Sofc Understanding Materials Compatibility

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18) P1: IKH 10.1146/annurev.matsci.33.022802.093856

Annu. Rev. Mater. Res. 2003. 33:581–610 doi: 10.1146/annurev.matsci.33.022802.093856 c 2003 by Annual Reviews. All rights reserved Copyright ° First published online as a Review in Advance on April 1, 2003

UNDERSTANDING MATERIALS COMPATIBILITY Harumi Yokokawa Energy Electronics Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki 305-8565, Japan; email: [email protected]

Key Words solid oxide fuel cell, thermodynamic stability, chemical potential, interface stability ■ Abstract The use of chemical potential diagrams to examine interface chemistry is discussed in terms of the chemical reactions among oxides and associated interdiffusion across the interface. The driving force for both processes can be determined from the chemical potential values. The geometrical features of the chemical potential diagrams can be related to the valence stability of binary oxides and the stabilization energy of double oxides from the constituent oxides. The materials compatibility in solid oxide fuel cell materials is discussed with a focus on a lanthanum manganite cathode and a yttria-stabilized zirconia (YSZ) electrolyte. Emphasis is placed on the valence numbers of manganese in the fluorite solid solution and the perovskite oxides, which have been derived by thermodynamic analysis of the magnitude of the stabilization energy/interaction parameters as a function of ionic size for respective valence numbers. The change of manganese valence on La2Zr2O7 formation and Mn dissolution in YSZ are discussed in relation with the oxygen evolution/adsorption process. Oxygen flow associated with electrochemical reactions exhibits markedly different features depending on the direction of the polarization, which can lead to drastic changes in the interface chemistry (precipitation or interdiffusion).

INTRODUCTION Solid oxide fuel cells (SOFCs), which convert chemical energy (fuels) to electricity, consist of an electrolyte, two electrodes, interconnects, and other relevant materials (1). SOFC stacks are fabricated around 1200–1400◦ C. Their desired operating time is more that 50,000 h at ∼1000◦ C. This implies that the material’s chemical behavior (2, 3) at such high temperatures is crucial for obtaining good performance and high reliability, in addition to the materials’ electrochemical properties that are required for the proper functioning of the electrochemical cells. Chemical compatibility with other materials and the environment (fuels or air) must be taken into account in designing SOFCs. For this purpose, thermodynamic considerations should be made properly. Advances have been made in constructing thermodynamic databases (4–6, 6a,b) and developing computer software for calculating complicated chemical equilibria (7) in order to understand the complex chemical processes at high temperatures. Although chemical equilibria 0084-6600/03/0801-0581$14.00

581

4 Jun 2003

12:51

582

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

calculations are powerful, they do not provide a good basis for examining the interface stability or interdiffusion across the interfaces among the cell components. This requires a thermodynamic analysis of equilibrium at the interfaces by using a generalized chemical potential diagram (8). Stack development of SOFCs with oxide interconnects can be categorized as the first generation of SOFCs (3, 9). In this development, thermodynamic considerations played important roles in obtaining an optimized solution to surmount materials problems. A similar approach can be applied to other high-temperature processes such as alkali metal thermo-electric converters (AMTEC) (10) or molten carbonate fuel cells (11) and also to some low-temperature devices such as lithium batteries (12). In this article, we review the recent advance in understanding materials compatibility in some high-temperature SOFCs. The thermodynamic approach to understand materials compatibility is discussed mainly in terms of the chemical potentials as a key concept in considering the chemical features of interfaces. Among the SOFC materials and their interfaces, air electrode/electrolyte interfaces are selected to examine in detail, using chemically useful quantities such as valence stability, stabilization energy in complex oxides, and related acid/base properties. Other topics associated with intermediate SOFCs, which can be categorized as second generation SOFCs, are also discussed briefly.

HOW TO UNDERSTAND MATERIALS COMPATIBILITY Chemical Potential and Interfacial Chemistry In interface systems, many of the most important chemical phenomena that occur involve precipitation of new phases at the interfaces and interdiffusion across the interfaces (13). Diffusion is usually discussed in terms of a concentration gradient. Strictly speaking, however, diffusion should be treated in terms of chemical potential because the chemical potential gradient provides the true driving forces for diffusion. Chemical potential is defined as µ ¶ ∂G , 1. µi = ∂n i j6=i where G is the Gibbs energy of a system consisting of several species, and ni is the mole number of species i. Here, chemical potential is defined as the differentiation of the Gibbs energy with respect to the mole number of species i, without changing the mole number of other species. In an equilibrium state, chemical potential species are defined as constant throughout the system. To properly treat diffusion, a local equilibrium approximation is adopted; the system is divided into small areas in which the thermodynamic quantities are defined as constant. From Equation 1, it is easily understood that when a small amount of species i is taken out of one phase, where its chemical potential is high, and placed on its adjacent point having

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

583

the lower value of chemical potential, the total Gibbs energy becomes lower as a result of two processes. This explains why the chemical potential gradient gives the true driving force for diffusion. Interdiffusion across the interface cannot be treated in terms of concentration in any case. Again, the chemical potential provides a basis for examining mass transfer across the interface or the precipitation of new phases. Under the local equilibrium approximation, chemical potential can be defined separately for two adjacent phases. Even when the chemical potential has the same value for two phases, interdiffusion can have taken place when there are chemical potential gradients in the same direction across the interface. New phase precipitation is usually discussed in terms of the Gibbs energy change for the proposed precipitation reactions. All possible chemical reactions should be counted and examined by using available thermodynamic data on the Gibbs energy change for formation. For example, the following possible reactions can be written for the interface between ZrO2 and LaCoO3; LaCoO3 + ZrO2 = 0.5 La2 Zr2 O7 + CoO + 0.25 O2 (g), LaCoO3 + ZrO2 = 0.5 La2 Zr2 O7 + 1/3 Co3 O4 + 1/12 O2 (g).

2. 3.

When some of those reactions have a negative Gibbs energy change for reaction, new phases can be precipitated. However, when there are many thermodynamically possible reactions, it is difficult to examine, from the Gibbs energy alone, which kinds of reactions do take place and how they proceed. Note that a similar examination can be made in terms of the chemical potential values in a more reasonable manner. When the chemical potential values of La, Co, Zr, and O at the LaCoO3/ZrO2 interface are determined as µ(La)∗ , µ(Co)∗ , µ(Zr)∗ , and µ(O)∗ , the precipitation reaction of La2Zr2O7 can be determined in the following quantities: 1G = µ◦ (La2 Zr2 O7 ) − [2µ(La)∗ + 2µ(Zr)∗ + 7µ(O)∗ ] = 1f G ◦ (La2 Zr2 O7 ) − {2[µ(La)∗ − µ(La)◦ ] + 2[µ(Zr)∗ − µ(Zr)◦ ] + 7[µ(O)∗ − µ(O)◦ ]}.

4.

Because the chemical potentials are key properties in this analysis, they are easily extended to combine with diffusion. When several phases can be precipitated, chemical potentials provide the basis for examining how those phases are arranged across the interfaces (8). The next question then is, how are the chemical potentials determined at the interfaces? In alloy systems, it is not difficult to imagine that the elements have their respective chemical potential values. In a system containing stoichiometric compounds, however, there remains some ambiguity in determining the chemical potentials. This problem is discussed further below.

4 Jun 2003

12:51

584

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

Construction of Chemical Potential Diagram The equilibrium state can be derived by several methods; one involves the chemical equilibrium calculations (14) based on the Gibbs energy minimization technique. When the mole numbers of elements are given at a constant temperature and pressure, the equilibrium state can be defined as the state in which the total Gibbs energy of the system becomes the minimum. Many computation programs for calculating complicated chemical equilibria have been developed in the last two decades on the basis of this technique (7). The same information as provided by the Gibbs energy minimization can be derived in the chemical potential space by adopting the duality theorem (see Appendix C in Reference 8). In chemical potential space, the equilibrium state, for example in the La-Co-O ternary system, can be defined as follows; from a computer-geometrical point of view, a stable compound, LaCoO3, in the La-Co-O ternary system can be represented by the following equation for a plane in chemical potential space: µ(La) + µ(Co) + 3µ(O) = µ◦ (LaCoO3 ),

5.

where the slope of the plane is determined from the stoichiometric numbers, [1,1,3], whereas the location of this plane relative to the origin of the space is determined from µ◦ (LaCoO3), which is easily derived form the Gibbs energy change for formation (see below). All stable compounds form a polyhedron, as shown in Figure 1 for the La-CoO system at 1273 K. Different compounds have different stoichiometric numbers and, therefore, different slopes on the faces of the polyhedron. Thus polygonal facets correspond to stable compounds, whereas edges and corners correspond to two-phase and three-phase boundaries, respectively. The dual theory indicates that when mole numbers are given for respective elements in a ternary system, the equilibrium state can be obtained from the condition that a plane having the slope corresponding to the given molar ratio is maximized within the polyhedron (8). When this plane is touched to a corner, an edge, or a facet, the equilibrium state can be derived as three-phase, two-phase, and single-phase state. Their features correspond well to those in the triangle compositional diagram as shown in Figure 1. Borderlines of polygonal phases represent two-phase tie lines in the triangle compositional diagram. When compared with the conventional chemical potential diagrams, the polyhedron shown in Figure 1 and its derivative can be regarded as a generalized −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→ Figure 1 Theremodynamic phase relations for the La-Co-O system at 1273 K. (a) Polyhedron in the three-dimensional chemical potential space. Respective planes are represented by planes whose slopes and location are determined from the stoichiometric numbers and the Gibbs energy changes for formation, respectively. (b) Phase relations in a normal triangle composition diagram.

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

585

4 Jun 2003

12:51

586

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

chemical potential diagram in the following sense. (a) There is no need to specify a redox element, which makes it possible to treat alloys and double oxides in a proper manner, (b) and there is no limitation of the coordinate axes to the environmental variables. Elemental chemical potentials are used as coordinate axes. This implicitly indicates that the diagram coordinates give the driving force for mass transfer inside condensed phases as well as through gas phases. Diffusion in condensed phases is driven by the chemical potential difference, which is represented by two different points in the chemical potential polyhedron. In the dilute solution approximation, the chemical potential of impurities such as La in Co can directly correspond to the concentration of impurities. µ(La) = µφ (La) + RT ln c(La in Co),

6.

where µφ (La) is the value for the Henrian reference state. The chemical potentials of the major components of the stoichiometric compound can also change inside the stability polygon, depending on the energetic state of their related defects. In LaCoO3, the concentration of oxide ion vacancies can be given as a function of oxygen potential. Cation defects are usually negligible in the B-site in the perovskite lattice. This implies that the difference in the µ(La) and µ(Co) inside the stability field of LaCoO3 corresponds to those in concentration of the A-site vacancies. To describe these features quantitatively, modeling based on defect chemistry is required. When concentrations of defects are significant beyond the stoichiometric approximation, the phase should be treated as a mixture; in such a case, the slope of the plane in the chemical potential space becomes curved. In practical materials applications where sophisticated doping techniques are frequently adopted, phase relations in multicomponent systems are required to account for the chemical behavior of dopants. In chemical equilibrium calculations, the complication caused from increasing the number of elements can be well treated without any difficulty. On the other hand, the chemical potential diagram has some weak points in displaying phase relations in multicomponent systems. Even so, the generalized chemical potential diagram provides a better solution for constructing diagrams for such systems. By setting some of chemical potentials at selected values, appropriate displaying can be made to some extent (8). Figure 2 gives an example for the phase relations in the La-Zr-Co-O system in order to examine interface stability. When the oxygen potential is fixed at 1 atm, the thermodynamic state of binary oxides is represented with the metallic chemical potential. In the La-Zr-Co-O system, the three elemental chemical potentials, µ(La), µ(Zr), andµ(Co), can be used to set up the three-dimensional chemical potential diagram. Two such diagrams are shown at (a) 1273 and (c) 1073 K; note that the Gibbs energy change for Equation 2 is negative at 1273 K and positive at 1073 K. Geometrical features in Figure 2 exactly correspond to the Gibbs energy values. At higher temperatures (1273 K), LaCoO3 reacts with ZrO2 to form La2Zr2O7 and CoO. Reaction experiments (14a) confirmed that the products are arranged in a layer structure: LaCoO3/CoO/La2Zr2O7/ZrO2. From this sequence, a possible reactive diffusion path can be drawn (Figure 2). The same phase relation and associated

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

587

Figure 2 The three-dimensional chemical potential diagram (a,c) and corresponding composition diagram (b,d) for the La-Zr-Co-O system at p(O2) = 1 atm. (a,b) 1273 K and (c,d) 1073 K. Reaction product arrangement, LaCoO3/CoO/La2Zr2O7/ZrO2, is represented by a continuous line in (a) and (b).

4 Jun 2003

12:51

588

AR

AR189-MR33-20.tex

YOKOKAWA

Figure 2 (Continued )

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

589

reactive diffusion path are also shown in the triangle composition diagram (Figure 2b,d ). It is clear that the reactive diffusion path is properly represented in the chemical potential diagram.

Which Thermodynamic Properties Dominate the Chemical Potential? The chemical potentials are practically derived from the Gibbs energy related functions. For a stoichiometric compound, the chemical potentials have the following relations with the Gibbs energy; 1f G ◦ (LaCoO3 ) = µ◦ (LaCoO3 ) − [µ◦ (La) + µ◦ (Co) + 1.5µ◦ (O2 )]

7.

G ◦ (LaCoO3 ) = µ◦ (LaCoO3 ) − (µ◦La + µ◦Co + 1.5µ◦O2 ).

8.

Here, µ◦ (La) and µ◦ La are the chemical potentials of element La in the reference state at a given temperature and at the standard temperature (298.15 K), respectively. 1fG◦ (LaCoO3) is the standard Gibbs energy change for formation and depends on how the reference state is selected at high temperatures for each element. G◦ (LaCoO3) is the Gibbs energy relative to the reference sate of elements at 298.15 K, which has been used by Barin (15). By combining Equations 7 and 8 with Equation 5, the basic equations for constructing chemical potential diagrams can be correlated with Gibbs energies. Through the Gibbs energy values, the chemical potentials are also closely related with enthalpy and entropy. Enthalpy is correlated directly with the energetics of the complex oxides, whereas entropy is related mainly with the redox of the constituent oxides. Both functions are important when macroscopic behavior such as chemical reactions or diffusion are discussed in terms of the physicochemical properties. Because there are large differences between gas and condensed phases in entropy values, valence changes accompanied with the evolution or absorption of oxygen gas usually exhibit a large entropy change. Note, however, that the entropy change per evolved oxygen mole number is about the same regardless of the binary oxide or complex oxides (16). In contrast, the enthalpy change on valence change is strongly affected by the ionic configuration in complex oxides. These features can be well characterized in terms of the valence stability for binary oxides, and the stabilization energy of double oxides from constituent oxides can be defined as MO0.5 n + 0.25 (m − n)O2 (g) = MO0.5 m , 1[Mn+ : Mm+ ] = 1f G ◦ (MO0.5 m ) − [1f G ◦ (MO0.5 n ) + 0.25(m − n)1f G ◦ (O2 )],

9.

AO1.5 + BO1.5 = ABO3 , δ(ABO3 ) = 1f G ◦ (ABO3 ) − [1f G ◦ (AO1.5 ) + 1f G ◦ (BO1.5 )].

10.

4 Jun 2003

12:51

590

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

By using the valence stability and the stabilization energy, the Gibbs energy change for Equation 2 can be rewritten as follows; 1G(Equation 2) = 0.51f G ◦ (La2 Zr2 O7 ) + 1f G ◦ (CoO) + 0.251f G ◦ (O2 ) −[1f G ◦ (LaCoO3 ) + 1f G ◦ (ZrO2 )] = 0.5[1f G ◦ (La2 Zr2 O7 ) − 1f G ◦ (La2 O3 ) − 21f G ◦ (ZrO2 )] −{1f G ◦ (LaCoO3 ) − 0.5[1f G ◦ (La2 O3 ) − 1f G ◦ (Co2 O3 )]} −{0.51f G ◦ (Co2 O3 ) − [1f G ◦ (CoO) + 0.251f G ◦ (O2 )]} 11. = 0.5 δ(La2 Zr2 O7 ) − δ(LaCoO3 ) − 1[Co3+ : Co2+ ]. Figure 1 shows that the cobalt trivalent ions are strongly stabilized in the perovskite phase, which widens the stability field of LaCoO3 down to the lower oxygen potential. In the presence of ZrO2, however, La2Zr2O7 phase is stabilized similarily so that 0.5 δ(La2Zr2O7) is cancelled out with δ(LaCoO3). As a result, the cobalt trivalent ions no longer are stable and CoO is formed instead. The features of stabilization energy and valence stability can be also correlated with geometrical features in the chemical potential diagram shown in Figure 2. For example, the magnitude of δ(La2Zr2O7) is represented by the location of a La2Zr2O7 plane relative to those for La2O3 and ZrO2. Because δ(La2Zr2O7) does not show strong temperature dependence, the relative location of the La2Zr2O7 plane does not change between 1273 and 1073 K. In other words, the width of the stability polygon of La2Zr2O7 between La2O3 and ZrO2 is about the same. However, the width of LaCoO3 between La2O3 and CoO becomes narrower at 1273 K than at 1073 K, which is the result of valence stability, 1[Co3+ : Co2+] because δ(LaCoO3) is essentially temperature independent.

Valence Stability: Key Role of Oxygen In many cases, the valence stability for binary oxides behaves in a complicated manner because the enthalpy term is determined as the energy difference among metallic substances, ionic substances, and covalent molecules. Thus it is difficult to find a good correlation between the enthalpy change for formation and fundamental properties such as ionic radius. Fortunately, valence stability has been experimentally determined and is available in various databases (6, 17). On the other hand, the stabilization energy of double oxide from the constituent oxides is not large compared with valence stability and exhibits good correlation with ionic radii and other ionic properties (18–20) because the stabilization energy of double oxides originates only from the energy difference between several ionic substances and can be well characterized in terms of the ionic packing features for respective crystals. Characterization of reactions in terms of valence stability and stabilization energy makes it easy to discuss reactivity in terms of the valence, the ionic size, and the coordination number or other lattice environments.

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

591

Whenever valence is changed in the oxide systems, oxygen has to be involved in the chemical process and thus is one of the most important reactants or products. From the equilibrium point of view, this suggests that phase relations should be investigated as a function of oxygen potential. The redox equilibria are shifted depending on the energetic states in double or complex oxides. Furthermore, to properly understand such processes kinetically, it is essential to consider the oxygen diffusion path through gaseous channels or inside materials. In electrochemical cells with the oxide-ion conductor as electrolyte, oxygen movement is particularly important for understanding the reactivity of the involved materials.

Acid-Base Properties Another approach for understanding reactivity is based on the acid-base concept (21–24). Although there are many definitions for determining acidity in oxide systems, the chemical nature for salt formation reactions is commonly considered; that is, the acidic oxide can easily react with the basic oxide. A typical example is given as (Zr0.8 Y0.2 )O2 + 0.2 V2 O5 = 0.8 ZrO2 (monoclinic phase) + YVO4 .

12.

Here, V2O5 is acidic and Y2O3 is basic so that the YVO4 formation can easily proceed, although Y2O3 is well stabilized in the cubic fluorite structure. From a chemical point of view, this is a convenient means of understanding solid-state reactions. This tendency can be correlated with the stabilization energy in double oxides, and these stabilization energies can be interpreted in terms of ionic size, valence, and other crystal chemical features as described above. For example, V2O5 has a small valence stability because the pentavalent V5+ requires a large number of coordinate oxide ions, which makes it difficult for oxide ions to be kept apart in order to maintain low electrostatic repulsive energy. When V2O5 forms a double oxide with those oxides having large ions with lower valency states, it possible for V5+ to have a large number of oxide ion coordinates, i.e., the oxide ions are separated by large cation neighbors. This explains why the stabilization energy of YVO4 is large and therefore Equation 12 can be easily processed.

Compound Formation Versus Solid Solution in Multicomponent Systems As shown above, significant stabilization can be expected when several cations with different valence and ionic size form double or complex oxides. By increasing the component number, however, these effects will be weakened. Thus it becomes difficult to obtain greater stabilization upon further addition of oxide components.

4 Jun 2003

12:51

592

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

SOLID OXIDE FUEL CELL MATERIALS Materials Issues in Solid Oxide Fuel Cells To realize electrochemical energy conversion, SOFCs consist mainly of an oxideion conductive electrolyte, two (air and fuel) electrodes, and interconnect materials (2). Because those materials are made of solid components, some technological difficulty of fabricating SOFC stacks is encountered compared with fabricating other types of fuel cells that contain liquid substances (3). To establish a good electrical contact at a solid-solid interface, high-temperature heat treatments or physically activated processes are needed. On the other hand, those stacks have to experience changes in temperature and ambients during fabrication, operation, and maintenance. When materials are tightly bonded, differences in volume expansion give rise to thermal stress so that thermal expansion matching among the cell components becomes another important requirement. The interface stability among the cell components is crucial. When interface reaction products have poor electrical properties, a degradation in electrochemical performance can occur. When interface reaction products have different thermal expansion coefficients, micro-crack formation at the interface may lead to mechanical instability (e.g., interface delamination). In the following, interface stability among the cell components is discussed from the thermodynamic point of view with emphasis on fluorite oxides and perovskite oxides.

Thermodynamic Representation of Solid Oxide Fuel Cells Materials In order to understand interdiffusion, as well as interface reactions, it is essential to thermodynamically represent materials including dopants or possible impurities. FLUORITE SOLID SOLUTION The fluorite-type oxides, YSZ and doped ceria used as an electrolyte, are well represented by a simple sub-regular solution model (20, 26). The Gibbs energy is given as follows

G total = xA G A + xB G B = xA (G ◦A + RT ln xA ) + xB (G ◦B + RT ln xB ) + xA xB [A0 + A1 (xA − xB )],

13.

where G◦ A and G◦ B are the lattice stability of the fluorite phase, xA and xB are the concentration of A and B, and A0 and A1 are the first and the second interaction parameters in the subregular solution. Figure 3 shows the calculated phase diagram and associated chemical potential behavior of a ZrO2-YO1.5 system. In pure ZrO2, the cubic fluorite phase is stable only above 2662 K. At low temperatures, the monoclinic structure, in which the Zr ions have seven coordinates, is stable. When the YO1.5 component is added to the cubic ZrO2 phase, the large stabilization is obtained as shown in Figure 3. This stabilization is expressed as the large negative

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

593

Figure 3 (a) Calculated phase diagram for the ZrO2-YO1.5 system and (b) Gibbs energy and chemical potentials of ZrO2 and YO1.5 at 1273 K.

value of A0 in Figure 4. Because the oxide ion vacancies are formed by YO1.5 doping, the cubic structure is stabilized. Recent investigations on the vacancy configuration by atomistic calculation (27, 28) and by EXAFS analyses (29) indicate that the oxide ion vacancies are formed selectively around the Zr ions, leading to the ionic configuration in which the Zr4+ ions have about seven coordinate oxide ions even in the cubic structure. With this in mind and with the aid of atomistic calculations, the macroscopic property can

4 Jun 2003

12:51

594

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

be well correlated with features of the ionic configuration. Successful attempts have been made to correlate the chemical potential of the YO1.5 component with the defects properties such as proton solubilities and hole conductivities (30–32). Although the phase diagrams of the MO2-M’On (M = Zr, Ce; M’ = dopant) systems exhibit a variety of behaviors, the interaction parameters for respective dopant valences show excellent regularity with the dopant radius (20). Figure 4 shows the interaction parameters for the ZrO2-MO and the ZrO2-MO1.5 systems. These parameters provide analyses for following chemical behaviors. REACTIVITY OF THE DOPANT COMPONENTS Chemical reactivity of doped fluorite oxides can be separately discussed for host materials and dopants. Although the chemical potential of host materials is not sensitive to the types of dopants,

Figure 4 Interaction parameters for the fluorite solid solution between the cubic zirconium oxide and metal oxides with divalent (upper) and trivalent (lower) ions, respectively. The coefficient a represents the lattice stability of MO (or MO1.5) relative to the most stable lattice structure. The parameters A0 and A1, are the first and second Redlich-Kister equations. Dopant radius for the 8 coordinates is used for each ion.

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

595

their chemical potential does depend on composition of the dopants and the hosts. For example, the YO1.5 component chemical potential values are quite different between the ZrO2-YO1.5 and the CeO2-YO1.5 systems (30). This gives rise to different reactivities with other substances; e.g., for Al2O3, the YO1.5 component in YSZ does not react, whereas that in the doped ceria does react to form YAlO3. Similarly, the basic oxide YO1.5 tends to react with the acidic oxides such as H2O, CO2, or SiO2. The reactivity with such substances also depends strongly on the chemical potential of YO1.5 in host materials (32).

Figure 5 Calculated solubility of the transition metal oxides and the associated equilibrated phase at 1273 K as a function of oxygen potential. The dash and dotted lines separate the contributions from the different valences.

4 Jun 2003

12:51

596

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA SOLUBILITY OF THE TRANSITION METAL OXIDES It is difficult to find correlations among solubilities of the transition metal oxides in YSZ. Even so, by using the present interaction parameters, the difference in solubilities can be well interpreted in terms of the valence number and the ionic size. Figure 5 shows the calculated solubility of transition metal oxides as a function of oxygen potential. The solubility of Mn oxide is larger than that of Fe oxide. This is because Mn2+ ion size is larger and closer to the critical dopant size, as shown in Figure 3, and also because of the valence stability of binary oxides; the Fe solubility decreases rapidly with lowering oxygen potential in the Fe metal stable region, whereas the stability region of MnO is large, which makes the solubility of Mn oxide constant over a wide range of oxygen potential. The valence state of transition metal ions in YSZ has been experimentally determined by Sasaki et al. (33). Figure 6 compares the calculated and observed values of transition metal ion concentration in YSZ. Except for V3+/V4+ ions, essentially the same behavior is obtained, which confirms that the valence number and the ionic size effects are important in determining the thermodynamic state of dopants. PEROVSKITE OXIDE The perovskite type oxides, ABO3, consist of two cations and oxide ions (34, 35). The B cations can have valences from +3 to +5 with 6 oxygen coordinates, whereas the A cations have the valences from +3 to +1 with 12 oxygen coordinates. The stabilization energy depends strongly on how the ionic configuration can be fitted to the perovskite structure; this is measured in terms of the Goldshmidt tolerance factor, t, which is determined from ionic radii, rA, rB, and rO, as follow: (rA + rO ) . 14. t =√ 2(rB + rO )

The most ideally packed ionic configuration can be obtained when t = 1; correspondingly, the stabilization energy shows an excellent dependence of t with the largest value around t = 1 as shown in Figure 7 (36). The large stabilization energy of LaMO3(M = transition metal) stabilizes these perovskite phases against reduction. Figure 8 shows that the reductive decomposition limit of LaMO3 is shifted to the more reductive side when compared with the corresponding M2O3 phase. Although LaCoO3 has a large stabilization energy, its decomposition oxygen potential is high compared with that of other perovskites because the valence stability of Co3+ is originally low. When comparison is made between LaMnO3 and LaFeO3, interesting differences can be seen in decomposition products and oxygen potential despite the fact that the magnitude of stabilization is about the same. Such differences originate from differences in the valence stability. The chemical features of perovskite oxides can be summarized as follows 1. A large oxygen nonstoichiometry; although the interstitial positions are not available for oxygen atoms in the perovskite structure, oxide ion vacancies can be formed to a large extent. Nondoped perovskites can have oxide ion vacancies when the valence of the B-site cation is changed.

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

597

Figure 6 Comparison between calculated and observed valence states of transition metal ions in YSZ as a function of oxygen potential.

2. The A-site ions can be substituted with cations with different valences. To compensate for the charge, the oxide ion vacancy formation or the hole (electron) formation will be accompanied by such a substitution. 3. An A-site deficiency can occur when the B-site cations have high valence states. Because the B-site ions change their valency, it is possible that oxide ion vacancies will form in addition to A-site vacancies.

Figure 7 The stabilization energy of perovskite oxides, AIIIBIIIO3 as a function of tolerance factor determined from the ionic size alone.

4 Jun 2003

12:51

598

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

Figure 8 The two-dimensional chemical potential diagram for the La-Mn-O, La-Fe-O, and La-Co-O systems at 1273 K.

In many cases, perovskite oxides exhibit a large mutual solubility despite the difference in the ionic size among respective A-site and B-site ions. This behavior suggests an ideal association model in which the constituents are assumed to have a perovskite-related structure. To investigate the Sr substitution on the La site, a mixture of LaCrO3 and SrCrO3 can be considered. When hypothetical (unstable) constituents such as CaCrO3 are considered, the good correlation between the stabilization energy and the ionic size allows for estimating the thermodynamic properties. To represent oxide ion vacancies, another type of hypothetical compound, LaMnO2.5, can be considered. For the A-site deficiency, La2/3MnO3 or La2/3MnO2.5 can be included as possible constituents (36). By adopting these constituents for LaMnO3 perovskites, attempts have been made to represent the thermogravimetric results on the A-site-deficient lanthanum manganite, which has been observed by Mizusaki and coworkers (37). As shown in Figure 9a, the reproducibility of the thermogravimetric results on LaMnO3-d, La0.95MnO3-d, and La0.9MnO3 is satisfactory. From the defect chemistry point of view, several investigators have taken into account the thermogravimetric results. Among them, Roosmalen & Cordfunke

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

599

(38–42) assumed that the disproportionation of manganese trivalence ions into divalent and tetravalences may occur; furthermore, they assumed that the tetravalence will exist at a selected concentration throughout the oxygen-deficient region. These assumptions have been adopted by Mizusaki et al. (43, 43a) to account for the electron conductivity and related defect properties of lanthanum manganite. In Figure 9c, the change of respective valence concentration is shown as a function of average Mn valence. The main point is that the tetravalent Mn4+ concentration is assumed to be constant over a wide range of Mn valence number; this is made to reproduce the thermogravimetric behavior. On the other hand, Figure 9b clearly indicates that without assuming the tetravalent manganese ions in the oxygen deficient region, the thermogravimetric results can be well taken into account within the ideal association model.

Valence Stability and Cathode/Electrolyte Interaction The transition metal elements play important roles in SOFC materials (44–46). In particular, air electrode and interconnect materials are made of manganese and chromium compounds, respectively. More recently, iron has attracted much attention as an intermediate SOFC cathode (47). The validity of such materials selection can be examined in terms of the valence stability and the stabilization energy. In what follows, cathode and electrolyte interaction are examined for the viewpoint of valence stability.

Figure 9 The nonstoichiometry of LaMnO3 and its relation with valence numbers. (a) Experimental thermogravimetric results (37) (square, circle and diamond) and model fitting results (solid lines), (b) calculated valence numbers as a function of average manganese valence, (c) proposed valence numbers by Mizusaki et al. (43, 43a).

4 Jun 2003

12:51

600

AR

AR189-MR33-20.tex

YOKOKAWA

Figure 9 (Continued)

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

601

The LaCoO3-based cathode was not selected for use with the YSZ electrolyte because the thermal expansion coefficient is too high and the reactivity with YSZ is severe. However, the LaMnO3-based perovskite has attracted much interest. Although the electrode performance of LaMnO3-based cathodes are relatively good, some reactions with YSZ have occurred. Lau & Singhal (48) investigated the interface stability between (La,Sr)MnO3 (LSM) and YSZ. They found that interdiffusion across the interface, particularly manganese dissolution into YSZ, is significant and also that La2Zr2O7 is formed at the interface. The thermodynamic considerations on reactivity with YSZ were made to optimize the LaMnO3-based cathode (49–59). As described above, the manganese ions can be dissolved into YSZ as Mn3+ and Mn2+ in the cathode atmosphere, and the concentration of Mn2+ will increase with decreasing oxygen potential. From the electrical measurements, some anomaly was observed in the oxygen potential where manganese valence changed from +3 to +2 (60). Reactions of LaMnO3-based perovskites with YSZ are not destructive so that the electrochemical activity does not degrade significantly compared with that of LaCoO3-based perovskite. As shown in Figure 10, the Gibbs energy change for the following reactions are essentially positive in the fabrication temperature and the operation temperature: LaMnO3 + ZrO2 = 0.5 La2 Zr2 O7 + MnO + 0.25 O2 ,

15.

LaMnO3 + ZrO2 = 0.5 La2 Zr2 O7 + MnO1.5 ,

16.

Figure 10 Temperature dependence of the possible reactions between LaMnO3 and ZrO2.

4 Jun 2003

12:51

602

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

LaMnO3 + ZrO2 + 0.25 O2 = 0.5 La2 Zr2 O7 + MnO2 .

17.

Since Lau & Singhal (48) actually observed the La2Zr2O7 formation at interfaces, other type of reactions should also be considered. To properly account for the reaction behavior of lanthanum manganite, it is necessary to consider the valence changes in the perovskite phase, which is related to the A-site deficiency as described above. The consideration of the A-site deficiency on the reactivity can lead to the following reaction: LaMnO3 + xZrO2 + 0.25 O2 (g) = 0.5xLa2 Zr2 O7 + La1−x MnO3 .

18.

Here, the average manganese valence changes from 3 to 3(1+x). When the initial A-site deficiency increases in La1-xMnO3, the driving force of reaction decreases and eventually the three-phase coexistence field, La1-x0 MnO3, ZrO2, and La2Zr2O7, appears. Below this critical A-site deficiency, x0 , La1-xMnO3 no longer reacts with YSZ. Note that the above relation is oxidative. In the A-site-deficient LaMnO3, the manganese dissolution into YSZ becomes significant because the MnOn activity in the A-site-deficient LaMnO3 becomes high. This Mn dissolution is reductive. Figure 11 shows that the chemical potential diagram for the La-Zr-Mn-O system at a constant oxygen potential. The solubility of Mn and La oxides in the cubic YSZ phase are included.

Chemical Reactions During Cell Fabrication and Cell Operation On La2Zr2O7 formation or Mn dissolution, the manganese valence changes are accompanied by evolving or adsorbing oxygen gas changes. Under electrochemical polarization, oxygen potential distribution is developed inside the cathode layer. This gives rise to an interesting situation in which the chemical reactivity at the LaMnO3/YSZ interfaces depends on the direction and extent of polarization. This reaction was experimentally confirmed by Tricker & Stobbs using TEM (61, 62). They observed the La2Zr2O7 formation along the LaMnO3/YSZ interface after high-temperature heat treatment. Because this reaction needs oxygen as a reactant, it is reasonable for La2Zr2O7 to extend along the interface, as shown in Figure 12a, because only the LSM/YSZ interface is an effective oxygen diffusion path. Note that the oxygen permeation rate through YSZ and LaMnO3 is very small because in the oxide-ion conductive YSZ, the electron conductivity is small and the oxide-ion conductivity is also small in the electron-conductive LaMnO3. Quite different features were observed after 24-h cell operation. When SOFC operation is performed, the oxygen flow and related oxygen potential distribution drastically change depending on the overpotential or the current density. Recent investigations on cathode reaction mechanism have revealed that the electrochemical active sites are distributed along the electrolyte/electrode/gas three-phase boundaries (TPBs). Oxygen gas is diffused on the surface of the LSM or through gaseous channels. Thus the oxygen potential in the gas phase or the LSM surface becomes minimal at

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

603

Figure 11 Chemical potential diagram for the La-Zr-Mn-O system at p(O2) = 1 atm. [Reproduced from (56) with permission from The Electrochemical Society]

the TPBs. Because oxide ions are formed at TPBs and are transported to the other side of the electrolyte, the inside LSM/YSZ interface becomes more reductive than the surface along TPBs, which provides the driving force for La2Zr2O7 to move to the oxidative side and disappear from the interface region as shown in Figure 12b (58). This phenomenon was first observed by Tricker & Stobbs (61) and later by Weber et al. (63). When the A-site LaMnO3 is adopted, the high-temperature heat treatment with YSZ produces two distinct interface areas (64, 65). Because the manganese dissolution is a reductive and therefore oxygen-evolving reaction, the manganesedissolved area can be well distinguished from the region where Zr or Y dissolved into perovskite phases (see Figure 12c). Under the cell operation, the Mn dissolution is enhanced at the entire interface, and the LSM/YSZ interface becomes homogeneously flat as shown in Figure 12d. On anodic polarization of the LSM/YSZ interface, the oxygen potential of the interface shifts to the more oxidative side. In the middle of the LSM interface

4 Jun 2003

12:51

604

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

Figure 12 (a) La2Zr2O7formation at high temperature; (b) La2Zr2O7 disappearing from the interface under a cell operation; (c) formation of spot-like Mn-rich region at convex parts of YSZ; (d) flattening of a YSZ/LaMnO3 interface under a cell operation.

islands, where the driving force of the La2Zr2O7 formation becomes large, the oxygen potential is the highest. Once the La2Zr2O7 phase is formed, this blocks the oxygen flow, which results in the increase of oxygen potential. Increased internal gas pressure leads to the detachment of LSM from YSZ. In view of this, it is essential in a water electrolyzer to avoid the composition region where the La2Zr2O7 formation can be expected (66).

Further Improvement of YSZ/(La,M)MnO3 Interface With the knowledge that the A-site-deficient lanthanum manganite does not react with YSZ, Suzuki et al. (67) fabricated the LSM/YSZ interface by electrochemical vapor deposition (EVD). The resulting cathode exhibits excellent performance as illustrated in Figure 13. Their investigation clearly indicated that (a) when TPBs are long enough, the activation overpotential is extremely small. For this purpose, the EVD process is excellent. (b) When air was used instead of pure oxygen, some overpotential appears. This is apparently due to the gas diffusion inside pores in a cathode layer. In practical stack fabrication, therefore, a key issue has been made clear: how to fabricate the cathode/electrolyte interface using inexpensive sintering procedures. Suzuki et al. (67) have revealed that the cathode overpotential is 200 times higher when the sintering method is adopted by using the same materials. Although the sintering behavior of LSM is not significant when sintering is made on LSM alone, the microstructure of LSM is usually damaged when LSM

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

MATERIALS COMPATIBILITY

605

Figure 13 (a) The cathode performance reported by Suzuki et al. (67) and (b) the temperature dependence of the reaction resistivity.

is sintered with YSZ (68). This is probably because the dissolution of Zr and/or Y into perovskite phase enhances the sinterability. In particular, this degradation due to the change in microstructure occurs after the heat treatment at temperatures higher than 1473 K. Several attempts have been made to overcome this issue: ■ ■ ■

■

The main efforts have involved the use the LSM/YSZ composite cathodes. Other efforts have sought alternative cathodes. Further doping to LSM has been tried: Mori et al. (69) found that the A-sitedeficient lanthanum manganite showed some degradation after heat treatment above 1473 K and attempted to improve this behavior by doping the chromium component with lanthanum manganite with the expectation that this would give rise to poor sinterability (69). Although some improvement was obtained, the overall results were not good. The most important demerit of Cr doping is the weakening of the cathode layer adhesion. Herbstritt and coworkers (70), using a Nobel technique, attempted to prepare a metal organic deposition thin (80 nm) layer of LSM on YSZ by spin coating process. Interestingly, this thin layer of LSM became porous during cell operation, providing a fine microstructure with long TPBs.

Composite Electrode Mizusaki et al. (68) also realized that the degradation of cathode can be derived from the microstructure change due to sintering. Their cathode, (La0.6Ca0.4)MnO3, did not react with YSZ so that the degradation is not due to the reactions among the components but rather to the morphological change, leading to the decrease in the TPBs. They found that addition of YSZ to cathode improved the degradation.

4 Jun 2003

12:51

606

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

Risø National Laboratory have made systematic investigations on the composite cathodes (71–75), whereas Virkar and coworkers (76, 77) and Bernett and coworkers (78–80) made efforts in obtaining excellent performance with composite electrodes at lower temperatures. From the thermodynamic and related points of view, the success of the LSM/ YSZ composite cathodes can be interpreted in terms of the following features: ■

■

■

■

During high-temperature heat treatment, La2Zr2O7 is formed if equilibrium is not achieved. Thus it is possible to avoid the La2Zr2O7 formation between LSM in the composite cathode and the YSZ electrolyte plate, which is the most important interface in view of the fact that it is here that the electrochemical active sites are distributed around this interface. The La2Zr2O7 phase, if formed inside the composite cathode, can inhibit sintering of LSM. Without La2Zr2O7, YSZ can also inhibit sintering. Recent attempts of using doped ceria (80, 81) instead of YSZ imply that doped ceria can also inhibit sintering of LSM. The LSM/YSZ composites exhibit higher oxide ion diffusivity and surface reaction rate. This assists the oxygen flow to the TPBs on the YSZ electrolyte plate.

Alternative Cathodes Many attempts have been made to clarify the interface stability of LSM on YSZ and to seek alternative cathodes that are chemically stable against YSZ (82–86). Initial reaction experiments were made around 1000 ◦ C, and essentially no new cathodes that stable against reactions with YSZ were proposed. Recently, attempts have been made to obtain cathodes for the intermediate temperature SOFCs. Among perovskite oxides, LaFeO3-based perovskites (47) have attracted much interest because reactions with YSZ are found to be insignificant. Reactivity and its relation with the A-site deficiency in those LaFeO3-based cathodes can be examined in terms of the valence stability and related properties in a manner similar to those of the LaMnO3-based cathodes.

SUMMARY A thermodynamic method of investigating interface stability has been discussed in terms of chemical potentials. Within the chemical potential space, phase relations associated with chemical reactions are represented using the stoichiometric numbers and the Gibbs energy for their respective compounds. Furthermore, diffusion through compounds can be well characterized in terms of the chemical potential values, which are given within the stable regions. Chemical potentials can also be correlated with valence stability and the stabilization energy through Gibbs energy because the stabilization energy is well explained in terms of the valence number, ionic size, and ionic-packing features. These considerations make it

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

P1: IKH

607

possible to investigate interface chemistry by using chemical means such as valence, ionic size, or ionic configuration. For the LaMnO3-based cathode/YSZ electrolyte interface, this approach made it possible to correlate those fundamental findings with the results of industrial efforts. This approach can be applied to other materials problems of SOFCs (87, 88), molten carbonate fuel cells (11), and other electrochemical devices (10, 12). The Annual Review of Materials Research is online at http://matsci.annualreviews.org

LITERATURE CITED 1. Minh NQ, Takahashi T. 1995. Science and Technology of Ceramic Fuel Cells. Amsterdam: Elsevier. 2. Kawada T, Yokokawa H. 1997. In Electrical Properties of Ionic Solids, ed. J Nowotny, CC Sorrell, pp. 187–248. Zurich:Trans Tech Publ. 3. Yokokawa H. 2003. History of High Temperature Fuel Cell Development: Handbook of Fuel Cells, ed. W Vielstich, H Gasteiger, A Lamm, pp 219–66. London: Wiley & Sons 4. Bale CW, Chartrand P, Degterov SA, Eriksson G, Hack K, et al. 2002. Calphad 26(2):189–228 5. Anderson JO, Helander T, H¨oglund L, Shi P, Sundman B. 2002. Calphad 26(2):273– 312 6. Yokokawa H, Yamauchi S, Matsumoto T. 2002. Calphad 26(2):155–66 6a. Yokokawa H, Yamaji K, Horita T, Sakai N. 2000: Calphad 24(4):435–48 6b. Yokokawa H, Yamauchi S, Matsumoto T. 1994. Thermochim. Acta 245:45–55 7. Eriksson G, Hack K. 1990. Metal. Trans B 21B:1013–23 7a. Bale CW, Eriksson G. 1990. Can. Metal. Q. 29:105–32 8. Yokokawa H. 1999. J. Phase Equilibria 20(3):258–87 9. Singhal SC. 2000. Solid State Ionics 135: 305–13 10. Yokokawa H, Horita T, Sakai N, Kawada T, Dokiya M. 1995. Solid State Ionics 78:203–10

11. Yokokawa H, Sakai N, Kawada T, Dokiya M, Ota K. 1993. J. Electrochem. Soc. 140: 3565–77 12. Yokokawa H, Sakai N, Yamaji K, Horita T, Ishikawa M. 1998. Solid State Ionics 113–115:1–9 13. Kirkaldy JS, Young DJ. 1987. Diffusion in the Condensed State. London: Institute of Metals 14. van Zeggeren F, Storey SH. 1970. The Computation of Chemical Equilibria, London: Cambridge Univ. Press 14a. Tagawa H, Mizusaki J, Katou M, Hirano K, Sawata A, Tsuneyoshi K. 1991. Proc. 2nd Int. Symp. Solid Oxide Fuel Cells, ed. F Grosz, P Zegers, SC Singhal, O Yamamoto, pp. 681–88. Brussels: Commision Eur. Communities 15. Barin I. 1993. Thermochemical Data of Pure Substances, Weinheim: VCH. 16. Spear KE. 1976. In Treatise on Solid State Chemistry, ed. NB Hannay, 4:115– 92. New York: Plenum 17. Wagman DD, Evans WH, Parker VB, Schumm RH, Halow I, et al. 1982. J. Phys. Chem Ref. Data 11:Suppl. 2 18. Yokokawa H, Kawada T, Dokiya M. 1989. J. Am. Ceram. Soc. 72:C152–53 19. Yokokawa H, Sakai N, Kawada T, Dokiya M. 1991. J. Solid State Chem. 94:106– 20 20. Yokokawa H. 1998. In Zirconia Engineering Ceramics: Old Challenges–New Ideas, ed. E Kisi, pp. 37–74. Zurich: Trans Tech Pub.

4 Jun 2003

12:51

608

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

21. Flood H, Førland T. 1947. Acta Chemica Scand. 1:592 22. Maier J. 2001. Chem. Eur. J. 7:4762–70 23. Yamaguchi S, Nakamura K, Higuchi T. Shin S, Iguchi Y. 2000. Solid State Ionics 136–137:191–95 24. Ingram MD, Janz GJ. 1965. Electrochim. Acta 10:783–92 25. Deleted in proof 26. Yokokawa H, Sakai N, Kawada T, Dokiya M. 1993. In Science and Technology of Zirconia V, ed. SPS Badwal, MJ Bannister, RHJ Hannink, pp. 59–68. Lancaster, Australia: Technomic Pub. 27. Minervini LM, Zacate MO, Grimes RW. 2000. Solid State Ionics 116:339– 49 28. Zacate MO, Minervini L, Bradfield DJ, Grimes RW, Sickafus KE. 2000. Solid State Ionics 128:243–54 29. Yoshida H, Deguchi H, Inagaki T, Miura K, Horiuchi M. 2001. Solid Oxide Fuel Cells VII, ed. H Yokokawa, SC Singhal, pp. 384–92. Pennington, NJ: Electrochem. Soc. 30. Yokokawa H, Sakai N, Horita T, Yamaji K, Xiong YP, et al. 2001. J. Phase Equilibria 22(3):331–38 31. Yokokawa H, Sakai N, Horita T, Yamaji K, Xiong YP, et al. 2001. In Solid Oxide Fuel Cells VII, ed. H Yokokawa, SC Singhal, pp. 339–48. Pennington, NJ: Electrochem. Soc. 32. Yokokawa H, Sakai N, Horita T, Yamaji K, Xiong YP. 2002. High Temperature Materials, pp. 26–37. Pennington, NJ: Electrochem. Soc. 33. Sasaki K, Claus J, Maier J. 1999. Solid State Ionics 121:51–60 34. Goodenough JB, Longo JM. 1970. Landolt-Bornstein, Group IV, Magnetic Oxides and Related Oxides, Vol. 4a, Chpt. 3. Berlin: Springer-Verlag 35. Nomura S. 1978. Landolt-Bornstein, Group IV, Magnetic Oxides and Related Oxides, Vol 12a. Chpt. 3. Berlin: SpringerVerlag 36. Yokokawa H, Horita T, Sakai N, Dokiya

37.

38. 39.

40. 41. 42. 43.

43a.

44.

45.

46.

47.

48. 49. 50. 51.

M. 1996. Solid State Ionics, 86–88:1161– 65 Mizusaki J, Tagawa H, Yonemura Y, Minamiue H, Nambu H. 1994. Proc. 2nd Ionic and Mixed Conducting Ceramics, ed. TA Ramanarayanan, WL Worrell, HL Tuller, pp. 402–11. Pennington, NJ: Electrochem. Soc. Van Roosmalen JAM, Cordfunke EHP. 1993. J. Solid State Chem. 93:212–23 Van Roosmalen JAM, Cordfunke EHP, Helmholdt RB. 1994. J. Solid State Chem. 110:100–5 Van Roosmalen JAM, Cordfunke EHP. 1994. J. Solid State Chem. 110:106–8 Van Roosmalen JAM, Cordfunke EHP. 1994. J. Solid State Chem. 110:109–12 Van Roosmalen JAM, Cordfunke EHP. 1994. J. Solid State Chem. 110:113–17 Mizusaki J, Mori N, Takai H, Yonemura Y, Minamiue H, et al. 2000. Solid State Ionics 129:163–77 Mizusaki J, Yonemura Y, Kamata H, Ohyama K, Mori N, et al. 2000. Solid State Ionics 132:167–80 Yokokawa H, Horita T, Sakai N, Dokiya M. 1998. Ceramic Interfaces: Properties and Applications, ed. RSC Smart, J Nowotny, pp. 171–201. London: IOM Commun. Sakai N, Tsunoda T, Kojima I, Yamaji K, Horita T, et al. 2001. Ceramic Interfaces 2, ed. HI Yoo, SJL Kang, pp. 135–56. London: IOM Commun. Sakai N, Tsunoda T, Fukumoto N, Kojima I, Yamaji K, et al. 1999. J. Electroceramics 4:119–26. Suppl. 1 Ralph JM, Vanghey JT, Krumpelt M. 2001. Solid Oxide Fuel Cells VII, ed. H Yokokawa, SC Singhal, pp. 466–75. Pennington, NJ: Electrochem. Soc. Lau SK, Singhal SC. 1985. Corrosion 85, Pap. 345:1–9 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1989. Denki Kagaku 57:821–28 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1989. Denki Kagaku 57:829–36 Yokokawa H, Sakai N, Kawada T, Dokiya

4 Jun 2003

12:51

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

MATERIALS COMPATIBILITY

52. 53. 54.

55.

56.

57. 58.

59.

60.

61.

62.

63.

M. 1990. Proc. Int. Conf. SOFC Nagoya, pp. 118–34. Tokyo: Science House Yokokawa H, Sakai N, Kawada T, Dokiya M. 1990. Denki Kagaku 58:161–71 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1990. Denki Kagaku 58:489–97 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1990. Solid State Ionics 40–41:398– 401 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1991. Proc. 2nd. Int. Symp. Solid Oxide Fuel Cells, ed. F Grosz, P Zegers, SC Singhal, O Yamamoto, pp. 663–70. Brussels: Commission Eur. Communities Yokokawa H, Sakai N, Kawada T, Dokiya M. 1991. J. Electrochem. Soc. 138:2719– 27 Yokokawa H, Sakai N, Kawada T, Dokiya M. 1991. Solid State Ionics 52:43–56 Yokokawa H, Horita T, Sakai N, Kawada T, Dokiya M. 1994. Proc. First Eur. Solid Oxide Fuel Cell Forum, ed. U. Bossel, pp. 425–34. Oberrohrdorf: Eur. Fuel Cell Forum Yokokawa H, Horita T, Sakai N, Kawada T, Dokiya M, et al. 1995. Solid Oxide Fuel Cells IV, ed. M Dokiya et al., pp. 975–84. Pennington, NJ: Electrochem. Soc. Kawada T, Sakai N, Yokokawa H, Dokiya M. 1992. Solid State Ionics 53–56: 418– 25 Tricker DM, Stobbs WM. 1993. In High Temperature Electrochemical Behavior of Fast Ions and Mixed Conductors, ed. FW Poulsen, JJ Bentzen, T Jacobsen, E Skou, MJL Østerg˚ard, pp. 453–60. Roskilde: RISØ National Laboratory Tricker DM. 1993. The microstructure of solid oxide fuel cells and related metal/oxide interfaces. PhD thesis. Cambridge, UK: Univ. Cambridge. 205 pp. Weber A, M¨anner R, Jobst B, Schiele M, Cerva H, et al. 1996. High Temperature Electrochemistry: Ceramics and Metals, ed. FW Poulsen, N Bonanos, S Linderoth, M Mogensen, B Zachar-Christianse, pp. 473–78. Roskilde: RISØ National Laboratory

P1: IKH

609

64. Horita T, Yamaji K, Negishi H, Sakai N, Yokokawa H, Kato T. 2000. Solid State Ionics 137:897–904 65. Mitterdorfer A, Gauckler LJ. 1998. Solid State Ionics 111:185–18 66. Kaiser A, Monreal E, Koch A, Stolten D. 1996. Ionics 2:184–89 67. Suzuki M, Sasaki H, Otoshi S, Kajimura A, Sugiura N, Ippommatsu M. 1994. J. Electrochem. Soc. 141(7):1928–31 68. Mizusaki J, Tagawa H, Tsuneyoshi K, Sawata A, Katou M, Hirano K. 1990. Denki Kagaku 58(6):520–27 69. Mori M, Sakai N, Kawada T, Yokokawa H, Dokiya M. 1990: Denki Kagaku 58:528– 32 70. Herbstritt D, Warga C, Weber A, IversTiff´ee E. 2001: Solid Oxide Fuel Cells VII, ed. H Yokokawa, SC. Singhal, pp. 349–57. Pennington, NJ: Electrochem. Soc. 71. Juhl M, Primdahl S, Manon C, Mogensen M. 1996. Solid State Ionics 61:173– 81 72. Østerg˚ard MJL, Mogensen M. 1993. Electrochim. Acta 38:2015–20 73. Østerg˚ard MJL, Clausen C, Bagger C, Mogensen M. 1995. Electrochim. Acta 40(2):1971–81 74. Jørgensen MJ, Mogensen M. 2001. J. Electrochem. Soc. 148:A433–42. 75. Jørgensen MJ, Bonanos N, Mogensen M. 2001: Solid Oxide Fuel Cells VII, ed. H Yokokawa, SC. Singhal, pp. 547–54. Pennington, NJ: Electrochem. Soc. 76. Virkar AV, Chen J, Tanner CW, Kim JW. 2000. Solid State Ionics 131:189–98 77. Tanner CW, Fung KZ, Virkar AV. 1997. J. Electrochem. Soc. 144:21–30 78. Tsai T, Barnett SA. 1997. Solid State Ionics 93:207–17 79. Murray EP, Tsai T, Barnett SA. 1998. Solid State Ionics 110:235–43 80. Murray EP, Barnett SA. 2001. Solid State Ionics 143:265–73 81. Hiwatashi K, Ueno A, Aizawa M. 2002. 5th. Eur. SOFC Forum, ed. J Huijsmans, pp. 1141–21. Oberrohrdorf: Eur. Fuel Cell Forum

4 Jun 2003

12:51

610

AR

AR189-MR33-20.tex

AR189-MR33-20.sgm

LaTeX2e(2002/01/18)

P1: IKH

YOKOKAWA

82. Sakaki Y, Takeda Y, Kato A, Imanishi N, Yamamoto O et al. 1999. Solid State Ionics 118:187–94 83. Mori M, Abe T, Itoh H, Yamamoto O, Shen GQ, et al. 1999. Solid State Ionics 123:113–19 84. Kindermann L, Das D, Nickel H, Hilpert K, Appel CC, Poulsen FW. 1997. J. Electrochem. Soc. 144:717–20. 85. Kindermann L, Das D, Bahadur D, Nickel H, Hilpert K. 1998. 106:165–72 86. Kindermann L, Das D, Bahadur D, Weiss

R, Nickel H, Hilpert K. 1997. J. Am. Ceram. Soc. 80:909–14 87. Yokokawa H, Horita T, Sakai N, Kawada T, Dokiya M. 1993. In High Temperature Electrochemical Behavior of Fast Ion and Mixed Conductors, ed. FW Poulsen, JJ Bentzen, T Jacobsen, E Skou, MTL Østerg˚ard, pp. 474–78. Roskilde: Risø National Laboratory 88. Horita T, Choi JS, Lee YK, Sakai N, Kawada T, et al. 1995. J. Am. Ceram. Soc. 78(7):1729–36