Investigation of Rapid, Moderate Temperature Change Thermal Cycles of a Micro-Tubular Flame-Assisted Fuel Cell

Solid oxide fuel cells (SOFCs) are solid-state electrochemical energy conversion devices that directly convert chemical energy to electricity and heat [1,2]. While SOFCs are promising in terms of high efficiency, low emissions, and no moving parts, the U.S. Department of Energy cites the major challenges for further development as (1) breakdown and corrosion of components at high temperatures, (2) slow startup, and (3) a limited number of shutdowns [3]. While the startup rate depends on the sealant, geometry (e.g., planar, tubular), and other system components, a recent review of the literature indicates that startup rates are typically around 5 °C · min⁻¹ with typical operating temperatures between 500 and 1000 °C [4]. As a result, some SOFC systems can take several hours or longer to reach operating temperatures. This characteristic is especially an obstacle in applications such as auxiliary power units (APUs), automobiles, aircraft, and portable power. SOFC micro-combined heat and power (mCHP) systems can take 1.5−3 h to reach operating temperature [5–7], and as a result, these systems are designed to operate continuously to achieve high electrical efficiency. Slow startup is a result of mismatches in thermal expansion between the SOFC, sealant, and materials used for fuel supply.

The third obstacle of limited shutdowns is a related challenge as repeated on/off cycling can cause damage to the SOFC system, especially to the sealant. While some on/off cycling has been reported in the literature, a more common test is to conduct thermal cycling of the SOFC between a low and a high temperature. Thermal cycling tests are typically less than 100 cycles in recent literature [4], with commercial SOFC mCHP systems reaching 50–80 cycles [5,8]. On/off cycling is also challenging because re-oxidation (i.e., redox cycling) of the anode can occur if exposed to oxygen while cooling [6,9,10].

Over the last two decades, many efforts have investigated ways of enhancing the startup rate and thermal cycling of SOFCs. Most of these efforts have been made in the interest of improving the sealant of SOFCs [11,12], reducing or eliminating the sealing constraint while improving the mechanical strength of the SOFC. Single chamber SOFCs (SC-SOFCs) [13–15] and direct flame fuel cells (DFFCs) [16–22] have removed the SOFC from the conventional dual-chamber configuration which has led to reduced efficiency, but improved startup and thermal cycling. One planar DFFC reported a heating rate of 1160 °C · min⁻¹ [23]. Micro-tubular SOFCs (mT-SOFCs) [24,25] have improved mechanical strength and reduced sealing constraint, which has led to improved heating rates (100–200 °C · min⁻¹ [26,27]) and thermal cycling (400 cycles [28]). More recently, Ceres Power has made significant progress with metal-supported planar SOFCs and has achieved over 3600 thermal cycles, which is believed to be equivalent to approximately 10 years of operation [29,30]. Another company, SOLID-Power, has also made progress demonstrating 124 thermal cycles between 25 °C and 750 °C [31].

Recently, a novel setup was developed for thermal cycling mT-SOFCs using a burner to achieve rapid heating. Rapid heating rates (966 °C · min⁻¹) and rapid cooling rates (353 °C · min⁻¹) were sustained for 3000 thermal cycles of 9 mT-SOFCs in a stack with no degradation in power density reported [4]. The maximum anode temperature was 752 °C. While these results indicated that rapid startup and cycling challenges can be overcome, thermal cycling at higher cathode temperatures was not characterized. In the previous test, the average air temperature around the cathode remained less than ∼500 °C due to non-uniform heating of the mT-SOFCs. Additional research is needed to assess the potential for rapid thermal cycling at higher cathode temperatures [32].

In this work, a rapid thermal cycling test of 3000 cycles of an mT-SOFC stack is carried out at an elevated cathode temperature to assess the impact of temperature on thermal cycling degradation. A different cathode material than was used in the previous test was selected based on the temperature. The stack polarization, power density, and impedance are characterized before and after the 3000 thermal cycles. The temperature at several points in the setup and the stack voltage is monitored throughout the experiment. A scanning electron microscope was used to investigate the degradation mechanisms including carbon deposition, delamination of the electrodes from the electrolyte, crack propagation, and breakdown of the current collector.

In Eq. (2), V₀ is the OCV, I is the maximum current density measured in the experiment, and V_i is the voltage at the corresponding current density, I.

In Eq. (2), n represents the number of moles of a species (e.g., fuel or air). The molar fuel/air ratio with the superscript S denotes the fuel/air ratio needed for stoichiometric combustion. The mT-SOFC stack was sealed to the end of the burner chamber using ceramic paste (Aremco ceramabond 552). The methane and air flowrates entering the partial oxidation burner are shown in Table 1. Air for the cathode is preheated over the partial oxidation chamber. The oxygen that is not reduced at the cathode is used for fuel-lean combustion of any fuel that is unreacted in the mT-SOFC stack. The equivalence ratio of the fuel-lean burner was fixed at 0.8, resulting in 25% excess air. The stack was arranged in a circular configuration with the SOFC at the top of the stack labeled “mT-FFC 1” in Fig. 1 and similarly for “mT-FFC 5” that is located at the bottom of the stack.

In the previous thermal cycling experiment, the anode temperature was 750–800 °C while the air around the cathode was ∼500 °C or less. To increase the temperature of the air around the cathode, the entire chamber was insulated resulting in a peak air temperature near the cathode of 756 °C. The temperature was relatively uniform around the cells after insulation with a minimum temperature of 750 °C.

The results of the mT-SOFC stack polarization and power density characterization prior to thermal cycling are shown in Fig. 2. The air temperature around the mT-SOFC cathode varied slightly with a typical average of around 750 °C. As the SOFCs operate directly in combustion exhaust the fuel cells have also been termed micro-tubular flame-assisted fuel cells (mT-FFC) in the literature [33]. For this test, polarization and power density data were obtained for equivalence ratios from 1.05 to 1.5. However, an equivalence ratio of 1.5 was relatively unstable as the flame was occasionally quenched near the upper flammability limit. For this initial test, a polarization curve at 1.5 could not be obtained before the flame quenched. Similarly, the reactions were relatively unstable at an equivalence ratio of 1.45, which explains the jumps in the polarization and power density curves at that condition. The results show a clear trend of decreasing polarization and increasing power density as the equivalence ratio increased which was predicted with previous studies on a model combustion exhaust [33,36].

The increase in power density with equivalence ratio is a result of increased synthesis gas concentration in the exhaust as the equivalence ratio increases. Table 2 summarizes the molar concentration of H₂, H₂O, CO, and CO₂ in the combustion exhaust that enters the mT-SOFC as fuel. As shown, the concentration of H₂ is less than 1 mol% of the total exhaust stream at an equivalence ratio of 1.05, which explains the low power density. As the equivalence ratio increases to 1.40, the concentration of H₂ increases to 6.32 mol% and the CO concentration increases to 8.84 mol%. Due to instabilities in the flame, accurate reading for the exhaust gas composition could not be obtained at an equivalence ratio of 1.45. The mT-SOFC showed a significant peak power density of 167.6 mW.cm⁻² at 0.6 V per fuel cell at an equivalence ratio of 1.45 prior to thermal cycling.

Table 2 also summarizes the change in OCV and ASR with equivalence ratio. As shown, the OCV increases with equivalence ratio due to the increase in H₂ and CO concentration and decrease of the other species concentration (H₂O, CO₂, and N₂), which do not participate in the electrochemical reactions. This behavior of increasing OCV with equivalence ratio can be predicted with the Nernst equation [33]. The ASR decreased with equivalence ratio with values varying from 0.94 to 1.95 Ω.cm². The ASR is high for current SOFCs, which typically have an ASR of 0.3–0.5 Ω.cm². The high ASR is due to the low concentration of H₂ and CO as the same fuel cells used in this experiment were utilized in other experiments with pure hydrogen [35] and achieved an ASR of 0.3 Ω.cm².

After obtaining the initial polarization and power density data reported in Fig. 2, the thermal cycling test was initiated. The equivalence ratio was fixed at 1.25 for the test because this condition resulted in fairly high synthesis gas concentration, which prevents re-oxidation of the anode. A higher equilvalence ratio could also be chosen, but the partial oxidation temperature decreases with increasing equivalence ratio, and therefore, 1.25 represented a good condition for high temperature and high synthesis gas concentration. The thermal cycling test was initiated by turning off the fuel and airflow, allowing the stack to cool naturally for 95 s, turning the fuel and airflow on, reigniting, and allowing the system to heat for 75 s, as was done in the previous experiment [4]. During the previous test [4], these conditions resulted in a maximum anode temperature of 752 °C when the fuel/air was reacting. The minimum anode temperature was 282 °C after cooling the setup. This resulted in an average anode heating rate of 966 °C · min⁻¹ and average anode cooling rate of 353 °C · min⁻¹.

Figure 3 shows temperature profiles obtained during two consecutive thermal cycles. For this thermal cycling test, the cathode temperature was monitored carefully with two K-type thermocouples. One thermocouple was located near the cathode of mT-SOFC 1 (Fig. 1) at the top of the cylindrical stack (denoted mT-FFC 1), and the other was located next to the cathode of mT-SOFC 2 (Fig 1) at the bottom of the stack (denoted mT-FFC 5). With the improvements in insulation, the cathode temperature remained much more stable throughout the testing with the maximum temperature reaching 680 °C and the minimum temperature reaching 460 °C. Based on the temperature measurements near the cathode, an average heating rate of 144 °C · min⁻¹ and an average cooling rate of 114 °C · min⁻¹ occurred. Some asymmetry in heating was observed with mT-FFC 1 reaching higher temperatures than mT-FFC 5. However, the asymmetry is expected as the cells at the top of the stack were at a slightly higher temperature than the cells at the bottom of the stack due to the warmer air rising in the chamber. Figure 4 shows seven cycles of the OCV after 15 h of the thermal cycling test. As the total temperature change was less than that found in a previously reported experiment [4] and did not approach room temperature during cooling, the thermal cycles are characterized as moderate thermal cycles for this experiment.

Table 3 summarizes the OCV degradation throughout the thermal cycling experiment. As shown, the voltage degradation rate is highest during the first 500 cycles at an average rate of 0.0231 V per 100 cycles per fuel cell. However, the degradation rate falls significantly from cycle 500 to 2000 and reaches a minimum of 0.0089 V per 100 cycles per fuel cell. Around the 2000th thermal cycle, the degradation rate begins to increase steadily. The reason for this increased degradation rate after 2,000 cycles appears to be linked to the interconnects based on the postmortem analysis of the stack.

Figure 5 shows polarization and power density curves obtained after completing 3000 moderate thermal cycles. The total test lasted just over 150 h. The initial results showed a significant decrease in power density at all equivalence ratios as well as a notable decrease in OCV of approximately 3 V. In this particular test, the flame was stable enough to obtain a polarization curve at an equivalence ratio of 1.50. Figure 5(b) compares the polarization and power density before and after the 3000 moderate thermal cycles at equivalence ratios of 1.05 and 1.45.

After allowing the experimental apparatus to cool, the mT-SOFC stack was investigated. Figure 6 shows one image of the stack after testing. The interconnects of three consecutive mT-FFCs broke and shorted out three fuel cells, which explains the ∼3 V decrease in OCV in Fig. 5. Based on the mT-SOFC stack OCV results summarized in Table 3, it appears that most of the damage to the interconnects occurred between cycles 2000 and 3000. No breakage in the interconnects of the other mT-SOFCs was observed, but the interconnects were more degraded as a result of the increased temperatures in the experimental apparatus. The breakage in the current collector in this test appears to be a result of increased temperatures and the mismatch in thermal expansion coefficients of the silver and steel wires. As shown in Fig. 3, mT-FFC 1, which is located at the top of the stack, was at a higher temperature throughout the experiment. From Fig. 6(b), more degradation of the interconnect is observed for this cell. For mT-FFC 5, which is located near the bottom of the stack, less degradation is noted in Fig. 6(c). The cells at the bottom of the stack had slightly lower temperatures as shown in Fig. 3.

To assess degradation in the mT-FFC stack without the influence of the three broken interconnects, the broken interconnects were removed and new interconnects were added. After fixing the interconnects and reigniting, the mT-FFC stack OCV increased by slightly less than 3 V, as expected. Figure 7 shows the mT-FFC performance after changing the three interconnects. A total decrease in OCV from 7.41 V to 7.19 V occurred, which is a 3.0% decrease over 3000 cycles. Without the degradation due to the interconnects, there was less OCV degradation in this test compared to the previous test at lower temperatures. With the nine cells in the stack, that is an average degradation rate of 0.0008 V per 100 cycles per fuel cell. The previous test (which occurred with an (La_0.8Sr_0.2)_0.95MnO_3−x cathode, no buffer layer, but the same anode, electrolyte, and test procedure) resulted in a degradation rate of 0.0018 V per 100 cycles per fuel cell [4]. A summary of this test and several previous investigations of the OCV degradation rate during thermal cycling are summarized in Table 4.

The voltage degradation at higher current density was more significant than the degradation at OCV. A comparison is provided here with a previously reported mT-SOFC stack. In that work, the 10 mT-SOFC stacks had no noticeable degradation in OCV over 100 cycles and reported a voltage drop of 0.06 V per 100 cycles per fuel cell at the standard current conditions [40]. For the mT-SOFC stack in this study, the standard current would be around 200 mA.cm⁻² at an equivalence ratio of 1.45. At that current density and equivalence ratio, the mT-SOFC showed an average voltage degradation rate of 0.0095 V per 100 cycles per fuel cell as the stack voltage degraded from 5.986 V to 3.424 V during the test. Another test of a single mT-SOFC at peak power density observed a voltage degradation rate of 0.073 V per 100 cycles per fuel cell [41]. The degradation rate observed in this study was approximately an order of magnitude lower than those previously reported experiments. At a current density of 100 mA.cm⁻², the voltage degradation rate was even lower at 0.0057 V per 100 cycles per fuel cell. A comparison of these studies and others is shown in Table 5.

As previously observed in other studies, the degradation rate is high for the first few cycles and decreases as the number of thermal cycles increases [40]. In other words, the most degradation occurs during the first few cycles and decreases with each additional cycle. This trend is similar to mechanical fatigue and was observed somewhat for the OCV as shown in Table 3. For example, in this study, the OCV degradation rate from 0 to 500 thermal cycles was 0.0231 V per 100 cycles per fuel cell; from 0 to 1500 thermal cycles, it was 0.0089 V per 100 cycles per fuel cell, and the degradation rate from 0 to 3000 thermal cycles was 0.0008 V per 100 cycles per fuel cell. Based on this study and the comparison to literature, it appears that the trend of decreasing degradation rate with an increasing number of cycles continues and may offer an explanation for the significant decrease in average degradation rate observed in this experiment. The high initial degradation is likely due to agglomeration of metallic Ni and loss of nickel–nickel contact in the anode which increases ohmic resistance [10,41]. The LSCF cathode is also known to have a higher degradation rate than other cathode materials due to interfacial reactions with YSZ and strontium diffusion out of the cathode [10]. These two degradation mechanisms are explored with postmortem analysis of the cells.

In addition to the agglomeration of Ni metal and LSCF degradation, the main mechanisms for fuel cell degradation during thermal cycling have been reported as the delamination of the electrodes from the electrolyte, crack initiation and propagation which can lead to leaks, and current collector disconnection from the electrodes [41,43]. In this study, the interconnect became disconnected from the cathode and resulted in significant degradation, as discussed previously. EIS and SEM were utilized at the end of the thermal cycling experiment to assess the other forms of degradation. Figure 8 shows the results plotted as a Nyquist plot and Fig. 9 as a Bode plot. The ohmic, activation, and concentration losses all significantly increased after thermal cycling.

As shown in Fig. 9, the jump in cathode polarization resistance, which is typically observed around a frequency of 100 Hz [36,44], was particularly large. After analyzing the mT-SOFC, some densification of the LSCF + SDC cathode appears to have occurred. Degradation of the LSCF has been shown to significantly increase the polarization resistance while having a much smaller effect on the ohmic resistance [10]. In addition to the densification, a greyish-brown layer was observed around the edges of the LSCF cathode after completing the cycling experiment. The greyish-brown layer on the outside of the cathode was observed previously [45] and has been identified as primarily Sr with some La. The transport is believed to be a combination of surface diffusion of Sr and possibly some vapor phase transport [45]. Degradation can also occur from Sr diffusion through the buffer layer to the YSZ electrolyte, which results in a SrZrO₃ interfacial layer that has higher resistance [35,46].

Changes in the diffusion processes at the anode are also evident due to the significant increase in impedance at low frequency. Carbon deposition on the anode was investigated with the SEM. Limited carbon deposition was observed at the anode surface (Fig. 10), but the anode remained porous (Fig. 11). The carbon deposition at the anode surface is likely one of the main contributors to the decreased transport in the anode as the middle section of the anode remained relatively porous and free of carbon after thermal cycling. Contact at the electrolyte/anode interface has been observed as a major cause of increased ohmic resistance [41], but no delamination was observed in this study. Figure 11 shows evidence of significant Ni metal agglomeration on the surface. Nickel agglomeration is known to cause a decrease in Ni–Ni contact, which results in higher ohmic resistance [10,41].

Previous research [4] has shown potential for mT-FFC operation in environments with rapid change in temperature and 3000 thermal cycles between 300 °C and 750 °C. In that previous work, the air around the cathode remained at a lower temperature (∼500 °C) with limited degradation. This study indicates that the degradation rates of the interconnects and LSCF cathode are more significant at a higher temperature (∼680 °C), as expected. Maintaining the cathode and interconnect at lower temperatures may offer opportunities to utilize mT-FFCs in environments where rapid startup and cycling are critical.

One of the key obstacles that mT-FFCs still need to overcome is limited electrical efficiency. Recent work has analyzed the theoretical electrical efficiency limits based on the amount of syngas that can be generated during partial oxidation at different equivalence ratios [47]. That work showed that the electrical efficiency is limited to less than 10% for methane/air partial oxidation at an equivalence ratio of 1.4. This study, which was not focused on improving the electrical efficiency, had a total efficiency of less than <0.5%. With proper optimization, the electrical efficiency has been improved to over 1.2% in other work [47] and can be increased further based on the theoretical analysis. As a result, the main applications for mT-FFCs are those that require significant thermal energy with small electric loads. For example, common residential combustion-based heating systems utilize electrically powered blowers and pumps (<500 W_el) that could be powered by the mT-FFCs while the heat (11,000–030,000 W_th) is recovered for space heating or domestic hot water. Applications for the technology have been investigated previously including a modified furnace [2] and boiler [4] setup which has relevant operating conditions as those explored in this paper. In the modified furnace or boiler setup, the fuel-rich combustion, or partial oxidation, of natural gas would generate synthesis gas for electrochemical conversion in the mT-FFC. The mT-FFCs can convert the synthesis gas to electricity to power small residential loads. Recent work showed that they can achieve a high (∼80%) combined (electrical plus thermal) efficiency for space heating [7]. The most suitable applications require significant thermal energy with limited electrical energy generation.

While these experiments investigated the stacks OCV and power density at the beginning and end of the test, future experiments should investigate degradation throughout the experiments to understand how the degradation rate changes. Such an approach will also allow for greater comparison with previous literature. Future work should also consider the degradation mechanisms in the current collector so that rapid cycling and high temperatures can be sustained.

Limited thermal cycling testing has been reported previously for SOFCs and is typically limited to slow startup rates (5 °C · min⁻¹), slow cooling rates, and less than 200 cycles. Using an experimental setup designed for rapid heating, cooling, and thermal cycling tests, an mT-SOFC stack consisting of nine fuel cells with LSCF + SDC cathode was tested. An experiment with 3000 moderate thermal cycles was conducted. When the interconnect degradation is considered, the OCV degradation rate was 0.0122 V per 100 cycles per fuel cell. When the interconnect was removed to isolate the other forms of degradation, the mT-SOFCs showed a much lower degradation rate. The results showed less degradation in OCV (0.0008 V per 100 cycles per fuel cell) during this test compared to a previous test when the degraded interconnect was removed. The degradation rate at a current density of 200 mA.cm⁻² was higher and averaged 0.0095 V per 100 cycles per fuel cell. The voltage degradation rates observed in this setup were lower than other previously reported studies and may be due to the change in degradation rate as the number of thermal cycles increases. Degradation of the mT-SOFC stack occurred due to some carbon deposition on the anode surface. Microstructural changes in the cathode and near the electrolyte/anode interface are also responsible for some of the degradation. The results indicate that the degradation of the interconnects accelerated with the temperature increase compared to the previous test. Future research opportunities were discussed.

This material is based upon the work supported by an Agreement with Syracuse University awarded by its Syracuse Center of Excellence in Environmental and Energy Systems with funding under prime award number DE-EE0006031 from the US Department of Energy and matching funding under award number 53367 from the New York State Energy Research and Development Authority (NYSERDA) and under NYSERDA contract 61736. This material is also based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. 1746928. This work is the final development of the original ASME Conference Proceedings, Denver, CO (Virtual Conference). Paper No. ES2020-1634.

There are no conflicts of interest.

n_fuel/n_air =

molar fuel/air ratio

n_fuel^S/n_air^S =

molar stoichiometric fuel/air ratio

Φ =

equivalence ratio