Designing the Platinum Catalyst Layer for Improved Performance and Durability in Anion Exchange Membrane Water Electrolysis

Abstract

To lower the cost of hydrogen produced by anion exchange membrane water electrolysis (AEMWE), it is critical to reduce the use of platinum group metal (PGM) catalysts within the device. While iridium has been successfully replaced with PGM-free catalysts at the anode, platinum-based (Pt) cathode catalysts are still required to meet the activity and durability targets. This study investigates the impact of commercial Pt/C catalyst loading, ionomer type and content, and electrode fabrication method on the cathode catalyst layer properties and AEMWE performance with the aim of determining the feasibility of reduced Pt loadings. While increased Pt loading is found to improve beginning-of-life performance, the effects are minimal above 0.6 mg/cm². Ink characterization shows that ionomer type and content affect the ink stability, particle size, and percent of unbound ionomer, which further impact the homogeneity of the sprayed catalyst layers. The 5% PiperION cathode exhibited the highest performance, which may be attributed to a balance between the small particle size and the low proportion of unbound ionomer, minimizing kinetic and transport losses. Theoretical calculations show that the ionomers interact differently with the Pt surface, causing different surface charges and water adsorption strength and activating different mechanisms for hydrogen evolution. Pt-PiperION lowered the enthalpy of water-splitting by 0.1 eV compared to Pt alone and allowed for equal site access between adsorbed H* and OH* (both adsorbed at circa −2.2 eV). Although catalyst-coated membrane (CCM) fabrication techniques are desirable for scale-up, no performance enhancement is observed compared with the catalyst-coated substrate approach. Durability, as measured by degradation rates, Pt loss, and catalyst layer restructuring, was found to improve with increased Pt loadings, higher ionomer content, and CCM architectures. These findings provide important insight into the significant role of the cathode in AEMWE and strategies for maintaining the performance with low Pt loading or PGM-free catalysts.

Introduction

Anion exchange membrane water electrolysis (AEMWE) is a promising technology for the production of low-cost hydrogen using platinum group metal (PGM)-free components. While first-row transition metal-based oxides have shown to be promising catalysts for the anodic oxygen evolution reaction (OER) in AEMWE, , Pt or PtRu PGM-based electrodes remain the most active and stable option for the cathodic hydrogen evolution reaction (HER). The reported HER kinetics on Pt are several orders of magnitude slower in alkaline than in acid, , meaning that AEMWE devices can be cathode limited and comparatively high Pt loadings are required. Therefore, to lower costs while maintaining high performance, it is critical to design the cathode catalyst layer to optimize the Pt utilization.

In the cathode catalyst layer, HER can only occur at catalyst sites where the reactants (water and electrons) are present and the products generated, i.e., hydrogen (H₂) bubbles and hydroxide (OH^–) ions, can be removed. Generally, performance is expected to improve with increased catalyst loading, although this is dependent on those active sites being accessible. Thus, the catalyst layer must be designed for effective electron conduction from the transport layer toward catalyst sites, water transport from the electrolyte or membrane (for cathode wet or dry operation, respectively), H₂ bubble removal from the catalyst layer, and OH^– ion transport through the liquid electrolyte and/or ionomer network, depending on the use of pure water or supporting electrolyte (e.g., potassium hydroxide). These important transport phenomena may be affected by the chemical and structural properties of the catalyst layer. For example, in anodic catalyst layers, the intrinsic catalyst conductivity, catalyst loading, ionomer content, and pore/fiber dimensions of the porous transport layer (PTL), among other properties, have been shown to impact catalyst utilization and overall device performance. ,− For the cathode, a few studies have identified mass transport of water and H₂ gas as the critical limiting factor, and strategies such as modifications of the gas diffusion layer (GDL), use of cathode dry operation to facilitate gas diffusion, and optimization of the water uptake of the ionomer have been studied to address these mass transport limitations. −

In addition to the role of the ionomer in transport, the interactions between the ionomer and catalyst in the ink impact their aggregation behavior, ionomer/catalyst local interfaces, and the overall structure and morphology of the catalyst layer and the accessibility of active sites once the electrode is fabricated. , Prior work in fuel cells has demonstrated that within a catalyst layer for a membrane electrode assembly (MEA), Nafion adsorbs to Pt/C primarily via hydrophobic interactions with the carbon support. ,, Relatively little is known about the interactions of anion exchange ionomers (AEIs) with the catalyst in the ink, and furthermore some differences in electrostatic interactions may be expected due to the negatively charged functional group of the AEI. Within the catalyst layer, the AEI/catalyst interface has been shown to impact activity and ionomer stability under OER conditions. The location of the ionomer in the catalyst layer has also been shown to affect performance, such as an ionomer layer at the interface with the membrane leading to improved interfacial contact. In summary, it is expected that catalyst–ionomer interactions in the cathode will impact performance, but the optimum catalyst layer ink design choices for Pt utilization, and especially durability, are not well understood.

The method of fabrication may also affect the catalyst layer morphology, performance, and durability. In PEMWE, it is standard to use a catalyst-coated membrane (CCM) approach, with the catalyst layer commonly deposited by ultrasonic spray coating or, at more manufacturing-relevant scales, by rod coating or decal transfer. , This is beneficial for PEMWE due to improved interfacial contact to the membrane that enhances ion transport, which is critical for operation in water. In AEMWE, a catalyst-coated substrate (CCS) approach has been more common because of challenges with membrane mechanical integrity and adhesion in CCM fabrication. , Additionally, in supporting electrolyte and when using catalysts with low electronic conductivity, electronic interfaces may be more limiting than ionic ones. However, CCM fabrication methods remain of interest due to improved scalability and uniformity, and there is a need for a comprehensive comparison with the “baseline” CCS configuration. ,,−

This study systematically investigates the roles of cathode catalyst layer composition and fabrication method in AEMWE performance and durability and connects the properties of the ink, catalyst layer morphology, and performance. Theoretical calculations further reveal the impact of the ionomer on the catalyst surface charge, the reaction enthalpy of HER mechanisms, and the binding of key intermediates such as H₂O*/H*. The impact of cathode properties on durability is further probed using chronopotentiometry (CP), with posttest microscopy and inductively coupled plasma-mass spectrometry (ICP-MS) characterization to understand the role of the cathode catalyst layer in degradation. These findings on ink formulations and catalyst layer properties provide crucial insight into how AEMWE performance can be maintained with low Pt loadings and provide guidance for future PGM-free cathode electrode fabrications.

Experimental Section

Materials

All materials were obtained from commercial suppliers and used without further treatment unless specified. Pt/C (TKK TEC10E50E, 47 wt % Pt), NiFe₂O₄ (US Research Nanomaterials, 98%), PiperION-A TP-85 membrane (Versogen, 80 μm) and ionomer (Versogen, 5% in ethanol), Nafion 117 ionomer (Sigma-Aldrich, 5% in alcohol/water), Ni fiber PTL (Bekaert, Currento PTL Ni-60/250), carbon GDL (Avcarb MGL280, 80280-40), KOH (EMD Millipore, Emsure grade) for single-cell testing, NaOH (Sigma-Aldrich, TraceSELECT grade, 99.9995%) for half-cell measurements, n-propanol (nPA, Sigma-Aldrich, OmniSolv, high-performance liquid chromatography grade), ASTM I quality deionized water (DIW, Milli-Q), and HNO₃ (Fisher Chemical, Optima grade, 67–70%).

Ink Formulations and Catalyst Layer Fabrication

For the multilayered fabrication of CCS and CCM electrodes, the catalyst inks consisted of 5 mg/mL Pt in a solvent mixture of DIW and nPA with a PiperION or Nafion ionomer. Note that as the ionomer content was varied, the Pt mass and solvent volumes were held constant, so the Pt concentration (5–5.7 mg/mL) and the total alcohol content (43–45%) varied slightly. The catalyst inks were bath sonicated using ice to control the temperature and prevent any dangerous overheating of Pt in the presence of alcohol. The spraying methods consist of using a pressurized nitrogen gas line to aerosolize the catalyst ink onto the surface of the substrate, either GDL or PiperION membrane, which is adhered to a vacuum hot plate at 80 °C to aid rapid solvent evaporation and inhibit ink flooding. For hand spraying, the airbrush (Grex Power Tools Tritium Tg5 Airbrush) nozzle was maintained at 10–15 cm from the substrate to reduce droplet aggregation and multiple spraying passes were conducted as described in Figure a–b to attain the desired catalyst loading. For ultrasonic spraying, a Sono-tek ExactaCoat ultrasonic spray system with a 25 kHz Accumist nozzle was used. For rod coating, the catalyst concentration was increased to 21.5 mg/mL. The ink was ball milled overnight with 5 mm silica spheres (Glenn Mills). The catalyst layer was fabricated by using an automatic film applicator (Qualitech Products Industry) at 55 mm/s using an RDS 20 Mayer rod with the hot plate maintained at 50 °C. The CCMs were fabricated by using a dry membrane. To prevent buckling of the membranes, they were secured on a vacuum plate, dried gently to maintain interior hydration, and quickly removed from the heat after ink deposition. Anode catalyst inks were created similarly, using a concentration of 11.4 mg/mL NiFe₂O₄ with 10% PiperION in a solvent of DIW and nPA; this was hand sprayed onto the Ni fiber PTL for a target loading of 0.5 mg/cm².

7.

Electrode fabrication techniques for single- and multiple-layer catalyst ink deposition. (a) Hand spraying method for CCS or CCM fabrication using an airbrush. The black lines indicate the crosshatch spray pattern for multilayer deposition of the catalyst ink. (b) Ultrasonic spray method for CCM fabrication using an automated spray pattern with an Accumist nozzle. (c) Rod coating method for CCM fabrication for a single-layer deposition. Images were generated using BioRender.com. Top-down SEM images (top) and EDS maps (bottom) of CCMs prepared by (d) ultrasonic spraying and (e) rod coating at 0.2 mg/cm² loading.

Electrochemical Testing

For AEMWE testing, the MEA consisted of a 5 cm² area sandwich of the sprayed cathode and anode catalyst layers with a PiperION membrane (80 μm). Prior to assembly, the membranes were pretreated in a 3 M KOH solution for 48 h. Tests were conducted using custom-built, 25 cm² hardware with stainless steel (316L) end plates, Au-coated current collectors, and triple-serpentine Ni flow fields under 20% compression. A 1 M KOH electrolyte was fed to both the anode and cathode at 50 mL/min. The electrolyte was preheated, with both electrolyte feeds and the cell held at 80 °C. Beginning-of-life performance was assessed with an Autolab PGSTAT302N potentiostat with a 20 A booster (Eco Chemie, Metrohm Autolab) after a short conditioning procedure (2 h hold at 2 V). Potentiostatic polarization curves were taken via 2 min holds at voltages between 1.4 and 2 V. The thermodynamic water-splitting potential under operating conditions was calculated to be 1.179 V. Electrochemical impedance spectroscopy (EIS) was performed at each voltage step, using frequencies from 18 kHz to 1 Hz with an AC amplitude of 10 mV; the impedance spectra fit well to an equivalent circuit model comprised of inductance, high-frequency resistance (HFR), and two circuits of a resistor in parallel with a constant phase element. Additional measurements taken at non-Faradaic voltages of 1.25, 1.3, and 1.35 V were used to assess the catalyst layer resistance (CLR). , HFR, Tafel slopes, and CLR were used to perform a voltage loss breakdown analysis, as reported previously. , 100 h chronopotentiometry durability tests were conducted at 1 A/cm² using a Rigol power supply. End-of-life diagnostics were subsequently performed with the potentiostat.

Inks for rotating disk electrode (RDE) testing were prepared in a mixture of DIW and nPA with 10 wt % ionomer. The inks were applied to polished gold (Au) working electrodes, initially rotating at 100 rotations per minute (RPM), with the speed increased to 750 rpm during the drying process. RDE experiments were conducted in a three-electrode cell consisting of a Au mesh counter electrode, a Hg/HgO reference electrode, and catalyst-coated Au discs as working electrodes with 0.1 M NaOH electrolyte. The electrochemical measurements were carried out using an Autolab PGSTAT302N potentiostat, with no booster. Linear sweep voltammograms were recorded from 0.2 to −0.8 V at a scan rate of 20 mV/s and an RDE rotation speed of 2500 rpm. This was preceded by five cyclic voltammetry cycles from 1.2 to 1.8 V to condition the catalyst and assess the hysteresis. The working electrode speed was controlled by a modulated speed rotator (Pine Research Instrumentation, AFMSRCE). Cyclic voltammograms were taken without rotation between 0 and 1.6 V at three different scan rates: 100, 50, and 20 mV/s. All potentials were referenced against the RHE, and overpotentials were calculated relative to the thermodynamic potential, adjusted for the altitude of Denver. The solution resistance was measured in situ using a built-in current interrupter, and the exchange current density was calculated using the Butler–Volmer equation.

Characterization

Samples were prepared for scanning electron microscopy (SEM) mounted with carbon tape to 0.25 in. Al SEM stubs. Top down images and energy dispersive X-ray (EDS) maps were collected using a Thermofisher Apreo Lo Vac SEM instrument operated at 15 kV and 6.4 nA with a Bruker Xflash 60 EDS detector. Cross-section SEM images were collected using an FEI Helios NanoLab 660 DualBeam SEM/focused ion beam (FIB) instrument at 5 kV and 0.2 nA and an Everhart–Thornley detector for secondary electron images. Cross-section samples were prepared using the FIB. Cross-sectional EDS elemental maps were collected using an Oxford 170 mm² Ultim Max SDD EDS detector at 15 kV and 6.4 nA with Oxford’s AZtec software. The FEI Helios NanoLab was also used to prepare scanning electron microscopy (STEM) lamella for analysis using an aberration-corrected Thermo Fisher Themis Z Scanning Transmission Electron Microscope at 300 keV. Atomic resolution high-angle annular dark field and bright field (HAADF/BF) STEM images and EDS data were collected using a large-solid angle (0.7 sr) silicon drift EDS detector (Super X). STEM images were segmented into their different components (e.g., catalyst, ionomer, GDL, etc.) using the free software ilastik. The porosity analysis was performed using Python 3.9 within the Anaconda distribution (version 2022.05) and executed in JupyterLab (version 3.4.4). The workflow included loading the segmented image parts, applying Otsu’s thresholding to segment the porosity channel, refining the binary mask using morphological operations (dilation, erosion, and opening), and calculating pore sizes and tortuosity. All analyses were conducted on a Windows 11 machine with 16 GB of RAM and an Intel Core i7 processor. For the ink analysis, all inks were prepared as defined above, with a different supplier for Nafion (Nafion D520 (Ion Power, Inc.)) due to a product discontinuation. To characterize particle size, inks without dilution were placed in a glass cuvette at 25 °C immediately after mixing and measured with a Malvern ZetaSizer Pro instrument using the backscattering angle. The pure solvent viscosity was used in the analysis. Data showed monomodal distributions, and the z-average was recorded at least 5 times per ink with at least one unique replicate per ionomer concentration. All inks were stable over the measurement period, except for the 1% PiperION ink. To measure zeta potential, the inks were diluted 100× and placed in a dip cell (Malvern), analyzed via the Smoluchowski equation, and otherwise used the same measurement conditions as for particle size. To quantify how much ionomer adsorbed to the catalyst particles within the inks, UV–vis (Cary 60 spectrophotometer, Agilent) of the supernatant was used following the procedure adapted from Rajupet et al. Briefly, to separate free ionomer from catalyst particles containing adsorbed ionomer, inks were centrifuged immediately after mixing/fabrication for 1 h. Catalyst particles plus adsorbed ionomer sedimented, while the free ionomer remained in the supernatant. The supernatant was removed and measured via UV–vis to quantify how much free ionomer remained in the solution. Unknown free ionomer concentration in a sample supernatant was quantitatively determined by measuring the absorbance relative to the absorbance of known concentrations to construct a calibration curve. The calibration curve was determined separately by plotting absorbance at a given wavelength versus known concentration of ionomer using a series of standards obtained via serial dilution. The R ² value on the calibration curve was over 0.998, allowing determination of free ionomer concentration for each sample with error smaller than the data points. The UV–vis peak at 300 nm was used for all PiperION absorbance measurements. Nafion is not inherently UV–vis active, and so 15 μM of a cationic dye malachite green oxalate in aqueous solution (>90% Sigma-Aldrich) was added, which has a peak absorbance at ∼617 nm. Once free ionomer was calculated from the supernatant, the adsorbed ionomer concentration was calculated from the mass balance (i.e., adsorbed ionomer = total ionomer added – free ionomer measured). Total adsorbed ionomer is then normalized by the total amount of carbon in the ink to better explain trends across different inks. This ratio is represented as Γ, which is the mass of adsorbed ionomer divided by the mass of carbon (which can be thought of as an effective adsorbed ionomer-to-carbon ratio). Error bars represent the standard deviation of multiple replicates. All measurements were made with at least two unique replicates. Catalyst ink viscosity was measured by using a modular advanced rheometer system (Thermo Scientific, Haake Mars) with a circular plate geometry of 35 mm. Pt dissolution during long-term AEMWE testing was assessed using a Thermo Scientific iCAP Q inductively coupled plasma–mass spectrometry (ICP-MS) instrument in the kinetic energy discrimination mode. Aliquots from the 1 M KOH electrolyte were taken at the end of testing and diluted 100× with 2% HNO₃ to acidify and reduce the K⁺ concentration. A calibration curve in the range of 1–50 ppb was prepared using a Pt standard (Inorganic Ventures, 10 ppm) for quantification. The calculated detection limit was 0.02 ppb. A commercial flow cell (BASi, MF1092) was used for the on-line ICP-MS experiments, with 60 μg/cm² Pt/C deposited on a gold disk substrate for the working electrode, a gold disk as the counter electrode, and a leakless Ag/AgCl reference electrode. 0.1 M KOH was flowed at 3.2 mL/min through the cell using a Kamoer M1-STP peristaltic pump; the flow was split and mixed with a 6% nitric acid internal standard to dilute the electrolyte prior to introduction to the nebulizer. A delay of 48 s between the flow cell and the detection was measured and accounted for.

Theoretical Calculations

Plane-wave density functional theory code Vienna Ab initio Simulation Package (VASP) 6.3.2 − utilized the Perdew–Burke–Ernzerhof (PBE) functional in conjunction with projector augmented wave method pseudopotentials. , In order to reflect the electrochemical environment of electrolysis, implicit solvation as implemented by VASPsol with a dielectric constant of water was applied. − All of the calculations presented here reflect well-converged, global break conditions for electronic (geometric) relaxation of 10^–6 (10^–5) eV. In order to accommodate the ionomer functional groups on Pt (111), a (3 × 3) surface was utilized for SO₃ and subsequent HER calculations and a (5 × 5) surface for piperidinium and subsequent HER calculations. The density of the Monkhorst–Pack grid centered at gamma was modified accordingly to be 4 × 4 × 1 for a Pt (111) surface grown to be (3 × 3) and 2 × 2 × 1 for a Pt (111) surface grown to be (5 × 5). A depth of 4 Pt layers was utilized for these surfaces and incorporated a vacuum gap >12 Å. A thorough minima search of the ionomer functional group was performed on atomic, bridging, and hollow sites with piperidinium and sulfur trioxide in monodentate, bidentate, and tridentate orientations rotated every 45°, resulting in 101 and 75 calculations, respectively. Utilizing the global minimum of the ionomer functional group, subsequent coadsorption with H₂O, OH, and H on neighboring Pt sites resulted in 300 calculations for piperidinium-O _x H _y and 227 calculations for sulfur trioxide-O _x H _y . Adsorption energies for the ionomer’s functional group were calculated from

$$E_{\mathrm{ion}}(\mathrm{eV}){=E}_{\mathrm{surface}+\mathrm{ion}}{-E}_{\mathrm{surface}}{-E}_{\mathrm{ion}}$$

where E _surface+ion refers to the total energy of the surface with the attached ionomer functional group, E _surface refers to the total energy of the surface alone, and E _ion refers to the ionomer functional group in an empty, asymmetric box. For HER intermediates such as H₂O, OH, and H, the adsorption energy arose from

$$E_{\mathrm{ads}}(\mathrm{eV}){=E}_{\mathrm{surface}+\mathrm{ion}+\mathrm{ads}}{-E}_{\mathrm{surface}+\mathrm{ion}}{-E}_{\mathrm{ads}}$$

where E _{surface+ion+ads} refers to the total energy of the surface with the attached ionomer functional group and HER intermediate, E _surface+ion refers to the total energy of the surface with the attached ionomer functional group, and E _ads refers to the HER intermediate specified (H₂O, OH, and H) in an empty, asymmetric box. Bader charge analysis provided an analysis of the charge-transfer characteristics at the interface between the surface, ionomer functional group, and HER intermediate. −

Results and Discussion

Mass Loading

To make AEMWE a cost-competitive and sustainable technology for H₂ production, it is critical that Pt use be eliminated or significantly limited. Toward this goal, it is important to understand the dependence of device performance on Pt loading. Many past studies have used high Pt loadings of 1 mg/cm² or higher, which is 10× higher than the 0.1 mg/cm² used in the “Future Generation MEA” standard developed for PEMWE, and 20× higher than the 0.05 mg/cm² target established by the United States Department of Energy (US DOE) for AEMWE. Here we demonstrate that varying the Pt loading from 0.05 to 1 mg/cm² causes drastic differences in catalyst layer structure and performance (Figure ). Top-down scanning electron microscopy (SEM) images show that the catalyst coverage of the carbon paper GDL is incomplete at 0.05 mg/cm², with complete coverage of the GDL fibers only at the higher loadings (Figure a–c). Energy dispersive X-ray spectroscopy (EDS) maps show the increasing coverage of Pt (green) and I (red); I corresponds to the ionomer, which has been ion exchanged from the carbonate form to the iodide form to allow for visualization. Cross-section images of the cathodic catalyst layers show that the thickness on coated fibers varies insignificantly between the different loadings, with lateral coverage as the larger effect (Figure S1).

1.

Top-down SEM-EDS images and maps showing Pt/C coverage on the C GDL at (a) 0.05, (b) 0.2, and (c) 1 mg/cm² Pt loadings. AEMWE performance for cathodes with Pt loadings between 0.05 (light blue) and 1 mg/cm² (dark blue). (d) Polarization curves, (e) EIS Nyquist plots at 1.9 V, (f) a summary of ohmic, kinetic, CLR, and residual/mass transport losses at 1 A/cm², and (g) mass activity. Polarization curves, voltage losses, and mass activity are reported as the average and standard deviation of tests performed in triplicate. (h) 100 h CP durability measurement at 1 A/cm² for the 0.05, 0.2 (black = standard), and 1 mg/cm² loadings. The rates of voltage increase are calculated over the entire test period.

Overall, the performance increases significantly as Pt loading increases (Figure d). From 0.05 to 0.6 mg/cm², the current density at 2 V (J@2) increases from 1.57 ± 0.03 to 2.58 ± 0.13 A/cm² and the cell voltage at 1 A/cm² decreases by 100 mV, representing an increase in efficiency from 78% to 83% (using the higher heating value of H₂) (Table S1). To better understand how Pt loading affects overall performance, voltage breakdown analysis (VBA) is used to identify the ohmic, kinetic, catalyst layer resistance, and mass transport or residual losses. As shown in Figure e, there is no significant change in HFR with loading, meaning that the ohmic loss at 1 A/cm² is the same for loadings up to 1 mg/cm² (Figure f). While the kinetic losses are lowered by increasing the loading from 0.05 to 0.1 mg/cm², there is no further improvement at higher loadings. The Pt mass activity is relatively constant from 0.05–0.2 mg/cm² but decreases at higher loadings (Figure g). Notably, the performance does not change between 0.6 and 1 mg/cm², indicating that the additional catalyst sites are not accessible. The most significant trend with loading is mass transport, where the improved uniformity and material density of the catalyst layer at the membrane interface with higher loadings (Figure a–c) may result in improved transport of water, KOH, and H₂ gas bubbles to and from these sites. These transport losses increase again between 0.6 and 1 mg/cm², indicating that increased density or intrusion into the pores of the GDL can also limit mass transport.

To further understand how Pt loading affects device stability, 100 h CP measurements were used at 1 A/cm² for loadings of 0.05, 0.2, and 1 mg/cm². These loadings were chosen to provide insight into degradation across the full loading range used in this study. For simplicity, the degradation rates will be reported and compared based on the total losses during the CP hold, although differences in initial loss rate and recovery behavior will be discussed where notable. For the 0.2 mg/cm² loading, the average degradation rate was 1.3 mV/h (Figure h). The degradation rate is higher for 0.05 mg/cm² (1.9 mV/h) and lower for 1 mg/cm² (0.5 mV/h), indicating that the instability is exacerbated at low Pt loadings. The performance losses in posttest polarization curves, particularly after exposure to open-circuit voltage (OCV), are much smaller, with the 0.2 and 1 mg/cm² samples showing improved performance compared to the beginning of the test (BOT) (Table S2, Figure S2). These improvements in performance after resting or OCV exposure are termed reversible losses and have been attributed to a variety of processes, including the reduction of oxidized components or bubble removal. Overall, AEMWE performance is found to be strongly dependent on Pt loading up to ∼0.6 mg/cm², primarily due to mass transport losses and poor catalyst layer coverage observed at low loadings, and higher loadings lead to lower degradation rates and more recovery.

Ionomer Effects

To improve performance without using additional Pt, we next aim to understand how the catalyst can be used more effectively through the engineering of the catalyst ink formulation. The catalyst ink is composed of Pt/C nanoparticles dispersed in a mixture of nPA and DIW, with a polymeric binder used to help uniformly disperse and bind the catalyst layer to the GDL substrate. AEIs are commonly used to serve these roles, as well as providing conduction of OH^– ions from the membrane toward the catalyst active sites. One common AEI is PiperION, a poly­(aryl piperidinium) polymer with a piperidinium cation and an ether-bond-free aryl backbone. An alternative approach when operating in a supporting electrolyte, which provides the necessary ionic conductivity and lessens the need for an AEI, is to use a nonanion conducting polymer, which can offer improved dispersion, binding, and chemical stability. − Nafion, a cation exchange polymer with a fluorocarbon backbone and a sulfonic acid functional group, has been shown to improve the durability of an AEMWE with a NiFe₂O₄ anode catalyst. Previous works focusing on the anode and cathode catalyst layers have shown that catalyst utilization can be improved by optimizing the ionomer content, typically between 10–20 total wt % of the catalyst layer, and ionomer chemistry for the catalyst and operating conditions. , Some of the identified effects of ionomer on catalyst layer structure have included agglomerate size, water transport, electrochemically accessible surface area, and adhesion. , Ionomer overlayers have also been identified as a route to improve the interface with the membrane, particularly for operation in pure water or dilute supporting electrolyte. , Here, we conduct a comprehensive analysis of the cathode ink formulation to understand the impact of catalyst–ionomer interactions on AEMWE performance by systematically tuning the ionomer content (between 1 and 30 total wt % of the catalyst layer) and ionomer identity (PiperION versus Nafion) with constant Pt mass loading (0.2 mg/cm²).

Figure a shows the impact of the ionomer content on performance. Intermediate ionomer contents have better performance, with a trend of 5% > 20% > 30% > 1% for PiperION. While the effect is smaller than that of Pt loading, decreasing PiperION content from 30% to 5% results in an increase in current density at 2 V from 2.32 ± 0.18 to 2.54 ± 0.04 A/cm² (Table S3). For Nafion, only samples with 5% and 30% ionomer were tested; these were chosen based on the PiperION results, which showed a significant impact on activity in this range, with very similar performance between 20% and 30%. 1% ionomer was excluded due to the low ink stability, which will be discussed later. The reverse loading trend is observed with Nafion, where increasing from 5% to 30% increases the current density at 2 V from 2.06 ± 0.03 to 2.37 ± 0.05 A/cm² (Figure b). Comparing the 5 and 30 wt % PiperION and Nafion cathodes, the VBA shows that the largest impacts are on kinetic and mass transport losses (Figures c–d, S3–S4). The PiperION cathodes have a small variation in kinetic losses, while the 30% Nafion cathode significantly improves kinetics compared to both 5% Nafion and all of the PiperION cathodes. Mass transport losses are lowest for 5% PiperION, increasing with higher ionomer content and with the change to Nafion, perhaps due to poor transport of OH^– with the acidic binder. Dry cathode operation is of interest for the production of dry and/or pressurized hydrogen, but there are concerns about sufficient water retention in the cathode catalyst layer. The ion exchange capacity of the ionomer, the amount of ionomer, and the catalyst layer porosity have all been shown to affect water retention. ,, Here, in short-term testing (4 h), the 1% PiperION cathode showed no significant difference in performance for wet and dry cathode operation up to 2 A/cm² (Figure S5); longer tests would be needed to determine whether this ionomer content is sufficient to prevent dry-out.

2.

AEMWE polarization curves for cathodes with (a) PiperION (red) content between 1 and 30 wt % and (b) Nafion (purple) contents of 5 and 30 wt %. Comparison of (c) kinetic and (d) mass transport losses for 5 and 30 wt % PiperION and Nafion. Polarization curves and voltage losses are reported as the average and standard deviation of tests performed in triplicate. (e) 100 h CP durability measurement at 1 A/cm² for cathodes with 5 and 30 wt % PiperION and Nafion ionomers. All cathodes have 0.2 mg/cm² Pt loading. 30% PiperION data (black) is reproduced from Figure .

While there are small differences in degradation rate as a function of ionomer, the performance trends are maintained over the 100 h durability test at 1 A/cm² (Figure e). The lowest degradation rate is found for 30% PiperION, with a small increase to 1.4 mV/h for 5% PiperION. The degradation rates are higher for Nafion, reaching 1.7 mV/h for the 30% sample, but overall, the ionomer is found to have minimal impact on stability. This is in contrast to a recent study of the anode catalyst layer, where Nafion improved stability compared to PiperION. Notably, both Nafion samples show larger improvements in the polarization curves at the end of the test (EOT) than the PiperION samples, which may indicate that the ionomer identity influences the reversible losses and recovery with exposure to the OCV (Figures S6–S7).

To better understand the impacts of the ionomer content and chemistry on the catalyst layer, the inks used to fabricate the above catalyst layers were analyzed with dynamic and electrophoretic light scattering (DLS/ELS) to assess particle size distributions and zeta potentials, respectively. UV–vis was also used to quantify the ionomer adsorption to the catalyst particles. The average particle size of the catalyst/ionomer aggregates within the ink increases for both types of ionomers as their content increases, as shown in Figure a. The exception to this is the 1% PiperION case, which was very unstable; therefore, the particle size is not reported. This demonstrates that small quantities of ionomer help stabilize the ink and break up particles, but increasing concentrations of ionomer induce further aggregation. When the two ionomers are compared, 5% PiperION and Nafion display near-identical particle sizes, while at 30% Nafion displays larger particle sizes than PiperION.

3.

(a) Ink z-average particle diameters and corresponding (b) zeta potential magnitudes, in which Nafion values are negative and PiperION values are positive, for inks with 1 to 30% PiperION (black) and 5 to 30% Nafion (purple) content. (c) Adsorption fraction, Γ, in grams of adsorbed ionomer/gram of carbon for each of the different ionomer weight % inks. (d) The total amount of ionomer added to the ink that is adsorbed (balance is free in the ink). (e) Correlation of AEMWE performance from Figure , as indicated by current density at 2 V, and the percent of adsorbed ionomer in the ink. Ink measurements are reported as the average and standard deviation of tests performed in triplicate. (f) Cross-sectional SEM images of the catalyst layers with (i) 5% PiperION, (ii) 30% PiperION, (iii) 5% Nafion, and (iv) 30% Nafion ionomer content. Cross-sectional STEM images of (g) 30% PiperION and (h) 30% Nafion catalyst layers.

Next, we measured the zeta potential, which is a measure of the effective surface potential and is reflective of the electrostatic environment around the catalyst particle. PiperION and Nafion display opposite value zeta potentials (Nafion is negative and PiperION is positive) due to their different functional groups, although their magnitudesin particular for the 30% ionomer caseare similar (Figure b). The 1% PiperION has a zeta potential value near zero, indicating that it is unstable, as corroborated by the lack of conclusive particle size data. As concentration increases, an inverted-U trend emerges (Figure b). This U-trend has been previously measured for Nafion-based Pt/C inks and was found to occur around the saturation point: at low ionomer contents, the ionomer adsorbs to the particles, causing the zeta potential magnitude to increase. Once the ionomer stops interacting as much with the Pt/C particle and instead remains free in solution (typically around the saturation point), the extra ionic strength due to additional ionomer screens the charge and causes the zeta potential magnitude to decrease. More systematic zeta potential measurements over a wider concentration range would be needed to definitively confirm this trend; however, this concept is supported by the adsorption measurements below.

UV–vis adsorption measurements (Figure c–d) demonstrate that while particle sizes and zeta potential magnitudes have similar trends across the two ionomers, they exhibit significantly different adsorption behaviors, particularly at low ionomer contents, as indicated by both the adsorption ratio Γ (g adsorbed ionomer/g carbon) in Figure c and the percent of the total ionomer added to the ink that adsorbs in Figure d. For PiperION, 100% of the ionomer adsorbs at 1% ionomer content (Figure d) and likely would continue to adsorb more if there were more ionomer available. This lack of free ionomer and lack of uniform ionomer coverage likely lead to the high instability/low zeta potential magnitude for the 1% PiperION cathode. As the ionomer content continues to increase, the ionomer continues to adsorb to the catalyst particles (Figure c), but the fraction that is free increases (% adsorbed decreased as shown in Figure d). Furthermore, the fact that the % adsorbed barely changes when going from 20% to 30% PiperION also supports the plateau/U-shaped trend exhibited by the zeta potential results (i.e., similar ratios of adsorbed versus free ionomer could cause a zeta potential magnitude plateau).

Despite the 30% PiperION and 30% Nafion displaying very similar adsorption amounts (i.e., comparable Γ values at 30% ionomer in agreement with the similar zeta potential magnitudes) and the generally similar trend of increasing Γ with increasing % ionomer in Figure c, the adsorption isotherm and mechanism are different between the two ionomers. For PiperION, increasing the total ionomer content in the ink decreases the fraction of adsorbed ionomer; the reverse is true for Nafion (Figure d). For Nafion, an equilibrium between free and adsorbed ionomer is always maintained, such that as total ionomer content increases, so does the % adsorbed (until the plateau value), as has been seen previously. In short, the Nafion partitions between free and adsorbed at all ionomer contents, whereas PiperION prefers to initially adsorb, and then the excess ionomer is free.

The opposing ionomer adsorption trends suggest different adsorption mechanisms and different adsorption conformations of the ionomers, which are likely to contribute to the observed performance differences in the catalyst layers. Comparing ink trends with MEA data for both ionomers, the highest-performing system has a greater fraction of adsorbed instead of free ionomer (Figure e). The negative impacts of excess free ionomer on catalyst layer morphology are shown in the heterogeneity and regions of low porosity of the dried catalyst layer for 30% PiperION (Figure f–g). For both ionomer types, we further find that kinetics improve with increased ionomer content due to improved ionomer/catalyst adsorption interactions, while mass transport is worsened due to larger agglomerate sizes and more heterogeneity in the catalyst layer (Figure c–d, Figure S8). These ink/MEA correlations have previously been demonstrated for other electrolyzer systems. For Nafion, the decrease in kinetic losses outweighs the increased transport loss, leading to an improved performance at 30% (Figure c–d). The opposite is true for PiperION, with only a small change in kinetics as the ionomer content is lowered, likely due to the initial adsorption and low free ionomer content, leading to optimal performance at 5% (Figure c–d).

Top-down SEM and cross-sectional STEM images of the 5% and 30% PiperION and Nafion catalyst layers show significant changes with the ionomer content. For both ionomers, increased ionomer content leads to increased coverage and density, particularly in the pores between carbon fibers, and an apparent decrease in the roughness of the catalyst layer surface (Figure S9). The elemental maps also indicate an increase in I and F signals for PiperION and Nafion as the ionomer content increased. There is some variation in catalyst layer coverage and thickness across the GDL, but the cross-section images indicate an increase in thickness with increased ionomer content (Figure f). For PiperION, the thickness increases from 4.3 to 6.1 μm for (i) 5% versus (ii) 30%, with the Nafion catalyst layers slightly thinner at 2.7 and 5.7 μm for (iii) 5% and (iv) 30%. There is also a visible difference in the pore structure, with Nafion providing a more uniform distribution. For the 30% PiperION sample, the high-resolution STEM lamella shows that the catalyst layer has significant inhomogeneity, with significant densification throughout the middle of the catalyst layer liftouts (Figure g). The overall porosity from the cross-sectional image was found to be 15%. STEM images at the GDL interface show that these pores are generally distributed along the interface and at the topmost layer of the lifted-out region (Figure S10). The uniform distribution of small Pt particles along the larger high-surface-area carbon support is also visible at high resolution. The 30% Nafion catalyst layer shows less densification with a slightly lower porosity of 9% and larger pores at the top and bottom interfaces (Figure h). This less dense structure is likely related to the larger fractal agglomerates in the ink, which have poorer packing density.

The preceding analysis has focused on the impacts of ionomer on the physical and mass transport properties of the catalyst layer. An additional effect of the ionomer may be chemical, impacting the surface coverage and binding energies of the reaction intermediates. To better understand this role of the ionomer in the HER kinetics, we conducted RDE testing and theoretical calculations. In RDE, we find HER activity is slightly higher using a Nafion binder compared to PiperION in the mass transport-limited region (Figure a). In the kinetic region, the calculated exchange current densities are 57.0 ± 0.6 A/g_Pt and 56.5 ± 0.2 A/g_Pt for Nafion and PiperION, respectively, indicating minimal effect of the ionomer on the mass-normalized kinetics (Figure b). The cyclic voltammograms from RDE show differences in electrochemically active surface area (ECSA), as assessed by double-layer capacitance and the hydrogen underpotential deposition (HUPD) peak at ∼0.25 V versus RHE (Figure c). Specifically, there is an increase in capacitance from 5.3 ± 0.2 to 6.5 ± 0.2 mF/cm² and an increase in ECSA from 29.7 ± 3.2 to 34.0 ± 1.8 m²/g_Pt between Nafion and PiperION, indicating a small increase in the surface area for both carbon and Pt with the PiperION ionomer. Thus, the ECSA-normalized activity is slightly higher for Nafion with an exchange current density of 1.92 ± 0.21 A/m² _Pt compared to 1.66 ± 0.09 A/m² _Pt for PiperION. While the differences are small, these results indicate that the ionomer impacts the fundamental electrochemical behavior of the Pt catalyst.

4.

RDE testing. (a) Polarization curves for Pt/C with Nafion (purple) and PiperION (black) ionomers. The center line and shaded region represent the average and standard deviation of four experiments, respectively. (b) A semilog plot of the average polarization curve with Butler–Volmer fits shown with dashed lines. (c) Cyclic voltammograms for Pt/C with Nafion and PiperION ionomers. Data shown was collected at a scan rate of 100 mV/s.

From the results of the ink analysis and electrochemical testing, it is clear that catalyst–ionomer interactions impact the catalyst layer formation and catalyst activity. To better understand the nature of these interactions and their impact on HER kinetics, theoretical calculations were performed. First, the theoretical surface model focused on interactions of the Pt (111) surface, the predominant facet found by X-ray diffraction (XRD, Figure S11), with the functional groups of Nafion (sulfur trioxide, SO₃) and PiperION (piperidinium, C₇H₁₆N). Pt (111) is an HER-active and well-studied facet. − Due to the complexity of the ionomer functional groups, an in-depth minima search of ca. 200 geometries was performed to determine the unique isomers summarized in Figure a. These geometries were theoretically assessed with implicit solvation but in the absence of OH^– and H₂, akin to the environment of the ink (Figure a).

5.

Theoretical calculations of the catalyst–ionomer interfaces in the ink and HER environment. (a) Catalyst-ionomer ink stability. In an alkaline environment, Nafion’s functional group is a deprotonated sulfur trioxide (SO₃), which can potentially bind to Pt (111) and poison 3 Pt-sites (Isomers I–II) or 1 Pt-site (Isomer III). PiperION showcases increased charge transfer of circa +0.9 e to the piperidinium functional group and stronger adsorption energies of circa −4 eV, suggesting that it can also potentially block Pt-site access. (b) HER reactivity of catalyst–ionomer interfaces. Isomers related to water adsorption and splitting via the Volmer (H₂O* → OH* + H*) mechanism; (c) Reaction profile for the Volmer mechanism on Pt (111) (gray), Pt (111) with Nafion (purple), and Pt (111) with PiperION (black). Bader charges (ΔQ), bond distances (Å), the adsorption energy of the functional group ( $E_{\mathrm{S}{\mathrm{O}}_3}$ , E _pip), and the adsorption energy relative to the global minimum ( ${\Delta E}_{\mathrm{S}{\mathrm{O}}_3}$ , ΔE _pip). Pt (light gray), adsorbate O (orange), S (yellow), C (dark gray), N (blue), and H atoms (white).

We note that our theoretical models that incorporate these functional groups onto the Pt (111) surface represent 18% Pt site coverage by Nafion’s SO₃ and 11% Pt site coverage by PiperION’s piperidinium. The SO₃ group of Nafion is found to bind to 3 Pt sites through a Pt–S bond and two Pt–O bonds, with a Bader charge of circa −0.7 e (Figure a). This is similar to the binding motif displayed by bisulfate (HSO₄) observed by in situ infrared spectroscopy, where 3 Pt sites are covered due to Pt–O bonds. In contrast, the piperidinium group has van der Waals interactions with the Pt surface, induced by a stronger Coulombic interaction between the negatively charged 0.9 e in Pt (111) and the positively charged 0.9 e in piperidinium (Figure a). This corresponds well to the zeta potentials, which were found to be opposite in sign due to the differences in charge on the functional groups (Figure b). Moreover, piperidinium-Pt isomers exhibit a lamellar, electrostatic attraction resulting in a more strongly bound but also fluxional interface where piperidinium can dock to Pt in multiple configurations (Isomers I–V). This may relate to the phenomenon observed in Figure c–d, where at 1% ionomer content, the PiperION seems to completely adsorb to the surface. At higher percentages, steric effects and the strong positive charge of the functional group (1⁺ e) may repel further ionomer adsorption, leading to greater amounts of free ionomer.

The strength of these interactions may impact the number of accessible active sites, and so theoretical calculations examined the enthalpy of the Volmer mechanism (H₂O* → OH* + H*, Figure b–c) and OH* versus H* adsorption (Figure ) in the presence of the functional group. Here, water adsorption and subsequent splitting into coadsorbed OH and H were examined explicitly (Figure b). In the alkaline environment of AEM electrolysis, the Volmer mechanism is often considered a possible rate-determining step for hydrogen evolution. , In an acidic environment, protons are plentiful for HER; in an alkaline environment, H must be supplied from either H₂O or OH and may also have to compete for metal sites with these other O _x H _y species. ,− Microkinetic analysis of the elementary oxygen reduction reaction on Pt suggests that the rate-limiting steps and the interaction of surface intermediates such as OH* and H₂O* can contribute to Tafel slopes. We note that DFT adsorption energies often directly correlate to temperature-programmed desorption studies, which estimate water desorption to require 0.50 to 0.55 eV on Pt (111), but functionals based on the generalized gradient approximation in DFT typically underestimate the adsorption energy of water. , Considerable discrepancies in adsorption and dissociation trends exist, depending on the dispersion correction applied. For the purpose of this study, the well-utilized GGA functional developed by Perdew–Burke–Ernzerhof (PBE) was implemented. In order to explicate in part the complexity of the catalyst–ionomer interactions on the hydrogen evolution reaction, theory examined a possible rate-determining step (Volmer step for water splitting; Figure b–c), the influence of H* binding on other HER mechanisms (Volmer, Heyrovsky, Tafel, Table ), and competition between H* and OH* (Figure ). These mechanistic factors may contribute in part to the HER kinetics and performance (overpotentials) observed in experiment.

6.

Theoretical calculations of Pt (111) surface coverages. (a) Pt (111) surface with Nafion and H*. Nafion’s SO₃ weakens H* binding, and the functional group easily desorbs from the surface with H* able to occupy many different sites. (c) Pt (111) surface with Nafion and OH*. Local minima of OH* isomers, including intact OH* hydrogen-bonded to SO₃ O atoms, split to form O* + H*, and desorbed, weakly attached to the functional group. PiperION’s piperidinium binds (b) H* and (d) OH* with comparable strength. Bader charges (ΔQ), bond distances (Å), adsorption energy of the functional group ( $E_{\mathrm{S}{\mathrm{O}}_3}$ , E _pip), and adsorption energy relative to the global minimum ( $\Delta E_{{\mathrm{SO}}_3}$ , ΔE _pip). Pt (light gray), adsorbate O (orange), S (yellow), C (dark gray), N (blue), and H atoms (white).

1. Reaction Enthalpies (ΔH _rxn) of the Electrolyte-Mediated Volmer, Heyrovsky, and Tafel Mechanisms with and without Ionomer .

Volmer, ΔH rxn (eV)				Heyrovsky, ΔH rxn (eV)	Tafel, ΔH rxn (eV)
Surface	Site Geometry	E H* (eV)	E surf + $E_{{\mathrm{H}}_2\mathrm{O}}$ → E H* + E OH	E H* + $E_{{\mathrm{H}}_2\mathrm{O}}$ → E surf + E OH + $E_{{\mathrm{H}}_2}$	E H* →E surf + $1/2E_{{\mathrm{H}}_2}$
Pt (111)	Pt–H*	–2.69	3.01	3.82	0.40
Pt (111) with Nafion’s SO3	Pt–H*	–1.69	4.01	2.82	–0.59
Pt (111) with PiperION’s Piperidinium	Pt–H*	–2.24	3.01	3.81	0.40

^a E _surf for HER mechanisms in the presence of a functional group includes the reference energy of a surface with the functional group already attached.

Despite differences in interaction type and strength between the ionomer and Pt, the enthalpy of the rate-determining Volmer step of water dissociation to OH and H on the Pt surface is the same for both ionomers, 1.07 eV in the presence of Nafion’s SO₃ and 1.01 eV in the presence of PiperION’s piperidinium (Figure b–c). These enthalpies are slightly lower compared with an enthalpy of 1.12 eV of the Volmer mechanism on Pt (111) without the ionomer. Water adsorbs more strongly in the presence of Nafion’s SO₃ due to advantageous hydrogen bonds and Coulombic attraction between SO₃ (ΔQ = −0.7 e) and H₂O (ΔQ = +0.2 e). In contrast, water adsorbs more weakly in the presence of PiperION’s piperidinium functional group due to the Coulombic repulsion between the positively charged functional group (ΔQ = +0.9 e) and the slightly positive H₂O (ΔQ = +0.2 e). Water attraction may become key to the Volmer step, since the enthalpy remains comparable for both Nafion and PiperION. Water desorption can also occur at ca. 0.4 eV higher in energy (Isomer V with PtNafion, Isomer VIII for PtPiperION). For PiperION, following water splitting, H* will preferentially bind further away from the functional group due to the repulsion between the positively charged piperidinium and the slightly positive hydrogen, leading to a difference of 0.2 eV between the reaction enthalpies of Iso. I → Iso. XII (ΔH = 1.01 eV) and Iso. I → Iso. XVII (ΔH = 1.21 eV). In contrast, H* can occupy multiple degenerate sites near SO₃ (Iso. VIIso. VIII).

Furthermore, theoretical calculations considered the impact of H* binding for all three mechanisms of Volmer (H₂O → H* + OH), Heyrovsky (H* + H₂O → OH + H₂), and Tafel (H* → 1/2 H₂). These electrolyte-mediated mechanisms specified in Table rely upon gaseous and liquid species such as hydroxide ions, water, and diatomic hydrogen to interact with a single H* bound to the surface to subsequently proceed to the HER. The presence of SO₃ considerably weakens H* binding, leading to an advantageous exothermic reaction for the Tafel mechanism (ΔH = −0.49 eV) and a considerably lowered enthalpy for the Heyrovsky mechanism (ΔH = 2.82 eV). This may explain in part the improvement in HER activity of a Nafion binder as compared to PiperION, where experiment observed improved kinetics in RDE tests (Figure ). It may also relate to active site access, shifting the thermodynamics of the reaction toward desorbed products and freeing the catalyst–ionomer interface’s Pt sites.

Another consideration for HER in an alkaline environment is the competition for Pt sites between H* and OH*. We contrast the binding trends of H* versus OH* on the Pt-ionomer systems in Figures and S12–S13. Both Nafion and PiperION’s functional groups weaken H* adsorption compared to H* on Pt (111) alone. Nafion’s influence is more advantageous than PiperION’s as summarized in Table and visually displayed in Figure a–b. Although H* adsorption is weakened on Nafion, it can also lead to the breaking of Pt–S (Figure a, Isomer I) and Pt–O (Figure a, Isomer II–VI) bonds, resulting in the desorption of the functional group. This trend occurs for coadsorption with OH*, where Pt–O bonds (Figure c, Isomers V–VII) or the Pt–S bond (Figure c, Isomer VIII–IX) can be broken. Moreover, theoretical calculations demonstrate that the splitting of OH is a possible, although higher energy (∼1.5 eV), pathway to provide H* for HER (Figure c, Isomer I → Isomer IX; Figure d, Isomer I → Isomer XV). Characterization of these adsorption trends for Pt-Nafion suggests that there are equilibrium amounts of free and adsorbed ionomer, which may be related to how easily Nafion’s SO₃ desorbs from the surface in the presence of other solvent species. In contrast, PiperION’s piperidinium remains rigid in its attraction to the surface with H* and OH* preferentially adsorbed 3–5 Å away; isomers adsorbed ∼2 Å are often higher in energy by 0.2–0.5 eV (Figure b, Isomers VIII–X; Figure d, Isomer XIV). We note that all HER species (H*, OH*, H₂O*) prefer to bind to Pt sites rather than the functional group (∼2 eV higher), suggesting that HER will be the preferred reaction as compared to other catalyst-ionomer combinations observed for OER. In an alkaline environment, Pt sites for the Pt–Nafion interface can be poisoned by OH due to the stronger binding energy of OH* (−2.26 eV) over that of H* (−1.69 eV). In contrast, H* and OH* have nearly degenerate binding energies of −2.2 eV for the Pt–PiperION interface. This may be advantageous for Pt-site access, where electrolysis favors the transport of OH to the anode. Overall, we find that the ionomer functional groups interact differently with the catalyst surface, contributing to differences in H* coverage and reaction energetics that lead to improved kinetics with Nafion compared to PiperION.

While the ionomer has a smaller effect on activity than Pt loading, we have found that it has significant impacts on the catalyst layer morphology, mass transport, and kinetics. Both theoretical calculations and RDE measurements indicate that there is a small improvement in HER kinetics for Nafion compared to PiperION. Theoretical calculations highlighted an exothermic reaction for the Tafel mechanism in the presence of Nafion. Moreover, the Pt-piperidinium surface features nearly degenerate H* and OH* adsorption, allowing for more favorable Pt–H* adsorption for subsequent access to different HER mechanisms compared with Pt (111) alone. For both ionomers, we find a trade-off for ionomer content, where higher ionomer content improves kinetics, lower ionomer content improves mass transport, and lower free ionomer content is necessary for catalyst layer uniformity. Although Nafion does improve the kinetics in MEA testing, this is outweighed by the higher mass transport losses and degradation rate. Overall, lowering the PiperION content to 5% is found to optimize the kinetics-mass transport trade-off, leading to improved activity and durability.

Fabrication Methods

For improved scalability of AEMWE technology, it is desirable to develop more consistent and higher-throughput methods for MEA fabrication, particularly those that utilize the CCM geometry. The CCS approach has been more common in AEMWE because of challenges with membrane stability and adhesion in CCM fabrication, , but CCMs have potential advantages, such as improved interfacial contact with the membrane and improved coverage at low loadings. For PEM fuel cells and electrolyzers, ultrasonic spraying and rod coating methods have been developed for depositing Pt/C catalyst layers using a Nafion ionomer and membrane. , In this study, we adapted three fabrication techniques using PiperION as the ionomer and membrane. For a comprehensive comparison of electrochemical performance and catalyst layer morphology for these fabrication techniques, we translated the standard ink formulation used for hand spraying a cathodic CCS to obtain a loading value of 0.2 mg/cm² with 30% PiperION content for the CCM cathodes. While 5% PiperION was found to give the highest performance for the CCS electrodes, it is hypothesized that higher ionomer contents may be beneficial for adhesion to the membrane in the CCM configuration. Figure shows an illustrative schematic of the fabrication techniques we adapted to demonstrate the differences in the deposition of the catalyst ink onto the substrate.

Hand spraying is a versatile method, allowing for the use of small quantities of catalyst and varying ink compositions, but it is difficult to produce reproducible and uniform catalyst layers. Ultrasonic spraying can address some of these limitations. Ultrasonication at a specific frequency allows for improved dispersion of the ink, and automation of the spray rate and pattern allows for improved loading control, uniformity, and reproducibility. This is a more resource-intensive technique, however, requiring large volumes of catalyst and more time for fabrication as well as having limits in catalyst size and properties. The hand spraying and ultrasonic spraying methods deposit the ink in a repetitive crosshatch pattern, as illustrated in Figure a–b. This creates a multilayered electrode that may have limitations in terms of catalyst layer adhesion to the GDL or membrane. Roll-to-roll fabrication is an alternative technique that has been implemented for upscaling electrode fabrication for PEM fuel cells and electrolyzers. In the rod coating technique, a Mayer rod with grooves of defined spacing is used to glide the catalyst ink across the membrane to form a single-layer electrode. The dried catalyst layer surface morphology can be optimized by varying the diameter of the grooves of the Mayer rod, the pressure exerted over the rod while gliding, rod speed, and the viscosity of the catalyst ink. To achieve the higher viscosity required for rod coating, the catalyst concentration was increased in the ink. Rheology measurements demonstrate that these changes successfully increased the viscosity for this ink compared to the spray inks, and all of the inks exhibited the expected Newtonian behavior (Figure S14). Top-down SEM images of the CCMs (Figure d–e) show more uniform coverage than the CCS electrodes at a comparable loading (Figure b). The ultrasonic CCM shows a few cracks as well as a region of segregated Pt and ionomer (Figure d). In contrast, the rod-coated CCM has more cracks but improved uniformity of the catalyst and ionomer (Figure e). Differences in crack size and connectivity may arise from variations in drying stresses as a function of the fabrication method. Specifically, crack behavior could be linked to three factors: wet film thickness, drying gradients, and solvent retention. − With spraying, the low wet film thickness of each pass minimizes drying gradients and reduces solvent retention. As a result, less defect formation and propagation are observed. On the other hand, the high wet film thickness of the rod coating prolongs solvent removal, creating nonuniform drying gradients across the catalyst layer. Under these conditions, additional solvent remains trapped within the compact particle network and capillary forces intensify during evaporation. The combined influence of such effects manifests as larger, interconnected cracks.

Next, we compare the AEMWE performance of these CCM cathodes to that of standard CCS fabrication. As shown in Figure a, the hand-sprayed CCS and CCM cathodes perform similarly, with CCM performance in the order of hand spray > rod coat > ultrasonic spray. The performance differences are relatively small, with the ultrasonic spray reaching 1.92 ± 0.14 A/cm² at 2 V compared to 2.28 ± 0.07 A/cm² for the hand-sprayed CCM (Table S4). The EIS at 1.9 V shows that the fabrication method and configuration have minimal impact on the HFR, but the hand-sprayed CCS and CCM samples show smaller charge transfer resistances (Figure b). Accordingly, the VBA summary in Figure c shows minimal difference in ohmic losses at 1 A/cm². The only significant change is higher transport losses for the ultrasonic spray and rod-coated samples as seen in Figure d; based on the previous loading and ink studies, this may relate to a decrease in coverage or an increase in particle size. Furthermore, this could indicate that there is a benefit to the intermittent drying and multilayer catalyst layer formation of the hand-spraying method. The inhomogeneities of hand spraying from having less control over ink atomization and application speed may actually improve catalyst layer morphology due to improved packing with heterogeneous particle sizes. The same trends are observed for CCMs fabricated at 0.05 mg/cm² and 1 mg/cm² loadings, with the CCS cathode significantly outperforming the rod-coated and ultrasonic-sprayed CCMs, respectively (Figures S15–S16). A potential cause of the decreased performance for the ultrasonic and rod-coated CCMs is the presence of cracks in the catalyst layer, which have been shown to impede mass transport and exacerbate instability in fuel cell studies.

8.

AEMWE performance for cathodes made by CCS (black) and CCM methodshand spray (pink), ultrasonic spray (orange), and rod coated (gold). (a) Polarization curves, (b) EIS Nyquist plots at 1.9 V, (c) summary of ohmic, kinetic, CLR, and residual/mass transport losses at 1 A/cm², and (d) mass transport losses. Polarization curves and voltage losses are reported as the average and standard deviation of tests performed in triplicate. (e) 100 h CP durability measurement at 1 A/cm². Data for CCS hand spray (black) is reproduced from Figure .

Figure e shows the durability profiles for the hand-sprayed CCS, hand-sprayed CCM, and ultrasonic-sprayed CCM over 100 h at 1 A/cm². Interestingly, the CCM cathodes have slightly lower overall degradation rates than the CCS for the same loading and ionomer content. At the end of the 100 h measurement, the hand-sprayed CCS and CCM have identical performances, while the ultrasonic CCM continues to have a higher voltage. The initial voltage change in the first few hours of the test is also different for the CCM, showing more gradual degradation compared to the CCS, which experiences most of the degradation on the first day. This suggests that there may be a different degradation mechanism for different catalyst layer architectures and interfaces. Overall, these results show that these CCM fabrication methods can be successfully employed with the PiperION membrane, although no significant advantage is found compared to the CCS approach, perhaps due to the small role of the ionomer in ionic conduction for tests in supporting electrolyte. Additional optimization of these fabrication methods may be required to exceed the CCS performance.

Degradation Mechanisms

Although the anode and membrane, among other components, contribute to performance changes over time, we have found that the overall cell degradation rate varies significantly with isolated changes to the cathode catalyst layer (Figures , , and ). To understand cathode degradation, Pt loss and catalyst layer structure changes were analyzed using ICP-MS and microscopy. Pt losses may occur through dissolution, which is likely to be driven by chemical and/or electrochemical processes, or through catalyst layer delamination due to shear stresses and gas evolution, which will be affected by the mechanical integrity of the catalyst layer. Overall, Pt loss, as measured from the recirculating electrolyte at the end of testing, is less than 10% but non-negligible for all tests (Figure a–c). The percentage of dissolved Pt decreases as the initial loading increases, from 6.5% at 0.05 mg/cm² down to 2.2% at 1 mg/cm² (Figure a). Dissolution also decreases slightly with increased ionomer content (Figure b), while no significant differences are observed as a function of the deposition method (Figure c). For the cathodes at different loadings, the ratio between the voltage increase and mass loss rates shows a much larger impact of dissolution at low starting loadings. For the cathodes with the same loading but varying ionomer or fabrication methods, the voltage gain/mass loss ratios are similar, suggesting a shared degradation mechanism and highlighting loading as the most significant factor in degradation. While the dissolution is small, it is indicative of possible redox reactions and catalyst layer restructuring, either during the chronopotentiometry hold or as part of the observed performance recovery upon exposure to low or open circuit cell voltages.

9.

Dissolved Pt mass percentage relative to initial catalyst loading as a function of (a) Pt loading (Figure ), (b) ionomer content and type (Figure ), and (c) the catalyst layer deposition method (Figure ). Error bars represent dilution and measurement of each sample in triplicate. (d) On-line ICP-MS flow-cell measurement showing (top) Pt dissolution as a function of time and (bottom) applied voltage and measured current density as a function of time in 0.1 M KOH at a scan rate of 10 mV/s. Cross-section STEM-EDS maps of (e) 30% PiperION and (f) 30% Nafion samples (i) before and (ii) after durability tests.

To better understand the dissolution behavior, an on-line ICP-MS measurement was conducted using a three-electrode flow cell. The potential window for these measurements, −0.6 to 1.0 V versus RHE, was chosen based on cathode potentials measured via reference electrode experiments in AEMWE ,− and open-circuit measurements in PEMWE. In cyclic voltammogram cycles at 10 mV/s, the majority of dissolution was found to occur at the upper and lower potential bounds (Figure d). Here we observe Pt dissolution, likely due to the transition between metallic Pt and PtO₂, between 0.75 and 1 V versus RHE, as has been observed previously. − However, there is an unexpected, larger dissolution feature at low, HER-relevant potentials, particularly below −0.5 V versus RHE. At these potentials, increased levels of continuous metal dissolution and full particle detachment, as indicated by the spikes in the signal, − are observed, particularly in the downward sweep. During the 3 cycles shown in Figure d, 45 nanoparticle events (with an estimated size of 1.3 ± 0.6 nm) were measured (Figure S17). This is comparable to the reported 2.4 ± 0.5 nm particle diameter for this catalyst, indicating that the spikes correspond to individual particle detachments; this is analogous to catalyst layer delamination in the MEA. While the precise amount and timing of dissolution are specific to this flow cell experiment, which is distinct from MEA testing in catalyst layer structure, substrate, applied potential, and reaction rate, this result indicates that Pt dissolution and detachment occur simultaneously with HER. Since Pt is expected to be thermodynamically stable in this pH and potential region, it may be assumed that the binding of HER intermediates and participation in HER is destabilizing. We hypothesize that Pt dissolution is thus an unavoidable degradation mechanism in AEMWE, even if the cathode is prevented from reaching oxidizing potentials.

To better understand the impact of Pt dissolution and mobility on the catalyst layer structure, top-down and cross-section microscopy analysis was conducted. It is important to note that some changes may have occurred during MEA disassembly and therefore would be unrelated to degradation. For the standard cathode, a significant portion of the catalyst layer was transferred to the membrane (Figure S18). Comparing top-down images of the membrane and GDL, a pattern related to the flow field geometry is observed with more Pt transferring from the GDL to the membrane in the “land” regions of the flow field and more Pt remaining on the GDL in the channels, indicating that this is likely a mechanical transfer process due to compression. In the channel regions, the coverage appears less complete than the untested electrode, with more carbon fibers visible, but both the catalyst and ionomer remain well-distributed. At the back side of the GDL, which was in direct contact with the flow field, there are more broken carbon fibers, and clusters of Ni and Fe are visible in the channel region. Since ICP-MS showed an increase in Ni and Fe concentrations in the electrolyte after testing and the electrolyte is recirculated through connected reservoirs, this finding is interpreted as evidence of catalyst dissolution/delamination at the anode and redeposition at the cathode.

Finally, comparison of pre- and post-test cross sections for the 30% PiperION and Nafion cathodes helps to explain differences in degradation rate. STEM analysis shows significant decreases in the I/Pt and F/Pt ratios from 1.4 to 0.4 and 0.4 to 0.2, respectively, for the PiperION cathode, indicating a loss of both the backbone and functional groups of the ionomer (Figure e). There is also a decrease in porosity to 4%, suggesting that the catalyst layer densifies during testing. This may be the cause of the observed increase in the mass transport losses after testing (Figure S6). In contrast, the Nafion catalyst layer shows a significant increase in porosity from 9% to 25% and a collapse of the catalyst layer structure with ionomer loss and an increase in oxidation (Figure f). The Nafion catalyst layer therefore shows more substantial morphological changes after testing, in agreement with the higher voltage increase rate and Pt dissolution. While Nafion provides some advantages for kinetics, the accelerated degradation may be a limitation for the use of Nafion in Pt cathodes for AEMWE.

Conclusions

In this work, we demonstrated the significant impact of cathode catalyst layer design on AEMWE performance and durability using a commercial Pt/C catalyst. Pt loading has the largest effect on performance, with a 100 mV decrease in overpotential at 1 A/cm² and a 4× decrease in degradation rate as Pt loading increases from 0.05 to 1 mg/cm². While performance was comparable for any loading between 0.2 and 1 mg/cm,² both kinetic and mass transport losses were severe at the DOE target loading of 0.05 mg/cm². The low areal coverage of the GDL at low loadings, leading to high transport losses, suggests that Pt-alloy materials may present an avenue to maintain performance while lowering PGM loading.

Analysis of catalyst ink properties demonstrated that the content and identity of the ionomer significantly impacted the agglomerate size, zeta potential, and fraction of adsorbed versus free ionomer in the ink. The catalyst-ionomer adsorption behavior varied between the two ionomers, with Nafion partitioning between free and adsorbed, whereas PiperION initially adsorbed and then the excess was free. Theoretical calculations further show that the strength of the interactions between Pt and the ionomer functional group affects the HER mechanism and accessible surface area, where Nafion’s SO₃ weakens H* binding and may favor an exothermic, Tafel mechanism and PiperION lowers the enthalpy of water-splitting by 0.1 eV compared to Pt alone. In an alkaline environment, Pt sites on the Pt–Nafion interface may more easily be poisoned by OH* due to its stronger binding than H*; whereas Pt-site access for the Pt–PiperION interface is equal for H* and OH* at circa −2.2 eV. A trade-off with ionomer quantity was identified, where catalyst-ionomer adsorption improves kinetics, while increased ionomer content increases particle size and mass transport losses. The high activity of the 5% PiperION cathode was attributed to a trade-off between these losses, with the fraction of free ionomer as an ionomer type-agnostic predictor of catalyst layer uniformity and performance. Comparison of catalyst layer fabrication methods showed little to no performance advantage to CCM approaches, either spraying or rod coating, indicating that ionic transport is not limiting for these cathodes. Evaluation of CCS versus CCM performance in DI water or lower supporting electrolyte concentration would provide more insight into the impact of the fabrication method on catalyst–ionomer interactions, interfacial resistances, and degradation. The cathode configuration further affected cell durability, with higher Pt loading, PiperION ionomer, and CCM fabrication, improving overall stability. Pt dissolution and delamination at high HER currents, as well as ionomer degradation and loss of catalyst layer structure, were identified as key degradation mechanisms.

Overall, this study reveals the important role of the cathode in AEMWE and the relationship among the ink, catalyst layer, and performance. In alkaline conditions, the slower HER kinetics and variety of reaction mechanisms make the cathode critical to both performance and stability. For PGM-free HER catalysts, there will be additional challenges due to their low intrinsic activity, larger nanoparticle sizes, and less optimized carbon supports. While higher catalyst loadings can still be economical, the thicker catalyst layers may pose further challenges with catalyst site accessibility and mass transport. The ideal ink and catalyst layer design is likely to vary for different catalysts and electrochemical systems, but this work identifies critical cathode variables that should be explored. Ultimately, cathode design choices, whether Pt or PGM-free, will need to be decided by the trade-off between cost, performance, and durability.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.5c20325.

• Additional electrolyzer performance, microscopy, theory, and ink characterization data (PDF)

The authors declare no competing financial interest.