4f-modified Ru-O polarity as a descriptor for efficient electrocatalytic acidic oxygen evolution

Abstract

The development of non-iridium-based oxygen evolution reaction (OER) catalysts is crucial for proton exchange membrane water electrolysis (PEMWE), but hydrogen production remains a great challenge because of sluggish OER kinetics and severe catalyst dissolution. Here, we present a 4f-induced covalent polarity modulation strategy for the construction of 4f-orbital-modified RuO₂ (4f-RuO₂) nanocatalysts with tunable Ru–O polarity. We find that the OER activity of 4f-RuO₂ shows a volcano shape as a function of the polarity of Ru–O bond. Consequently, the best 4f-Nd-RuO₂ catalyst possesses an ultra-low overpotential of 214 mV at 10 mA cm⁻² and robust electrochemical stability in 0.1 M HClO₄. Theoretical calculations coupled with in situ synchrotron infrared and X-ray absorption spectroscopy analyses reveal that the modulation of Ru–O polarity in RuO₂ by the valence f−p−d gradient orbital coupling can modify the adsorption energy of the reaction intermediates and suppress the participation of lattice oxygen to avoid over-oxidation of Ru, which can thus serve as an effective descriptor for fine tuning the activity and durability of acidic OER nanocatalysts.

The sluggish oxygen evolution reaction kinetics and intense catalyst degradation under acidic conditions limit the practical application of electrolyzers. Here, the authors construct 4f-orbital-modified RuO₂ catalysts with tunable Ru–O polarity that enhance catalytic activity and durability.

Introduction

Electrocatalytic water splitting is considered a promising and sustainable technology for electrochemical energy storage and conversion^1–4. Compared with alkaline water electrolysis (AWE), acidic proton exchange membrane water electrolysis (PEMWE) has gained greater attention because of its advantages, including higher voltage efficiency, higher current density, more compact configuration, and lower ohmic loss^5–7. However, the harsh acidic media and the strong oxidative environment of the sluggish oxygen evolution reaction (OER) lead to severe degradation of anode catalysts^8–10. Currently, precious iridium (Ir)-based catalysts (such as IrO₂) are recognized as the only practical anode catalysts for PEMWE with the compromise between stability and activity in acidic electrolytes^11–14. Nevertheless, the high cost of these catalysts impedes their widespread use in practical electrolyzers^13,15. Although ruthenium (Ru)-based oxides have been acknowledged to be a potential alternative owing to their relatively high intrinsic activity and abundant reserves, the unsatisfactory stability and dramatically decreased activity at a high potential or current density (≥100 mA cm⁻²) remain serious problems as a result of the participation of lattice oxygen (oxygen vacancy) and over-oxidation of Ru (soluble RuO₄)^16–18. Thus, developing Ru-based OER electrocatalysts with enhanced stability and activity to improve the efficiency of acidic electrochemical water splitting is still highly challenging.

To date, numerous strategies have been adopted to realize acid-viable OER properties of Ru-based electrocatalysts, especially their stability, with a particular focus on element doping, strain effects, structure tuning, and surface engineering (Mn-doped RuO₂, Ru₁-Pt₃Cu, CaCu₃Ru₄O₁₂ perovskite, Ru/TiO_x, and Ru@IrO_x)^19–23. Unfortunately, most reported Ru-based catalysts operate only at a low current density of ~10 mA cm⁻² for tens of hours with the standard three-electrode setup, which still leaves much room for enhancing the stability and activity under high current density conditions for industrial applications. Recently, it has been found that tuning the active metal–oxygen (M–O) covalent polarity can strongly affect the OER performance via an electronic modulation strategy^24–28. Interestingly, the modification of Ru–O polarity by using 3d transition metals can suppress the lattice oxygen mechanism (LOM) and the dissolution of Ru sites^16,29. Fu et al. reported that the introduction of Co–N₄ sites into Ru/Co–N–C can optimize the interaction between Ru 4d and O 2p orbitals and the bonding strength of the intermediates with Ru sites through the adsorbate evolution mechanism (AEM), which results in an increased resistance of Ru sites to over-oxidation and corrosion¹⁸. However, it is difficult to manipulate and then maximize the modification of Ru–O polarity by 3d transition metals, because 3d orbitals are susceptible to the crystal field and coordination environment³⁰. Unlike 3d orbitals, the 4f orbitals of rare earth elements (REs) in the sixth period with unique 4f ^x5d ^y6s² valence-electron configurations can effectively participate in the optimization of electrocatalytic performance owing to the valence f–p–d gradient orbital coupling. It is anticipated that the 4f orbitals can effectively adjust the local electronic structure of the surrounding atoms and maintain a stable chemical state because of the distinctive loose 4f orbitals^31–34. This provides a high possibility of engineering Ru–O polarity via 4f-orbital-modification to stabilize lattice oxygen and improve the intrinsic activity of Ru sites for Ru-based oxides in electrocatalytic reactions. Moreover, to effectively direct catalyst design, the catalytic mechanism based on the RE 4f–O 2p–Ru 4d hybridization needs to be explored but has rarely been systematically investigated in the field of acidic OER. Therefore, it is quite important to develop 4f-orbital-modified Ru-based catalysts and identify the relationship between Ru–O polarity and catalytic performance.

In this work, theoretically and experimentally, we successfully constructed 4f-orbital-modified RuO₂ (4f-RuO₂) nanocatalysts via a 4f-induced covalent polarity modulation strategy as a model catalyst to investigate the effect of Ru–O polarity on the acidic OER. Bader charges, soft X-ray absorption spectroscopy (sXAS), and X-ray photoelectron spectroscopy (XPS) revealed that the stable RE–O–Ru network promoted electron donation from the RE to Ru, thus weakening Ru−O 4d–2p hybridization and decreasing Ru–O polarity. Interestingly, we found that the OER activity exhibits a classic volcano relationship as a function of the Ru–O polarity, which can be a powerful OER activity descriptor. The best catalyst, 4f-Nd-RuO₂, delivers an overpotential of 214 mV to achieve a current density of 10 mA cm⁻² in 0.1 M HClO₄, with a lifetime ten times longer than that of commercial RuO₂. More impressively, the 4f-Nd-RuO₂ catalyst showed stable water electrolysis for 200 h at a high current density of 0.1 A cm⁻² in a PEM electrolyzer, suggesting the potential for practical water-splitting applications. Moreover, by combining in situ synchrotron radiation Fourier-transform infrared (SR-FTIR) and in situ X-ray absorption fine structure (XAFS) spectroscopies, we confirmed that the optimum Ru−O polarity of 4f-Nd-RuO₂ can modify the adsorption energy of the intermediates and suppress the participation of lattice oxygen to avoid over-oxidation of Ru, which is responsible for the improved acidic OER performance.

Results and discussion

Synthesis and characterization of catalysts

In light of the Ru−O polarity, we synthesized a series of 4f-RuO₂ catalysts with different REs (see “Methods” for details). A three-step method involving H₂ reduction, nanoparticle oxidation, and acid-leaching processes was used to prepare these electrocatalysts (Fig. 1a). As shown in Fig. 1b, after annealing reduction under a H₂ atmosphere, the NdRu nanoparticles were uniformly distributed on the carbon substrate. Subsequently, the obtained NdRu nanoparticles were annealed in an air atmosphere and converted to 4f-Nd-RuO₂ nanoparticles, which have an average particle size of only ~3.9 nm (Fig. 1c, d). In addition, commercial RuO₂ (c-RuO₂) and unmodified RuO₂ nanoparticles (n-RuO₂) were used as control samples (Supplementary Figs. 1 and 2). The X-ray diffraction (XRD) pattern of the resulting 4f-Nd-RuO₂ is essentially identical to that of n-RuO₂ (Fig. 1e), revealing that the incorporation of Nd cannot significantly affect the RuO₂ structure. As determined from the spherical aberration corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image (Fig. 1f and Supplementary Fig. 3a), the 4f-Nd-RuO₂ nanoparticles exhibit high crystallinity. The well-defined lattice fringes are assigned to the (101) and (110) crystal planes of RuO_2, with lattice spacings of 0.257 and 0.322 nm, respectively^35,36. The energy-dispersive spectroscopy (EDS) mapping images show that Nd, Ru, and O are uniformly distributed throughout the entire nanocrystals (Fig. 1g and Supplementary Fig. 3b), confirming the successful modification of RuO₂ (Supplementary Figs. 4 and 5). Furthermore, the results from inductively coupled plasma‒optical emission spectrometry (ICP‒OES) characterization revealed that the measured content of Nd in 4f-Nd-RuO₂ was 0.8 wt%. Accordingly, by using additional metal salts as precursors, 4f-RuO₂ with different REs were successfully synthesized and denoted as 4f-La-RuO₂, 4f-Ce-RuO₂, 4f-Pr-RuO₂, 4f-Sm-RuO₂, 4f-Gd-RuO₂, 4f-Dy-RuO₂, 4f-Er-RuO₂, 4f-Yb-RuO₂ and 4f-Lu-RuO₂ (4f-La-RuO₂ without 4f electrons as a non-4f control, Supplementary Figs. 6, 7 and Table 1).

Fig. 1. Synthesis and characterization of the catalysts.

a Schematic diagram of the synthetic process. TEM images of b NdRu/C, c, d 4f-Nd-RuO₂. e XRD patterns of 4f-Nd-RuO₂ and n-RuO₂. f HAADF-STEM image and g EDS mapping images of 4f-Nd-RuO₂. The average particle size of 4f-Nd-RuO₂ is ~3.9 nm, as shown in the inset of (c). Source data are provided as a Source Data file.

To determine the surface electronic structure and chemical state of 4f-Nd-RuO₂, X-ray photoelectron spectroscopy (XPS) was performed. Compared with the Ru 3d signals of n-RuO₂, those of 4f-Nd-RuO₂ shift to lower binding energies with two sets of doublets, which could be assigned to the doublet peaks of Ru 3d_5/2, 3d_3/2 and their satellite peaks (Fig. 2a)^35,37,38. This result can be attributed to the strong electronic interaction between Nd and Ru, resulting from the formation of the Nd–O–Ru local structure. For the Nd 3d_5/2 peak of 4f-Nd-RuO₂ (Supplementary Fig. 8b), a shift to higher binding energy was observed in comparison with that of commercial Nd₂O₃ (c-Nd₂O₃). Notably, the peaks corresponding to Nd 3d_5/2 and O KLL overlapped between 970 and 986 eV^39–41. In addition, the electron paramagnetic resonance (EPR) spectra reveal the non-existence of oxygen vacancies in 4f-Nd-RuO₂ (Supplementary Fig. 9)⁴². We further employed X-ray absorption fine structure (XAFS) spectroscopy characterization to investigate the influence of Nd. As seen from X-ray absorption near-edge structure (XANES) spectroscopy of the Ru K-edge (Fig. 2b), the absorption edge of 4f-Nd-RuO₂ shifts to a lower energy compared with that of n-RuO₂, indicating a decrease in the oxidation state, which agrees well with the XPS analysis. Moreover, the corresponding Fourier transform extended XAFS (FT-EXAFS) spectrum of 4f-Nd-RuO₂ is almost identical to that of n-RuO₂ (Fig. 2c). The dominant peak at ~1.5 Å can be assigned to shell coordination of the nearest Ru–O bond (1.98 Å), which is slightly longer than that in n-RuO₂ (1.97 Å) according to quantitative fitting, suggesting that Nd can stretch the Ru–O bond (Supplementary Figs. 10, 11 and Table 2)^18,19,43. Furthermore, the wavelet transform (WT) EXAFS analysis provided additional insights into the local bonding environment. The Ru K-edge WT plots of 4f-Nd-RuO₂ and n-RuO₂ reveal a maximum WT value of ~4.7 Å⁻¹, which is attributed to the bonding between Ru and O atoms in the first shell. The maximum WT value of ~9.5 Å⁻¹ is associated with the scattering between Ru and Ru atoms in the first shell of the Ru foil (Fig. 2d). Accordingly, compared with that of c-Nd₂O₃, the intensity of the Nd L₃-edge white lines of 4f-Nd-RuO₂ increases (Fig. 2e), which clearly reveals that the Nd in 4f-Nd-RuO₂ is in a relatively lower electron density state, resulting in strengthened Nd–O bonding. Apparently, the incorporation of Nd atoms tunes the local electronic and atomic structure of the Ru–O bond; that is, the strengthened Nd–O and stretched Ru–O bonds of the Nd–O–Ru network could guarantee high electron densities around Ru sites, which may directly influence the OER performance^44–46.

Fig. 2. Electronic structure of the catalysts.

a Ru 3d XPS spectra of 4f-Nd-RuO₂ and n-RuO₂. b Ru K-edge XANES spectra, c Fourier transforms, and d wavelet transforms of the Ru K-edge EXAFS oscillations of 4f-Nd-RuO₂, n-RuO₂, and Ru foil. e Nd L₃-edge XANES spectra of 4f-Nd-RuO₂ and c-Nd₂O₃. Source data are provided as a Source Data file.

Electrocatalytic OER in a three-electrode system

The OER activity of 4f-RuO₂ was studied in strongly acidic media (0.1 M HClO₄) with a three-electrode system. The working electrodes were prepared by drop-casting water/ethanol and Nafion-based ink of 4f-RuO₂ powder on a glassy carbon disk. Figure 3a and Supplementary Figs. 12 and 13 show the linear sweep voltammetry (LSV) curves for the 4f-RuO₂ electrocatalysts and reference samples. Obviously, the overpotential for 4f-Nd-RuO₂ to deliver 10 mA cm⁻² is 214 mV, which is superior to those of n-RuO₂ (276 mV), c-RuO₂ (338 mV) and other 4f-RuO₂ (218–255 mV) and lower than those of the most recently reported electrocatalysts in acidic media (Supplementary Table 5 and Fig. 14). Importantly, Fig. 3b indicates that the Tafel slope of 4f-Nd-RuO₂ (65 mV dec⁻¹) is significantly lower than that of n-RuO₂ (129 mV dec⁻¹) and c-RuO₂ (130 mV dec⁻¹). As a result, the OER process of 4f-Nd-RuO₂ accelerates rapidly with increasing overpotential, which may lead to great potential in industry. The fast kinetics of 4f-Nd-RuO₂ was further confirmed by the electrochemical impedance spectroscopy (EIS) measurements (Fig. 3c and Supplementary Fig. 15). The high-frequency semicircle of the Nyquist plot represents the charge transfer process during the OER⁴⁷. Compared with n-RuO₂ and c-RuO₂, 4f-Nd-RuO₂ presented the lowest charge transfer resistance (R_ct) at 1.45 V, suggesting much faster OER kinetics.

Fig. 3. Electrocatalytic OER in a three-electrode system.

a LSV curves with 95% iR correction (scan rate = 10 mV s⁻¹, solution resistance = 40.5 ± 2.5 Ω), b Tafel slopes, c Nyquist plots at 1.45 V, d capacitive current at 1.006 V against the scan rates and the corresponding C_dl value, e normalized LSV curves to the electrochemically active surface area, f MA and TOF at 1.53 V, g chronopotentiometry curves, and h comprehensive comparison of the OER performance of 4f-Nd-RuO₂, n-RuO₂, and c-RuO₂. An FTO substrate was used as the catalyst support for the stability tests. All the measurements were performed in 0.1 M HClO₄ solution (pH = 1.0 ± 0.1). The error bars are the standard deviations of three individual calculations. Source data are provided as a Source Data file.

To better understand the intrinsic activity of 4f-Nd-RuO₂, the electrochemically active surface area (ECSA) was evaluated by the electrochemical double-layer capacitance (C_dl) for activity normalization (Fig. 3d and Supplementary Figs. 16, 17a). 4f-Nd-RuO₂ has the highest C_dl, which is almost 2.3 times greater than that of n-RuO₂, indicating that 4f-Nd-RuO₂ possesses much more available catalytic active sites. Considering the ECSA-normalized OER activity, the specific activity of 4f-Nd-RuO₂ still outperforms those of n-RuO₂ and c-RuO₂, which reveals that the enhanced activity of the OER performance of 4f-Nd-RuO₂ is not just enhanced by the ECSA and that the intrinsic activity arising from the Nd ions plays a more important role (Fig. 3e). In particular, 4f-Nd-RuO₂ shows the highest specific activity among all the 4f-RuO₂ catalysts (Supplementary Fig. 17b), which is consistent with our computational results. Moreover, the mass activity (MA) and turnover frequency (TOF) were also calculated on the basis of the Ru sites. As shown in Fig. 3f, the MA and TOF of 4f-Nd-RuO₂ are ~580 A g_Ru⁻¹ and 546 h⁻¹ at a potential of 1.53 V, which are ~7.8 and 29 times greater than those of n-RuO₂ (74 A g_Ru⁻¹, 70 h⁻¹) and c-RuO₂ (20 A g_Ru⁻¹, 19 h⁻¹), respectively.

On the other hand, operation stability is also an indispensable index for evaluating catalysts for practical water-splitting applications. As illustrated in Fig. 3g, we performed chronopotentiometry measurements at a current density of 10 mA cm⁻², a benchmark criterion suggested in previous reports. To avoid carbon corrosion under strongly acidic and anodic conditions, an FTO substrate was used as the catalyst support (Supplementary Figs. 18 and 19). Obviously, the potential of c-RuO₂ sharply jumps from 1.6 to 2.0 V within 12 h, essentially losing all of its activity. In contrast, during the 40 h long-term test for 4f-Nd-RuO₂, the potential required to reach 10 mA cm⁻² was almost constant (increasing by only ~68 mV), well surpassing those of n-RuO₂ (~190 mV) and c-RuO₂. The notable stability of 4f-Nd-RuO₂ outperformed most of the reported Ru-based OER electrocatalysts, suggesting that the modification of Nd can markedly mitigate the performance degradation. We also used inductively coupled plasma-mass spectrometry (ICP-MS) to quantify the mass loss of Nd and Ru during chronopotentiometry measurements (Supplementary Table 3). After stability testing, there was a slight loss of 0.35 wt% Ru and 2.17 wt% Nd in the electrolyte for 4f-Nd-RuO₂. Note that the amount of Ru leached from 4f-Nd-RuO₂ is smaller than that from n-RuO₂ and c-RuO₂, demonstrating that the modification of Nd inhibits the oxidative release of lattice oxygen and the dissolution of Ru, which may be the origins of the stability. The stability number (S-number) is defined as the ratio between the amount of evolved oxygen and the amount of dissolved active metal, which also serves as a metric for benchmarking the stability of electrocatalysts⁴⁸. The S-number was 3.3 × 10⁵ at 40 h for 4f-Nd-RuO₂, which is comparable to that of rutile IrO₂ (~10⁵; Sigma-Aldrich)⁴⁹, further demonstrating the excellent stability of 4f-Nd-RuO₂. Furthermore, the morphology, crystalline phase, and electronic structures of post-catalysis 4f-Nd-RuO₂ were well retained, as demonstrated by the HRTEM, XRD, and XPS results (Supplementary Figs. 20 and 21). Therefore, a comprehensive analysis clearly suggests that 4f-Nd-RuO₂ has excellent OER activity and stability (Fig. 3h), which may be attributed to the unique Nd incorporation for optimizing the reaction mechanism.

Ru–O polarity analyses

To determine the polarity of Ru−O, the Bader charges of Ru were calculated. Starting from the pure rutile RuO₂ structure, the Ru atoms are replaced with REs (Supplementary data 1 and Fig. 22). Figure 4a shows the charge density difference of 4f-Nd-RuO₂. Nd shares more electron density depletion, and O serves as the electron acceptor because more electron density accumulates at the O sites of the Nd–O bond. The Bader charges of Ru (Fig. 4b) indicate that the number of electrons donated by RE decreases slightly with increasing atomic number, whereas the number of electrons donated by Ru increases gradually, suggesting an increased valence state of Ru⁴⁶. Notably, the introduction of RE decreases the Bader charge of Ru, leading to a reduction in the Ru–O polarity and the formation of the RE–O–Ru local structure, which may suppress the involvement of lattice oxygen during the OER process⁵⁰. Furthermore, the near-surface electronic structure was assessed via soft X-ray absorption spectroscopy (sXAS) with a total-electron-yield (TEY) signal. The O K-edge XAS spectra shown in Fig. 4c and Supplementary Fig. 23 are powerful experimental evidence to reveal the Ru−O polarity, by showing electronic transitions originating from the O 1s orbitals to unoccupied energy levels⁵¹. The two peaks A₁ and A₂ represent the excitations of the O 1s core electrons into hybridized states of O 2p–Ru 4d t_2g (<530 eV) and O 2p–Ru 4d e_g (>530 eV) orbitals, respectively, owing to splitting by the crystal field^52,53. The shapes of the O K-edge XAS spectra were similar to those for n-RuO₂ and 4f-Nd-RuO₂, and there was no obvious formation of an additional peak after the introduction of Nd. However, the peaks A₁ and A₂ of 4f-Nd-RuO₂ are observed to slightly move toward higher energy regions relative to those of n-RuO₂, indicating weakened Ru–O 4d–2p hybridization and reduced polarity of the Ru–O bond^50,54. Likewise, 4f-orbital-modification with Ce, Pr, Sm, Gd, Dy Er, Yb or Lu also leads to weaker Ru–O polarity, which dictates that the polarity of the Ru 4d and O 2p states can be modulated by the involvement of unique 4f electrons. The decreased polarity of the Ru–O bond can suppress the participation of lattice oxygen and thus restrict the over-oxidation of Ru during the OER, which directly improves the catalyst stability^33,55. More importantly, the metal−oxygen polarity reflects the adsorption strength of oxygen-related intermediates, which was verified to be an important descriptor in alkaline and neutral environments^24,56,57. Namely, the optimal Ru–O polarity, that is neither too weak nor too strong, favors OER. The 4f-RuO₂, with altered electronic structures, provides an opportunity for establishing an OER activity descriptor of metal−oxygen polarity in an acidic environment. We plotted the Ru–O polarity of 4f-RuO₂ (the position of peak A₁) as a function of its specific OER activity (the current density at 1.45 V), yielding a volcano-type trend (Fig. 4d). The 4f-Nd-RuO₂ located at the top of the volcano has optimum Ru–O polarity and the most appropriate adsorption of intermediates, which is responsible for the highest OER activity. Correspondingly, this trend is also in good agreement with the Ru 3d XPS, 3p XPS, and XANES results (Supplementary Figs. 24 and 25) and Bader charge analysis of 4f-RuO₂. The number of electrons donated from RE atoms to O decreases slightly with increasing electronegativity (from La to Lu), whereas the number of electrons donated from Ru to O increases, resulting in a gradual increase in the valence state of Ru.

Fig. 4. Ru−O polarity analyses.

a Charge density difference of 4f-Nd-RuO₂. The yellow area shows electron accumulation, and the blue area represents electron depletion. b Bader charges of Ru in 4f-RuO₂ and RuO₂. c O K-edge XAS spectra of 4f-Nd-RuO₂ and n-RuO₂. d Correlation between the OER activity and the Ru−O polarity of 4f-RuO₂. e In situ SR-FTIR spectra of 4f-Nd-RuO₂ in the range of 900–1300 cm⁻¹. f OER current density at 1.45 V plotted on a log scale as a function of the pH of 4f-Nd-RuO₂ and n-RuO₂. g In situ Ru K-edge XANES spectra and h corresponding FT-EXAFS spectra of 4f-Nd-RuO₂. Source data are provided as a Source Data file.

To experimentally clarify the effect of the modulation of Ru–O polarity, we performed in situ synchrotron radiation Fourier transform infrared (SR-FTIR) spectroscopy measurements under different applied potentials using a home-made electrochemical cell^58,59. New absorption bands at a vibration frequency of ~1076 cm⁻¹ are observed for 4f-Nd-RuO₂ when the potential increases above 1.25 V, which could be attributed to the characteristic vibration of the surface-adsorbed oxygen-related species (*OOH) during the OER process (Fig. 4e), a typical intermediate for AEM pathway^9,60–62. In contrast, a weak peak appeared for n-RuO₂ at a high applied potential of 1.60 V and blue shifted to ~1110 cm⁻¹ (Supplementary Fig. 26), indicating that the LOM pathway may dominate the OER for n-RuO₂ and that the 4f orbitals optimize the adsorption energy of the oxygenated intermediates (*OOH)⁶³. Furthermore, according to the calculated proton reaction orders (ρ = ∂log(i)/∂pH), n-RuO₂ shows significant pH-dependent OER activity, whereas 4f-Nd-RuO₂ reflects a pH-independent characteristic in which the current density nearly remains constant as the pH value changes (Fig. 4f and Supplementary Fig. 27)^64,65. These results reveal that the decreased polarity of the Ru–O bond can inhibit the LOM and follow the AEM process and that optimizing the adsorption energy of key intermediates over that of 4f-Nd-RuO₂ can reduce the reaction energy barrier, resulting in faster OER kinetics. To further investigate the potential-dependent structural evolution of Ru during the OER, in situ XAFS measurements were carried out⁶⁶. The XAFS results are shown in Fig. 4g, h and Supplementary Fig. 28. As the potential increased to 1.50 V, no obvious changes were observed in the Ru K-edge XANES and EXAFS spectra of 4f-Nd-RuO₂, which implies that the oxidation state and coordination structure of Ru remained stable throughout the reaction (Supplementary Figs. 29–31 and Table 4). In addition, the data in Supplementary Fig. 30b indicate that the valence state of Ru in n-RuO₂ increased significantly after stability testing compared with that in 4f-Nd-RuO₂. Above all, the 4f-orbital modification results in the formation of an RE–O–Ru network. The 4f orbitals of RE and 4d orbitals of Ru can form π bonds with the O 2p orbitals, facilitating valence electron transfer from RE to O 2p π orbitals, creating the RE 4f−O 2p−Ru 4d gradient orbital coupling and decreasing the polarity of the Ru–O bond (Supplementary Figs. 32 and 33). The decreased Ru–O polarity suppresses the loss of lattice oxygen and drives the AEM (Supplementary Fig. 34). During the reaction process, the RE could act as an electron reservoir to donate electrons toward surrounding Ru atoms through bridging oxygen and help prevent the over-oxidation of Ru, enhancing the stability of the catalysts. On the other hand, in the gradient-regulated 4f-RuO₂, the optimum Ru–O polarity modulated by Nd can balance the adsorption energy of critical reaction intermediates, supporting the improved activity of the 4f-Nd-RuO₂ catalyst.

Electrocatalytic OER in a PEM electrolyzer

Finally, to further evaluate the practical application potential of the catalysts, we examined the catalytic performances of 4f-Nd-RuO₂, n-RuO₂, and c-RuO₂ in a PEM electrolyzer at 60 °C using commercial Pt/C as the cathode catalyst and a Nafion 115 membrane (Fig. 5a and Supplementary Fig. 35). The polarization curves show that the performance of the 4f-Nd-RuO₂/PEM/Pt/C electrolyzer is superior to that of the n-RuO₂/PEM/Pt/C and c-RuO₂/PEM/Pt/C (Fig. 5b). The best performing 4f-Nd-RuO₂ can reach 0.1, 0.5, and 1 A cm⁻² at cell voltages of 1.829 V, 2.138 V, and 2.385 V, respectively. Moreover, upon applying a constant current of 0.1 A cm⁻² as shown in Fig. 5c, a cell voltage increase of only ~70 mV is observed during 200 h of electrolysis, indicating considerably superior stability under realistic operation conditions. We hypothesize that the degradation of cell performance may be due to catalyst detachment during prolonged solution cycling and bubble evolution (Supplementary Fig. 36), as we detected only ~20 ppb of dissolved Ru in the solution after stability testing. In addition, the H₂ generated from the 4f-Nd-RuO₂/PEM/Pt/C electrolyzer was collected via gas drainage. The average rate of H₂ generation is 5.133 × 10⁻⁷mol s⁻¹ and the Faraday efficiency is ~99.1%.

Fig. 5. Electrocatalytic OER in a PEM electrolyzer.

a Schematic diagram of the PEM electrolyzer (CP, carbon paper; Pt@Ti foam, Pt-plated Ti foam). b Polarization curves of PEM electrolyzers using 4f-Nd-RuO₂, n-RuO₂, and c-RuO₂ as anodic catalysts and commercial Pt/C as a cathodic catalyst. c Chronopotentiometry curves of 4f-Nd-RuO₂ and n-RuO₂ operated at 0.1 A cm⁻². All measured in PEM electrolyzers were without iR correction. Source data are provided as a Source Data file.

In summary, we uncover that the electrocatalytic OER performance of RuO₂ is significantly improved by a 4f-induced covalent polarity modulation strategy, and the activities of 4f-RuO₂ exhibit a volcano trend as a function of the Ru−O polarity based on the results of sXAS. In particular, 4f-Nd-RuO₂ exhibited an ultralow overpotential of 214 mV for delivering a current density of 10 mA cm⁻² attributed to the moderate Ru−O polarity modulated by Nd intercalation, which optimized the adsorption energy of the oxygen intermediates during the acidic OER process. More importantly, in contrast to n-RuO₂, the incorporated REs donated electrons that reduced the valence state of Ru, weakened the Ru 4d–O 2p hybridization, and decreased the Ru–O polarity via the valence f–p–d gradient orbital coupling. Thus, the 4f-orbital-modification effect greatly inhibits the participation of lattice oxygen to avoid over-oxidation of Ru in acidic media, resulting in enhanced catalytic stability of the 4f-RuO₂ catalysts. This work not only introduces a strategy for engineering Ru–O polarity to enhance the catalytic performance of Ru-based catalysts but also develops an activity descriptor of the metal–oxygen polarity to predict effective OER catalysts in acidic media.

Methods

Materials

La(NO₃)₃·6H₂O (≥99.9%), Ce(NO₃)₃·6H₂O (≥99.9%), Nd(NO₃)₃·6H₂O (≥99.9%), Pr(NO₃)₃·6H₂O (≥99.9%), Sm(NO₃)₃·6H₂O (≥99.9%), Gd(NO₃)₃·6H₂O (≥99.9%), Dy(NO₃)₃·5H₂O (≥99.9%), Er(NO₃)₃·6H₂O (≥99.9%), Yb(NO₃)₃·5H₂O (≥99.9%), Lu(NO₃)₃·6H₂O (≥99.9%), RuCl₃·xH₂O (≥99.9%), HClO₄ (70.0–72.0%), Nafion (5%) and commercial RuO₂ (≥99.9%) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Carbon black (BP2000) was purchased from Cabot Corporation. Pt/C (20 wt%) was purchased from Suzhou Sinero Technology Co., Ltd. All the chemicals were used directly without further purification. Deionized (DI) water (18.2 MΩ cm resistivity) was used in all experiments.

Synthesis of 4f-RuO₂

Typically, 0.3 mmol of RuCl₃ and 0.1 mmol of rare earth element salts (La(NO₃)₃·6H₂O, Ce(NO₃)₃·6H₂O, Nd(NO₃)₃·6H₂O, Pr(NO₃)₃·6H₂O, Sm(NO₃)₃·6H₂O, Gd(NO₃)₃·6H₂O, Dy(NO₃)₃·5H₂O, Er(NO₃)₃·6H₂O, Yb(NO₃)₃·5H₂O, and Lu(NO₃)₃·6H₂O) were dissolved in 50 ml of 1.0 M HCl solution and sonicated for 1 h. Carbon black (0.18 g) was added and dispersed by vigorous stirring for 12 h. Then, the mixture was dried at 70 °C for 12 h by using a rotary evaporator to obtain the powder. The powder was annealed in a flowing Ar/H₂ (5% H₂) atmosphere at 900 °C for 2 h and further annealed in air at 400 °C for 3 h at a heating rate of 5 °C min⁻¹. Finally, the obtained product was treated with 1.0 M HCl for 12 h, then rinsed with DI water 3 times and dried in a vacuum oven at 60 °C overnight, yielding the target catalysts. La (4f⁰5d¹6s²), Ce (4f¹5d¹6s²), Pr (4f³5d⁰6s²), Nd (4f⁴5d⁰6s²), Sm (4f⁶5d⁰6s²), Gd (4f⁷5d¹6s²), Dy (4f¹⁰5d⁰6s²), Er (4f¹²5d⁰6s²), Yb (4f¹⁴5d⁰6s²), and Lu (4f¹⁴5d¹6s²) were selected to be introduced into RuO₂. These REs trivalent ions have the different number of 4f electrons covering fully empty to fully filled 4f orbitals, and empty 5d/6s orbitals, which contributes to studying the influence of RE 4f–O 2p–Ru 4d orbitals coupling on the electrocatalytic performance. Notably, the introduction of La ions without 4f electrons is a non-4f control in this work. For the synthesis of n-RuO₂, the general procedures were similar to those of the synthesis of 4f-RuO₂, besides the addition of carbon black (0.12 g) and the absence of rare earth element salts.

Morphology and structural characterization

Transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HRTEM), and energy dispersive spectroscopy (EDS) were performed on a JEM-2100F microscope with an acceleration voltage of 200 kV. The powder X-ray diffraction (XRD) patterns were measured on a Philips X’Pert Pro Super X-ray diffractometer with Cu Kα radiation (λ = 1.54178 Å). Spherical aberration corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy dispersive spectroscopy (EDS) were conducted on a Themis Z (300 kV) microscope. X-ray photoelectron spectroscopy (XPS) was recorded on a Thermo ESCALAB 250Xi with Al K_α (hν = 1486.6 eV) as the excitation source. With reference to C 1s to 284.8 eV, the binding energies obtained in XPS analysis were corrected. The O K-edge X-ray absorption spectra (XAS) were measured at the BL12B station of National Synchrotron Radiation Laboratory (NSRL) in the total electron yield mode in a vacuum chamber (<5 × 10⁻⁸ Pa). Inductively coupled plasma atomic emission spectroscopy (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) were performed on an Optima 7300 DV (Perkin-Elmer) and a PlasmaQuad 3 (VG Instruments), respectively. Electron paramagnetic resonance (EPR) spectra of samples were obtained using a JEOL JES-FA200 electron spin resonance spectrometer.

Electrochemical measurements in a three-electrode system

All the electrochemical measurements of hydrogen evolution reaction (OER) performance were performed in a standard three-electrode system with an electrochemical workstation (CHI760E, CH instruments) operated with a glassy carbon electrode (GCE) as the working electrode, a carbon electrode as the counter electrode, and an Ag/AgCl electrode (saturated KCl solution) as the reference electrode. In a typical preparation procedure for the working electrode, 5 µL of the homogeneous catalyst solution, which was prepared by dispersing 5.0 mg of catalyst and 40 µL Nafion solution (5 wt%) in mixed solution containing 750 μL DI water and 250 μL ethanol, was loaded onto glassy carbon electrode (GCE) of 3 mm in diameter (catalyst loading is about 0.34 mg cm⁻²). All the tests were carried out in 0.1 M HClO₄ solution at room temperature (25 °C). A quantity of HClO₄ (70.0–72.0%) is slowly added to the deionized water. The mixture was stirred and cooled, and then sealed and stored in a dry place at room temperature (25 °C). Multiple pH tests were performed on the configured electrolyte, and the average pH is 1.0 ± 0.1. The measured solution resistance is 40.5 ± 2.5 Ω after several measurements to obtain the average value. All final potentials were converted to reversible hydrogen electrode (RHE) with the conversion E (vs. RHE) = E (vs. Ag/AgCl) + 0.197 V + 0.059 × pH, which was corrected experimentally by using Pt/C as the working electrode potential in a H₂-saturated solution. Linear sweep voltammetry (LSV) curves were obtained at a rate of 10 mV s⁻¹ with 95% iR correction. The electrochemical double-layer capacitance (C_dl) was calculated by cyclic voltammetry curves in the region of between 0.956 and 1.056 V (vs. RHE) with scanning rates of 10, 20, 40, 60, 80, and 100 mV s⁻¹. The electrochemically active surface area (ECSA) values were calculated from the measured double layer capacitance divided by the specific capacitance of an atomically smooth material: ECSA = C_dl/C_s (C_s = 0.035 mF cm⁻²). The measured potential of the electrochemical impedance spectroscopy (EIS) instrument was 1.450 V (vs. RHE) in the frequency range of 0.1–100,000 Hz. For stability tests, the catalyst ink was coated on FTO (1 × 1 cm², catalyst loading is about 0.34 mg cm⁻²). After stability testing, there was a slight loss of 0.35 wt% Ru and 2.17 wt% Nd in the electrolyte for 4f-Nd-RuO₂ (Supplementary Table 3).

Electrochemical measurements analysis

Contributions from the capacitive current of remaining carbon and the back reaction need be neglected under low overpotentials. To assess the Tafel slopes of n-RuO₂ and 4f-Nd-RuO₂, overpotential intervals corresponding to current densities of ~3–50 mA cm⁻² were chosen. For c-RuO₂, the selection was limited to an overpotential interval at current densities ranging from ~1–20 mA cm⁻², due to the relatively small capacitive current of remaining carbon and its slow kinetic response.

Typically, a constant potential beyond the onset is chosen for the EIS measurement, where all catalysts show catalytic activities. However, the bubbles generated at large potentials or high current densities may result in the jittering of the data during the testing. Therefore, trade-offs between different catalysts, a potential of 1.45 V vs. RHE in the EIS measurement was used.

The mass activity (MA, A g_Ru⁻¹) can be calculated according to the equation: MA = (j × A)/m, where j is the current density at 1.53 V (vs. RHE), A is the surface of the electrode, and m is the active metal loading mass on the electrode.

The turnover frequency (TOF, s⁻¹) can be calculated according to the equation: TOF = (j × A)/(4 × F × M), where j is the current density at 1.53 V (vs. RHE), A is the surface of the electrode, F is the Faradaic constant (96,485 C mol⁻¹), and M is the number of moles of active metal on the electrode. It is assumed that every Ru atom of the catalyst participates in the reaction, which might underestimate the real TOF.

The S-number can be calculated according to the equation: S-number = (j × A ×  t)/(4 × F × n), where j is the current density (10 mA cm⁻²), A is the surface of the electrode, t is the time of reaction, F is the Faradaic constant (96,485 C mol⁻¹), and n is the number of moles of dissolved Ru in the solution (extracted from ICP-MS data).

The H₂ generated from the 4f-Nd-RuO₂/PEM/Pt/C electrolyzer was collected via gas drainage. The electrolyzer was operated continuously at 60 °C under a current density of 0.1 A cm⁻². After the operation stabilized, we collected H₂ for ~80 min and repeated the experiment three times. The average rate of H₂ generation can be calculated according to the equation: Average rate = V/(V_m × t), where V is the volume of gas, V_m is the molar volume of gas (24.5 L mol⁻¹, 25 °C), and t is the time of gas collection. The Faraday efficiency can be calculated according to the equation: FE = (2 × V × F) / (V_m × j × t) × 100%, where V is the volume of gas, F is the faradaic constant (96,485 C mol⁻¹), V_m is the molar volume of gas (24.5 L mol⁻¹, 25 °C), j is the current density (0.1 A cm⁻²), and t is the time of gas collection.

In situ XAFS measurements

The XAFS measurements of the Ru K-edge and Nd L₃-edge were carried out at the 1W1B station in Beijing Synchrotron Radiation Facility (BSRF, China) and NW10A station in Photon Factory Advanced Ring (PF-AR), High Energy Accelerator Research Organization (KEK, Japan). The storage ring of the BSRF was operated at 2.5 GeV with a maximum current of 250 mA. The electrochemical in situ Ru K-edge XAFS tests were performed with a homemade cell in a 0.1 M HClO₄ electrolyte in fluorescence mode. The electrocatalyst ink was dropped on a carbon cloth (1 × 1 cm²) as the working electrode. Similarly, a carbon rode is the counter electrode, and a Ag/AgCl electrode (saturated KCl solution) electrode is the reference electrode. First, the spectra were collected for the sample under potential free conditions (Ex situ). In situ spectra were collected by chronoamperometry measurements at a series of representative potentials (1.25–1.50 V vs. RHE).

XAFS data analysis

To obtain the detailed structural parameters around Ru atoms in the samples, the EXAFS data were processed according to the standard procedures using the ATHENA and ARTEMIS module implemented in the IFEFFIT software packages⁶⁷. The k²-weighted χ(k) data in the k-space were Fourier-transformed to real R-space using hanning windows (dk = 1.0 Å⁻¹). For the fitting of Ru K-edge data, the amplitude reduction factor S₀² was treated as adjustable variable and the obtained value of 0.73 for Ru foil, which was fixed in fitting the subsequent Ru K-edge data for 4f-Nd-RuO₂ and n-RuO₂. The Ru–O scattering path was considered to fit the EXAFS data.

In situ SR-FTIR measurements

In situ synchrotron radiation FTIR measurements were collected at the BL01B station in National Synchrotron Radiation Laboratory (NSRL, China) through a homemade top-plate cell. A platinum wire was used as the counter electrode, and an Ag/AgCl electrode and a catalyst-coated carbon cloth (1 × 1 cm²) served as the reference and working electrodes, respectively. Each infrared absorption spectrum was acquired by averaging 514 scans with a resolution of 2 cm⁻¹. To avoid signal differences caused by sample shedding, infrared spectra were collected after a constant potential was applied to the catalyst electrode for 5 min. Before each systemic OER measurement, the background spectrum of the electrocatalyst electrode was acquired under potential free conditions (Ex situ), and then the measured potential ranges of the OER were 1.10–1.50 V (vs. RHE).

Electrochemical measurements in PEM electrolyzers

To bridge the gap between laboratory research and industrial application, it is imperative to conduct device-level testing of catalysts under industrially relevant conditions, namely high current densities (>0.1 A cm⁻²) and high temperature (>60 °C). Before constructing the single cell, the Nafion 115 membrane (N115, DuPont, thickness 127 μm, size 2 × 2 cm²) was first sequentially treated by 3 wt% H₂O₂, DI water, 0.5 M H₂SO₄, and DI water at 80 °C for 0.5 h each step. After cooling to room temperature (25 °C), the treated N115 membrane was preserved in DI water. The commercial Pt/C catalyst (20 wt%, ~1.0 mg cm⁻²) was sprayed onto a carbon paper (1 × 1 cm²) as the hydrogen evolution reaction (HER) cathode. The 4f-Nd-RuO₂ catalyst (~2.0 mg cm⁻²) was sprayed onto a platinized titanium fiber felt (1 × 1 cm²) as the OER anode. After drying, the anode was annealed in air for 30 min at 300 °C. During the test, the cell was operated at the temperature of 60 °C using 0.1 M HClO₄ as the aqueous solution under the flowing rate of 1 ml min⁻¹. The polarization curves were measured in galvanostatic mode at 0.05–1.0 A cm⁻². The stability was evaluated by chronopotentiometry at 0.1 A cm⁻². All cell voltages measured in PEM electrolyzers were reported without iR correction.

Author contributions

Q.L. supervised the project. X.Z., Y.Z., and W.C. designed and performed the experiments. J.Z., S.B., and H.W. performed the characterizations. C.Y., B.P., M.S., and W.C. carried out in situ spectroscopy measurements and data analysis. X.C. and C.W. performed the density function theory calculations. X.Z. and Q.L. wrote the manuscript. All authors discussed the results and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available in the main text and Supplementary information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-62258-z.