Electrochemical Performance and Degradation Analysis of an SOFC Short Stack Following Operation of More than 100,000 Hours

From August 2007 to January 2019, a two-layer solid oxide fuel cell (SOFC) short stack of a planer design with zirconia-based, anode-supported cells (ASC) and ITM interconnectors (with 26% chromium content) was tested with hydrogen and compressed air at a furnace temperature of 700°C for more than 100,000 hours, of which ∼93,000 were in constant current mode, with a current density of 0.5 Acm−2 and with a fuel utilization of 40%. The calculated voltage degradation rate slowly decreased, from ∼8.0 mV/kh (∼1.0%/kh) for the first 40,000 h to ∼1.4 mV/kh (∼0.2%/kh) for the subsequent operation under load, indicating different dominating degradation mechanisms. The average voltage and area-specific resistance (ASR) degradation rates for the complete operating period under electrical load were 0.5%/kh and 2.5%/kh, respectively. Electrochemical impedance spectroscopy (EIS) was also implemented at the end of the testing period for the purpose of electrochemical characterization and a degradation analysis. The post-mortem analysis of the stack is currently in preparation. In this study, the performance and degradation behavior of the stack and cells are analyzed and discussed on the basis of the electrochemical measurements collected. © The Author(s) 2019. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0751916jes]

In order to make solid oxide fuel cells (SOFCs) a commerciallycompetitive technology, their lifetime and cost must reach acceptable targets. According to the U.S. Department of Energy (DOE), small residential and light commercial combined heat and power (CHP) fuel cell systems should achieve a lifetime of 60,000 hours and degradation rate of 0.3%/kh. In general, the lifetime is extrapolated from the shortperiod experimental measurement, based on certain assumptions and previous understanding of the degradation mechanisms. The methodology of accelerated stress testing using advanced mathematical models has also been intensively and repeatedly discussed. However, none of these methods has been able to provide a reliable and convincing prediction of the lifetime of an SOFC until now. Due to the high demand on testing infrastructure and other technical difficulties arising from cells and stacks themselves, the time-consuming endurance tests have been fairly limited, especially at the stack and system levels. [1][2][3][4][5][6][7][8][9][10][11] From 2004 to 2008, a series of tests with SOFC short stacks were carried out within the framework of the European project, REAL-SOFC, aimed at further illuminating the degradation mechanisms of SOFCs for stationary applications. 1 One of the stacks (F1002-97), commenced in August 2007, continued to be operated after the end of the project. In October 2015, this stack reached 70,000 hours of operation under load, which superseded the longest ever reported operation time of SOFC cells and stacks at the time, as reported by Blum et al. 2 After about 11.5 years (i.e., 100,000 hours) of operation at a furnace temperature of 700°C, the stack was shut down in January 2019 for a post-mortem analysis. In this work, the electrochemical performance and degradation of this stack will be presented in detail.

Experimental
The stack configuration and testing conditions were essentially described by de Haart et al. 1  • Glass sealants for cells and interconnectors. 14 • Silver gaskets between the stack and test bench, compressed with a 50 kg dead weight.
• Stationary operation with a constant current density of 0.5 Acm −2 at a furnace temperature of 700°C.
• One Type K thermocouple with a diameter of 1 mm was inserted ∼10 mm-deep into the intermediate interconnector, as shown in Figure 1.
After the initial conditioning and characterization, the stack was operated under constant current mode at a constant furnace temperature of 700°C. Under the electrical load, the measured stack temperature was initially 713°C and later 720°C after more than 93,000 hours of operation. During operation, the electrical load was interrupted multiple times due to different fault functioning of the testing infrastructure, as well as planned maintenance of the laboratory facilities. The operation under load had to be ended because drastic drops in cell voltages frequently occurred, which was assumed to be due to a disturbance of the electrical controlling system of the test bench. This was also  why the stack was kept in the open circuit voltage (OCV) condition for a long period, anticipating the possible new start of the operation once the problem could be solved. One thermal cycle was performed shortly before 100,000 hours of operation was achieved for checking the possible impact of the thermal cycle on the stack after such a long operating period. The electrochemical impedance (EIS) measurement using a Zahner IM6 Workstation was first implemented after ∼90,000 hours of operation. The distribution function of the relaxation times (DRT) was calculated using the Matlab toolbox DRTTOOLS. 15 Figure 1 shows stack F1002-97 in the test bench before and after the 100,000 hours of operation at 700°C. Three Pt wires were welded to the stack for voltage measurement. One thermocouple was then inserted into the interconnector, where it was closer to the fuel inlet side. The stack was then operated in counter-flow mode. Because of the cooling effect of the air, the temperature near the air inlet was slightly lower than the rest of the stack. According to CFD modelling by Peksen, 16 the maximum temperature difference inside the stack under the similar conditions is less than 30°C. As can be seen in Figure 1, the stack did not suffer any oxidation or damage after operation. It must also be emphasized here that there was no cleaning or movement of any parts before taking the photos, which means that there was no spallation of the oxide scales from the ITM interconnectors or glass.

Results and Discussion
The stack was heated up to 850°C for the joining process, and then cooled down to 800°C for reduction of the anode. After that, the initial performance was characterized by I-V curves at three furnace temperatures of 800°C, 750°C and 700°C, respectively. For I-V curve measurements, a gas mixture of 2.8 SLM (standard liters per minute) H 2 and 18 ml min −1 of deionized water was used on the anode side, while 6.6 SLM of compressed air was used on the cathode side. The amounts of H 2 and air were chosen such that the fuel utilization and air utilization could reach 40% at a current density of 1 A cm −2 (i.e., 80 A). During the measurements, both the gas fluxes and furnace temperature were kept constant. The ramp of the electrical load was 4 A min −1 . The I-V curves at three temperatures are shown in Figure 2. Both layers showed nearly identical performance at the beginning of the operation. The stationary operation at 700°C was commenced directly after the characterization. The water amount was kept unchanged, while the flow rates of H 2 and air were decreased to 1.4 SLM and 5.28 SLM, so that the fuel utilization and air utilization would be 40% and 25% (i.e., λ∼4) at 0.5 A cm −2 , respectively. The complete evolution of the voltages of each layer, stack temperature, calculated area-specific resistances (ASR) and local voltage degradation rate are shown in Figure 3. The ASRs were calculated from the difference of the measured voltage and the calculated average Nernst voltage at the given current density, as described by Fang et al. 13,17 During stationary operation at 0.5 A cm −2 , the electrical load was interrupted many times for different reasons, as mentioned above. For safety reasons, the fuel gas had to be switched to forming gas (i.e., 96% Ar and 4% H 2 ) in case of any alarms. The stack was never cooled down under any circumstances during operation until it was decided to perform a thermal cycle at the end of it. It can be clearly seen in Figure 3 that both layers showed similar degradation behavior during the stationary operation. The average degradation rates for different time domains were calculated on the basis of the initial values and are shown in Table I. From Figure 3 and Table I, the following information can be observed or speculated upon without further post-mortem analysis of the stack: • The degradation rates were neither constant nor linear. Despite the small variation in the degradation rate, the stack showed quasilinear behavior until 40,000 hours, with an average voltage degradation rate of 1.0%/kh. After that, the degradation rate continuously decreased, to 0.1%/kh. If the voltage curves can be treated as two quasilinear domains separated at 40,000 h, the stack only had a degradation rate of 0.2%/kh for the last 53,000 hours of operation under load. The total degradation rates, taking into consideration the starting and ending values only, are 0.5%/kh for voltage and 2.5%/kh for ASR, respectively.
• Starting from ∼30,000 h, it can be noticed that the cell voltages always received a slight improvement directly after an interruption of the electrical load, and it took days to return to the values prior to the interruption. Such behavior became more pronounced after 40,000 h, and it could take thousands of hours for the cells to return to their old level. However, such hysteresis phenomena did not change the general degradation trend. A preliminary analysis based on the EIS measurements showed a possible correlation with cathodic polarization, which still requires further analysis and proof.
• The long-term performance of the F10-design stacks with different configurations but under the same testing conditions has been presented many times previously. [1][2][3]18 It has been proven that the MnOx protective coating used in F1002-97 is not sufficient to prevent Cr from evaporating from the interconnectors. The observed degradation rate of 1.0%/kh for the first 40,000 hours corresponds well to all of the other tested stacks with the same type of protective coating. With the improved MCF spinel coating prepared by atmospheric plasma  spraying (APS), a low degradation rate of 0.3%/kh was achieved for 34,000 hours. 19,20 Based on these results, it can be assumed for the time being that the degradation under the given conditions is dominated by Cr-poisoning in the cathodes and the oxidation of metallic interconnectors for the first 40,000 hours. Figure 4 shows the stack's behavior during the last 10,000 hours. The operation under load had to be ended due to a malfunction in the test bench. The stack was then kept under the OCV condition with dry H 2 for the remaining period. A comparison of the I-V curves before and after the stationary operation is shown in Figure 5. The EIS measurement was first performed near the end of the stationary operation. Figure 6 shows the Nyquist plots of the cells taken with dry H 2 at 700°C. For a rough qualitative comparison, a measurement of another stack with the same type of cells and similar configurations is also shown. However, this cannot fully represent the initial performance of the cells in F1002-97. Nevertheless, the following insights can be drawn from the comparison: • Both layers in F1002-97 degraded in a similar manner. • There was no degradation in the OCVs. • The low-frequency arc, which corresponds to gas conversion and gas diffusion in the anode substrate, exhibits few differences between the two stacks. Assuming both stacks were tight enough that the low frequency arc was not influenced by the humidity caused by the leakage, then the slightly smaller arc after 90,000 hours of operation corresponds well to the possible morphology change in the anode structure, which favors the gas diffusion process.
• Compared to the increase in ohmic resistance, electrode polarization dominated the degradation. Based on the DRT analysis, it could be concluded that both the anodic and cathodic polarizations increased, as shown in Figure 7. Only the results of layer 2 and the reference (the same measurement as shown in Figure 6) are plotted for the purpose of simplification. To further distinguish the cathodic and anodic polarizations from each other, the impedance measurement was also conducted with humidified H 2 , where the anodic polarizations relating to the gas diffusion (i.e., near 1 Hz and 20 Hz) and charge transfer (i.e., near 800 Hz) processes were obviously smaller than those with dry H 2 , while the cathodic polarization (i.e., near 100 Hz) did not show any difference between the measurements with dry and humidified H 2 .
As can be seen from Figure 4, the OCVs of both layers with dry H 2 were ∼1.2 V directly after 93,000 hours of operation under load. Figure 8 shows all OCV measurements during operation with dry H 2 . Apart from the first measurement, all values were taken at 700°C. No degradation in the OCV could be noticed until the end of the operation under load. During the last 5,000 hours of the OCV period, the OCVs of both layers, especially layer 2, decreased very slowly but continuously, as shown in Figures 4 and 8. Starting from 99,000 h, the OCV of layer 2 decreased rapidly to ∼1,030 mV, which was close to the internal criteria of 1 V for ending the test. Based on the current results, we believe that the continuous voltage decrease during the OCV period, as well as the failure in layer 2 after 99,000 h, were not because of the degradation in glass sealants, but mainly related to the possible damage in cell 2 after the repeated malfunction of the test bench between 94,300 h and 94,500 h. The related degradation in layer 2 can also be seen through the impedance measurement EIS6 in Figure 9.
It is also important and interesting to see in Figure 9 that there was no difference in the ohmic resistance between EIS2 and EIS5  (∼4,500 hours of operation under load in between), which indicates that the degradation during that period was mainly/only dominated by the electrode polarization. The DRT analysis of the measurements EIS2, EIS5 and EIS6 are shown in Figure 10. The damage in layer 2 after the failure of test bench can be seen mainly at the anode side. The continuous increases of both high frequency peaks of layer 1 can be also noticed, indicating a slight degradation of the anode. The possible change in the cathodic peak (i.e. near 60 Hz) can be hardly confirmed, since the width of the peak changes between EIS2 and EIS5 as well. Based on these results, together with the information displayed in Figures 3, 6 and 7, the following assumptions were made: • Assuming that the degradation after 40,000 hours was dominated by the anodic polarization alone and the degradation rate was linear throughout the entire operation, the degradation rate of the anodic polarization can be calculated to be 2∼3 m cm 2 /kh, which led to a degradation of 200∼280 m cm 2 after 93,000 hours of operation • The degradation during the first 40,000 hours was dominated by Cr-poisoning in the cathodes and the oxidation of metallic interconnectors, which led to an increase in both cathodic polarization and ohmic resistance. Assuming that the difference of ∼180 m cm 2 in the ohmic resistance between F1002-97 and the reference stack in Figure 6 was of the same magnitude as the degradation of the ohmic resistance in F1002-97, and assuming that the degradation in ohmic resistance and cathodic polarization only occurred during the first 40,000 hours of operation, then the degradation in cathodic polarization can be estimated to be ∼310 m cm 2 . It must be noted that the assumption here is only based on the observations and analysis under current operating conditions. We have also observed from another stack that the anodic polarization could increase from the beginning of the test under high current densities and fuel utilizations (not published yet).
As is shown in Figure 4, EIS measurements were performed regularly during the OCV period, as the OCVs kept decreasing slightly.       Figure 11 shows the Nyquist plots of measurements EIS6, EIS11 and EIS15 before the thermal cycle. Due to the continuously increased leakage rate in layer 2, the total impedance of layer 2 correspondingly decreased as a result of the reduced diffusion polarization. On the contrary, there was virtually no change in layer 1 between EIS6 and EIS11. The increased polarization in layer 1 by EIS15 could be a consequence of the increased leakage rate in layer 2. The EIS18 measurement was recorded after the thermal cycle, and is also shown in Figure 11 for comparison with EIS15. Both layers showed a slight increase in the ohmic resistance, indicating a possible contact loss after the thermal cycle. The I-V curves before and after the thermal cycle are shown in Figure 12. There was no obvious change in the OCV and stack temperature. The thermal cycle had few impacts on the mechanical stability of the cells and the glass sealants after nearly 100,000 hours of operation under working conditions at 700°C.

Conclusions
A solid oxide fuel cell short stack in a planar design was tested at a constant furnace temperature of 700°C for more than 100,000 hours, of which ∼93,000 were under a constant current density of 0.5 Acm −2 with a fuel utilization of 40%. H 2 with 21% humidity and compressed air were used for stationary operation. Under current conditions, the evolution of the cell voltages under load can be separated at ∼40,000 h into two periods, in which the degradation rate can be treated as quasi-linear for the purpose of simplification. Figure 11. Nyquist plots of the impedance measurements EIS6, EIS11, EIS15 (before the thermal cycle) and EIS18 (after the thermal cycle) showed changes during the OCV period and after the thermal cycle. During the first 40,000 hours of operation, the average cell voltage decreased from ∼820 mV to ∼510 mV, with a nominal degradation rate of ∼1.0%/kh. Such a degradation rate has been proven to be a reproducible value for stacks with the MnO x protective coating. Therefore, it is assumed that the degradation during the first 40,000 hours was dominated by Cr-poisoning in the cathodes and the oxidation of the metallic interconnectors. The cell voltages dropped from ∼510 mV to ∼430 mV during the next ∼53,000 hours, with a reduced degradation rate of ∼0.2%/kh. Based on the EIS measurements during the last ∼4,500 hours of operation under load and the preliminary DRT analysis, it seems that the degradation during ∼53,000 hours was mainly dominated by the anodic polarization. Accordingly, the overall contributions of the ohmic, cathodic and anodic polarizations in observed total degradation could be estimated as ∼180 m cm 2 , 310 m cm 2 and 280 m cm 2 , respectively. The average voltage and ASR degradation rates for the complete operating period under electrical load were 0.5%/kh and 2.5%/kh. Both OCVs were still more than 1.2 V after 93,000 hours of operation. The thermal cycle at the end of the test did not deteriorate the tightness of the stack, either.
Although the stack was mostly operated below a cell voltage of 0.7 V, which is far below the normal working voltages, the 100,000 hours of operating time was essential in view of the following considerations: • As a proof of concept, demonstrating the feasibility of current material and stack technology to achieve such a long lifetime.
• The degradation rate is hardly linear or constant. Lifetime predictions, by extrapolating short-term tests, can easily be misleading, even with a testing time of over tens of thousands of hours. The understanding of the degradation phenomena and mechanisms at different time domains will help to better predict the lifetime.
Based on the previous and current results, we believe that a lifetime of 60,000 hours with a low degradation rate of 0.3%kh can be achieved with current stack technology by properly choosing the stack components and operating parameters.