Performance Evaluation and Degradation Mechanism for Proton Exchange Membrane Fuel Cell with Dual Exhaust Gas Recirculation

Fuel gas utilization and water management are particularly challenging integrated engineering problems in hydrogen–oxygen proton exchange membrane fuel cell (H2/O2 PEMFC) systems. Herein, a standardized process is adopted to evaluate the performance and investigate the degradation mechanisms of a PEMFC with dual exhaust gas recirculation. The purpose of incorporating recirculation subsystems in the fuel cell is to achieve a high fuel gas utilization rate and realize effective water management inside the stack, which consists of 3D‐printed ejectors and a customized recirculation pump. Evaluation of the electrochemical performance degradation and morphological characterization of the fuel cells under different operating strategies are performed after 50 h durability experiments. At a current density of 400 mA cm−2, the performance degradation rates of the stack decrease from 16.50% to 7.49% and 0.71% in the ejector and recirculation pump operation strategies, respectively. The results show that using exhaust gas recirculation devices (ejector/pump) in the fuel cell stack can help in effectively mitigating water flooding and chemical degradation of the membrane electrode assembly. The findings of the study provide a perspective on the exhaust gas recirculation behavior and contribute to the engineering application of H2/O2 PEMFCs.

DOI: 10.1002/aesr.202200180 Fuel gas utilization and water management are particularly challenging integrated engineering problems in hydrogen-oxygen proton exchange membrane fuel cell (H 2 /O 2 PEMFC) systems. Herein, a standardized process is adopted to evaluate the performance and investigate the degradation mechanisms of a PEMFC with dual exhaust gas recirculation. The purpose of incorporating recirculation subsystems in the fuel cell is to achieve a high fuel gas utilization rate and realize effective water management inside the stack, which consists of 3D-printed ejectors and a customized recirculation pump. Evaluation of the electrochemical performance degradation and morphological characterization of the fuel cells under different operating strategies are performed after 50 h durability experiments. At a current density of 400 mA cm À2 , the performance degradation rates of the stack decrease from 16.50% to 7.49% and 0.71% in the ejector and recirculation pump operation strategies, respectively. The results show that using exhaust gas recirculation devices (ejector/pump) in the fuel cell stack can help in effectively mitigating water flooding and chemical degradation of the membrane electrode assembly. The findings of the study provide a perspective on the exhaust gas recirculation behavior and contribute to the engineering application of H 2 /O 2 PEMFCs.
PEMFC system and obtained the spatial distribution of water content in the membrane electrode by analyzing the two-phase flow transfer in the fuel cell. Xu et al. [26] investigated the selfhumidification effect of the fuel cell with cathode exhaust gas recirculation. Kim et al. [27] investigated the effects of blowers and backpressure control valves in 80 kW class PEMFCs with dual-recirculation subsystems and developed a performance map of the fuel cell stack for different loads. Jiang et al. [28] proposed a PEMFC system with anode and cathode recirculation to improve the durability and startup capability and showed that cathode recirculation could cause self-humidification. Zhao et al. [29] investigated the dynamic performance of a PEMFC system with dual recirculation using an orthogonal experimental design and showed that voltage clamping at low current densities can inhibit catalyst degradation under the low-load operating strategy. To quantitatively compare the self-humidification effect of a fuel cell stack, Shao et al. [30] proposed a dynamic model of a PEMFC system with anodic and cathodic pump-based exhaust gas recirculation and demonstrated that anode recirculation was a better solution within their energy efficiency and mechanical constraints. Dynamic characteristics of fuel cells based on dual exhaust gas recirculation subsystems have received sufficient attention, particularly in the context of self-humidification and system efficiency. Nevertheless, only a few studies have focused on the water management and chemical degradation mechanisms of key components after the continuous operation of PEMFC systems with dual gas recirculation. [31] The major obstacles affecting the industrialization and commercial application of fuel cells are degradation and the limited catalyst lifetime. [32,33] In this study, we conducted a standardized process to investigate the performance and degradation mechanism of PEMFCs with dual exhaust gas recirculation. The recirculation subsystems of the fuel cell were designed to achieve a high fuel gas utilization rate and effective water management inside the stack, which consisted of 3D-printed ejectors and a customized recirculation pump. The electrochemical performance degradation and morphological characterization of the fuel cells under different operating strategies were performed after 50 h of durability experiments. Static contact angle (CA) measurements, scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS), transmission electron microscopy (TEM), and X-ray photoelectron spectroscopy (XPS) were performed to characterize the chemical degradation mechanism of each component of the MEA under different operating strategies. In addition, constant-current polarization curves were obtained to evaluate the performance degradation of the PEMFCs.

Electrochemical Performance Degradation
Polarization curves are the most direct tools for characterizing the output performance of PEMFC stacks. The polarization curves of the fuel cells were measured before and after the durability experiment, while the test conditions were consistent. Figure 1a-c shows the polarization and power curves of the stack obtained under three different operating strategies. The performance degradation rate of the fuel cell stack under different current densities is shown in Figure 1d. After prolonged continuous operation, the stack performance decreased to varying degrees under different operating strategies. Moreover, the performance degradation rate of the stack showed a positive correlation with the operating current density in the DEAC and the ejector modes. At a current density of 400 mA cm À2 , the stack performance degradation rates reached 16.50%, 7.49%, and 0.71% in the DEAC, ejector, and recirculation pump modes, respectively, indicating that the cathodic and anodic dualrecirculation operation strategy could significantly reduce the performance degradation rate of the stack after prolonged operation. Forced convection of gases occurred inside the stack, which purged the generated liquid water and alleviated water flooding in the stack. Finally, the performance degradation rate of the stack was negligible when the cathode was assembled using a recirculation pump; however, the parasitic power of the pump reduced the efficiency of the entire system.

Morphological and Hydrophobicity Changes in Gas Diffusion Layer (GDL)
The pore structure and hydrophobic condition of the pore surface impact water management in the gas diffusion layer (GDL). Some loss of the carbon fiber and hydrophobic polytetrafluoroethylene (PTFE) was detected after the durability test, which could lead to local water transmission and performance degradation of the GDL. [34] Figure 2 shows the results of static contact angle measurements of the GDLs separated from the MEA measured after the durability tests under different operating strategies. The gas diffusion backing surface of the GDL was the contact angle measurement side. The static contact angle of the GDL was %150°i n freshly prepared MEA. The GDL contact angle decreased significantly after the 50 h durability test, more notably under the DEAC operation strategy. The contact angle of the cathode GDL in cell 10 decreased to 133.0°, and the contact angle of the remaining cells decreased to %144°. The hydrophobicity of the GDL was significantly degraded after the long operation, resulting in a decrease in the water management capacity and an increase in the GDL resistance. The hydrophobicity of the GDLs in the ejector and pump modes barely changed, verifying that the recirculation strategy mitigated water flooding in the PEMFC stack.
The hydrophobicity of the cathode GDLs exhibited the same trend for single cells within the stack under the same operating strategy, indicating that the single cell near the cathode exit bears a greater risk of degradation and severe water flooding. Therefore, during the actual operation of high-power fuel cell stacks, more attention should be paid to the water management of the MEA at the cathode outlet to ensure the performance homogeneity of every single cell.
To further investigate the changes in the hydrophobicity of the GDLs, the surface morphology of the GDLs was characterized after durability testing, as shown in Figure 3. No significant differences were observed in the pores and the carbon fibers on the surface of the GDLs after the experiment; however, substantial cracks and PTFE shedding occurred on the surface of the substrate layer of GDLs under the DEAC operation strategy.
PTFE plays an important role in maintaining the hydrophobicity of GDL, and its loss may be one of the factors contributing to the reduced water management capacity of GDLs.

Dimensional Changes in Catalyst Layer (CL) and Degradation of Pt
To further investigate the mechanism of the performance degradation of the PEMFC stack under different operation strategies, downstream regions of the MEA near the oxygen outlet MEA were examined with SEM after the experiments. Different single cells inside the fuel cell stack were also analyzed. A schematic of the MEA and the SEM images of freshly prepared MEA are shown in Figure 4. A diagram of the membrane cross section and average thickness of the cathode catalyst layer (CL) in MEA after 50 h of continuous operation under different operation strategies is shown in Figure 5. During the continuous operation of the stack, cathode-generated liquid water was dislodged by gravity and accumulated at the bottom of the flow channel, obstructing the oxygen transfer channels and causing starvation in the local area. At this point, the oxygen supply was insufficient to maintain the electrochemical reaction, causing the cathode to experience high potentials and consequently leading to carbon corrosion and loss of the Pt catalyst. [35] Cross-sectional analysis  www.advancedsciencenews.com www.advenergysustres.com results revealed that the thickness of the cathode CL was unchanged in the pump and the ejector operating strategies, but the thickness of the exit area decreased slightly (cell 10). The decrease in the cathode CL thickness in the DEAC operating strategy was easily visible, especially at the oxygen exit area (cell 10), where the average thickness of the cathode CL decreased from 7.81 to 5.52 μm at a decay rate of 29.3%. When water flooding occurs on the cathode side, the oxygen reduction reaction (ORR) cannot meet the demand of the external current. Thus, to maintain the charge balance, the carbon carrier reacts with water and generates hydrogen gas locally, resulting in a sharp decline in the performance of the PEMFC. The reaction mechanism is as follows. [36] Anode∶ Continuous water flooding in the PEMFC stack causes severe degradation of the Pt particles in the CL and a significant reduction in the electrochemical surface area due to the Ostwald ripening effect; consequently, small Pt particles tend to dissolve and deposit on larger Pt particles. This process is driven by the surface energy of the Pt particles, which makes the whole system more stable. [37,38] Pt dissolution∶ Pt À 2e À ! Pt 2þ (3) The Ostwald ripening effect is exacerbated by increased platinum ion concentration after the dissolution of metallic Pt. In addition, platinum ions detach from the carbon carrier migrate and deposit onto the ionomer or the proton exchange membrane. Pt particles without a carbon carrier lose their catalytic effect because of the lack of electron transfer channels. [34] After the durability test, the cathode CL of the MEA was gently scraped off, and the particle size of the catalyst was measured by TEM. In a freshly made catalyst, the diameter of the Pt nanoparticles well dispersed in an ethanol solution was %3-5 nm. Figure 6 shows the TEM images of the cathode catalyst after being used under the three operating strategies. We counted about 100 Pt particles in the TEM images to digitize the particle size distribution. Pt particle size in the cathode CL of the fuel cell was %4 nm, with no significant agglomeration after operating in a dual recirculation strategy, which is the best state for the Pt particles to demonstrate catalytic activity. In the case of the DEAC operating strategy, agglomeration was observed at the 20 nm scale, with the largest catalyst particle size exceeding 10 nm. Agglomeration seriously impacted the specific surface area of Pt nanoparticles, leading to severe degradation of the electrochemical surface area and loss of mass specific activity. [39] In addition, EDS data showed that under the DEAC operating www.advancedsciencenews.com www.advenergysustres.com strategy, the Pt particles in CL dissolved, oxidized to Pt 2þ , and migrated out of the MEA (Figure 7). The elemental composition of the surface and the electronic properties of the catalyst significantly influences the catalytic performance of the ORR. [40,41] XPS was performed on the cathode CL of the MEA to characterize the relationship between the microscopic properties of the catalyst and its performance. In addition, the results obtained from our tests were sorted to better depict the distribution of different valence states of Pt. Figure 8 shows the high-resolution XPS results for Pt 4f under the various operating strategies. The high-resolution Pt 4f spectrum was deconvoluted to Pt (0) peak (71.6/74.95 eV) and Pt (þ2) couples (72.5/75.7 eV). [42,43] The relative atomic percentages of the Pt (0) state and the Pt (2þ) state are calculated in Table 1. The relative atomic percentages of the Pt (2þ) state for cell 1 under the DEAC operating strategy were calculated to be higher (47.87%) than those under the ejector operating strategy (46.36%) and pump operating strategy (43.62%). A similar trend was observed for the relative atomic percentages of the Pt (2þ) state in cell 10. There were significant differences in the ORR catalytic activity of Pt with different metal valence states, with Pt (0) state exhibiting the highest activity for the ORR. [44] This trend implies that the cathode CL will be flooded for a long time when the fuel cell stack is operated continuously under the DEAC strategy, leading to catalyst oxidation and dissolution. At this time, the surfaceactive sites of the catalyst are significantly reduced, which is consistent with the macroscopic degradation of the polarization properties exhibited by the fuel cell stack.

Degradation of Polymer Electrolyte
Nafion membranes are often combined with catalyst materials to form the MEA of fuel cells, and electrochemical reactions occur at the phase interface. The complex environment of MEA increases the sensitivity of the membrane structure to chemical/ electrochemical conditions. Studies have shown that membranes are subject to some degree of chemical/electrochemical degradation during operation. [45,46] In general, the attack of free radicals (including hydroxyl radical (·OH) and hydroperoxyl radical www.advancedsciencenews.com www.advenergysustres.com (·OOH)) is the main cause of the chemical degradation of perfluorinated sulfonic acid (PFSA) membranes. [47][48][49] Hydrogen peroxide, which is produced in fuel cells via the ORR, is the source of these attacking groups.
During the operation of a PEMFC, metal ions are generated by the corrosion of metallic materials in the stack, which then diffuse into the proton exchange membrane via the water transfer process and react with hydrogen peroxide to produce free ·OH and ·OOH radicals. Figure 9 depicts the main mechanisms underlying the chemical degradation of the PFSA membranes. Chemical degradation of the proton exchange membrane results in the loss of F and C elements. Hence, EDS characterization was performed on the proton exchange membrane after the durability test in order to monitor the chemical degradation of the membrane under different operating strategies of the PEMFC stack, as shown in Figure 10. Degradation in cell 10 located near the oxygen outlet of the stack was the most severe; therefore, we only analyzed the elemental distribution of this cell. Figure 10 shows the nonuniform concentration distribution of F and C elements in the proton exchange membrane after 50 h of accelerated experiments. In addition, elemental F in the membrane was lost or diffused into the CL because of the attack of free radicals on the membranes molecular structure. The most significant loss of elemental F and C was observed in the proton exchange membrane operating in the DEAC strategy, indicating that more free radicals were generated due to prolonged water flooding. In contrast, the F and C concentrations in the proton exchange membranes operating under the dual-recirculation strategy were more evenly distributed and showed lower losses as compared to those in the DEAC mode.
In addition, all described membrane degradation reactions produce hydrofluoric acid (HF), which makes the fluorine emission rate a reliable indicator of membrane stability. This parameter has been widely used in existing studies. [34,46,50,51] Figure 11 shows the changes of fluorine release rate in the cathode and anode side drainage during different operating conditions. The fluorine ions in the cathode and anode water are the products of the degradation of PFSA resin in the proton exchange membrane, and the presence of fluorine ions in the wastewater indicates the degradation of the proton exchange membrane. The concentrations of fluoride in the cathode and anode did  not show a clear pattern, which is consistent with the results of Inaba et al. [52] However, the fuel cell operation mode has a significant effect on the concentration of fluoride in the generated water, which complements the EDS results described earlier.

Conclusion
In this study, a standardized process was conducted to investigate the performance and degradation mechanisms of a PEMFC with dual exhaust gas recirculation. Analysis of the electrochemical performance degradation and morphological characterization of the PEMFCs was performed under different operating strategies after 50 h of durability experiments. The results showed the following. 1) The cathodic and anodic dual-recirculation operation strategies could significantly reduce the performance degradation of the stack after long-term operation. The degradation rates decreased from 16.50% to 7.49% and 0.71% under the ejector and the recirculation pump operation strategies, respectively, at a current density of 400 mA cm À2 .
2) Visible degradation of the MEA occurred in the oxygen exit area (cell 10). Continuous water flooding of the fuel cell stack caused cracks and PTFE shedding on the substrate layer of the GDLs, and a significant reduction in the thickness of the   www.advancedsciencenews.com www.advenergysustres.com cathode CL was observed.
3) The MEA will be inevitably flooded for a long time when the fuel cell stack is operated continuously under the DEAC operation strategy, causing severe degradation of the Pt particles in the CL dissolution, oxidation, and agglomeration. Membrane degradation was also confirmed by the loss of elemental F and C from the proton exchange membrane. 4) The exhaust gas recirculation devices (ejector/pump) in the fuel cell stack can effectively mitigate water flooding and chemical degradation of the MEA. However, the performance improvement is compromised by the shortcomings of the recirculation device itself, such as the parasitic power and manufacturing cost.

Experimental Section
Design of the Recirculation Subsystem: A gas recirculation strategy is usually adopted to ensure the continuous operation of H 2 /O 2 PEMFCs. [53] The ejector enables the full use of the pressure difference caused by the high-speed primary fluid to continuously entrain the exhaust fuel gas at the outlet of the PEMFC. However, only one optimal operating point exists for a fixed-size ejector, and only one good entrainment ratio is possible within a particular work range. [54] Therefore, it is necessary to determine the optimal design conditions of the ejector according to the operating range of the PEMFC system.
The performance of the ejector was defined by the entrainment ratio ε.
where W P and W S denote the mass flow rates of the primary and secondary fluids, respectively. The flow of the convergent nozzle was divided into two different regions, subsonic and sonic, based on the flow characteristics of the convergent nozzle. The critical pressure ratio v cr of a convergent nozzle was described as where κ is the specific heat ratio of the gas. W P can be obtained using the isentropic flow relations and the energy balance law [55] for sonic flow ðP S =P P ≤ ν cr Þ for subsonic flow P S =P P > ν cr ð Þ (10) where ψ P is the isentropic coefficient.
The mass flow rate of the secondary flow is expressed as [56] where ρ S is the average density of the secondary flow and n S is the molar flow rate of secondary flow. M 0 represents the molecular weight. n V is the exponent of the velocity function which is expressed as a function of the pressure ratio and diameter ratio of the ejector.
where β P ¼ P 0.8 S =P 1.1 P and β D ¼ D m =D t . For an ideal gas, the energy balance of the primary and secondary flows in the ejector can be described by where the energy loss, E loss , of the primary and secondary flows in the ejector can be approximated as The design of the ejector requires the knowledge of four parameters: pressures of the primary fluid, entrained fluid, mixed fluid, and the mass flow rate of the primary fluid, which is used to maximize the entrainment ratio of the ejector. The important structural dimensions of the ejector were then obtained based on the empirical formulations earlier. [57,58] A 1 kW PEMFC (N cell Â I = 10 Â 150 A) was used for the experiment, and the gas pressure in the front of the stack inlet was kept constant at 100 kPa. The operating temperature of the fuel cell stack under all test conditions was set to 40°C. The design dimensions of the ejectors used in this study are listed in Table 2.
The ejector used in this study was fabricated by 3D printing, [59] and direct metal laser sintering (DMLS) was used to manufacture the ejector, which prints devices by heating and melting metal powder layers with a high-power laser beam. The primary advantage of the DMLS technology is that it does not require expensive and time-consuming pre-and postprocessing treatments, and the manufacturing accuracy could reach AE 0.02 mm. [60] Unlike other laser 3D printing technologies, DMLS technology allows the material structure and mechanical properties of the printed part to be adjusted by adjusting the printing parameters. [61][62][63] After 3D printing, the ejector was finalized at Hubei Huacheng 3D Technology Co., Ltd. The inner and outer surfaces of the ejector were polished to meet precision requirements, particularly at the nozzle. The final processed ejectors are shown in Figure 12a.
Despite the poor adaptability of the ejector to the operating conditions, the mechanical recirculation pump enabled sufficient gas supply and high efficiency under various operating conditions. The recirculation pump used in the experiments was purchased from Jiangsu Shen Hygen Technology Co., Ltd., with a customized structure size and rated power. The structure of the fuel cell exhaust gas recirculation pump is shown in Figure 12b.
Experimental System and Scheme: For this experiment, a commercial MEA produced by Wuhan WUT New Energy Co., Ltd., consisting of a CCM and a GDL, was used. The catalyst slurry was prepared by mixing a certain amount of the catalyst, dispersant, and Nafion solution and was directly coated on the Nafion XL proton exchange membrane. The platinum loading was 0.4 mg cm À2 for both the cathode and anode of the single cells. The graphite bipolar plates were designed and manufactured in-house. The PEMFC stack was assembled using ten pieces of MEA. The geometric parameters of the PEMFC stack are presented in Table 3. After assembly, the fuel cell stack was tested for gas tightness, that is, no air leakage or crossover in the stack. The experiments were conducted using a homemade test bench, designed especially for hydrogen-oxygen fuel cell stacks (Figure 13a), with precise control of various operating parameters, including the electronic load, operating temperature, and the inlet pressure of the stack. The electronic load provided a dynamic load for the fuel cell within a current load range of 0-300 A. The single-cell voltage monitoring range was À1.5 to þ1.5 V with a   measurement accuracy of 0.1% V. The cooling water circuit of the fuel cell stack was equipped with a condenser to regulate the operating temperature of the stack. The PEMFC stack and the water-cooled graphite bipolar plate with an electrochemical active area of 500 cm 2 are shown in Figure 13b,c, respectively. The experimental procedure of the study is shown in Figure 14.
During the experiment, the PEMFC stack was placed vertically to take advantage of the gravity-assisted removal of the water generated inside the fuel cell. The electronic loading current of the stack was maintained at 150 A per the optimal operating conditions of the ejector. The anode and cathode were fed with 99.999% pure hydrogen and oxygen, respectively, without humidification. Then, 50 h durability tests were conducted on the PEMFC stack with dual-gas recirculation. The test configurations were divided into two cases: a 3D-printed ejector or mechanical recirculation pump was adopted in the cathode exhaust gas recirculation subsystem, and an ejector was always used in the anode exhaust gas recirculation subsystem. A durability test was conducted in a PEMFC stack with a DEAC for comparison. Under the three test conditions, the cathode and anode outlets were equipped with solenoid valves, which were triggered for 2 s every hour during the continuous operation test to purge the accumulated liquid water and impurity gases. The experiments were followed by hydrophobicity testing of the GDLs in the MEA (LAUDA Scientific LSA100), as well as morphological characterization and elemental analysis of different parts of the MEA to investigate the effects of various operating strategies on fuel cell performance decay and MEA degradation. The cross sections of initial and degraded MEAs were analyzed by SEM via riving them in a liquid nitrogen atmosphere measured by Sirion 200 (FEI Co., Holland). TEM was carried out by TecnaiG2 20 (FEI Co., Holland)  Figure 13. a) Schematic of the PEMFC with dual recirculation (cathode and anode recirculation subsystems were removed when operating in the DEAC mode, and the solenoid valve at the stack outlet was kept closed). b) The PEMFC stack was assembled using ten straight-flow channel graphite plates. c) Water-cooled graphite bipolar plate with an electrochemically active area of 500 cm 2 .