magnitude slower (usually >50 µs).

Fig. 9-4 Schematic view of the measurements setup for the determination of current - voltage characteristics and electrode overpotentials.

As shown in chapter 8, the measured overpotential just after current interrupt, when ohmic processes are already relaxed, is given by

η δC th CV C+ = −, 4 η δA th CV A+ = −, 4 (9-17)

Cathode 130

where C4 and A4 are constants reflecting the fact, that we measure against an already polarized reference electrode (Ref C, Ref A). These constants correspond to the overpotential at the reference electrodes under open circuit conditions (when the cell current It vanishes) due to ionic and electronic leaks through the MIEC which originate at the reference electrodes. The interpretation is different from the interpretation of current interrupt measurements on purely ionic conductors, where the current through the reference electrodes vanishes and the reference electrodes do not contribute therefore to the overpotential. The electrochemical properties of the samples were measured under fuel cell operating conditions with a high oxygen partial pressure at the cathode side and a low oxygen partial pressure at the anode side. The test cell was placed in a tubular furnace (Gero, Neuhausen, FRG) and the test gases were supplied to it by quartz glass tubes. The oxygen partial pressure of the fuel gas was monitored close to the anode by a zirconia oxygen sensor. A mixture of 90% Ar with 10% H2, saturated with water at ambient temperature was used as fuel. Air was supplied to the cathode side and in some cases also pure oxygen or special mixtures of Ar with O2, ranging from 1% O2 to 0.1% O2 were supplied. Since no seals were used in the cell, a slight overpressure both, for air and the fuel ensured the desired atmospheres at the electrodes. This was also verified by measuring cells with zirconia electrolytes which showed an open circuit voltage within 1 % of the theoretical Nernst voltage. The amount of gases supplied to the electrodes was regulated with needle valves and was monitored by rotameters (Vögtlin V100, Aesch BL, CH). No attempt was made to determine the fuel or air utilization, but it was ensured, that the gases were supplied in excess. Additionally, an increased air flow was used to compensate Joule heating at high cell currents. Platinum wires (0.35 mm) were attached to the electrodes. The wires were shielded by thin tubes of Ni- superalloy (HS 600, Haynes International) to minimize electric noise originating from the furnace. These shieldings were grounded. Separate wires for current and voltage on either electrode were used and the electrode overpotential was measured with two reference electrodes on the cathode and anode side, respectively. The wires were connected to the cell by electrical spot welding, which ensured good and reproducible contacts (Resistronic 3201, Brügg, CH). Steady state electrode overpotential, cell voltage, the voltage between the reference electrodes and the electromotive force of the oxygen sensor were measured with a high impedance scanning digital multimeter (Keithley DMM 2001). The cell current was measured with a digital multimeter (Keithley 197A). The current was regulated with a laboratory power supply which was used in its galvanostatic mode. The test cells were heated to 800 °C and were equilibrated at this temperature until they gave stable cell current and voltage. Subsequently the cells were measured at 800 °C, 700 °C and 600 °C to cover the desired range of intermediate operating temperatures.

131 Cathode

The temperature of the cell was measured with two thermocouples (type K) close to the cathode and close to the anode side and was continuously monitored during the measurements with a thermocouple scanner (Keithley 740 System Scanning Thermometer). A schematic picture of the sample holder setup is given in Fig. 9-5. The electrode overpotentials were characterized by time dependent and steady state measurements. For this purpose, the cell current was interrupted by a fast electronic switch and the time dependence of the voltage between reference and working electrode was monitored on a digital storage oscilloscope (LeCroy 9450 A, 300 MHz). The oscilloscope was connected to the test cell by a coaxial cable. To suppress electrical noise, each measurement was repeated at least 10 times and the averaging stacking function of the oscilloscope was used to determine the noise-free signal. Normally, only the very fast decay (attributed to ohmic processes) was measured and subtracted from the overpotential measured in the steady state to yield the iR-free overpotential. For this purpose, only the first 100 µs of the signal were followed with a very high time resolution (2.5 ns/pt), in order to avoid interpretation errors by overestimating the iR-drop. Between every single measurement the system was allowed to reach a stable steady state value until the current was interrupted again (some seconds at 800 °C up to some 10 seconds at 600 °C). The measurement setup is displayed in Fig. 9-6a (current circuit) and Fig. 9-6b (voltage and diagnostic circuit). By interrupting the current, simultaneously the measurement on the oscilloscope is triggered. After current interrupt, the switch exhibits a ringing with a duration of about 10 µs. Therefore, to determine the IR-drop the signal was extrapolated from the time just after the ringing to the time where the current was interrupted (Fig. 9-7). A similar effect has already been reported by Büchi et al. [14]. The estimation of the ohmic drop by this procedure causes an absolute error of approximately 2 mV, while the IR drop goes up to ~300 mV.

Cathode 132

Fig. 9-6 Schematics of the measurement setup, with the current measuring circuit and gas supply a and the voltage and diagnostics circuit b.

133 Cathode

0 5 10 15 2040

60

80

100

120

140

160

η C(t)

[mV

]600 °C

t [µs]

ohmic drop

0 250 5000

40

80

120

160

t [µs]η C

(t) [m

V]

Fig. 9-7 Current interrupt measurement of a Pt-cathode on Ce0.8Gd0.2O1.9 in air at 600 °C. The

cell current before interrupt is It = 0.015 A. The ohmic drop is estimated from the

extrapolation of the smooth curve just after the ringing. The inset shows the time

dependence of the overpotential up to 500 µs. Note the ringing in the first 10 µs.


4.1 Fitting steps in the evaluation of overpotentials

Using four point measurement for fuel cells based on MIECs as electrolytes described in the previous section, the overpotential can only be measured up to a constant, denoted in Eq. (9-18) as C4 and A4. It should be emphasized, that we do not measure the absolute overpotential, but only the overpotential difference between the working electrode where Ii >> Ie and the reference electrode where It = 0 (Ii = Ie). This has already been described in chapter 8. To obtain the relation between effective electrode overpotentials and ionic current, one fitting step is necessary: By varying the parameters S, C4 and A4 (Eq. 9-18), the theoretical cell current given in an explicit form in Eq. (9-18) is fitted to the measured cell current (It) and voltage VCell and

Cathode 134

measured overpotentials ηC+ and ηA

+ (Eq. (9-4a), (9-4b), (9-4c) and (9-17)) by a least square method.

( )

( )

ISL

V V C A

p(Op(O

e

e

ti

th app Cell C A

p O L

high

qV

q V V V Ve

high

i

Cell


= − − − − −

−

−

−

+ +

−

− − − −

ση η

σσ

β

β δ δ

,

( ( ) ) )) , , ,

4 4

2

21

1

1

2

14

(9-18)

The parameter S is a fitting parameter, which reflects the fact, that not the whole electrode area is used for the electrochemical reaction and that therefore a so called constriction of the current occurs [15 - 1617]. p(O2)L is the oxygen partial pressure at the location L just inside the MIEC. It is related to the cathode overpotential δVth,C by

δVkT

qp(Op(Oth C

high

L, ln))

=

42

2 (9-19)

The exchange current density J0,C is obtained by division of the exchange current I0,C by the constriction factor S (0<S<1). The constriction depends strongly on the current collector used for the measurements and on the electrode itself. Normally, S ranges from ~0.3 to 1. The parameters C4 and A4, which reflect the polarization of the electrodes due to the ionic current at VOC, are of the order of 30 to 100 mV for electrodes with a high electrochemical performance, but for electrodes with a poor performance they can reach up to 500 mV! The ionic and electronic conductivities of the electrolyte were determined by four point measurements and are described in chapter 5. According to Eq. (9-17) we now can relate the effective electrode overpotential to the ionic current, or by using the parameter S we can relate it to the ionic current density. In a second fitting step, the relevant electrochemical parameters are determined (αC, I0, diffusion limitation constants Cn and n). This is done by a least square fit of the effective electrode overpotential to:

( )δα

VkT

qp(O n p(O

kTq

IIth C

high high C IC

i

C

nn i,

,ln( ) ln ) arsinh= −

+

−

4 2 22 20

1

(9-20)

where n is either 1 for diffusion of molecular oxygen or 2 for diffusion of atomic oxygen. This fit may not always produce unambiguous results. However, further selectivity of the parameters is obtained by comparing with fitted parameters for electrochemical cells of identically prepared cathodes having a different MIEC thickness and/or anode. The fitted parameters must be quite similar.

135 Cathode

4.2 La0.84Sr0.16CoO3 cathode

LSC exhibits a low overpotential for oxygen reduction and has therefore turned out to be the standard cathode material for SOFC with ceria based electrolytes [18, 19, 20]. The requirements for a cathode with a high electrochemical performance are • a small thermal expansion mismatch with the electrolyte • no chemical reaction between electrolyte and cathode • a high ionic and electronic conductivity LSC fulfills most of these requirements. The thermal expansion coefficient of some La(1-

x)SrxCoO3 compositions and their ionic and electronic conductivities are summarized in Table 9-3. For comparison values of Ce0.8Sm0.2O1.9 and LSM are given also.

Material

thermal expansion coefficient

10-6/K

electronic conductivity

104 S/m

600 °C 700 °C 800°C

ionic conductivity

S/m

La0.85Sr0.15CoO3 18.7 [21]

La0.85Sr0.15MnO3 11.7 [25] 1.4 104 [25]

La0.8Sr0.2CoO3 19.7 [22] 5.6 [23] 7 [22] 12.6 [21, 22]

1.6 (850 °C)

[24]

La0.7Sr0.3CoO3 18.3 [21]

La0.6Sr0.4CoO3 1.4 [22] 1.5 [22] 1.6 [22] 1.5 (700 °C) [25]

La0.5Sr0.5CoO3 22.0 [21]

La0.65Sr0.35MnO3 1.10-5 [26]

Ce0.8Sm0.2O1.9 12.8 4.10-10 6.10-9 9.10-8 3.4 (700 °C)

Table 9-3 Thermal expansion coefficient ionic and electronic conductivity of some selected oxide compositions.

LSC cannot be used on zirconia electrolytes, since it readily forms poorly conducting interphases (La2Zr2O7, SrZrO3) [27 - 282930]. However, on ceria electrolytes no such phases are found [18]. The only drawback of LSC is its high thermal expansion coefficient of 15 to 20 10-6/K compared to ceria electrolytes having thermal expansion coefficients of 12 to 13.5 10-6/K (chapter 7). However, for electrolyte layers with a thickness of 10 to 20 µm a very good adherence to the electrolyte was found, even for electrode areas of more than 30 cm2.

Cathode 136

The microstructure of a La0.84Sr0.16CoO3 cathode prepared by screen-printing is shown in Fig. 9-8. The cathode was sintered at 1050 °C and has a thickness of 15 µm. The grain size is 0.3 µm. This cathode, prepared from a ball milled powder shows a homogeneous microstructure. In Fig. 9-9, a cathode made from a powder with the same composition, but not ball milled shows an inhomogeneous microstructure. In Fig. 9-10, the current-voltage characteristics of a fuel cell with a LSC cathode a Ce0.8Sm0.2O1.9 electrolyte with a thickness of 240 µm, and a Ni-Ce0.9Ca0.1O1.9-x anode are shown in the range of 600 °C to 800 °C. The solid lines correspond to fit according to the first fitting step outlined in the preceding section (Eq. 9-18). The parameter S is determined as equal to 1, which means that no constriction occurred. The cell shows a maximum power output of more than 0.45 W/cm2 at 800 °C and 0.09 W/cm2 at 600 °C (Fig. 9-11). For technical application, a cell current density of at least 0.2 W/cm2 is envisaged, which is obtained already at temperatures below 700 °C for this fuel cell.

Fig. 9-8 Fracture surface of a La0.84Sr0.16CoO3 cathode (top) on a Ce0.8Sm0.2O1.9 electrolyte (bottom) prepared by screen-printing from a homogeneous, ball milled starting powder. The cathode shows a homogeneous microstructure with an average grain size of 0.3 µm.

137 Cathode

Fig. 9-9 Fracture surface of a La0.84Sr0.16CoO3 cathode (top) on a Ce0.8Sm0.2O1.9 electrolyte

(bottom) prepared by screen-printing from an inhomogeneous starting powder. The

cathode shows an inhomogeneous microstructure.

0.0 0.3 0.6 0.9 1.20.0

0.2

0.4

0.6

0.8

1.0

600 °C

700 °C

800 °C

It [A]

V Cel

l [V

]

Fig. 9-10 VCell vs. It of a fuel cell:

air/La0.84Sr0.16CoO3/CSO/Ni-CCO/Ar-H2-H2O. The electrolyte thickness is 240 µm and

the apparent electrode area, S, is 1 cm2.

Cathode 138

0.0 0.3 0.6 0.9 1.20.0

0.1

0.2

0.3

0.4

0.5

600 °C

700 °C

800 °C

It [A]

pow

er o

utpu

t [W

/cm

2 ]

Fig. 9-11 Power output of the fuel cell of Fig. 9. No constriction was determined (S = 1) and the apparent electrode area corresponds to the effective one.

0.0 0.2 0.4 0.6 0.8 1.0 1.20.00

0.04

0.08

0.12

0.16

0.20

0.24

800 °C

700 °C

600 °C

η C+ +

C4

[V]

Ii [A]

Fig. 9-12 LSC cathode overpotential characteristic in air.

139 Cathode

In Fig. 9-12 the electrode overpotential is plotted against ionic current, after the second fitting step. The observed relation between (ηC

+ + C4) and the ionic current (Ii) is comparable to the pure charge transfer reaction overpotential depicted in Fig. 9-2b. The solid lines indicate a fit to the Butler-Volmer equation without any diffusion overpotential contribution. The parameters C4, S, αC and J0,C are summarized in Table 9-4. It is seen that the parameter C4, the initial overpotential at VOC due to the ionic leakage current, increases somewhat with increasing temperature, while the exchange current density rises exponentially as well as the ionic leak current under open circuit conditions. The increase can be understood by the mixed conducting properties of the electrolyte, which at higher temperature has a higher ionic leakage current at VOC compensating for the lowered electrochemical impedance resulting in a larger overpotential.

T [°C] α C S J0,C [A/cm2] C4 [V] Ji at VOC

[A/cm2]

600 0.262 1 0.094 0.0193 0.045

700 0.45 1 0.18 0.031 0.145

800 0.65 0.8 0.55 0.03 0.45

Table 9-4 Electrochemical parameters for the system air/La0.84Sr0.16CoO3/CSO.

Since LSC electrodes are the most promising cathodes for ceria based fuel cells it is important to determine the rate limiting steps in the overall oxygen reduction reaction. The overpotential-ionic current characteristic of LSC cathodes in air can be fitted very well to a Butler-Volmer type equation. We suggest a charge transfer controlled reaction at this electrode. In the range from VOC to VCell = 0, no diffusion limitation was detected. Previously, the charge transfer step was also suggested as rate determining step for LSC, with a preceding dissociative adsorption step of oxygen on the electrode [4]. The activation energy of the exchange current density is found to be 0.7 eV. This is in good agreement with the values found by Inoue et al. [4] for La0.6Sr0.4CoO3. As pointed out in section 2.2 of this chapter, the dependence of the exchange current and the limiting currents on oxygen partial pressure at the cathode are related to specific elementary processes. In order to clarify the rate determining step here, measurements with different oxygen partial pressures at the cathode were performed on LSC electrodes. The dependence of the cathode overpotential on oxygen partial pressure at 700 °C is shown in Fig. 9-13.

Cathode 140

The solid lines again represent a fit according to the second fitting step with Eq. (9-20). At lower oxygen partial pressures the overpotential - ionic current relations show a behavior as depicted in Fig. 9-2c and a clear limiting current is observed at higher current densities. A plot of the exchange current density I0,C of LSC cathodes versus p(O2) is given in Fig. 9-14. For comparison data of I0,C for La0.6Sr0.4CoO3 on a Ce0.9Ca0.1O1.9 electrolyte given in Ref. [4] are also displayed. Although these results were obtained by measurements in electrochemical half cells with the same oxygen atmosphere on cathode and anode (oxygen pump mode), the exchange current density is comparable to the present results obtained under fuel cell operating conditions. The dependence of the exchange current density on the oxygen partial pressure follows a dependence of J0,C ~ p(O2)-¼ . As pointed out in section 2.2, it is suggested for this dependence, that the oxygen reduction reaction at the cathode consists of a dissociative adsorption step of oxygen molecules (Eq.(9-7)) followed by a charge transfer step. Diffusion plays a minor role in the overall cathode reaction. It is only important at low oxygen partial pressures (< 0.21 atm). In the high p(O2) regime (0.2 to 1 atm ), the cathode reaction is determined by charge transfer in the whole current density range from VOC to short circuit.

0.0 0.2 0.4 0.60.0

0.1

0.2

0.3

0.4

0.5

0.6

η C+ +

C4

1 atm

0.21 atm

0.0113 atm

0.001 atm

Ii [A]

Fig. 9-13 Cathode overpotential vs. ionic current as function p(O2) at 700 °C. The solid lines are fits to Eq. (20). (p(O2)/La0.84Sr0.16CoO3/CSO)

Charge transfer processes described by the Butler-Volmer equation are characterized by a second parameter, the charge transfer coefficient αC. This transfer coefficient is a symmetry

141 Cathode

parameter, describing the location of the maximum electrochemical potential within the electrochemical double layer (Fig. 9-3). For La0.84Sr0.16CoO3, the transfer coefficient was found to depend on the oxygen partial pressure of the gas fed to the cathode. In Fig. 9-15 the dependence of the transfer coefficient αC on p(O2) is given at 600, 700 and 800 °C. The variation of the transfer coefficient with oxygen partial pressure at 700 and 800 °C is attributed to a change in the oxygen stoichiometry of the La0.84Sr0.16CoO3 cathode, thereby modifying the shape of the potential barrier the oxygen ions have to overcome. At 600 °C it is constant over most of the p(O2) range measured. The LSC cathode sample on which the oxygen partial pressure dependent electrochemical measurements were performed had the same composition as the other samples, however with a somewhat lower exchange current density.

0 1 2 3 4

0.01

0.1

1/4

J 0,C

[A

/cm

2 ]

800 °C

700 °C

600 °C

- log (p(O2)/atm)

Fig. 9-14 Exchange current density of La0.84Sr0.16CoO3 cathodes as a function of p(O2), showing a dependence of log(J0,C) vs. log p(O2) of 1/4. Solid symbols: this work, open symbols: measurements of Inoue et al. (Ref. [4], La0.6Sr0.4CoO3).

Cathode 142

0.0 0.5 1.0 1.5 2.0 2.5 3.00.0

0.1

0.2

0.3

0.4

0.5

0.6 800 °C 700 °C 600 °C

α C

- log (p(O2)/atm)

Fig. 9-15 Dependence of αC on p(O2) for a La0.84Sr0.16CoO3 cathode on CSO

The reproducibility of the electrode deposition on one hand and the reproducibility of the fitting procedure on the other hand are depicted in Fig. 9-16. Most of the overpotential ionic current characteristics are within a range of ± 5 to 15 mV. Only one curve shows a clearly diffusion limited behavior. This behavior can be explained by the different morphologies of the starting powders used for the electrode preparation: Whereas the electrode with the diffusion limited behavior was made from a strongly agglomerated powder, the other ones were made from ball milled powders leading to a homogeneous cathode microstructure. Similar results have already been reported on LSM electrodes on zirconia electrolytes [16]. However, both types of electrodes revealed the same parameters for αC and J0,C, i.e. the increase in overpotential for the cathode made from agglomerated powder is solely due to a gas diffusion limitation. Moreover the fact that for both electrodes the same charge transfer parameters were determined indicates the usefulness of our approach. The accuracy of the measurement and of the fitting procedure was estimated as ± 0.05 for αC , ± 30 % for J0,C and ± 0.1 for S. The mixed ionic electronic conductivity of LSC and the fact, that the transfer coefficient changes at high operating temperatures and low oxygen partial pressures, suggests that the oxygen reduction reaction on LSC electrodes takes place on the surface of the electrode material, since a change in oxygen partial pressure strongly changes the defect concentration in LSC. The relatively high ionic conductivity of LSC (Table 9-3) also supports an oxygen

143 Cathode

reduction reaction which at least partly takes place on the cathode surface (reaction c3, Fig. 9-1). Diffusion polarization for LSC cathodes plays only a minor role in the current density range which is interesting for technical applications. Only at low applied oxygen partial pressures (<< 0.21 atm) diffusion polarization is remarkable. From the data shown in Fig. 9-13 the limiting current - p(O2) dependence is roughly proportional to p(O2)1/2.

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

η C+ +

C4

[V

]

700 °C

Ii [A]

Fig. 9-16 Overpotential vs. ionic current of several La0.84Sr0.16CoO3 cathodes on CSO. The lines correspond to a fit to Eq. (20) with S = 1 , I0,C = 0.18 A and αC = 0.45. For a cathode with inhomogeneous microstructure (solid squares) an additional diffusion limitation had to be introduced (C2 = 1 atm½A).

4.3 Platinum cathode

Platinum is the most widely investigated electrode material for solid electrolytes or mixed ionic electronic conductors. Due to its low impedance for oxygen reduction (Eq. (9-6)) it has been used as cathode material for SOFC [1 - 7, 31 - 33]. However, there is much controversy about the reaction rate-limiting step. Either gas phase or surface diffusion has been suggested [3, 5, 3131 - 3233] or the charge transfer was concluded to be rate limiting [1, 2, 4]. However, different authors agree upon oxygen being delivered to the reaction sites by dissociative

Cathode 144

adsorption [1, 2, 4, 31]. Due to the low diffusivity of oxygen in Platinum also oxygen supply to the reaction site by diffusion through Pt-bulk can be neglected [34, 35]. In Fig. 9-17, the microstructure of the sputtered layer (0.5 µm thickness) without the contact layer of Pt-paste is shown after a heat-treatment at 800 °C for 5 hrs. From this plot the triple phase boundary length was estimated by picture analysis as 2.3.104 cm/cm2 with a surface coverage of Pt of 60.7 %. The micrograph (Fig. 9-17) shows isolated droplets of platinum with a diameter of 1 to 3 µm. Fig. 9-18 shows a SEM-micrograph of the Pt-paint contact layer. This 5 µm thick layer exhibits a good connectivity and an adequate porosity after a heat treatment at 800 °C. Fig. 9-19 shows the overpotential characteristics of sputtered Pt-cathodes in air. These Pt cathodes exhibit an overpotential-ionic current characteristics which fit very well to the Butler-Volmer equation in the temperature regime from 600 to 800 °C. This suggests a charge-transfer governed behavior. The apparent electrode area was 1 cm2. By the use of Eq. (9-20), the transfer coefficient was estimated to be 0.14 for all temperatures. No diffusion limitation was detected.

Fig. 9-17 Sputtered Pt-cathode after a heat-treatment at 800 °C for 5 hrs. The cathode consists of isolated Pt-droplets. The triple phase boundary length is 2.3.104 cm/cm2.

145 Cathode

Fig. 9-18 Micrograph of the painted Pt-paste contact layer on top of the sputtered Pt layer.

0.0 0.1 0.2 0.3 0.4 0.50.0

0.1

0.2

0.3

0.4

0.5

0.6

η C+ +

C4

[V]

800 °C

700 °C

600 °C

Ii [A]

Fig. 9-19 Overpotential vs. ionic current of Pt cathodes on CGO, showing charge transfer controlled characteristics. The solid lines are fits to Eq. (20).

Cathode 146

For electrodes which exhibit no mixed ionic and electronic conductivity or bulk diffusion of oxygen, the charge transfer reaction takes place at the triple phase boundary or on the electrode surface followed by surface diffusion of oxygen ions. However, all the ionic current passes by the triple phase boundary. Therefore, one expects a slight constriction of the ionic current and an effective electrode area S (Eq. (9-18) which is smaller than the apparent one. For sputtered Pt-electrodes the parameter S was found to be S = 1 at 600 °C and decreased gradually with increasing temperature (S = 0.92 at 700 °C and S = 0.77 at 800 °C). The overpotential at open circuit voltage C4 and effective electrode area S are summarized in Table 9-5.

T [°C] α C S J0,C [A/cm2] C4 [V]

600 0.14 1 0.012 0.07

700 0.14 0.92 0.103 0.06

800 0.14 0.77 0.649 0.03

Table 9-5 Electrochemical parameters for the system air/Pt/CGO.

The activation energy of the exchange current density in the range from 600 to 800 °C in air was found as 1.6 eV. This value is higher than ~1 eV for Pt-cathodes on ceria-based electrolytes, found by Wang and Nowick [1] on Ca doped CeO2 and by Inoue et al. [4] on Sm doped CeO2 , but it is comparable to the activation energy reported by Verkerk et al. who found 1.5 - 2 eV for Pt on Gd doped CeO2 [7].

4.4 Silver cathode

Silver is an interesting electrode material, since it possesses a good electronic conductivity as well as a high oxygen diffusivity [47]. Therefore, it might be speculated, that the oxygen uptake could take place at the entire metal surface and not only at the triple phase boundary. However, the material possesses a considerable vapor pressure at elevated temperatures and cannot be used as a high temperature (900 to 1000 °C) SOFC cathode. On the other, hand silver might be suitable at intermediate (~700 °C) operating temperatures, as in ceria based SOFCs. Fig. 9-20 shows a SEM-micrograph of the sputtered Ag-layer after a heat-treatment of 5 hrs at 800 °C, the surface coverage of silver is 59 %. The layer consists of isolated silver droplets with a diameter of 1 to 15 µm. In Fig. 9 21, the Ag contact layer (Heraeus C 1075, Hanau,

147 Cathode

FRG) on top of the sputtered layer is shown. The heat-treatment temperature of 800 °C led to a complete densification and recrystallization of this layer. The thickness of the contact layer after densification was 1 to 5 µm. In Fig. 9-22 the cathode overpotential characteristics of a sputtered Ag-cathode on a Ce0.8Sm0.2O1.9 electrolyte is displayed for 600 °C, 700 °C and 800 °C in air. The solid lines correspond to the fitting using Eqs. (9-18) to (9-20). The cathode shows a clearly diffusion limited behavior, and the exponent n was determined as 2 by fitting to Eq. (9-20). This indicates current limitation due to diffusion of atomic oxygen. This diffusion limitation can be explained with the tendency of the silver cathode to densify almost completely at elevated temperatures, due to the high mobility of Ag at elevated temperatures [36]. Also in the VCell - Ii characteristics of fuel cells with Ag-cathodes and Ni-Ce0.9Ca0.1O1.9 anodes a clear diffusion limitation is observed (Fig. 9-23). However, Ag-cathodes show a considerable exchange current density, which was estimated as 0.069 A/cm2 at 600 °C, 0.169 A/cm2 at 700 °C and 0.71 A/cm2 at 800 °C. The Ag-cathodes exhibited a remarkable constriction of S = 0.42 at 600 °C, 0.68 cm2 at 700 °C and 0.65 at 800 °C, which correlates well to the measured surface coverage of silver in Fig. 9-20. The electrochemical parameters for Ag-cathodes are summarized in Table 9-6.

Fig. 9-20 SEM-micrograph of the sputtered Ag layer after a heat treatment at 800 °C.

Cathode 148

Fig. 9-21 Micrograph of the dense contact layer on top of the sputtered Ag layer.

0.0 0.1 0.2 0.3 0.4 0.50.0

0.1

0.2

0.3

0.4

0.5

Ii [A]

η C+ +

C4

[V]

800 °C

700 °C

600 °C

Fig. 9-22 Overpotential characteristics of sputtered Ag cathodes on CSO, showing diffusion limitation due to diffusion of atomic oxygen. The apparent electrode area is 1 cm2.

149 Cathode

0.0 0.1 0.2 0.3 0.40.00

0.15

0.30

0.45

0.60

0.75

0.90

V Cel

l [V

]

It [A]

800 °C

700 °C

600 °C

Fig. 9-23 Current-voltage characteristics of fuel cells with sputtered Ag cathode (500 nm), CSO electrolyte (240 µm) and Ni-CCO anode (15 µm). The apparent electrode area is 1 cm2.

T [°C] α C S I0,C [A] C4 [V] ILim,C [A]

600 0.3 0.42 0.069 0.02 0.055

700 0.33 0.68 0.156 0.077 0.224

800 0.33 0.65 0.71 0.076 0.527

Table 9-6 Electrochemical parameters for the system air/Ag/CSO.

Due to its high exchange current density, silver is an interesting cathode material for SOFC operating at intermediate temperatures. For a technical application, the densification of silver electrodes, leading to diffusion limitation has to be avoided. Therefore, we suggest for example the use of Ag-cermets, e.g. with LSC or LSM as oxide component in order to avoid complete densification of the Ag-cathode during sintering and fuel cell operation.

4.5 Gold cathode

Cathode 150

The oxygen diffusion in Au is assumed to be orders of magnitude lower than in silver. Eberhart et. al [37] concluded from model calculations that in gold oxygen strengthens the nearest neighbor bonds in the lattice, thereby hindering any interstitial diffusion. This is in contrast to Cu and Ag in which oxygen weakens the nearest neighbor bonds, enabling interstitial movement of oxygen. MacDonald and Hayes [38] determined the adsorption of oxygen on gold surfaces up to temperatures of 500 °C. Compared to silver they found a 5 times lower concentration of oxygen on gold at this temperature. This finding supports at least partly the general assumption, that the activity for oxygen reduction on gold is low, since it is obvious that adsorption on the surface is needed for the oxygen reduction reaction. Gold electrodes were prepared by sputtering a 0.5 µm thick Au-layer on ceria substrates. A SEM-micrograph of this electrode is given in Fig. 9-24. The microstructure consists of a network of connected Au-grains with a diameter of 2 to 10 µm. The triple phase boundary has a length of 4600 cm per cm2 of electrode area. The surface coverage of gold was estimated as 50 % from SEM micrographs. The overpotential-ionic current relation of these gold electrodes with a thickness of 0.5 µm on Ce0.8Sm0.2O1.9 shows a behavior, which can be fitted to the Butler-Volmer equation and which therefore is attributed to a charge-transfer process (Fig. 9-25) at the triple phase boundary. The rather large values of S indicate that the triple phase boundary is quite wide, of the order of the pore size. No diffusion limitation was detected. The exchange current density, derived by fitting the measured data to Eq. (9-20) was 1.8⋅10-

4 A/cm2 at 600 °C, 1⋅10-3 A/cm2 at 700 °C, and 4⋅10-3 A/cm2 at 800 °C. The transfer coefficient was αC = 0.55 and was found to be independent of temperature. The low exchange current density causes a high cathode overpotential. This leads to a low open circuit voltage VOC and a VCell - Ii characteristic (Fig. 9-26) which is different from those with Pt, Ag, LSC and LSM cathodes. The slope of VCell vs. Ii curves decreases with increasing Ii. In Fig. 9-27, the partial electronic and ionic currents are shown, which lead to this unusual VCell - It characteristic.

151 Cathode

Fig. 9-24 SEM-micrograph of a sputtered Au-cathode after a heat treatment at 800 °C.

Gold electrodes have been reported to have a very high impedance for oxygen reduction [39] and are impermeable for oxygen. For these reasons, they were used to determine the effect of Fe-ion implantation in zirconia on the oxygen take-up [40], since the reaction can only take place at the gas/electrolyte/cathode triple phase boundaries. It is hypothesized that the electrolyte itself plays an important role in the oxygen reduction reaction. The activation energy of the exchange current density found for the oxygen reduction reaction of Au on Fe-implanted zirconia of 1.1 - 1.5 eV [40] and the charge transfer coefficient of αC = 0.6 [40] are comparable to the values found in this study (1.09 eV and 0.55).

Cathode 152

0.00 0.02 0.04 0.06 0.080.0

0.1

0.2

0.3

Ii [A]

η C+ +

C4

[V]

600 °C700 °C

800 °C

Fig. 9-25 Overpotential - ionic current characteristics of an Au electrode on CSO in air, the apparent electrode area is 1 cm2.

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.070.0

0.1

0.2

0.3

0.4

0.5

800 °CV Cel

l [V

]

It [A]

700 °C

600 °C

Fig. 9-26 Current-voltage characteristic of a fuel cell with sputtered Au cathode and Au anode (0.5 µm). The apparent electrode area is 1 cm2.

153 Cathode

Fig. 9-27 Partial ionic (Ii) and electronic (Ie) currents for the cell shown in Fig. 25.Ii, Ie and It are plotted on the x-axis against VCell on the y-axis. The solid dots are the measured VCell vs. It values.

However, the exchange current density we found was more than one order of magnitude higher than that for Fe-implanted zirconia and more than two orders of magnitude higher than for not implanted zirconia [40]. This suggests, that the reduction reaction in case of the mixed ionic electronic conducting ceria-based electrolyte of this study proceeds via a similar charge transfer mechanism, but with a possibly much broader triple phase boundary. The electrochemical parameters of Au-cathodes are summarized in Table 9-7.

T [°C] α C S J0,C [A/cm2] C4 [V]

600 0.55 0.7 0.0003 0.02

700 0.55 0.7 0.001 0.053

800 0.55 0.7 0.004 0.072

Table 9-7 Electrochemical parameters for the system air/Au/CSO.

4.6 La(1-x)SrxMnO3 cathodes

The most widely used cathode material for SOFC based on zirconia electrolytes is Sr-doped LaMnO3 (LSM). It exhibits a high electronic conductivity and shows much less tendency to react with the commonly used zirconia-based electrolytes than other perovskite materials do. Additionally, the thermal expansion coefficient matches very well with zirconia electrolytes. However, LSM has only a low ionic conductivity in air compared to zirconia-based

Cathode 154

electrolytes or ceria-based electrolytes [26]. Only at high overpotentials, LSM becomes reduced and develops considerable ionic conductivity [41]. For the overpotential characteristics of LSM-cathodes we therefore expect a similar behavior as for Pt-cathodes, since both material exhibit a high activity for oxygen reduction, but almost no bulk diffusion of oxygen or ionic conductivity. A SEM-micrograph of the fracture surface of a sintered La0.8Sr0.2MnO3 cathode is given in Fig. 9-28. The cathode shows a homogeneous microstructure with a grain size of 0.3 µm, which is very similar to the microstructure of the La0.84Sr0.16CoO3 cathodes (Fig. 9-8). In Fig. 9-29, the overpotential characteristic of a La0.8Sr0.2MnO3 cathode on Ce0.8Gd0.2O1.9 is given for 600 °C, 700 °C and 800 °C. The cathode shows a decreasing slope of overpotential vs. ionic current, indicating a charge-transfer governed reaction. The exchange current density J0,C (Eq. (9-20)) is 0.0082 A/cm2 at 600 °C, 0.053 A/cm2 at 700 °C and 0.151 A/cm2 at 800 °C. The activation energy of the exchange current density is 1.20 eV. The charge transfer coefficient αC, equally derived by fitting the experimental data to Eq. (9-20) was found to be 0.2 at 600 °C and 700 °C, 0.3 at 750 °C, and increased to αC = 0.62 at 800 °C. The cathodes exhibit a considerable constriction, for an apparent electrode area of 1 cm2, the effective electrode area was found as S = 0.67 at 600 °C, decreasing to 0.33 at 800 °C. This decrease might be explained by experimental difficulties. A partial detachment of the current collector would e.g. result in such an additional constriction. However, this fact shows the necessity of an analysis with respect not only to overpotentials but also to constriction. In Table 9-8, the electrochemical parameters for La0.8Sr0.2MnO3 are summarized.

T [°C] α C S J0,C [A/cm2] C4 [V]

600 0.2 0.67 0.0082 0.097

700 0.2 0.57 0.053 0.097

750 0.3 0.34 0.103 0.1

800 0.62 0.33 0.151 0.026

Table 9-8 Electrochemical parameters for the system air/La0.8Sr0.2MnO3/CGO.

155 Cathode

Fig. 9-28 Fracture surface of a sintered La0.8Sr0.2MnO3 (1100 °C / 2 hrs). The thickness of the

cathode is 15 µm with an average grain size of 0.3 µm.

The increase of the charge transfer coefficient αC is attributed to a higher degree of reduction of the LSM at increasing temperatures leading to a higher ionic conductivity. At 800 °C the charge transfer coefficients of LSM and LSC cathodes become almost equal. At lower temperatures, αC of LSM is comparable to the charge transfer coefficient of Pt. This material has neither ionic conductivity nor significant oxygen bulk diffusion but exhibits oxygen diffusion on the surface. Therefore, we suggest, that at lower temperatures the charge transfer reaction takes place on the cathode not far from the triple phase boundary (αC ~ 0.1 - 0.2) but not on the electrolyte. At elevated temperatures and high cathodic overpotentials the reaction increasingly takes place on the whole electrode surface with a charge transfer coefficient of around 0.6. THIS seems to be typical to the cathodic oxygen transfer process into an ionic conductor whether YSZ, CSO, LSM, or LSC. For gold electrodes on CSO, αC was equally determined as 0.55. Here also an oxygen transfer reaction directly on the CSO is suggested.

Cathode 156

0.0 0.1 0.2 0.3 0.4 0.5 0.60.0

0.1

0.2

0.3

0.4

0.5

0.6

Ii [A]

η C+ +

C4

[V]

800 °C

700 °C

600 °C

Fig. 9-29 Overpotential characteristics of La0.8Sr0.2MnO3 cathodes on CGO electrolyte in air (apparent electrode area 1 cm2).

In our analysis the charge transfer coefficient as it is applied in Eq. (9-12), is strictly valid only for charge transfer reactions where the electrons are transferred in a single step. It would be possible to further deconvolute αC by considering multistep reactions, as has been done before [13, 42]. However, in this work we dispense a further deconvolution, in favor of a more technical view of the overpotential. For a detailed description of the reaction mechanism, one would have to perform not only current-interrupt measurements in air, but also in atmospheres with other oxygen partial pressures and preferably by the use of impedance spectroscopy. Further work in this direction is in progress.

4.7 Comparison of cathodes

For technical applications, such as in SOFC systems, a minimal impedance is desired. In terms of the electrochemical parameters in Eq. (19) this means: a) the exchange current density J0,C as well as the charge transfer coefficients αC, should be high.

157 Cathode

b) the effective electrode area S should be equal to the apparent electrode area, i.e., the whole electrode surface should be active (minimal constriction). c) the electrode should have no diffusion limitation in the targeted operating regime. In this study, diffusion limitation occurred only for dense Ag-cathodes and for LSC electrodes at low oxygen partial pressures. The exchange current density and the charge transfer coefficient depend on the nature of the electrode and electrolyte materials used. The effective electrode area S depends not only on the electrode and the electrolyte, but also on the microstructure of the electrode [15]. If the cathode material has no oxygen diffusivity or mixed ionic electronic conductivity, the triple phase boundary per unit area should be as long as possible. If the electrode material, however, exhibits high oxygen permeability, the triple phase boundary length is no more limiting and one might even favor a considerable part of the electrolyte covered with electrode material. The cathode materials examined in this study can be classified into

a) materials with a low activity for oxygen reduction and no oxygen permeability (e.g. Au). b) materials with high activity for oxygen reduction and no or limited oxygen permeability (e.g. Pt, LSM). c) cathodes with mixed ionic electronic conductivity, or materials with high oxygen diffusivity and high activity for oxygen reduction (e.g. Ag, LSC).

Diffusion data for the investigated cathode materials are listed in Table 9-9, where diffusivity and solubility of the metal electrodes is taken as a measure for the bulk oxygen transport capacity. For the perovskite electrodes ionic conductivities are given. Gold cathodes belong to class a. We suggest, that the oxygen reduction in this case takes place on the ceria surface itself, however with a very small exchange current density. The transfer reaction takes place on the whole not gold-covered electrolyte surface, the triple phase boundary is very broad. LSM belongs to class b. These electrodes are known to have a high exchange current density. However, the oxygen permeability of Pt and LSM is very small and the charge transfer reaction takes place at the triple phase boundaries. At operating temperatures above 700 °C, the charge transfer coefficient is found to increase for LSM, suggesting a different charge transfer reaction mechanism. The ionic conductivity of LSM is still very small compared to LSC, so that the charge transfer still only takes place at a quite narrow region near the triple

Cathode 158

phase boundary. Only at very high cathodic overpotentials and/or temperatures, LSM becomes ionically conducting [43]. LSC and silver cathodes belong to class c. These electrodes have either mixed ionic electronic conductivity (LSC) or in case of Ag a good electronic conductivity combined with a high oxygen diffusivity. For LSC compositions, the ionic conductivity is of the order of the conductivity of the electrolyte (~2 S/m at 700 °C for LSC [25] compared to 3.5 S/m for CSO). Therefore, the oxygen reduction reaction can take place on the whole cathode/gas interface and the charge transfer on the electrolyte becomes negligible. Since the reaction takes place on the whole cathode surface, no constriction is observed for LSC cathodes. For Ag cathodes on the other hand a considerable constriction is observed. However, the effective area equals roughly the metal coverage of the sputtered Ag-layer. A diffusion limitation exists also due to the existence of a continuous dense contact layer of Ag on top of the sputtered porous layer. The classification is summarized in Table 9-9. As a parameter for the suitability in terms of activity for oxygen reduction, the exchange current density is displayed.

oxygen transport properties

Electrode Material

diffusivity 700 °C [cm2s-1]

solubility 700 °C [atom %]

ionic con-ductivity [S/m]

electronic conductivity

[S/m]

electrochem.

activity

(J0,C )

[A/cm2]

Au (--) [37] (--) -- 4.5.107 [44]* 0.001

LSM -- 10-5 [26] 1.4.104 [21] 0.053

Pt 2.7.10-17 [45] 3.10-12 -- 0.9.107 [45]* 0.103

Ag 1.2.10-5 [46] 0.018 -- 6.3.107 [45]* 0.156

LSC -- 80 [25] (900 °C)

7 .104 [22] 0.18

* 25 °C

Table 9-9 Transport properties of the examined cathode materials. Conductivity and diffusivity values are given at 700 °C if not stated otherwise.

4.8 Overpotential depending on the electrolyte material

It has been suggested previously by van Hassel et al. [40], that the electrolyte material itself might have an influence on the oxygen reduction reaction in combination with different cathode

159 Cathode

materials. After all, mixed conductivity or redox pair introduced i.e. by ion implanting were hypothesized to promote the cathodic reaction [40]. A marked increase of the exchange current density of more than one order of magnitude was only found in combination with gold electrodes on iron implanted zirconia. However, these exchange current densities are very low. Since gold is known to have only poor activity for oxygen reduction, the increase of the exchange current density has to be attributed to the electrolyte surface modification, i.e. the electrolyte itself is more active for the cathodic reaction and seems to be directly involved in the oxygen reduction reaction. This, however, is only observed in case of "bad" Au-cathodes. The exchange current densities for Au electrodes found in this work are higher than those given by van Hassel et al., however, they are much lower than the ones found for all other cathode materials examined. It seems, that for electrodes with a reasonable activity for oxygen reduction the influence of the electrolyte surface on the oxygen reduction reaction is very small. To examine the influence of the electrolyte surface and to clarify, whether ceria-based SOFC systems exhibit a higher performance at intermediate temperatures due to a higher surface exchange rate of ceria, experiments with LSM cathodes with identical microstructure and preparation conditions were made. In Fig. 9-29, the overpotential characteristics of La0.8Sr0.2MnO3 cathodes on Ce0.8Gd0.2O1.9, Ce0.8Sm0.2O1.9 and 8 mol% Y2O3 stabilized zirconia in air are shown at 700 °C and 800 °C. The reaction for all three electrolytes appears to be charge transfer controlled with almost the same exchange current density and electrochemical activity. The electrolyte, despite its mixed ionic electronic conductivity, shows little influence on the rate of the oxygen reduction. It is therefore likely, that the charge transfer reaction takes place on the cathode and not on the electrolyte. Similar results are also reported by Chen et al. [8], who found basically the same electrode resistance for LSM on Sm-doped ceria and yttria stabilized zirconia. Similar results are reported by Verkerk et al [2]. For Pt electrodes on either Gd-doped ceria or yttria-stabilized zirconia, they found similar electrochemical properties and proposed the same cathodic reaction mechanism for both electrolytes.

Cathode 160

0.0 0.1 0.2 0.3 0.4 0.5 0.60.00

0.04

0.08

0.12

0.16

0.20

0.24

LSM-YSZLSM-CSOLSM-CGOη C

+ + C

4 [V

]

Ii [A]

800 °C

Fig. 9-31 Overpotential of La0.8Sr0.2MnO3 cathode (screen-printing, 15 µm) on ceria and zirconia electrolytes at 800 °C. Apparent electrode area is 1 cm2.

5. Summary and Conclusions

The present analysis allows to determine whether an overpotential versus ionic current characteristic of an electrode is dominated by charge transfer or diffusion. By considering the transport and oxygen adsorption/reduction properties of the electrode material it is possible to limit the number of possible reaction mechanisms to a few. However, to identify the exact reaction mechanism and location of the single reaction steps additional investigations are necessary. These might for example include multilayer combinations of several cathode materials with different oxygen transport properties and the subsequent observation of exchange current density and constriction. Preferably such experiments are carried out with impedance spectroscopy under varying oxygen partial pressures. It was shown that a wide range of overpotential data gathered for different electrode materials on CSO MIECs at different temperatures can be analyzed in terms of charge transfer and diffusion impedance using I-V relations developed in chapter 8. The oxygen reduction reaction at cathodes on Ce0.8Sm0.2O1.9 and Ce0.8Gd0.2O1.9 electrolytes exhibits a charge transfer controlled characteristic in most cases. Only for Ag electrodes and for LSC cathodes a significant diffusion polarization was found at low oxygen partial pressures.

161 Cathode

The best electrochemical performance was obtained for fine grained, porous La0.84Sr0.16CoO3 cathodes. These cathodes show a high charge transfer coefficient and a high exchange current density. p(O2) dependent measurements of the overpotential, suggest a charge transfer controlled reaction for which oxygen is supplied by dissociative adsorption of O2 on the electrode surface. No additional constriction resistance was found for these electrodes. This suggests, that a large fraction of the surface is involved in the charge transfer reaction. LSC is known to have a considerable ionic conductivity, we therefore conclude, that the charge transfer reaction takes place on the LSC cathode/gas atmosphere interface rather than on the electrolyte surface or at the triple phase boundary gas/cathode/electrolyte. Silver electrodes show a high exchange current density. However, these cathodes tend to densify at operating temperatures. Due to this densification of the cathode surface oxygen is supplied to the electrolyte by diffusion through the dense Ag-cathode layer. On Pt and LSM cathodes, the cathode reaction is charge transfer controlled. Pt electrodes exhibit a high exchange current density, but a small exchange coefficient (αC = 0.14), which is temperature independent. For LSM, the charge transfer coefficient is temperature dependent (αC = 0.2 at 600 °C - 700 °C, increasing to 0.62 at 800 °C). This implies a different charge transfer reaction mechanism at elevated temperatures, due to changes in the oxygen stoichiometry of LSM. The ratio of effective to apparent electrode area is found to decrease with increasing temperature for Pt as well as for LSM. This could possibly be explained by an increasing influence of the electrolyte material itself on the charge transfer reaction. At lower operating temperatures, the influence of the electrolyte surface on the electrochemical reaction becomes more important, since the activation energy for the oxygen reduction on the ceria surface (1.09 eV, measured with Au-electrodes) is smaller than the activation energy for oxygen reduction Pt (1.6 eV) or LSM (1.18 eV). At higher operating temperatures however, the ionic current density increases, the cathode reaction is much faster near the triple phase boundary, and the influence of the electrolyte surface is negligible. The question whether the electrolyte surface has an influence on the electrochemical reaction at the cathode is important to clarify. Fuel cells with mixed conducting ceria-based electrolytes exhibit a much better performance than fuel cells with zirconia electrolytes which show pure ionic conductivity only. By comparison of LSM cathodes on zirconia as well as on ceria electrolytes, it might be concluded, that the electrolyte material has a negligible influence on the electrochemical reaction. This holds as long as the electrodes exhibit reasonable activity for oxygen reduction. In the other case like with Au electrodes, the oxygen reduction reaction is influenced by the electrolyte surface. In this case, ceria-based electrolytes seem to be favorable over zirconia-based electrolytes. The question, why fuel cells with ceria-based electrolytes show a better electrochemical performance compared to ZrO2-based SOFC, can be answered as follows:

Cathode 162

a) Fuel cells with ceria-based electrolytes show a better electrochemical performance compared to zirconia-based systems, since its ionic conductivity is two to three times higher than the conductivity of zirconia electrolytes, specifically in the intermediate operating temperature range. b) Ceria-based materials do not form low conducting intermediate compounds with perovskite cathodes. These are observed in case of zirconia electrolytes with LSM-cathodes and even more with LSC on zirconia where La2Zr2O7 and SrZrO3 are formed.

6. References


2. M.J. Verkerk, M.W.J. Hammink, and A.J. Burggraaf, J. Electrochem. Soc., 130, 70 (1983).

3. D. Braunsthein, D.S. Tannhauser, and I. Riess, J. Electrochem. Soc., 128, 82, (1981).

4. T. Inoue, N. Seki, K. Eguchi and H, Arai, J. Electrochem. Soc., 137, 2523 (1990).

5. T. Kenjo, Y. Horiuchi and S. Osawa, J. Electrochem. Soc., 137, 2423 (1990).


7. M.J. Verkerk and A.J. Burggraaf, J. Electrochem. Soc., 130, 78 (1983).

8. C.C Chen, M.N. Nasrallah, H.U. Anderson, in Ionic and Mixed Conducting Ceramics,

T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller, Editors, PV 94-12, p. 252, The Electrochem. Soc.

Proceedings Series, Pennington, NJ (1994).

9. N. Nardella, D.V. Ho and S.P.S. Badwal, in Ceramic Developments, p. 237-41, C.C. Sorrell

and B. Ben-Nissan, Editors, Materials Science Forum Volumes 34-36 (1988),


11. I. Riess, Solid State Ionics, 52, 127 (1992). 12 J. Gland, B.A. Sexton, and G.B. Fisher, Surf. Sci., 95, 587 (1980).

13. J. O'M. Bockris and A.K.N. Reddy, Modern electrochemistry, Vol. 2, p. 877, Plenum Press

New York, (1970).

14. F.N. Büchi, A. Marek, and G.G. Scherer, in Proc. Symposium on Electrode Materials and

Processes for Energy Conversion and Storage, S. Srinivasan, D.D. Macdonald and A.C. Khandkar, Eds.,

PV 94-23, p. 236, The Electrochem. Soc., Pennington NJ (1994).

15. C.S. Tedmon, Jr., H.S. Spacil, and S.P. Mitoff, J. Electrochem. Soc., 116, 1170 (1969).

16. K. Sasaki, J.P. Wurth, R. Gschwend, M. Gödickemeier, and L.J. Gauckler, J. Electrochem.

Soc., to be published.

17. F.P.F. van Berkel, F.H. van Heuvelen and J.P.P. Huijsmans, Solid State Ionics, 72, 240

(1994).

163 Cathode

18. C. Milliken, S. Elangovan and A.C. Khandkar, in Proc. Ionic and Mixed Conducting

Ceramics, T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller, Editors, PV 94-12, p. 466, The Electrochem.

Soc. Proceedings Series, Pennington, NJ (1994).

19. M. Gödickemeier, K. Sasaki and L.J. Gauckler, in Proc. 4th Internat. Symp. Solid Oxide

Fuel Cells, M. Dokyia, O. Yamamoto, H. Tagawa and S.C. Singhal, Editors, PV 95-1, p. 1072, The

Electrochem. Soc. Proceedings Series, Pennington, NJ (1995).


21. K. Sasaki, phD. Thesis No. 10331, Swiss Federal Institute of Technology, ETH-Zürich,

1993.

22. L.-W. Tai M.M. Nasrallah, H.U. Anderson, D.M. Sparlin, S.R. Sehlin, Solid State Ionics,

76, 259 (1995).

23. Ch. Ftikos, S. Carter and B.C.H. Steele, J. Europ. Ceram. Soc., 12, 79 (1993).

24. K. Eguchi, H. Mitsuyasu and H. Arai, in Proc. 2nd Intl. Symp. Ionic and Mixed

Conducting Ceramics, PV 94-12, p. 307, T.A. Ramanarayanan, W.L. Worrell, H.L. Tuller, The

Electrochemical Soc., Pennington, NJ (1994).

25. Y. Teraoka, H.M. Zhang, K. Okamoto and N. Yamazoe, Mat. Res. Bull., 23, 51 (1988).

26. B.C.H. Steele, S. Carter, J. Kajada, I. Kontoulis and J.A. Kilner, in Proc. 2nd Internat.

Symp. SOFC, p. 517, F. Grosz, S.C. Singhal and O. Yamamoto, Editors, Commission of the European

Communities, Athens, Greece (1992).

27. T. Setoguchi, T. Inoue, H. Takebe, K. Eguchi, K. Morinaga and H. Arai, Solid State

Ionics, 37, 217 (1990).

28. Y. Takeda, R. Kanno, M. Noda, Y. Tomida, O. Yamamoto, J. Electrochem. Soc., 134, 2656

(1987).

29. E. Ivers-Tiffée, M. Schiessl, H. Oel, W. Wersing, in Proc. Third Intl. Symp. SOFC, S.C.

Singhal and H. Iwahara, Editors, PV 93-4, p. 613, The Electrochem. Soc. Proceedings Series, Pennington,

NJ (1993).

30. C. Clausen, C. Bagger, J. Bilde-Sorensen, A. Horsewell, in Proc. 14th Riso Internat. Symp.

Mat. Sci., F.W. Poulsen, J.J. Bentzen, T. Jacobsen, E. Skou and M.J.L: Ostergard, Editors, p. 237, Riso, DK

(1993).

31. J. Mizusaki, K. Amano, S. Yamauchi, and K. Fueki, Solid State Ionics, 22, 313 (1987).

32. J. Mizusaki, K. Amano, S. Yamauchi, and K. Fueki, Solid State Ionics, 22, 323 (1987).

33. N. Robertson and J. Michaels, J. Electrochem. Soc., 137, 129 (1990)

34. F. Norton, J. Appl. Phys.,29, 1122 (1958).

35. F. Moghadam and D. Stevenson, J. Electrochem. Soc.,133, 1329 (1986).

36. J. Van herle, Oxygen Reduction Reaction Mechanisms at Solid Oxide Fuel Cell Cathodes,

Ph.D. Thesis, EPF-Lausanne, Switzerland (1993).

37. M.E. Eberhart, M.M. Donovan, and R.A. Outlaw, Phys. Rev. B, 46, 12744 (1992).

Cathode 164

38. W.R. MacDonald and K.E. Hayes, J. Catal. 18, 115 (1970).

39. S.P.S. Badwal, M.J. Bannister and M.J. Murray, J. Electroanal. Chem., 168, 363 (1984).

40. B.A. van Hassel, B.A. Boukamp and A.J. Burggraaf, Solid State Ionics, 53-56, 890 (1992).

41. B. Gharbage, T. Pagnier, and A. Hammou, J. Electrochem. Soc., 141, 2118 (1994).

42. B.A. van Hassel, B.A. Boukamp and A.J. Burggraaf, Solid State Ionics, 48, 139 (1991).

43. E. Siebert, A. Hammouche and M. Kleitz, Electrochim. Acta, 40, 1741 (1995).

44. Handbook of Chemistry and Physics, R.C. Weast and D.R. Lides, Eds., 70, p. F-146,

CRC, Press, Boca Raton FL, USA [1989/90).

45. L.R. Velho and R.W. Bartlett, Met. Trans., 3, 65 (1972).

46. T.A. Ramanarayanan and R.A. Rapp, Met. Trans., 3, 3239 (1972).

10. Electrochemical Characteristics of Mixed Ionic Electronic Conducting Ni-Ce0.9Ca0.1O1.9 Cermet Anodes for Ceria-based SOFC

Abstract

Ni-Ce0.9Ca0.1O1.9 cermet anodes with 54 vol. % nickel were tested in ceria-based solide oxide fuel cells (SOFC) under 90% Ar - 10% H2 atmosphere. The anodes exhibited a overpotential - ionic current relation which was attributed to a superposition of a charge transfer overpotential and a diffusion overpotential due to diffusion of atomic hydrogen. The exchange current density for the charge transfer reaction at 700 °C was found as 0.3 A/cm2 with an activation energy of 0.73 eV. For a similarly prepared Ni-YSZ anode on a YSZ electrolyte the exchange current density was one order of magnitude lower. This study on the anode electrochemistry is far from being complete and reflects only the present state of experimental technique at our institute. Further studies, preferably using hydrocarbons as fuel and impedance spectroscopy for the characterization are needed to fully understand the anode.

1. Introduction

A striking point in favor to solid oxide fuel cells (SOFC) is their versatility in fueling. SOFC can be operated with hydrogen, any hydrocarbons and natural gas. However, the electrode kinetics and reaction mechanisms at the anode are by far not as well known as the cathode reactions are. It has been argued, that the reaction rate at SOFC anodes is a function of the ionic conductivity of the electrolyte [1], and especially mixed ionic electronic conductivity within the electrolyte at the anode side was assumed to improve the reaction kinetics [2, 3, 4]. The standard anode material for present SOFC is a porous nickel cermet, where the oxide component of the cermet is usually yttria-stabilized zirconia (YSZ) or ceria. To achieve a sufficient electrical conductivity, the nickel content of the anode has to be higher than 30% in volume [5, 6]. However, with increasing Ni content, the thermal expansion coefficient increases also, leading to possible delamination of the anode [5] due to increasing mismatch between the cermet anode and the electrolyte. The optimum nickel content is a compromise between the thermal expansion mismatch to the electrolyte and the conductivity and is at about 50 vol. % of Ni. In contrast to the characterization of SOFC cathodes which can be performed in oxygen pump mode, i.e. with galvanic half cells in one ambient atmosphere, it is necessary to

Anode 166

characterize the anode in fuel cell mode, i.e. with oxidizing atmosphere at the cathode side and fuel gas at the anode side. Otherwise the supply of oxygen ions necessary for the anodic reaction can only be delivered by reduction of the electrolyte resulting in a reduction of the electrolyte material. The investigation of SOFC anodes under operating conditions introduces experimental difficulties, like sealing and gas handling. Therefore, the number of investigations on the anode reactions is small and often limited to the description of microstructural aspects only. For mixed ionic electronic conductors the situation is even more complicated, since the correct interpretation of anode overpotentials implies an exact knowledge of the experimental conditions not only at the anode side of the system but also at the cathode side. The interpretation involves a fitting procedure due to the fact that on mixed ionic electronic conductors electrode overpotentials are not directly measurable. This procedure has already been pointed out in chapter 8 and 9. The aim of the present chapter is to characterize anodes of Ni-cermets with Ce0.9Ca0.1O1.9 (CCO) as oxide component and to give an interpretation of the electrochemical performance of SOFC anodes on mixed ionic electronic conducting ceria-based electrolytes. For the evaluation of overpotential - ionic current characteristics, current interruption measurements were performed using diluted hydrogen at the anode side. The measured data were interpreted by the use of the equations for mixed ionic electronic conducting electrolytes developed in chapter 8, since the as measured data alone would not allow directly to draw conclusions about the reaction mechanism. The overpotential measurements are compared to measurements with Ni-YSZ cermet anodes on YSZ electrolyte. The electrochemical measurements were carried out in a temperature range from 600 to 800 °C with a fuel consisting of 90% Ar and 10% H2 saturated with H2O at ambient temperature. In section 2 of this chapter a survey of the literature on SOFC anodes is given, including possible reaction mechanisms for hydrogen and hydrocarbon oxidation. Section 3 outlines the procedure by which the chemical potential drops are derived from the measurements. In this section also a mathematical model for the interpretation of the measurements in terms of the different possible rate determining steps is explained. This model allows to determine charge transfer coefficients and exchange current densities in case a charge transfer reaction is rate determining. The experimental description for the preparation of the electrodes is given in section 4. In section 5, a screening test for the determination of the optimum anode composition and sintering temperature is described. In section 6, results of the electrochemical measurements with Ni-ceria-cermet, Au, and Pt anodes on ceria-based electrolytes are given. These results are compared to measurements with Ni-YSZ-cermet anodes on YSZ electrolytes. A summary is given in section 7. It has to be mentioned again, that the interpretation of the anode reactions applied in this study are far from being complete and the microstructure of the anodes are by far optimized. However, the aim of this study was only to derive an anode suitable for the use with ceria-based electrolytes. Exact studies of the anode

167 Anode

reactions would have to be performed with different concentrations of reactants and reaction products to describe the reaction in detail. Such investigations are now being performed in our labs and are not subject of this study.

2. Anodes for SOFC, materials and reaction mechanisms

2.1 Hydrogen oxidation

The following requirements must be fulfilled by an anode for hydrogen reduction: • high activity for the oxidation of H2 • stability against oxidation in fuel atmosphere • stable microstructure in fuel atmosphere • chemical compatibility with the electrolyte • mechanical compatibility, i.e. similar thermal expansion coefficient in fuel atmosphere at and up to operating temperature • high electronic (and preferably ionic or protonic) conductivity Possible candidate materials for the oxidation of hydrogen are transition metals like Co, Fe, Ni or noble metals like Pt, Ru and Pd. The lowest anodic overpotentials at 1000 °C were found for Ni followed by Ru, Fe, Pt and Co [7, 8]. However, the transition metals which show the highest activities for the oxidation of hydrogen tend to change their microstructure in fuel atmosphere at operating temperatures, i.e. coalescence of metal particles or densification of the electrode [9]. To prevent morphology changes and to stabilize the anode in terms of porosity and thermal expansion coefficients, mostly cermets are employed, where the oxide component provides a stable network, hindering metal particles from coalescence. Furthermore, the oxide component ensures a thermal expansion coefficient similar to the one of the electrolyte. For reasons of chemical and mechanical compatibility mostly YSZ is employed as oxide component in the cermet in case of YSZ electrolytes. Consequently, pure or acceptor doped ceria is employed as oxide component for ceria electrolytes [10]. Ceria has also been proposed as cermet oxide component for anodes on YSZ electrolytes because its mixed ionic electronic conductivity lowers the anode impedance [7, 11]. However, at high temperatures and low oxygen partial pressures ceria shows considerable lattice expansion [2, 12]. This expansion due to nonstoichiometry restricts the use of ceria in SOFC anodes to intermediate temperatures below 800 °C (chapter 7). For Ni-cermets as anodes it is generally agreed upon, that a small particle size ratio of Ni to oxide component is preferable for a low anode impedance [5, 8, 13].

Anode 168

Whereas in the oxygen reduction reaction at cathodes of SOFC only oxygen, the cathode and the electrolyte are involved, the hydrogen oxidation reaction at the anode is much more complicated to describe. It involves more chemical species and a wide variety of reaction mechanims is possible. The current density - overpotential relationship was found to follow a Butler-Volmer equation. However, the basic approach, described i.e. by Vetter [14] is insufficient, since it is only valid for a single step charge transfer reaction. Instead, the analysis of Bockris and Reddy [15] is more promising as it better reflects the complicated anodic reactions. Kawada et. al.[16] described the overpotential - ionic current characteristics of Ni-YSZ-cermet anodes in humidified hydrogen by a Butler Volmer equation with a charge transfer coefficient αA for the anodic reaction of αA = 2 and for the cathodic reaction of αC = 1.

J Ji AV

kTq V

kTth ACT

th ACT

=

− −

0

2q, exp exp, ,δ δ

(10-1)

In Eq. (1), according to the notation used already in chapter 8 and 9, J0,A denotes the exchange current density, Ji the ionic current density, and δVth,A

CT the overpotential of the anode due to charge transfer. T, q and k have their usual meanings. Similar approaches were used by Guindet et al.[17] for ball-shaped Ni-electrodes, and Uchida et al. [1] for Pt electrodes, who proposed a Tafel-type law

J Ji AF V

RTth ACT

=

0

2, exp ,δ

(10-2)

Mizusaki et al. [18] reported anode overpotentials vs. ionic current for a Ni electrode on YSZ which follow a dependency similar to the one described in Eq. (10-1). In addition to the Butler-Volmer like charge transfer overpotential, overpotentials due to chemical reaction and diffusion have to be considered, too. Diffusion processes are mostly surface and/or bulk diffusion of species involved in the anode reaction. Gas diffusion is of major importance only for anodes with a low porosity or for low concentrations of reactants, and at high current densities under low gas exchange conditions [19]. The reaction overpotential is considered as important, if a chemical reaction step is rate determining for the anode reaction. Both, reaction as well as diffusion overpotential result in a limiting current, i.e. the overpotential rises exponentially with increasing ionic current. It was noted already, that conditions under which a simple charge transfer overpotential characteristic is predominant can only be obtained for electrodes with an optimized microstructure [20]. Possible reaction mechanisms for the anode reaction on Ni-YSZ-cermets have been proposed by Mogensen and Lindegaard [21]. They propose a reaction where first hydrogen adsorbs dissociatively (2Had,Ni) on the surface of Ni particles (Eq. (10-3a)). This step is followed by a

169 Anode

charge transfer step to produce protons H+ad,Ni (Eq. (10-3b)) which then diffuse on the Ni

surface to the Ni/YSZ boundary (Eq. (10-3c)). Here the protons react with oxygen ions of the YSZ bulk forming hydroxyl ions OH- at the YSZ surface (Eq. (10-3d)) which react to O= and H2O (Eq. (10-3e)).

H Had Ni2 2↔ , 10-3a

( )2 ⋅ ↔ ++H H ead Ni ad Ni, , 10-3b

Had Ni,+ → Ni / YSZ boundary 10-3c

( )2 ⋅ + ↔+ = −H O OHad Ni YSZ YSZ, 10-3d

2 2OH H O OYSZ YSZ− =↔ + 10-3e

In the reactions described by Eqs. (10-3a) to (10-3e) a charge transfer step taking place on the Ni-surface (Eq. (10-3b)) and a diffusion step (Eq. (10-3c)) are involved. Although this reaction scheme is very complex it was not able to explain all the experimental findings for the anode reaction [21], since additional reactions like the oxidation of Ni to NiO or other adsorbed oxygen intermediates might play a role in the overall reaction. This analysis illustrates the complexity of the reaction occurring at the anode side of SOFC. For a detailed understanding of the anode reactions investigations with different water and hydrogen concentrations and with different anode microstructures would have to be performed. The aim of the present study, however, was to develop anodes for the use in SOFC with ceria-based electrolytes. We therefore concentrate further on a rather technical analysis of the measured anode overpotential characteristics, considering only a charge-transfer reaction and concentration polarization due to either limited gas diffusion of water or hydrogen or slow surface diffusion/desorption of these components.

3. Interpretation of current interruption measurements on ceria-based electrolytes

The overall reaction at the anode in a hydrogen fueled SOFC is described as follows:

O H g H O g eceria anode= −+ → +2 2 2( ) ( ) (10-4)

O=ceria denotes an oxygen ion within the electrolyte, H2(g) and H2O(g) denote gaseous

hydrogen and water, and e-anode denotes an electron within the anode. Possible reaction steps

as given e.g. by Mogensen and Lindegaard [21] (Eqs. (10-3a) to (10-3e)) are illustrated in Fig. 10-1.

Anode 170

Fig. 10-1 Schematic picture of possible reaction steps for the anodic reaction in hydrogen

atmosphere in a cermet anode with Ni (gray), pores, electrolyte (white) and CCO or

YSZ particles (white):

1) Dissociative adsorption of hydrogen on the Ni surface and charge

transfer to H+.

2) Surface diffusion of atomic hydrogen or protons to the reaction site at the

Ni/electrolyte interface.

3) Formation of OH- and desorption of water

4) Gas diffusion of molecular hydrogen into the pores and diffusion of water out of

the pores

Possible rate limiting steps in this reaction are: limited transport of hydrogen to the anode limited surface diffusion of H+ to the reaction sites and hindered transport of reaction products (H2O) from the anode. These two steps cause diffusion limitation. Reaction overpotential can arise from slow dissociation of H2 at the anode surface and from slow desorption of H2O. Charge transfer overpotential arises from the electron transfer step Had,Ni -> H+ + e- (Eq. (10-3b). Due to the relatively high electronic conductivity of CCO under SOFC operating conditions (~10 S/m at 700 °C and p(O2) = 10-22 atm) it might be speculated, that the adsorption of hydrogen a subsequent charge transfer and a reaction with O= can additionally occur at the CCO surface. However, for the analysis of the overpotential versus ionic current relation a simplified model, describing the anode rather phenomenologically is applied further. It is not

171 Anode

the intention of this analysis to describe, where the reaction occurs but what kind of reaction step limits the anode performance. In analogy to the procedure presented in chapter 9, a model is applied which allows to distinguish between anode overpotential due to charge transfer or diffusion of any species involved in the reaction. If the transport of gaseous hydrogen to the anode is hindered, a diffusion overpotential increasing the effective oxygen chemical potential at the anode is encountered, assuming local equilibrium for the reaction in Eq. (10-4):

p(O Kp(H Op(H2 1

2

2

12)

))

= (10-5)

Following the analysis presented in chapter 8, the overpotential due to increased oxygen chemical potential near the electrolyte/gas interface in the cermet anode can be expressed as

δVkT

qp(H

p(H A Ith AD

ext

exti

, ln)

)=

−

22

2 3 (10-6)

where δVth,AD denotes the overpotential at the anode due to diffusion, p(H2)ext denotes the

externally applied hydrogen partial pressure, A3 is a constant and Ii is the ionic current. T, k and q have their usual meanings. This diffusion limitation leads to a limiting current according to

Ip(H

Ai Lim

ext

,)

= 2

3 (10-7)

A possible gas diffusion limitation can also occur due to an increase in water concentration in the porous anode microstructure. This leads to an additional overpotential,

δVkT

qp(H O ext A I

p(H O extth AD i, ln

))

=+

42 1

2 (10-8)

since an increase in ionic current would increase the water concentration in the anode if the transport of water is impeded. In Eq. (8) p(H2O)ext denotes the water vapor pressure in the fuel gas which is applied to the anode. An additional diffusion overpotential can arise from limited diffusion of monoatomic species to the reaction sites, as it is e.g. proposed by Mogensen and Lindegaard [21] (Eq. 10-3c). This leads to an overpotential

( )( )

δVkTq

p(H

p(H A Ith A

Dext

exti

, ln)

)=

−

2

2 2

12

12

(10-9)

The corresponding limiting current is expressed as

Anode 172

( )I

p(H

Ai Lim

ext

,)

=2

2

12

(10-10)

According to our analysis in chapter 8 and 9 the overpotential due to charge transfer is expressed by a Butler-Volmer equation which is approximated by a hyperbolic sinus function:

δα

VkTq

IIth A

CT

A

i

A,

,arsinh=

2 2 0 (10-11)

The overpotential measured between anode and reference electrode just after current interruption (ηA

+) has to be enhanced by a constant A4 (see chapter 8) reflecting the fact, that in the usually applied 4-pt measurement for mixed conductors the reference electrode is already polarized:

δ ηV Ath A A, = ++4 (10-12)

The constant A4 is determined by fitting the measured cell voltage VCell, ionic current It and overpotentials ηC

+ and ηA+ to Eq.(8-18), varying the current constriction factor S and the

constants C4 and A4. This means in other words, that although the effective overpotential is considerably higher than the one affecting the anode, the cell voltage is not lowered this much in a mixed ionic electronic conductor (MIEC), since the overpotential also changes the defect concentration within the MIEC towards a lower electronic defect concentration, increasing thereby the ionic transference number. This has already been described in chapter 8. For the current analysis, only the overpotentials described in Eqs. (10-6), (10-8), (10-9), and (10-11) are considered. It is obvious, that also other effects like oxidation of the Ni-metal in the anode could play an important role. However, these contributions are difficult to quantify and will be subject of further studies. Similar as in Eq. (8-19) the anode overpotential can be expressed by considering the contributions of Eqs. (10-6), (10-8), (10-9), and (10-11):

( )δ

α

VkT

qp(H O A I

p(H OkTq

p(H

p(H A I

kTq

II

th A

exti

ext

ext

extn i

n

A

i

A

n,

( )

,

ln)

)ln

)

)

arsinh

=+

+

−

+

−4 2 2

2 2

2 1 2

2 4

0

1

(10-13)

The parameter n describing atomic or molecular diffusion can be either equal to 1 for molecular diffusion or 2 for atomic/ionic diffusion.

173 Anode

4. Experimental

4.1 Anode preparation

Anode layers were prepared on sintered membranes of doped ceria by screen-printing mixtures of Ce0.9Ca0.1O1.9 with NiO. A Ni-cermet anode should have a microstructure consisting of three interpenetrating networks, one being a network of the sintered oxide (YSZ or Ce0.9Ca0.1O1.9) to enhance the possible reaction area and to give a structure which hinders Ni from coalescence, the second being a metallic network consisting of Ni particles ensuring a high electrical conductivity and a network of pores ensuring supply of reactants to active sites. The oxide network should consist of rather coarse particles, however still having a high sintering activity to keep the sintering temperature for the anode as low as possible. For the Ni-ceria cermets Ce0.9Ca0.1O1.9 (CCO) powder prepared by oxalate coprecipitation (chapter 4) was used. Calcia doped ceria was applied instead of pure ceria, since this powder was sinterable already at temperatures below 1400 °C whereas for pure ceria and rare earth doped ceria usually sintering temperatures of about 1650 °C were necessary. To obtain a powder consisting of coarse particles with, however, a high sintering activity, the material was die pressed and sintered at 1650 °C for 2 hrs, subsequently the sintered body was crushed in a vibratory mill and wet-sieved to obtain a sieving fraction of 0 - 20 µm. The NiO powder was a commercial product (J.T. Baker, 1313-99-1, 99.0%, Deventer, The Netherlands) with a narrow particle size distribution and a mean particle size diameter of 1 µm and a content of sub-micron particles of about 50%. The particle size distributions of the different powders were measured by sedigraphy (Sedigraph, Micromeritics®, Norcross GE, USA) and are given in Fig. 10-2. The CCO powder has a mean particle diameter of 3 µm and contains 15% of sub-micron particles. To obtain a homogeneous microstructure and a good adherence to the ceramic membrane, it is important, that the printed electrode sinters basically only in one direction, i.e. perpendicular to the substrate. To ensure this, the printed layer must not exceed a certain thickness [22], a green thickness of about 50 µm was found to be optimal. In addition, the screen-printing paste has to have a high solids loading to minimize shrinkage parallel to the substrate. To meet these requirements a screen-printing paste with the following formulation was used: 50 wt. % NiO (J.T. Baker, 1313-99-1, 99.0%, Deventer, The Netherlands) 25 wt. % CCO (NW-ETHZ, see chapter 4) 20 wt. % solvent (Diethylene glycol monobutyl ether acetate, Fluka, Buchs, CH) 0.7 wt. % dispersant (Furan-2-carboxylic acid, Fluka; Buchs, CH) 4.3 wt. % binder (Ethocel 9004-57-3, Aldrich, Milwaukee, WI, USA)

Anode 174

The weight ratio of NiO to CCO was 2 : 1 corresponding to a volume fraction of Ni of 54% after reduction. The ceramic powders and organic ingredients were mixed in a planetary mill for 1 hr. The paste was stable and exhibited no tendency of sedimentation.

10 1 0.10

20

40

60

80

100

equivalent spherical diameter [µm]

cum

ulat

ive

mas

s pe

rcen

t

CCO

NiO

Fig. 10-2 Particle size (agglomerates) distribution (by Sedigraph) for the different powders used

in the anode preparation.

Cermet anodes of 1 x 1 cm2 were screen-printed on ceria substrates by the use of a screen with a thickness of 50 µm (Fig. 10-2). Pt-mesh current collectors were attached on top of the green anodes. These cermet anodes were compared to Pt and Au anodes prepared by sputtering as described in chapter 9.

Fig. 10-3 Schematic illustration of the screen printing process.

175 Anode

To determine the optimum microstructure for Ni-cermet anodes a screening test was carried out. For this test, Ni-CCO cermet anodes were sintered at temperatures of 1250 to 1450 °C in air and tested in fuel cell mode. For these tests, the electrolyte thickness was kept constant, a Pt electrode, prepared as described in chapter 9 was used as cathode. The microstructure of the anodes was examined at surfaces and fracture surfaces of as sintered anodes and anodes which were reduced in humidified 10% H2 at 750 °C (Ni-CCO) for 1 hr. Subsequently, anodes were characterized electrochemically by current-interruption measurements in humidified 90% Ar - 10% H2 mixtures in the temperature range from 600 to 800 °C against Pt/air. The platinum cathodes were prepared according to the procedure described in chapter 9.

4.2 Electrochemical characterization

The electrodes were electrochemically characterized using a 4-electrode setup as described in chapter 8. The test cell was placed in a tubular furnace (Gero, Neuhausen, D) and the test gases were supplied to it by quartz glass tubes. Usually, the anodes were tested with a gas mixture consisting of 90% of argon and 10% of hydrogen. This gas mixture was saturated with water by a water bubbler at ambient temperature. This procedure results in a water content of the gas of 3% at 298 K [23]. The amount of gases supplied to the electrodes was regulated with needle valves and was monitored by rotameters (Vögtlin V100, Aesch BL, CH). Hydrogen mass flow rates were varied between 0.3 and 0.8 g/h to investigate the influence of increased mass flow on the electrode overpotential. The applied fuel mass flow corresponds to an excess of 60 to 170 times the amount of hydrogen needed for an ionic current of 1 A. Additionally for one experiment pure hydrogen saturated with water was used as fuel. The oxygen partial pressure in the anode compartment was monitored with a zirconia oxygen sensor and only very little influence of the fuel mass flow on the EMF of the sensor was observed. Steady state electrode overpotential, cell voltage, the voltage between the reference electrodes and the electromotive force of the oxygen sensor were measured with a high impedance scanning digital multimeter (Keithley DMM 2001). The cell current was measured with a digital multimeter (Keithley 197A). The current was regulated with a laboratory power supply which was used in its galvanostatic mode. The test cells were heated to 800 °C and were equilibrated at this temperature until they gave stable cell current and voltage. Subsequently the cells were measured at 800 °C, 700 °C and 600 °C to cover the desired range of intermediate operating temperatures. Besides the steady state measurements, the electrode overpotentials were characterized by time dependent measurements. For this purpose, the cell current was interrupted by a fast

Anode 176

electronic switch and the time dependence of the voltage between reference and working electrode was monitored on a digital storage oscilloscope (LeCroy 9450 A, 300 MHz). The oscilloscope was connected to the test cell by a coaxial cable. To suppress electrical noise, each measurement was repeated at least 10 times and the averaging stacking function of the oscilloscope was used to determine the noise-free signal. Normally, only the very fast decay (attributed to ohmic processes) was measured and subtracted from the overpotential measured in the steady state to yield the iR-free overpotential. For this purpose, only the first 100 µs of the signal were followed with a very high time resolution (2.5 ns/pt), in order to avoid interpretation errors by overestimating the iR-drop. Between every single measurement the system was allowed to reach a stable steady state value until the current was interrupted again (some seconds at 800 °C up to some 10 seconds at 600 °C). The 4-electrode setup for the electrochemical measurements is displayed in Fig. 9-3.

5. Screening test

5.1 Microstructure

Ni-CCO cermet anodes with the composition described in section 4 were sintered at temperatures ranging from 1250 °C, 1300 °C, 1350 °C and 1450 °C for 2 hrs in air. In Fig. 10-4 to 10-7 SEM micrographs are shown to demonstrate the microstructural development of sintered electrodes. Whereas for the oxide component almost no grain growth is found, the growth of Ni-particles is considerable. The grain growth of Ni-particles determined from SEM micrographs is shown in Fig. 10-8. For good electrochemical performance it is necessary to have an intimate contact between the anode and the electrolyte and a good grain to grain connectivity of Ni and of the oxide component. It is seen, that a good connectivity between Ni-grains is only achieved at sintering temperatures above 1250 °C. Also only at temperatures above 1250 °C, the bonding of the electrode to the electrolyte was found to be sufficient, i.e. it was not possible to scratch the electrode off with a knife after cooling to room temperature. By sintering at 1450 °C however, the anode is almost dense and the electrochemical performance is very poor. Additionally, the grain size of Ni is very large after reduction, leading to a large volume change during the reduction. This volume change and the fact, that the ceria particles are almost wrapped by Ni particles leads again to a poor adherence of the anode to the electrolyte.

177 Anode

Fig. 10-4 Fracture surface of a Ni-CCO cermet anode sintered at 1250 °C/2 h, reduced at

750 °C/1 h.


750 °C/1 h.

Anode 178


750 °C/1 h.


800 °C/1 h.

179 Anode

1250 1300 1350 1400 14500

2

4

6

8

10

Ni p

artic

le d

iam

eter

[µm

]

TS [°C]

Fig. 10-8 Average Ni particle size as a function of sintering temperature (Ts). The average

diameters are derived from SEM micrographs, NiO : CCO = 1 : 1 (vol.).

At temperatures above 1400 °C grain growth of NiO leads to increasing Ni particle size after reduction with subsequent reduction of three phase contacts. At a sintering temperature of 1450 °C the anode is almost dense and grain growth of Ni leads to a insufficient connectivity of Ni (Fig. 10-6).

5.2 Current-voltage characteristics

An electrochemical screening test was carried out to find the optimum microstructure for the given cermet composition. As pointed out in chapter 8, the open circuit voltage VOC can be regarded as a measure for the quality of the electrodes for a given applied oxygen chemical potential and a given electrolyte thickness. In this screening test calcia doped ceria membranes with a thickness of 215 µm ± 5 µm were used as electrolyte. Since also the cathode (Pt) and the applied atmosphere were kept constant, the open circuit voltage is a direct measure for the electrochemical performance of the cermet anodes in the low ionic current regime.

Anode 180

1250 1300 13500.66

0.68

0.70

0.72

0.74

0.76V O

C [V

]

800 °C

700 °C

600 °C

TS [°C]

Fig. 10-9 Open circuit voltage of fuel cells: air / Pt / CCO / Ni-CCO / 90% Ar-10%H2+H2O as a

function of sintering temperature of the cermet anode. Closed symbols: Ni-CCO, open

symbols: Ni-YSZ (TS = 1350 °C).

By analysis of the total cell voltage (VCell) versus cell current (It) regime a possible diffusion limitation at higher current densities could also be detected. However by the use of these screening tests it is not possible to conclude about reaction mechanisms. In Fig. 10-9 the open circuit voltage of fuel cells with CCO electrolyte, Pt cathodes, and Ni-cermet anodes is depicted depending on the sintering temperature of the anode. At a sintering temperature of 1350 °C, the highest open circuit voltage was achieved. At 1450 °C, the open circuit voltage decreased further to values around 0.59 V. A comparison with a similarly prepared Ni-YSZ cermet anode on ceria electrolytes (open symbols) revealed a significantly lower VOC. By comparison of these findings with the microstructures of Fig. 10-4 to 10-7, it is obvious, that a bad grain to grain connection of Ni leads to a bad electrode performance and, hence to a low open circuit voltage. Increasing the sintering temperature to 1350 °C leads to a better electrode performance due to better grain to grain connection of the Ni and the CCO network. The mixed ionic electronic conductivity of the CCO leads to a higher electrochemical performance of an anode with this material compared to a similar anode with YSZ as oxide component.

181 Anode

In Fig. 10-10 and Fig. 10-11 the whole VCell - It relations are shown for the anodes of Fig. 10-9. At 700 °C the performance of the anodes sintered at 1300 °C is equal to those sintered at 1350 °C, at 800 °C, however, the anode sintered at 1350 °C shows a slightly better performance. From these screening tests a sintering temperature of 1350 °C was found to be optimum for Ni-CCO cermet anodes with a Ni content of 50 vol. %. This anode was selected because it had the best adherence to the electrolyte, the grain to grain connection of Ni and also the electrochemical performance was the highest of all investigated samples.

0.00 0.05 0.10 0.15 0.20 0.250.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

V Cel

l [V

]

Sintering temperature:1250 °C1300 °C1350 °C

700 °C

Jt [A/cm2]

Fig. 10-10 Cell current - voltage characteristics of fuels cells: air / Pt / CCO / Ni-CCO / 90% Ar-

10%H2+H2O, as a function of sintering temperature of the cermet anode. T = 700 °C.

Anode 182

0.0 0.2 0.4 0.6 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 800 °C

Sintering temperature:1250 °C1300 °C1350 °C

V Cel

l [V]

Jt [A/cm2]

Fig. 10-11 Current voltage characteristics of fuels cells air / Pt / CCO / Ni-CCO / 90% Ar-

10%H2+H2O as a function of sintering temperature of the cermet anode at 800 °C.

6. Electrochemical characterization by current interruption measurements: Results and Discussion

According to the results of the screening test, the Ni-CCO anodes presented in this section had a Ni content after reduction of 50 vol. % of the solids and were sintered at 1350 °C for 2 hrs. The anodic overpotential of Ni-CCO cermet anodes is shown in Fig. 10-12 for operating temperatures of 600, 700, and 800 °C using humidified argon-hydrogen atmosphere. The overpotential is separated from the electrolyte resistance by current interruption. The solid lines represent a fit according to Eq. 10-13, δVth,A, i.e. the deviation from the equilibrium potential is given as ηA

+ + A4, i.e. the measured overpotential just after current interrupt (ηA+) plus a

constant A4 reflecting the polarized state of the reference electrode in the 4 pt arrangement. The fitting procedure in detail is described in chapter 9. It is assumed, that the anode overpotential can be described as the sum of a charge transfer overpotential and a diffusion overpotential. Fig. 10-12 shows a overpotential versus ionic current relation with an increasing slope towards increasing current densities. Such a behavior can not be explained by charge-transfer

183 Anode

overpotential, since for this kind of overpotential the slope must decrease with increasing ionic current density. It is therefore concluded, that the overpotential relation is diffusion controlled at higher current densities. Analysis of the data revealed a superposition of a charge transfer overpotential and a diffusion overpotential which is controlled by diffusion of an atomic hydrogen species. The diffusion limitation found at higher ionic currents favors a diffusion step of dissociatively adsorbed hydrogen to the reaction sites (Ni-ceria triple phase boundary), as proposed in Eq. (10-3c). For this electrode a charge transfer coefficient of αA around 1 (Eq. (10-11) at temperatures from 700 to 800 °C with a corresponding activation energy for the exchange current density of 0.73 eV. The parameters for the charge transfer contribution to the overpotential are in accordance with the findings of Kawada et al. [16] and Guindet et al. [17] who obtained an αA of 2/n, where n is the number of electrons transferred (n = 2 for the oxidation of hydrogen). At 600 °C a charge transfer coefficient of αA of 0.23 was found. This suggests a change in reaction mechanism, however data are too limited to conclude about this. For the diffusion overpotential of atomic hydrogen the analysis revealed limiting of 0.68 A/cm2 at 600 °C up to more than 6 A/cm2 at 800 °C.

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

0.3

η A+

+ A 4

[V

]

800 °C

700 °C

600 °C

Ii [A]

Fig. 10-12 Anodic overpotential of Ni-CCO cermet anodes on a Ce0.8Sm0.2O1.9 electrolyte in

90% Ar - 10 & H2, humidified at ambient temperature. The solid lines represent a fit to

Eq. 10-13. Apparent electrode area: 1 cm2.

Anode 184

Since the anode shows a clearly diffusion controlled characteristic, the parameters for diffusion limitation are not unambiguous. The exchange current density, charge transfer coefficients and limiting currents (Eq. (10-10) found for these anodes from best fits to Eq. (10-13) are summarized in Tab 10-1.

T [°C] A4 [V] J0,A [A/cm2] α A Ji,Lim [A/cm2]

800 0.03 0.6 1 6.4

700 0.034 0.3 1 0.98

600 0.043 0.1 0.23 0.68

Tab. 10-1 Initial overpotential, exchange current density, transfer coefficient and limiting current

describing the Ni-CCO cermet anodes presented in Fig. 10-12.

It was argued [10], that also an enrichment of water in the microstructure of the anode could increase the overpotential (Eq. 10-8), but no evidence for such an effect was found. It is, however, possible, that water adsorbs on the electrode surface similar to hydrogen. Thereby the surface diffusion of adsorbed hydrogen to the reaction sites is hindered. In such a case one would expect a decreasing diffusion limitation for an increasing ratio of H2 to H2O in the fuel. In Fig. 10-13 a comparison between a Ni-CCO cermet anode fueled with a hydrogen to water ratio of 32 is compared to an anode fueled with H2 : H2O = 3. Whereas for both anodes the same charge transfer parameters were found (αA = 1 and J0,A = 0.3 A/cm2), the anode fueled with the higher hydrogen to water ratio showed no diffusion limitation. A similar effect was found for an increase in hydrogen mass flow. Fig. 10-14 shows the overpotential ionic current relations for Ni-CCO anodes with a flow rate of humidified 90% Ar - 10% H2 corresponding to a hydrogen mass flows of 0.34, 0.54, and 0.78 g/h at 700 °C. At currents around 1 A this corresponds to an excess in fuel of about a factor 20 to 40. For these anodes, again the same charge transfer parameters were found, only the term attributed to the diffusion of atomic hydrogen changed. Therefore, it is suggested, that the change in diffusion overpotential can be attributed to a removal of adsorbed water, enhancing the mobility of adsorbed hydrogen.

185 Anode

0.0 0.2 0.4 0.6 0.80.00

0.05

0.10

0.15

0.20

0.25

0.30700 °C

η A+

+ A 4

[V

]

H2:H2O=32

H2:H2O=3

Ii [A/cm2]

Fig. 10-13 Comparison between two Ni-CCO cermet anodes fueled with different hydrogen to

water ratios. 700 °C, Ce0.8Sm0.2O1.9 electrolyte, La0.84Sr0.16CoO3 cathode.

0.0 0.2 0.4 0.6 0.8 1.00.00

0.05

0.10

0.15

0.20

0.25

0.30

η A+

+ A 4

[V

]

0.78 g/h

0.54 g/h

0.34 g/h

700 °C

Ii [A/cm2]

Fig. 10-14 Dependence of the anode overpotential of hydrogen mass flow at 700 °C.

air / La0.84Sr0.16CoO3 / Ce0.8Sm0.2O1.9 / 90 Ar-10 H2 - H2O.

Anode 186

Fig. 10-15 shows the comparison of a Ni-YSZ cermet anode on a YSZ electrolyte and a Ni-CCO cermet anode on a Ce0.8Sm0.2O1.9 electrolyte. Both anodes were sintered at 1350 °C and had a Ni content of 50 vol. % after reduction. For the Ni-YSZ cermet anode a charge transfer coefficient of 1 and an exchange current density of 0.023 A/cm2 were found compared to αA = 1 and J0,A = 0.3 A/cm2. For both anodes, a limiting current due to diffusion of atomic hydrogen was found at ~ 0.7 A/cm2.

0.00 0.15 0.30 0.450.0

0.1

0.2

0.3700 °C

Ni-CCO

Ni-YSZ

η A+

+ A 4

[V

]

Ii [A/cm2]

Fig. 10-15 Comparison between a Ni-YSZ cermet anode on a YSZ electrolyte and a Ni-CCO

cermet anode on a Ce0.8Sm0.2O1.9 electrolyte. 700 °C.

7. Summary

Ni-Ce0.9Ca0.1O1.9 anodes were prepared by mixing coarse Ce0.9Ca0.1O1.9 with fine NiO. For compositions which correspond to a volume fraction of50% nickel metal after reduction of NiO, a sintering temperature for the anodes of 1350 °C was found to give the best electrochemical performance. For a NiO with an average particle diameter of d50 = 1 µm and a CCO grain size of d50 = 3 µm these sintering conditions exhibited the best grain to grain connection of Ni and an excellent adherence of the anode to the substrate. The anodes were tested under fuel cell operating conditions in an atmosphere humidified at ambient temperature with 90% Ar and 10% H2. The overpotential versus ionic current relations of these electrodes were interpreted as the sum of a charge transfer controlled

187 Anode

reaction a diffusion limited reaction. Charge transfer dominates at low current densities whereas diffusion processes dominate at high current densities. At higher fuel mass flow rates or a higher hydrogen to water ratio, the diffusion limitation disappeared. From the shape of the overpotential relations and the hydrogen mass flow dependence of the diffusion limitation it was concluded, that surface diffusion of atomic hydrogen to the reaction sites at the Ni/CCO/gas phase boundary followed by charge transfer is rate limiting for the overall anode reaction. A change in the charge transfer coefficient at 600 °C indicates a change in the reaction mechanism, however, in the anode reaction a variety of chemical species can be involved and the proposed reaction model might not be the only possible explanation. To clarify the reaction mechanism in detail, further studies are needed. A comparison of Ni-YSZ anodes on YSZ electrolyte to Ni-CCO anodes on ceria-based electrolytes revealed for the latter an exchange current density which was one order of magnitude higher.

8. References

1. H. Uchida, M. Yoshida, and M. Watanabe, in Solid Oxide Fuel Cells IV, M. Dokiya, O.

Yamamoto, H. Tagawa and S.C. Singhal, Eds., PV 95-1, p. 712, The Electrochem. Soc., Pennington NJ

(1994).

2. M. Mogensen, T. Lindegaard, and U.R. Hansen, G. Mogensen, J. Electrochem. Soc., 141,

2122 (1994).

3. B.C. Nguyen, T.A. Lin, and D.A. Mason, J. Electrochem. Soc., 133, 1807 (1986).

4. K. Eguchi, T. Setoguchi, T. Inoue and H. Arai, Solid State Ionics, 53 165 (1992).

5. D.W. Dees, T.D. Claar, T.E. Easler, D.C. Fee, F.C. Mrazek, J. Electrochem. Soc., 134, 2141

(1987).

6. T. Kawada, N. Sakai, H. Yokokawa, and M. Dokiya, J. Electrochem. Soc., 137, 3042

(1990).

7. T. Setoguchi, T. Inoue, H. Takebe, K. Eguchi, K. Morinaga, and H. Arai, Solid State

Ionics, 37, 217 (1990).

8. T. Setoguchi, K. Okamoto, K. Eguchi, and H. Arai, J. Electrochem. Soc., 139, 2875 (1992).

9. J. Van herle, Oxygen Reduction Reaction Mechanisms at Solid Oxide Fuel Cell Cathodes,

Ph.D. Thesis, EPF-Lausanne, Switzerland (1993).

10. C. Milliken, S. Elangovan, and A.C. Khandkar, in Solid Oxide Fuel Cells IV, M. Dokiya, O.

Yamamoto, H. Tagawa and S.C. Singhal, Eds., PV 95-1, p. 1049, The Electrochem. Soc., Pennington NJ

(1995).

11. M. Watanabe, H. Uchida, M. Shibata, N. Mochizuki, and K. Amikura, J. Electrochem.

Soc., 141, 342 (1994).

Anode 188

12. D. Schneider, Nichtstöchiometrie und Thermische Ausdehnung von Ceroxid

Membranen, Diplomarbeit, Nichtmet. Werkstoffe, ETH-Zürich, Switzerland (1995).

13. F.P.F. van Berkel, F.H. van Heuveln and J.P.P. Huijsmans, Solid State Ionics, 72, 240

(1994).

14. K.J. Vetter, Electrochemical Kinetics, Acad. Press, New York (1967), p. 117.

15. J. O'M Bockris and A.K.N. Reddy, Modern Electrochemistry, Vol. 2, Plenum Publishing

Corp., New York (1977), p. 991.

16. T. Kawada, N. Sakai, H. Yokokawa, M. Dokiya, M. Mori and T. Iwata, Solid State Ionics,

40/41, 402 (1990).

17. J. Guindet, C. Roux, and A. Hammou, in Proc. 2nd Intl. Symp. SOFC, p. 533, F. Grosz, S.C.

Singhal and O. Yamamoto, Eds., Commission of the Europ. Communities, Athens Greece (1991).

18. J. Mizusaki, H. Tagawa, T. Saito, T. Yamamure, K. Kamitani, K. Hirano, S. Ehara, T.

Takagi, T. Hikita, M. Ippomatsu, S. Nakagawa and K. Hashimoto, Solid State Ionics, 70/71, 52 (1994).

19. M. Mogensen, in Proc. 14th Riso Intl. Symp. Mat. Science, F.W. Poulsen, J.J. Bentzen, T.

Jacobsen, E. Skou and M.J.L. Ostergard, Eds., p. 117, Roskilde, Denmark (1993).

20. T. Kawada, N. Sakai, H. Yokokawa, M. Dokiya, M. Mori and T. Iwata, J. Electrochem.

Soc., 173, 3042 (1990).

21. M. Mogensen and T. Lindegaard, in Proc. 3rd Intl. Symp. SOFC, S.C. Singhal and H.

Iwahara, Eds., PV 93-4, p. 484, The Electrochem. Soc., Pennington NJ (1993).

22. K. Sasaki, J.-P. Wurth, M. Gödickemeier, A. Mitterdorfer and L.J. Gauckler, in Proc.1st

Europ. SOFC Forum, U. Bossel, Editor, p. 475, Lucerne (Switzerland) (1994).

23. G.W.C. Kaye, T.H. Laby, Tables of Physical and Chemical Constants, 14th Ed.,

Longman, 1971.

11. Engineering of Solid Oxide Fuel Cells with Mixed Ionic Electronic Conductors

Abstract

Solid oxide fuel cells that contain mixed ionic electronic conductors (MIECs) as electrolytes exhibit a maximum in efficiency at high ionic current density due to the relations between ionic and electronic current. This leads to restrictions in cell operation and cell design. Provided the necessary materials and cell design parameters are known, the optimum cell design can be obtained. The optimization of fuel cells with mixed ionic electronic conducting electrolytes is outlined for an example of a fuel cell with a mixed conducting ceria electrolyte. The cell operation and design parameter optimization regarded in this study includes operating range (V - I), MIEC thickness and electrode electrochemical parameters. Defect chemical considerations predict an opti-mum in MIEC thickness, depending on operating temperature and electrode properties.

1. Introduction

State-of-the-art solid oxide fuel cells are using zirconia electrolytes. They operate at temperatures of 900-1000 °C, due to the low ionic conductivity of the electrolyte. Lowering the operation temperature below 800 °C improves the lifetime of SOFC systems, and enables less expensive alloys to be used as current collectors [1]. CeO2 has been suggested as an alternative electrolyte to stabilized zirconia by several authors [234567891011-12]. Like stabilized zirconia it possesses the fluorite structure. Mobile oxygen vacancies are introduced by substituting ceria with di- or trivalent metal oxides such as CaO or rare earth oxides. The highest ionic conductivities were reported for a doping level of 20 mol% rare earth or 10 mol% of alkaline earth addition leading to 5% oxygen vacancies. At high oxygen partial pressures the material is reported to be a purely ionic conductor [3, 4]. At low oxygen partial pressures, as at the anode side of SOFCs, CeO2 becomes partially reduced. This leads to predominately electronic conductivity at low oxygen partial pressures as they prevail at the anode side of SOFCs. The influence of mixed conductivity in electrolytes on current-voltage characteristics, efficiency and performance has been analyzed before [1314-15]. However, these investigations were based on ideal fuel cell systems without the consideration of electrode overpotentials. Therefore, they deserve a further refinement considering non-ideal real operating conditions. The theory describing such non-ideal systems was derived in chapter 8.

191 Engineering

The present chapter elucidates engineering aspects of SOFC development with mixed ionic electronic conducting membranes as electrolytes. This includes the question of the useful operating range of such cells in terms of power output and efficiency as well as the optimization of the electrode/MIEC element in terms of electrode overpotential and MIEC thickness. Ceria-based ceramics were shown to exhibit different thermal expansion coefficients under oxidizing or reducing conditions (chapter 7). Therefore, it was studied whether upscaling of fuel cells from the laboratory scale (30 mm in diameter, 1cm2 electrode area) to prototype cells (70 mm in diameter, 30 cm2 electrode area) introduces mechanical failure of the electrode/MIEC elements. In section 2 a short summary of the theory given in chapter 8 is presented. Section 3 provides the necessary cell and system parameters used for the calculation of oxygen chemical potential profiles and efficiencies. The effect of upscaling from 1 cm2 to 30 cm2 electrode area is presented in section 4. In section 5, the efficiency of fuel cells with MIECs as electrolytes is interpreted by use of the theory given in section 2. Section 6 presents oxygen chemical potential profiles across the MIEC under fuel cell operating conditions. In section 7 the procedure for the determination of optimum cell design parameters is outlined. A summary is given in section 8. Note that although the term electrolyte is commonly used only for purely ionic conducting substances it is used here for mixed ionic electronic conductors also if it describes the structural element of a solid oxide fuel cell.

2. Ionic and electronic currents in fuel cells with mixed conducting electrolytes

2.1 Partial ionic and electronic currents

The current-voltage characteristics of SOFCs based on mixed ionic electronic conductors with non reversible electrodes have been derived in chapter 8. The equations are valid for a mixed conductor, which conducts ions and only one type of electronic charge carries (e.g. electrons) and whose electronic conductivity is induced by deviation from oxygen stoichiometry. For the partial ionic and electronic currents we can write [15, 16]:

IV MC V MC

Rith

i=

−( ) ( )

(11-1)

and

IV MC

Ree

= −( )

(11-2)

Engineering 192

Ii, the ionic current is expressed as a function of Vth(MC), V(MC) and Ri. Ri is the resistance to ionic current and is treated here as a constant due to doping of the electrolyte with lower valent cations (Sm or Gd). Vth(MC) is a virtual electromotive force given by the oxygen chemical potentials just inside the MIEC, just beyond a possible space charge region. V(MC) is the corresponding voltage drop across the MIEC. It is related to the measurable cell voltage by

V MC V I R RCell t C A( ) ( )= + + (11-3)

where RC and RA are the electrode resistances (in plane and cross plane) and It is the cell current It=Ii+Ie. In most cases, after all by the use of metallic electrodes or electrodes with a high electrical conductivity the electrode resistances can be neglected and V(MC)=VCell. The resistance of the MIEC to electronic current Re (Eq. (11-2)) is not a constant. Instead, it is a function of cell geometry, the voltage V(MC) on the MIEC, the applied oxygen partial pressures at the cathode and the anode, and the cathode and anode overpotentials. It is given by (chapter 8)

( )

( )

IS

LV V V V

p(Op(O

p(Oe

e

e th app Cell th C th A

ehigh

L

high

qV

q V V V V

Cell


= − − − −

−

−

−

− − − −

, , ,

( ) ))

) , , ,

δ δ

σβ

β δ δ22

2

14 1

1 (11-4a)

Here, Vth,app denotes the Nernst potential given by the applied outer atmospheres p(O2)high at the cathode and p(O2)low at the anode and S is the effectively active cross section of the electrodes.

p(O2)L is related to the cathode overpotential by p(O p(O eL high VqkT th C

2 24

) ) ,= ⋅ − δ

and Vth,app is related to Vth(MC) by

V MC V V Vth th app th C th A( ) , , ,= − −δ δ (11-5)

δVth,C and δVth,A are the overpotentials at cathode and anode, respectively. They correspond to a drop in oxygen chemical potential. For the ionic current one can write with Eq. (11-1), Eq. (11-5) and Vth(MC) = VCell :

IV V V V

Rith app Cell th C th A

i=

− − −, , ,δ δ

(11-6)

A schematic plot of the electronic current Ie versus cell voltage, the ionic current vs. cell voltage and the sum of Ii + Ie is given in Fig. 11-1.

193 Engineering

Fig. 11-1

Cell voltage versus electronic

current (left), cell voltage versus

ionic current (middle) and cell

voltage versus cell current (right).

VCell versus Ie

At open circuit voltage, the

electronic current is not zero. To-

wards lower cell voltages it

decreases and at short circuit

current it is zero .

VCell versus Ii At the theoretical electromotive

force Vth,app the Ii = 0. At VOC it is

equal to Ie.

VCell versus It

The total current is the sum of the

ionic and the electronic current.

Therefore, at open circuit voltage the

ionic and the electronic current are

equal

Engineering 194

2.2 Electrode overpotentials

Electrode overpotentials can be determined by either impedance spectroscopy or current interruption measurements, preferably using a 4-electrode setup with working and reference electrodes on either side of the electrolyte. From the preceding subsection 2.1 it is obvious (Eqs. (11-5) and (11-6)), that under open circuit conditions the electrodes and reference electrodes are already polarized to some extent, since It = Ii + Ie ≠ 0. Assuming that the polarization of the reference electrodes does not change with increasing It, the electrode overpotential measured by current interruption or impedance spectroscopy can be formulated as

η δC th CV C= −, (11-7)

η δA th AV A= −, (11-8)

where ηC is the measured voltage drop and δVth,C the overpotential for the cathode. C is a constant reflecting the polarization of the cathode under open circuit conditions in a MIEC (ηA, δVth,A and A for the anode). Further, we will apply an interpretation of the electrode overpotentials based on the assumption, that the total electrode overpotential can be expressed by the sum of an overpotential due to charge transfer reactions and diffusion polarizations. Generally, the charge transfer reaction's characteristic is a decreasing interfacial resistance with increasing electrode overpotential. The characteristic of a diffusion polarization is an increase of electrode resistance with increasing Ii resulting in a limiting current at high current densities. Therefore the cathode overpotential is given as

( )δα

VkT

qp(O n p(O

kTq

IIth C

high high C IC

i

C

nn i,

,ln( ) ln ) arsinh= −

+

−

4 2 22 20

1

(11-9)

where the Butler-Volmer equation is approximated with a arsinh-function. The first term of Eq. (11-9) reflects possible diffusion limitations due to molecular oxygen (n = 1, C1) or atomic oxygen (n = 2, C2). The cathode can therefore be described by the parameters αC (charge transfer coefficient) I0,C (exchange current density) and n and Cn For the anode a similar equation can be formulated, where again charge transfer and diffusion limitation due to limited supply of molecular or atomic hydrogen is assumed. Eq. (11-10) describes the anode overpotential as the sum of diffusion overpotentials due to diffusion of atomic hydrogen (n = 2, A2) or molecular hydrogen (n = 1, A1).

195 Engineering

( )δ

αV

kTq

p(H

p(H A I

kTq

IIth A

ext

extn i

nA

i

An,

,ln

)

)

arsinh=

−

+

2 2 22

201

(11-10)

The anode can be described by the parameters αA , I0,A , n , and An . In Eq. (11-10) p(H2)ext is the hydrogen and water partial pressures in the fuel atmosphere just outside the anode. 3. Electrical and electrochemical cell parameters For the full description of a fuel cell with a mixed ionic electronic conductor as electrolyte with Eqs. (11-4) (11-6) and (11-7) to (11-9), the following materials and cell parameters are necessary: A) The dependence of the ionic and electronic conductivity of the electrolyte on the oxygen partial pressure. B) The electrochemical parameters cathode and anode overpotential δVth,C and δVth,A, as function of the ionic currant density. These are described by the parameters n , Cn , An , I0,C , I0,A , αC , and αA. C) The fuel cell dimensions , such as L, the electrolyte thickness and S the effective electrode cross section (which might be smaller than the apparent cross section due to the electrode microstructure). D) The fuel cell operating parameters p(O2)high on the cathode and p(H2)ext and p(H2O)ext the fuel gas composition. In this chapter we describe the fuel cell engineering of a fuel cell with a MIEC as electrolyte using an example of a fuel cell with a samaria doped ceria electrolyte of the composition Ce0.8Sm0.2O1.9-x (CSO) where x is the oxygen nonstoichiometry of the MIEC depending on p(O2). As cathode a La0.84Sr0.16CoO3 (LSC) cathode prepared by screen-printing is applied. The anode is a nickel cermet with calcia doped ceria as oxide component. Its composition is Ni-Ce0.9Ca0.1O1.9-x, (NCC) with a volume content of 54% Ni. Air is supplied to the cathode and a mixture of argon, hydrogen and water is supplied to the anode (H2 : H2O = 10 : 3). As current collectors from the electrodes Pt-mesh was used. A) The dependence of the electrical conductivity of the electrolyte on the oxygen partial pressure can be derived from electrical conductivity measurements as a function of oxygen partial pressure as shown in Fig. 11-2 for CSO at 600, 700 and 800 °C. At high oxygen partial pressures, the total electrical conductivity, σt , is predominately ionic (σt ≈ σi) whereas at low p(O2) the conductivity is predominantly electronic (σt ≈ σe). The electronic conductivity is

Engineering 196

proportional to p(O2)-1/4. The relevant electrolyte parameters for the calculation of the I - V characteristics following Eqs. (11-4) and (11-6) are summarized in Tab. 11-1.

30 25 20 15 10 5 01

10

30

-log(pO2 / atm)

σto

t [S

/m]

Ce0.8Sm0.2O1.9-x

800 °C

700 °C

600 °C

Fig. 11-2 Total electrical conductivity of CSO at 600, 700 and 800 °C as a function of oxygen

partial pressure. Solid lines show the dependence of σt = σi + k .p(O2)-1/4.

T

[°C]

σi

[S/m]

σe in air

[S/m]

dependence of

log(σe) on log (p(O2)

EDB

[atm]

600 1.65 3.55.10-6 -1/4 4.48.10-24

700 3.47 5.61.10-5 -1/4 1.48.10-20

800 7.2 8.86.10-4 -1/4 4.82.10-17

Tab. 11-1 Ionic conductivity, electronic conductivity at p(O2)high = 0.21 atm (air), dependence of

electronic conductivity on oxygen partial pressure and the electrical domain boundary

(EDB), where the electronic conductivity equals the ionic conductivity for CSO at 600,

700, and 800 °C.

B) The electrochemical parameters of the applied LSC cathode and NCC anode are shown in Fig. 11-3 (cathode) and Fig. 11-4 (anode).

197 Engineering

0.0 0.2 0.4 0.6 0.8 1.0 1.20.00

0.04

0.08

0.12

0.16

0.20

0.24

800 °C

700 °C

600 °Cη C

+ C

[

V]

ji [A/cm2]

Fig. 11-3 Cathode overpotential as a function of ionic current density at 600, 700, and 800 °C in

air. The lines are a fit to Eq. (11-9) with δVth,C = ηC + C.

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

0.3

η A +

A

[V]

800 °C

700 °C

600 °C

ji [A/cm2]

Fig. 11-4 Anode overpotential as a function of ionic current density at 600, 700, and 800 °C in

air. The lines are a fit to Eq. (11-10) with δVth,A = ηA + A.

Engineering 198

The current overpotential characteristics of Fig. 11-2 and 11-3 are derived from current interruption measurements of the electrodes against reference electrodes of the same composition on either side of the electrolyte. The effective overpotential is derived from the measured one following Eqs. (11-7) and (11-8). The LSC cathode shows a decreasing electrode resistance with increasing ionic current density. A least square fit to Eq. (11-9) revealed pure charge transfer overpotential, no diffusion limitation was detected. The NCC anode shows a charge transfer governed characteristic at low ionic current densities and a diffusion limitation at high current densities. A least square fit to Eq. (11-10) revealed diffusion limitation due to an atomic hydrogen species (H or H+). The electrochemical parameters of the LSC cathode presented in Fig. 11-2 and the electrochemical parameters of the NCC anode presented in Fig. 11-3 are summarized in Tab. 11-2. The effective electrode cross section S is also included.

T [°C]

αC J0,C [A/cm2]

S

C

[V]

n

LSC 600 0.26 0.094 1 0.019 -

700 0.45 0.18 1 0.031 -

800 0.65 0.55 0.8 0.03 -

T [°C]

αA J0,A [A/cm2]

S

A [V]

n A2 [cm2atm1/2/A]

NCC 600 0.23 0.1 1 0.043 2 0.458

700 1 0.3 1 0.034 2 0.318

800 1 0.6 0.8 0.03 2 0.049

Tab. 11-2 Charge transfer coefficient and exchange current density, effective electrode cross-

section, initial overpotential, and parameters for diffusion limitation for the applied LSC

cathode and NCC anode.

C) The fuel cell dimensions used for the further calculations and considerations are an apparent electrode area of 1 cm2 and an electrolyte thickness of 239 µm. For the prototype cell an electrode area of 30 cm2 and an electrolyte thickness of 250 µm was used. The calculations as function of the electrolyte thickness in section 7 base on the electrode and electrolyte properties presented in this section.

199 Engineering

D) The fuel cells were operated with air on the cathode side. A mixture of 87.3% Ar + 9.7% H2 + 3% H2O was used as fuel unless stated otherwise. The fuel utilization was kept very low (< 5%) to fix the fuel atmosphere at the anode side. Only for the prototype cell, a higher fuel utilization was used. The fuel cells were tested at 600, 700, and 800 °C.

4. Upscaling to prototype size

Nonstoichiometric oxides such as ceria or lanthanum chromite exhibit an increase of the lattice constant at low oxygen partial pressures [17, 18] and hence a different expansion on the cathode side and the anode side of a SOFC. A rapid upscaling already during the development process of new materials is therefore desirable, to investigate whether the desired fuel cell design and configuration can withstand the mechanical stresses due to the expansion behavior. Normally, laboratory scale SOFC have a size of 10 to 30 mm in diameter. Their electrode area ranges from 0.2 to 3 cm2. As current collector from the electrodes Pt-meshes are applies generally. This avoids difficulties due to gas distribution and ensures a homogeneous current collection. In real SOFC systems, however Pt-mesh would be too expensive as current collector and therefore, either La-chromite, superalloys or electrically conducting ceramic foams are used as current collectors. To ensure the compatibility of the laboratory scale cell design with industrial cell design a small prototype cell with 70 mm in diameter and an electrode area of 30 cm2 was tested in the Sulzer HEXIS SOFC environment [19]. As current collector an electrically conducting ceramic foam was used on the cathode side and Ni-mesh / Ni-felt was used on the anode side. Air was fed to the cathode side and H2 + 3% H2O was fed to the anode side. The electrolyte thickness was 250 µm. The electrodes and the electrolyte were the same as presented in section 3. A schematic picture of the setup for prototype testing compared to laboratory scale test setup is shown in Fig. 11-5, and the test stand with the 70 mm cell is shown in Fig. 11-6.

Engineering 200

Fig. 11-5 Schematic setup of a prototype cell and a laboratory scale test cell with cathode,

electrolyte, anode and current collectors.

Fig. 11-6 Photograph of the prototype cell with 70 mm in diameter and an effective electrode

area of 30 cm2.

201 Engineering

0.0 0.3 0.6 0.9 1.20.0

0.2

0.4

0.6

0.8

1.0

p [W/cm

2]700 °C

800 °C

Jt [A/cm2]

V Cel

l [V]

0.0

0.1

0.2

0.3

0.4

0.5

Fig. 11-7 Current - voltage characteristic of a laboratory scale test cell (1 cm2 active area):

air / LSN / CSO / NCC / (87.3% Ar + 9.7% H2 +3% H2O).

0.0 0.1 0.2 0.3 0.4 0.50.0

0.2

0.4

0.6

0.8

Jt [A/cm2]

V Cel

l [V

]p [W

/cm2]

755 °C

715 °C

715 °C

0.00

0.05

0.10

0.15

0.20

0.25

Fig. 11-8 Current - voltage characteristic of a prototype cell (30 cm2 active area):

air / LSN / CSO / NCC / (97% H2 +3% H2O).

Engineering 202

The power output and current - voltage characteristic of a laboratory scale test cell with Pt-mesh current collector is shown in Fig. 11-7. The cell was operated with a mixture of 90% Ar and 10% H2,saturated with 3% of water. The mass flow of hydrogen was 0.54 g/h corresponding to a fuel utilization of about 3.7%. This cell shows a maximum power output of 0.24 W/cm2 at 700 °C. No attempt was made to raise the fuel utilization in order to minimize diffusion limitations at the anode. The prototype cell was driven with hydrogen saturated with water (3%) at a hydrogen mass flow of 1.55 g/h. This corresponds to a fuel utilization of 30%. The maximum power output of this cell was 0.18 W/cm2 at 715 °C (Fig. 11-8) which is only about 20% lower than the output of the laboratory scale cell of Fig. 11-7. At 755 °C this cell shows a power output of 0.22 W/cm2.

5. Efficiency of fuel cells with mixed conductors

The in-cell efficiency of a fuel cell can be given as the voltage produced by the cell VCell under operating conditions It > 0 divided by the theoretical value. This theoretical value is the electromotive force Vth,app of the cell given by the applied chemical potential at cathode and anode, i.e. zFVth,app = -∆GR(T) . The so-defined efficiency is the voltage efficiency εV of the cell and is given by

εVCell Cell

R

VV

zFVG T

= =−th,app ∆ ( )

(11-11)

where z is the number of electrons transferred in the reaction, F is the Faraday constant and ∆GR(T) is the Gibbs free enthalpy of reaction at the operating temperature given by the inlet concentration of reactants to the fuel cell. To compare the efficiency of different fuel cell types operating at different temperatures it might be useful to define the maximum efficiency in terms of the enthalpy of the considered overall fuel cell reaction. It is given as the ratio ∆GR(T)/∆H°, where ∆H° is the enthalpy of reaction at 298 K. This is the maximum available efficiency εmax for the considered fuel cell reaction. For example

H O H O l212 2 2+ = ( ) (11-12a)

or

H O H O g212 2 2+ = ( ) (11-12b)

Eqs. (11-12a) and (11-12b) are different with respect to whether the reaction product is liquid (11-12a) or gaseous (11-12b). For the reaction of Eq. (11-12a) ∆H°a = -286 kJ/mol (called

203 Engineering

HHV for higher heating value) and for Eq. (11-12b) ∆H°b = -242 kJ/mol (LHV, lower heating value) [20]. The maximum efficiency is therefore defined by

ε εmax max( )( )

( )( )

HHVG TH

or LHVG TH

R

a

R

b=

°=

°∆∆

∆∆ (11-13)

Additionally to εV and εmax the fuel utilization uf , i.e. the ratio between fuel fed to the system and reacted fuel coming from the system has to be considered. For fuel cells with purely ionic conducting solid electrolytes, the overall efficiency εtot with respect to the enthalpy of combustion of the overall cell reaction is given by

ε ε εtot V fCell

fuzFV

Hu= ⋅ ⋅ =

− °⋅max ∆ (11-14)

For fuel cells with MIECs as electrolytes, the total efficiency has to be modified by an additional term, the faradaic or current efficiency εf . As shown in chapter 8, the total cell current It is the sum of the ionic current Ii and the electronic current Ie, hence fuel is consumed proportional to Ii , but useful electric energy is only delivered by It . Since It is always smaller than Ii for the operating range of a SOFC the faradaic efficiency is given by

ε ft

i

e

i

II

II

= = +1 (11-15)

In Eq. (11-15) use was made of the relation It = Ii + Ie. At open circuit voltage (VOC) -Ie = Ii and the faradaic efficiency is zero, whereas at VCell = 0 Ie = 0 and the faradaic efficiency is maximal. The overall efficiency of a fuel cell with a MIEC as electrolyte is therefore given as

ε ε ε εtot V f fCell t

ifMIEC u

zFVH

II

u( ) max= ⋅ ⋅ ⋅ =− °

⋅ ⋅∆ (11-16)

and the efficiency with respect to ∆GR(T)is given by

ε ε ε∆ ∆G V fCell

R

t

i

Cell t

i

zFVG T

II

V IV I

= ⋅ =−

⋅ =⋅⋅( ) th, app (11-17)

The following calculations and consideration are based on this efficiency ε∆G . ε∆G can be calculated from Eqs. (11-4), (11-6) and (11-17). It exhibits a maximum at a total cell current density corresponding to about 1/3 of the short circuit current density. This maximum results from the finding, that εV decreases with increasing total cell current density, whereas εJ increases. The voltage efficiency, the faradaic efficiency and ε∆G are shown in Fig. 11-9 for a fuel cell LSC/CSO/NCC at 700 °C.

Engineering 204

0.0 0.2 0.4 0.6 0.80.0

0.2

0.4

0.6

0.8

1.0

ε∆G

εj

εV

Jt [A/cm2]

ε

Fig. 11-9 Efficiency of a fuel cell LSC/CSO/NCC at 700 °C versus total cell current Jt. εV is given

as VCell/Vth,app; ε j is given as Jt/Ji (Eqs. (11-4) and (11-6)) and ε∆G is the product of ε j

and εV (Eq. 11-17).

Whereas the maximum efficiency of the fuel cell is at about 1/3 of the maximum current density, the maximum power output is close to 1/2 of the short circuit current density. Therefore, the useful operating range of fuel cells with MIECs as electrolytes is restricted to the range in between the maximum in power output and the maximum in efficiency. The efficiency with respect to ∆GR(T) and the power output are shown in Figs. 11-10 to 11-12 for the fuel cell presented in section 3 at 600, 700 and 800 °C. The maximum efficiency ε∆G(max) of these cells is 0.5 at 800 °C, 0.57 at 700 °C and 0.6 at 600 °C, whereas the maximum power output is 0.45 W/cm2 at 800 °C, 0.24 W/cm2 at 700 °C and only 0.09 W/cm2 at 600 °C. For useful technical application a power output of at least 0.15 W/cm2 is necessary. To fulfill this requirement, this cell has to be operated at around 700 °C.

205 Engineering

0.0 0.2 0.4 0.6 0.8 1.00.0

0.1

0.2

0.3

0.4

0.5p [W

/cm2]ε∆G

800 °C

Jt [A/cm2]

0.0

0.1

0.2

0.3

0.4

0.5

Fig. 11-10 Efficiency and power output of a fuel cell with an electrolyte thickness of 250 µm:

800 °C, air / LSN / CSO / NCC / (87.3% Ar + 9.7% H2 +3% H2O).

0.0 0.2 0.4 0.6 0.80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

p [W/cm

2]

ε∆G

700 °C

Jt [A/cm2]

0.00

0.05

0.10

0.15

0.20

0.25



Engineering 206

0.0 0.1 0.2 0.3 0.40.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7p [W

/cm2]

ε∆G

600 °C

Jt [A/cm2]

0.00

0.02

0.04

0.06

0.08

0.10



6. Oxygen chemical potential across the MIEC

According to [21] the variation of the oxygen chemical potential across the MIEC used as electrolyte here can be expressed for the assumption of reversible electrodes as

( )( ) ( )µµ

( )( )( )( )

lnexp ( )

exp ( )exp

,

,,

O xO

kTV V

xL

V VV

qkT th app Cell

qkT th app Cell

qkT th app

2

2 04 1

1

11= − −

− − −

− − −− −

(11-18)

Where µ(O2)(x) is the oxygen chemical potential at the location x across the MIEC, is the oxygen chemical potential at the anode (x = 0) and x/L is the dimensionless coordinate. The cathodes is at x = L. For the case of non-reversible electrodes Eq. (11-18) has to be modified, since electrode overpotentials lead to a change in µ(O2) at the interfaces just inside the MIEC. This jump in oxygen chemical potential is located within the electrochemical double layers at cathode and

207 Engineering

anode. According to chapter 8 the variation of µ(O2) across the MIEC can be given for non-reversible electrodes, provided the overpotentials are known as:

( )µµ

ββ

ββδ( )( )

( )( )( )

( ( ) ( ))

( ( ) ( )),O x

OkT e

e

eeq V MC

q V MC V MC

q V MC V MCq Vth

thxL

th

th A2

2 04 1 1

1

1= − − −

−

−

−− −

− −−

(11-19)

where β = 1/kT. Eq. (11-19) describes the variation of the oxygen chemical potential where µ(O2)(0) is the oxygen chemical potential in the fuel atmosphere. In Fig. 11-13 to 11-15 the variation of the oxygen chemical potential versus dimensionless coordinate x/L is presented as log(p(O2)) (with µ(O2) = kT ln(p(O2))) using the electrode overpotential of a fuel cell with CSO electrolyte with a thickness of 239 µm, LSC cathode and NCC anode. The electrolyte parameters used for the calculations are from Tab. 11-1. Small numbers in the plot indicate the corresponding cell voltage VCell in [V]. The dotted line indicates the location of the electrolytic domain boundary (EDB) in the MIEC. In Fig. 11-13 the oxygen chemical potential is plotted as log(p(O2)) vs. x/L at 800 °C. The cell voltage corresponding to the maximum in efficiency (see Fig. 11-10) is 0.614 V. This corresponds to ~60% of the MIEC being predominantly ionic conducting (arrow in Fig. 11-12) under these operating conditions. At 700 °C, the maximum efficiency is at a voltage VCell = 0.656 V, ~75% of the MIEC is predominantly ionically conducting. At 600 °C, the EDB is located in the electrical double layer at the anode side and the cell voltage corresponding to the maximum efficiency is 0.698 V. Under the given operating conditions the location of the EDB is shifted towards the anode side, i.e. the MIEC is gradually shifted to more ionic than electronic conductivity. Therefore, a lower operating temperature is favorable from this point of view.

Engineering 208

0.0 0.2 0.4 0.6 0.8 1.0-20

-15

-10

-5

0

log(

p(O 2

) / a

tm)

EDB

x/L

0.445 0.6140.793

Fig. 11-13 Variation of ∆µ(O2) expressed as log(p(O2) across a CSO electrolyte at 800 °C.

Numbers at the curves indicate the corresponding VCell in volts. The EDB is located at

x/L = 0.38. The cathode is at x/L = 1, the anode at x/L = 0.

0.0 0.2 0.4 0.6 0.8 1.0-25

-20

-15

-10

-5

0

log(

p(O 2

) / a

tm)

EDB

x/L

0

0.6560.864

Fig. 11-14 Variation of ∆µ(O2) expressed as log(p(O2) across a CSO electrolyte at 700 °C. The

EDB is located at x/L = 0.24. Numbers indicate the corresponding cell voltage in volts.

209 Engineering

0.0 0.2 0.4 0.6 0.8 1.0-25

-20

-15

-10

-5

0

log(

p(O 2)

/ at

m)

EDB

x/L

0.001

0.698

0.931

Fig. 11-15 Variation of ∆µ(O2) expressed as log(p(O2) across a CSO electrolyte at 600 °C. The

EDB is located in the electrochemical double layer at the anode side. Numbers

indicate the corresponding cell voltage in volts.

7. Cell design and operating conditions

As shown in the preceding section, fuel cells based on MIEC as electrolytes exhibit maximum efficiency and power output within a narrow operating range. To lower the necessary operating temperature of fuel cells based on purely ionic conducting solid electrolytes (e.g. ZrO2), it is useful to minimize the electrolyte thickness. By this the electrolyte resistance can be minimized and the power output rises. For MIECs a minimization of the electrolyte thickness would also lead to an increase of the ionic current density. However, by combining Eqs. (11-4), (11-6) and It = Ie + Ii one can see, that the electronic current density at open circuit voltage would increase also. This is calculated for reversible electrodes (δVth,C = δVth,A = 0) in Fig. 11-16 at 600, 700 and 800 °C using the parameters in Tab. 11-1 and Eqs. (11-4) and (11-6). For reversible electrodes this would only affect the maximum power output, but not the maximum efficiency. However since under fuel cell operating conditions the ionic current is always larger then the ionic current at VOC (Ii > Ii(VOC)) a down-thinning of the MIEC

Engineering 210

requires the handling of large current densities. For a CSO fuel cell with an electrolyte thickness of 5 µm the ionic current would always exceed ~10 A/cm2.

10 100 500

10-1

100

101

J i (V

OC) [

A/c

m2 ]

reversible electrodes

800 °C

700 °C

600 °C


Fig. 11-16 Ionic current density vs. MIEC thickness calculated for CSO by the use of the

parameters in Tab. 11-1 and Eqs. (11-4) and (11-6). Reversible electrodes are

assumed here.

For non-reversible electrodes on the other hand, the overpotential depends on the ionic current density. Therefore, a high ionic current at VOC would lead to strong decay in efficiency since for the higher ionic current densities also the overpotential increases. This leads to a decrease in oxygen chemical potential at the cathode side and to an increase in oxygen chemical potential at the anode side of the MIEC and, hence, to a lower Vth(MC) the virtual electromotive force just inside the MIEC which is the driving force for the ionic current. In Fig. 11-17, the ionic current density at VOC of a fuel cell with a CSO electrolyte is calculated assuming real electrodes, i.e. including electrode overpotential into the calculation. For this calculation the electrochemical parameters of the LSC cathode and NCC anode presented in Tabs. 11-1 and 11-2 were used as well as Eqs. (11-4), (11-6), (11-9), and (11-10). The ionic current density at open circuit voltage as presented in Fig. 11-17 is much lower than for the case of reversible electrodes.

211 Engineering

10 100 500

10-1

100

101J i

(VO

C)

[A/c

m2 ]

real electrodes

800 °C

700 °C

600 °C


Fig. 11-17 Ionic current density vs. MIEC thickness calculated for CSO with a LSC cathode and a

NCC anode. The parameters for the calculation are from Tab. 11-1 and 11-2.

Eqs. (11-4), (11-6), (11-9) and (11-10) were used.

From these considerations it is obvious, that the efficiency of these fuel cells decreases with decreasing electrolyte thickness. For fuel cell design optimization with MIECs as electrolytes in terms of efficiency, power output and operating temperature the following procedure is suggested: 1. Determine the electrical conductivity of the MIEC as a function of oxygen partial pressure. 2. Determine the electrochemical properties of the electrodes over the desired operating temperature range by current interruption or impedance measurements. This can be done on electrolytes with any thickness, provided it is possible to reach useful current densities. To determine the electrochemical parameters Eqs. (11-4), (11-6), (11-9), and (11-10) are used. 3. Calculate the power output at maximum efficiency, pmax(ε∆G) as a function of MIEC thickness (Eqs. (11-4), (11-6, (11-9), (11-10), and (11-17). 4. Define the desired MIEC thickness (Fig. 11-18), the desired pmax(ε∆G) (Fig. 11-19) or the desired ε∆G(max) (Fig. 11-20) and determine the other cell design parameters.

Engineering 212

An example for such a cell design optimization is shown in Figs. 11-18 to 11-20, where the efficiency and the corresponding power output are shown as a function of MIEC thickness at 600, 700 and 800 °C. For a desired MIEC thickness Fig. 11-18 yields a ε∆G(max) of 0.43 and a pmax(ε∆G) of 0.34 W/cm2. On the other hand, a pmax(ε∆G) of 0.2 W/cm2 (Fig. 11-19 (700 °C)) yields a MIEC thickness of 250 µm and a ε∆G(max) of 0.57. For a ε∆G(max) of 0.5 the results in Fig. 11-20 for 600 °C yield a MIEC thickness of 45 µm and a pmax(ε∆G) of 0.175 W/cm2. As already mentioned, these consideration have to be made for every MIEC/electrodes combination under the desired fuel and cathode atmospheres.

10 100 5000.0

0.1

0.2

0.3

0.4

0.5

0.6

ε ∆G(m

ax) o

r

pm

ax(ε

∆G)

[W/c

m2 ]

pmax(ε∆G)

ε∆G(max)

800 °C


Fig. 11-18 Maximum efficiency with respect to ∆G and power output at this efficiency maximum

as a function of MIEC thickness at 800 °C. Arrows see text.

213 Engineering

10 100 5000.0

0.1

0.2

0.3

0.4

0.5

0.6

pmax(ε∆G)

ε∆G(max)ε ∆G

(max

) or

pm

ax(ε

∆G)

[W/c

m2 ]

700 °C




10 100 5000.0

0.1

0.2

0.3

0.4

0.5

0.6

pmax(ε∆G)

ε∆G(max)

ε ∆G(m

ax) o

r

pm

ax(ε

∆G)

[W/c

m2 ]

600 °C




Engineering 214

For a fuel cell with reversible electrodes, it is obvious, that the efficiency is not a function of the MIEC thickness. This can be verified by combining Eq. (11-6) with Ri = L/σi, Eq (11-4) and It = Ii + Ie and substituting of these terms in Eq. (11-17). Then the MIEC thickness L vanishes. However, the electrode overpotential is a function of the ionic current density which increases for a thin MIEC thickness. Fig. 11-21 shows the maximum efficiency as a function of MIEC thickness at 700 °C for reversible electrodes, for good electrodes (LSC + NCC) given in Table 11-2, and for bad electrodes. As bad electrodes we consider an Au cathode and an Au anode. The Au cathode exhibits an exchange current density I0,C of 0.001 A/cm2 and a transfer coefficient αC of 0.55. The Au anode has an αA of 1 and an I0,A of 4.86.10-9 A/cm2. As expected, the maximum efficiency is independent of the MIEC thickness for reversible electrodes. For non ideal electrodes the maximum efficiency is a function of the MIEC thickness. The maximum attainable efficiency is varying in a wide range for the same fuel and oxidant atmospheres and for the same MIEC thickness.

10 100 5000.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8


LSC/Ni-CCO

Gold

reversible electrodes

ε ∆G(m

ax)

700 °C

Fig. 11-21 Maximum efficiency of ceria based fuel cells as a function of MIEC thickness for

reversible electrodes and for electrodes with high performance (LSC/NCC) and for

electrodes with a low performance (Au).


215 Engineering

Fuel cells with mixed ionic electronic conductors as electrolyte exhibit a maximum in efficiency at fuel cell current densities corresponding to ~1/3 of the short circuit current density. The maximum in power output for these cells is at ~1/2 of the short circuit current density. This somewhat unexpected behavior is explained by variation of the ionic and electronic currents in such MIECs. At high cell voltages and low total cell currents the ionic and the electronic currents are similar, however flowing in opposite direction. Driving the cell in this range results in a very low faradaic efficiency. Driving the cell at very high total current densities results in a low cell voltage and hence in a low voltage efficiency. For these reasons, fuel cells with mixed ionic electronic conductivity should be operated between the maximum in efficiency and the maximum in power output. Lowering of the cell operating temperature results in lower electronic conductivity of the MIEC. However, at these operating temperatures the MIEC ionic resistance becomes too high for fuel cell operation. An obvious method to decrease the electrolyte resistance would be to make it thinner, as it is practicable in conventional zirconia based fuel cells. However, for fuel cells with MIECs as electrolytes it can be shown that this results in low efficiencies due to the non-reversibility of real electrodes. Therefore, for every electrode overpotential - ionic current relation and every MIEC operating temperature and thickness, there exists an optimum between power output and efficiency. Thinner electrolytes are not necessarily the best choice. A procedure for designing fuel cells to yield an optimum efficiency and power output is suggested.

9. References

1. H. Altdorfer, K. Honegger, in Proc. 1st Swiss Conference on Materials Research for

Engineering Systems, B. Ilschner, M. Hofmann and F. Meyer-Olbersleben, Eds., p.56, Technische

Rundschau, Sion, Switzerland (1994).

2. D. Singman, J. Electrochem. Soc., 113, 502 (1966).

3. R.N. Blumenthal, F.S. Brugner, and J.E. Garnier, J. Electrochem. Soc., 120, 1230 (1973).

4. H.L. Tuller and A.S. Nowick, J. Electrochem. Soc., 122, 255 (1975).

5. T. Kudo and Y. Obayashi, J. Electrochem. Soc., 123, 415 (1976).

6. R.T. Dirstine, R.N. Blumenthal and T.F. Kuech, J. Electrochem. Soc., 126, 264 (1979).

7. M.D. Hurley and D.K. Hohnke, J. Phys. Chem. Solids, 41, 1349 (1980).


9. D.L. Maricle, T.E. Swarr and S. Karavolis, Solid State Ionics, 52, 173 (1992).

10. C. Milliken, S. Elangovan and A.C. Khandkar, in Solid Oxide Fuel Cells IV, M. Dokiya, O.

Yamamoto, H. Tagawa and S.C. Singhal, Eds., p. 1049, PV 95-1, The Electrochem. Soc. Inc., Pennington NJ,

USA (1995).

Engineering 216

11. F.P.F. van Berkel, G.M. Christie, F.H. van Heuveln and J.P.P. Huijsmans, ibid., p. 1062.

12. M. Gödickemeier, K. Sasaki and L.J. Gauckler, ibid., p. 1072.


14. P.N. Ross and T.G. Benjamin, J. Power Sources, 1, 311 (1976/77).


16. I. Riess, Solid State Ionics, 52, 127 (1992).


2122 (1994).

18. P.V. Hendriksen, J.D. Carter and M. Mogensen, in Solid Oxide Fuel Cells IV, M. Dokiya,

O. Yamamoto, H. Tagawa and S.C. Singhal, Editors, PV 95-1, p. 934, The Electrochem. Soc., Pennington

(NJ) (1995).

19. R. Diethelm and K. Honegger, in Proc. 3rd Internat. Symp. SOFC, S.C. Singhal and H.

Iwahara, Eds., PV 93-4, p. 822, The Electrochem. Soc., Pennington NJ, USA (1993).

20. E. Barendrecht, in Fuel Cell Systems, L.J. Blomen and M.N. Mugerwa, Eds., p. 75, Plenum

Press New York, USA (1993).

21. I. Riess, J. Phys. Chem. Solids, 47, 129 (1986).

Further Work Intermediate temperature solid oxide fuel cells are receiving intensive attention all over the world. Although many publications on this topic have been presented in the past three years, still very little is known about the mechanisms of the electrochemical reactions at the electrode/electrolyte interfaces. Also the performance of these systems has to be improved to allow commercialization. On the base of investigations of ECN in the Netherlands, on the work of Imperial college in London, UK, and on the present work, a large European research program is to be launched aiming at a 1 kW SOFC with ceria-gadolinia electrolytes. Several important aspects of ceramic fuel cell research and engineering are listed below. Cathodes with improved electrochemical activity. The intermediate temperature operation of SOFC imposes requirements on the cathode material, that can not be fulfilled by the presently applied ones. For the intermediate temperature range perovskite cathodes with the compositions La(1-x)SrxCo(1-y)NiyO3, La(1-x)SrxCo(1-y)FeyO3 or La(1-x)CaxCo(1-y)FeyO3

have been proposed [1-234]. However, still very little is known on their performance at intermediate temperatures and their chemical compatibility to other fuel cell components. These materials are mixed ionic electronic conductors like La(1-x)SrxCoO3 presented in chapter 9. This fact seems to be important for the improvement of the effective active cross sectional area of the cathode and for the oxygen reduction reaction. For improving the cathode performance it is necessary to identify the rate determining electrochemical steps. It is suggested to perform impedance spectroscopic measurements as a function of oxygen partial pressure under fuel cell operating conditions to further investigate the reaction mechanisms. Additionally, new cathode materials, e.g. in the system rare earth (La, Pr or Gd) - alkaline earth (Ca or Sr) - transition metals (Mn, Co, Fe) with still higher rates for the oxygen reduction reaction can be expected. For example perovskite systems (Pr,Sr)(Mn,Co)O3 or (Gd,Ca,Sr)(Mn,Co)O3 have been suggested [5, 6]. Anodes for direct hydrocarbon oxidation. For a commercialization of natural gas fueled SOFCs it is necessary to reduce overall system costs. Therefore, additional system components such as, e.g. an external reformer have to be avoided. From this a need for either internal reforming or direct hydrocarbon oxidation at the anode is obvious. To achieve this, I see two possible strategies: One is, to allow direct hydrocarbon oxidation at the anode. For this purpose, new anode materials have to be found, since the currently applied Ni-cermet acts itself as a hydrocarbon cracking catalyst. This results in carbon being readily deposited in the pores of the anode. As possible alternative materials carbides, e.g. WC has been proposed recently [7].

Further work 218

The other strategy is internal reforming. This should not be performed directly on the cell, since the endothermic reforming reaction would badly influence the thermal balance of the system, but in a separated yet thermally integrated reformer in the fuel gas stream. To compensate for the endothermic reduction reaction, partial oxidation of the fuel is suggested. This partial oxidation delivers the thermal energy needed for the reforming reaction and eliminates the otherwise necessary steam addition to the fuel. Such investigations are currently starting in most intermediate temperature research programs, since it obviously has been realized that there is no need for anodes for the laboratory, but for real fuel cell systems. New fuel cell design with thin film electrolytes. A promising way of decreasing the resistance of fuel cell elements is to reduce the thickness of the electrolyte. However, it has been shown in the present work, that this is only promising for pure solid ionic conductors. Reducing the thickness in case of MIECs as electrolytes would result in a breakdown of the cell voltage. Only at low temperatures – around 500 °C – application of thin film ceria electrolytes seems to be possible. In case thinner electrolytes are used, these can no more be the load bearing structural element of a single cell. Therefore, substrates have to be developed allowing a deposition of the thin electrolyte. This substrate can be either the cathode or the anode. For reasons of gas permeability the structure has to be porous and for the thin film deposition it has to be dense at one surface. The most promising methods for the electrolyte deposition are magnetron sputtering, electrochemical vapor deposition or electrophoretic deposition of fine ceramic particles. Currently a new research program has been initiated in our laboratory dealing with the development of such a thin film cell design. Oxygen ion conductors. For almost 20 years it seemed that there is no possibility to further improve the ionic conductivity of the existing oxygen ionic conductors. Only recently, however, a Japanese group presented a new ionic conductor with exceptional ionic conductivity. This material Mg and Sr doped LaGaO3 has the perovskite structure and exhibits ionic conductivities of 10 S/m already at 700 °C [8, 9]. It might be speculated that there exist still other materials in the systems La - group IIIa elements, since also LaAlO3 is known to exhibit (however small) ionic conductivity. Little is known on the stability of these materials in terms of reducibility and mechanical strength and further investigations are needed. Current-collector integrated fuel cell stack-repeat-elements. As shown in the present work it is necessary to further integrate the fuel cell components by regarding it as a materials system. From the current state of development this means, that the current collector has not be applied onto the electrodes/electrolyte element only during stacking, but already during the

219 Further Work

development of the electrodes. Only by doing so, the fractal problem of the current distribution from a coarse current collector to a fine grained electrode can be treated adequately. Losses arising from the sudden transition from a coarse to a fine structure can be avoided by applying a structural gradient material. A promising possibility for this is the application of electrically conducting foams with 30 to 60 pores per inch as current collectors. These foams build an intermediate structure between the electrode and the gas separating bipolar plate and reduce the current constriction. Such current collector developments are already under study at our institute and promising results are obtained. One example for such current collectors has already be presented in this dissertation, it is the current collector of the prototype cell shown in chapter 11.

1. Y. Teraoka, H.M. Zhang, K. Okamoto, and N. Yamazoe, Mat. Res. Bull., 23, 51 (1988).

2. T. Inoue, N. Seki, K. Eguchi, and H. Arai, J. Electrochem. Soc., 137, 2523 (1990).

3. L.-W. Tai, M.M. Nasrallah, H.U. Anderson, D.M. Sparlin, S.R. Sehlin, Solid State Ionics,

76, 259 (1995).

4. C. Ftikos, S. Carter and B.C.H. Steele, J. Europ. Ceram. Soc., 12, 79 (1993).

5. O. Yamamoto, S. Watanabe, H. Ueno, N. Imanishi, Y. Takeda, N. Sammes, and M.B.

Phillips, in Solid Oxide Fuel Cells IV, PV 95-1, p. 414, M. Dokiya, O. Yamamoto, H. Tagawa and S.C.

Singhal, Editors, The Electrochemical Society Proceedings Series, Pennington NJ, USA (1995).

6. R. Chiba, T. Ishii, in Solid Oxide Fuel Cells IV, PV 95-1, p. 482, M. Dokiya, O. Yamamoto,

H. Tagawa and S.C. Singhal, Editors, The Electrochemical Society Proceedings Series, Pennington NJ,

USA (1995).

7. A. Naoumidis, German Patent, DE 43 40 486 C1, Germany (1995).

8. T. Ishihara, H. Matsuda and Y. Takita, in 2nd Internat. Symp. Ionic and Mixed

Conducting Ceramics, PV 94-12, p. 85, T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller, Editors, The

Electrochem. Soc. Proceedings Series, Pennington, NJ, USA (1994).

9. T. Ishihara, H. Matsuda, and Y. Takita, J. Am. Chem. Soc., 116, 3801 (1994).

Curriculum Vitae

Personal

Date and place of birth: March 5, 1967, in Winterthur, Switzerland

Marital status: Married with Dorothée

Nationality: Swiss, Citizen of Rickenbach, ZH

Education

1980 - 1986 Kantonsschule Rychenberg in Winterthur

1986 Matura Typus B

1987 Military service

1987 - 1992 Student of Materials Science, Swiss Federal Institute of Technology, Abt. III D, ETH-Zürich

1992 Diploma in Materials Science

July 1992 - present Research associate and Ph. D. Student, Chair of Nonmetallic Materials, Dept. of Materials Science,

ETH-Zürich

Publications

In the course of this dissertation the following publications were accomplished:

1. M. Gödickemeier, B. Michel, A. Orliukas, P. Bohac, K. Sasaki, L.J. Gauckler, H. Heinrich,

P. Schwander, G. Kostorz, H. Hofmann and O. Frei, "Effect of Intergranular Glass Films on the Electrical

Conductivity of 3Y-TZP", J. Mat. Res., 9, 1228 (1994).

2. M. Gödickemeier, K. Sasaki, P. Bohac and L.J. Gauckler, "Current-Voltage Characteristics

of Fuel Cells with Doped CeO2-Electrolytes", in Advanced Solid Oxide Fuel Cells, Proc. 6th IEA Workshop

SOFC, ENEA, p. 225, Rome Italy (1994).

3. M. Gödickemeier, K. Sasaki, P. Bohac, A. Mitterdorfer, and L.J. Gauckler, "Development

of Alternative Electrolyte Materials and Production Methods for IT-SOFC", in Proc. 1st Swiss

Conference Mat. Research for Engineering Systems, p. 51, B. Ilschner, M. Hofmann, F. Meyer-

Olbersleben, Editors, Technische Rundschau, CH (1994).

4. M. Gödickemeier, K. Sasaki and L.J. Gauckler, in Solid Oxide Fuel Cells IV, "Current-

Voltage Characteristics of Fuel Cells with Ceria-based Electrolytes", M. Dokiya, O. Yamamoto, H. Tagawa

and S.C. Singhal, Editors, p. 1072, The Electrochem. Soc. Proceedings Series, Pennington, NJ (1995).

221 CV

5. M. Gödickemeier, L.J. Gauckler and I. Riess, "Electrochemical Characteristics of

Cathodes in Solid Oxide Fuel Cells based on Ceria Electrolytes", submitted to Solid State Ionics.

6. I. Riess, M. Gödickemeier and L.J. Gauckler, "Characterization of Solid Oxide Fuel Cells

Based on Solid Electrolytes or Mixed Ionic Electronic Conductors", submitted to Solid State Ionics.

7. K. Sasaki, J.P. Wurth, M. Gödickemeier, A. Mitterdorfer, and L.J. Gauckler, in Solid Oxide

Fuel Cells IV, "Processing-Microstructure-Property Relations of Solid Oxide Fuel Cell Cathodes",

M. Dokiya, O. Yamamoto, H. Tagawa and S.C. Singhal, Editors, p. 625, The Electrochem. Soc. Proceedings

Series, Pennington, NJ (1995).

8. K. Sasaki, M. Gödickemeier, P. Bohac, A. Orliukas and L.J. Gauckler, "Microstructure

Design of Mixed-Conducting Solid Oxide Fuel Cell Cathodes", in Proc. 5th IEA -Workshop SOFC,

Materials, Process Engineering and Electrochemistry, P. Biedermannn and B. Krahn-Urban, Editors, Jülich,

FRG, p. 187 (1993).

9. L.J. Gauckler, K. Sasaki, A. Mitterdorfer, M. Gödickemeier and P. Bohac, "Processing of

SOFC Ceramic Components", in 1st European Solid Oxide Fuel Cell Forum, U. Bossel, Editor, p. 545, Baden

Switzerland (1994).

Presentations

1. M. Gödickemeier, K. Sasaki, P. Bohac and L.J. Gauckler, "Current-Voltage

Characteristics of Fuel Cells with Doped CeO2-Electrolytes", 6th IEA Workshop Advanced SOFC, Rome,

Italy, Feb. 24, 1994.

2. M. Gödickemeier, P. Bohac, A. Mitterdorfer, K. Sasaki and L.J. Gauckler, "Development

of Alternative Electrolyte Materials for IT-SOFC", 1st Swiss Conference on Materials Research for

Engineering Systems, Sion, Switzerland, Sept. 9, 1994.

3. M. Gödickemeier, K. Sasaki and L.J. Gauckler, "Current-Voltage Characteristics of Fuel

Cells with Ceria-based Electrolytes", 4th Internat. Symposium on Solid Oxide Fuel Cells, Yokohama,

Japan, June 21, 1995. Poster, awarded for presentation design and scientific content, The SOFC

Society of Japan, 1995.

4. M. Gödickemeier, L.J. Gauckler and I. Riess, "Electrode Impedance and Electro-chemical

Properties of Fuel Cells with Mixed Ionic Electronic Conductors", 188th Electrochemical Society Meeting,

Chicago, USA, Oct. 10, 1995 .

Date post:	08-Dec-2016
Category:	Documents
Upload:	ngodien
View:	219 times
Download:	0 times

Mixed Ionic Electronic Conductors for Solid Oxide Fuel Cells

Documents