Modulating the covalency of Ru-O bonds by dynamic reconstruction for efficient acidic oxygen evolution

Abstract

Developing ruthenium-based oxide catalysts capable of suppressing lattice oxygen participation in the catalytic reaction process is crucial for maintaining stable oxygen evolution reaction (OER) under acidic conditions. Herein, we delicately construct a RuO₂ nanoparticle-anchored LiCoO₂ nanosheet electrocatalyst (RuO₂/LiCoO₂), achieving dynamic optimization of RuO₂ during the reaction process and improving catalytic stability. Benefiting from the unique electrochemical delithiation characteristics of the LiCoO₂ support, the covalency of the Ru-O bond is effectively regulated during the OER process. The weakened Ru-O covalent bond inhibits the participation of lattice oxygen in the catalytic reaction and ensures the continuous operation of the Ru active sites. Moreover, the extended Ru-O bond in the optimized RuO₂/LiCoO₂ catalyst reduces the formation energy barrier of the *OOH intermediates, accelerating the progress of the OER. As a result, the RuO₂/LiCoO₂ catalyst requires only an overpotential of 150 ± 2 mV at 10 mA cm⁻² in 0.5 M H₂SO₄ and operates stably for 2000 h at 1 A cm⁻² in a proton exchange membrane water electrolysis. This work opens new avenues for designing efficient ruthenium-based catalysts.

Introduction

Proton exchange membrane water electrolysis (PEMWE) with high current density and low resistance loss is regarded as a promising hydrogen production technology in the future¹. Currently, noble metal ruthenium (Ru) and iridium (Ir) oxide catalysts are extensively used in the anodes of PEMWE^2,3. Compared to IrO₂, RuO₂ offers high activity and cost-effectiveness, presenting significant potential for application in acidic oxygen evolution reactions (OER)^4,5. However, RuO₂ catalysts tend to follow the lattice oxygen mechanism (LOM) and generate numerous oxygen defects during the OER reaction, leading to crystal structure collapse^6,7. Additionally, the metal Ru sites can be over-oxidized into soluble RuO₄ species, which separate from the crystal lattice under high oxidation potential, resulting in poor catalytic stability^8,9. From a crystal structure perspective, the instability of RuO₂ is closely related to the charge distribution of the Ru-O bond^10,11,12,13. Therefore, regulating the electronic state of the Ru-O bond to suppress the involvement of lattice oxygen in the OER process is an effective strategy for enhancing the activity and stability of RuO₂ catalysts.

Generally, electron-rich Ru sites in RuO₂ activate lattice oxygen and generate defects, while electron-deficient states tend to oxidize to excessively high valence states and dissolve^14,15. Traditional electron-donating support strategies modulate the electron distribution of the Ru-O bond, preventing the dissolution of the metal center and following a relatively stable adsorption oxygen mechanism (AEM)^16,17. At present, electron-donating supports with limited regulatory capacity cannot meet the demand for manipulating dynamically changing Ru-O bonds under complex high-potential OER conditions^18,19. Therefore, designing supports capable of dynamically regulating the electronic structure of RuO₂ in response to changes in the catalytic reaction process is highly attractive. To address this challenge, transition metal oxides with cation intercalation dynamically optimize the local microenvironment of metal-oxygen bonds during the catalytic reaction through the extraction and insertion of cations^20,21. Notably, lithium cobalt oxide (LiCoO₂) with a unique layered structure exhibits an ordered cation arrangement and superior thermodynamic stability^22,23,24. The tight edge-shared CoO₆ octahedral structure in LiCoO₂ reduces the migration energy barrier of Li⁺ and ensures two-dimensional diffusion of Li⁺ in the plane, facilitating dynamic reconstruction during the catalytic process^25,26. However, the multi-layered bulk structure of LiCoO₂ exhibits slow electron transport capability and limited surface area, which cannot meet the requirements for efficient catalyst support²⁵. In this regard, two-dimensional LiCoO₂ nanosheets markedly shorten the intercrystalline electron transmission path and enhance conductivity²⁷. Additionally, the two-dimensional nanosheet structure forms a fast electron transmission channel perpendicular to the exposure plane, resulting in strong metal-support interaction^28,29. Therefore, using two-dimensional LiCoO₂ nanosheets as a support for RuO₂ is expected to achieve dynamic regulation of the electronic structure during catalytic reactions and exhibit high catalytic activity.

Herein, we propose an effective strategy to improve the charge distribution of RuO₂ through the unique dynamic evolution process of the LiCoO₂ support (RuO₂/LiCoO₂). The LiCoO₂ nanosheet support with electron-donating ability induces electron transfer from Co to Ru sites, providing electron compensation to stabilize the valence state of RuO₂. More importantly, the two-dimensional diffusion and extraction of Li ions within the interlayer of the LiCoO₂ nanosheet under OER potential cause the dynamic reconstruction and evolution of the RuO₂/LiCoO₂ catalyst interface. The unique dynamic self-optimization process moderately weakens the covalency of the Ru-O bond, suppressing the participation of lattice oxygen and achieving a good balance between catalytic activity and stability. The optimized Ru sites facilitate the formation of the *OOH intermediate, significantly lowering the catalytic energy barrier of the rate-determining step. Consequently, the RuO₂/LiCoO₂ electrocatalyst provides a current density of 10 mA cm⁻² at an overpotential of 150 ± 2 mV and maintains stability for over 2300 h in 0.5 M H₂SO₄. In addition, RuO₂/LiCoO₂ as anode can operate continuously for 2000 h at 1 A cm⁻² in a PEM electrolyzer. This work advances the application of ruthenium-based catalysts in PEMWE.

Results

Reaction mechanism and structural analysis

Due to the differences in the covalency of Ru-O bonds, Ru oxides follow either the lattice oxygen mechanism (LOM) or the adsorption oxygen mechanism (AEM) during the acidic OER process (Fig. 1a, b)^7,30. In the LOM pathway, activated lattice oxygen with stronger covalency participates in the OER process, resulting in the generation of oxygen defects³¹. An excessive number of oxygen vacancies can induce detachment from the crystal lattice at the Ru metal sites, which greatly reduces the catalytic stability. However, Ru oxides with weak covalent Ru-O bonds follow the AEM mechanism, achieving a stable acidic OER process¹³. Therefore, dynamically regulating the covalency of Ru-O bonds during the catalytic reaction is crucial for enhancing catalyst stability. The O3-type LiCoO₂ material with Li element intercalation exhibits superior conductivity and structural stability. More importantly, the extraction of interlayer lithium ions at the OER potential triggers electron transfer, enabling real-time regulation of the covalency of metal-oxygen bonds (Fig. 1c). Additionally, we employ density functional theory (DFT) calculations to study the stability of different Co-based oxides in acidic environments. LiCoO₂ presents a thermodynamically unfavorable dissolution barrier, indicating that it can still maintain stability in an acidic environment (Fig. 1d). Inspired by these results, dynamically regulating the Ru-O bond covalency in RuO₂ using the layered LiCoO₂ support is expected to achieve high catalytic stability.

Fig. 1: Schematic diagram of OER mechanism and structural analysis of lithium cobalt oxide.

a LOM schematic of RuO₂ in acidic OER. b AEM schematic of RuO₂ in acidic OER. c Structural models of LiCoO₂ and Li_1−xCoO₂. d Dissolution energy barriers of cobalt-based oxides.

Materials characterization

A facile adsorption calcination strategy is employed to prepare LiCoO₂ nanosheets with loaded RuO₂ nanoparticles (RuO₂/LiCoO₂). Briefly, LiCoO₂ nanosheets are obtained through ultrasonic-assisted exfoliation of solid-phase synthesized bulk LiCoO₂ (Supplementary Figs. 1, 2). Subsequently, ruthenium ions are adsorbed on the LiCoO₂ surface through electrostatic binding and then calcined to form the RuO₂/LiCoO₂ electrocatalyst. According to the scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images, RuO₂/LiCoO₂ exhibits an ultrathin nanosheet structure (Fig. 2a and Supplementary Figs. 3–5). Interestingly, the RuO₂ nanoparticles are tightly confined in the LiCoO₂ lattice in the high-resolution TEM (HRTEM) image, which is different from simple physical adsorption (Fig. 2b). A clear lattice contact exists between RuO₂ nanoparticles and the LiCoO₂ support, in which lattice stripes with interplanar spacings of 0.31 and 0.20 nm correspond to the RuO₂ (110) and LiCoO₂ (104) planes, respectively (Fig. 2c, h)²⁵. The LiCoO₂ support provides lattice confinement for the RuO₂ nanoparticles to form a unique metal-support interaction. Atomic force microscopy (AFM) shows that the thickness of the RuO₂/LiCoO₂ nanosheet is 5.1 nm in Fig. 2d. The ultrathin nanosheet structure of RuO₂/LiCoO₂ maximizes the exposure of more Ru active sites. Atomic-resolution aberration-corrected high-angle annular dark-field scanning TEM (HAADF-STEM) is used to analyze the atomic structure of the RuO₂/LiCoO₂ composite. The Co atoms are arranged in an αβγ stacking mode in the HAADF-STEM image along the [010] zone axis, indicating the formation of O3-type LiCoO₂ and consistent with the XRD results (Fig. 2e and Supplementary Figs. 6, 7)²³. The spacing of about 0.47 nm corresponds to the typical interlayer spacing of LiCoO₂. In addition, the bright Ru atomic array is embedded in the lattice of LiCoO₂ and exhibits superb lattice matching at the interface (Fig. 2f). The distribution of Ru, Co, and O elements in the energy dispersive spectroscopy (EDS) spectrum further confirms the uniformity of the heterostructure (Fig. 2g). The above analysis shows that the structurally ordered LiCoO₂ nanosheet support is closely connected to the RuO₂ nanoparticles, which provides the possibility for the optimization of the electronic structure.

Fig. 2: Characterizing the structure of RuO₂/LiCoO₂.

a TEM, (b, c) HRTEM, and (d) AFM images of RuO₂/LiCoO₂ (inset: height profile of RuO₂/LiCoO₂). e, f HAADF-STEM images of RuO₂/LiCoO₂. g The EDS elemental mapping of RuO₂/LiCoO₂. h The atomic model of RuO₂ and LiCoO₂. i Normalized Ru K-edge XANES spectra and (j) Ru K-edge FT-EXAFS spectra of RuO₂/LiCoO₂, RuO₂, and Ru foil, respectively. k Normalized Co K-edge XANES spectra and (l) Co K-edge FT-EXAFS spectra of RuO₂/LiCoO₂, LiCoO₂, Co₃O₄, and Co foil, respectively.

To further understand the coordination structure and electronic states of RuO₂/LiCoO₂, X-ray absorption spectroscopy (XAS) was performed. As shown in Fig. 2i, the oxidation state of RuO₂/LiCoO₂ is lower than that of RuO₂ in the Ru K-edge X-ray absorption near-edge structure (XANES) spectra, implying the formation of an electron transfer channel between RuO₂ and LiCoO₂³². This is also consistent with the results of the first derivative of the Ru K-edge XANES spectra and high-resolution X-ray photoelectron spectroscopy (XPS) (Supplementary Figs. 8, 9). The LiCoO₂ nanosheet support with charge compensation effectively regulates the oxidation state of Ru sites, preventing excessive oxidation during the catalytic process. According to the Fourier transformed Ru K-edge extended X-ray absorption fine structure (FT-EXAFS) spectra, the Ru-O bond length is slightly shortened after contact between RuO₂ and LiCoO₂ (from 1.93 to 1.91 Å), which is attributed to the formation of interfacial interactions (Fig. 2j and Supplementary Figs. 10, 11 and Supplementary Tables 1, 2)³³. The Co K-edge XANES of RuO₂/LiCoO₂ is transferred to higher energies compared to LiCoO₂ in Fig. 2k, which is consistent with the XPS valence band spectrum and electron paramagnetic resonance (EPR) results. This verifies the electron-donating ability of the LiCoO₂ carrier (Supplementary Figs. 12, 13). Moreover, the main peaks of RuO₂/LiCoO₂ at 1.4 and 2.4 Å correspond to Co-O and Co-Co coordination shells in the Co K-edge FT-EXAFS spectra, respectively (Fig. 2l)^34,35. It is worth noting that the Co-O bonds in RuO₂/LiCoO₂ are stretched more than LiCoO₂, which is consistent with the wavelet transform spectral results (Supplementary Figs. 14–16). The electron transfer from Co to Ru are verified by collecting electron energy loss spectra (EELS) of Co-L_2,3 and Ru-M_2,3 at the RuO₂/LiCoO₂ interface (Supplementary Fig. 17). Therefore, the interfacial interaction between RuO₂ and LiCoO₂ electronically compensates the Ru sites and improves the stability of the electrocatalyst.

OER electrocatalytic performance

Inspired by the structural advantages of the RuO₂/LiCoO₂ catalyst, the OER performance was evaluated via a three-electrode system in 0.5 M H₂SO₄. In Fig. 3a, the RuO₂/LiCoO₂ electrocatalyst exhibits an overpotential of 150 ± 2 mV at 10 mA cm⁻², which is much smaller than commercial RuO₂ (260 ± 4 mV) (denoted Com-RuO₂) and RuO₂ (210 ± 5 mV) (Supplementary Figs. 18–24). In comparison, the highly acidic OER activity of RuO₂/LiCoO₂ also surpasses most previously reported noble metal-based electrocatalysts (Supplementary Figs. 25 and Supplementary Table 3). The Tafel slope of RuO₂/LiCoO₂ (51.97 mV dec⁻¹) shows improved reaction kinetics compared to RuO₂ (62.06 mV dec⁻¹) and Com-RuO₂ (92.80 mV dec⁻¹) (Fig. 3b)³⁶. At the OER potential, RuO₂/LiCoO₂ exhibits fast intermediate conversion efficiency. Meanwhile, the phase angle of RuO₂/LiCoO₂ rapidly decreases at different potentials in the Bode plot, further confirming the rapid charge diffusion on the catalyst surface (Supplementary Figs. 26–28 and Supplementary Table 4). The mass activity of the RuO₂/LiCoO₂ catalyst is 41.81 times that of RuO₂ at an overpotential of 240 mV in Fig. 3c. In addition, the turnover frequency (TOF) and OER current normalized by the electrochemically active surface area (ECSA) of RuO₂/LiCoO₂ electrocatalysts are significantly higher than RuO₂ (Supplementary Figs. 29–32 and Supplementary Table 5). The catalyst durability is tested using chronopotentiometry to evaluate the potential for practical applications. The RuO₂/LiCoO₂ electrocatalyst can operate stably for 2300 h at a current density of 10 mA cm⁻², while RuO₂ almost loses its activity after 400 h (Fig. 3d and Supplementary Fig. 33). The above results show that the LiCoO₂ support in the RuO₂/LiCoO₂ electrocatalyst avoids the dissolution of RuO₂ in acidic OER to achieve sustained acidic OER catalytic activity.

Fig. 3: Electrocatalytic performance of RuO₂/LiCoO₂.

a OER polarization curves and (b) Tafel plots of RuO₂/LiCoO₂, RuO₂, Com-RuO₂, and LiCoO₂ in 0.5 M H₂SO₄ electrolyte, respectively. The voltage is corrected by an automatic 90% of iR compensation (R is 1.10 ± 0.02 Ω). c Mass activity of Ru atoms in RuO₂/LiCoO₂ and RuO₂ as a function of overpotential. d Chronopotentiometric curves of RuO₂/LiCoO₂ and RuO₂ at 10 mA cm⁻², respectively (Inset: Potential changes of RuO₂/LiCoO₂ before and after stabilization). e The polarization curves of PEMWE with RuO₂/LiCoO₂ | |Pt/C and RuO₂ | |Pt/C catalyst in pure water at 80 °C without iR-correction. f Chronopotentiometric curve of PEMEW using RuO₂/LiCoO₂ | |Pt/C catalyst at 1 A cm⁻².

To evaluate the industrial application potential of RuO₂/LiCoO₂, a PEM electrolyzer was assembled with RuO₂/LiCoO₂ and Pt/C as anode and cathode catalysts, respectively. Specifically, RuO₂/LiCoO₂ | |Pt/C only requires a cell voltage of 1.68 V to reach a current density of 1 A cm⁻², which is significantly lower than RuO₂ | |Pt/C (1.84 V) (Fig. 3e). Overpotential analysis shows that improved mass transport alleviates concentration polarization effects in the local reaction environment, indirectly enhancing catalytic kinetics (Supplementary Fig. 34). Furthermore, the mass activity of RuO₂/LiCoO₂ | |Pt/C (2.56 A mg_Ru⁻¹) is approximately 21.33 times greater than that of RuO₂ | |Pt/C (0.12 A mg_Ru⁻¹) at a cell voltage of 1.7 V (Supplementary Fig. 35). The mass activity and cost of RuO₂/LiCoO₂ | |Pt/C are also much lower than those of commercial IrO₂ | |Pt/C (Supplementary Fig. 36). Strikingly, the PEM electrolyzer using RuO₂/LiCoO₂ | |Pt/C can operate stably for 2000 h at a current density of 1 A cm⁻² with negligible decay in Fig. 3f, indicating the potential of the RuO₂/LiCoO₂ catalyst for practical applications. The long-term durability of RuO₂/LiCoO₂ | |Pt/C in PEMWE also exceeds that of most recently reported various high-performance electrocatalysts (Supplementary Table 6). These findings demonstrate that the RuO₂/LiCoO₂ electrocatalyst exhibits outstanding performance as a choice for hydrogen production in the PEM electrolyzer, providing strong support for sustainable energy production.

Structural transformation after stability

The physical structure of the RuO₂/LiCoO₂ electrocatalyst after OER stability testing was investigated to explore the influencing factors for the improved catalytic activity. SEM and TEM images show that RuO₂ nanoparticles remain tightly supported on the surface of LiCoO₂ nanosheets in the RuO₂/LiCoO₂ electrocatalyst after stability testing (Supplementary Figs. 37, 38). Besides, a reconstructed amorphous layer can be clearly observed on the surface of LiCoO₂ nanosheets in RuO₂/LiCoO₂. This layer is attributed to the migration and extraction of interlayer Li-ions during voltage application, resulting in the formation of a surface Li_1−xCoO₂ amorphous layer. Compared with the original RuO₂/LiCoO₂, the (003) crystal plane peak of RuO₂/LiCoO₂ after OER stability shifts to a lower angle, which may be attributed to the lattice distortion caused by the extraction of Li ions (Fig. 4a and Supplementary Fig. 39). Additionally, the Raman spectrum of RuO₂/LiCoO₂ detects two characteristic peaks of E_g and A_1g modes at 484 and 594 cm⁻¹, corresponding to O-Co-O bending and Co-O stretching, respectively (Fig. 4b and Supplementary Fig. 40). Due to the enhanced polarity of the Co-O bond induced by Li extraction, the E_g and A_1g characteristic peaks of RuO₂/LiCoO₂ exhibit a blue shift after OER stability (Supplementary Fig. 41)²³. In particular, the new peak at 665 cm⁻¹ after stabilization is attributed to the Co-O bonds of the spinel structure Li_1−xCoO₂ (LiCo₂O₄) of the surface reconstruction layer, which enhances the polarity of the Co-O bonds. The interlayer Li ions detach from RuO₂/LiCoO₂ through a two-dimensional diffusion path under the OER potential, performing dynamic structural reconstruction and altering the interface coordination environment of the catalyst.

Fig. 4: Reconstruction of RuO₂/LiCoO₂ catalyst after stabilization.

a XRD patterns and (b) Raman spectra of RuO₂/LiCoO₂ and RuO₂/LiCoO₂ after stability, respectively. c The content of Co²⁺, Co³⁺, Ru³⁺, and Ru⁴⁺ in RuO₂/LiCoO₂ before and after OER stability. d Bader charge analysis of RuO₂, LiCoO₂, RuO₂/LiCoO₂, and RuO₂/Li_1−xCoO₂, respectively. e Comparison of charge amounts of RuO₂, LiCoO₂, RuO₂/LiCoO₂, and RuO₂/Li_1−xCoO₂, respectively. f COHP of Ru-O in RuO₂/LiCoO₂ and RuO₂/Li_1−xCoO₂, respectively. g Schematic diagram of the dynamic evolution of RuO₂/LiCoO₂.

The electronic states of RuO₂/LiCoO₂ catalysts before and after OER stability are further analyzed to understand the influence of the reconstruction process. The Co³⁺ ratio of RuO₂/LiCoO₂ after OER stabilization increases compared with the pristine RuO₂/LiCoO₂, implying the reduction of electron density for Co sites (Fig. 4c and Supplementary Fig. 42)^37,38. The dynamic extraction of Li ions induces electron transfer, leading to partial oxidation of Co sites and increased covalency in the Co-O bond. Additionally, the proportion of Ru⁴⁺ in RuO₂/LiCoO₂ slightly increases from 58.02% to 61.17% after the stability measurement³⁹. As a comparison, the proportion of Ru⁴⁺ in RuO₂ after the stability measurement is 85.24%, which is much higher than that of RuO₂/LiCoO₂ (Supplementary Figs. 43, 44). Despite the slight oxidation of the LiCoO₂ support, it can maintain the oxidation state of RuO₂ by continuously donating electrons, thereby preventing the dissolution of Ru sites (Supplementary Figs. 45, 46). A structural model of RuO₂/LiCoO₂ after partial delithiation (RuO₂/Li_1−xCoO₂) is constructed to reveal the changes in electronic structure by DFT calculations (Supplementary Fig. 47). The Bader charge indicates that Ru in RuO₂/LiCoO₂ obtains about 0.169 e of electrons from the LiCoO₂ support (Fig. 4d)^40,41. Due to the extraction of Li ions, the charge density at the Ru sites in RuO₂/Li_1−xCoO₂ (6.584 e) is lower than that of RuO₂/LiCoO₂ (6.626 e), while remaining higher than that of RuO₂ (6.457 e) (Fig. 4e). Crystal orbital Hamiltonian population (COHP) analysis shows that the Ru-O bond covalency of RuO₂/Li_1−xCoO₂ (−4.07 eV) is lower than that of RuO₂/LiCoO₂ (−4.35 eV) (Fig. 4f)^42,43. The unique delithiation process ensures dynamic regulation of the covalency of the Ru-O bond during the catalytic process to suppress the participation of lattice oxygen (Fig. 4g).

In situ characterization of catalyst structural changes

In-situ XAS was implemented to explore the dynamic evolution mechanism of electrocatalysts in the OER process (Supplementary Fig. 48). In the fitted Ru K-edge FT-EXAFS spectra, the Ru-O bond of RuO₂/LiCoO₂ is appropriately extended as the potential increases from open circuit voltage (OCV) to 1.7 V (from 1.90 to 1.99 Å), indicating that the covalency of Ru-O is weakened (Fig. 5a, b and Supplementary Fig. 49 and Supplementary Table 7)⁴⁴. In contrast, the Co-O bond tends to be more covalent during the delithiation process, resulting in a negative shift of the Co-O coordination shell in RuO₂/LiCoO₂ and the formation of a surface reconstruction layer (from 1.92 to 1.82 Å) (Supplementary Figs. 50, 51 and Supplementary Table 8)⁴⁵. The wavelet transform of Ru K-edge and Co K-edge EXAFS spectra also intuitively confirm this result (Supplementary Figs. 52–54). Thus, the dynamic extraction of lithium ions regulates the covalency of the Ru-O bond through interfacial interactions, which can restrict the participation of lattice oxygen and maintain the structural stability of RuO₂ (Fig. 5c). Notably, the Ru-O bond in RuO₂ undergoes a significant shortening from 1.93 Å to 1.81 Å as the voltage transitions from OCV to 1.7 V (Fig. 5d, e and Supplementary Figs. 55–57 and Supplementary Table 9). The excessive enhancement of the covalency of the Ru-O bond provides conditions for triggering lattice oxygen to participate in the OER reaction⁴⁴. Moreover, the slight change in the Ru-O coordination number in RuO₂/LiCoO₂ indicates the stability of the coordination structure of RuO₂. As a comparison, a reduced coordination number can be observed for the Ru-O bonds in RuO₂, which is attributed to the presence of defective oxygen (Fig. 5f). Due to the strong covalency of the Ru-O bond, RuO₂ follows the LOM path and generates a large number of oxygen defects during the OER process, leading to the collapse of the catalyst surface structure.

Fig. 5: In-suit XAS characterization of RuO₂/LiCoO₂ confirming the dynamic process.

a In situ Ru K-edge EXAFS spectra of RuO₂/LiCoO₂ with applied potentials from OCV to 1.7 V. b Changes in bond length and intensity of the Ru-O shell of RuO₂/LiCoO₂ at different potentials. c Schematic diagram of the structural evolution of RuO₂/LiCoO₂ at different potentials. d In situ Ru K-edge EXAFS spectra of RuO₂ with applied potentials from OCV to 1.7 V. e Changes in bond length and intensity of the Ru-O shell of RuO₂ at different potentials. f Schematic diagram of the structural evolution of RuO₂ at different potentials. g In situ Ru K-edge XANES spectra of RuO₂/LiCoO₂. Inset: Oxidation state changes of Ru in RuO₂/LiCoO₂. h In situ Ru K-edge XANES spectra of RuO₂. Inset: Oxidation state changes of Ru in RuO₂.

To elucidate the electronic structure changes of the electrocatalyst at high potentials, the oxidation states of RuO₂/LiCoO₂ and RuO₂ are detected by in situ XANES spectra. The average valence state of Ru species in RuO₂/LiCoO₂ rises gently from +3.82 to +4.21 as the potential transitions from OCV to 1.7 V vs. RHE in Fig. 5g (Supplementary Fig. 58). This further confirms that the strong electronic interaction between the LiCoO₂ and RuO₂ interface inhibits the excessive oxidation of RuO₂. In addition, the Co species in the RuO₂/LiCoO₂ catalyst can still maintain an oxidation state of +3.16 at a high voltage of 1.7 V (Supplementary Figs. 59, 60). As a result, the extraction of lithium ions can moderately regulate the electron distribution of LiCoO₂ support without destroying the main structure. Remarkably, the valence state of Ru rapidly increases from +4.02 to +5.57 in RuO₂ switching the voltage from OCV to 1.7 V vs. RHE, which is attributed to the excessive oxidation originating from the damage of the RuO₂ structure (Fig. 5h, Supplementary Figs. 61, 62). Based on the above results, the dynamic reconstruction process of RuO₂/LiCoO₂ can effectively modify the charge environment of the Ru-O bond and stabilize the lattice oxygen, resulting in improved OER performance.

Origin of high catalytic performance

To investigate the transfer phenomena of reaction intermediates on the electrocatalyst surface, a series of electrochemical measurements are performed. The cyclic voltammetry curves of RuO₂/LiCoO₂ and RuO₂ can observe two pairs of redox peaks around 0.5 V and 1.0 V respectively, which correspond to the formation of *OH reaction intermediates with H₂O deprotonation (Fig. 6a). Obviously, the redox peak of RuO₂/LiCoO₂ moves downwards than that of RuO₂, indicating favorable H₂O deprotonation and promoting the reaction process⁴⁶. Subsequently, methanol is used as a probe molecule to detect the coverage of *OH intermediates on the catalyst surface (Supplementary Fig. 63). The methanol oxidation reaction (MOR) is more active on OH*-dominated catalyst surfaces due to the mechanism of nucleophilic attack on electrophilic sites¹². RuO₂/LiCoO₂ exhibits high surface OH* coverage compared to RuO₂ in Fig. 6b, implying that RuO₂/LiCoO₂ contains more effective OER active sites. Furthermore, the change in catalytic activity of the RuO₂/LiCoO₂ catalyst at different pH values is negligible, which is a characteristic of the typical AEM mechanism (Fig. 6c and Supplementary Fig. 64)⁴⁷. In contrast, RuO₂ exhibits a pH-dependent behavior of OER activity, demonstrating the occurrence of an unstable LOM mechanism. The rotating ring-disk electrode test and the adsorption/desorption energy barrier analysis of Li-containing intermediates indicate that it is difficult for the extracted Li ions to participate in the OER process of RuO₂/LiCoO₂ (Supplementary Figs. 65–68). Furthermore, RuO₂/LiCoO₂ maintains the same potential in electrolytes with different Li-ion concentrations after 50 h of stability testing, further confirming that the small amount of extracted Li has a negligible effect on the performance of the catalyst (Supplementary Fig. 69).

Fig. 6: Reaction mechanism on RuO₂/LiCoO₂.

a Cyclic voltammetry curves of RuO₂/LiCoO₂ and RuO₂ catalysts in 0.5 M H₂SO₄. b The function plot between MOR current density and methanol concentration on RuO₂/LiCoO₂ and RuO₂ catalysts. c The pH dependence of the OER potential at various current densities for RuO₂/LiCoO₂ and RuO₂. Operando ATR-SEIRAS spectra of d RuO₂/LiCoO₂ and e RuO₂ at various applied potentials. f Schematic diagrams of the OER mechanism on RuO₂/LiCoO₂ and RuO₂, respectively. g OER Gibbs free energy diagrams of RuO₂/Li_1−xCoO₂, RuO₂/LiCoO₂, and RuO₂, respectively.

The mechanism of the acidic OER reaction of RuO₂/LiCoO₂ and RuO₂ is further monitored by Operando attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) (Supplementary Fig. 70)⁴⁸. A strong absorption band can be observed in RuO₂/LiCoO₂ at about 1033 cm⁻¹, which is identified as O-O stretching in the *OOH intermediate in Fig. 6d⁴⁹. Since *OOH is a typical intermediate of the AEM mechanism, the OER process of RuO₂/LiCoO₂ is mainly dominated by AEM. In contrast, the characteristic peaks at 1033 and 1122 cm⁻¹ in RuO₂ correspond to *OOH and *OO intermediates, respectively, indicating a combined path of LOM and AEM (Fig. 6e)⁵⁰. Thus, the dynamic reconstruction effect realizes the complete transformation of RuO₂/LiCoO₂ electrocatalysis from LOM to AEM mechanism, thereby preventing the dissolution of the catalyst and enhancing OER stability (Fig. 6f). The Gibbs free energy barriers of OER reaction intermediates on the electrocatalyst are calculated to explore the activity differences (Supplementary Figs. 71–73). According to Fig. 6g, the OER rate-determining step (RDS) of RuO₂/Li_1−xCoO₂, RuO₂/LiCoO₂, and RuO₂ is the formation energy barrier of the intermediate *OOH. The RuO₂/Li_1−xCoO₂ electrocatalyst only requires an energy barrier of 1.77 eV to overcome RDS, which is much lower than that of RuO₂ (2.19 eV). The formation barrier of the *OO intermediate of RuO₂/Li_1−xCoO₂ is much higher than that of *OOH, which is more conducive to the AEM mechanism (Supplementary Figs. 74–81). In addition, the overlap between Ru d and O p orbitals is reduced, further confirming the weakened covalency of the Ru-O bond. This hinders lattice oxygen from participating in the OER reaction process, thereby enhancing structural stability (Supplementary Figs. 82, 83). As a result, the modification of Li_1−xCoO₂ support optimizes the binding energy of RuO₂ with key intermediates, resulting in a favorable thermodynamic process and enhanced intrinsic activity.

Discussion

In summary, we have successfully manipulated the bonding environment of RuO₂ during the catalytic reaction by utilizing the dynamic reconstruction of the LiCoO₂ support to achieve a balance between activity and stability in acidic OER. Dynamic electrochemical delithiation regulates the electron distribution and coordination structure at the interface, promoting the self-optimization of the RuO₂/LiCoO₂ catalyst during the OER process. The weakened covalency of the Ru-O bond triggers a complete transition of the RuO₂/LiCoO₂ electrocatalyst from the LOM to the AEM reaction pathway, improving the catalytic stability. As a result, the RuO₂/LiCoO₂ electrocatalyst reaches a current density of 10 mA cm⁻² at a low overpotential of 150 ± 2 mV. In particular, the PEM electrolyzer using RuO₂/LiCoO₂ operates stably for 2000 h at a high current density of 1 A cm⁻². This work provides a method to solve the balance between the activity and stability of ruthenium-based oxide electrocatalysts in acidic OER.

Methods

Experimental section

Materials preparation

Preparation of LiCoO₂ Nanosheets

The stoichiometric amount of Co₃O₄ (solid, ACS reagent, ≥ 98%) and 4% excess Li₂CO₃ (solid, ACS reagent, ≥ 98%) was mixed by ball milling for 6 h at 4.47 × g. Subsequently, the precursors were calcined for 4 h at 950 °C in air with a heating rate of 5 °C min⁻¹ to obtain bulk LiCoO₂ powder. The bulk LiCoO₂ powders were dispersed in distilled water and sonicated for 4 h in an ice-water bath, maintaining the temperature at 15 °C. The supernatant was collected by centrifugation at 698.75 × g. Finally, LiCoO₂ nanosheet powders were obtained by freeze-drying the supernatant at −40 °C.

Preparation of RuO₂/LiCoO₂

The 50 mg of LiCoO₂ and 10 mg of RuCl₃ (99.9%) were dissolved in 20 mL of deionized water and magnetically stirred at room temperature for 12 h. Then, the precursor powder was collected by centrifugation at 4025 × g. Finally, the obtained powder was calcined in air at 300 °C for 3 h with a heating rate of 5 °C min⁻¹ to obtain RuO₂/LiCoO₂. For comparison, the amount of RuCl₃ was changed to 5 mg and 15 mg to obtain a RuO₂/LiCoO₂ electrocatalyst with different RuO₂ loadings.

Characterization

The morphology of the catalysts was characterized by a scanning electron microscope (SEM JEOL JSM-6700F) and transmission electron microscope (TEM FEI Tecnai G2 F20). Aberration-corrected high-angle annular dark-field scanning transmission electron microscope (AC HAADF-STEM) images were taken at JEM-ARM200F equipped with a JED-2300T SDD. X-ray diffraction (XRD) data obtained from Bruker D8 Advance equipment was used to analyze the crystal structure. The elemental compositions were analyzed by ICP (ICP-MS, Inductively coupled plasma-mass spectrometry). X-ray photoelectron spectroscopy (XPS) analysis was performed on an Escalab 250Xi system using Al Kα X-rays. The Co K-edge and Ru K-edge X-ray absorption spectroscopy (XAS) was measured at the beamline of TLS 17C1 and TPS 44A1 at the National Synchrotron Radiation Research Center (NSRRC) in Taiwan.

Electrochemical measurements

The electrochemical performance was tested on an electrochemical workstation using a three-electrode system (Autolab PGSTAT302, Metrohm). Graphite rod and saturated Hg/HgSO₄ electrode were used as counter electrode (CE) and reference electrode (RE), respectively. The catalyst, carbon black, and polyvinylidene fluoride (PVDF) were mixed in a weight ratio of 7:2:1, and N-methyl-2-pyrrolidone (NMP) was used as the solvent. The viscous slurry was evenly coated on carbon paper and dried under a vacuum. The catalyst loading is 1 mg_cat cm⁻² and the loading area is 1 cm². The measured potential was converted into a reversible hydrogen electrode according to the equation E_RHE = E_Hg/Hg2SO4 + 0.656 V + 0.0591 pH. Subsequently, CV was measured at a scan rate of 1 mV s⁻¹. The linear sweep voltammetry (LSV) was performed in N₂-saturated 0.5 M H₂SO₄ solution (pH is 0.3 ± 0.1) at a scan rate of 5 mV s⁻¹. The electrolyte was prepared and used immediately and stored in a glass bottle at room temperature. The potential was corrected according to the E=E_applied - iR formula. The electrochemical impedance spectroscopy (EIS) was performed in the frequency range of 0.01-100 kHz with an amplitude of 5 mV. The double-layer capacitance (C_dl) was obtained by collecting CV curves with scan rates of 10 to 60 mV s⁻¹. OCP tests were performed with H₂SO₄ saturated with high-purity hydrogen to confirm the reference electrode potential. Calibration experiments were performed at room temperature (25 °C) to reduce temperature effects.

Rotating ring disk electrode (RRDE) test of RuO₂/LiCoO₂ catalyst. The 5 mg RuO₂/LiCoO₂ catalyst was dispersed in a mixed solution of 980 μL isopropyl alcohol and 20 μL Nafion (5 wt%), followed by ultrasonication for 1 h to obtain a uniformly dispersed catalyst ink. 20 μL of catalyst ink was dropped onto a disk electrode (area 0.2475 cm²) and dried under vacuum at room temperature to obtain a working electrode. A carbon rod and Hg/HgSO₄ were used as the counter electrode and reference electrode, respectively. The linear sweep voltammetry was carried out in 0.5 M H₂SO₄ solution with different Li-ion concentrations at a scan rate of 5 mV s⁻¹ and a rotation speed of 1600 rpm.

The electrochemical surface area (ECSA) of an electrocatalyst can be evaluated by electrochemical double capacitance (C_dl) according to the following equation:

$${\rm{ECSA}}=\frac{{{\rm{C}}}_{{\rm{dl}}}}{{{\rm{C}}}_{{\rm{s}}}}$$

(1)

where C_dl was determined by taking half the slope of the current differences (Δj = j_anodic - j_cathodic) that were plotted as a function of the scan rate in a CV experiment. C_s is the general surface specific capacitance (0.04 mF cm⁻²).

The turnover frequency (TOF) of the electrocatalyst was calculated by the following equation:

$${\rm{TOF}}=\left(\frac{{\rm{Formate}}\; {\rm{turnovers}}\; {\rm{per}}{{\rm{A}}}_{{\rm{geo}}}}{{\rm{Active}}\; {\rm{sites}}\; {\rm{per}}{{\rm{A}}}_{{\rm{geo}}}}\right)$$

(2)

The formate turnover per geometric area was obtained from the geometric current density for the LSV polarization curves according to the equation:

$${{\rm{A}}}_{{\rm{geo}}}{={\rm{j}}}_{{\rm{geo}}}\times \frac{1\,{\rm{C}}\,{{\rm{s}}}^{-1}}{1000\,{\rm{mA}}}\times \frac{1\,{\rm{mol}}}{96485.3\,{\rm{C}}}\times \frac{1}{4}\times \frac{6{.023\times 10}^{23}}{1\,{\rm{mol}}\,{{\rm{O}}}_{2}}$$

(3)

Electrochemical in situ ATR-SEIRAS experiments

ATR-SEIRAS measurements were conducted using a Nicolet iS50 FT-IR spectrometer, with each spectrum obtained by accumulating 32 interferograms, achieving a spectral resolution of 8 cm⁻¹. The working electrode preparation involved two key steps. First, an ultra-thin Au film was chemically deposited onto a silicon crystal to enhance infrared signal sensitivity and electron conductivity. Then, a catalyst slurry with a loading of 0.1 mg cm⁻² was applied onto the Au surface. This slurry was prepared by dispersing 7 mg of catalyst and 3 mg of carbon black in 1 mL ethanol, followed by the addition of 50 μL of Nafion after 30 min of sonication. The assembled working electrodes were placed in a three-electrode electrochemical cell, where Hg/HgSO₄ served as the reference electrode, a graphite rod as the counter electrode, and Ar-saturated 0.5 M H₂SO₄ as the electrolyte for the OER reaction⁵¹. All measurements were performed using linear sweep voltammetry (LSV) to investigate OER reaction intermediates at various applied potentials.

In situ XAFS measurements

In-situ X-ray absorption spectroscopy (XAS), including both XANES and EXAFS at the Ru and Co K-edges, was conducted in total fluorescence yield mode under ambient conditions at BL-12B2 of SPring-8, NSRRC. The measurements were carried out using a custom-designed Teflon container equipped with a Kepton tape-sealed window, allowing X-rays to pass through both the tape and electrolyte. This setup ensured that XAS signals were effectively captured in total fluorescence yield mode at the National Synchrotron Radiation Research Center (NSRRC), SPring-8. The experiments were performed under a three-electrode configuration, consistent with the electrochemical characterization conditions⁵². For data processing, spectral normalization was achieved by removing the pre-edge baseline and adjusting the post-edge region. The k²-weighted EXAFS oscillations underwent Fourier transformation to facilitate EXAFS analysis, with all EXAFS spectra presented without phase correction. The Fourier-transformed (FT) data fitting was conducted using Artemis (version 0.9.25), employing a k³ weighting factor with a k-range of 3–12 Å⁻¹ and an R-range of 1.0–4.0 Å. The coordination number, bond length, Debye-Waller factor, and energy shift (CN, R, σ², and ΔE₀) were determined through fitting without any fixed parameters, while the amplitude reduction factor (S₀²) was set to 0.85.

PEMWE measurements

RuO₂/LiCoO₂ was used as anode catalysts in PEM electrolyzers. Commercial Pt/C was used as a cathode catalyst. The membrane electrode assembly was fabricated via the catalyst-coated membrane technique, covering a geometric area of 2 cm × 2 cm (4 cm²). The catalyst powder was dispersed in isopropanol, deionized water, and Nafion solution to prepare the ink. The uniformly dispersed ink was obtained by slice emulsification and ultrasonic cell disruption. The well-dispersed RuO₂/LiCoO₂ and Pt/C catalyst inks were sprayed on both sides of the PEM, respectively. The loadings of RuO₂/LiCoO₂ anode and Pt/C cathode are 4 mg_cat cm⁻² and 1 mg_cat cm⁻², respectively. Nafion 115 was used as a proton exchange membrane (PEM) and was treated with H₂O₂ and 0.5 M H₂SO₄ at 80 °C for 1 h in sequence. The size of the proton exchange membrane was 2.6 cm² and the membrane thickness was 127 μm. The sprayed membrane, anode gas diffusion layer (Ti felt), and cathode gas diffusion layer (carbon paper) were hot-pressed at 130 °C and 10 MPa pressure to obtain a membrane electrode assembly. Subsequently, a PEM water electrolyzer was assembled and tested at 80 °C using pure water as the electrolyte. The polarization curve of PEMWE was obtained at a scan rate of 5 mV s⁻¹, and a chronopotentiometric test was performed at 1 A cm⁻² to evaluate the stability.

DFT calculations

All density functional theory (DFT) calculations were conducted using the Vienna Ab initio Simulation Package (VASP)⁵³. The projector augmented wave (PAW) pseudopotential was applied in combination with the PBE generalized gradient approximation (GGA) exchange-correlation functional⁵⁴. To appropriately describe the localized d-electrons of Co, the DFT + U method was employed, incorporating a Hubbard-U correction of Ueff(Co) = 3.32 eV, determined via linear response theory. The calculations were performed with a plane wave basis set energy cutoff of 500 eV, and a Monkhorst-Pack k-point grid of 3 × 3 × 1 was used for Brillouin zone sampling. Spin polarization was considered, and full structural relaxation was conducted until the energy convergence criterion reached 10⁻⁵eV per atom, with the final force acting on each atom kept below 0.05 eV Å⁻¹. Additionally, Pourbaix diagrams were generated using the Atomic Simulation Environment (ASE), where input formation energies were derived from DFT calculations of bulk and surface models⁵⁵.

The adsorption energy of reaction intermediates can be computed using the following Equation:

$${\Delta \text{G}}_{\text{ads}}={\text{E}}_{\text{ads}}-{\text{E}}_{*}+\Delta {\text{E}}_{\text{ZPE}}-\text{T}\Delta \text{S}$$

(4)

Where ads = OH*, O*, OOH*, an ${\text{E}}_{\text{ads}}-{\text{E}}_{*}$ is the binding energy, $\Delta {\text{E}}_{\text{ZPE}}$ is the zero-point energy change, $\Delta \text{S}$ is the entropy change. In this work, the values of $\Delta {\text{E}}_{\text{ZPE}}$ and $\Delta \text{S}$ were obtained by vibration frequency calculation.

The Gibbs free energy of the reaction steps can be calculated by the following four Equations:

$${\text{H}}_{2}\text{O}{+}^{*}\leftrightarrow \text{H}{\text{O}}^{*}+{\text{H}}^{+}+{\text{e}}^{-}$$

(5)

$$\Delta {\text{G}}_{1}=\Delta {\text{G}}_{{\text{HO}}^{*}}+{\text{G}}_{\text{H}}-\Delta {\text{G}}_{*}-{\text{G}}_{{\text{H}}_{2}\text{O}}-\text{eU}$$

(6)

$${\text{HO}}^{*}\leftrightarrow {\text{O}}^{*}+{\text{H}}^{+}+{\text{e}}^{-}$$

(7)

$$\Delta {{\rm{G}}}_{2}=\Delta {{\rm{G}}}_{{{\rm{O}}}^{\ast }}-\Delta {{\rm{G}}}_{{{\rm{HO}}}^{\ast }}{+{\rm{G}}}_{{\rm{H}}}-{\rm{eU}}$$

(8)

$${\text{O}}^{*}+{\text{H}}_{2}\text{O}\leftrightarrow {\text{OOH}}^{*}+{\text{H}}^{+}+{\text{e}}^{-}$$

(9)

$$\Delta {{\rm{G}}}_{3}=\Delta {{\rm{G}}}_{{{\rm{OOH}}}^{\ast }}{+{\rm{G}}}_{{\rm{H}}}-\Delta {{\rm{G}}}_{{{\rm{O}}}^{\ast }}{-{\rm{G}}}_{{{\rm{H}}}_{2}{\rm{O}}}-{\rm{eU}}$$

(10)

$${{\rm{OOH}}}^{\ast }{\leftrightarrow }^{\ast }{+{\rm{O}}}_{2}{+{\rm{H}}}^{+}{+{\rm{e}}}^{-}$$

(11)

$$\Delta {{\rm{G}}}_{4}=\Delta {{\rm{G}}}_{\ast }-\Delta {{\rm{G}}}_{{{\rm{OOH}}}^{\ast }}{+{\rm{G}}}_{{\rm{H}}}{+{\rm{G}}}_{{{\rm{O}}}_{2}}-{\rm{eU}}$$

(12)

In this work, $\Delta {\text{G}}_{1-4}$ were calculated at U = 0.