2018 Volume 59 Issue 2 Pages 244-250
The structural irregularities of solid-oxide layers for fuel cells prepared by electrophoretic deposition and co-sintering were examined using electron-probe microanalysis (EPMA) and X-ray analysis. The solid-oxide electrolyte layer with Sr- and Mg-doped lanthanum gallate (LSGM) prepared on a NiO-yttria-stabilized zirconia (YSZ) anode with an inserted Gd-doped ceria (GDC) buffering interlayer was not rigid. EPMA of the cross-section showed diffusion of La and Sr out of LSGM, and Ni, Y, and Zr out of NiO-YSZ. Using synchrotron radiation, X-ray absorption near-edge structures of the layer cross-sections were examined using an X-ray fluorescence yield method. The spectral features supported the formation of SrLaG3O7 and La-doped GDC by a reaction between the layers. The formation of these and other oxides was also confirmed by X-ray powder diffraction patterns. Because the electrophoretic deposition layers were co-sintered, elemental diffusion must have occurred before the synthetic powders were well fused and fixed in each layer. As an alternative to GDC, La-doped ceria (LDC) was synthesized, and La diffusion between LSGM and LDC was examined using X-ray powder diffraction. LDC, which contains 40% La, seems to be the best material to suppress La migration.
Fuel cells have been developed as high-efficiency generators of electricity from chemical energy. Solid-oxide fuel cells (SOFCs) have attracted attention because they provide high conversion efficiencies, can use a variety of fuels, and do not require noble metal catalysts, such as Pt. SOFCs are composed of a solid electrolyte (usually an oxygen-ion conductor), an anode, and a cathode, all of which are made of heat-resistant ceramics or metals.1,2) However, reducing the size and weight of SOFCs, lowering their running temperature, and improving their long-term durability are important issues that must be addressed. Thus, Sr- and Mg-doped LaGaO3 (LSGM), which has a perovskite-type structure and high oxygen-ion conductivity at intermediate temperatures (600–800℃), is expected to replace yttria-stabilized zirconia (YSZ), which is a well-known and commercially used electrolyte material3). In a previous study4), an SOFC composed of La0.8Sr0.2Ga0.8Mg0.2O3−δ (LSGM8282) and Gd-doped ceria (GDC) solid electrolytes was prepared by electrophoretic deposition (EPD) and subsequent sintering. Here, GDC was inserted between LSGM8282 and the NiO anode containing YSZ (NiO-YSZ) or Co-containing cathode to prevent their reactions and the cracking caused by their different thermal expansions. EPD is a colloidal process used in ceramics, in which ceramic bodies are directly shaped from a stable colloidal suspension. It has many advantages for fabricating multi-configuration cells. Additionally, EPD makes it easy to control the thickness and morphology of a deposited film. A cross-sectional scanning electron microscopy image of the EPD cell4) showed layers in close contact. The performance of the cell was stable, but the power was below that theoretically expected for the cell. The electrolytes were reasonably dense and gas-tight, but the cell had a relatively high electrical resistance.
In this study, we performed electron-probe microanalysis (EPMA) and X-ray fluorescence (XRF) mapping of the cross-section of the cell to examine the layered structure. Scanning XRF microscopy was previously used to analyze the elemental composition of various samples at microscopic scales5). Through XRF observation, an X-ray absorption spectrum can be obtained when the energy of the incident X-rays is scanned. By scanning the X-ray beams, it is also possible to construct two-dimensional maps of the electronic and chemical states of materials.6) We also used X-ray diffraction (XRD) to assess the reactivities of the constituent compounds. La-doped ceria (LDC) was examined as a substitute for the GDC interlayer and was synthesized by oxalate co-precipitation. The LSGM reactivity of LDC was investigated with respect to the degree of La doping.
An SOFC layer composed of NiO-YSZ/GDC/LSGM8282/GDC was prepared by an EPD method and subsequent sintering as follows. NiO-YSZ and LSGM8282 were used as the anode and electrolyte, respectively, and GDC was used as the buffer layer to sandwich LSGM8282. First, a presintered NiO-YSZ plate was prepared, and polypyrrole (a conducting polymer) was coated onto the plate to create a conductive surface. LSGM8282 (AGC Seimi Chemical Co., Ltd.; average particle size, d = 0.57 μm) and GDC (Anan Kasei Co., Ltd., d = 0.30 μm) powders were separately dispersed in ethanol with a small amount of polyethylenimine as a dispersant, and a layered structure was formed by sequential deposition of GDC, LSGM8282, and GDC (in that order). After the lamination process, the deposited layers were dried in air and then co-sintered at 1400℃ for 2 h. More details of the cell preparation procedure are available in the literature4).
In this study, the cell was used for analysis before arranging the cathode. For elemental analysis, part of the cell was embedded in Wood's metal and cut across the stacked layers at a 45° angle. Wood's metal is a fusible alloy and was used here to fix the cell at the proper angle and to provide conductivity to prevent charge up. The declination was set so that it enlarged the outcrop area of each layer, which were <20-μm thick. The surface of the cross-section was polished using a lapping film sheet. EPMA with a wavelength-dispersive spectrometer (JEOL, Tokyo, Japan, JXA-8200) was used to determine the distribution of the constituent elements of the layers. A 15-kV electron beam of approximately 1 μm was scanned in 0.50-μm steps in both X and Y.
The cross-section slice was also analyzed using X-ray absorption near-edge structure (XANES) analysis. The experiments were carried out using highly brilliant synchrotron X-rays on the BL-4A beamline at the Photon Factory, KEK (High Energy Accelerator Research Organization) in Tsukuba, Japan. Horizontal X-ray beams from a bending magnet were monochromatized using a Si(111) double-crystal and focused using polycapillary half-lens optics (X-Ray Optical Systems, Inc., USA). The beam size at the focal point was less than 30 μm. The specimen surface was set vertically at the focal point of the X-ray. The angle of X-ray incidence was adjusted to about 45° to the surface. The specimen had layers inclined at 45° with respect to the surface, so the orientation of the plane of the layers was set parallel to the beam. A Si(Li) detector was set in the horizontal plane and received XRF perpendicular to the incident X-ray beam. The X-ray beam and focal point were fixed throughout the measurements, and the specimen was scanned in a plane parallel to the surface. The elemental distribution was observed across the SOFC layers using the K or L XRF line intensities of the elements. XANES spectra were measured at various points on the layers using the XRF yield mode. The incident X-ray energy was scanned around the Sr K and La LIII absorption edges. The XRF intensity was integrated for 10 or 50 s for each energy point, depending on the intensity. As a reference compound, SrLaGa3O7 was synthesized from a mixture of La2O3 (High Purity Chemicals, 99.99%), CaCO3 (High Purity Chemicals, 99.99%), and SrCO3 (High Purity Chemicals, 99.9%) by firing in air at 1100℃ for 6 h, and then at 1370℃ for 12 h. LDC was prepared as described in section 2.2. XANES spectra of the pure powders of some reference compounds were also obtained by measuring their X-ray absorption using transmission or XRF yield mode.
2.2 X-ray diffraction analysis of compounds formed between the different layersThe SOFC cell was also analyzed by XRD using a powder diffractometer (Rigaku, MiniFlex600) with CuKα radiation (λ = 0.15418 nm). The cathode-side GDC layer surface was set parallel to the sample holder; that means, it was in contact with the focusing circle of the diffractometer and diffractions were observed from the surface in reflection geometry.
The reactivities of the different layers were also examined by thermal treatment of the raw material powder mixture. LSGM8282 (AGC Seimi Chemical Co., Ltd., d = 0.57 μm, referred to simply as LSGM hereafter) and GDC (Anan Kasei Co., Ltd., d = 0.30 μm) or other constituting oxide powders were mixed in certain molar ratios in a mortar and pelletized using a press. The pellets were heated at 1000℃ for 12 h in air. After removal from the furnace, they were pulverized, and the resulting powders were characterized by XRD using a powder diffractometer (MAC Science, MX-Labo) with CuKα radiation.
As an alternative to GDC, LDC (LaxCe1−xO2−δ) samples with several different La:Ce ratios were synthesized by an oxalate co-precipitation method. LaCl3·7H2O (Wako Pure Chemical Industries, Ltd., Osaka, Japan; 99.9%) and CeCl3·7H2O (Wako Pure Chemical Industries, Ltd., >97%) were dissolved in purified water and mixed in appropriate proportions. Aqueous solutions containing 0.04 mol/L (La, Ce)Cl3 were dropped into a 0.1 mol/L oxalic acid ethanol solution, which was prepared from oxalic acid dehydrate (Wako Pure Chemical Industries, Ltd., 99.5%) and ethanol (Wako Pure Chemical Industries, Ltd., 99.5%). The resulting precipitates were fired at 1100℃ for 10 h. The products were identified by XRD.
The obtained LDC and LSGM powders were mixed in 1:1 molar ratios, and pelletized using a press. The pellets were heated at 1100℃ for 10 h in air. The resulting powders were characterized by XRD.
A compositional image was obtained in backscattered electron mode for a cross-section of the SOFC cell, which was simultaneously observed using EPMA, and shows a layered structure corresponding to NiO-YSZ/GDC/LSGM/GDC (Fig. 1(a)). Although the layers in the sample are not parallel to the direction of the incident electron beam, contamination from the neighboring layers was not a problem because the analytical depth, calculated from the depths of electron penetration and escape of characteristic X-rays, was less than 1 μm. The distributions of Gd and Ce in the EPMA images shows no serious diffusion of the elements from either of the GDC layers (Fig. 1(b)). Because the ionic charge of Ce (tetravalent) is large, it is firmly fixed. For LSGM, almost no diffusion of Ga and Mg from the layer is detected. However, as shown in Fig. 1(c), extensive diffusions of La and Sr from LSGM are observed. Both La and Sr are detected outside the LSGM and form new layers. On the cathode side (right side in the figures), Sr forms a concentrated layer near the GDC border. Ga accumulation is observed in the Sr-concentrated layer, while La is more diluted than in the LSGM layer body. The sharp layer composed of La, Sr, and Ga suggests chemical interaction between these elements. A third phase, SrLaGa3O7 (discussed below) forms at the LSGM/GDC interface. It is generated from LSGM by drawing out La. Some La moves from LSGM and penetrates GDC because La can dissolve in fluorite-type ceria to form a solid solution, i.e., La-doped GDC or ceria codoped with La- and Gd.
Elemental distribution on a cross-section of solid-oxide layers, as observed by EPMA: (a) compositional image in backscattered electron mode, (b) distribution of Gd, Ce, and Mg, (c) distribution of La, Sr, and Ga, and (d) distribution of Ni, Y, and Zr. The line patterns are the signal intensities along the transversal line in the center.
On the opposite anode side, diffusion is more severe, and a lot of La spreads toward the anode. La diffuses into GDC and becomes concentrated at the GDC/NiO-YSZ interface, in particular. La appears to react with Zr in NiO-YSZ. The migration of Sr is greater than that of La, and the high-level of Sr originally in the LSGM passes through the GDC and penetrates deeper into the NiO-YSZ anode. Ionic transfer is easier for Sr and Ni than the other constituent ions because Sr and Ni have smaller ionic charges (divalent). Additionally, it is easier for divalent Sr to form certain compounds, such as SrCeO3 and Sr3Zr2O7 (which are identified in section 3.2), because the divalent-ion sites in these compounds are a better fit for Sr than the trivalent La site in LaGaO3 (LSGM). Thus, Sr elutes out of LSGM. Near NiO-YSZ, Sr reacts with Zr to form a complex oxide, such as SrZrO3 or Sr3Zr2O7. EPMA images of Ni, Y, and Zr (Fig. 1(d)) show that the NiO-YSZ layer is not rigid, even though it is a presintered substrate. Zr is exuded from the NiO-YSZ surface and is concentrated inside of GDC, forming a new interface. The concentrated edge overlaps with those of La and Y. YSZ has a fluorite-type structure and Zr and Y both dissolve in fluorite-type ceria with La to form a solid solution. Additionally, La and Zr can form complex oxides such as pyrochlore-type La2Zr2O77) or defect-fluorite-type LaxZr1−xO2−x/28). Ni that was exuded by Ni-YSZ diffuses through GDC and aggregates at the GDC/LSGM interface. Ni reacts with Ga from LSGM and forms a spinel-type NiGa2O4, as described in section 3.2.
3.2 XRD analysis of the compounds formed from layer componentsXRD patterns of the SOFC cell, thermally-treated-mixture of LSGM, GDC, and NiO (1:1:1 w/w/w), and that of LSGM, YSZ, and NiO are shown in Fig. 2. Though the XRD pattern of SOFC is complicated with overlapping reflection lines of multiple compounds, each compound is identified while comparing the XRD data. SrLaGa3O7, Sr3Zr2O7, SrZrO3, La2Zr2O7, SrCeO3, and NiGa2O4 are detected in addition to the oxides used to fabricate the SOFC (i.e., LSGM, GDC, NiO, and YSZ). Reflection lines of fluorite-type compounds overlap each other; therefore, separation discrimination is impossible. However, low-angle-shifted GDC and YSZ XRD lines could be observed for LSGM+GDC+NiO and LSGM+YSZ+NiO, respectively, which must be La-doped GDC and YSZ, respectively. La-doped GDC, La-doped YSZ, and Zr-doped GDC can form in the SOFC cell. La2Zr2O7 forms a prochlore-type structure, which is a superlattice structural derivative of the simple fluorite structure. The main reflections of La2Zr2O7 overlap with the reflections of the fluorite-type compounds, but superlattice reflections, such as 311 (2θ = 27.2°), 331 (36.0°), and 733 (70.9°) in Fig. 2 indicate the existence of La2Zr2O7. However, other cations are likely also blended in these fluorite-type and fluorite-related compounds. A binary mixture of LSGM and NiO can form complex oxides, such as LaNiO39); however, LaNiO3 is not detected in this study. The reflection lines of LaNiO3 would be superimposed on those of LSGM and would be observed as split lines because of distortion of the perovskite lattice.
XRD patterns of an SOFC cell and thermally treated powder of LSGM with other SOFC constituting oxides.
When the 2θ angles of the LSGM lines of SOFC, LSGM+GDC+NiO, and LSGM+YSZ+NiO are compared (Fig. 2), those of SOFC and LSGM+GDC+NiO shift toward a slightly higher angle, while those of LSGM+YSZ+NiO are not affected. This may be caused by the escape of La from LSGM. To confirm the reactions of LSGM and GDC, the raw powders were dispensed and thermally treated. The XRD pattern of a heated powder consisting of LSGM:GDC = 7:3 (molar ratio) is shown in Fig. 3. The XRD lines of GDC shift toward significantly lower 2θ angles. Extra lines appear with the GDC and LSGM lines; these are attributed to a melilite-type oxide, SrLaGa3O710). When the mixing ratio of LSGM:GDC is 7:3, the XRD intensities of SrLaGa3O7 are higher and the shifts of the GDC lines are larger than those obtained using other ratios. La can easily be removed from the A sites of perovskite-type compounds. La3+ can substitute for Ce4+ in the eight-coordinated sites of CeO2, as observed for Gd3+. The solid solubility of La3+ in CeO2 can be up to 50 mol%11). Therefore, La diffusion from LSGM to GDC could be easily induced. According to the phase diagram of LaO1.5–SrO–GaO1.512), a La deficiency in the cubic perovskite phase (ABO3-type LaGaO3) forms SrLaGa3O713). The ionic radius of eight-coordinated La3+ is 0.116 nm, whereas those of Gd3+ and Ce4+ are 0.105 nm and 0.097 nm, respectively14). Although electroneutrality forms oxygen vacancies with the substitution of Ce4+ by lower-valence La3+, the larger cation expands the cell. Therefore, diffusion of La into GDC expands the unit cell.
XRD patterns of a thermally treated mixture of LSGM and GDC in a ratio of 7:3 and the raw powders.
XRF spectra were observed from synchrotron X-ray beam scans across the SOFC cross-section layers. Because the cross-section of the incident X-ray beam is larger than that of each layer, the observed boundaries of the layers are not clear, but the distribution features of each element could be reconfirmed by the XRF intensities. Diffusion of Ni and Zr toward the LSGM and GDC layers is obvious and was also observed using EPMA. The distributions of La and Sr are displaced from the distribution area of Ga (LSGM electrolyte area).
At several points across the layers, XRF spectra are observed while scanning the energy of the incident X-rays. The integrated intensity of the Sr Kα line is collected for each energy point around the Sr K absorption edge and plotted against the energy, thus giving Sr K XANES spectra. La LIII XANES was also observed by measuring the La Lα intensities. Figures 4 and 5 show the Sr K and La LIII XANES spectra, respectively, for several points across a cross-section of the GDC/LSGM/GDC/NiO-YSZ layers. The spectra were processed using the moving-average method to smooth the statistical fluctuations attributed to weak XRF intensities.
Sr K-edge XANES spectra across a cross-section of SOFC layers. A to E correspond (in alphabetical order) to the region from the cathode side to the NiO-YSZ anode side. LSGM and SrLaGa3O7 were measured as references in transmission mode.
Figure 4 shows the spectral set observed by scanning the X-ray irradiation points on a cross-section of the cell from the cathode side (A) through LSGM to the NiO-YSZ anode side (E) in 20-μm steps. Reference spectra of LSGM and SrLaGa3O7 are also shown in the figure. A marked difference is observed between the spectra for the first and second peak positions at 16.12 and 16.15 keV (2.583 × 10−15 and 2.588 × 10−15 J), respectively. These peaks are shifted to higher energies for SrLaGa3O7. Spectrum A was measured at a point outside the GDC on the cathode side (on the Wood's metal). At this point, the incident X-ray, which has a rather wide beam (30 μm in diameter), grazes the GDC edge. Therefore, the Sr XRF intensity is weak, but the signal from the Sr that was transferred toward the GDC could be selectively detected at some level. Spectrum A shifts slightly to higher energies. The second peak of spectrum A is broader toward higher energy than in spectrum C, which was recorded at the center of LSGM. This is consistent with the EPMA result, which suggests formation of SrLaGa3O7 on the cathode side. In LSGM, Sr is surrounded by 12 O atoms at 0.2768 nm (calculated from the crystal data15)). In contrast, in SrLaGa3O7, Sr is surrounded by eight O atoms at distances ranging from 0.251 to 0.294 nm16) (average of 0.2658 nm). Shorter atomic distances correspond to longer oscillation periods in the X-ray absorption spectra. The shift of the second peak indicates an extension of the period and shortening of the atomic distance. Spectrum E, which was recorded at a point on NiO-YSZ, show a large shift in magnitude of the second peak. It is attributed to the formation of SrZrO3, which was also suggested by the EPMA and XRD analysis. In SrZrO3, shorter atomic distance can be observed17); four of the Sr–O distances are 0.2094 nm and two of them are 0.2084 nm. The characteristic small peak that was reported for SrZrO3 at about 20 eV above the absorption edge18) was not observed in the spectra because the energy resolution in the spectra was reduced and the fine structure was lost because of statistical fluctuations and smoothing of the XRF signals.
The La LIII XANES spectra (Fig. 5) were measured in the same manner as that for Sr K. The SOFC was observed across the cell from the cathode side (F) to the NiO-YSZ anode side (I) in 20-μm steps, although the analysis positions are not identical to those used for Sr XANES. The LDC10 (La0.1Ce0.9O2−δ) reference spectrum shows a small hump at 20 eV above the absorption edge. In addition, the peak at 5.52 keV is higher in energy than that in LSGM. Spectra F and G are analogous to that of LSGM. In contrast, the features of spectra H and I, which are near the NiO-YSZ point, are similar to those of the LDC10 spectrum, although they show large fluctuations because of the weak XRF intensity. The XANES study supports that La diffuses from LSGM and dissolves in GDC, forming a solid solution, i.e., La-doped GDC.
La LIII-edge XANES spectra across a cross-section of SOFC layers. F to I correspond (in alphabetical order) to the region from the cathode side to the NiO-YSZ anode side. Spectra of LSGM and LDC10 (La0.1Ce0.9O2−δ) are used as references.
As described above, SrLaGa3O7 is likely formed at the cathode side. However, the La LIII spectrum of SrLaGa3O7 is similar to that of LSGM. In addition, it would be buried by the spectra from La in LSGM because major amounts of La remain in LSGM. Ni K XANES spectra were also measured, but the shapes of the spectra are the same as those of the Ni oxides. Spectral differences among NiO and other oxides, such as NiGa2O4, could not be detected under the present conditions.
3.4 LDC synthesis and reaction with LSGMThe XRD patterns of LaxCe1−xO2−δ (x = 0.1–0.7, LDC10–LDC70) prepared by the oxalate method are shown in Fig. 6. Here, the composition is based on the mixing ratio of the starting materials (LaCl3 and CeCl3), and LDCX indicates X% substitution of Ce by La. Because of La doping, the XRD lines shift to lower angles than those for pure CeO2. The shift becomes larger as x increases. When x is greater than 0.60, XRD lines of undissolved La2O3 are observed. The unit cell parameter a of each LDC sample (LDC10–LDC60) calculated from the angles of the 111, 200, 220, and 311 lines of the XRD patterns, are listed in Table 1. Because the value of a for each value of x is almost consistent with that reported in the literature11), La substitution using the oxalate method was judged to be successful. The upper limit of the solid solubility using the oxalate method is around x = 0.60.
XRD patterns of LaxCe1−xO2−δ (x = 0.1–0.7, LDC10–LDC70) synthesized by oxalate co-precipitation and subsequent sintering. The pure CeO2 pattern is for reference.
| Composition, x | Cell parameter, a/nm |
|---|---|
| 0 | 0.54090(7) |
| 0.1 | 0.54419(9) |
| 0.2 | 0.54759(8) |
| 0.3 | 0.5501(5) |
| 0.4 | 0.5541(2) |
| 0.5 | 0.5585(3) |
| 0.6 | 0.5625(2) |
In Fig. 7, the cell parameters (a) are plotted against the solid solubility (x). The cell parameter increases approximately linearly with x. The linearity between the cell parameters and compositions is known as Vegard's law for alloys and solid solutions. Based on Pauling's rules, the sum of the ionic radii determines the bond length; thus, the larger ionic radius of La3+ causes a longer bond length and larger unit cells. According to the radius ratio of La3+ and O2−, a fluorite-type structure with coordination number eight would be stable, although it would be a defect-type structure (C-type rare-earth structure) under Pauling's electrostatic valence rule. However, oxides of light rare-earth metals have an A-type structure at ambient temperatures; in La2O3, La3+ is surrounded by seven O2− with different bond lengths. For substitutional solid solutions, the formation rules that require matching ionic radii, crystal structures, and valences are applicable, and are similar to the well-known Hume–Rothery rules for binary alloys. If these conditions are all satisfied, complete solubility may occur throughout the composition. However, in the present solid solution, the crystal structures of CeO2 and La2O3 are different; thus the compositional range of the solubility is limited for LDC.
Compositional dependence of the cell parameter (a) of LaxCe1−xO2−δ.
XRD patterns of heated powder mixtures of synthesized LDC and LSGM are shown in Fig. 8. The formation of SrLaGa3O7 and cell expansion of the unit cell of LDC are observed when x of LDC is small, which is the same as that observed for the GDC and LSGM mixtures. Therefore, La defects in LSGM and diffusions to the ceria phase could not be inhibited. With increasing x, the intensities of the SrLaGa3O7 lines decrease. SrLaGa3O7 is undetectable for x ≥ 0.50. However, instead of SrLaGa3O7, SrLaGaO4 forms, and the XRD lines of LDC shift in the opposite direction, i.e., to higher 2θ. As shown in Fig. 8, all the diffraction lines of LDC20–LDC60 move toward the angles seen for LDC40 (Fig. 6). SrLaGaO4 is an A2BO4-type perovskite. According to the phase diagram13), the A2BO4 phase appears if the La concentration in LSGM (ABO3-type perovskite) is enriched.
XRD patterns of thermally treated mixtures of LSGM and LDC in a 1:1 ratio.
From the above, LSGM is unstable if it is buffered by a ceria-based interlayer; however, LDC40 seems to be better because the La concentration in LDC40 is in equilibrium with that in LSGM. Although ceria-based interlayer contacts are close and tight between the electrolyte and the electrode, and can work as buffer layers, this is not sufficient to inhibit the elemental diffusion and the reaction between LSGM and NiO-YSZ.
The formation of extra phases such as SrLaGa3O7 is inconvenient. Recently, ionic conductivity originating from two-dimensional structures was reported for SrLaGa3O719). However, this phase has a lower conductivity than LSGM and degrades the SOFC performance. The formation of perovskite-type SrZrO3 and spinel-type NiGa2O4 can cause insufficient cell conductivity. The former is a dielectric20), and the latter is a wide-gap semiconductor21).
EPMA, XANES, and XRD studies of SOFC layers of NiO-YSZ/GDC/LSGM8282/GDC prepared by sequential EPD stacking and co-sintering indicate significant diffusion of elements and formation of extra phases. La ions diffused from LSGM and dissolved in GDC; they moved easily in the ceria-based phases. Other constituent ions also dissolved or passed through GDC and formed complex oxides. The ceria-based buffer layer kept a large portion of La in the LSGM but allowed some La to migrate, forming extra phases, namely SrLaGa3O7 or SrLaGaO4. Along with the LSGM-derived phases, elemental (e.g., Sr, Ni, Y, and Zr) diffusion also formed resistive phases. As described above, the ceria-based layer shows insufficient resistance to elemental diffusion to act as a blocking layer between LSGM and NiO-YSZ. When the SOFC structure was prepared, each layer was sequentially deposited by EPD, and all layers were sintered at the same time. Because of the monolithic firing, elemental diffusion may have occurred before the synthetic powders in each layer were well fused and fixed. If the layer stack was produced by deposition and sintering in a layer-by-layer manner, elemental diffusion should be suppressed and high-performance SOFCs should be produced. However, as an efficient industrial process, monolithic firing is a key process; therefore, the EPD process and the co-sintering temperature and times should be optimized. Moreover, to improve the performances of lanthanum-gallate SOFCs, the composition of the buffer layer or electrode must be explored further while considering long-term durability, because elemental diffusion is detrimental to the performance.
The authors would like to thank Drs. H. T. Suzuki, F. Munakata, and T. Uchikoshi for supplying the SOFC sample. We would also like to thank Prof. A. Iida (Photon Factory, BL-4A) for assistance during the synchrotron experiments. The experiments were performed with the approval of the Photon Factory Program Advisory Committee (Proposal No. 2010G558).
