Heterophase RuO₂ oxygen evolution catalyst for durable proton exchange membrane water electrolysis

Abstract

Dissolution of ruthenium (Ru) species under harsh anode operating conditions severely impedes the commercialization of non–iridium (Ir) proton exchange membrane water electrolyzers (PEMWEs). Here, we propose a catalyst design strategy that engineers phases beyond conventional atomic ordering, constructing a heterophase MoRuO_x catalyst (AC-MoRuO_x). This catalyst achieves 10 milliamperes per square centimeter in an acidic oxygen evolution reaction with merely 180-millivolt overpotential while operating stably for >3000 hours. In PEMWE devices, AC-MoRuO_x delivers record durability at industrial current densities: ≥2000 hours at 1.0 ampere per square centimeter and 1000 hours at 1.5 amperes per square centimeter. We demonstrate that heterophase architecture and atomic arrangements reconfigure electronic structures and phase distribution, simultaneously optimizing active-site density and structural integrity. This breakthrough resolves the long-standing activity-stability trade-off in ruthenium-based catalysts.

A heterophase RuO₂ catalyst delivers industrial-grade durability, overcoming a key barrier to green hydrogen commercialization.

INTRODUCTION

Proton exchange membrane (PEM) water electrolyzers (PEMWEs), as the core of green hydrogen production technology, have attracted much attention because of the high current density, fast response, and high-purity hydrogen production characteristics (1–5). However, the harsh acidic environment at the anode necessitates the use of precious metal–based catalysts for the oxygen evolution reaction (OER), which suffers from intrinsically slow kinetics (6–8). Now, iridium (Ir)–based catalysts have excellent performance, but their scarcity and high cost severely limit the large-scale application of PEMWEs (9–14). Ruthenium (Ru)–based materials are considered as ideal alternatives because of their cost advantage (only one-eighth of Ir) and the excellent OER activity of RuO₂ (15–17). However, RuO₂ is susceptible to peroxidation in acidic media to generate soluble RuO₄, leading to structural collapse and dissolution of Ru species, and the long-term stability is far from meeting the practical needs (10, 18–21). How to synergistically enhance the activity and stability of Ru-based catalysts has become a key challenge to promote the commercialization of non-Ir PEMWEs.

To address this challenge, researchers have explored various strategies. Metal doping (e.g., with Mo) has been shown to modulate the local electronic structure of RuO₂, suppressing overoxidation to some extent but often at the cost of sacrificed intrinsic activity (22–25). On the other hand, constructing amorphous-crystalline (AC) heterophases can generate abundant interfacial active sites and enhance structural flexibility (26, 27). However, a heterophase structure lacking effective electronic modulation may still be insufficient to completely inhibit Ru dissolution at high anodic potentials. This dilemma underscores a prominent deficiency in current material design: These strategies have frequently been used in isolation, resulting in a lack of intrinsic synergy between electronic modulation and structural engineering. Given this, we propose a previously unknown design principle that moves beyond the application of individual strategies to their deliberate synergistic integration. We hypothesize that coupling the electronic effect of Mo doping with the structural advantages of an AC heterophase could be the key to breaking the activity-stability trade-off. Crucially, in our design, Mo doping is not merely an additive component but plays an instrumental role in inducing and stabilizing the beneficial heterophase interface during synthesis. This synergistic design is envisioned to enable a spatial separation of functions: The crystalline domains ensure efficient electron transport, the amorphous matrix acts as a flexible buffer to accommodate structural stress, and the Mo dopants, strategically located at the interfaces, electronically passivate the Ru sites by lowering their oxidation state and weakening the Ru─O covalency, thereby raising the energy barrier for peroxidation.

On the basis of the above strategy, we successfully constructed a MoRuO_x catalyst (AC-MoRuO_x) with the phases beyond ordered atomic arrangements. Experiments showed that Mo doping effectively inhibited Ru peroxidation and solvation by lowering the Ru charge state and weakening the Ru─O bond covalency, while the AC interface stabilized the valence state of Ru through separating charge aggregation and intermediate adsorption processes, and the simultaneous optimization of the activity and stability was achieved by taking advantage of the high electrical conductivity of the crystalline region and the structural flexibility of the amorphous region. The catalyst exhibits an ultralow overpotential of 180 mV (10 mA cm⁻²) in acidic OER and operates stably at 10 mA cm⁻² for more than 3000 hours with negligible Ru dissolution. The PEMWE device based on this catalyst achieves a current density of 2.0 A cm⁻² at 1.65 V and operates stably at ultrahigh current densities of 1 and 1.5 A cm⁻² for no less than 2000 and 1000 hours, respectively, which is far superior to existing Ru-based catalysts. Moreover, the estimated cost of producing 1 kg of hydrogen from this PEM electrolyzer is approximately $0.91, well below the US Department of Energy’s target of $2 per kilogram of hydrogen. Theoretical calculations demonstrated that the heterophase significantly reduces the OER decisive speed-step energy barrier by modulating the electronic energy band structure and interfacial charge distribution. This breakthrough provides a promising design paradigm for the development of non-Ir catalysts that combine high activity with industrial-grade stability.

RESULTS

Synthesis and characterizations of catalysts

The MoRuO_x catalysts appearing in this work were obtained by a simple mixing-evaporation-annealing process (Materials and Methods and Fig. 1A). First, we dispersed ruthenium chloride, molybdenum acetylacetonate, and poly(vinyl pyrrolidone) in an ethanol solution, which was then evaporated to obtain a gelatinous precursor and lastly annealed in air to remove the carbon source and simultaneously convert the precursor into the corresponding MoRuO_x catalysts, where poly(vinyl pyrrolidone) not only acts as a steric stabilizer to prevent nanoparticle agglomeration but also serves dual roles as a pore-forming template and a crystallization regulator. This is crucial for the formation of catalysts with high specific surface area. By controlling the annealing temperature, we synthesized three distinct forms of Mo-doped Ru oxide: A-MoRuO_x with an amorphous structure at 300°C, AC-MoRuO_x with a heterophase structure exhibiting AC coexistence at 400°C, and C-MoRuO_x with a crystalline structure at 500°C. Notably, at an annealing temperature of 400°C, the crystallinity of RuO₂ was significantly enhanced in the absence of molybdenum acetylacetonate. This observation suggests that in addition to temperature, the incorporation of molybdenum atoms influences the atomic arrangement and crystallinity of RuO₂. X-ray diffraction (XRD) patterns (Fig. 1B) showed that the MoRuO_x catalysts prepared at different annealing temperatures had the same rutile RuO₂ crystal structure. As the annealing temperature increased, the diffraction peaks in the (110) and (101) facets shifted to a lower angle and became sharper, indicating that the large-radius element (Mo⁶⁺, 0.62 Å) was successfully introduced into the structure of RuO₂ (Ru⁴⁺, 0.54 Å) (Fig. 1C) and the crystallinity of the catalysts increased with increasing annealing temperature (23, 28). Specifically, when annealed at 300°C, the prepared A-MoRuO_x had no obvious RuO₂ peaks, indicating that the catalyst was amorphous. When the temperature increased to 400°C, the obvious diffraction peaks of RuO₂ were observed, but the peak shapes were slightly broader and not sharp, implying that AC-MoRuO_x was an AC coexisting phase with the heterophase. As the annealing temperature was increased to 500°C, the diffraction peaks of RuO₂ were obviously sharper than those of AC-MoRuO_x at 400°C and the half peak widths were very narrow, which indicated that C-MoRuO_x had good crystallinity.

Fig. 1. Synthesis and characterizations of the prepared AC-MoRuO_x catalyst.

(A) Schematic illustration for the synthesis of the MoRuO_x catalyst. (B) XRD patterns of synthesized MoRuO_x catalysts with different crystallinity, like C-MoRuO_x, AC-MoRuO_x, A-MoRuO_x, and homemade pure RuO₂. a.u., arbitrary units. (C) Extended XRD patterns of MoRuO_x and pure RuO₂ derived from (B). (D) HRTEM image and (E) enlarged image of AC-MoRuO_x. (F) Corresponding energy dispersive spectral mapping images of AC-MoRuO_x. Scale bars, 50 nm (D), 5 nm (E), and 50 nm (F).

The effect of calcination temperature on the morphology and crystallinity of the MoRuO_x catalysts was further elucidated using high-resolution transmission electron microscopy (HRTEM) and aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (STEM). As depicted in Fig. 1D, AC-MoRuO_x displayed abundant nearly transparent nanoareas resulting from defects or structural porosity. The HRTEM image and enlarged image further revealed that defects or structural porosity were mainly distributed in amorphous regions (marked by white dashed lines) (Fig. 1E and figs. S1 and S2). At the same time, the lattice spacing (0.334 nm) of the crystal region was assigned to the (110) crystal plane of RuO₂ (fig. S1) (12). This was in agreement with the selected area electron diffraction results (fig. S3), in which several discrete spots assemble with a diffractive halo-like pattern, suggesting the coexistence of amorphous and crystalline phases, thus further confirming this well-defined boundary (AC) on AC-MoRuO_x (19). Notably, the strain values in the crystalline region did not vary much away from the boundary, as shown by the results of the geometrical phase analysis, in contrast to the previously reported results that AC boundaries induce surface tensile strains (fig. S4) (27). The corresponding energy-dispersive x-ray element mappings (Fig. 1F) confirmed the homogeneous distribution of Ru, Mo, and O elements on AC-MoRuO_x. For A-MoRuO_x, the chemical composition was consistent with the amorphous region in AC-MoRuO_x, besides the near absence of crystalline domains, indicating an amorphous structure (fig. S5), and contained almost no unconverted organic groups (fig. S6). On the contrary, C-MoRuO_x showed a high degree of crystallinity, with no obvious defects detected, and the overall structure and composition remain consistent with the crystalline regions in AC-MoRuO_x species (fig. S7). In addition, this result also applies to homemade RuO₂ (figs. S8 to S10). The elemental composition of the MoRuO_x catalysts was also confirmed by x-ray photoelectron spectroscopy (XPS) results (fig. S11). The above results indicate that the atomic arrangements of MoRuO_x catalysts can be adjusted by tuning the calcination temperature and introducing foreign Mo atoms, which are consistent with the XRD results.

Acidic OER performance evaluation of the catalysts

To evaluate the OER activity of the as-fabricated catalysts, we performed OER measurements in acidic media using a three-electrode system. Linear sweep voltammetry curves showed that AC-MoRuO_x presented intriguing OER activity with an overpotential of 180 mV at a 10 mA cm⁻² current density than homemade RuO₂ (190 mV) and commercial RuO₂ (Com RuO₂; 280 mV) (Fig. 2A). AC-MoRuO_x also showed the lowest Tafel slope of (48.3 mV dec⁻¹) than the homemade RuO₂ (66.4 mV dec⁻¹) and Com RuO₂ (85.7 mV dec⁻¹), implying superior OER kinetics (Fig. 2B. Similarly, AC-MoRuO_x had the smallest charge transfer resistance in electrochemical impedance spectroscopy (EIS) (Fig. 2C and table S1), which also demonstrated the superior OER kinetics of AC-MoRuO_x (29). The AC coexisting AC-MoRuO_x with a heterophase exhibited much higher OER activity than crystalline C-MoRuO_x (193 mV at 10 mA cm⁻²) and amorphous A-MoRuO_x (no superior OER activity) when elemental compositions and contents were identical, indicating that the heterophase optimized the OER activity of MoRuO_x (figs. S12 and S13). The above results implied that the heterophase simultaneously promotes the OER process of AC-MoRuO_x. Besides, the superior activity of AC-MoRuO_x can be reliably repeated (fig. S14).

Fig. 2. Electrocatalytic OER performance of the catalysts in a 0.1 M HClO₄ electrolyte.

(A) Polarization curves and (B) Tafel slopes of AC-MoRuO_x and the references. (C) EIS of AC-MoRuO_x, RuO₂, and Com RuO₂. CPE, constant phase element. (D) C_dl plots derived from CV curves. (E) Chronopotentiometry of AC-MoRuO_x and the benchmarks at 10 mA cm⁻² for 3000 hours in a 0.1 M HClO₄ electrolyte. h, hours. (F) Comparison of the durability of our catalysts in an acidic electrolyte with that of reported OER catalysts.

To better explore the origin of the superior OER performance of AC-MoRuO_x, the electrochemically active surface area (ECSA) and roughness factor (R_f) were calculated through electrochemical double-layer capacitance (C_dl) for activity normalization (Fig. 2D, fig. S15, and table S2). Apart from amorphous A-MoRuO_x, AC-MoRuO_x showed higher C_dl and ECSA than the crystalline C-MoRuO_x, homemade RuO₂, and Com RuO₂, reflecting that the heterophase and atomic arrangements can significantly increase the density of active sites. Considering ECSA-normalized OER activity, specific activity was still in the order AC-MoRuO_x > homemade RuO₂ > Com RuO₂ > C-MoRuO_x > A-MoRuO_x (fig. S16), suggesting that the heterophase and atomic arrangements resulted in an intrinsic enhancement in the OER activity of AC-MoRuO_x. Even after normalization by total anodic charge, AC-MoRuO_x maintained the highest intrinsic OER activity among all samples, confirming that its superior performance originates from heterophase engineering rather than surface area effects (fig. S17). A Brunauer-Emmett-Teller specific surface area of 153.8 m² g⁻¹ improves the accessibility of reactants to the catalytic sites and facilitates water-oxygen mass transport in and out of the catalyst structure (fig. S18). In addition, AC-MoRuO_x was also superior to those of the prepared TMRuO_x (TM = Ti, Cr, Mn, Ga, Zr, and Ce) catalysts, and the OER performance decreased gradually with the increase in Mo content, suggesting the key role of element selection and contents (figs. S19 and S20) despite the turnover frequencies or mass activity (fig. S21). The superior performance of Mo is attributed to its stable +6 oxidation state, which optimally modulates the electronic structure of Ru by lowering its oxidation state and enhancing stability, a synergistic effect not achieved by other dopants under acidic conditions.

Favorable stability is equally important for OER catalysts (20). The OER durability of the catalysts was evaluated by chronopotentiometry tests in a 0.1 M HClO₄ electrolyte at a 10 mA cm⁻² current density (Fig. 2E); Com RuO₂ showed better durability compared with the homemade RuO₂, but its performance still degraded rapidly and lasted no longer than 30 hours. Intriguingly, AC-MoRuO_x presented a negligible potential loss and Ru dissolution after 3000 hours at the current density of 10 mA cm⁻² and was stable during 1000 hours at 200 mA cm⁻² (Fig. 2E and figs. S22 and 23), which was significantly better than C-MoRuO_x (180 hours) and AC-RuO₂ (220 hours) (figs. S24 and 25). Moreover, the stability number (S) was 1.11 × 10⁶ at 3000 hours for AC-MoRuO_x (table S3), which was comparable with rutile IrO₂ (~10⁶; Alfa-Aesar) (30) and other reported IrO₂ (~10⁵; Sigma-Aldrich) (31), further demonstrating the much better durability of AC-MoRuO_x with the heterophase and atomic arrangements under acidic OER conditions. The synergistic effect is unambiguously demonstrated by direct comparison with our control samples (figs. S24 and S25). The crystalline C-MoRuO_x and the heterophase catalyst without Mo doping (AC-RuO_x) each showed significantly inferior stability compared to AC-MoRuO_x. This critical evidence proves that neither Mo doping nor the heterophase structure alone is sufficient; their combination is essential to achieve breakthrough durability. These results were also confirmed by the cyclic voltammetry (CV) tests (fig. S26). The long-time durability of AC-MoRuO_x was superior to that of the state-of-the-art noble metal–based acidic OER catalysts that have been reported (Fig. 2F and table S4). In addition, the structure and properties of AC-MoRuO_x were well preserved after the catalytic reaction (figs. S27 and S28). To elucidate the stability of the Ru sites during the OER process, the oxidation behavior of the Ru sites at high overpotentials was analyzed by CV measurements (fig. S29). Compared with A-MoRuO_x, C-MoRuO_x, homemade RuO₂, Com RuO₂, and the controls, the Ru species (Ru³⁺/⁴⁺) in AC-MoRuO_x were more dominant than Ru⁴⁺/Ru⁶⁺, while the redox peak (1.33 V) of metastable dissolution Ru⁶⁺/Ru⁸⁺ disappeared (23), indicating that the overoxidation of the Ru sites to a highly oxidized state was suppressed during the OER process after obtaining heterophase and atomic arrangements. Therefore, this unique electronic state resulting from the disordered phase increase can lower the OER potential barrier and inhibit the overoxidation behavior of the Ru site during the OER process, thus improving the OER activity and stability. Together, the heterophase and atomic arrangements due to Mo doping synergistically optimized the stability of the RuO₂ catalyst, resulting in AC-MoRuO_x exhibiting excellent acidic OER durability.

Correlation between the local electronic structure and OER performance

For insight into the source of activity and stability, we performed a series of characterizations of AC-MoRuO_x and corresponding control catalysts to elucidate the effect of heterophase and atomic arrangements on the microstructural evolution and electronic structure variation of AC-MoRuO_x. From XPS results, the Ru 3d peaks of AC-MoRuO_x were slightly shifted, the binding energies decreased by 0.3 eV compared to the homemade RuO₂ and crystalline C-MoRuO_x, and the binding energy decreased by 0.4 eV compared to the amorphous A-MoRuO_x, which suggested that the oxidation state for the Ru sites in AC-MoRuO_x was obviously reduced with the disordered phase appropriately increasing (fig. S30). XPS analysis of the Ru 3d core level provides critical insight into the oxidation state evolution (figs. S31 and S32). For AC-MoRuO_x, the binding energy after OER testing increased slightly by 0.2 eV but remained lower than that of the pristine RuO₂ benchmark. This indicates that the Ru sites in AC-MoRuO_x were effectively maintained in a relatively stable, lower average oxidation state, resisting overoxidation. In stark contrast, for C-MoRuO_x, a 0.15-eV positive shift resulted in a final binding energy that exceeded that of pristine RuO₂, signifying that Ru was driven to a less stable, higher oxidation state, consistent with its inferior durability. Furthermore, as shown in fig. S33, the Ru 3p peak position and Ru³⁺/Ru⁴⁺ ratio of AC-MoRuO_x before and after OER stability measurements remained almost unchanged compared to the homemade RuO₂ and crystalline C-MoRuO_x, which was attributed to the effective inhibition of peroxidation of the Ru site by increasing the heterophase and atomic arrangements. The Mo 3d XPS spectra of AC-MoRuO_x reveal peaks corresponding to Mo 3d_3/2 at 230.9 eV and Mo 3d_5/2 at 234.0 eV, with peak positions situated between those of Mo⁴⁺ and Mo⁶⁺ (fig. S31). This indicates that electron transfer from molybdenum to neighboring ruthenium atoms occurs, which helps to stabilize the Ru valence state under OER conditions. Consequently, this electron transfer mitigates the dissolution and overoxidation of ruthenium. In addition, the Mo 3d XPS spectrum shows that the Mo⁶⁺ content gradually increases with the increase in the Mo:Ru ratio in AC-MoRuO_x (fig. S35). Notably, the Mo:Ru ratio determined by XPS was consistently higher than that obtained from inductively coupled plasma mass spectrometry (ICP-MS), indicating the surface enrichment of Mo species, which likely contributes to modulation of the local Ru─O coordination and stabilization of the catalyst under OER conditions (tables S5 and S6). To provide a more complete assessment, ICP-MS analyses were performed for both the anode and cathode compartments, and the counter electrode was digested to quantify possible Ru redeposition, ensuring that the reported dissolution values reflect total Ru loss across the entire cell (table S7). In addition, the peaks in the O 1s XPS spectra could be deconvoluted into the lattice oxygen (O_L─Ru), hydroxyl (OH), and adsorbed oxygen (O_ad) (fig. S36) (32, 33). We found that the peak of O_L─Ru was shifted from 529.15 to 529.25 eV since the increase in heterophase and atomic arrangements, which means that more electrons are taken away from the oxygen atoms by the neighboring metal atoms in AC-MoRuO_x, decreasing the covalency of Ru─O bonds (23, 28). The soft x-ray absorption spectra at the O K-edge were collected to further interpret the structural evolution. The relative spectral weight of AC-MoRuO_x decreased and shifted to high energy compared with the homemade RuO₂, suggesting weak covalence with O 2p as a result of incorporating Mo atoms (fig. S26), which agreed with those of O 1s XPS results (34). Besides, the peak position of O_L─Ru remains almost unchanged after OER stability tests (fig. S37).

To precisely analyze the electronic and atomic structure analysis of AC-MoRuO_x, we measured synchrotron x-ray absorption spectra. From Fig. 3A and fig. S38, we can find that the Ru K-edge absorption position of AC-MoRuO_x was located between Ru foil and Com RuO₂ and lower than RuO₂, while the Mo K-edge absorption position was located between MoO₂ and MoO₃ and closer to MoO₃ (Fig. 3B and fig. S39). The Fourier-transformed extended x-ray absorption fine structure (FT-EXAFS) (Fig. 3C) of the Ru K-edge suggested that the bond length of Ru─O in AC-MoRuO_x was slightly stretched to 1.50 Å compared with Com RuO₂ (1.47 Å) because of the existence of a lower Ru^x<+4 valence state (29). In addition, the EXAFS spectra of the Mo K-edge in AC-MoRuO_x exhibited an oxidation behavior of Mo sites compared with that in MoO₃, causing the bond length of Mo─O to be stretched (fig. S40). In addition, the wavelet-transformed EXAFS Mo K-edge in AC-MoRuO_x showed that the oxidation behavior of Mo sites caused the bond length of Mo─O to be stretched, further demonstrating the weakening of the Ru─O bond covalency, increasing the electron density near Ru sites (fig. S41). In addition, we later fitted the EXAFS spectra by screening multiple structures to obtain the structural parameters around Ru and Mo atoms (figs. S42 to S47 and tables S8 and S9); the increase in the Ru─O interatomic distance and the decrease in the coordination number together indicate that the covalency of the Ru─O bond is optimized and the formation of Ru─O─Mo is in favor of the long-term OER stability (Fig. 3D). Therefore, we have a reason to believe that this stability stems from synergistic effects: The Mo doping electronically suppresses Ru overoxidation by modulating the Ru─O covalency, while the heterophase interface structurally accommodates the reaction-induced stress. The Mo species play a pivotal role not only as an electronic modulator but also in stabilizing the heterophase boundary itself, as evidenced by the stable Ru─O─Mo units under operating conditions (Fig. 3G). Furthermore, the wavelet transform for Ru K-edge EXAFS (fig. S48) was applied to indicate the variables of Ru─O─Ru/Mo and Ru─O bonds in AC-MoRuO_x. Two distinct intensities at 9.32 and 5.82 Å⁻¹ of wavelet-transformed EXAFS Ru K-edge, assigned to Ru─O─Ru/Mo and Ru─O, respectively, were attributed to the strong interaction between Mo and Ru sites (10). The above comprehensive conclusions provide qualitative evidence for the existence of the Ru─O─Mo structural unit. Furthermore, we attempted a semiquantitative contribution analysis of Ru─O─Mo, demonstrating that Ru─O─Mo constitutes a locally obvious structural feature with a substantial contribution within the AC-MoRuO_x catalyst (table S10). It is precisely this appreciable proportion of Ru─O─Mo units that plays a pivotal role in modulating the electronic structure and enhancing stability.

Fig. 3. Electronic structure characterizations of the AC-MoRuO_x catalyst before and after acidic OER.

(A) Ru K-edge XANES spectra of AC-MoRuO_x, Ru foil, and RuO₂. The inset is the extended image of the selected area. (B) Mo K-edge XANES spectra of AC-MoRuO_x, Mo foil, MoO₂, and MoO₃. The inset is the extended image of the selected area. (C) Ru K-edge FT-EXAFS in R spaces of AC-MoRuO_x, Ru foil, and fresh RuO₂. (D) Ru K-edge EXAFS fitting analyses for AC-MoRuO_x in the R space. (E) In situ XANES spectra of the Ru K-edge for AC-MoRuO_x at different voltages. The inset is the magnified image. (F) Ru valence states of AC-MoRuO_x fitted by Ru K-edge absorption energy at different potentials. (G) Ru K-edge FT-EXAFS 2D spectra of AC-MoRuO_x at different potentials.

The in situ x-ray absorption near-edge structure (XANES) was performed to further investigate the evolution of the electronic structure of metal sites during OER qualitatively and quantitatively to obtain additional direct evidence of the superior stability of AC-MoRuO_x. Ru K-edge and Mo K-edge x-ray absorption spectroscopy (XAS) data were recorded under test conditions where the open circuit potential (OCP) went to 0.4 and 1.0 V and then jumped to 1.8 V at intervals of 0.2 V versus reversible hydrogen electrode (RHE), and the final potential returns to the OCP. The relative absorption edge positions in the Ru K-edge shifted to higher energy, showing discontinuous changes with the potential raised (Fig. 3E and fig. S49). Specific valence fitting results indicate that switching the applied potential from the OCP to 1.4 V versus RHE causes the oxidation state of Ru in AC-MoRuO_x to increase from +3.0 to +3.1 slowly and the oxidation states steadily near +3.1 as the potential is further raised to 1.8 V versus RHE (Fig. 3F). The stable Ru sites at high potential prevent overoxidation and dissolution of the catalyst, thereby maintaining stable catalytic activity. However, the Mo K-edge positions show continuous changes as the voltage increases (fig. S50), corresponding to the increasing valence state of Mo. In addition, an unnoticeable increase in the valence state of Mo at large potentials and the initial valence state upon return to the OCP were observed. These findings are consistent with CV results (fig. S29, A and B). Consequently, the oxidation state dynamics of Mo maintained the stability of the oxidation state of Ru, thus improving the stability of the catalyst. It was further demonstrated that the increase in heterophase and atomic arrangements overcame the limitations of the stability-activity trade-off of Ru-based catalysts by regulating the proton and electron transfer processes during OER.

The FT-EXAFS curve-fitting analysis was carried out to provide more insight into the evolution of the local coordination structure of Ru active sites under the applied potentials (Fig. 3G and figs. S50 to S53). The Ru─O bond distance at AC-MoRuO_x is stretched from 1.89 to 1.90 Å as the OCP increased up to 1.8 V (Fig. 3G, figs. S50 to S53, and table S11), which could effectively adjust the electronic structure of RuO₂, causing a weak covalency of Ru─O bonds. In situ XAS confirmed that the asymmetric Ru─O─Mo structural units remain stable during the reaction, ensuring the catalyst’s prolonged operation. In summary, the heterophase and atomic arrangements inhibit the maintenance of Ru in AC-MoRuO_x in a low valence state during the OER process while maintaining the vibrational balance of the Ru─O bond lengths, which enables the catalyst to exhibit extremely excellent stability.

We then carried out in situ Raman spectroscopy to further clarify the structure evolution during OER. For both AC-MoRuO_x and homemade RuO₂, two major peaks at ~524 and 631 cm⁻¹ assignable to E_g and A_1g vibration modes, respectively, were observed (fig. S54) (35, 36). With the potential increases from OCP to 1.6 V versus RHE, the AC-MoRuO_x sample maintains a constant Raman shift, suggesting the consistency in the Ru─O bonding structure during OER (37). However, a positive shift (~6 cm⁻¹) in RuO₂ was noticed, implying the shrinkage in Ru─O bonding length during OER. Therefore, the increase in heterophase and atomic arrangements kept the Ru species in a low valence state, at which time the Mo atoms and the AC boundaries served as electron-feedback centers, which fed back the electrons to the Ru sites in time to avoid their overoxidation under the OER conditions (16, 22).

More insights into the catalytic mechanism

We further performed first-principle density functional theory (DFT) calculations to understand the influence of Mo incorporation on the activity and stability of RuO₂ under acidic conditions. Considering the Mo adatoms doped in RuO₂, the resultant AC-MoRuO_x demonstrated highly disordered phase and atomic arrangements (Fig. 4A), displaying a nonstoichiometric and crystal/amorphous configuration after the thermal process. We divided the complex process in AC-MoRuO_x into two steps, successively showing the increase in heterophase and atomic arrangements. These steps improve catalytic activity from the electronical (mechanism 1) and structural (mechanism 2) standpoints. Centering on uncoordinated sites for catalyzing OER, we explored three potential doped sites to accommodate adatoms (Fig. 4B) and constructed a series of C-TMRuO_x (TM denotes the commonly used transitional metals).

Fig. 4. Mechanism investigations of AC-MoRuO_x.

(A) Illustration of catalytic activity increases with the increase in the disordered phase. The black arrows represent the enhanced role of adatom or configuration change. (B) Top view of the possible doped site of transition metals and uncoordinated site (US) for catalyzing OER. The red and grayish-white spheres represent the O and Ru atoms, respectively. (C) 2D volcano of OER activity against the adsorption of *OOH (ΔG_*OOH) and *OH (ΔG_*OH). The black line is the scaling relation between ΔG_*OOH and ΔG_*OH, namely, ΔG_*OOH = 0.71ΔG_*OH + 2.89, R² = 0.58. The yellow, blue, and black fonts label the doped A, B, and C sites, respectively. A-TM, B-TM, C-TM, and A(TM) represent the Ru atoms adjacent to TM atoms in A, B, and C sites and the TM atom in the doped A site, respectively. The red font highlights the catalyst with the highest OER activity. (D) Bader charge of Ru sites adjacent to doped C sites. (E) DEMS signals of ³²O₂ (¹⁶O + ¹⁶O) and³⁴O₂ (¹⁶O + ¹⁸O) collected during the CV cycle. (F) Ratio of ³⁴O₂:³²O₂, which remained unchanged with the increase in potential, suggesting that ³⁴O₂ was derived from background ¹⁸O rather than from the lattice. (G) Free energy profile of OER on C-RuO₂(110) and C/A-RuO₂(110) with and without Mo adatoms. The inset is a comparison of activity (namely, overpotential) in C-RuO₂(110) and C/A-RuO₂(110).

First, we investigated the elemental dopants and calculated the energy of TM species in B/C sites relative to the A site (fig. S55). Most 3d adatoms prefer to reside at doped A site, while 4, 5d metals tend to dope into C sites. It validated that 3d metal atoms at A sites not only tailor the adsorption of the adjacent Ru atom but also possibly serve as the catalytic sites. We calculated the OER process on Ru/TM sites (fig. S56). Notably, the 4, 5d metal atoms, including Nb, Mo, and W, in C sites primarily regulate the catalytic sites. The two-dimensional (2D) volcano curve (Fig. 4C and fig. S57) quantitatively showed the OER activity (overpotential) of all potential sites. Along with the scaling relation ΔG_*OOH = 0.71ΔG_*OH + 2.89, the highly active sites are mainly the Ru atoms close to C sites, and TM atoms in A sites exhibit high OER activity instead of Ru atoms in the vicinity of B and C sites. Among them, Ru sites adjacent to the Mo atom in C sites (C-Mo) displayed the highest OER activity. This prediction indicates that introducing Mo could enhance catalytic activity.

To deeply exploit the intrinsic nature, DFT calculations were performed to study the adsorption behavior of related intermediates (*OOH, *OH, and *O) on such active sites. Two types of scaling relations are present on the Ru atom close to the doped A, B, and C sites and the adatom in the C site (fig. S58). In the A(TM) site, the great difference in the slope of the scaling relation between ΔG_*OOH, ΔG_*O, and ΔG_*OH states that the rate-determining step is limited to the adsorption energy of *O relative to *OH (ΔG_*O − ΔG_*OH). Comparably, other sites exhibit similar scaling relations between ΔG_*OOH, ΔG_*O, and ΔG_*OH, implying that the rate-determining step is restrained in the adsorption energy of *OOH relative to *OH (ΔG_*OOH − ΔG_*OH). Consequently, two-type volcanoes against (ΔG_*O − ΔG_*OH) and (ΔG_*OOH − ΔG_*OH) were built to evaluate the catalytic activity of adatom and Ru sites, respectively. Besides, the tailoring of C-TM on the Ru site can be ascribed to altering of the *OOH adsorption behavior (fig. S59) via the H─O hydrogen bond for *OOH dehydrogenation, while *OH and *O adsorption remains unchanged. Therefore, C-TM provided practical tactics to improve OER activity.

Moreover, we also calculated the charge state of Ru adjacent to varying TM species in the doped C site using Bader charge (Fig. 4D). Compared to pristine Ru, most doped TM species, especially 3d metal atoms, enable the upshift of Ru charge state toward higher charge. This increase suggests that the Ru site at a high charge state proceed with OER, potentially leading to acceleration of the dissolution of Ru. C-Mo significantly decreased the Ru charge state, guaranteeing high stability as a result of the OER process on a low Ru charge state, which effectively avoided the transition of a low-charge-state Ru site to a high-charge-state Ru site, suggesting higher stability.

To experimentally confirm this adsorbate evolution mechanism (AEM) pathway, we conducted operando differential electrochemical MS (DEMS) measurements via the isotope ¹⁸O-labeling method (Fig. 4, E and F, and figs. S60 and S61). First, AC-MoRuO_x was labeled with ¹⁸O by performing OER under potentiostatic control (Materials and Methods) to study whether lattice oxygen was exchanged by ¹⁸O. We then measured the isotope signal of evolved O₂ during OER in H₂¹⁶O (0.1 M HClO₄). With a gradual increase in OER potential, a corresponding increase in both ³²O₂ and ³⁴O₂ signals was detected (Fig. 4E). The signal ratio of ³⁴O₂:³²O₂ within the OER potential range remained relatively constant and was comparable to that of the natural abundance of ¹⁸O in deionized water within experimental uncertainty (Fig. 4F) (32, 38, 39), suggesting that the lattice oxygen oxidation mechanism did not occur over AC-MoRuO_x. The DEMS analysis, coupled with CV measurement, further confirmed this conclusion (fig. S60), in which the constant ³²O₂ and ³⁴O₂ signals and ³⁴O₂:³²O₂ ratio were clearly observed. Our DEMS results were consistent with a previous report that also suggested the exclusivity of AEM over RuO₂-based OER catalysts (40).

On the basis of the crystalline C-MoRuO_x, we further considered the increase in heterophase and built a mixed structure of AC-MoRuO_x using the method of Parrinello and Rahmans. We calculated the free energy change of OER on Ru sites (Fig. 4G and fig. S62). The amorphous A-MoRuO_x exhibited inferior catalytic performance with a high overpotential of 1.23 V, much higher than homemade pristine RuO₂ (0.63 V). Once introducing amorphous A-MoRuO_x, AC-MoRuO_x showed high catalytic activity (0.50 V), exceeding pure crystalline C-MoRuO_x (0.61 V). The increase in heterophase can provide new approaches to further tailor the OER activity. In addition to redox-induced solvation at high overpotential, we further consider the potential peroxidation of the catalytic site during redox, which may lead to active site degradation. With increasing potential, this degradation can also lead to rapid catalyst dissolution collapse. To address this issue, we introduced the concept of a mechanism for separating charge accumulation and intermediate adsorption processes, which can break the traditional OER behavior of charged sites directly involved in catalysis. On the basis of this mechanism, we calculated the formation energies of C-RuO₂, AC-RuO₂, and AC-RuMoO_x. Our results show that the introduction of the amorphous phase greatly improves stability (fig. S63). In addition, the accompanying elemental dopants further improve stability (fig. S64). These findings highlight the critical role of heterophase and atomic arrangements in improving catalyst durability.

To further prove the AEM for the generation of oxygen molecules, we further performed in situ attenuated total reflection infrared (ATR-IR) spectroelectrochemical measurements to capture the OOH intermediate (fig. S65). The recorded peak at ~1161 cm⁻¹ is assigned to the characteristic vibration adsorption of the surface-adsorbed superoxide (*OOH) (41, 42). This indicates that the AEM rather than the lattice oxygen oxidation mechanism dominated O₂ generation over AC-MoRuO_x, which certainly improved its stability under acidic conditions.

PEMWE device performance

The activity under practical operation conditions was one of the most important indicators of OER catalysts. Given that, we prepared membrane electrodes using AC-MoRuO_x and control catalysts as anodes, commercial Pt/C as a cathode, and Nafion 117 membrane as a separator for acidic PEM electrolyzers (Fig. 5A and figs. S66 to S68). The current-voltage (I-V) curves collected in Fig. 5B clearly showed that the AC-MoRuO_x‖Pt/C electrolyzer had higher water electrolysis activity compared to the Com RuO₂‖Pt/C combination. Specifically, the AC-MoRuO_x–based electrolyzer (at 80°C) required only 1.57 and 1.65 V to reach current densities of 1000 and 2000 mA cm⁻², respectively, lower than that of the Com RuO₂ electrolyzer voltage (Fig. 5B). It should be noted that the relatively high ohmic resistance of ~0.3 Ω·cm² arises from the use of a thick Nafion 117 membrane and the catalyst-coated substrate fabrication method. Nevertheless, the fact that AC-MoRuO_x sustains excellent durability under these deliberately conservative, nonoptimized conditions highlights the intrinsic structural resilience of the catalyst. Future work with optimized membrane electrode assembly (MEA) architectures (e.g., catalyst-coated membrane fabrication and thinner, high-conductivity membranes) is expected to further reduce ohmic losses and unlock the full performance potential of AC-MoRuO_x. This did not affect our evaluation results of the intrinsic activity and stability of AC-MoRuO_x using a standard three-electrode setup, much less the stability testing of AC-MoRuO_x under actual operating conditions. The cell potential was decoupled to provide different contributions, including kinetic overpotential, ohmic potential, and mass transfer ohmic potential, in which a low kinetic overpotential indicates the high OER kinetics (fig. S69). For industry, the stability of the OER catalysts was far more important than even their activity. To measure the ability of our prepared AC-MoRuO_x to work for a long time in a PEM electrolyzer close to the practical PEM electrolyzer, we set the electrolyzer for continuous operation at current densities of 200, 1000, and even up to 1500 mA cm⁻² (never achieved with Ru-based acidic OER catalysts) (Fig. 5, C to E). Compared with Com RuO₂, we found that the constructed AC-MoRuO_x–based electrolyzer operated continuously for 2000 hours at current densities of 200 and 1000 mA cm⁻² and 1000 hours at a current density of 1500 mA cm⁻² without obvious performance degradation and even showed a negligible amount of ruthenium dissolution (fig. S70), meaning the great application potential of AC-MoRuO_x, representing one of the best records among the reported state-of-the-art catalysts (Fig. 5F and table S12). However, the performance of the Com RuO₂–based MEA electrolyzer decayed significantly after less than 100 hours of operation at a 200 mA cm⁻² current density. Similar stability was also observed in pure water as the electrolyte (fig. S71). According to the calculation from the US Department of Energy (43), the estimated cost of producing 1 kg of hydrogen from the AC-MoRuO_x–based PEM electrolyzer was approximately $0.91, well below the US Department of Energy’s target of $2 per kilogram of hydrogen. The above results clearly confirm that AC-MoRuO_x has great potential for application in practical PEMWEs.

Fig. 5. PEMWE device performance using AC-MoRuO_x as an acidic OER catalyst in a 0.1 M HClO₄ electrolyte at 80°C.

(A) Schematic diagram of the PEM electrolyzer, a typical single PEMWE setup comprising a MEA and bipolar plates (BPs) with a flow field presented, wherein the MEA comprises gas diffusion layers with a Ti felt and a carbon paper (CP) at the anodic and cathodic sides, respectively; anodic and cathodic catalyst layers; and a PEM. (B) Polarization curves of the PEM electrolyzers using AC-MoRuO_x or Com RuO₂ as the anode catalyst and Com Pt/C as the cathodic catalyst at 80°C. (C to E) Chronopotentiometry curve of the PEM electrolyzer using AC-MoRuO_x and Com RuO₂ nanocatalysts operated at 200 mA cm⁻² (C), 1000 mA cm⁻² (D), and 1500 mA cm⁻² (E). The small fluctuations in cell voltage might be caused by the generation of bubbles that could not dissipate from the surface of MEAs. (F) Performance comparison of the PEM electrolyzer activity and stability of AC-MoRuO_x with those of previously reported state-of-the-art catalysts.

DISCUSSION

In conclusion, we have developed a non-Ir AC-MoRuO_x acidic OER catalyst that uses a previously unknown concept of synergistic enhancement of heterophase and atomic arrangements, with an overpotential of only 180 mV at a 10 mA cm⁻² current density and continuous operation for at least 4.5 months at this current density (3000 hours, 100 times that of Com RuO₂). In practice, PEMWE devices with AC-MoRuO_x as the anode can operate stably for at least 2000 hours (more than 2 months) at 200 and 1000 mA cm⁻² current densities, and even at current densities as high as 1500 mA cm⁻², and still operate stably for at least 1000 hours (more than 1.5 months) with negligible cell voltage loss. The discovery of AC-MoRuO_x is expected to provide a feasible strategy for the development of Ru-based acidic OER catalysts with both high activity and stability, which will greatly contribute to the promotion of low-cost non–Ir-based PEM electrolyzers.

MATERIALS AND METHODS

Preparation of MoRuO_x

In a typical procedure, 0.018 mmol of ruthenium(III) chloride hydrate (RuCl₃·H₂O), 0.002 mmol of molybdenyl(VI) oxide bis(2,4-pentanedionate) [Mo(acac)₂], and 120 mg poly(vinyl pyrrolidone) were mixed in a glass bottle containing 10 ml of ethanol and stirred to obtain a clear solution. Then, the bottle was sealed and stirred at 60 rpm until the ethanol evaporated. Next, the obtained precursors were annealed in air at 300° to 500°C (400°C as the optimized temperature) for 2 hours. Homemade RuO₂ and Mo_δRu_10−δO_x (δ = 1 to 5) were prepared by the same synthetic procedure with different Ru/Mo feeding ratios.

Characterization

The microstructure and morphology of the prepared materials were observed by transmission electron microscopy (TEM; HT7800) and HRTEM [FEI-Themis Z TEM/STEM operated at 300 kV and equipped with double-spherical aberration (Cs) correctors and Themis Z TEM/STEM equipped with high-angle annular dark-field and annular bright-field detectors]. The high-angle annular dark-field images were acquired using the Themis Z with a 59- to 100-mrad inner-detector angle. The attainable resolution of the probe defined by the objective prefield is 60 pm. A Super X Windowless x-ray detector was used to collect the energy dispersive spectra. XPS was performed on a KULVAC PHI Quantera x-ray photoelectron spectrometer and analyzed by Avantage software. Powder XRD patterns were obtained by a Bruker D8 Advances using Cu Kα radiation (λ = 1.5406 Å). The x-ray absorption fine structure spectra (Ru K-edge and Mo K-edge) were collected at the 1W1B station in Beijing Synchrotron Radiation Facility (BSRF) and at the BL14W1 station in Shanghai Synchrotron Radiation Facility (SSRF). The storage rings of BSRF were operated at 2.5 GeV with an average current of 250 mA. Using the Si(111) double-crystal monochromator, data collection was carried out in S4 transmission/fluorescence mode using an ionization chamber. All spectra were collected under ambient conditions.

Details regarding the XAS experiments and data analysis methods

In situ XAS, including XANES and EXAFS at the Ru K-edge and Mo K-edge, was performed in total fluorescence yield mode at ambient air at the BL14W1 station in SSRF. The measurement in a typical three-electrode setup, the same condition as the electrochemical characterization case, was performed in a specially designed Teflon container with a window sealed by Kempton tape. X-ray was allowed to be transmitted through the tape and electrolyte, so the XAS signal would be collected in total fluorescence yield mode at the BL14W1 station in SSRF. Subtracting the baseline of pre-edge and normalizing that of postedge obtained the spectra. EXAFS analysis was conducted using the Fourier transform on k³-weighted EXAFS oscillations. All EXAFS spectra are presented without phase correction. The obtained XAFS data were processed in Athena (version 0.9.25) for background, pre-edge line, and postedge line calibrations. Then, FT fitting was carried out in Artemis (version 0.9.25). The k³ weighting, k range of 3 to 11.5 Å⁻¹, and R range of 1.0 to 4.0 Å were used for the fitting. The four parameters, coordination number, bond length, Debye-Waller factor, and E₀ shift (N, R, DW, and ΔE₀, respectively), were fitted without anything being fixed, and $S_0^2$ (0.9) was set. For wavelet transform analysis, the χ(k) exported from Athena was imported into the Hama Fortran code. The parameters were listed as follows: R range, 1.0 to 3.5 Å; k range, 0 to 12.4 Å⁻¹ for samples; k weight, 3; Morlet function with κ = 10. σ = 1 was used as the mother wavelet to provide the overall distribution.

Calculation of oxidation states from XANES

First, background deduction and normalization of χμ(E) data were executed before E₀ calibration with foil samples as standard references. Then, E₀ values of each sample studied were determined by their first derivative vertex (second peak for the Ru element). Taking Ru foil and RuO₂ as references, the oxidation states of the samples with different E₀ values can be obtained through linear fitting.

In situ Raman spectra

Raman spectra were recorded with an Xplora confocal microprobe Raman system (HORIBA JobinYvon). A 50×-long-working-distance (8 mm) objective was used. The wavelength of the Ar-ion excitation laser was 532 nm (the power was about 6 mW). Raman frequencies were calibrated using a Si wafer. Each Raman spectrum shown here was acquired over a collection time of 240 s and is the average of three measurements.

In situ ATR-SEIRAS measurement

The in situ ATR-SEIRAS (ATR surface-enhanced infrared absorption spectroscopy) monitoring process was carried out using a Nicolet 6700 Fourier transform infrared spectrometer armed with a mercury cadmium telluride detector and a PIKE VeeMAX III ATR accessory. A single cell with 25 ml of 0.1 M HClO₄ electrolyte was used as the reactor. The working electrode was glass carbon. Ag/AgCl was used as the reference electrode, while a Pt wire was used as the counter electrode. Before starting the tests, the device was purged with ultrapure water, and the electrolyte was treated for 30 min. Then, N₂ was introduced into the electrolyte for 30 min to achieve N₂ saturation. The background spectrum was collected with no potential applied, and spectra were recorded during the electrolysis procedure under different potentials. A 4-cm⁻¹ resolution and 32 scans were applied in the collection of all the spectra. Potentials were corrected with respect to an RHE.

In situ XAS measurement

In situ XAS, including XANES and EXAFS at the Ru K-edge and Mo K-edge, was performed in total fluorescence yield mode at ambient air at the BL14W1 station in SSRF. The measurement in a typical three-electrode setup, the same condition as the electrochemical characterization case, was performed in a specially designed Teflon container with a window sealed by Kempton tape. X-ray was allowed to be transmitted through the tape and electrolyte, so the XAS signal would be collected in total fluorescence yield mode at the BL14W1 station in SSRF. The Ru K-edge and Mo K-edge of AC-MoRuO_x were polarized at different anodic potentials in this system, respectively. After the corresponding current density reaches a steady state, the XAFS data were collected at the indicated potentials. Subtracting the baseline of pre-edge and normalizing that of postedge obtained the spectra. EXAFS analysis was conducted using the Fourier transform on k³-weighted EXAFS oscillations. All EXAFS spectra are presented without phase correction. The obtained XAFS data were processed in Athena (version 0.9.25) for background, pre-edge line, and postedge line calibrations. Then, FT fitting was carried out in Artemis (version 0.9.25).

Electrochemical measurements

Electrochemical measurements were conducted on a CHI 760E electrochemical workstation (Shanghai Chenhua Instrument Corporation, China) in a conventional three-electrode cell. The graphite rod electrode was the counter electrode, a Ag/AgCl (saturated KCl solution) or saturated calomel electrode (SCE; Hg/Hg₂Cl₂) was the reference electrode, and the glassy carbon electrode (diameter: 5 mm; area: 0.1963 cm²) was the working electrode. Conversion from the SCE reference electrode to the RHE was done according to E (versus RHE) = E (versus SCE) + 0.0591 × pH + 0.244. The working electrode was prepared as follows: 2.5 mg of catalyst was dispersed in the mixture of ethanol (0.4 ml) and Nafion (5 wt %, 30 μl) under ultrasonication for 1 hour, and a catalyst with the concentration of 5.9 mg ml⁻¹ was obtained. Ten microliters of the catalyst was dropped onto the glassy carbon electrode surface for further electrochemical tests. All the potentials reported in this work were converted to the RHE. All measured potentials in rotating disc electrode tests were 95% iR compensated, unless otherwise specified. Linear sweep voltammetry tests were recorded at a scan rate of 5 mV s⁻¹. Stability was examined by chronopotentiometry testing at 10 and 200 mA cm⁻² in 0.1 M HClO₄. CV was performed in a N₂-saturated 0.1 M HClO₄ solution from 1.3 to 1.5 V (RHE) at a scan rate of 50 mV s⁻¹. EIS measurements were measured at 50 mV versus RHE in the frequency range from 10 kHz to 0.01 Hz in a N₂-saturated 0.1 M HClO₄ solution. ECSA was estimated from the electrochemical C_dl from Eq. 1 (21)

$$\text{ECSA}=\frac{C_{\text{dl}}}{C_{\mathrm{s}}}$$ (1)

The value for C_dl was determined via CV scanning in a non-Faradic region from 1.18 to 1.28 V versus RHE with scanning rates of 2.5, 5.0, 7.5, 10, 15, and 20 mV s⁻¹. The value of the slope was determined from fitting of the data to obtain C_dl. The specific current density per ECSA was computed by normalizing the current by ECSA, and a general specific capacitance of Cs = 0.035 mF cm⁻² was used on the basis of typical reported values (32).

The stability number (S) was computed from Eq. 2 (30)

$$S=\frac{n_{{\mathrm{O}}_2}}{n_{\text{Ru}}}$$ (2)

Electrochemical measurements in the PEM electrolyzer

For electrolyzer tests, a self-made cell was used as the PEMWE device and a cation exchange membrane (Nafion 117) as the membrane electrolyte. In our initial trial, 20 wt % commercial Pt/C (around 0.5 mg cm⁻²) and OER catalyst (~2 mg cm⁻²) with a 20 wt % polytetrafluoroethylene (PTFE) binder were sprayed onto two sheets of PTFE (surface area, 2 cm by 2 cm). Subsequently, the PTFE-supported Pt/C cathode catalyst, pretreated Nafion 117 membrane, and PTFE-supported AC-MoRuO_x anode catalysts were hot pressed together at 100°C for 5 min under a low pressure of 5 MPa. Afterward, the MEA was obtained by the catalyst-supported Nafion 117 membrane in the middle of the CP and Pt-plated Ti foam gas diffusion layers before a hot-press process with the same protocols, which was placed in the PEMWE and then circulated with a 0.1 M HClO₄ aqueous solution at 400 ml min⁻¹. The polarization curves were collected from 50 to 2500 mA cm⁻² at 80°C and ambient pressure. The stability test was carried out at 200, 1000, and 1500 mA cm⁻² at 80°C and ambient pressure.

ICP-MS analyses of Ru dissolution

The dissolution of catalysts during the OER process at 250 mA cm⁻² in an electrolyte with a total volume of 200 ml was quantified by ICP-MS (Agilent 8900 ICP-QQQ). Two milliliters of the electrolyte was obtained from the anodic cell following electrolysis at 24, 72, 240, 480, 960, and about 3000 hours in an H-type cell. In the PEMWE, the anodic cell following electrolysis at 24, 48, 120, 288, 480, and about 1000 hours was collected for ICP-MS measurement.

DEMS tests

One milligram of the catalyst was mixed with 10 wt % Nafion in 0.5 ml of ethanol and then drop-cast onto the glassy carbon (0.5-mm diameter) working electrode. The electrode was dried at room temperature for at least 1 hour before testing. Measurement was performed on an HPR-40 DEMS (Hiden Analytical) coupled to the flow cell, a CHI 760E electrochemical workstation (Shanghai Chenhua Instrument Corporation, China), an OER catalyst modified working electrode, a Ag/AgCl reference electrode, and a Pt wire counter electrode. The electrolyte flow rate was controlled at 0.5 ml min⁻¹. All electrochemical results were recorded versus the reference electrode and converted to the RHE following the relationship E_RHE = E_Ag/AgCl + 0.197 + 0.0591 × pH. Catalysts were first labeled with ¹⁸O in a 0.1 M HClO₄ solution created with ¹⁸O-labeled water (97% ¹⁸O; Cambridge Isotope Laboratories) for 10 min at 1.41 V versus RHE. Electrodes and cells were then rinsed with ¹⁶O water for 20 min to remove any H₂¹⁸O residual and scanned in a 0.1 M HClO₄ solution of H₂¹⁶O at 0.5 mV s⁻¹ from 1.2 V versus RHE until the current density reached 60 mA cm⁻². An electron energy of 70 eV was used for ionization of all species, with an emission current of 450 μA. All mass-selected products [mass/charge ratio (m/z) = 32 and 34] were detected by a secondary electron multiplier with a detector voltage of 1200 V. Data were averaged every 10 points to yield a smoother MS signal, and m/z 34/32 signals were calculated with the fitting data. For DEMS measurements using the potentiostatic method, after labeling and rinsing (as mentioned above), the background was collected for 10 min before the application of 10 mA cm⁻² for 30 min to probe potential oxygen lattice exchange.

Estimation of hydrogen produced and cost in the PEMWE

The calculation of hydrogen cost in the PEMWE was based on refs. (43, 44). Specifically

$$\begin{array}{c}\text{Mass of} {\mathrm{H}}_2\hfill \\ =\frac{j\left(\text{current density}\right)\times A\left(\text{electrolyzer area}\right)\times t\left(\text{working time}\right)\times \text{Molar mass} {\mathrm{H}}_2}{2\times F}\hfill \\ =\frac{1.0 \mathrm{A} {\text{cm}}^{-2}\times 4 {\text{cm}}^2\times 2000\times 3600 \mathrm{s}\times 2 \mathrm{g} {\text{mol}}^{-1}}{2\times \mathrm{96,485} \mathrm{c} {\text{mol}}^{-1}}=298.5 \mathrm{g}\hfill \end{array}$$ (3)

$$\begin{array}{c}\text{Cost} \text{per} \text{kilogram of} {\mathrm{H}}_2=\text{Energy consumption}\times \text{Electricity bill}\hfill \\ =\frac{1.7 \mathrm{V}\times 4.0 \mathrm{A}\times 2000 \text{hours} }{0.298 \text{kg} {\mathrm{H}}_2}\times \frac{\text{USD}0.02}{\text{kWh}}=\frac{\text{USD}0.91}{\text{kg} {\mathrm{H}}_2}\hfill \end{array}$$ (4)

Computational details

DFT calculations were performed by using the ab initio simulation package (VASP) (45–47). The generalized gradient approximation in the Perdew-Burke-Ernzerhof functional was adopted to describe the electron exchange and correlation energy (48), and the frozen-core projector-augmented wave method with a cutoff energy of 500 eV was chosen to describe the interaction between core electrons and valence electrons (49). The long-range van der Waals interactions between atoms are finely described by the DFT-D3 correction method in Grimme’s scheme (50). The criteria of energy and force convergence are set to 1.0 × 10⁻⁵ eV per atom and 0.02 eV Å⁻¹, respectively, for geometry optimization. In addition, a Γ-centered Monkhorst-Pack k-point mesh grid of 3 by 3 by 1 was used for all structural optimizations (51). Bader charge was used to explore the charge change of Ru sites in RuO₂ and TM-doped RuO₂ (52).

The OER performance was explored via the four-electron pathway as follows

$${\mathrm{H}}_2\mathrm{O}\mathbf{+}\ast \to \ast \text{OH}+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$ (5)

$$\ast \text{OH}\to \ast \mathrm{O}+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$ (6)

$$\ast \mathrm{O}+{\mathrm{H}}_2\mathrm{O}\to \ast \text{OOH}+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$ (7)

$$\ast \text{OOH}\to \ast +{\mathrm{O}}_2+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$ (8)

where * represents an active site. OOH*, O*, and OH* are the active sites with OOH, O, and OH intermediate adsorption, respectively. The free energy of the intermediates is defined as

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\text{ΔZPE}-T\mathrm{\Delta }S+\mathrm{\Delta }{G}_{\text{pH}}+\mathrm{\Delta }{G}_{\mathrm{U}}$$ (9)

where ΔE is the reaction energy of each step, obtained from DFT calculations; ΔZEP is the change of zero-point energies in the reactions; TΔS is the entropy contribution at 300 K; and ΔG_pH is the correction of H⁺ concentration. In this work, we considered the acidic condition, meaning that ΔG_pH was taken as zero. ΔG_U is the influence of applied potential, defined as ΔG_U = −eU, where U is the potential at the electrode and e is the transferred charge. For the small difference between the vibrational frequencies of the adsorbents on the surface, ∆ZPE and TΔS were taken from the previous work (53). The overpotential ( $\mathrm{\eta }$ ) for OER is evaluated as

$$\mathrm{\eta }=\frac{\text{min}(\mathrm{\Delta }{G}_1,\mathrm{\Delta }{G}_2,\mathrm{\Delta }{G}_3,\mathrm{\Delta }{G}_4)}{e}-1.23$$ (10)

where ΔG_i (i = 1 to 4) is the free energy of Eqs. 5 to 8.

Adsorption energy

The adsorption energy (ΔE_ads) of the key OER intermediates, including *OOH, *O, and *OH, was calculated relative to H₂O and H₂ under conditions of T = 298.15 K, pH 0, and U = 0 V (versus SHE) according to the following equations

$$\mathrm{\Delta }{E}_{\ast \text{OOH}}=E_{\ast \text{OOH}}+3/2E_{\text{H2}}-E_{\ast }-2E_{\mathrm{H}2\mathrm{O}}$$ (11)

$$\mathrm{\Delta }{E}_{\ast \mathrm{O}}=E_{\ast \mathrm{O}}+E_{\text{H2}}-E_{\ast }-E_{\mathrm{H}2\mathrm{O}}$$ (12)

$$\mathrm{\Delta }{E}_{\ast \text{OH}}=E_{\ast \text{OH}}+1/2E_{\text{H2}}-E_{\ast }-E_{\mathrm{H}2\mathrm{O}}$$ (13)

where * represents the adsorption sites. The above ΔE_ads is defined as the reaction free energy of the following reactions

$${}^{\ast }+{\text{2H}}_2\mathrm{O}\to {}^{\ast }\text{OOH}+3/2{\mathrm{H}}_2$$ (14)

$${}^{\ast }+{\mathrm{H}}_2\mathrm{O}\to {}^{\ast }\mathrm{O}+{\mathrm{H}}_2$$ (15)

$${}^{\ast }+{\mathrm{H}}_2\mathrm{O}\to {}^{\ast }\text{OH}+1/2{\mathrm{H}}_2$$ (16)

Models

The RuO₂ slab was modeled from the RuO₂ cell (lattice volume: 3.12 Å by 4.52 Å by 4.52 Å). To construct the crystal RuO₂ phase, we adopted three-layer (3 × 2) RuO₂(110) slabs, in which the upper two layers were relaxed, and the bottom layer was fixed, which contains 72 O and 36 Ru atoms within these supercells (lattice: 9.35 Å by 12.72 Å by 26.62 Å). In addition, the vacuum space was 15 Å to avoid artificial interactions between periodic images in the z direction. The amorphous RuO₂ phase was generated using ab initio molecular dynamics following the method of Parrinello and Rahman, starting from crystalline RuO₂ configurations. The initial structures were randomized at 3000 K for 10 ps, quenched from 3000 to 300 K within 5 ps, and further equilibrated at 300 K for an additional 10 ps to obtain amorphous geometries. For the crystal/amorphous heterophase model, two layers of RuO₂ were fixed, while the other two layers were relaxed during ab initio molecular dynamics simulations. To screen suitable transition metals for doping on RuO₂(110), Ru atoms were substituted with transition-metal atoms at the doped A, B, and C sites.

Supplementary Materials

This PDF file includes:

Figs. S1 to S71

Tables S1 to S12

References