Orbital-level band gap engineering of RuO₂ for enhanced acidic water oxidation

The oxygen evolution reaction (OER) plays an indisputably critical role in energy conversion systems, underpinning processes such as hydrogen production via water electrolysis, CO₂ reduction to generate clean small-molecule fuels, and the utilization of energy conversion technologies such as metal‒air batteries^1,2,3. As an essential half-reaction, the OER requires more energy to surmount the kinetic barrier owing to its sluggish four-electron transfer process, which is a bottleneck for realizing the industrialization of those conversion technologies^4,5,6. Noble metal Ir-based nanomaterials are currently regarded as the most effective catalysts for the acidic OER, but their scarcity and exorbitant cost pose significant challenges to widespread implementation^7,8,9,10. Moreover, owing to their high activity and relatively low cost (approximately one-sixth of the price of Ir), Ru-based nanomaterials hold tremendous potential in the field of proton-exchange membrane water electrolysers (PEMWEs), but the stability of RuO₂ in acidic environments remains unsatisfactory^10,11,12. To increase the overall efficiency of PEMWEs, it is necessary to develop efficient and stable Ru-based catalysts for the oxygen evolution reaction in acidic media.

Ru-based catalysts, which tend to follow the lattice oxygen oxidation mechanism (LOM) pathway, can bypass the typical thermodynamic constraints of the traditional adsorption evolution mechanism (AEM) to increase OER kinetics^13,14,15. However, the persistent involvement of lattice oxygen (O_L) during the continuous LOM pathway (LOM-LOM) can lead to excessive oxidation of Ru and its dissolution into the electrolyte, resulting in poor stability^16,17,18,19. Currently, transforming the LOM pathway of Ru-based catalysts to the AEM pathway to suppress the emergence of high-valence Ru^n>4+ species is considered an effective means to improve the stability of Ru-based catalysts, but the OER activity is inevitably constrained by the thermodynamic linear relationship^13,20,21. Thus, combining the advantages of the AEM and LOM, i.e., promoting catalytic activity by initiating the OER through lattice oxygen and improving stability by inhibiting the continuous participation of lattice oxygen in the OER, is an effective strategy for designing high-performance Ru-based catalysts. In an exciting recent breakthrough, the lattice oxygen-mediated mechanism-oxygen vacancy site mechanism (LOM-OVSM) was proposed, which could bypass the linear relationship of oxygen intermediates and avoid the generation of Ru^n>4+, thereby enabling the activities and stabilities of the OER to be even superior to the theoretical limit, offering a promising advanced pathway for the design of improved Ru-based catalysts^22,23,24. As demonstrated in the pioneering research reported by Wang et al., an electrocatalyst satisfying the requirements for the LOM-OVSM-type OER can potentially overcome the so-called scaling limit without sacrificing stability²⁵. During the LOM-OVSM pathway, the LOM-mediated process can increase the kinetic rate, and the OVSM process can prevent the formation of excessive high-valence Ru species, which requires the rapid diffusion of electrons to the external circuit and their swift return. However, the lack of comprehensive research on the relationship between the electronic structure of Ru-based materials and the OER mechanism may inevitably lead to additional challenges in constructing the LOM-OVSM pathway for Ru-based catalysts. Owing to the spatial overlap between orbitals and similar energy distributions of electronic states, d–p orbital hybridization naturally occurs following the introduction of light elements (F, C, O, N, P, S, Se, etc.) and provides substantial flexibility for optimizing the electronic structure, thereby contributing to high catalytic performance^26,27. Therefore, a more in-depth understanding of the electronic structure‒mechanism relationship can be established by providing an unequivocal structural analysis of Ru-based materials with appropriate doped anions; however, research on this topic is limited.

In this work, we pioneer the doping of different p-orbital atoms R (N, P, S, Se) into RuO₂ (R-RuO_x) with a simple one-step calcination synthesis method to investigate the electronic structure‒mechanism relationship of Ru-based materials and the construction principle of the LOM‒OVSM path. In situ and ex situ measurements along with theoretical calculations demonstrated that the notable OER activity and stability of Se-RuO_x originate from the dynamic adjustment of the band gap between the Ru-e_g and O-p orbitals during the OER process. Notably, Se-RuO_x catalyst also translates into promising performance in PEMWE devices, exhibiting a minimal charge overpotential of 1.67 V to deliver 1 A cm⁻² and maintaining cyclability for over 1000 h. Overall, a Se-RuO_x catalyst that dynamically adjusts the band gap between the Ru-e_g and O-p orbitals is essential for creating LOM-OVSM pathway catalysts that notably enhance both OER activity and stability, presenting a promising avenue for the strategic development of high-performance metal oxide catalysts.

Materials characterization

X-ray diffraction (XRD) measurement was conducted to identify the phase information for all the catalysts. As shown in Fig. 1a, b, all the XRD patterns display the same group of peaks in agreement with the RuO₂ phase (JCPDS No. 40-1290). Moreover, with the introduction of R (R = N, P, S, or Se), the peak positions of R-RuO_x tend to shift towards lower angles than those of RuO₂, without the appearance of any additional diffraction peaks, indicating the successful doping of N, P, S, or Se without damaging the structure. The transmission electron microscopy (TEM) images shown in Fig. 1c and Figs. S1–5 demonstrate the flexible nanoparticle morphology of the as-prepared R-RuO_x with smooth and evenly dispersed surfaces. Compared with that of RuO₂, the lattice spacing of R-RuO_x slightly expands, further indicating the successful introduction of anionic R, which is consistent with the XRD results. Then, spherical aberration-corrected (HAADF-STEM) was employed to investigate the Se atom distribution in the Se-RuO_x hybrid nanostructures. Energy-dispersive X-ray spectroscopy (EDS) elemental mapping images revealed a uniform distribution of all the elements within Se-RuO_x (Fig. 1d). Se-RuO_x nanoparticles with an average diameter of 10 nm were further determined from HAADF-STEM images, as shown in Fig. 1e and Fig. S6. The high-resolution TEM images obtained via spherical aberration correction show interplanar spacings of 0.32 and 0.26 nm, which correspond to the (110) and (101) lattice fringes of face-centred cubic (FCC) RuO₂ (Fig. 1f–h). This suggests that Se-RuO_x closely resembles the crystal structure of RuO₂, with Se atoms exclusively located at the O position, and the corresponding crystal structure model of Se-RuO_x simulates consistent element spread (Fig. 1i).

a XRD patterns of RuO₂ and R-RuO_x (R = N, P, S, Se). b Extended XRD patterns derived from (a). c TEM images of Se-RuO_x. d Elemental mappings of Se-RuO_x at a 1 nm scale by HAADF-STEM. e–g AC-HAADF-STEM images and the corresponding lattice fringe profile of Se-RuO_x. h FFT pattern in (g). i HAADF-STEM image and the simulated molecular structure of Se-RuO_x.

Full size image

X-ray absorption spectroscopy (XAS) techniques were utilized to investigate the electronic configuration of R-RuO_x catalysts, providing valuable insights into the lattice oxygen activity and OER catalytic pathways in R-RuO_x catalysts. X-ray absorption near-edge structure spectroscopy (XANES) of the Ru K edge, as shown in Fig. 2a, b, reveals variations in the adsorption threshold position and oxidation state of the R-RuO_x catalyst with different R values. Se-RuO_x has the lowest valence state, indicating the highest covalency of the Ru-O bond and the most electron-rich state in Se-RuO_x, thereby facilitating the LOM process^28,29. Furthermore, Ru K-edge Fourier transform extended X-ray absorption fine structure (FT-EXAFS) spectroscopy was used to investigate the local coordination environment of R-RuO_x. The main peak at 1.967 Å for pure RuO₂ corresponds to the first Ru-O coordination shell of the Ru cations (Fig. 2c)^30,31. Compared with those of RuO₂, increases in the length of the Ru-O bond are observed in Se-RuO_x (1.978 Å), S-RuO_x (1.976 Å), P-RuO_x (1.971 Å) and N-RuO_x (1.973 Å), suggesting the enhanced reactivity of lattice oxygen, particularly in Se-RuO_x (Figs. 2c, d, S7–8 and Table S1). Furthermore, wavelet transform analysis corroborated the results from the FT-EXAFS spectra and revealed the longest Ru-O bond length in Se-RuO_x (Fig. 2e). Additionally, the O K-edge XANES spectrum for R-RuO_x catalysts, as shown in Fig. 2f, reveals two distinct sharp peaks at 531.7 and 534.7 eV, corresponding to the excitation of O 1 s core electrons into hybridized states between O 2p and Ru 4 d t_2g and e_g states caused by the splitting of an octahedral field^32,33,34. Another broad peak at higher energy is attributed to the hybridization of the O 2p orbital with Ru 5sp states³⁵. Compared with that of RuO₂, the increased t_2g/e_g intensity further reveals the increased Ru-O covalency in R-RuO_x, with Se-RuO_x being particularly notable, indicating the highest degree of lattice oxygen activation in Se-RuO_x and thus making it more conducive to promoting the formation of V_O (Fig. 2g).

a Ru K-edge XANES spectra, (Inset: Partially enlarged image). b Oxidation state of ruthenium species obtained from Ru K-edge XANES. c Ru K-edge FT-EXAFS spectra. d The fitted Ru‒O bond length (horizontal axis R: coordination bond length). e Wavelet transform for Ru K-edge EXAFS signals (k: Wave vector number). f O K-edge XANES spectra. g t_2g/e_g intensities of RuO₂ and R-RuO_x. h Ru 3p XPS spectra. i O 1 s XPS spectra (Vo: oxygen vacancy; O_L: Lattice oxygen; H₂O: absorbs oxygen). j EPR spectra. k Inverse susceptibility 1/χ against temperature. l Effects of Se doping on the Ru-d orbital structure and the changes in the active centre (LOM lattice oxygen oxidation mechanism, LOM-OVSM lattice oxygen-mediated mechanism-oxygen vacancy site mechanism).

Full size image

To verify the impact of R doping on the formation and stability of V_O during the OER process, the electronic structures of R-RuO_x after cyclic voltammetry (CV) activation (V_O-R-RuO_x) in the potential range of 1.1 to 1.6 V vs. RHE were investigated. First, X-ray photoelectron spectroscopy (XPS) was used to determine the V_O concentration of V_O-R-RuO_x. The Ni 1 s, P 2p, S 2p and Se 3 d XPS spectra shown in Fig. S9 further demonstrate the successful introduction of anionic R (R = N, P, S, Se). Figure 2h shows that the Ru 3p spectra of all the V_O-R-RuO_x samples are almost identical, but the two peaks of Ru 3p_3/2 and Ru 3p_1/2 in V_O-Se-RuO_x shift towards lower binding energies than those of the other V_O-R-RuO_x samples do, indicating that V_O-Se-RuO_x possesses more low-valence Ru ions and V_O. In the O 1 s spectra, the relative contribution of the peak related to V_O in V_O-Se-RuO_x is 55.03%, which is markedly greater than that in the other samples, demonstrating a richer presence of V_O in V_O-Se-RuO_x (Fig. 2i)^36,37. Moreover, V_O-Se-RuO_x shows a stronger electron paramagnetic resonance (EPR) signal at g  =  2.004, which also proves that V_O-Se-RuO_x contains more V_O than other V_O-R-RuO_x (Fig. 2j)³⁸. The results from XAS, XPS, and EPR analyses indicate that Se-RuO_x is more prone to facilitate the LOM-mediated process and produce abundant V_O. Additionally, the temperature-dependent magnetizations (M-T) curves for V_O-R-RuO_x were collected with a magnetic field of H = 500 Oe under field-cooling procedures to probe the spin structures of the Ru centres because of the high sensitivity of magnetism to unpaired electrons (Fig. S10). The magnetic susceptibilities above 150 K obey a paramagnetic Curie–Weiss law, and the e_g-electron filling state of the sample can be calculated by fitting³⁹. As expected, an impressive relationship with the trend of e_g-electron filling state after activation, V_O-Se-RuO_x (1.16) > V_O-S-RuO_x (0.95) > V_O-P-RuO_x (0.73) > V_O-N-RuO_x (0.54) > V_O-RuO₂ (0.43), can be observed (Fig. 2k, and Table S2), demonstrating faster electron compensation from the outer circuit back to the e_g orbitals of V_O-Se-RuO_x, thus eliminating the overoxidation of Ru and promoting the OVSM process⁴⁰. In summary, as shown in Fig. 2l, when O atoms are replaced by Se atoms, the d_z² (e_g) energy decreases, allowing electrons from the O-p orbital to transfer more rapidly to the e_g orbital and subsequently to the external circuit, thereby enhancing the LOM-mediated process. After the LOM pathway occurs, the electrons of V_O-Se-RuO_x return quickly, thereby stabilizing the V_O, leading to increased energy of the t_2g orbital and decreased energy of the O-p orbital, thereby promoting the OVSM process^41,42.

Electrochemical and spectroscopic techniques were further utilized to identify the reaction kinetics, validating the potential LOM-OVSM pathway as considered above, which facilitates deciphering the notable OER performance. Linear sweep voltammetry (LSV) curves of R-RuO_x and RuO₂ in H₂SO₄ (pH 0-1) at a scan rate of 5 mV s⁻¹ were recorded (Fig. 3a and S11). All the R-RuO_x and RuO₂ samples demonstrated obvious pH dependence, suggesting nonconcerted proton‒electron transfer (NCPET) processes^7,43. Furthermore, good linear relationships were observed between the corresponding log(j) values at different potentials and the electrolyte pH values, emphasizing the mediated LOM pathway during the OER in R-RuO_x and RuO₂ (Fig. 3b). Notably, the pH correlation of Se-RuO_x is much weaker than that of RuO₂, indicating that Se-RuO_x may not exclusively follow the LOM pathway during the OER process. In addition, to gain insight into the OER mechanism, the oxidation behaviour of Ru sites at high overpotentials was investigated via cyclic voltammetry (CV) analysis between 0.3 and 1.2 V vs. SCE at a scan rate of 10 mV/s for R-RuO_x and RuO₂ (Fig. 3c). Two groups of redox peaks at approximately 0.72 and 0.98 V could be observed for RuO₂, which could be assigned to Ru³⁺/Ru⁴⁺and Ru⁴⁺/Ru⁶⁺, respectively^44,45. The reduction peak at 1.12 V corresponds to Ru⁶⁺/Ru⁸⁺ and is attributed to the excessive accumulation of the metastable dissolution intermediate *RuO₄^2- (continuous LOM pathway)⁴⁶. Compared with those of other R-RuO_x materials, the peaks of Ru³⁺/Ru⁴⁺ in Se-RuO_x are more dominant than those in Ru⁴⁺/Ru⁶⁺, and the peak potentials of Ru³⁺/Ru⁴⁺ (0.61 V) and Ru⁴⁺/Ru⁶⁺ (0.91 V) are lower, indicating that Se-RuO_x has rapid electron diffusion capability, thereby improving the OER activity⁹. Furthermore, the peak of Ru⁴⁺/Ru⁶⁺ in Se-RuO_x exhibited a distinctly decreased area accompanied by the disappearance of the redox peak for Ru⁶⁺/Ru⁸⁺, indicating that Se doping promoted electron feedback and suppressed the overoxidation of Ru sites, thereby improving OER stability. Furthermore, in situ Raman spectroscopy was employed to probe the surface state of Se-RuO_x and RuO₂ during the OER. As depicted in Fig. 3d, three main Raman features of rutile RuO₂ can be observed on both Se-RuO_x and RuO₂, including the E_g, A_1g and B_2g vibration modes⁴⁷. In addition, when the applied potential exceeds 1.20 V (Se-RuO_x) or 1.40 V (RuO₂), a Raman band can be found at approximately 640 cm⁻¹, which can be assigned to the presence of Vo, demonstrating that Se-RuO_x can form Vo more rapidly during the OER^48,49. Importantly, when the applied voltage exceeds 1.35 V, another Raman band belonging to *OOH (791 cm⁻¹) appears in Se-RuO_x, further indicating that Se-RuO_x does not follow the traditional LOM pathway during the OER but rather follows the LOM-OVSM pathway. To further confirm that the LOM-OVSM pathway is involved in Se-RuO_x during the OER process, operando ¹⁸O isotope labelling differential electrochemical mass spectrometry (DEMS) measurements were performed (Fig. S12). The RuO₂ and Se-RuO_x were labelled by conducting CV (1.2-1.65 V vs. RHE) in H₂¹⁸O aqueous sulfuric acid electrolyte, followed by OER testing in the nonlabelled aqueous (H₂¹⁶O) sulfuric acid electrolyte. The results for both RuO₂ (Fig. S13) and Se-RuO_x (Fig. 3e) show that the main mass signal is ³²O₂, followed by ³⁴O₂, with no ³⁶O₂ signal. Compared with that of RuO₂, the significant decrease in the ratio of ³⁴O₂ to ³²O₂ demonstrates that the majority of the intrinsic lattice oxygen in Se-RuO_x does not participate in O‒O coupling, thereby excluding the traditional LOM^24,25. In situ attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) was employed to further identify the reaction intermediates during the OER. For RuO₂ and N-RuO_x, P-RuO_x, and S-RuO_x, an obvious absorption band emerges at ~1200 cm⁻¹ at high overpotentials (Fig. 3f and Fig. S14), which is attributed to the stretching vibration of the *OO intermediate⁵⁰. This indicates that the OER process of RuO₂ and N-RuO_x, P-RuO_x, and S-RuO_x is predominantly governed by continuous LOM at high overpotentials, consequently accelerating dissolution and diminishing stability. In contrast, the spectrum of Se-RuO_x shows that the dynamic *OO absorption band appears at a lower voltage ( ~ 1.05 V), indicating the mediation of the LOM mechanism (Fig. 3g). Moreover, the vibrational band of the crucial *OOH intermediate (at ~1050 cm⁻¹) in the Se-RuO_x electrocatalyst emerges at ~1.15 V and notably intensifies with increasing potential, excluding the traditional LOM⁵¹. The combined findings from in situ ATR-SEIRAS, in situ Raman spectroscopy and DEMS indicate that Se-RuO_x involves a unique LOM-OVSM pathway different from the LOM pathway of RuO₂, which not only promotes the reaction kinetics but also inhibits the dissolution of the Ru-based catalyst.

a pH dependence of Se-RuO_x. b The linear relationship of onset potential vs. pH. c CV analysis of the redox peaks of RuO₂ and R-RuO_x measured from 0.0 to 1.2 V vs. SCE. d In situ Raman spectra of RuO₂ and Se-RuO_x during the OER. e DEMS tests of ¹⁶O¹⁶O, ¹⁸O¹⁶O and ¹⁸O¹⁸O signals from the catalysts for ¹⁸O-labelled R-RuO_x in 0.5 M H₂SO₄ in H₂¹⁶O. In situ ATR-SEIRAS spectra of f RuO₂ and g Se-RuO_x during the OER.

Full size image

Electrocatalytic OER performance

Encouraged by the unique LOM-OVSM pathway, we further surveyed potential applications for the acidic OER. The electrochemical OER performance of the R-RuO_x electrocatalyst was assessed by a standard three-electrode system in a 0.5 M H₂SO₄ electrolyte. As shown in Fig. 4a, b and Fig. S15, Se-RuO_x displays notable catalytic activity compared with the other R-RuO_x electrocatalysts, reflecting the smallest required overpotential of 188 mV to deliver 10 mA cm⁻², which is better than to the benchmarks of RuO₂ (279 mV), N-RuO_x (215 mV), P-RuO_x (209 mV), S-RuO_x (201 mV) and most of the Ru-based catalysts reported to date (Table S3). In addition, Se-RuO_x has a lower Tafel slope (64.6 mV dec⁻¹) than RuO₂ (80.2 mV dec⁻¹), S-RuO_x (65.2 mV dec⁻¹), P-RuO_x (66.4 mV dec⁻¹) and N-RuO_x (76.5 mV dec⁻¹) (Fig. 4c). The lower Tafel slope reveals more facile OER kinetics for Se-RuO_x, and the potential-determining step (PDS) tends towards the *O-OH coupling step during the OVSM pathway, indicating that the Se-doping system can accelerate the corresponding electron reaction. Notably, Se-RuO_x exhibited the highest turnover frequency (TOF) value among R-RuO_x at 1.55 V vs. RHE, indicating its enhanced intrinsic electrocatalytic activity (Fig. S16). Moreover, Se-RuO_x displays a lower charge transfer resistance (R_ct), which is less than to that of other R-RuO_x, as evidenced by a smaller Nyquist cycle diameter from electrochemical impedance spectroscopy (EIS) measurements (Fig. S17 and Table S4). Furthermore, Additionally, in situ EIS analyses at various applied potentials were utilized to survey the electrocatalytic kinetics occurring during OER (Figs. S18-19). In the Bode phase diagrams, the observed maxima at elevated frequencies demonstrate the electrocatalyst’s inherent electron transport properties, while corresponding peaks detected in lower frequency domains characterize interfacial charge transfer processes between electrolyte and catalyst materials^52,53. The observed electrochemical behavior in Fig. 4d, e and Fig. S20 demonstrates that R-RuO_x catalysts consistently display amplified phase signal intensity within the low-frequency ___domain ( ≈ 0.01-10 Hz) relative to their high-frequency regions ( ≈ 100—10000 Hz). This frequency-dependent response suggests that interfacial charge transfer mechanisms are principally governed by resistive interactions at the electrolyte‒catalyst interface. As operational voltage escalates from 1.25 V to 1.55 V, pronounced phase angle attenuation emerges specifically in the low-frequency regime of Se-RuO_x, contrasting with other R-RuO_x materials, which demonstrates that Se doping can reduces interfacial charge transfer resistance between electrocatalyst and electrolyte, resulting in optimized electron transport efficiency that directly improves oxygen evolution reaction performance. In addition, the electrochemical surface area (ECSA) was determined through the calculation of the double-layer capacitance (C_dl). As shown in Fig. S21, the C_dl of Se-RuO_x is 64.7 mF cm⁻², which is close to those of RuO₂ (36.3 mF cm⁻²),S-RuO_x (57.8 mF cm⁻²), P-RuO_x (46.7 mF cm⁻²) and N-RuO_x (44.2 mF cm⁻²), indicating the greater number of exposed active sites in Se-RuO_x. To elucidate the TOF progression pattern undetected in conventional polarization measurements (Fig. S16), the voltage polarization curves normalized by ECSA were constructed. (Fig. S22). Interestingly, a comparable trend to that identified in the TOF calculations is evident, clearly indicating the combined role of the intrinsic catalytic characteristics on determining the practical performance of Se-RuO_x. Given the limited stability of traditional RuO₂ under acidic oxygen evolution environments, the stability assessment of the synthesized Se-RuO_x was investigated in 0.5 M H₂SO₄. Se-RuO_x exhibits fully retained activity following 1000 CV cycles relative to its original state (Fig. 4f). In comparison, a significant increase in the overpotential is found for the com. RuO₂ electrocatalyst (Fig. S23). Furthermore, Se-RuO_x exhibited nearly unaltered OER performance during the chronopotentiometric test over a period of approximately 150 h at 100 mA cm⁻², which was much better than that of the other R-RuO_x ( < 10 h) (Fig. 4g, h). ICP-MS test was further performed to quantify the dissolution rates of Se and Ru. As illustrated in Figs. S24-25 and Table S5, Se-RuO_x demonstrates notably low dissolution rates of 0.91% for Ru and 2.29% for Se after 100 h stability testing, with the stability number (S-number) determined to be 2.2*10⁵, indicating its comparable long-term durability. In situ PXRD was performed to provide in-depth information and realistically assess the durability of Se-RuO_x at different applied potentials (Fig. 4i). Se-RuO_x retains its high crystallinity under OER operating potentials, further revealing its electrochemical stability.

a LSV curves of RuO₂ and R-RuO_x in a 0.5 M H₂SO₄ solution. b The overpotentials at 10 mA cm⁻² and 100 mA cm⁻² of RuO₂ and R-RuO_x. c Corresponding Tafel plots according to the LSV curves in (a). d Bode phase plots of Se-RuO_x. e Summarized phase peak angles of RuO₂ and R-RuO_x at 1.25–1.55 V. f LSV curves of initial and after 1000 CV cycles on Se-RuO_x. g Chronopotentiometric curves of RuO₂ and R-RuO_x at a current density of 10 mA cm⁻² for 10 h. h Chronopotentiometric curves of Se-RuO_x at a current density of 100 mA cm⁻² for 150 h. i In situ XRD patterns of Se-RuO_x at 1.3 V-1.8 V.

Full size image

Theoretical insights into intrinsic activity and stability

Our experimental results verify that Se-RuO_x changes the OER pathway from the traditional LOM to LOM-OVSM, thereby notably enhancing the OER performance. To determine the reasons behind the enhanced activity and stability of Se-RuO_x and to understand why the LOM-OVSM pathway is favoured in Se-RuO_x, density functional theory (DFT) calculations were carried out. Based on the classical theory, the metal site tends to bond with oxygen through hybridization between the p and d orbitals, resulting in d orbitals of active centres splitting into e_g and t_2g orbitals (Fig. 5a)⁵⁴. When Se is substituted for O, the O-p orbitals upshift to E_f, and the d_z² (e_g) orbitals move down, making it easier for electrons to transfer from the nonbonding O-p band to the e_g orbitals and then diffuse to the external circuit, thereby promoting the kinetics of the LOM-mediated process. When abundant V_O form during the OER process in the LOM pathway, Se doping can increase the proportion of empty e_g orbitals, promote electron return, prevent the formation of high-valence Ru species and stabilize the V_O. When stable vacancies form, the O-p orbital energy decreases, and the t_2g orbital energy increases compared with that of RuO₂, allowing electrons to more easily transfer from t_2g to e_g orbitals, thereby promoting the OVSM pathway (Fig. 5b). In particular, the introduction of an appropriate element is reasonable for broadening the e_g band and upshifting the t_2g orbitals, resulting in the conversion of the active centre from the O centre to the M centre (Fig. 5c) and conducting a unique LOM-OVSM route different from the LOM path of RuO₂ (Fig. 5d)⁵⁵.

a Ru-d orbital splitting and hybridization schematic. b Band structures of the compounds synthesized from RuO₂ and Se-RuO_x during the OER process. The position of the O₂/H₂O redox couple is 1.23 V versus RHE, as shown schematically on the right. The relationship between the voltage under the RHE and standard hydrogen electrode (SHE) scale is Ɛ_RHE=Ɛ_SHE + 59 mV×pH (pH=0 in 0.5 M H₂SO₄). c Schematic band structure of different active centres. d Adsorption/desorption and electron transfer processes of OER intermediates on the surface of the traditional LOM pathway (on RuO₂ surface) and LOM-OVSM pathway (on Se-RuO_x surface) (LOM lattice oxygen oxidation mechanism, LOM-OVSM lattice oxygen-mediated mechanism-oxygen vacancy site mechanism).

Full size image

For the computational studies, we began by building R-RuO_x (R = N, O, S, P, Se) models based on an established RuO₂ compactional framework (Figs. S26, 27). First, the partial orbital density of states (PDOS) of Ru atoms and O atoms for R-RuO_x samples were analysed and are presented in Fig. 6a and S28. The results show that the energy gap (∆ɛ) between the O-p band and the Ru-e_g band follows the order RuO₂ > N-RuO_x > P-RuO_x > S-RuO_x > Se-RuO_x, indicating increased lattice oxygen activity in Se-RuO_x compared with other R-RuO_x and RuO₂ (Fig. 6b)⁵⁶. This is further confirmed by the ∆ɛ between the O-p band and Ru-d band, which shows a similar trend (Fig. S29). Furthermore, the free energy profiles indicate that the OER on all R-RuO_x materials follows the LOM pathway rather than the AEM pathway and that Se-RuO_x has the highest OER kinetics (Fig. S30). Next, the structural properties following the formation of lattice oxygen (Vo-R-RuO_x) during the LOM process were examined, thereby identifying the reasons for the disruption of the traditional LOM pathway (Figs. S31, 32). The PDOSs for the e_g and t_2g orbitals of the Ru site for the Vo-R-RuO_x samples were analysed to verify the orbital structures in Fig. 2l and are presented in Fig. 6c and S33–37. In particular, Vo-Se-RuO_x has the highest Ru-e_g empty rate and the highest Ru-t_2g band centre, as well as the smallest gap between the e_g and t_2g orbitals (Figs. 6c, d and S34, 35). This facilitates the electron transition from the t_2g orbital to the e_g orbital and promotes electron feedback from the external circuit to the e_g orbital, thereby suppressing the continuous oxidation and dissolution of Ru. Moreover, the gap between the O-p band and the Ru-e_g band increased from Vo-RuO₂ to Vo-N-RuO_x, Vo-P-RuO_x, Vo-S-RuO_x, and Vo-Se-RuO_x (Fig. 6e, f and Fig. S36). The increased gap between the O-p band and the Ru-e_g band in Vo-Se-RuO_x promotes electron transfer from the t_2g orbital to the e_g orbital and then diffuses to the external circuit, converting the redox centre from the lattice oxygen to the metal centre, thereby transforming the traditional LOM pathway into the OVSM pathway. The lowest O-p band and the largest ∆ɛ between the O-p band and Ru-d band in Vo-Se-RuO_x also confirm the transformation of redox centres from the lattice oxygen to the metal centre (Fig. S37). In addition, compared with other Vo-R-RuO_x, Vo-Se-RuO_x possesses the lowest Ru positive charge enrichment (Figs. 6g and S38), the regression of more Ru electrons (Fig. S39), the rapid electron returns to the surface of Ru atoms (electron localized function, ELF, Fig. S40) and the smallest band gap (Fig. S41), which all further confirm the above conclusions. As expected, the optimized LOM-OVSM path on Se-RuO_x is favourable for changing and decreasing the rate-determining step (RDS) (*OH → *O) ( ~ 1.66 eV), which results in the lowest required RDS energy compared with that of RuO₂ (1.73 eV), N-RuO_x (1.71 eV), P-RuO_x (1.73 eV) and S-RuO_x (1.69 eV), indicating better reaction kinetics (Figs. 6h and S42)⁹. On the basis of these results, the notable OER activity and stability of Se-RuO_x originate from the following. (1) The decreased ∆ɛ between the O-p band and Ru-e_g bands promotes electron diffusion into the external circuit and then facilitates the initiation of the LOM pathway, thereby increasing the OER catalytic activity. (2) The increased ∆ɛ between the O-p band and Ru-e_g band and e_g orbital broadening promote electron return from the external circuit to avoid the formation of high-valence Ru species and change the traditional LOM path to OVSM, thereby improving OER stability.

a PDOS plots of the Ru-e_g and O-p states in Se-RuO_x. b Band centres of the Ru-e_g orbital and O-p orbital in RuO₂ and R-RuO_x (R = N, S, P, Se). c PDOS plots of the Ru-e_g and Ru-t_2g states in V_O-Se-RuO_x. d Ru-e_g empty rates of V_O-RuO₂ and V_O-R-RuO_x. e PDOS plots of the Ru-e_g and O-p states in V_O-Se-RuO_x (Inset: the specific atoms on Vo-Se-RuO _x surface). f Band centres of Ru-e_g and O-p in V_O-RuO₂ and V_O-R-RuO_x. g Differential charge density of RuO₂ and Se-RuO_x. h Free energy profiles of different OER intermediates of V_O-Se-RuO_x.

Full size image

In addition, DFT calculations for Se-RuO_x with different doping sites indicate that Se doping along the z-direction can lead to the optimal ∆ɛ between the O-p band and Ru-e_g bands and promote the LOM-OVSM pathway, thereby having the optimal OER performance (Figs. S43, 44). DFT calculations show that with a doping ratio of 1/12, the LOM-OVSM pathway occurs, resulting in increased OER activity and stability (Figs. S45–47). The electrocatalytic performance was assessed in 0.5 M H₂SO₄ utilizing a standard three-electrode system. As shown in Fig. S48, Se-RuO_x-2 displays the best acidic OER activity, with an overpotential of 189 mV to achieve a current density of 10 mA cm⁻², which is much lower than those of Se-RuO_x-1 (238 mV), Se-RuO_x-4 (219 mV) and Se-RuO_x-6 (233 mV). Furthermore, with increasing Se doping ratio, the stability of the Ru-based catalyst initially increases but then decreases. An excessively high ratio of Se leads to accelerated dissolution of Ru, thereby reducing the stability of the catalyst, which is consistent with the DFT calculations.

PEMWE electrolyzer

Finally, we sought to operate the Se-RuO_x catalyst in practical applications. A membrane electrode assembly (MEA)-based PEMWE was constructed with Se-RuO_x as the anode catalyst, commercial Pt/C as the cathode catalyst, and Nafion®211 as the proton-exchange membrane (Fig. 7a and Fig. S49). As a benchmark, a similar PEMWE was fabricated using other R-RuO_x as the anode catalyst. All device performances were evaluated in 0.5 M H₂SO₄ at 25 °C without iR correction. Compared with Fig. S50, to reach a water-splitting current density of 1 A cm⁻², the Se-RuO_x-based PEMWE required a full-cell voltage of 1.84 V, which was much lower than those of other R-RuO_x-based PEMWE. In addition, the activity of Se-RuO_x gradually increased with increasing temperature (Fig. 7b). When the temperature reaches 80 °C (the industrial level), the full-cell voltage of the Se-RuO_x-based PEMWE is 1.59 V at 0.5 A cm⁻², 1.67 at 1 A cm⁻² and 1.74 V at 1.5 A cm⁻², highlighting its competitive performance in PEM electrolysers. Moreover, the Faraday efficiencies for O₂ generation were evaluated and showed excellent agreement with the theoretical values based on the O₂ molar ratio, thereby confirming a Faraday efficiency approaching almost 100% (Fig. S51). More impressively, the Se-RuO_x-based PEMWE demonstrated markedly improved device stability at 1 A cm⁻² in 0.5 M H₂SO₄. As shown in Fig. 7c, the Se-RuO_x-based PEMWE can operate stably for 1000 h, with a degradation rate of only 58 μV•h⁻¹. The voltage of the RuO₂-based PEMWE increased by 600 mV in just a few hours. In contrast, Se-RuO_x demonstrates notable performance over the reported Ru-based OER catalysts in terms of the ampere-level current density and long-term stability (Table S6). These results underscore the notable OER activity and stability metrics of the Se-RuO_x catalyst at an industrial-level current density in PEM devices.

a Schematic illustration of the designed PEMWE water electrolyzer. b LSV curves of the PEMWE water electrolyzer for Se-RuO_x at 25, 40, and 80 °C. c Durability cell voltage‒time plots for the PEMWE water electrolyzer at a constant current density of 1000 mA cm⁻² for 1000 h.

Full size image

In summary, we pioneer the doping of different p-orbital atoms (N, P, S, Se) into RuO₂ (R-RuO_x) with a simple one-step calcination synthesis method to investigate the electronic structure‒mechanism relationship of Ru-based materials and the construction principle of the LOM‒OVSM path. Extensive experimental and theoretical evidence has confirmed that when p-orbital Se atoms are doped, electrons are more likely to diffuse from the O-p orbitals to the Ru-e_g orbital, accelerating their diffusion to the external circuit and resulting in a faster LOM-mediated process. When abundant oxygen vacancy (V_O) forms during the LOM process, Se doping can induce an increase in Ru-e_g band broadening, promoting electron return and stabilizing the V_O, leading to increased energy of the Ru-t_2g orbital and decreased energy of the O-p orbital, causing electrons to transfer from the Ru-t_2g orbital to the Ru-e_g orbital and thereby promoting the OVSM process. As a result, the representative Se-RuO_x exhibited a low overpotential (188 mV at 10 mA cm⁻²) and stable PEMWE performance (1.67 V to deliver 1 A cm⁻²) for 1000 h. Overall, the synergistic interplay of the Se-RuO_x catalyst with rapid electron diffusion and recovery capabilities provides prospective insight into rationally creating LOM-OVSM catalysts that notably enhance both OER activity and stability.

Chemicals and materials

All chemicals were used as received. Ruthenium(III) chloride (RuCl₃, ≥99.9%), selenium (Se, ≥99%), sulfur sublimation (S, ≥99%), sodium monophosphate (NaH₂PO₂, ≥99%), Vulcan XC 72 carbon, urea (CH₄N₂O, ≥99%), absolute ethanol (C₂H₆O, ≥99.7%), Nafion117 (≈5%), and isopropanol (C₃H₈O, ≥99.7%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Carbon cloth (WOS 1011) was obtained from CeTech. High-purity water was acquired from the Milli-Q purification system.

Fabrication of RuO₂

First, 0.5 mmol of RuCl₃ was dissolved in 4 mL of absolute ethanol, and then the solution was mixed well by stirring. The product was collected by heating to 600 °C using a muffle furnace at a ramp rate of 5 °C/min for 120 min and then cooling to room temperature at a ramp rate of 5 °C/min.

Fabrication of R-RuO_x (R = P, S, N, Se) and Se-RuO_x-y (y = 1, 2, 4, 6)

The synthesis of R-RuO_x followed the same procedure as the previously described synthesis method, with an additional 0.02 mmol of Se (or S, CH₄N₂O, NaH₂PO₂) in anhydrous ethanol.

The synthesis of Se-RuO_x-y (y = 1, 2, 4, or 6) followed the same procedure as the synthesis method, with different Se powder contents (0, 0.01, 0.02, 0.04 and 0.06 mmol, respectively) in anhydrous ethanol.

Characterizations

X-ray diffraction (XRD) patterns were collected on an Empyrean PANalytical diffractometer with Cu-Kα radiation (40 kV, 40 mA, λ = 1.5418 Å) at a scanning rate of 5° min⁻¹. JEM-2100F was used to acquire the TEM, HAADF-STEM, and EDS mapping images. A Thermo Scientific K-Alpha+ instrument was used to collect X-ray photoelectron spectrometry (XPS) information. The calibration peak data were based on the C 1 s peak at 284.8 eV. All vibrating sample magnetometer (VSM) data were collected via a Quantum Design-PPMS9. The magnetic hysteresis (M-H) curve was measured at ±3 T magnetic field strength at room temperature, and the temperature-dependent magnetization (M-T) curve was measured at H = 1 K Oe magnetic field strength at temperatures ranging from 2–300 K. Electron paramagnetic resonance (EPR) was performed on a Bruker A300 at 77k temperature to obtain oxygen vacancy correlation information. The g value can be calculated from the original data via the following equation: g = hν/βH, where h is the Plank constant with a value of 4.135 × 10⁻¹⁵eV s, ν is the microwave frequency of the X-band spectrometer, β is the electron Bohr magneton with a value of 5.788×10⁻⁵eV T⁻¹, and H is the applied magnetic field. For X-ray absorption spectroscopy (XAS) analysis, test samples were prepared on titanium foil and subsequently brushed off to collect powders. The resultant powders were evenly applied onto a 10 cm long and 1 cm wide piece of plastic, radiation-resistant Scotch tape (commonly known as Kapton tape, often referred to as the duct tape of synchrotrons). In order to obtain a compact particle distribution on the Scotch tape without any pinholes, following the removal of larger particles by shaking, the sample tape was divided into 10 smaller segments. These segments were then layered one on top of the other onto a separate piece of Scotch tape and subsequently encapsulated with an additional layer of Scotch tape. The Ru K-edge X-ray absorption spectra (XAS) were collected at the 1W1B beamline of the Beijing Synchrotron Radiation Facility in the People’s Republic of China, which operates at a current of 200 mA and an energy of 2.5 GeV. Ru foil served as the calibration standard, with all specimen analyses conducted using transmission-mode spectroscopy. The XAFS data were processed utilizing the Demeter software package. Normalization of the spectra was initially carried out using Athena, followed by shell fitting analyses performed with Artemis. The χ(k) function was subjected to Fourier transformation (FT) employing k3 weighting, and all fitting procedures were conducted in R space.

In situ Raman

In situ Raman spectroscopic measurements were conducted through a confocal microscopy system (DXR platform) interfaced with an electrochemical flow cell. Experimental configurations utilized a 10× magnification objective and 2.0 mW laser excitation energy. The working electrode was Se-RuO_x/Au, the counter electrode was a platinum wire, and the reference electrode was a saturated calomel electrode. The flow rate of the electrolyte was 5 mL min⁻¹ to remove the oxygen produced in the reaction process. After applying a constant potential for 20 s, the sample information was gathered, and the in situ Raman spectra of each working electrode were captured three times. Within the voltage range of 0.9–2.0 V vs. SCE, a variation in the Raman spectra became evident.

In situ ATR-SEIRAS

In situ FTIR experiments were carried out with a custom-designed cell. The catalysts were deposited on a GCE positioned at the center of the cell, and a Pt wire as well as a saturated calomel electrode were used as the counter and reference electrodes, respectively. In order to avoid electrolyte evaporation that could damage the microscope lens, a BaF₂ window was positioned above the cell. The OER was conducted with linearly increasing potentials from open-circuit potential (OCP) to 1.85 V vs. RHE at fixed polarization, during which synchrotron radiation Fourier-transform infrared (SR-FTIR) spectral data were collected through 100-scan signal accumulation with 1 cm⁻¹ spectral resolution. First, the background spectrum was recorded without applying any voltage. During data processing, the background spectrum was subtracted from the infrared spectrum of the sample.

Electrochemical measurements

The OER catalytic performance was evaluated in a three-electrode system on a CHI760e electrochemical workstation at 25 °C. Electrolyte solutions are generally prepared fresh before use. Three milligrams of R-RuO_x were dispersed in 600 μL of 0.1% Nafion solution (diluted with isopropanol) and ultrasonicated for 20 min to disperse uniformly, after which 21 μL of the resulting suspension was deposited onto a glassy carbon (GC, d = 0.5 cm) electrode. All control samples exhibited comparable mass loading. Briefly, the synthesized catalysts were used as working electrodes, while a carbon rod and a saturated calomel electrode (SCE, Hg/Hg₂Cl₂) functioning as the counter and reference electrode, respectively. Linear sweep voltammetry (LSV) curves were conducted in a 0.5 M H₂SO₄ solution (pH≈0.3) at a scan rate of 5 mV/s with 85% inner resistance (iR) compensation. E (vs. RHE) = E (vs. SCE) + 0.059*pH + 0.241. All the chronopotentiometric stability tests were evaluated in 0.5 M H₂SO₄ using a standard three-electrode system with a carbon cloth electrode. All OER performances were evaluated in 0.5 M HSO₄ with iR correction (85% iR-correction set via the instrument). For Tafel analysis, the potential corresponding to a current density near 10 mA cm⁻² was chosen. The Faradaic efficiency was determined by gathering O₂ and H₂ at a constant current of 0.1 A (25 °C, 1 atm). Electrochemical impedance spectroscopy (EIS) was conducted across a frequency spectrum extending from 100 kHz to 0.01 Hz with applied voltages set at 1.25 V, 1.3 V, 1.35 V, 1.4 V, 1.45 V, 1.5 V, and 1.55 V. Electrochemically active surface areas (ECSAs) were determined using cyclic voltammetry (CV) within a potential range of 1.0–1.1 V versus SCE, at scan rates of 10, 20, 30, 40, and 50 mV s⁻¹. The ECSA was calculated based on the current density measured at 1.05 V versus SCE. The electrical double-layer capacitance (C_dl) was estimated using the formula C_dl = (j_a-j_c)/2ν, where ja and jc represent the anodic and cathodic current densities, respectively, and ν denotes the scan rate. Consequently, C_dl corresponds to the slope of the linear relationship between (j_a-j_c)/² and the scan rate. The ECSA was then derived using the equation ECSA = C_dl/C_s (C_s = 0.040 mF cm⁻²).

PEMWE electrolyser

Membrane electrode assemblies (MEAs) were fabricated utilizing a Nafion®211 polymer membrane (active area of 1 cm², film thickness: 25.4 μm). Se-RuO_x (2 mg cm⁻²) and commercial Pt/C (1 mg cm⁻²) catalysts were used as the anode and cathode, respectively. Electrochemical evaluation was carried out in a single-cell PEM water electrolyzer (PEMWE) configuration. Porous titanium felt layers served as both anode and cathode gas diffusion media. For durability analysis, stability measurements were executed using a Keithley 2461 Source Meter instrumentation platform.

DFT calculation details

DFT calculations were carried out using the VASP, employing the Perdew–Burke–Ernzerhof (PBE) functional within the generalized gradient approximation (GGA) and the projector augmented wave (PAW) method. The interaction between core and valence electrons was described using PAW approach⁵⁷. The crystal structure of RuO₂ is shown in Fig. S52, and DFT-optimized equilibrium lattice constants are a = 4.492, b = 4.492, c = 3.107 Å, which is inconsistent with previous reports⁵⁸. To discuss the impact on the doping sites of Se atom, we substituted three Se atoms along the x, y, and z directions to simulate Se-RuO_x. Additionally, we substituted 1, 2, and 3 Se atoms at the Ru-Ru sites along the z-direction to verify the effect of different doping ratio on Ru. Periodically repeated three-layer slab is modeled for RuO₂ (110) as the active surfaces to investigate the OER performance. The bottom layer of the slab was fixed during structural optimization, while the top two layers were relaxed, using a 4×4×1 Monkhorst–Pack k-point mesh for Brillouin zone sampling. A plane-wave basis set with a kinetic energy cutoff of 450 eV was employed to expand the valence electron wavefunctions, and a Gaussian smearing width of 0.05 eV was utilized during structural optimization. The electronic self-consistent field (SCF) calculations were considered converged when the total energy change was below 10⁻⁵ eV, and atomic structures were optimized until the residual forces were smaller than 0.02 eV/Å. For Ru atoms, an effective Hubbard parameter (U–J) of 4.2 eV was applied.

The OER process via LOM-mediated process can be represented by the following stepwise reactions^59,60:

As a result, the Gibbs free energy variations were determined as follows:

The free energy of O₂ was derived from the reference reaction: H₂O(l) → H₂(g) + O₂(g).

The overpotential (η), a key descriptor of OER activity, can be directly obtained based on the following proposed OVSM pathway for the oxygen evolution reaction^59,60:

where Vo-M* denotes the active site involved in the reaction; G₁, ΔG₂, ΔG₃, and ΔG₄ correspond to the free energy changes for each individual step; and

and η = max (ΔG₁, ΔG₂, ΔG₃, ΔG₄) – 1.23.