Durable Thin‐Film Porous Transport Electrodes for High Current Density PEM Water Electrolysis

Abstract

Proton exchange membrane water electrolyzers rely on relatively expensive Ir‐based catalysts for efficient and durable hydrogen production. To reduce system costs, Ir loadings can be reduced if performance and durability are maintained. Sputter deposition is a readily scalable method to synthesize uniform, low‐loading catalyst layers with controlled composition. A catalyst applied directly to the porous transport layer can have advantages for performance, manufacturing simplicity, and catalyst recovery. Suitable porous transport layer porosity can minimize activity losses when reducing loadings. Here, methods are presented to deposit metallic Ir as well as amorphous and rutile Ir oxides. The activity and durability of these materials in the porous transport electrode architecture is evaluated. The metallic and amorphous forms have better initial activity, however, operation at 3 A cm⁻² with 0.1 mg Ir cm⁻² shows that only rutile IrO₂ maintains performance beyond 100 h with a 50 mV improvement after 700 h. A >10x reduced dissolution rate is shown for rutile IrO₂. With a low‐porosity transport layer and 0.4 mg Ir cm⁻², a steady‐state voltage decay rate of 6 µV h⁻¹ is achieved. The results demonstrate that sputter‐deposited rutile IrO₂ porous transport electrodes with low Ir loading can be operated at high current density to reduce hydrogen production costs.

Water electrolysis using porous transport electrodes with sputter‐deposited, ionomer‐free thin films of rutile IrO₂ catalyst suppresses Ir dissolution by >10x over other forms of IrO_x. The rutile IrO₂ catalyst prepared by this readily scalable electrode synthesis method provides stable cell operation at 3 A cm⁻² while using low Ir loading (0.4 mg Ir cm⁻²) and achieves a voltage decay rate of 6 µV h⁻¹.

1. Introduction

Hydrogen can be produced using a variety of energy sources, and from the energy carrier electricity, by using an electrolyzer that splits water molecules into hydrogen and oxygen gas.^{[ 1 ]} The hydrogen can be stored and used for various purposes: as a fuel to create electricity in a fuel cell with water as the only by‐product, or as a critical chemical input for hydrocarbon fuel refining and heat‐intensive industries such as steel and ammonia production.^{[ 2 ]} Consumption of hydrogen fuel is expected to increase to 150 MMT by 2030,^{[ 3 ]} and demand for hydrogen fuel is expected to increase anywhere from 73 to 568 MMT by 2050,^{[ 4 ]} making it an important part of the United States’ energy portfolio.

Various water electrolysis technologies exist that either use an alkaline solution as an ion‐conducting medium and a porous separator for gas separation, or that utilize alkaline or proton exchange membranes for this purpose. Proton exchange membrane water electrolysis (PEMWE) offers various beneficial features. It can operate at high current densities greater than 2 A cm⁻², has high conversion efficiencies of 80%–90% at temperatures of ≤80 °C, produces hydrogen of greater than 99% purity, and has promising durability.^{[ 5 ]} PEMWE can further operate at differential pressure, produce pressurized hydrogen, and run in compact stack configurations. To scale up hydrogen production via PEMWE, capital and operating costs must be reduced. This can be accomplished through: i) operation at high current density, which can proportionally reduce the cost of materials (e.g., catalyst, membrane, bipolar plates, diffusion media) with regards to the amount of hydrogen produced and ii) reduction of the Ir‐based oxygen evolution reaction (OER) catalyst^{[ 6} , ⁷ , ^{8 ]} which accounts for 17% of current stack cost.^{[ 9 ]} However, reducing the amount of iridium in an electrolyzer cell without compromising activity and durability is challenging. Researchers have shown that reducing the electrode loading may lead to reduced catalyst layer (CL) homogeneity, durability, conductivity, and overall performance.^{[ 10} , ^{11 ]}

Generally, the core of a PEMWE cell consists of five layers. From anode to cathode side these are: a titanium‐based anode porous transport layer (PTL), an iridium‐based anode catalyst layer, a proton conductive membrane, a platinum‐based cathode catalyst layer, and a carbon fiber‐based gas diffusion layer (GDL) that is often also called a PTL. These layers can be integrated into a membrane electrode assembly (MEA) using two different strategies: the porous transport electrode (PTE) or the catalyst‐coated membrane (CCM). The PTE differs from the traditionally used CCM in that the catalyst is coated onto the PTL, rather than the membrane.^{[ 12} , ^{13 ]} PTEs can improve catalyst utilization even at reduced catalyst loading^{[ 13} , ^{14 ]} by providing a direct electronic interface between the catalyst and PTL. In addition, manufacturing costs can be lowered through simplified coating processes, reduced precursor complexity, and avoiding ink formulation challenges.^{[ 15 ]} PTEs with a thin‐film, ionomer‐free approach^{[ 16 ]} can further allow for a wider variety of membrane options,^{[ 15 ]} have reduced catalyst layer deformation due to membrane swelling,^{[ 12} , ¹³ , ^{15 ]} and can facilitate recovery and recyclability of the Ir catalyst.^{[ 13 ]} It is challenging to successfully deposit the catalyst layer on the PTL, whether this is accomplished by coating a catalyst‐containing ink or via vapor deposition techniques. For ink‐based approaches, PTL properties, such as porosity and pore size, are as important as ink viscosity and solid content.^{[ 17 ]} Catalyst ink may settle into the pores of the PTL, inhibiting mass transport or resulting in non‐continuous catalyst layers, especially at low loadings.^{[ 18 ]} Physical vapor deposition approaches can provide low‐loading catalyst layers that are highly uniform and have good electrical contact to the PTL substrate. However, catalyst material that is deposited too far into the pores may not be able to contribute to the electrochemical reaction. Nonetheless, the presence of water has been shown by Lee et al.^{[ 19 ]} to provide sufficient conductivity for ionomer‐free PTEs to achieve current densities of 10 A cm⁻² without significant transport limitations.

Sputter deposition is a form of physical vapor deposition. It is a scalable method that has been used for semiconductors, medical devices, and high‐throughput applications such as thin film solar cells.^{[ 20 ]} During deposition, atoms of a target material are ejected by ions formed and energized in a plasma adjacent to the target material. When an inert process gas is used (e.g., Ar) to form the plasma, the ejected atoms are deposited onto a substrate with the same composition as the target, while the deposited film properties, such as porosity and texture, can be tuned by the deposition parameters. A reactive gas may be added (e.g., O₂) to form compound species, such as metal oxides, in the deposited films. Sputter deposition has been used in electrochemical applications to create catalyst layers with finely tuned catalyst thickness, composition,^{[ 21} , ^{22 ]} and morphology in order to achieve high performance with low loadings.^{[ 23} , ²⁴ , ^{25 ]} While very promising, the sputter‐deposited PTE has a membrane/catalyst interface that is dictated by, and in some cases limited to, the PTL surface because common sputter deposition approaches, such as DC magnetron sputtering, yield thin films with limited internal porosity. Although not a strict limitation toward achieving performance comparable to that of CCMs,^{[ 19} , ^{26 ]} sputter‐deposited PTEs may have reduced triple‐phase boundary regions, which causes higher local kinetic overpotentials and may increase catalyst dissolution.

Previous studies on sputter‐deposited iridium and iridium oxide catalysts have demonstrated the feasibility of using sputter deposition techniques for PEMWE applications. The studies focused either on ex‐situ half‐cell studies reporting initial kinetic performance at low current density,^{[ 6} , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ^{31 ]} on in‐situ studies of CCM architectures, or on the metallic form of Ir and Ir‐based catalyst,^{[ 26} , ³² , ^{33 ]} which is significantly less stable in PEMWE than IrO_x materials. Sputter‐deposited metallic Ir PTEs were studied by Lee et al.,^{[ 13} , ^{19 ]} reporting on performance results and durability testing at a maximum of 1.8 A cm⁻². PTEs with electrodeposited IrO_x were developed by Ding et al.^{[ 34 ]} and durability tested for 80 h at 1.8 A cm⁻². In other work, Geuß et al.^{[ 35 ]} introduced a PTE half‐cell to investigate catalyst dissolution. The study compared the dissolution of different IrO_x forms in ink‐based, particulate catalyst layers, and their results indicated that rutile IrO₂ materials are two orders of magnitude more stable than amorphous IrO_x materials, when tested at a maximum of 1 A cm⁻². At these conditions, a 10‐year lifetime was projected to be possible for rutile IrO₂ at loadings of just 0.1 mg Ir cm⁻².

The present investigation focuses on sputter‐deposited, thin‐film IrO_x catalysts directly deposited on PTLs to form PTEs. It reports on the durability and dissolution behavior of: i) metallic iridium (m‐Ir), ii) amorphous iridium oxide (a‐IrO _x ), and iii) crystalline rutile iridium oxide (r‐IrO₂). Each catalyst was evaluated by physical characterization, half‐cell dissolution studies, and in‐situ experiments, i.e., in an MEA configuration at Ir loadings of 0.1 and 0.4 mg cm⁻² and at a current density of 3 A cm⁻² over 700 h of operation. Post‐durability microscopy and ex‐situ dissolution studies are used to investigate the stability and Ir dissolution behavior for each PTE type. The r‐IrO₂ is further durability tested using two PTL types to assess the influence of PTL properties on steady‐state PTE performance. Specifically, results are presented from synthesis and physical characterization (Subsection 2.1), beginning of life (BOL) performance (Subsection 2.2), in‐situ PTE durability testing (Subsection 2.3), half‐cell electrochemical dissolution (Subsection 2.4), post‐durability microscopy (Subsection 2.5), and a comparison of PTE durability (Subsection 2.6). Subsequently, the results will be discussed in Subsection 2.7.

2. Results and Discussion

2.1. Synthesis and Physical Characterization

The physical properties of sputter‐deposited m‐Ir, a‐IrO_x, and r‐IrO₂ thin films were characterized for crystal structure, composition, coverage uniformity, and morphology using X‐ray diffraction (XRD), Rutherford backscattering (RBS), X‐ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM) with energy dispersive X‐ray spectroscopy (EDS), and high‐resolution scanning transmission electron microscopy (STEM). Results from each technique are reported in the Subsections 2.1.1–2.1.4 below.

2.1.1. Synthesis of m‐Ir, a‐IrO _x , and r‐IrO₂

While m‐Ir thin films can be readily synthesized using standard sputter deposition conditions under an inert process gas (e.g., Ar), multiple pathways are available to form a‐IrO _x and r‐IrO₂. Common among these latter approaches is the need to incorporate oxygen throughout the internal volume of the thin film. We refer to this internal volume as the “bulk” to distinguish it from the surface, or outermost layer, of the thin film. Synthesizing r‐IrO₂ may further require additional energy during or after deposition to drive the formation of the rutile crystal structure. As described in the Experimental section, the a‐IrO _x in this work was synthesized by reactive sputtering of Ir with oxygen, using a relatively low deposition energy to ensure a pure amorphous phase. To controllably stimulate r‐IrO₂ formation, post‐deposition anneals were conducted on m‐Ir and a‐IrO _x thin films. During synthesis development, glass substrates were employed to avoid potential interference with substrate diffraction peaks and to identify the minimum temperature and time for transformation of m‐Ir and a‐IrO _x into r‐IrO₂. During this process, the m‐Ir must undergo oxidation and diffusion of oxygen into the bulk for full layer conversion to r‐IrO₂. Identifying and using the minimum thermal budget for each pathway can avoid oxidation of the desired target substrate, i.e., the Ti PTL surface, so that a low electrical interface resistance can be maintained for efficient electrolyzer cell operation.

2.1.2. X‐Ray Diffraction

During the annealing process (that was applied to convert m‐Ir and a‐IrO _x into r‐IrO₂), in‐situ XRD was conducted throughout the applied temperature range to study the progressive conversion to r‐IrO_2. The results are shown in Figure 1a,b for starting materials a‐IrO_x and m‐Ir, respectively. The as‐deposited a‐IrO _x shows no diffraction peaks, as expected for fully amorphous material, up to at least 400 °C (Figure 1a, red). Upon ramping, the onset of crystallization was first observed at around 470 °C–480 °C (Figure S1, Supporting Information). The ramp was stopped upon reaching 500 °C (Figure 1a, orange) and held constant. Within 30 min at 500 °C (Figure 1a, green), the rutile crystal peaks at 28°, 35°, 40°, and 54° were fully present and peak growth had saturated, indicating that the phase transition to r‐IrO₂ had equilibrated under these conditions. Figure S1 (Supporting Information) shows a detailed set of XRD spectra and a plot of the diffraction peak full width at half maximum (FWHM) evolution throughout the anneal. While the data suggest that the initial transformation of a‐IrO _x to r‐IrO₂ requires <10 min at 500 °C, an anneal time of 1 h was used throughout this work to ensure thoroughly crystallized r‐IrO₂ layers. The formation of r‐IrO₂ on PTL substrate, rather than glass, was characterized by ex‐situ XRD. The results, which are shown in Figure S2 (Supporting Information), confirmed that the chosen r‐IrO₂ synthesis approach translated well to fabricate the r‐IrO₂‐based PTE material.

Figure 1.

In‐situ XRD during anneal at a temperature ramp rate of 5 °C min⁻¹ for a) a‐IrO_x and b) m‐Ir on glass substrates. The XRD measurement sequence proceeds from bottom to top with the anneal temperature/time (vertical arrow labeled “T/t”). Upon reaching 500 °C (orange), the temperature was held constant. For m‐Ir b), the anneal was continued for 18 h at 500 °C (blue). Diffraction peaks for r‐IrO₂ (28°, 35°, 40°, 54°) and m‐Ir (41°, 47°) are indicated at the top of each plot.

To evaluate the feasibility of m‐Ir to be used as the starting material for an alternative pathway to forming r‐IrO₂, the as‐deposited m‐Ir was subjected to the same anneal sequence and in‐situ XRD measurement. The results are shown in Figure 1b. For temperatures below 500 °C, peaks that correspond to metallic Ir are observed at 41° and 47° (Figure 1b, red). Upon first reaching 500 °C (orange), no additional peaks are observed initially, but holding the temperature at 500 °C for 30 min resulted in the development of a small rutile peak at 35° (green). Holding the temperature at 500 °C (blue) even longer, i.e., for 18 h, resulted in the growth of the first rutile peak at 35° and the development of a second one at 28°. It is interesting to note that the m‐Ir peak at 41° is not significantly diminished due to the annealing process. This indicates that attempting to form r‐IrO₂ from m‐Ir thin films can yield incomplete conversion of m‐Ir to r‐IrO_2, and a significant amount of residual metallic Ir can remain. This may suggest that the observed r‐IrO₂ is limited to the surface and that insufficient oxygen diffusion into the m‐Ir restricted r‐IrO₂ formation beyond the surface. In other words, the results suggest that the bulk m‐Ir did not oxidize and thus could not convert to r‐IrO₂.

2.1.3. X‐Ray Photoelectron Spectroscopy

XPS was used to measure the surface atomic concentrations of the m‐Ir, a‐IrO _x , and r‐IrO₂. The results are depicted in Figure S3 (Supporting Information). No underlying Pt or Ti was detected in the survey spectra, indicating uniform and continuous coverage of all three forms of the sputter‐deposited catalyst layers even for the lower 0.1 mg Ir cm⁻² loadings. The m‐Ir sample has Ir metal (60.9 eV, Ir⁰) and Ir‐O present (61.4 eV) with the Ir⁰ signal dominating. Comparing the Ir 4f spectra of the oxides to the Ir metal, we observe a shift to higher binding energy. Both the a‐IrO _x and r‐IrO₂ samples feature Ir‐O environments only and no detectable Ir⁰. This indicates that for the a‐IrO _x and r‐IrO₂ forms, which were synthesized via oxygen reactive sputter deposition, Ir⁰ is avoided. The as‐deposited catalyst layers contain only the oxide phases at the surface. In general, the m‐Ir PTE surface has less O and more Ir content than the oxide samples. The surface O:Ir ratios were ≈1.0, 3.1, and 2.4 for m‐Ir, a‐IrO _x , and r‐IrO₂, respectively (Table S2, Supporting Information). Note that some of the oxide could also be chemically bound to C or N. Both elements are also detected in the surface composition, though in relatively small amounts.

RBS was used to determine the bulk oxygen content of the a‐IrO _x film. The results are shown in Figure S4 (Supporting Information). From RBS, the ratio of oxygen to iridium for the as‐deposited a‐IrO_x was found to be about 2.9:1. This excess of O relative to the 2:1 stoichiometry expected for pure r‐IrO₂ indicates that a portion of adjacent Ir atoms do not share bonds with a common O atom, as would be the case for a crystalline material.

2.1.4. Electron Microscopy

Scanning electron microscopy (SEM) was used to analyze the morphology and uniformity of the IrO_x films. Characteristic secondary electron (SE) images for each of the surfaces of the PTEs are shown in Figure 2a–c, i.e., the top row of Figure 2. The m‐Ir has a cauliflower‐like morphology at the sub‐micron scale, whereas the a‐IrO_x is smoother, and the r‐IrO₂ exhibits micro‐cracks. Elemental analysis with SEM‐EDS was used to further investigate the coating coverage and uniformity (Figure S5, Supporting Information). Qualitative and quantitative insights were obtained by evaluating two sets of Ti:Ir ratios; atomic percentages (at. %) were derived from EDS spectra; and surface area percentage (SA%) ratios were derived from analysis of each individual EDS map. This data showed that Ir and IrO_x catalyst layers formed homogeneous coatings. Cross‐sectional views of the catalyst coatings, prepared by FIB lift out, are presented in Figure 2d–f, i.e., the center row of Figure 2. The m‐Ir layer was ≈100 nm thick, and the image shows similar contrast for the Ir and the underlying metallic Pt layer. The Ir, however, appears to be more porous with a branching structure consistent with the cauliflower‐like morphology observed in the plan‐view SEM image (Figure 2a). The a‐IrO_x and r‐IrO₂ layers show lower contrast than the m‐Ir due to the lower average atomic number of these layers. They are also thicker at around 150 to 175 nm, which is likely related to their oxygen content and lower density. Atomic‐resolution STEM images of catalyst layer scraped from the PTL surfaces are presented in Figure 2g–i, i.e., the bottom row of Figure 2. The images confirm the amorphous nature of the a‐IrO_x and the crystalline structure of the metallic and rutile catalysts at the nanoscale. They further indicate that the r‐IrO₂ had a larger crystallite size (> 5 nm) than the m‐Ir. Analysis of the fast Fourier transforms (FFTs) of the images (Figure S6, Supporting Information) showed lattice spacings consistent with metallic Ir and rutile IrO₂ crystal structures, respectively.

Figure 2.

Electron microscopy analysis of the m‐Ir, a‐IrO _x , and r‐IrO₂ PTE catalyst layers prepared by sputter deposition. Labeled at the top, the three columns of images correspond to m‐Ir, a‐IrO _x , and r‐IrO₂ from left to right. The top row a–c) of images shows plan‐view characteristic secondary electron (SE) images. The second row d–f) shows cross section STEM images of the Ir catalyst layers on top of the Pt‐coated PTLs. The bottom row g–i) shows atomic‐resolution STEM images of the Ir catalysts scraped from the PTE surfaces.

2.2. BOL Performance

2.2.1. Influence of PTL type and Ir Loading

Figure 3 shows a comparison of polarization curves from sputter‐deposited m‐Ir PTEs with the cell voltage and high‐frequency resistance (HFR)‐free cell voltage plotted versus current density. The bottom of each plot shows the HFR plotted versus current density. The m‐Ir was deposited on two different PTL materials: i) a Ti fiber‐based PTL (Bekaert 2GDL10N) with 56% porosity and ≈20 µm Ti fiber diameter referred to as PTL (purple squares), and ii) a Ti sinter PTL (Toho 303‐K02‐2231001‐01‐02) with 38% porosity and 3 µm pore size (blue circles). The latter PTL is referred to as a low‐porosity PTL (lp‐PTL). Detailed characterization of these PTLs was reported by Duarte et al.,^{[ 36 ]} in which the fiber‐based 2GDL10N PTL and the Toho sinter lp‐PTL are referred to as PTL #1 and PTL #3, respectively. In the present work, the PTEs were prepared and compared at two loadings: i) “low” Ir loading of ≤0.4 mg Ir cm⁻² (Figure 3b), and ii) “ultra‐low” Ir loadings of ≤0.1 mg Ir cm⁻² (Figure 3a). In both figures, the performances of the samples are compared to benchmark data (black triangles) with the same Ir loading. The reference MEA was the DOE H2NEW consortium's Future Generation MEA or “FuGeMEA”, a catalyst‐coated membrane prepared by ultrasonic spray coating with each Ir loading.^{[ 37 ]}

Figure 3.

Polarization curves and HFR of sputter‐deposited m‐Ir at loadings of a) 0.1 mg Ir cm⁻² and b) 0.4 mg Ir cm⁻² on two PTL types: Bekaert 2GDL10N ("PTL", purple squares) and Toho ("lp‐PTL", blue circles) with the FuGeMEA benchmark reference data (black triangles). In the polarization curve plots, solid markers indicate the total cell voltage and open markers indicate the HFR‐free voltage.

Figure 3b shows the data for the cells with 0.4 mg Ir cm⁻² anode loadings. The FuGeMEA CCM outperformed the m‐Ir PTE samples over the entire current range. At 4 A cm⁻² the FuGeMEA cell displayed a ≈75 mV lower cell voltage than either of the PTE based cells, which performed similarly regardless of the morphology of the base PTL material. This performance difference was also reflected in the HFR‐free performance (open markers), though the lp‐PTL‐based PTE slightly outperformed the PTL‐based PTE. At loadings of 0.1 mg Ir cm⁻² (Figure 3a), all three MEA types displayed increased cell voltages and HFRs when compared to the 0.4 mg Ir cm⁻² loading. The FuGeMEA and the PTL‐based PTE required ≈100 and 150 mV higher voltages at 4 A cm⁻², respectively, while the lp‐PTL‐based PTE required a voltage increase of <50 mV at the same current density. The HFR‐free voltage of the FuGeMEA was 100 mV higher at the low loading, while those of the PTL‐based and lp‐PTL‐based PTEs increased by 50 and 10 mV, respectively. The relatively small increase in HFR‐free cell voltage with reduced Ir loading enabled the lp‐PTL‐based PTE to largely preserve its level of performance despite the four‐fold loading reduction from 0.4 to 0.1 mg Ir cm⁻². These results demonstrate that PTE architectures with sputter‐deposited, ionomer‐free catalyst layers can retain relatively high performance at ultra‐low Ir loadings if a sufficiently high interfacial contact area between the electrode and the membrane is provided by a PTL with low porosity.

2.2.2. Comparison of m‐Ir, a‐IrO _x , and r‐IrO₂

The m‐Ir, a‐IrO _x , and r‐IrO₂ thin films were evaluated by sputter‐depositing each on the lower porosity Toho lp‐PTL at loadings of 0.4 and 0.1 mg Ir cm⁻². Preliminary studies indicated that the reactive deposition of IrO_x can lead to elevated PTE through‐plane resistance (Table S5, Supporting Information), as measured by the interfacial contact resistance technique,^{[ 38 ]} likely due to oxidation of the Ti PTL surface during deposition. This challenge was mitigated by first sputter depositing a Pt interlayer with loading of 0.2 mg Pt cm⁻² onto all base PTLs. Note that the Pt interlayer may not have been required for the m‐Ir deposition, as it uses inert process gas, but it was applied nonetheless for consistency.

The polarization curves in Figure 4 show that all PTEs have better total cell voltage at the 0.4 mg Ir cm⁻² loading (Figure 4b) relative to the same material at 0.1 mg Ir cm⁻² loading (Figure 4a). Among the higher loading samples, the m‐Ir has the highest performance (2.07 V at 4 A cm⁻²), the a‐IrO _x the second best (2.12 V) and the r‐IrO₂ the third best (2.16 V). Reductions in loading to 0.1 mg Ir cm⁻² resulted in 40, 250, and 20 mV cell voltage increases for m‐Ir, a‐IrO _x , and r‐IrO₂, respectively, at 4 A cm⁻². In other words, the m‐Ir and the r‐IrO₂ PTEs both show small performance tradeoffs when reducing the Ir loading. In contrast, the a‐IrO _x shows a significant performance decline and an increased HFR that displays a dependence on current density. The HFR increase for a‐IrO_x is 15 mΩ cm² at reduced loading, while those for m‐Ir and r‐IrO₂ are only 5 mΩ cm² higher. The a‐IrO_x was observed to be less stable during conditioning, particularly during the 1.85 V hold, which may have caused its atypical polarization curve shapes. To exclude any influence of the conditioning procedure on the results, subsequent BOL characterizations employed a milder conditioning protocol, as described in the Experimental Section, to mitigate potential instability of the a‐IrO_x PTE.

Figure 4.

Polarization curves and HFR of sputter‐deposited m‐Ir, a‐IrO _x , and r‐IrO₂ at loadings of a) 0.1 mg Ir cm⁻² and b) 0.4 mg Ir cm⁻². Solid markers indicate the total cell voltage, and the open markers indicate the HFR‐free voltage.

2.3. Durability Testing

Figure 5a shows the cell voltage progression of m‐Ir, a‐IrO_x, and r‐IrO₂ PTEs (0.1 mg Ir cm⁻² on 2GDL10N PTL) during a 700 h current hold at 3 A cm⁻², along with ascending polarization curves measured at beginning‐of‐test (BOT) and end‐of‐test (EOT) and corresponding HFR values for each sample (Figure 5b–d). Note that additional diagnostics were conducted at 250 and 500 h. Generally, BOT performances and HFR values were similar for the three samples, but as mentioned above, a gentler conditioning procedure that employed lower current densities (see Experimental Section for details) mitigated the detrimental performance effects that a‐IrO_x experienced in the results shown in Figure 4. The impact of the conditioning procedure suggests that stability of the PTE materials may depend on the operating time spent at specific current densities.

Figure 5.

Cell voltage over time is depicted for a 3 A cm⁻² hold for sputter‐deposited m‐Ir, a‐IrO_x, and r‐IrO₂. Ir loadings on the anode PTEs were 0.1 mg cm⁻² with a 0.2 mg cm⁻² Pt‐interlayer on 2GDL10N PTL material. Polarization curves from intermittent diagnostic measurements are shown for beginning‐of‐test (BOT) and end‐of‐test (EOT) for b) m‐Ir, c) a‐IrO_x, and d) r‐IrO₂, along with corresponding HFR values.

Despite similar BOT performances, the cell voltages of the three samples differed shortly after beginning the 3 A cm⁻² current hold. The a‐IrO_x and m‐Ir performed similarly within the first 100 h of operation, starting at 2 V and increasing in cell voltage to ≈2.1 V by t = 100 h. Meanwhile, the r‐IrO₂ PTE operated at a slightly higher cell voltage of 2.1 V. By t = 150 h, the performances of the three samples crossed, and the m‐Ir and a‐IrO_x cell voltages begin to increase dramatically by ≈500 mV within the next 100 h. The r‐IrO₂ performance, however, remained relatively stable, displaying a slight increase in cell voltage (≈50 mV) within the same period, but settling back in at ≈2.1 V following each diagnostic measurement at t = 250 h and t = 500 h.

The cell voltages of the a‐IrO_x and m‐Ir PTEs also began to diverge from each other following the diagnostic measurement at 250 h. While the cell voltage of the a‐IrO_x sample stabilized at 2.5 V at a low degradation rate on par with those observed in benchmarking tests (≈60 µV h⁻¹),^{[ 39} , ^{40 ]} the cell voltage of the m‐Ir PTE continued to increase over the course of the current hold. Further, while the cell voltage of the m‐Ir PTE exhibited some recovery during polarization curve/impedance diagnostics, its degradation rate increased dramatically with each subsequent portion of the current hold. For instance, the cell voltage of the m‐Ir PTE reached a degradation rate of 1.9 mV h⁻¹ between 488 and 500 h, which increased by a factor of 6.8 to reach 12.9 mV h⁻¹ between 646 and 653 h. The data suggest that, although an initial recovery in cell voltage was observed after the applied performance diagnostics, m‐Ir degradation was also exacerbated.

The resulting EOT polarization curves and HFR data (Figure 5b–d) indicate significant performance loss and resistance increases for the m‐Ir and a‐IrO_x samples. The data demonstrate that the form of the anode catalyst layer strongly affects the extent and reversibility of performance degradation. Metallic Ir shows an increase in cell voltage from 2.0 V at BOT to 2.4 V by EOT at 3 A cm⁻² and an HFR increase of 40 mΩ cm². Despite better performance than m‐Ir during the current hold, a‐IrO_x showed similar EOT performance, with a cell voltage increase from 2.0 to 2.4 V at 3 A cm⁻², although the increase in HFR was only 12 mΩ cm² at 3 A cm⁻². The polarization curve data combined with the current hold data highlight how challenging it is to employ polarization curves and impedance diagnostics to describe durability performance. The degradation rates exhibited by the m‐Ir were prohibitively high, but they were not represented to the same extent in the performance EOT diagnostics, a phenomenon previously observed by others.^{[ 40 ]} While the a‐IrO_x and m‐Ir samples exhibited dramatic degradation, the r‐IrO₂ sample displayed a slight improvement in performance relative to BOT, with the cell voltage at 3 A cm⁻² decreasing by 50 mV and the HFR remaining nearly identical during the 700 h test.

2.4. Half‐Cell Electrochemical Dissolution Studies

Complimentary to the in‐situ experiments, the dissolution behavior of the three forms of IrO_x was studied in a half‐cell configuration using the same PTE architecture. Details of the experimental setup are depicted in Figure S8 (Supporting Information). Figure 6a shows the catalytic activity normalized to geometric area, at 1.7 V_RHE measured before and after a 4 h cyclic voltammogram (4h‐CV) test consisting of cycling between 0.1 and 1.7 V_RHE at 50 mV s⁻¹. In good agreement with the literature, a‐IrO_x showed the highest activity, followed by m‐Ir, and r‐IrO₂.^{[ 41} , ⁴² , ^{43 ]} After the 4h‐CV test, a‐IrO_x lost 43% of its initial OER activity, whereas m‐Ir and r‐IrO₂ lost 32%, and 58%, respectively. Quantification of the electrochemically active surface area (ECSA) using common peak integration methods before and after the 4h‐CV test was not feasible because the contribution of the Pt interlayer to the cyclic voltammetry masked Ir redox features (see Figure S9, Supporting Information). In addition, using the voltage/current to assess stability can lead to erroneous conclusions due to convolution of effects such as substrate passivation and bubble accumulation.^{[ 44} , ⁴⁵ , ^{46 ]} Instead, measurements of dissolved Ir were used as the primary indicator of degradation. Figure 6b shows dissolution results obtained by inductively coupled plasma mass spectrometry (ICP‐MS) analysis of the electrolyte before and after the 4h‐CV test. The largest mass losses were observed for m‐Ir (3.5%) and a‐IrO_x (2%), while a much smaller loss was measured for r‐IrO₂ (0.15%). This trend in dissolution behavior for these IrO_x thin‐film PTEs is consistent with studies of powder‐based catalyst layers, where metallic Ir is less stable than both a‐IrO_x and r‐IrO₂ catalyst layers.^{[ 41 ]}

Figure 6.

a) Electrochemical activity measured as the current density at 1.7 V_RHE before and after the 4h‐CV test consisting of cycling between 0.1 and 1.7 V_RHE at 50 mV s⁻¹; and b) Ir mass loss as calculated from the change in total amount of Ir in the 0.5 m HClO₄ electrolyte after the 4h‐CV test expressed as a wt% of the total mass of Ir in the working electrode.

2.5. Post Durability Catalyst Layer Microscopy

Post‐durability microscopy was used to examine the PTE surface (Figure 7a–c) and the catalyst/membrane interface (Figure 7d–i) after MEA disassembly. Figure 7j–l shows additional high‐resolution images of the catalyst structure. In regions where the catalyst layer was transferred from the PTL to the membrane during disassembly, cross‐sections were examined using HAADF‐STEM (Figure 7d–l) and corresponding elemental maps of the Ir distribution in catalyst layer and membrane were created (Figure 7g–i). For m‐Ir (Figure 7d,g), these images show a significant loss of Ir in the catalyst layer while the a‐IrO_x (Figure 7e,h) and r‐IrO₂ (Figure 7f,i) appear to remain intact. The extent of Ir migration into the membrane differs; m‐Ir and a‐IrO_x show a pronounced band of Ir within 1 µm of the surface, suggesting significant Ir dissolution and migration into the membrane. The Ir band for the m‐Ir is slightly denser than that for the a‐IrO_x. In contrast to these two materials, the r‐IrO₂ shows limited to no formation of an Ir band with sparsely distributed Ir oxide agglomerates in the membrane. The degree of Ir migration into the membrane is consistent with the ex‐situ Ir dissolution study (Figure 6) in which m‐Ir had the highest dissolution, a‐IrO_x the second highest, and r‐IrO₂ featured >10x lower dissolution than both m‐Ir and a‐IrO_x. The durability test results (Figure 5) follow the same trend, suggesting low Ir dissolution as a criterion for durable PTEs. Additional observations of Ir, Pt, and Ti migration are summarized Table S3 (Supporting Information). In short, the r‐IrO₂ PTE showed that no Ir had migrated to the cathode and that no Pt from the PTL coating had migrated into the adjacent anode catalyst layer. In contrast, the m‐Ir and a‐IrO_x PTEs showed that ≈10% Ir relative to Pt had migrated to the cathode and ≈10% Pt relative to Ir had migrated from the PTL coating into the adjacent anode catalyst layer.

Figure 7.

Post‐durability comparison of m‐Ir (left column), a‐IrO_x (middle column), and r‐IrO₂ (right column) catalyst layers with a–c) SEM plan view of PTE surface, d–f) HAADF‐STEM cross section of catalyst/membrane interface with g–i) corresponding EDS Ir elemental maps, and j–l) HAADF‐STEM images of the catalyst at atomic resolution.

Figure 7j–l shows atomic‐resolution HAADF images of the three catalysts after durability testing. The crystallite size of the r‐IrO₂ is larger than that of m‐Ir and a‐IrO_x. Fast‐Fourier transform (FFT) analysis (Figure S10, Supporting Information) indicated that all three catalysts, regardless of their respective structures at BOT (Figure S6, Supporting Information), have been converted to r‐IrO₂ at the catalyst layer surface where atomic imaging was possible. A pronounced rutile (110) plane with d‐spacing of 0.33 nm^{[ 47 ]} was distinguished from all FFT patterns (Figure S10, Supporting Information).

In other regions of the PTE surfaces, the catalyst layer remained attached to the PTL after MEA disassembly. These regions appear as lighter contrast areas of Figure 7a–c while the darker contrast areas correspond to the underlying PTL. Elemental analysis with SEM‐EDS was used to quantify the catalyst coverage and uniformity of the disassembled PTEs. The results are summarized in Table 1 , and images of the measurements are shown in the Supporting Information (Figure S11, Supporting Information). The Ir content of the r‐IrO₂ was 1.3% and the highest that was observed. Those for a‐IrO_x and m‐Ir were slightly lower with 1.1% and 1.2%, respectively. As this measurement quantifies the amount of Ir remaining on the PTE after durability testing, the observation that the r‐IrO₂ PTE had the highest Ir is consistent with the half‐cell Ir dissolution measurements (Figure 6b) as well as with the post‐durability microscopy (Figure 7d–i), which visualize the degree of Ir migration into the membrane. Ti‐Pt‐Ir images from the SEM‐EDS analysis were overlaid with each other. Those images indicated that the a‐IrO_x retained the highest combined Pt and Ir content, while the m‐Ir had the most exposed Ti. This result was confirmed by quantifying the atomic percentage (at %) acquired by the EDS software. Complementary to the individual atomic quantities, the Ti to (Pt+Ir) ratios were determined for the three samples. The a‐IrO_x sample featured the lowest ratio (8.7), which was about doubled for the r‐IrO₂ sample (18.3). The m‐Ir sample had the most elevated ratio (21.0), indicating the highest Ti exposure. The SEM‐EDS results further indicated that the m‐Ir sample had the lowest oxygen presence, while a‐IrO_x and r‐IrO₂ exhibited similar oxygen content, i.e., ≈28%–29 at %. Note that the a‐IrO_x sample had more localized concentrations of oxygen, while the r‐IrO₂ sample displayed a more uniform distribution, which aligned with the Ir distribution.

Table 1.

Post‐durability comparison of PTE surface composition (SEM‐EDS, at.%).

	m‐Ir		a‐IrO2		r‐IrO2
	Average	SD	Average	SD	Average	SD
O	18.5	0.2	27.6	1.3	29.6	2.0
Ti	76.5	0.2	62.2	1.4	65.4	1.9
Ir	1.2	0.1	1.1	0.1	1.3	0.1
Pt	2.5	0.0	6.1	0.2	2.3	0.1
Ti:(Pt+Ir)	21.0	0.8	8.7	0.1	18.3	0.3

2.6. Durability of r‐IrO₂ PTEs: Influence of PTL Porosity

Figure 8a shows the cell voltage of two r‐IrO₂ PTEs during a 700 h current hold at 3 A cm⁻². The loading is 0.4 mg Ir cm⁻², which is used here because it corresponds to the current “low” Ir loading target of the H2NEW consortium. The purple data represent results with a high porosity PTL base material, i.e r‐IrO₂@PTL, and the blue data represent results from a low porosity PTL base material, i.e., r‐IrO₂@lp‐PTL. Figure 8b shows polarization curves and HFR of the same samples measured at BOT and EOT. BOT polarization curves show higher cell voltage, HFR, and HFR‐free voltage for the r‐IrO₂@PTL than for the r‐IrO₂@lp‐PTL. For example, the cell voltage at 3 A cm⁻² was 2.046 V for the r‐IrO₂@PTL and 1.986 V for the r‐IrO₂@lp‐PTL. At the beginning of the 3 A cm⁻² current hold (Figure 8a), the starting cell voltages were similar to those measured in the BOT polarization curves. However, over the first 20 h, the cell voltage for each PTE increased by ≈30 mV before settling into less rapid voltage degradation trends that were approximately linear.

Figure 8.

Cell voltage over time is depicted for a) a 3 A cm⁻² hold for the r‐IrO₂ on PTL (2GDL10N, purple) and on lp‐PTL (Toho, blue) with loadings of 0.4 mg Ir cm⁻². Polarization curves and HFR are shown for b) BOT and EOT measurements. Literature‐reported voltage decay rates are plotted for c) the current density used in durability testing and d) Ir loading. The result for the r‐IrO₂@lp‐PTL from this work is indicated by the blue diamond marker, and the ultimate H2NEW consortium target of 2.3 µV h⁻¹ with 0.4 mg Ir cm⁻² is indicated with a target symbol. For each reported voltage decay rate, detailed test parameters and results are available in Table S4 (Supporting Information).

Following the diagnostic measurements at t = 250 h, a similar transient behavior was observed for each cell upon restart. Interestingly, after returning to a linear voltage decay at t = ≈300 h, the cell voltage of the r‐IrO₂@lp‐PTL sample was ≈30 mV improved compared to its initial steady state voltage. Ultimately, the sample reached a stable voltage decay rate of 6 µV h⁻¹ between 500–700 h. The r‐IrO₂@PTL sample had a significantly higher steady state voltage decay rate of ≈125 µV h⁻¹ which was calculated from the 300–500 h data segment.

Polarization curve and HFR measurements at EOT (Figure 8b, square markers) indicated only small differences from the measurements at BOT (circle markers). The r‐IrO₂@PTL performance declined by ≈25 mV while the r‐IrO₂@lp‐PTL displayed a performance improvement of ≈10 mV. Both PTE's HFR‐free polarization curves (dashed lines) show minimal differences at EOT versus BOT, particularly at high current density. The HFR at 3 A cm⁻² increased by 2 mΩ‐cm² for the r‐IrO₂@PTL sample, and it decreased by 5 mΩ‐cm² for the r‐IrO₂@lp‐PTL sample. The latter HFR change accounted for a ≈15 mV improvement, which is half of the ≈30 mV improvement of the r‐IrO₂@lp‐PTL sample's steady‐state cell voltage observed during restart at t = 250 h. The other half of this improvement is unaccounted for and may have been due to catalyst utilization, local contact area changes or other phenomena that are challenging to track. As also noted in Section 2.3, the performance degradation determined from BOT and EOT polarization curve/impedance diagnostics underestimates the voltage decay rates extracted from steady state operation. Future investigations may apply in‐situ diagnostics, such as impedance spectroscopy, without interrupting steady state operation to better understand the observed changes.

In addition to the results from this work (blue diamonds), Figure 8c,d display the results of a literature survey (black circles) on work reporting voltage decay rates as the metric for durability performance. Survey criteria included the use of an Ir‐based anode catalyst, Nafion membrane with >50 µm thickness, test temperatures of 80 °C–90 °C, and the use of V versus time durability test data to determine the voltage decay rate. The literature‐based voltage decay rates are plotted for the current density of the respective durability test (Figure 8c) and for the Ir loading of the respective cell (Figure 8d). The cell materials, test parameters, and method used to calculate the voltage decay rate are recorded in Table S4 (Supporting Information). A notable study by Lewinski et al.^{[ 33 ]} reported a negative voltage decay rate of −6.8 µV h⁻¹ (an overall performance improvement) when tested at 2 A cm⁻². However, voltage fluctuations of ≈100 mV during portions of the test create significant uncertainty in the decay rate determination, so this result was omitted from Figure 8c,d.

The r‐IrO₂@lp‐PTL sample tested in this work demonstrated a voltage decay rate of 6 µV h⁻¹ (Figure 8c,d, blue diamond markers). The chosen material approach appears to be more stable than those tested recently by the electrolysis community at comparable conditions, i.e., at high current density (3 A cm⁻²) and low Ir loading (0.4 mg Ir cm⁻²).

2.7. Discussion

This work focused on the PTE performance and durability effects of the PTL base material and catalyst material variations. Varying activity tradeoffs were observed in initial performance that resulted from the reduction of Ir loading from 0.4 to 0.1 mg cm⁻². For the m‐Ir on two PTL types (Figure 3) and for each form of IrO_x (Figure 4), total and HFR‐free cell voltages and HFR are higher when the Ir loading is reduced. Most notable for the comparison of the PTL base material is that the PTL with lower porosity (38% versus 56%) and smaller Ti particle size (3 µm versus 20 µm) showed minimal tradeoff in HFR‐free cell voltage. Nearly all total cell voltage losses that occurred when reducing the Ir loading were associated with an HFR increase of ≈20 mΩ cm². When a Pt interlayer (0.2 mg Pt cm⁻²) was used, the same Ir loading reduction (Figure 4a,b pink traces) resulted in a smaller increase in HFR (<5 mΩ cm²). This suggests that the Pt mitigated interfacial resistances, either through lowering resistances between the Ir coating and the Ti and/or by increasing the contact surface area with the membrane. Thus, when reducing Ir loadings, the more cost‐effective Pt may be used to minimize non‐kinetic interfacial losses.

The Pt interlayer also showed a benefit during PTE synthesis of oxygen‐containing catalyst materials. The Pt interlayer limited Ti oxidation during reactive IrO_x deposition and post‐processing at elevated temperature. Such Ti oxide growth, known to also occur during cell operation without proper mitigation, can lead to high PTL and PTE contact resistances even for small oxide thicknesses.^{[ 48 ]} The chosen synthesis path thus included the formation of a Pt interlayer to prevent undesired impacts on the activity and durability studies of the m‐Ir, a‐IrO_x, and r‐IrO₂ PTE samples (Figures 4 and 5). The PTEs with m‐Ir and a‐IrO_x catalyst had better initial performance, which is consistent with our measured kinetic parameters and known intrinsic activity of each.^{[ 41 ]} However, reducing the Ir loading came with significant tradeoffs, particularly for the a‐IrO_x sample. The performance of this material was unstable during our routine conditioning hold at 1.85 V, had an unusual progression of the HFR with current density, and featured a 0.35 V higher cell voltage at 4 A cm⁻² (Figure 4). The r‐IrO₂ PTE sample, while having the lowest performance at the beginning of testing, offered a tradeoff of as little as ≈20 mV increase in cell voltage when reducing the Ir loading to 0.1 mg cm⁻². Thus, the ability for PTEs to retain performance when Ir loadings are reduced might be linked to catalyst stability and may also depend on specific conditioning and operating points.

Before durability testing, the gentler conditioning procedure yielded similar BOT performance for all three forms of IrO_x (Figure 5b–d). During the durability test, m‐Ir and a‐IrO_x slightly outperform r‐IrO₂ for the first ≈100 h at 3 A cm⁻² (Figure 5a). Thus, the BOT and initial durability performances agree with the intrinsic activity trends (Figure 6a). However, at 150 h of durability testing, voltages for both m‐Ir and a‐IrO_x exceeded that for r‐IrO₂, and throughout the first 250 h of operation, both m‐Ir and a‐IrO_x exhibited an S‐shaped increase in voltage. Such behavior has been observed in ultrasonic spray‐coated a‐IrO_x systems and has been shown to be dependent on anode catalyst loading, as well as operating current density and temperature.^{[ 40 ]} This behavior is thus not restricted to the PTEs used in this study and likely depends on a variety of MEA parameters and operating conditions. Its dependence on loading, in particular, where a significant portion of losses have been observed to be irreversible at low anode catalyst loadings (0.1 mg Ir cm⁻²), suggests that Ir dissolution from the catalyst layer may be the primary driver of the observed losses. Any such dissolution effects on performance are expected to be masked by higher catalyst loadings and cannot be reversed via a change in operating conditions, such as a return to low current densities or a change in temperature.^{[ 39} , ⁴⁰ , ⁴⁹ , ^{50 ]} It is additionally well‐known that metallic Ir and amorphous IrO_x are less stable under PEMWE conditions than rutile IrO₂, and a higher degradation rate would thus be expected for m‐Ir materials relative to a‐IrO_x and r‐IrO₂ materials, as has been observed in the PTE architecture developed here (Figure 6).

Notably, the performance of a‐IrO_x stabilized during the degradation experiment of this work, while the m‐Ir continued to degrade at an increasing rate. This may indicate the ability of a‐IrO_x to transform, at least partially, to a more stable crystalline form, as is commonly observed following operation over extended time periods.^{[ 39} , ⁵¹ , ⁵² , ^{53 ]} Some r‐IrO₂ was also observed in the surface of the m‐Ir sample after durability testing (Figure 8g; Figure S10, Supporting Information). This transformation may be limited to the surface due to slow oxygen diffusion into the bulk, a hypothesis that is consistent with our exploration of two synthesis pathways for creating the r‐IrO₂ PTE. During an anneal, the a‐IrO_x crystallized within minutes at 500 °C, while the m‐Ir showed minimal formation of r‐IrO₂ and significant residual m‐Ir after 18 h.^{[ 54} , ⁵⁵ , ⁵⁶ , ^{57 ]} Thus, stable operation may be facilitated by the rate of formation of r‐IrO₂ during operation, which ideally exceeds that of irreversible Ir dissolution. The extent of this transformation to r‐IrO₂ during operation, however, is unclear. For example, although microscopy suggested that all three anode catalyst layers were transformed to r‐IrO₂ (Figure 7), this transformation may have been limited to surface‐active sites, where remaining domains of the less stable phases would present an ongoing risk for dissolution losses. The continued increase in degradation rate observed for m‐Ir, for instance, suggests that the transformation to r‐IrO₂ likely did not extend throughout the entirety of the anode catalyst layer. Thus, the fabrication of phase‐pure r‐IrO₂ as the native catalyst is helpful in limiting irreversible degradation and enabling the high current density durability of PTEs with low and ultra‐low Ir loadings.

It is also noteworthy that the form of the IrO_x appears to influence the intermixing of the Pt into the IrO_x layers during durability testing. Before durability testing, the pristine sample STEM cross sections (Figure 2d–f) show distinct and abrupt contrast at the interface between Pt and both the a‐IrO_x (Figure 2e) and r‐IrO₂ (Figure 2f) layers, indicating that intermixing of the two layers appears to be insignificant initially. The Pt and m‐Ir (Figure 2d) have low contrast such that a distinct interface is not observable, but it may also be expected to be abrupt because the m‐Ir deposition conditions are similar to those of a‐IrO_x, which shows an abrupt interface with Pt. After durability testing, there appears to be some intermixing of the Pt into the m‐Ir and a‐IrO_x layers (9%–10% Pt relative to Ir), while no Pt was detected in the r‐IrO₂ layer (Table S3, Supporting Information). The absence of Pt intermixing in the r‐IrO₂ layer could be due to multiple reasons, such as the higher chemical stability or higher density and crystallinity of r‐IrO₂ acting as a barrier to Pt diffusion. The Pt intermixing mechanism and the degree to which it influences activity and durability warrant further investigation.

3. Conclusion

As a pathway to durable, low‐Ir loading PEM electrolyzers, we demonstrated the use of sputter deposition to fabricate thin, uniform, and ionomer‐free porous transport electrodes with controlled composition. Going beyond previous work on m‐Ir, we utilized methods that enabled depositing thin films of pure a‐IrO_x and r‐IrO₂. PTE samples produced with these methods were evaluated through in‐situ activity and durability testing, ex‐situ dissolution measurements, and electron microscopy. Two potential synthesis pathways for r‐IrO₂ thin film synthesis, having starting points of m‐Ir and a‐IrO _x , were evaluated with the criteria of minimizing time and thermal energy input for full formation of rutile crystallinity while avoiding undesirable oxidation of the Ti PTL base material. The a‐IrO_x film could be converted to r‐IrO₂ within 10–30 min at 500 °C, while attempting to form r‐IrO₂ from m‐Ir yielded minimal r‐IrO₂ after 18 h, and significant residual m‐Ir remained.

The PTL base material was shown to have an influence on performance when reducing the overall Ir loading of the cell. Results showed that sputter‐deposited PTE catalyst layers can retain high performance when using a PTL with low porosity, which appears to be required to provide a high interfacial contact area with the membrane. In‐situ activity and durability testing and ex‐situ dissolution studies yielded consistent results for m‐Ir and a‐IrO_x. Both offered higher initial activity than r‐IrO₂ but were surpassed within 150 h of operation at 3 A cm⁻². The r‐IrO₂ material featured superior stability and displayed more than 10x lower dissolutions rates than m‐Ir and a‐IrO_x. With a low‐porosity transport layer and 0.4 mg Ir cm⁻², PTEs with the r‐IrO₂ catalyst layer achieved a steady‐state voltage decay rate of 6 µV h⁻¹ at 3 A cm⁻². A literature survey of durability test results shows that this electrode stability is state‐of‐the‐art, particularly for durability testing conducted at high current density and low Ir loading.

This work demonstrates that sputter‐deposited thin films of r‐IrO₂ are a means to form durable PTEs for high current density PEM water electrolysis. These PTEs are a readily scalable electrode option for reducing electrolyzer stack and hydrogen production costs. Future efforts may focus on further improving the initial activity of r‐IrO₂ catalyst layers by increasing their electrochemically active area by using advanced PTLs, integration strategies, and deposition techniques.

4. Experimental Section

Sputter‐Deposited Catalyst Layers

Metallic Pt and Ir were DC magnetron sputter deposited onto PTL materials to form PTEs in a Denton Desktop Pro system. Sputter targets were 2″‐diameter Pt (99.99%) and Ir (99.95%), respectively. The deposition chamber was evacuated to at least 5E‐4 Torr before flowing Ar gas (UHP 99.999% pure) at 50 sccm to establish a deposition pressure of 12–13 mTorr. The sputtering power was 20 W. Pre‐sputtering at the same conditions was conducted for 2 min. In addition to depositing on glass slides, two porous transport layers were used as substrates: 2GDL10N – 0.25 Bekaert Ti fiber PTL (56% Porosity) or Toho Ti sinter PTL (model 303‐K02‐2231001‐01‐02). The Toho PTL has a much finer pore size (3 µm) and lower porosity (38%), compared to the Bekaert PTL with an average pore size of 17 µm and a porosity of 56%. The as‐received PTL samples were cleaned by bath sonication for 10 min in 1:1 DI:IPA and then dried with a N₂ stream without additional treatments, such as etching to remove native Ti oxide, before coating. Some samples contained a Pt interlayer, where indicated, between the PTL material and Ir coating. The Pt loadings for these samples were 0.2 mg Pt cm⁻². Loadings were controlled by the sputter deposition time, which was found to have a linear relationship. All IrO _x catalyst layers were either 0.1 or 0.4 mg Ir cm⁻², and all loadings were confirmed by x‐ray fluorescence (XRF) spectroscopy (Fischer XDV‐SDD). The a‐IrO _x was deposited by reactive sputtering, and the r‐IrO₂ PTEs were obtained by annealing the a‐IrO _x as described in the Results section.

Fabrication of Catalyst Coated Membranes (CCMs)

The PTE anodes were paired with half‐CCM cathodes that were prepared by ultrasonic spray coating. Cathode catalyst, TKK TEC10E50 E, 46.7 wt% Pt was mixed with deionized water (DI, 18.2 MΩ cm) and HPLC‐grade n‐propanol (n‐Pa, OmniSolv). Nafion D2020 ionomer was added in an ionomer‐to‐catalyst ratio of 0.3. The water to n‐Pa ratio of the ink was 1.63. The resulting catalyst inks were dispersed using 2 min of horn sonication followed by 20 min of sonication in an ice bath and subsequently sprayed onto Nafion 115 membranes (Chemours) using a Sono‐Tek spray station featuring a 25 kHz accumist nozzle. During the spray process the membrane was held at 80 °C on a vacuum plate. Target catalyst loading was 0.1 mg Pt cm⁻², and the spray area was slightly larger than 5 cm² area. Actual platinum Pt loadings were determined by XRF (Fisher XDVSDD). For details on the ink‐making process for the cathode, refer to Parimuha and Young et al.^{[ 37 ]}

Membrane Electrode Assembly (MEA)

The CCMs were assembled in a standard electrolysis hardware^{[ 37 ]} with the Pt catalyst layer serving as the cathode and AvCarb MGL280 as the cathode gas diffusion layer (GDL). The PTEs were used as a combined anode catalyst and porous transport media. Skived polytetrafluoroethylene (PTFE) gaskets were added to both sides. The anode gasket thickness matched the PTE thickness, while the cathode gasket thickness allowed for a ≈20% compression of the GDL material. The cathode gasket window was exactly 5 cm² and thus reduced the active area of the oversized cathode electrode to the desired value. The reference baseline MEA is the U.S. Department of Energy's H2NEW research consortium standard, Future Generation Membrane Electrode Assembly (FuGeMEA). It contained a two‐sided CCM structure, i.e., ultrasonically sprayed anode and cathode electrodes. The configuration of the FuGeMEA is 0.4 mg cm⁻² ultrasonic spray‐coated IrO _x (iridium(IV) oxide, Premion 99.99%, Alfa Aesar) operated with a Bekaert 2GDL10‐0.25 PTL at the anode. For the cathode, 0.1 mg cm⁻² Pt/C and a AvCarb MGL280 GDL is used, once again with PTFE gasketing to achieve ≈20% compression of the GDL.

In‐Situ Electrochemical Diagnostics

During in‐situ electrochemical conditioning and BOT characterization (Section 2.2), the test cell was held at 80 °C with a dry cathode and 50 mL min⁻¹ DI water supplied to the anode. The cell pressure was 0.83 bar under ambient conditions in Golden, CO, USA. The cell conditioning procedure consisted of a voltage hold and two sequences of polarization curves. After an initial 12 h 1.85 V hold, each cell underwent a series of 10x increasing and decreasing current‐controlled polarization curves. Current densities from 0.01 to 4 A cm² were held for 3 and 0.5 min, respectively, for the first and second sequences of polarization curves. The entire conditioning protocol lasted 40 h. Upon completion of the conditioning protocol, the cell was held at 1.4 V. Performance characterization consisted of a single increasing polarization curve with paired EIS conducted on a Gamry Reference 3000 potentiostat/galvanostat with Reference 30 K Booster. Each current density step of the polarization curve was held for 1 min immediately followed by the corresponding EIS sequence, which spanned a frequency range of 40 kHz–0.05 Hz and used a perturbation of 5% of the current set point. Last, a 100 cm³ min⁻¹ stream of hydrated H₂ at 80 °C was supplied to the cathode for a cyclic voltammetry series from 0.01–1.3 V at 25, 50, 75, 100, and 125 mV s⁻¹ scan rates. For comprehensive details, refer to the protocol section in Parimuha and Young et al.^{[ 37 ]}

Durability Testing

For durability testing (Section 2.3), PEMWE cells were assembled as described in Section 4.3 and operated simultaneously under galvanostatic conditions in a four‐cell test station at atmospheric pressure (0.85 bar), 80 °C, an anode inlet water flow of 50 mL min⁻¹, and a dry cathode. Long‐term current holds were performed using the test station power supply, while diagnostic measurements were performed on the same Gamry potentiostat/galvanostat with booster.

Conditioning

The cells were conditioned on the test station simultaneously using a gentle galvanostatic conditioning procedure consisting of six‐minute holds at 0.05, 0.25, 0.5, and 0.75 A cm⁻², followed by a hold at 1 A cm⁻² for ten hours. This was followed by four ascending/descending polarization curves up to 1 A cm⁻² (three minutes at steps 0.012, 0.018, 0.042, 0.06, 0.078, 0.1, 0.12, 0.14, 0.16, 0.18, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1 A cm⁻²), four ascending/descending polarization curves up to 2 A cm⁻² (same steps as previous with three minutes at 1.5 and 2 A cm⁻² added), four ascending/descending polarization curves up to 3 A cm⁻² (three minutes at 2.5 and 3 A cm⁻² added), and one ascending/descending polarization curve up to 4 A cm⁻² (3 min at 4 A cm⁻² added).

Diagnostic Measurements

Following cell conditioning and before the start of the current holds, polarization curve/impedance diagnostics, and voltage‐controlled non‐Faradaic impedance measurements were performed at 1.35 V with an Autolab potentiostat/galvanostat and 20 A booster (PGSTAT302N, Metrohm, 40 kHz to 0.4 Hz), while galvanostatic polarization curve/impedance diagnostic measurements up to 4 A cm⁻² were performed on all four cells simultaneously using a Gamry Reference 3000AE (with auxiliary electrometer) and 30 k booster (30 A) at frequencies from 300 kHz to 0.1 Hz. The cells were electrically disconnected from the test station each time diagnostic measurements were performed.

Durability Holds

The four cells were held at 3 A cm⁻² for 750 h, with polarization curve/impedance diagnostics measured at BOL, 250, 500, and 750 h. Cell operation was interrupted near 660 h due to the metallic Ir sample reaching a cell voltage of 3 V, which exceeded the voltage limit set for the test station. Upon reaching this limit, the cells were maintained at 80 °C with DI water flowing at the anode, with the cells uncontrolled by the power supply. Gaps in operation time have been removed in Figure 8 to facilitate data presentation. Similarly, an outage due to test stand maintenance issues occurred at the 250 h mark and lasted for nine days, during which the test stand was fully shut down with no water flow or cell heating.

Electrochemical Half‐Cell Tests

Sputter‐deposited m‐Ir, a‐IrO_x, and r‐IrO₂, and spray‐coated IrO_x (Alfa Aesar) were evaluated using an electrochemical half‐cell shown in Figure S8 (Supporting Information). The IrO_x was coated on 25 cm² PTL (Bekaert 2GDL10N) at a loading of 0.1 mg Ir cm⁻². The PTL material was then laser cut to 1 cm² disc samples and soaked for 24 h in 0.5 m HClO₄ before electrochemical investigation to minimize the amount of bubbles trapped in the PTL pores.

The electrolyte solution was 100 mL of 0.5 m HClO₄ with a Nafion 212 (Chemours) membrane separating the working and counter electrode compartments. The membrane was soaked in 0.5 m HClO₄ overnight to ensure full hydration before testing. A Pt counter electrode and working electrode were placed adjacent to the membrane, along with a RE‐5B Ag/AgCl reference electrode (BASi). The reference electrode was calibrated to a reversible H₂ electrode and determined to have a potential of 0.25 V versus RHE. In the working electrode compartment, the PTE was fixed to a Pt wire. A glass bubbler allowed for purging of the cell and electrolyte solution with nitrogen gas (N₂), and a stir bar at 75 rpm was used to agitate the solution. All labware materials, except the reference electrode and PTEs, were cleaned in a Miele industrial dishwasher to minimize contamination.

A Biologic SP‐300 Potentiostat was used for electrochemical characterization. Potentiostatic electrochemical impedance spectroscopy (PEIS) was measured at open circuit with an AC voltage perturbation of 5 mV, to determine the ohmic correction employed during electrochemical testing, which was ≈20 Ω. Cyclic voltammetry (CV) between 0.1–1.7 V_RHE at 5 mV s⁻¹ was conducted for electrochemical characterization, and the current at 1.7 V_RHE served as a measure of electrochemical activity extracted from the 5th cycle. Dissolution cycling testsDCT were applied using CVs with the same potential window but at a scan rate of 50 mV s⁻¹and over a 4 h period after the initial CV was conducted. Aliquots of electrolyte were collected after the initial CV (t = 0 h), after the first hour of the test, and after four hours of the test. 1 mL electrolyte aliquots were collected, diluted in triplicate to 5 mL with MilliQ water (18.2 MΩ cm), and analyzed using a Thermo Scientific iCAP Q ICP‐MS instrument to detect Ti, Pt, and Ir dissolution products in the electrolyte.

Physical Characterization

In‐situ structural characterization was performed on the m‐Ir and a‐IrO_x films that were deposited on a glass slide using a Bruker D8 Discover diffractometer with Cu Kα radiation to determine crystallization while heating. The hot stage was an Anton Paar DHS900, used without a dome for ease of access. Individual XRD scans were 1.5 min long, taken 2 min apart, and the ramp rate used was 5 °C min⁻¹ up to 500 °C, and held for 30 min or 18 h as indicated. The 2D Vantec detector was at a set location, while the sample stage (w (omega) axis) rotated through the full q (theta) range covered by the detector. Data was then integrated in c (chi) over a range of 21° to 56° in 2q (theta).

The composition of the as‐deposited a‐IrO _x was characterized by Rutherford Backscattering (RBS) in a 168° backscattering configuration with a 2 MeV He⁺ beam energy. The instrument used was a model 3S‐MR10 RBS system from National Electrostatics Corporation. Oxygen composition was determined by fitting using RUMP analysis software.^{[ 58 ]}

Using XPS, the surface (≈10 nm) atomic concentrations of the m‐Ir, a‐IrO _x , and r‐IrO₂ PTEs (Figure S3, Supporting Information) were measured. XPS data were obtained on a PHI VersaProbe III using Al Kα radiation (1486.7 eV). The XPS data were calibrated with Au and/or Cu metal, which was cleaned via Ar‐ion sputtering. The raw atomic concentration has a 5% error due to surface inhomogeneities, surface roughness, and literature sensitivity values for peak integration.

A Thermo Fisher Scientific Helios 5 CX was used to create plan‐view secondary electron images, employing an accelerating voltage of 20 kV. A Hitachi NB5000 focused ion beam was used to prepare cross sections of the surfaces of the catalyst layers, which were subsequently imaged in a JEOL 2200FS STEM. To capture atomic‐resolution images of the pristine IrO_x, catalyst was scraped from the PTE surface, suspended in IPA, then drop‐cast onto lacey carbon‐coated TEM grids. Post‐test PTEs were imaged using a Hitachi S‐4800 SEM operated at 30kV. Energy dispersive X‐ray spectrum (EDS) images were recorded to analyze the surface composition. Cross sections of the tested MEAs were prepared using diamond knife ultramicrotomy and imaged in an aberration‐corrected JEOL NEOARM STEM operated at 200 kV. EDS maps were processed using standardless quantification routines in the JEOL Analysis Station software.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

J.L.Y conceived and directed this work and led the manuscript writing. D.B contributed to writing, testing, data analysis, and editing (electrochemical testing), H.Y. (microscopy), S.B. (durability testing), R.P.M.D (ex‐situ dissolution), L.v.E. (microscopy), M.R.P (experimental, literature survey), E.M.M. (XPS), J.D.L (ex‐situ XRD), K.N.H. (in‐situ XRD and RBS), and D.A.C (microscopy). Additional experimental contributions and manuscript reviewing were provided by J.M. (ex‐situ dissolution), B.A.C (electrochemical testing), J.W. (electrochemical testing), M.E.K (ICP‐MS), S.P. (microscopy), T.G. (FIB liftouts), and K.S.R (microscopy). G.B. reviewed and edited the manuscript.

Supporting information

Supporting Information

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.