Synthesis and characterization of Cr0.2Ru0.8O2-x
The Cr0.2Ru0.8O2-x catalyst was synthesized via a simple and scalable sol-gel method (Fig. 1a). Briefly, the metal ion precursors were first gelated to produce a liquogel. After an air annealing treatment, the liquogel was activated and transformed to metal oxides powder. The obtained powder was then acid-leached to remove unstable species and form the final catalyst. For comparison, catalysts with different Cr: Ru mole ratios and annealing temperatures were also synthesized using the same method.Fig. 1: Synthesis scheme and physical characterizations of Cr0.2Ru0.8O2-x.a Schematic synthesis of Cr0.2Ru0.8O2-x. b XRD patterns and c Raman spectra of Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2. d TEM image and diameter distribution (inset) of Cr0.2Ru0.8O2-x. e HADDF-STEM images of Cr0.2Ru0.8O2-x under low and high (inset) magnifications. f Corresponding fast Fourier transform pattern of inset in (e). g EDS mapping of Cr0.2Ru0.8O2-x.In Fig. 1b, the X-ray diffraction (XRD) pattern shows that the crystalline phase of Cr0.2Ru0.8O2-x matched the rutile-structured RuO2 without phase segregation. We noted that the diffraction peaks of homemade RuO2 were shifted to lower angles than those of commercial RuO2, presumably owing to the lattice expansion induced by the formation of oxygen vacancies56,57, an important attribute in lowering the OER activation energy and stabilizing the intermediates during the reaction11. With the increased amount of Cr dopant, the diffraction peaks were gradually shifted to higher angles, suggesting a lattice contraction due to the smaller atomic radius of Cr (1.27 Å) compared to Ru (1.32 Å) (Supplementary Fig. 1). Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2 present the same characteristic Raman peaks at 521, 638, and 705 cm−1, corresponding to the Ru-O vibrational modes of Eg, A1g, and B2g, respectively (Fig. 1c), which confirms the incorporation of Cr in RuO2. Transmission electron microscopy (TEM) and high-angle annular dark-field scanning TEM (HAADF-STEM) images (Fig. 1d and Supplementary Fig. 2) of Cr0.2Ru0.8O2-x clearly show an interconnected and uniformly sized nanoparticles morphology with an average diameter of ~6.5 nm (inset in Fig. 1d), slightly smaller than that of homemade RuO2 (Supplementary Fig. 3a). Each nanoparticle exhibits good crystallinity and well-resolved lattice fringes corresponding to the (110) planes of rutile RuO2 (Figs. 1e, f), testifying the lattice contraction caused by the incorporation of Cr (Supplementary Figs. 3b–d), in agreement with the XRD analysis. The lattice contraction represents a shortened interatomic Ru-Cr distance in Cr0.2Ru0.8O2-x, which potentially contributes to promote dopant/host interaction. Energy-dispersive spectroscopic (EDS) mapping images display the homogenous distribution of Ru, Cr, and O elements throughout the catalyst (Fig. 1g and Supplementary Figs. 4, 5). The co-existence of Ru and Cr was also evidenced by the electron energy-loss spectroscopy (EELS) analysis of two randomly selected regions in a single nanocrystal (Supplementary Fig. 6). Brunauer-Emmett-Teller (BET) adsorption/desorption isotherms disclose that the Cr0.2Ru0.8O2-x nanoparticles possess a relatively high surface area of 40.0 m2 g−1 (Supplementary Fig. 7). The overall Ru percentages and Cr/Ru mole ratios of different catalysts were further determined by inductively coupled plasma mass spectrometry (ICP-MS) (Supplementary Fig. 8 and Supplementary Table 1).Electronic properties of Cr0.2Ru0.8O2-x
The local structure and chemical states of Ru, Cr, and O in Cr0.2Ru0.8O2-x were investigated by X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 9). Given that the Ru 3d region is overlapped with C 1 s peak of adventitious carbon, we calibrated all the binding energies based on the lattice oxygen signal centered at 529.8 eV and analyzed the Ru 3p region instead38,58. Figure 2a shows the doublet peaks of Cr 2p3/2 and 2p1/2 orbitals of Cr0.2Ru0.8O2-x. The Cr 2p3/2 spectrum could be fitted with the peaks at 576.0 and 577.9 eV, ascribable to Cr3+-O and Cr3+-OH59,60. For Ru 3p3/2, the fitted peaks at 463.1 and 465.1 eV could be assigned to Ru4+ and Ru3+, respectively (Fig. 2b)50,61. Compared to homemade RuO2, Cr0.2Ru0.8O2-x exhibited a lower integral area ratio of Ru4+ and Ru3+ peaks (Ru4+/Ru3+), implying that Ru species in Cr0.2Ru0.8O2-x possess a slightly lower oxidation state than that in homemade RuO2. As the content of Cr dopant increased from 0% to 40%, the Ru4+/Ru3+ value decreased from 2.09 to 1.35 (Supplementary Fig. 10a), which exhibited a good linear relationship (Supplementary Fig. 11a), suggesting that Cr doping increased the concentration of Ru3+ species. In Fig. 2c, the O 1 s spectrum of Cr0.2Ru0.8O2-x could be deconvoluted into three peaks located at 529.8, 531.1, and 533.0 eV, corresponding to lattice oxygen (OL), oxygen vacancies (OV), and surface-adsorbed hydroxyl species (OHad), respectively53,54,55. We noted that the proportion of surface oxygen vacancies and lattice oxygen (OV/OL) increased markedly after doping (Supplementary Fig. 10b). A higher OV/OL ratio implies preferential adsorption of water at the intrinsic OV sites due to the decrease of steric hindrance54,62, which pertains to optimizing the adsorption of oxygenated intermediates. Moreover, a linear trend was also found between the OV/OL value and the doping ratio (Supplementary Fig. 11b), indicating that more oxygen vacancies were induced by the incorporation of Cr. The electron paramagnetic resonance (EPR) spectra confirmed the higher OV concentration in Cr0.2Ru0.8O2-x compared with that in homemade RuO2 (Supplementary Fig. 12)63. According to the neutrality principle, oxygen vacancies will be formed to maintain the electrostatic balance when a lower valence cation dopant (Cr3+) is incorporated into the host metal ions (Ru4+) in the oxide, and the valence of the host metal ions (Ru4+) will be reduced correspondingly64,65.Fig. 2: Electronic structure characterizations of Cr0.2Ru0.8O2-x.a Cr 2p core level XPS spectra of Cr0.2Ru0.8O2-x. b Ru 3p3/2 and c O 1 s core level XPS spectra of Cr0.2Ru0.8O2-x and homemade RuO2. d Normalized Ru K-edge XANES spectra and e Ru K-edge FT-EXAFS spectra of Cr0.2Ru0.8O2-x, RuO2 and Ru foil. f Ru K-edge WT-EXAFS spectra of Cr0.2Ru0.8O2-x and RuO2. g Normalized Cr K-edge XANES and h Cr K-edge FT-EXAFS spectra of Cr0.2Ru0.8O2-x, Cr2O3, and Cr foil. i Cr K-edge WT-EXAFS spectra of Cr0.2Ru0.8O2-x and Cr2O3.We further employed X-ray absorption spectroscopy (XAS) to evaluate the valence state and coordination environment of Ru and Cr sites in Cr0.2Ru0.8O2-x. Figure 2d depicts the Ru K-edge X-ray absorption near-edge structure (XANES) spectrum of Cr0.2Ru0.8O2-x, along with Ru foil and commercial RuO2 as references. The absorption edge of Cr0.2Ru0.8O2-x was found to be slightly shifted to lower energy compared with that of commercial RuO2 (inset in Fig. 2d), hinting at a relatively lower Ru oxidation state in Cr0.2Ru0.8O2-x, consistent with the XPS results. The Fourier-transformed extended X-ray absorption fine-structure (FT-EXAFS) spectrum of Cr0.2Ru0.8O2-x at the Ru K-edge showed a prominent peak at ~1.50 Å affiliated with the first Ru-O coordination shell, identical to that of RuO2 (Fig. 2e). Quantitative FT-EXAFS fitting yielded a Ru coordination number of 5.67 (Supplementary Fig. 13 and Supplementary Table 2), attributing to the presence of oxygen vacancies, which serve as the nucleophilic sites in promoting O-O bond formation66. We also analyzed the XANES and EXAFS spectra of Cr0.2Ru0.8O2-x and reference samples at the Cr K-edge, showing that the oxidation state of Cr in Cr0.2Ru0.8O2-x was very close to Cr2O3 of +3 (Fig. 2g). Additionally, the distinct FT-EXAFS spectrum compared with that of the Cr2O3 or Cr reference revealed that no obvious Cr2O3 or Cr metal nanoparticles were formed in Cr0.2Ru0.8O2-x (Fig. 2h), indicating the incorporation of Cr inside the RuO2 lattice, in agreement with the HAADF-STEM results. The wavelet transform (WT) EXAFS spectra confirmed that the Cr dopants were uniformly distributed in the RuO2 matrix (Figs. 2f, i, and Supplementary Fig. 14).Electrocatalytic performance in solution and PEMWEThe OER catalytic activity of Cr0.2Ru0.8O2-x was assessed using the rotating disk electrode (RDE) method in 0.1 M HClO4. Linear sweep voltammetry (LSV) of Cr0.2Ru0.8O2-x exhibits an overpotential of 170 ± 6 mV at 10 mA cm−2 (Fig. 3a, Supplementary Figs. 15–17 and Supplementary Note 1), outperforming that of homemade RuO2 (260 ± 10 mV) and commercial RuO2 (330 ± 13 mV). The Tafel slope of Cr0.2Ru0.8O2-x in the second region was 75.3 mV dec-1 (Fig. 3b), much smaller than that of homemade RuO2 (118.9 mV dec-1) and commercial RuO2 (127.3 mV dec-1), suggesting a possible change in the rate-determining step (RDS) at high overpotential67,68. Electrochemical impedance spectroscopy (EIS) shows the lowest charge-transfer resistances (Rct) for Cr0.2Ru0.8O2-x, indicating a superior charge transfer and enhanced OER kinetics (Supplementary Fig. 18)34,46,49. The electrochemically active surface areas (ECSA) of different catalysts were also calculated by estimating the double-layer capacitance (Cdl) from the cyclic voltammogram (CV) curves in the non-Faradaic region (Fig. 3c, Supplementary Fig. 19 and Supplementary Note 2). Cr0.2Ru0.8O2-x exhibits the largest ECSA (220.3 cm2) compared with that of homemade RuO2 (115.7 cm2) and commercial RuO2 (39.4 cm2) (Supplementary Table 3). Figure 3d compares several important performance metrics of different studied catalysts (Supplementary Table 4 and Supplementary Notes 3–5). Cr0.2Ru0.8O2-x yields a turnover frequency (TOF) of 0.127 ± 0.010 s-1, which is 9.1 and 25.4 folds greater than that of homemade RuO2 and commercial RuO2, respectively. Additionally, the specific activities of Cr0.2Ru0.8O2-x normalized by the Ru loading mass, the BET surface area, or the ECSA were also much higher than those of homemade RuO2 and commercial RuO2. Overall, the Cr0.2Ru0.8O2-x catalyst exhibits the highest intrinsic activity, possibly caused by a different OER mechanism. We further utilized in-situ gas chromatography (GC) to analyze the O2 gas produced from the Cr0.2Ru0.8O2-x decorated carbon paper electrode at 10, 100, 200, and 500 mA cm−2 (Supplementary Fig. 20). The quantified O2 amount at these current densities all matched well with the theoretical values, representing an average Faradaic efficiency (FE) of 99.8 ± 0.3% (Supplementary Fig. 21 and Supplementary Note 6), indicating that the O2 formation through the four-electron transfer during OER is the only reaction pathway over Cr0.2Ru0.8O2-x.Fig. 3: OER performance.a LSV curves and b Tafel plots of Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2 on a 0.196 cm2 electrode (100% iR correction, where R was determined to be 28.5 ± 0.4 Ω). c Cdl plots of Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2. d Radar diagram of some major OER performance metrics of Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2. e Chronopotentiometry curves of Cr0.2Ru0.8O2-x, homemade RuO2, and commercial RuO2 at a constant current density of 10 mA cm−2. f Polarization curves of the PEMWE device with Cr0.2Ru0.8O2-x and commercial RuO2 as anode catalysts at 60 °C. g Chronopotentiometry curves of the PEMWE device with Cr0.2Ru0.8O2-x and commercial RuO2 as anode catalysts operated at 1 A cm−2 and 60 °C. Insets show photographs of the PEMWE device.Next, we investigated the acidic OER stability of Cr0.2Ru0.8O2-x by employing the accelerated degradation tests (ADT). After 5000 cycles, the LSV curve of Cr0.2Ru0.8O2-x experienced a minor shift, with the overpotential at 10 mA cm−2 increased by mere 5 mV (Supplementary Fig. 22). In contrast, the homemade RuO2 displayed a pronounced positive shift of 81 mV, while the commercial RuO2 almost lost its OER activity entirely. Additionally, the chronopotentiometry (CP) curves of Cr0.2Ru0.8O2-x shows that the overpotential at 10 mA cm−2 increased 26 mV after 2000 h (Fig. 3e), representing a degradation rate of 13 μV h−1. In stark contrast, the activity of homemade RuO2 and commercial RuO2 decayed rapidly within 120 h and 30 h, respectively. At a higher current density of 100 mA cm−2, Cr0.2Ru0.8O2-x also demonstrated superior stability with an additional overpotential of 96 mV for over 300 h (Supplementary Fig. 23). The metal dissolution during CP test at 10 mA cm−2 was checked by ICP-MS (Supplementary Fig. 24). It shows that the concentrations of Ru and Cr in the electrolyte were 168 and 91 ppb after 2000 h stability test, corresponding to the mass loss of ~1.67% with Ru and ~7.79% with Cr over their stoichiometric loadings in Cr0.2Ru0.8O2-x (Supplementary Note 7). Accordingly, the stability number (S-number) of Cr0.2Ru0.8O2-x at 2000 h was determined to be 1.1 × 106 noxygen nRu-1 (Supplementary Note 8), 35.5 times as that of homemade RuO2 at 100 h (3.1 × 104 noxygen nRu-1). These results place Cr0.2Ru0.8O2-x among the most active and stable Ru-based catalysts documented in acidic electrolyte (Supplementary Table 5).Encouraged by the three-electrode test results, we further assembled a PEMWE device with Cr0.2Ru0.8O2-x as anode catalyst using deionized water as feed at 60 °C (Supplementary Fig. 25). The polarization curves in Fig. 3f show that Cr0.2Ru0.8O2-x requires a cell voltage of 1.77 V at 1 A cm−2, which is 260 mV lower than that of commercial RuO2. More impressively, Cr0.2Ru0.8O2-x can be operated steadily at 1 A cm−2 over 200 h, with potential increase of only 64 mV, which is in stark contrast with the commercial RuO2 that deactivated within mere 5 h (Fig. 3g). Subsequent structural characterizations of the used Cr0.2Ru0.8O2-x catalyst revealed that the structure, morphology and composition were well preserved (Supplementary Fig. 26). The performance of Cr0.2Ru0.8O2-x in PEMWE surpasses most Ru-based catalysts reported previously and even some Ir-based catalysts (Supplementary Table 6), demonstrating its prospect for industrial-scale water electrolysis.Structure-stability relationshipTo detect the structural evolution of Cr0.2Ru0.8O2-x, we performed a separate CP test in 0.1 M HClO4 electrolyte for 300 h at 10 mA cm−2. The surface structure and morphology of the catalyst after OER were found to be nearly unchanged from the pristine state (Supplementary Figs. 27, 28). We noted that the lattice spacing of Cr0.2Ru0.8O2-x after OER was slightly increased (Supplementary Figs. 28b, e), probably because the mass loss of Cr was more than that of Ru during the reaction, giving rise to a decrease of Cr doping in Cr0.2Ru0.8O2-x. Nevertheless, such slight lattice expansion caused by the surface metal dissolution seemed to not affect the OER performance, showing the robustness of Cr0.2Ru0.8O2-x against structural perturbation. Intriguingly, the Raman spectrum exhibits a new vibration peak of Cr-OH at ~800 cm−1 after OER (Supplementary Fig. 29a)69,70, indicating that Cr dopant may serve as a Lewis acid site to dynamically split water molecules and capture hydroxyl species during the reaction59. To verify the existence of Cr-OH, the Raman spectra were recorded after another CP test in D2O supported 0.1 M HClO4 electrolyte for 3 h at 10 mA cm−2. In the deuterium isotopic measurement, the peak of Cr-OH at ~800 cm−1 was shifted to lower wavenumbers of ~580 cm−1 (Supplementary Fig. 29b), in accordance with the expected shifts from the mass formula (γ = 72.76%) (Supplementary Note 9), confirming that the peak at ~800 cm−1 belongs to the Cr-OH vibration on Cr0.2Ru0.8O2-x.We also analyzed the chemical states of Ru, Cr, and O in Cr0.2Ru0.8O2-x before and after OER by XPS (Supplementary Fig. 30a) to demonstrate the over-oxidation and dissolution resistance of Cr0.2Ru0.8O2-x. The slightly decreased area ratio of Ru4+/Ru3+ together with the positive shift of Cr3+-O after stability test revealed that Cr dopants donate partial electrons to Ru active sites in the Ru-O-Cr structural motif to prevent Ru over-oxidation (Supplementary Figs. 30b, c)43. Noticeably, the contribution of Cr3+-OH to the Cr 2p3/2 peak increased compared with that of the pristine sample, confirming the capture of OH* on Cr dopants from water proposed by the Raman results. The O 1 s (Supplementary Fig. 30d) and EPR (Supplementary Fig. 31) spectra show that the OV/OL ratio was significantly reduced after OER. We thus surmise that, with the presence of Cr dopant and oxygen vacancy, the formation energy of oxygen vacancy was raised and the O–O coupling from lattice oxygen via LOM was inhibited. These observations provide a preliminary explanation for the stability of the Cr0.2Ru0.8O2-x catalyst.OER enhancement mechanismOur Raman and XPS analysis on Cr0.2Ru0.8O2-x revealed the capture of hydroxyl species on Cr dopants from water. We thus speculated that the surface Cr dopants could participate in the catalytic OER pathway and enhance the kinetics. To this end, we carried out quasi in-situ XPS to probe the variation in adsorbed hydroxyl species on Cr dopants from open-circuit potential (OCP) to 1.8 V. Results show that the area ratio of Cr3+-OH in Cr 2p3/2 peak displays an opposite trend to the LSV curve with the increase of anodic potential (Fig. 4a, b). In the pre-catalysis stage (< 1.3 V), the Cr3+-OH ratio was maintained at a high level and hardly decreased. However, when OER commenced, the Cr3+-OH ratio declined rapidly as the current intensity increased, indicating the accelerated participation of OH* on Cr dopants in the reaction. Operando surface-enhanced Raman scattering (SERS) was applied to investigate the dynamic surface structural evolution (Supplementary Figs. 32–34). For Cr0.2Ru0.8O2-x, the vibration peak area of Cr-OH at ~800 cm−1 presents a similar decreasing tendency with increasing potential (Fig. 4c, d), affirming the involvement of Cr dopants in OER. In addition to the vibration peaks of Ru-O bands appears at 521, 638, and 705 cm−1 throughout the potential interval, a new Raman peak at ~720 cm−1 could be deconvoluted above 1.4 V, assignable to the generation of Ru-OOH intermediates38,71. These OOH* species gradually enriched on the Ru sites as the potential increased (1.3-1.6 V) and then deprotonated to form O2 at high overpotential (> 1.6 V), when the OER proceeds much more rapidly, resulting in a “volcano curve” of surface coverage. In contrast, the amount of Ru-OOH on homemade RuO2 kept on increasing during OER, with a slower growth rate than that of Cr0.2Ru0.8O2-x (Fig. 4d and Supplementary Fig. 35). We noted that the decrement of Cr-OH combined with the intensified enrichment of Ru-OOH suggest a different mechanism in which OOH* on active site can be converted from O* via the dynamic adsorption of OH* desorbed from adjacent dopant. To further corroborate the source and amount of Ru-OOH on Cr0.2Ru0.8O2-x, operando SERS were performed on Cr2O3 (Supplementary Fig. 36a). As the potential increased, the peak area of Cr-OH initially unchanged and then slightly increased at high overpotential, indicating that OH* adsorbed on Cr site should not be oxidized to O* or OOH* and thus tends to act as a provenance to form Ru-OOH, consistent with the LSV curve of Cr2O3 within the same potential range (Supplementary Fig. 36b). The operando attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) spectra (Supplementary Figs. 37, 38) of Cr0.2Ru0.8O2-x show an apparent signal peak at ~1230 cm−1 above 1.4 V, corresponding to the generation of OOH*, which can be regarded as Ru-OOH since the formation of Cr-OOH is excluded by the operando SERS and LSV results on Cr2O3. Compared with homemade RuO2, Cr0.2Ru0.8O2-x shows stronger peak intensities of OOH*. Moreover, the peak area variation with increasing potential (Supplementary Fig. 39) exhibits a similar trend as that of Ru-OOH measured by operando SERS. These findings together evidence the faster generation of Ru-OOH on Cr0.2Ru0.8O2-x.Fig. 4: Quasi in situ XPS, operando SERS, and operando DEMS measurements understanding the mechanism on Cr0.2Ru0.8O2-x catalyst during OER.a Quasi in situ XPS spectra of Cr 2p3/2 for Cr0.2Ru0.8O2-x under applied potentials from OCP to 1.8 V. b Area ratio of Cr3+-OH and OER current (without iR correction) as a function of applied potential for Cr0.2Ru0.8O2-x. c Operando SERS spectra of Ru-O and Cr-OH bands for Cr0.2Ru0.8O2-x under applied potentials from OCP to 1.8 V. d Normalized peak areas of Cr-OH and Ru-OOH as a function of applied potential for Cr0.2Ru0.8O2-x and homemade RuO2. e Operando DEMS measurements for Cr0.2Ru0.8O2-x. Top: CV cycles (without iR correction); middle: DEMS signals of 32O2 (16O + 16O) and 34O2 (16O + 18O) collected during CV cycles; bottom: signal intensity ratio of 34O2:32O2. f Schematic illustration of dopant-mediated hydroxyl spillover mechanism on Cr0.2Ru0.8O2-x for acidic OER.We used 18O isotope-labeled operando differential electrochemical mass spectrometry (DEMS) to study the OER mechanism on Cr0.2Ru0.8O2-x (Supplementary Fig. 40), following a test protocol45,54. First, we labeled the catalyst with 18O by operating CP test. We then detected the isotope signal of evolved O2 by scanning CV in H216O electrolyte. If OER occurs through LOM pathway, 34O2 can be produced by the combination of 18O labeled lattice oxygen and 16O in water, thus the signal ratio of 34O2/32O2 would be significantly higher than the natural abundance determined 34O2/32O2 (Supplementary Note 10) and gradually decrease as the reaction proceeds due to the consumption of labeled 18O in lattice. In contrast, in AEM pathway, 34O2 can only be generated from 18O and 16O in water because lattice oxygen does not participate in OER, therefore the 34O2/32O ratio would be identical to the natural abundance determined 34O2/32O2 and remain constant. As Fig. 4e shows, the 34O2/32O2 ratio was maintained at around 0.4% during OER, suggesting the OER mechanism on Cr0.2Ru0.8O2-x is similar to AEM rather than LOM72,73,74.In light of the above analysis, we propose a mechanism with dopant-mediated hydroxyl spillover on Cr0.2Ru0.8O2-x catalyst for acidic OER (Fig. 4f), which is different from conventional AEM and LOM (Supplementary Figs. 41a, b). Dual-site oxide path mechanism (OPM) (Supplementary Fig. 41c) can also be excluded because distinctive OOH* intermediates involved in the AEM-like OER were detected in both operando SERS and operando ATR-SEIRAS measurements. It has been recognized that rutile RuO2 follows AEM in acidic media, with the attack and oxidation of water on O* to form OOH* as the RDS45,53,54,55. Our modified OER mechanism involves the accelerated formation and increased coverage of OOH* intermediates on the catalyst surface via the dynamic hydroxyl spillover from dopant (Cr) to the adjacent active site (Ru), which lowers the energy barrier of RDS in conventional AEM and greatly improves the OER activity. Apart from the structural integrity inherited from AEM, the lowered energy barrier of RDS under the modified mechanism mitigates the over-oxidation of Ru active sites, thus the OER stability is also improved.Furthermore, our quasi in-situ XPS analysis exhibits that the peak of Cr3+-O shifted continuously towards higher binding energy when the applied potential increased from OCP to 1.8 V (Fig. 4a), confirming that Cr dopants protect Ru active sites from over-oxidation by donating electrons to Ru. The area ratio of Ru4+/Ru3+ increased in the pre-catalysis stage due to the oxidizing potential, suggesting an activation of Ru sites prior to OER (Supplementary Figs. 42a, 43a)43. The Ru4+/Ru3+ value then restored to its original level at low overpotential and stayed nearly constant at high overpotential, confirming the stable Ru chemical state during OER. Additionally, the OV/OL ratio initially decreased and then showed little changes at high overpotential (Supplementary Figs. 42b, 43b), indicating the crucial role of oxygen vacancy in stabilizing the Ru-O-Cr structural motif.To better understand the OER mechanism and the source of boosted activity and stability, density functional theory (DFT) calculations were performed. Based on the XRD and HAADF-STEM observations, we constructed three models, including perfect RuO2 (RuO2), Cr-doped RuO2 (Cr-RuO2), and Cr-doped RuO2 with oxygen vacancy (Cr-RuO2-OV) on RuO2 (110) surface (Supplementary Figs. 44-47 and Supplementary Data 1), to investigate the synergistic effect of Cr dopant and oxygen vacancy. We first calculated the Gibbs free energies for RuO2 (AEM) and Cr-RuO2-OV (AEM and dopant-mediated hydroxyl spillover mechanism) to confirm the favorable OER pathway (Fig. 5a, Supplementary Figs. 48, 49 and Supplementary Note 11). In AEM pathway, the formation of OOH* from O* is the RDS for both RuO2 and Cr-RuO2-OV, and the limiting free energy barrier for Cr-RuO2-OV (1.90 eV) is even higher than that for RuO2 (1.87 eV). Nevertheless, when taking the dopant-mediated hydroxyl spillover into account, the RDS for Cr-RuO2-OV changes from OOH* formation to O2 formation, and the limiting free energy barrier decreased drastically to 1.76 eV, in line with the Tafel analysis (Fig. 3b), suggesting that the OER activity of RuO2 can be indeed enhanced by activating OH* at adjacent doping sites. Besides, the OH* deprotonation energy (1.82 eV) on Cr site of Cr-RuO2-OV is much higher than the spillover energy (1.07 eV) from Cr to Ru (Fig. 5b), corroborating that OH* on Cr cite tends to spillover to Ru site rather than to be oxidized to O*. The projected density of states (PDOS) reveals that the introduction of Cr dopant and oxygen vacancy pushes the Ru d-band center (εRu) close to the Fermi level, moving from −3.38 eV (RuO2) to −3.17 eV (Cr-RuO2), then to −3.07 eV (Cr-RuO2-OV) (Fig. 5c). The up-shift of Ru d-band center increases the binding strength between Ru active sites and oxygenated intermediates, which is deleterious to the OER activity due to the scaling relationship in conventional AEM, explaining the higher limiting free energy barrier for Cr-RuO2-OV compared to RuO2 following AEM pathway. However, the enhanced binding strength of oxygen species on Ru site would accelerate the hydroxyl adsorption from adjacent Cr dopant to form Ru-OOH. Meanwhile, the presence of oxygen vacancy significantly down-shifts the Cr d-band center (εCr) from −3.36 eV (Cr-RuO2) to −4.20 eV (Cr-RuO2-OV), indicating much weaker adsorption and more difficult deprotonation of OH* on Cr site, conducive to the efficient hydroxyl desorption. Therefore, we theoretically validated that Cr dopant and oxygen vacancy synergistically modulate the adsorption of oxygenated intermediates and the hydroxyl spillover from Cr to Ru is energetically favorable.Fig. 5: DFT calculations.a Free energy profiles of RuO2 and Cr-RuO2-OV with different OER pathways. b OH* deprotonation energy on Cr site of Cr-RuO2-OV. Insets show the optimized structures in the corresponding steps. The silver, red, golden, and white balls represent Ru, O, Cr, and H atoms, respectively. c PDOS of Ru 4d and Cr 3d-bands for RuO2, Cr-RuO2, and Cr-RuO2-OV; corresponding d-band centers are denoted by dashed lines. d Vacancy formation energies of surface Ru and lattice oxygen for RuO2, Cr-RuO2, and Cr-RuO2-OV.The vacancy formation energies of surface Rucus (ΔEVRu) and Olat (ΔEVO) for different models were also calculated (Supplementary Figs. 50, 51 and Supplementary Note 11). Figure 5d shows that ΔEVRu increases from 1.68 eV for RuO2 to 2.53 eV for Cr-RuO2, indicating that surface Ru can be stabilized by Cr dopant. With the presence of oxygen vacancy, ΔEVRu decreases to 2.04 eV for Cr-RuO2-OV due to lower overall Ru-O chelation number, which, however, is still 0.36 eV higher than that for RuO2, suggesting a more stable surface Ru in Cr-RuO2-OV than in RuO2. The ΔEVO values for RuO2, Cr-RuO2 and Cr-RuO2-OV were calculated to be 3.55, 3.63 and 3.75 eV, respectively, manifesting that both Cr dopant and oxygen vacancy can stabilize surface lattice oxygen in the Ru-O-Cr structural motif. Our DFT results thus predict the enhanced OER stability of Cr0.2Ru0.8O2-x, agreeing with experimental observations.
Cr dopant mediates hydroxyl spillover on RuO2 for high-efficiency proton exchange membrane electrolysis
