Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013) 0013-4651/2013/160(11)/F1293/12/$31.00 © The Electrochemical Society

F1293

Three-Dimensional Microstructural Evolution of NiYttria-Stabilized Zirconia Solid Oxide Fuel Cell Anodes At Elevated Temperatures David Kennouche,a Yu-chen Karen Chen-Wiegart,b J. Scott Cronin,a Jun Wang,b and Scott A. Barnetta,∗,z a Department b Photon

of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208, USA Science Directorate, Brookhaven National Laboratory, Upton, New York 11973, USA

The microstructural evolution of Ni - Yttria-Stabilized Zirconia (YSZ) anode functional layers in anode-supported solid oxide fuel cells was studied after aging in humidified hydrogen at temperatures from 900–1100◦ C and times up to 500 h. The relatively large (∼6000 μm3 ) three-dimensional transmission X-ray microscopy images provided good statistics in measured microstructural data. Feature sizes in the YSZ and Ni phases changed little until the highest temperature and time, whereas the pore phase feature size increased for all temperatures and times. Ni-YSZ interfacial area increased at the expense of pore-YSZ interfaces after moderate annealing, with a general decrease in interfacial areas observed at longer times and higher temperatures. Three-phase boundary (TPB) density decreased rapidly initially, but then more slowly, with increasing annealing temperature and time. Electrochemical impedance spectroscopy measurements showed a corresponding increase in the anode response associated with electrochemical hydrogen oxidation at TPBs. A higher fraction of isolated pores and larger pore tortuosity was observed at intermediate annealing temperatures and times, with a corresponding increase in the gas diffusion impedance response. © 2013 The Electrochemical Society. [DOI: 10.1149/2.084311jes] All rights reserved. Manuscript submitted August 22, 2013; revised manuscript received October 1, 2013. Published October 10, 2013.

Solid oxide fuel cell (SOFC) systems are being developed mainly for stationary power generation, where they should operate with minimal performance loss at elevated temperature for at least 40,000 hours.1 Achieving minimal long-term degradation is thus critical to the commercialization these systems. Coarsening of Ni particles in nickel – yttria stabilized zirconia (Ni-YSZ) anodes has been reported as a reason for SOFC performance degradation.2–4 While these studies have mostly quantified Ni coarsening, it is also desirable to have a more complete description of Ni-YSZ anode structural evolution including the YSZ and pore phases, since the electrode polarization resistance depends on many structural factors. Furthermore, Ni-YSZ anodes with different characteristic microstructures – including the volume fractions of the solid and pore phases, particle sizes, etc. – will presumably evolve differently for a given set of operating conditions. For example, Cronin et al.5 suggested that anodes with low Ni content and low pore volume are less susceptible to Ni coarsening. Thus, it is important to provide a full description of the anode microstructure in studies of microstructural evolution. Focused ion beam – scanning electron microscopy (FIB-SEM) and transmission X-ray microscopy (TXM) tomography have recently yielded quantitative three-dimensional (3D) electrode microstructures, from which electrochemical performance can be predicted either via 3D simulations6,7 or by extracting macrohomogeneous structural parameters that can be input into electrochemical models.8–10 FIBSEM has been used to quantify microstructural changes due to elevated temperature in composite Ni-GDC11,12 and Ni-YSZ anodes.5,13 However, there are few reports where anode microstructural evolution was correlated with electrochemical performance. Recently Jiao et. al used FIB-SEM to analyze microstructural changes of Ni-YSZ electrodes that had been operated with current, and correlated with electrochemical performance changes.14 Recently, experiments have been done observing the same Ni-YSZ sample volume before and after annealing, taking advantage of the non-destructive TXM measurement.15 Such experiments are difficult and have not yet been used for systematic studies of structural evolution. This paper describes the effects of annealing temperature and time on the microstructure of Ni-YSZ anode functional layers (AFLs) in anode-supported SOFCs, as measured using TXM 3D tomography. Electrochemical impedance spectroscopy (EIS) measurements of SOFC button cells containing annealed anodes are also presented, and discussed in terms of the structural changes. Compared with a ∗ z

Electrochemical Society Active Member. E-mail: [email protected]

prior report where imaging was done using FIB-SEM,5 the present results provide a more detailed and accurate description of the structural changes and their effect on anode polarization resistance. The present TXM images are of considerably larger volumes, providing more accurate structural information, especially phase intra-connectivity. Furthermore, more temperatures and times are analyzed, and a detailed explanation of how the structural changes impact specific responses in electrochemical impedance spectra is developed. Effects of cell operating currents and high steam content, which are known to affect anode degradation,11,14,16 are not considered. Experimental Cell preparation and annealing.— Anode-supported SOFCs were prepared as described elsewhere.5 The support was a 750 μm thick dry-pressed pellet with composition 50:50 wt% NiO:YSZ along with 10 wt% graphite pore former. After bisque firing the support at 1135◦ C, a 20 μm thick 50:50 wt% NiO:YSZ anode functional layer and then a 10 μm thick YSZ electrolyte layer were deposited by drop coating. This structure was subsequently fired at 1400◦ C for 4 h. A 15–20 μm porous YSZ layer was then deposited on the YSZ electrolyte by screen-printing and fired at 1200◦ C for 2 h. Note that all thicknesses given above are measured after firing. Note that a number of prior Ni-YSZ 3D tomography studies have utilized Ni-YSZ processed by the above methods;5,10,17,18 in all cases, cell-to-cell variations were small enough to allow useful comparisons between differentlyprocessed anodes. The above half-cells were then reduced at 800◦ C for 3h in a 4% H2 / 3% H2 O / 93% Ar gas mixture, long enough to fully reduce the NiO. One cell was not annealed further, providing the baseline case of an as-reduced AFL at the beginning of cell operation. Other halfcells were annealed in the same gas mixture for the various times and temperatures listed in Table I. This gas composition is consistent with previous reports on Ni-YSZ annealing;2–4,19,20 the mixture is reducing enough to maintain the Ni in the metallic state and has a H2 content below the explosion limit, conducive to laboratory safety. Cathodes were formed, after annealing or reduction, using an identical procedure for all cells. The porous YSZ scaffold was infiltrated with a 2M solution containing samarium, strontium, and cobalt nitrates with amounts designed to yield the composition Sm0.5 Sr0.5 CoO3-δ (SSC).21 Each sample was infiltrated 6 times, resulting in a 2.5 mg SSC deposition over the 0.5 cm2 cathode. Cathode calcination was done in air at 800◦ C at the beginning of cell testing; subsequent analysis using X-ray diffraction indicated that the desired perovskite SSC phase was

Downloaded on 2013-10-24 to IP 130.199.3.165 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

F1294

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

Table I. Half-cell annealing or reduction conditions utilized in the study. Temperature

Time

800◦ C 900◦ C 1000◦ C 1100◦ C 1100◦ C 1100◦ C

3h 500h 500h 500h 100h 20h

produced. This procedure avoided changes to the anode that might occur during conventional cathode processing with a higher temperature firing step, and was shown to produce low-resistance cathodes. The low resistance and similar nature of the cathodes makes it more straightforward to discern the changes in cell performance caused by anode annealing. Samples were prepared for TXM measurements by first fracturing the cells with a diamond tip, and mounting one of the resulting pieces on a stub for focused ion beam milling. Cylindrical samples with a diameter of ∼35 μm and height of ∼80 μm were produced, and then lifted out and attached using a Pt weld to a W needle, suitable for mounting on the TXM sample manipulator.10 Additional electrolyte supported symmetrical cathode cells were processed in order to isolate the electrochemical response of the SSC cathode. 750 μm thick YSZ pellets were pressed and sintered at 1400◦ C for 4h. Porous YSZ scaffolds (15–20 μm) were then deposited on each side of the YSZ pellet by screen-printing and fired at 1200◦ C for 2 h. The porous YSZ scaffolds were infiltrated with the same SSC nitrate solution and procedure as described above. Electrochemical characterization.— A Zahner IM6 electrochemical workstation was utilized to collect Electrochemical Impedance Spectroscopy (EIS) spectra in the frequency range from 100 mHz to 1 MHz. Cells were maintained at open circuit during the measurements. EIS testing was done prior to any electrode polarization, starting at 800◦ C and reducing to 600◦ C, with an anode fuel flow rate of 50 sccm and a fuel composition of 3% H2 O, 10–97% H2 , balance Ar. After EIS measurements were completed, current–voltage characteristics were collected at 10 mV increments from open circuit voltage to 0.4 V. Structural characterization.— Nano-tomography experiments were carried out at the transmission X-ray microscopy (TXM) beam-

line X8C, National Synchrotron Light Source, Brookhaven National Laboratory.22 TXM is a full-field imaging technique providing a sub30 nm 2D resolution with tomography capability. Details on the use of this method for 3D tomography of Ni-YSZ anodes have been reported previously.10,17 Two tomography datasets were collected on the same sample, with different X-ray energies below (8300 eV) and above (8400 eV) the Ni K-edge, to facilitate the segmentation process. The alignment of the X-ray projections was performed automatically using a capacitive sensor-based metrology system.22 A standard Filtered Back Projection Reconstruction algorithm was used to reconstruct the 3D structure, and extract a stack of virtual cross-section slices from each tomography dataset.23 Each reconstructed volume was a cylinder 40 μm in diameter and height, with the voxel size of 39.6 nm. Image processing.— Prior to segmentation, the highest quality volumes were selected and cropped from each of the TXM datasets. Figure 1a shows an example of a typical 2D image section along with the corresponding gray scale histogram. For all data sets, a normalization algorithm was run such that the histograms contained no lateral shifts from one image to the next; to do so a Gaussian was fit to the largest peak in each histogram and the grayscale values shifted so that the peaks all matched. Next, the brightness variation of the images was normalized along the vertical and horizontal axis to further improve consistency in the histograms. Finally, the peaks in image histograms were fit and stretched to specific gray scale values, as seen in Figure 1b, to ensure similar histograms for all images in a data set, thus allowing a constant threshold value to be utilized in segmentation. This also allowed similar threshold values to be used for all data sets analyzed in this study, improving segmentation consistency. Two different methods were used for segmentation. When the above-edge image contrast was high enough, standard thresholding was employed. Figure 2a shows an example of an un-processed crosssection image for the AFL annealed at 1100◦ C for 20h, and Figure 2b shows the corresponding histogram that exhibits clear separation of the three peaks. Segmentation using the threshold values shown yielded the image shown in Figure 2c. For some data sets, the separation of the three histogram peaks was not as clear and so both above- and belowedge images were used in segmentation, as illustrated in Figure 3 for the AFL annealed at 1100◦ C for 500 h. The above-edge image, shown in Figure 3a, was used to identify and segment the Ni phase. In this data set, the contrast between porosity and YSZ was not sufficient for threshold segmentation. Thus, the above-edge image was subtracted from an inverted below-edge image (Figure 3b), yielding accurate identification of pores. The resulting segmented pore and Ni phase

Figure 1. Example virtual 2D image from a TXM data set along with the corresponding grayscale histogram (a). Altered version of the same image after stretching of the grayscale histogram (b).

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

F1295

Figure 3. Visual description of the segmentation method using image data sets taken with X-ray energies above and below the Ni K-edge, for the AFL annealed at 1100◦ C for 500 h. (a) shows a typical 2D image from the 3D data set, (b) shows the inverted below edge image, and (c) shows the segmented image (Ni = white, YSZ = gray, and pore = black).

Figure 2. Visual description of the segmentation method using only the image data sets taken with X-ray energy above the Ni K-edge, for the AFL annealed at 1100◦ C for 20h. (a) shows a typical 2D image from the 3D data set, (b) shows the corresponding grayscale histogram, and (c) shows the segmented image (Ni = white, YSZ = gray, and pore = black).

were then added to a gray background to complete the segmentation process, with the resulting image shown in Figure 3c. In some cases, both of the above methods were used and they were found to yield nearly identical results. Microstructural quantification.— Macrohomogeneous morphological information including phase volume fraction, surface areas, particle size distributions, phase connectivity and TPBs, were calculated using methods described elsewhere.5,9,10,24 Algorithms were written in-house for the image process programming language, Interactive Data Language (IDL). Phase intra-connectivity was analyzed for all three phases of each composite electrode. ‘Connected’ phases were defined as follows: electronic conductor and pore networks that connected to the current collector, and ionic conductor networks that connected to the electrolyte. This is the so-called directional method,5,25 where the connection must be in one specific direction; this is different than

the standard criterion for percolation which can be in any direction. A network was labeled ‘isolated’ if it did not satisfy the above directional criterion, unless it intersected an edge of the measured volume such that the true connectivity was indeterminate and it was labeled ‘unknown’. Since unknown networks appear near the edges of the measured volume, larger-volume data sets have reduced unknown fractions, allowing for more accurate determination of the true quantity of connected vs. non-connected phases. The present TXM tomography data provides a significant increase in the measured volume, while maintaining good resolution, compared to prior FIB-SEM data.5,26 This made possible a further measure to eliminate most of the unknown connectivity networks, by removing the portions near the edges of the measured volume. The resulting extracted inner volumes were still 2500 to 6000 μm3 , large enough to maintain good statistical accuracy27 and were used for subsequent analysis and calculations. The volume fraction of unknowns was <5%. The electrochemicallyactive component of the TPB density was determined by calculating the true fraction of active and inactive TPBs by splitting the unknowns into active and inactive portions.9 Results and Discussion Microstructural measurements.— Figure 4 compares typical 2D images taken from the 3D data sets. The non-annealed AFL shown in (a) appears visually to have smaller features than the AFLs annealed at

F1296

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

Figure 4. Typical 2D segmented images (Ni = white, YSZ = gray, and pore = black) taken from AFLs after different annealing treatments: (a) nonannealed AFL, (b) 1100◦ C–100h, (c) 1000◦ C–500h, and (d) 1100◦ C–500h.

1100◦ C–100h (b), 1000◦ C–500h (c) and 1100◦ C–500h (d). Annealed AFLs exhibit larger features, especially the nickel and pores phases. Higher temperature and longer times induce stronger microstructural changes. Figure 5 shows the phase volume fractions for the different AFLs, normalized to the values expected based on the starting NiO and YSZ fractions, and assuming that the as-fired NiO-YSZ was fully dense such that the porosity derived only from NiO reduction. The deviations from the expected volume fractions were ≤3%, within the error levels previously observed in Ni-YSZ measured by 3D methods,9 and were presumably due to a combination of processing and measurement errors. There was no evidence of Ni loss even at the highest temperature, an effect that has been reported for high temperature and high steam conditions due to Ni-hydroxide formation and evaporation,11 but apparently is not important under the present conditions. The pore volume fraction was unchanged, indicating that there was no sintering during the anneals. Specific surface areas for each phase are shown in Figure 6. The pore surface area decreased with increasing temperature and time, with the most rapid decrease at early times. The Ni surface area increased initially with increasing time at 1100◦ C, before decreasing. The effect of annealing temperature was similar, with a higher Ni surface area for 900◦ C, but a lower value for higher temperatures. The YSZ surface area changed little at the lower temperatures, but decreased with increasing time at 1100◦ C, especially for shorter times. Overall, the Ni, pore and YSZ surface areas decreased respectively by ∼18%, ∼33%, and ∼19% (Table II) between the as-reduced state and the 1100◦ C–500h anneal.

Figure 5. Deviations of the phase volume fractions from expected values (27.61 vol% Ni, 53.06 vol% YSZ, 19.33 vol% porosity) for the various asreduced and annealed AFLs.

Figure 6. Specific surface areas of Ni, YSZ and pore phases versus temperature for 500 h (a) and time at 1100◦ C (b).

Figure 7 shows the evolution of the interfacial areas with temperature (a) and time (b). Ni-Pore interface area, which was relatively low because of the low Ni and pore volume fractions, decreased with both increasing temperature and time. The YSZ-pore interface area decreased similarly. The Ni-YSZ interface area increased after 100 h at 1100◦ C, and also for 900◦ C–500 h. For longer times and higher temperatures, however, the Ni-YSZ interface area decreased, perhaps indicating that coarsening became important. This is also indicated by Figure 6, which shows a general decrease in all surface areas at the highest times and temperatures. Figure 8 shows feature size distributions in the Ni, YSZ, and pore networks, obtained using the method described by Holzer et al.11 Figure 8 indicates that the YSZ phase grew measurably only after annealing at the longest and highest temperature, 1100◦ C for 500 h. Decreases in YSZ surface area were detected for shorter times at 1100◦ C (Figure 6b), so it can be concluded that the YSZ phase did not coarsen significantly below 1100◦ C. The feature size in the Ni phase initially remained constant and then grew with increasing time at 1100◦ C (Figure 8d), but there was no significant coarsening below 1100◦ C. The Ni and YSZ coarsening at longer times at 1100◦ C agrees with the decrease in Ni and YSZ surface areas in Figure 6. Pore network evolution was clearly occurring below 1100◦ C, as the feature size grew with increasing temperature and time. The considerable changes in interfacial areas at temperatures below 1100◦ C show that substantial structure changes were occurring even in the absence of Ni and YSZ coarsening. Perhaps the most surprising change is the increase in Ni surface area (Figure 6) and Ni-YSZ interfacial area (Figure 7) at intermediate temperatures and times. These changes apparently took place mainly in the Ni and pore phases, with little change in the YSZ. Figure 7a shows that the YSZ-Ni interfacial area increase occurs at the expense of YSZ-pore and Ni-pore. These shifts can presumably be explained by the system decreasing its overall surface energy by eliminating the higher surface energy solid-vapor interfaces; this process apparently involved an increase in lower-energy solid-solid (i.e., Ni-YSZ) interfacial area. Given that this change occurred while the average Ni feature size

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

F1297

Table II. Summary of the measured 3D structural data.

Volume (μm3 ) Ni Vol% YSZ Vol% Pore Vol% Ni SA (μm−1 ) YSZ SA (μm−1 ) Pore SA (μm−1 ) Ni Pore SA (μm−1 ) YSZ-Pore SA (μm−1 ) Ni-YSZ SA (μm−1 ) Pore Connected Total TPB Density (μm−2 ) Active TPB Density (μm−2 ) % Active TPB

800◦ C-IC

900◦ C–500h

1000◦ C–500h

1100◦ C–500h

1100◦ C–100h

1100◦ C–20h

5696 28.0 52.7 19.3 5.73 5.20 9.14 0.31 1.45 1.29 95.6 2.98 2.59 87

5978 27.8 53.3 18.9 6.59 5.39 7.38 0.17 1.22 1.66 86.7 2.09 1.65 79

5962 28.1 52.4 19.5 5.58 5.03 6.59 0.11 1.18 1.46 87.3 1.58 1.18 74

7340 26.8 53.4 19.8 4.71 4.21 6.16 0.12 1.10 1.15 94.6 1.21 1.04 85

5903 27.5 53.3 19.2 5.59 4.69 6.88 0.18 1.14 1.36 89.0 1.88 1.42 75

6148 28.4 52.6 19.1 5.20 5.28 8.62 0.17 1.47 1.30 95.5 2.21 1.91 86

remained constant (Figure 8), this indicates a change in shape of the Ni network. Analysis of the AFL Ni and YSZ phase networks showed that they were nearly 100% connected to the anode support and electrolyte, respectively (Figure 9). The AFL pore phase generally exhibited lower fraction connected to the anode support, with a minimum connectivity at intermediate annealing temperatures and intermediate times at 1100◦ C. These are the same conditions where the Ni and pore phase evolution, discussed above, occurs. Figure 10 depicts 3D reconstructions of the AFL pore phase showing the active, unknown, and isolated portions. The increased fraction of isolated pores at intermediate temperature is evident in the images. The pore connectivity changes shown in Figures 9 and 10 are relatively small but well above measurement errors (<1%).10 One other piece of evidence supporting a pore connectivity minimum comes from FIB-SEM measurements of annealed Ni-YSZ,5 which showed a maximum in isolated pores un-

Figure 7. Ni-YSZ, Ni-pore, and YSZ-pore interfacial surface areas versus temperature for 500 h (a) and time at 1100◦ C (b).

der these conditions, observed directly because they did not fill with impregnated epoxy. Figure 11 shows measured total and electrochemically-active TPB densities. The active TPB density is slightly lower, due to the isolated pores, but follows the same trend. The minimum in connected pore fraction (Figures 9 and 10) results in a relatively low active TPB density at 100 h for the 1100◦ C anneals (Figure 11b). The TPB density decreases rapidly initially with time and then more slowly (Figure 11b). Similarly, much of the TPB decrease occurs at annealing temperature well below 1100◦ C (Figure 11a). Thus, much of the TPB density decrease for moderate annealing must be related to the re-arrangements that occur in the Ni and pore phases, as discussed previously. For example, changes in interface roughness and network shapes – indicated by changes in Ni surface area that occur even while the feature size distribution is relatively unchanged – should also affect TPB length. The lesser TPB density decrease during the later stages of annealing, i.e., after 100 h in Figure 11b, should be directly related to the coarsening observed at this stage of annealing. Figure 12 shows tortuosity values that were computed for all phases using a method where a distance map is created and the tortuosity is computed by fitting the curve, as described elsewhere.28 The values given are averages of calculations that were done in all three directions in order to obtain better statistical accuracy; the values appeared to be isotropic. Ni and YSZ tortuosity values remained relatively low for all annealing conditions. The relatively low YSZ tortuosity values are as expected given the high YSZ volume fraction of ∼53%. But the pore tortuosity was higher for anodes annealed at intermediate temperatures and times, i.e., those that had a high fraction of isolated pores. This is reasonable because those AFLs had a lower connected pore volume fraction. Electrochemical measurements.— Figure 13 presents the currentvoltage characteristics of cells with as-reduced and annealed anodes measured at 800◦ C. The cell performance is similar to many reported anode-supported SOFCs, with maximum power densities 1 W/cm2 in most cases. There was some indication of a negative curvature of the potential curves at higher current densities, even though measurements were not taken below 0.4 V, perhaps indicating concentration polarization. Figure 14 shows Nyquist and Bode plots of the Electrochemical Impedance Spectroscopy (EIS) data from the same cells. All of the cells exhibited four clear impedance responses with magnitudes that varied considerably from cell to cell. This is similar to prior reports for anode-supported SOFCs where four or five responses were identified.29–31 The cell with the as-reduced anode had a lower overall resistance and yields higher power density than the annealedanode cells, but there was no obvious trend in these cell performance measures with annealing time or temperature. Equivalent circuit fitting, also shown in Figure 14, was used to quantify the individual EIS responses. Four RQ elements were used,

F1298

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

Figure 8. Normalized size distributions of the Ni, pore, and YSZ phases after annealing at different temperatures for 500 hours (a, b, and c) and for different times at 1100◦ C (d, e, and f).

Figure 9. Evolution of the phase intra-connectivity versus annealing temperature at 500 h.

matching to the four different responses seen by visual inspection. The response peaking at ∼ 104 Hz can be attributed to the SSC cathode, as indicated by separate measurements done on cathode-symmetric cells (Figure 15). Besides peaking at ∼ 104 Hz, the SSC response also

had a maximum –Zimag value of 0.025 cm2 , similar to the high frequency responses in Figure 14. This high frequency response did not vary much with annealing, as expected since the cathodes were prepared identically. The only exception is for the 1100◦ C 500 h anneal, discussed further below. Figures 16 and 17 show the EIS responses and circuit-model fits of the as-reduced cell measured at varying temperature and hydrogen partial pressure. The highest frequency peak increases and shifts with decreasing temperature but not with varying hydrogen partial pressure, as expected for cathode response. The responses at ∼0.5 Hz and 10 Hz vary with fuel composition but not temperature (Figures 16 and 17), suggesting they are related to gas diffusion. A similar ∼10 Hz response was observed previously for Ni-YSZ anode-supported cells and associated with gas diffusion.32,33 This response did appear to track with the AFL pore connectivity, showing a maximum value for 900◦ C and 1000◦ C anneals (Figure 14) where connectivity was lowest (Figure 9). Low connectivity reduces the active pore volume fraction ε, and also increases tortuosity τ due to a lower connected pore volume fraction, decreasing the effective gas diffusion coefficient Deff given by:

De f f =

εD τ2

[1]

Figure 10. 3D image renderings of the pore phase for the as reduced AFL (a) and for AFLs annealed for 500 h at 1000◦ C (b) and 1100◦ C (c). The colors indicate pore phase connectivity: red – isolated; yellow – unknown; and green - connected.

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

F1299

Figure 11. Evolution of total and electrochemically-active TPB density versus temperature for 500h anneals (a), and annealing time at 1100◦ C (b).

Figure 12. Evolution of tortuosity versus temperature for 500h anneals (a), and annealing time at 1100◦ C (b).

F1300

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

Figure 15. Bode plot of EIS data taken at open circuit in air at 800◦ C for a SSC cathode symmetric cell.

Figure 13. Potential and power density versus current density for cells with as-reduced and annealed anodes.

where D is the binary (H2 /H2 O) gas diffusion coefficient. Although the decrease in ε is relatively minor, the increase in τ is substantial, as shown in Figure 12. Note that this increased τ and Deff is associated

with the AFL. The AFL has been observed to contribute significantly to gas diffusion resistance in anode-supported cells,31 and the relatively low pore volume fraction of ∼19% in the present anodes (Table II) will yield a relatively low Deff . On the other hand, it is possible that structural changes occur in the thick Ni-YSZ anode support that are similar to those observed in the Ni-YSZ AFL, contributing to the observed changes in gas diffusion resistance. The response at

Figure 14. Nyquist (left) and Bode (right) plots of EIS data measured at 800◦ C at open circuit for cells with various as-reduced and annealed anodes. Also shown are fits made using an equivalent circuit (see inset) with four R-CPE elements, along with the contributions of each individual element.

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

F1301

Figure 16. Nyquist and Bode plots of the EIS data taken from the as-reduced cell at 750◦ C at open circuit voltage at different fuel Ar/H2 ratios (top panels). The data at each PH2 are shown individually in the lower panels, along with fits done using the same equivalent circuit as shown in Figure 14.

∼ 0.5 Hz does not vary with annealing and may be associated with gas conversion29 or gas diffusion through the anode support.31 The response at ∼300 Hz increases strongly and decreases in frequency with deceasing measurement temperature (Figure 17) and increases slowly with decreasing partial hydrogen pressure (Figure 16), suggesting that it is an anode electrochemical process. However, its evolution with annealing (Figure 14) doesn’t correlate with active TPB density. According to prior observations, it may also be associated with gas diffusion,32,33 and the changes with annealing do appear to correlate with the tortuosity. Leonide et al.31 indicated that responses at ∼103 and 104 Hz are associated with charge transfer and oxygen ion diffusion in Ni-YSZ. The former could be similar to

the 300 Hz response in the present EIS data. The latter could be a smaller response buried under the cathode peak. If this is the case, it could explain the larger response at 104 Hz for 1100◦ C 500h as a combination of the cathode response and an anode response that becomes large enough to observe where the TPB density is lowest (Figure 11). As a test of the above interpretation, Figures 18 and 19 present EIS data for as-reduced and annealed cells measured under conditions where a specific response should dominate. At high measurement temperature and low PH2 , the lower frequency gas diffusion response is large and the electrochemical processes at higher frequency relatively small, such that the cell resistance should follow the variation

F1302

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

Figure 17. Nyquist and Bode plots of the EIS data taken from the as-reduced cell at 97% H2 at open circuit voltage at different temperatures (top panels). The data at each temperature are shown individually in the lower panels, along with fits done using the same equivalent circuit as shown in Figure 14.

in pore connectivity and tortuosity with annealing. Indeed, the total polarization resistance in Figure 18 peaks for intermediate annealing temperature/time, showing a strong correlation with connectivity (Figure 9) and tortuosity (Figure 12). The other measurement condition, shown in Figure 19, is low temperature and high PH2 , where the higher frequency electrochemical resistance becomes large relative to gas diffusion. In this case, the total cell polarization resistance does generally follow the variation of TPB density with annealing (Figure 11), with the 1100◦ C–100h case an apparent outlier.

Electrochemical modeling.— The above AFL microstructural data were used in a simple electrochemical model that accounts for the TPB hydrogen oxidation process and ionic transport in the YSZ phase, to provide estimates of anode polarization resistance.5,34 In order to best compare with the experimental data, the calculations were done for 600◦ C where, as discussed above, the TPB-related processes were dominant. Figure 20 shows the anode polarization resistance, estimated from the total polarization resistances at 600◦ C by subtracting the separately-measured cathode polarization resistance of 0.75  · cm2 . The model simplifies the structure by assuming

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

F1303

Figure 20. Measured and computed anode polarization resistance versus annealing temperature. The measured anode polarization resistance was an estimate made by subtracting the separately-measured cathode polarization resistance of 0.75  · cm2 from the total polarization resistance measured at 600◦ C and PH2 = 0.97 atm (Figure 19). The computed values were obtained from the electrochemical model using measured structural data.

Figure 18. Nyquist and Bode plots of the EIS data from the as-reduced and annealed cells, taken at 750◦ C and a fuel H2 content of 9.7%.

one-dimensional columns of ionic conductor (YSZ) with a surface resistance determined by distributing TPBs uniformly over the column surfaces. Column dimensions were estimated from the measured AFL thickness (20 μm) along with YSZ volume fractions and particle sizes, while the column surface resistance was calculated from the TPB density and literature values of TPB length-specific resistance (2.00 × 105  · cm at 600◦ C in 84% H2 / 17% H2 O).35 A YSZ con-

ductivity of 0.0142 S/cm was used. The effect of YSZ tortuosity was included by dividing this YSZ conductivity by the tortuosity factor. The model does not account for gas diffusion or electron transport resistance in Ni. Figure 20 shows the predicted anode polarization resistances for the as-reduced anode and those annealed for 500 h at various temperatures. The trend of increasing resistance with increasing annealing temperature is in good agreement with the experimentally measured resistances. Although the computed resistances are a factor of ∼2 below the measured values, this is well within the range of reported values in the input TPB length-specific resistance values. A calculation of the concentration polarization associated with gas diffusion was not attempted since detailed structural data for the anode support, which provided a component of this resistance, was not available. Conclusions

Figure 19. Nyquist and Bode plots of the EIS data from the as-reduced and annealed cells, taken at 600◦ C and a fuel H2 content of 97%.

Feature sizes in the YSZ and Ni phases changed little until coarsening began at times > 100 h at 1100◦ C. On the other hand, substantial changes in pore and Ni phase surface areas were observed at lower temperatures and times. A higher fraction of isolated pores and larger pore tortuosity was observed at intermediate annealing temperatures and times. These early-stage changes are probably explained by noting that, although the as-fired NiO-YSZ was probably quite stable, NiO reduction to Ni and accompanying pore formation resulted in a highly unstable morphology. Three-phase boundary (TPB) density decreased rapidly initially, again probably due to the rapid rearrangement of the pore and Ni phases. The decrease continued much more slowly at longer times and temperatures, probably due to coarsening. These results suggest that attempts to establish Ni-YSZ coarsening rate laws, which are needed to predict AFL stability for long SOFC operation times, must be based on measurements done after the earlystage microstructural changes are completed. The present EIS results show that electrochemical hydrogen oxidation processes are strongly affected by annealing. The overall polarization resistance, when measured at low temperature where these processes dominate, correlated reasonably well with the measured active TPB density. The EIS results also show that gas diffusion is an important component of the cell resistance. The overall polarization resistance, when measured at low PH2 where gas diffusion processes dominates, correlated reasonably well with annealing-induced changes in pore connectivity and tortuosity. It should be noted that the present AFLs were at a limit of low pore volume fraction compared to most AFLs, because of the relatively low Ni content and the fact that no pore former was utilized; thus, it is perhaps reasonable that gas diffusion was important. However, it seems likely that structural

F1304

Journal of The Electrochemical Society, 160 (11) F1293-F1304 (2013)

changes in both the AFL and the anode support played a role in this gas diffusion resistance. Acknowledgments This material is based upon work supported by the National Science Foundation under grant Number DMR-0907639. Use of the National Synchrotron Light Source, Brookhaven National Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC0298CH10886. We thank Dr. Fernando Camino (BNL) for assisting the sample preparation using FIB/SEM at the Center for Functional Nanomaterials, Brookhaven National Laboratory, which is supported by the U.S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. The input and contribution from Kyle Yakal-Kremski, Northwestern University, was greatly appreciated. References 1. N. Q. Minh, Journal of the American Ceramic Society, 76, 563 (1993). 2. D. Simwonis, F. Tietz, and D. St¨over, Solid State Ionics, 132, 241 (2000). 3. A. Faes, A. Hessler-Wyser, D. Presvytes, C. G. Vayenas, and J. Van Herle, Fuel Cells, 9, 841 (2009). 4. P. Tanasini, M. Cannarozzo, P. Costamagna, A. Faes, J. Van Herle, A. Hessler-Wyser, and C. Comninellis, Fuel Cells, 9, 740 (2009). 5. J. S. Cronin, J. R. Wilson, and S. A. Barnett, Journal of Power Sources, 196, 2640 (2011). 6. H.-Y. Chen, H.-C. Yu, J. S. Cronin, J. R. Wilson, S. A. Barnett, and K. Thornton, Journal of Power Sources, 196, 1333 (2011). 7. Z. Jiao and N. Shikazono, Journal of The Electrochemical Society, 160, F709 (2013). 8. J. R. Wilson, W. Kobsiriphat, R. Mendoza, H. Y. Chen, J. M. Hiller, D. J. Miller, K. Thornton, P. W. Voorhees, S. B. Adler, and S. A. Barnett, Nature materials, 5, 541 (2006). 9. J. S. Cronin, K. Muangnapoh, Z. Patterson, K. J. Yakal-Kremski, V. P. Dravid, and S. A. Barnett, Journal of the Electrochemical Society, 159, B385 (2012). 10. Y. C. K. Chen-Wiegart, J. S. Cronin, Q. X. Yuan, K. J. Yakal-Kremski, S. A. Barnett, and J. Wang, Journal of Power Sources, 218, 348 (2012). 11. L. Holzer, B. Iwanschitz, T. Hocker, B. Munch, M. Prestat, D. Wiedenmann, U. Vogt, P. Holtappels, J. Sfeir, A. Mai, and T. Graule, Journal of Power Sources, 196, 1279 (2011).

12. B. Iwanschitz, L. Holzer, A. Mai, and M. Schutze, Solid State Ionics, 211, 69 (2012). 13. G. J. Nelson, K. N. Grew, J. R. Izzo Jr., J. J. Lombardo, W. M. Harris, A. Faes, A. Hessler-Wyser, J. Van Herle, S. Wang, Y. S. Chu, A. V. Virkar, and W. K. S. Chiu, Acta Materialia, 60, 3491 (2012). 14. Z. J. Jiao, N. Shikazono, and N. Kasagi, Journal of the Electrochemical Society, 159, B285 (2012). 15. P. R. Shearing, R. S. Bradley, J. Gelb, S. N. Lee, A. Atkinson, P. J. Withers, and N. P. Brandon, Electrochemical and Solid-State Letters, 14, B117 (2011). 16. R. R. Liu, S. H. Kim, S. Taniguchi, T. Oshima, Y. Shiratori, K. Ito, and K. Sasaki, Journal of Power Sources, 196, 7090 (2011). 17. J. S. Cronin, Y. C. K. Chen-Wiegart, J. Wang, and S. A. Barnett, Journal of Power Sources, 233, 174 (2013). 18. J. R. Wilson, J. S. Cronin, and S. A. Barnett, Scripta Materialia, 65, 67 (2011). 19. S. P. Jiang, Journal of Materials Science, 38, 3775 (2003). 20. A. Hagen, R. Barfod, P. V. Hendriksen, Y. L. Liu, and S. Ramousse, Journal of the Electrochemical Society, 153, A1165 (2006). 21. J. D. Nicholas and S. A. Barnett, Journal of the Electrochemical Society, 157, B536 (2010). 22. J. Wang, Y. C. K. Chen, Q. X. Yuan, A. Tkachuk, C. Erdonmez, B. Hornberger, and M. Feser, Applied Physics Letters, 100, 143107 (2012). 23. F. Natterer, in The Mathematics of Computerized Tomography, Society for Industrial and Applied Mathematics (2001). 24. J. R. Wilson, J. S. Cronin, A. T. Duong, S. Rukes, H. Y. Chen, K. Thornton, D. R. Mumm, and S. Barnett, Journal of Power Sources, 195, 1829 (2010). 25. P. R. Shearing, J. Gelb, J. Yi, W. K. Lee, M. Drakopolous, and N. P. Brandon, Electrochemistry Communications, 12, 1021 (2010). 26. P. R. Shearing, Q. Cai, J. I. Golbert, V. Yufit, C. S. Adjiman, and N. P. Brandon, Journal of Power Sources, 195, 4804 (2010). 27. C. Metcalfe, O. Kesler, T. Rivard, F. Gitzhofer, and N. Abatzoglou, Journal of the Electrochemical Society, 157, B1326 (2010). 28. Y. C. K. Chen-Wiegart, R. DeMike, C. Erdonmez, S. Barnett, K. Thornton, and J. Wang, Tortuosity Characterization of 3D Microstructure at Nano-scale for Energy Storage and Conversion Materials, in preparation. 29. S. Primdahl and M. Mogensen, Journal of the Electrochemical Society, 145, 2431 (1998). 30. S. H. Jensen, J. Hjelm, A. Hagen, and M. Mogensen, in Handbook of Fuel Cells (2006). 31. A. Leonide, A. Weber, and E. Ivers-Tiff´ee, in, p. 1183 (2010). 32. A. Leonide, V. Sonn, A. Weber, and E. Ivers-Tiffee, Journal of the Electrochemical Society, 155, B36 (2008). 33. H. Y. Zhu, A. Kromp, A. Leonide, E. Ivers-Tiffee, O. Deutschmann, and R. J. Kee, Journal of the Electrochemical Society, 159, F255 (2012). 34. C. W. Tanner, K. Z. Fung, and A. V. Virkar, Journal of the Electrochemical Society, 144, 21 (1997). 35. W. G. Bessler, M. Vogler, H. Stormer, D. Gerthsen, A. Utz, A. Weber, and E. Ivers-Tiffee, Physical Chemistry Chemical Physics, 12, 13888 (2010).