Durable acidic water oxidation ruthenium based electrocatalyst by fluorination induced symmetry breaking

Abstract

Ru–based materials exhibit high electrocatalytic activity for the acidic oxygen evolution reaction, but they are prone to transform into soluble RuO₄ species via traditional lattice oxygen mechanism at high oxidation potentials, resulting in rapid inactivation. Herein, density functional theory calculations predict that the F induced symmetry–breaking can alter the oxygen evolution reaction route of RuO₂ from the lattice oxygen mechanism route to the stable adsorption evolution mechanism pathway. Consequently, we fabricate an efficient F–RuO₂/FC electrocatalyst by substituting a portion of O in RuO₂ with F. Specifically, it can operate continuously for over 1440 h (2 months) at 100, 500 and 1000 mA cm^–2 when using F–RuO₂/FC as anode electrocatalyst in proton exchange membrane water electrolyzer. Detailed in situ experiments demonstrate that the adsorption of intermediates is impacted by the presence of F in RuO₂, thus forcing the oxygen evolution reaction to proceed via the adsorption evolution mechanism. This study provides experimental and theoretical insights for customizing stable Ru–based electrocatalysts for water splitting and beyond.

Introduction

Hydrogen energy has received considerable attentions in scientific research and industry attributed to its environmentally friendly and sustainable characteristics over the past decades^1,2,3,4. Electrocatalytic water splitting based on polymer electrolyte membrane (PEM) is considered a promising strategy for hydrogen production due to their minimal gas crossover, large current density, and high gas purity^5,6,7,8. Regrettably, the demanding acidic environment and large oxidation potential make it necessary to use corrosion–resistant yet expensive iridium (Ir) ( ~ US$140 g^–1) as anode electrocatalysts in practical industrial applications^9,10,11. It is predicted that the requirement for Ir in all proposed PEM electrolyzers is ninefold the current global output for 2030 in the Net Zero Emissions Scenario, undoubtedly exacerbating the critical situation of low iridium reserves¹². Therefore, the construction of robust Ir–free OER electrocatalysts is highly desired yet challenging for PEM electrolyzers.

Ruthenium (Ru), as an acidic OER electrocatalyst, is anticipated to supplant Ir because of its elevated electrocatalytic activity, greater earth abundance, and economical cost ( ~ US$16 g^–1)^13,14,15. However, the characteristics of thermodynamic instability under acidic OER conditions severely limit their industrialization process, primarily due to the fact that traditional Ru–based materials tend to follow the lattice oxygen mechanism (LOM) that lattice oxygen (O_L) is participates in the reaction process, which inevitably leads to structural collapse^16,17. Initially, researchers sought to enhance the OER performance of Ru–based materials by introducing oxygen vacancy (O_V)¹⁸. Regrettably, it has been argued that the generated O_V leads to the exposure of Ru, which tends to over–oxidize to form soluble high valence Ru^x>4, ultimately compromising stability¹⁹. Therefore, several schools of thought believe that making the formation energy of O_V higher than the redox H₂O/O₂ energy may be a promising strategy to break the stability bottleneck of Ru–based materials^20,21. As so far, numerous feasible approaches have been reported to significantly enhance the formation energy of O_V, including heterometal doping (SnRuO_X²², Ni–RuO₂²³, Ta₁/RuO₂²⁴), interface engineering (Co–Ru@RuO₂ 9:1²⁵, GB–RuO₂²⁶), strain regulation (Li_0.52RuO₂²⁷, RuSnO_X²⁸), high–entropy compounds (M–RuIrFeCoNiO₂²⁹), anion regulation (LD–B/RuO₂³⁰, Ru/Se–RuO₂³¹, RuO₂–Cl³²) and so on. Among these, the electronegativity and electron affinity of anions are more conducive to optimizing the adsorption/desorption behavior of reaction intermediates on the electrocatalyst surface³³. Nevertheless, most recently reported anion–modulated catalysts for OER have predominantly focused on performance enhancement while neglecting the intrinsic connection between the reaction mechanism and microstructure, which impedes the in–depth understanding of highly efficient electrocatalysts. Fundamentally, the intrinsic properties of materials overwhelmingly depend on their crystal symmetry and atomic arrangement. The defects in a lattice may reduce its crystal symmetry, which is named the symmetry–breaking effect^{34,35,36,37,38}. Consequently, it is reasonable to speculate that the energy barrier for O escape can be regulated to suppress the participation of O_L in the OER process by breaking the symmetry of the octahedral configuration. Simultaneously, the reconfigured electrons on Ru will alter the adsorption of intermediates, thus forcing the Ru–based materials from the inherent LOM pathway toward a more robust and stable route.

Herein, the acidic OER mechanism of RuO₂, predicted by using density functional theory (DFT), has been successfully switched by highly electronegative F–induced symmetry–breaking. As a result, the optimized F–RuO₂/FC exhibits robust OER performance with low overpotential and long–term stability in acidic media. In addition, only cell voltages as low as 1.58, 1.70, 1.89, and 2.05 V are required to reach current densities of 500, 1000, 2000, and 3000 mA cm^–2, respectively, in a scalable PEM water electrolysis (PEMWE) device. It is worth noting that it can operate stably for up to 1440 h (2 months) at 100, 500, and 1000 mA cm^–2. In situ experiments demonstrate that the reaction mechanism follows the adsorption evolution mechanism (AEM) pathway when part O in RuO₂ is replaced by F, ensuring long–lasting stability. This work pioneers a fresh horizon for constructing robust catalysts in the electrochemical field by anion–induced symmetry–breaking.

Results

Theoretical simulations

The widely recognized acidic OER mechanism on RuO₂ is LOM pathway, in which O_L interacts with water–oxygen (O_W) to facilitate the formation of O–O intermediate species, concurrently creating O_V (Fig. 1a). The generated O_V can also serve as active site to accelerate the OER process, ultimately leading to structural collapse and leaching of metal Ru, thus presenting rapid deactivation. In contrast, converting the traditional LOM pathway to a more stable AEM route by anion–induced symmetry–breaking may be a feasible strategy to improve the stability of Ru–based materials (Fig. 1b). Evidently, the oxygen involved comes from water rather than O_L, thereby ensuring robust structural integrity during the OER process. Subsequently, DFT simulations were used to accelerate the discovery of high–performance electrocatalysts, thus M–RuO₂ (M = F, S, Se, P, N, Cl, and Br) was constructed to evaluate their OER performance. According to the results, a more energy–favorable structural model was adopted for subsequent calculations (Supplementary Figs. 1–7 and Supplementary data 1). The Gibbs free energy of rate–determining step (RDS) for F–RuO₂ is 1.80 eV, lower than that of N–RuO₂ (1.83 eV), S–RuO₂ (1.86 eV), P–RuO₂ (2.60 eV), Se–RuO₂ (2.76 eV), Cl–RuO₂ (3.16 eV), Br–RuO₂ (4.28 eV) and RuO₂ (1.89 eV) (Fig. 1c and Supplementary Figs. 8–15). In addition, the observed hydrogen bonds between H and adjacent O indicate a low value of the calculated enthalpy change, which serves to stabilize OOH adsorption at the coordinatively unsaturated Ru site on the F–RuO₂(110) surface (Supplementary Fig. 16). The charge distinction for the two O species adsorbed on F–Ru and O–Ru is about 0.06 e, which demonstrates the improved O atom binding on F–RuO₂(110), thereby fostering the genesis of O−O coupling and ultimately leading to the production of molecular O (Supplementary Fig. 17). Furthermore, the electron localization function (ELF) of the *OOH adsorbed on F–RuO₂ surfaces shows a higher ELF_max value than RuO₂, which indicates an enhanced covalent bond of OOH intermediate induced by the stronger interaction of *OOH on the F–RuO₂ (Supplementary Fig. 18). Moreover, the limiting potential for OER on the F–RuO₂(110) surface was estimated at 1.80 V, representing a reduction of approximately 0.11 V compared to RuO₂ and implying enhanced OER activity (Supplementary Fig. 19). The outcomes of the calculations revealed that the d–band structure of Ru and the 2p–band structure of O undergo modulation due to the incorporation of F (Fig. 1d). Density of states (DOS) indicates that O 2p band center moves from –2.96 eV (RuO₂) to –3.23 eV with the introduction of F, signifying an elevated formation energy for O_V, which consequently suppresses the dissolution of Ru in acidic media. Notably, a decrease of the Ru d–band center relative to fermi energy level (E_f) is observed after the partial substitution of O with F, potentially influencing the adsorption/desorption of intermediates on the catalyst surface³⁹. Therefore, compared with the traditional RuO₂ following LOM route, the introduction of F gives RuO₂ more energy advantages on the AEM path (Supplementary Fig. 20). Furthermore, similar phenomena have been observed on (111) and (100) planes of F–RuO₂ and RuO₂, proving its universality (Supplementary Fig. 21).

Fig. 1: DFT calculations.

Schematic illustration of a LOM and b AEM route for OER on RuO₂ and M–RuO₂ surface. (c) Gibbs free energy diagram for different catalysts. d DOS of Ru 3d and O 2p for F–RuO₂ and RuO₂; corresponding d–band centers are denoted by white lines. e Calculated enthalpy changes for subsurface oxygen loss of various catalysts. f Reaction free energy and subsurface O loss enthalpy change volcano against the symmetry–breaking. The gray, red, blue, pink, yellow, orange, green, cyan and brown balls represent Ru, O, F, N, S, P, Se, Cl and Br atoms, respectively. Source data for this figure are provided as a Source Data file.

To further assess the stability of F–RuO₂, the enthalpy of F within bulk C–centered orthorhombic structure was used to compute their chemical potential, indicating resistance to the detachment of F from RuO₂ (Supplementary Fig. 22). Additionally, the enthalpy change for subsurface oxygen loss was calculated in Supplementary Figs. 23–30. The enthalpy change of subsurface O loss on the F–RuO₂ was significantly higher than that of RuO₂ modified by other anions, indicating that symmetry breaking also affects the stability of O_L in RuO₂ (Fig. 1e). The examination of the enthalpy change associated with the demetallation process of the surface Ru atom from both RuO₂(110) and F–RuO₂ (110) was evaluated to further investigate the stability (Supplementary Fig. 31). It can be found that the energy expenditure for Ru demetallation elevated from 1.72 to 2.52 eV after the symmetry of octahedral configuration in RuO₂ is broken by partial substitution of O with F, indicating increased stability of the surface Ru in F–RuO₂ compared to RuO₂. It is clear that the Gibbs free energy of the RDS varies with symmetry breaking in an inverted volcano shape (Fig. 1f and Table S1). In contrast, the enthalpy change related to subsurface oxygen loss displays a volcano–shaped profile as a function of symmetry breaking. Note of that F–RuO₂ is located at the bottom/top of the volcano due to the presence of high electronegativity but small atomic radius F, which induce slight symmetry breaking in the octahedral configuration of RuO₂, consequently reducing the reaction energy barrier and enhancing its stability.

Catalyst synthesis and characterization

Of the candidate electrocatalysts predicted by theoretical calculations, F–RuO₂ was considered to be the most promising for OER activity and stability. Therefore, the spherical FC and RuCl₃ were ultrasonicated at room temperature (25 °C) and then dried to obtain FC supported RuCl₃ (RuCl₃/FC) (Supplementary Figs. 32, 33). Subsequently, the F–RuO₂/FC was successfully achieved by heat treating the RuCl₃/FC at 400 °C in air for 4 h. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) characterization revealed that the spherical FC emerged with abundant pore structures (Supplementary Fig. 34). However, the intact spherical can be maintained for FC without the addition of RuCl₃, indicating that the presence of Ru will accelerate the local pyrolysis of FC and thus lead to a rich pore structure (Supplementary Fig. 35). Therefore, the pyrolysis temperature and time of FC will be ameliorated once Ru is present, thus the specific surface area of F–RuO₂/FC can be significantly enhanced as evidenced by N₂ absorption–desorption isotherm (Supplementary Figs. 36, 37). In addition, long–range ordered atomic arrangements were successfully detected using a spherical aberration–corrected high–angle annular dark–field scanning TEM (AC–HAADF–STEM) (Fig. 2a). The well–defined lattice spacings (0.22 and 0.18 nm) of (020) and (21$\bar{1}$) planes for RuO₂ unequivocally indicate the presence of lattice strain compared to RuO₂, which may be caused by symmetry breaking due to introduction of heterogeneous atoms (Fig. 2b). The fast Fourier transform (FFT) pattern, inset in Fig. 2b, reveals the exposed (020) and (21$\bar{1}$) planes along the [$\bar{1}$0$\bar{2}$] zone axis of F–RuO₂/FC. To confirm the heterogeneous atoms in F–RuO₂/FC, STEM energy dispersive spectrometer (STEM–EDS) elemental mappings were conducted in Fig. 2c. In addition to the uniformly distributed Ru and O elements, conspicuous F signals can be observed, suggesting that the lattice strain in F–RuO₂/FC may be caused by the incorporation of F. Subsequently, geometrical phase analysis (GPA) was performed to evaluate the strain level and distribution in the F–RuO₂/FC. Evidently, the reduced and enhanced strain can be observed along the horizontal (ε_xx) and vertical (ε_yy) direction, respectively (Fig. 2d, e). The strain profiles of line1 and line2 acquired from strain mappings further confirm that the compressive stress is predominantly along the ε_xx direction and the tension stress along the ε_yy direction (Fig. 2f)⁴⁰. To investigate the critical insight of chemical distributions for F–RuO₂/FC, time of flight secondary ion mass spectrometry (TOF–SIMS) analysis was performed in Fig. 2g. The element intensity versus sputter time reveals that the concentration of O^– is notably greater than that of Ru⁺, being consistent with the characteristics of RuO₂. In addition, the detected RuF^– signals indicate that F indeed replaces part of O and then bonds with Ru during the heat treatment process. Two–dimensional (2D) TOF–SIMS elemental mappings also confirm the presence of significant RuF^– species alongside the abundant Ru and O species (Fig. 2h). Clearly, the distribution of Ru and O signals throughout the space in three–dimensional (3D) TOF–SIMS views indicates that the predominant species is RuO₂, consistent with the AC–HAADF–STEM analysis (Fig. 2i). The uniform distribution of RuF^– signals throughout the entire 3D space further confirms the partial substitution of O by F, thus leading to the formation of Ru–F bonds. X–ray diffraction (XRD) with Rietveld refinement was conducted to obtain the detailed unit cell parameter (Fig. 2j and Table S2). The sharp diffraction peaks correspond to the tetragonal RuO₂ with the space group P42/mnm, indicating that the adsorbed RuCl₃ on the surface of FC gradually crystallizes into RuO₂ during the heat treatment process. The presence of Ru accelerates the local pyrolysis of FC and resulting in F occupying part of the position of O in F–RuO₂/FC (inset in Fig. 2j). The F elements were removed by calcining the FC at 400 °C for 12 h (denoted as (F)C) (Supplementary Fig. 38). Subsequently, M–RuO₂/(F)C was successfully prepared by adsorption of RuCl₃ on the (F)C combined with heat treatment process (Supplementary Figs. 39–46). Furthermore, the Baur distortion index (D_i) was employed to quantify the asymmetry of virtual octahedral units within the RuO₂ crystal structure:

$${D}_{{{\rm{i}}}}=\,\frac{1}{6}{\sum }_{{{\rm{i}}}=1}^{6}\left(\frac{|{b}_{{{\rm{i}}}}-\bar{b}|}{\bar{b}}\right)$$

(1)

where ${b}_{{{\rm{i}}}}$ and $\bar{b}$ represents the individual bond length and the average of six bond lengths, respectively⁴¹. Notably, the D_i, based on the Rietveld refinement of XRD patterns results, were increase in the order F < N < S < P <Se <Cl <Br, indicating an increase of octahedral symmetry breaking degree, consistent with DFT predictions (Supplementary Fig. 47 and Table S3–S8).

Fig. 2: Structural characterizations.

a,b spherical aberration–corrected HAADF–STEM images, c STEM–EDS elemental mappings of F–RuO₂/FC. Strain mappings of F–RuO₂/FC along d ε_xx and e ε_yy direction from GPA and f line profiles of strain along the dotted white lines in (d) and (e). g TOF–SIMS depth analysis of Ru⁺, O^–, and RuF^– for F–RuO₂/FC and h corresponding 2D elemental mapping images and i 3D render overlay images. j XRD pattern of F–RuO₂/FC. Inset in (b) and (j) are the FFT pattern and structural model of F–RuO₂/FC, respectively. The gray, red and blue balls represent Ru, O and F atoms, respectively. Source data for this figure are provided as a Source Data file.

Subsequently, Raman spectra were carried out in Fig. 3a, where the typical peaks at 506.8, 622.2, and 693.4 cm^–1 are attributable to the vibration modes of E_g, A_1g, and B_2g for Ru–O bonds, respectively⁴². Compared to RuO₂, a noticeable red shift can be observed in the vibration peaks of F–RuO₂/FC, which may be caused by the shortened Ru–O bond owing to the partial substitution of O by the highly electronegative F⁴³. Therefore, F 1s XPS spectra of F–RuO₂/FC present a wider peak than that of FC, corresponding to the C–F and Ru–F species (Fig. 3b). Furthermore, the content of F in F–RuO₂/FC varies with the calcination temperature, which can be confirmed by EDS, XPS, and Raman analysis (Supplementary Figs. 48–54 and Table S9, S10)^44,45. Additionally, O 1s XPS peaks of RuO₂ are located at 529.0, 530.0, and 531.7 eV, corresponding to O_L, O_V, and surface–adsorbed H₂O, respectively (Fig. 3c)^46,47. Evidently, a notable shift of the O species towards higher binding energy can be observed in F–RuO₂/FC due to the presence of high electronegativity F results in an altered electronic interaction between Ru and O⁴³. Compared with RuO₂, Ru 3p XPS peaks for F–RuO₂/FC are negative shift of 0.35 eV, revealing a lower valence state than Ru⁴⁺ (Fig. 3d). Furthermore, X–ray absorption near–edge structure (XANES) and extended X–ray absorption fine structure (EXAFS) was executed to scrutinize the electronic structure. The lower absorption energy for F–RuO₂/FC compared to RuO₂ suggests that the average valence state of Ru in F–RuO₂/FC is below +4, which is attributed to the partial substitution of O by highly electronegative F (Fig. 3e). The chemical structure and coordination environment of the F–RuO₂/FC was analyzed by Fourier transform of the EXAFS (FT–EXAFS) spectra, where the peak at ~1.42 Å in F–RuO₂/FC can be attributed to Ru–O or Ru–F coordination environment (denote as Ru–O/F) (Fig. 3f and Supplementary Fig. 55). Obviously, the diminished intensity of Ru–O/F bonds for F–RuO₂/FC as opposed to RuO₂ suggests that the introduction of F leads to an unsaturated coordination environment (Table S11). In addition, charge density difference and Bader charge analysis also reveals that the valence state of Ru (+1.34) in F–RuO₂ decreases with increasing surrounding electron density, attributing to the electron density surrounding Ru was readjusted as a result of the octahedral symmetry breakdown induced by the highly electronegative F, potentially altering the adsorption of intermediates on RuO₂ and thereby switching the reaction route (Fig. 3g–j).

Fig. 3: Electronic structure investigation.

a Raman spectra of F–RuO₂/FC and RuO₂. b F 1s XPS spectra of F–RuO₂/FC and FC. c O 1s and d Ru 3p XPS spectra of F–RuO₂/FC and RuO₂. Ru K–edge e XANES and f EXAFS of F–RuO₂/FC, Ru foil and RuO₂. Electron density difference of g F–RuO₂/FC and h RuO₂. The blue and yellow shaded area mean the electron density accumulation and donation and the isosurface value is 0.065 e/ų. Bader charge analysis for Ru (gray), F (blue), and O (red) sites of i F–RuO₂/FC and j RuO₂. Source data for this figure are provided as a Source Data file.

Evaluation of electrocatalytic activity

To verify the prediction of DFT calculation, the catalytic activities of various catalysts towards water oxidation were evaluated using a representative three–electrode setup in 0.5 M H₂SO₄ (Fig. 4a and Supplementary Fig. 56). Notably, the F–RuO₂/FC exhibit a lower overpotential (192, 243, and 296 mV) than those of RuO₂/C (230, 306, and 370 mV) and RuO₂ (270, 357, and 428 mV) for achieving a current density of 10, 50, and 100 mA cm^–2 (Fig. 4a). The rapid OER kinetics of F–RuO₂/FC is evidenced by a low Tafel slope value of 41.6 mV dec^–1 compared to RuO₂/C (69.7 mV dec^–1) and RuO₂ (76.0 mV dec^–1) (Fig. 4b). The OER activities of overpotential and Tafel slope for various catalysts are visually presented in Fig. 4c, where F–RuO₂/FC exhibit the smallest overpotential and the Tafel slope value, exhibiting competitive activity among the Ru–based catalysts reported recently (Table S12). Furthermore, electrochemically active surface area (ECSA), which is related to the electrochemical double–layer capacitances (C_dl), was conducted to evaluate the OER activity (Supplementary Fig. 57). The F–RuO₂/FC (29.3 mF cm^–2) exhibit higher C_dl values compared to RuO₂/C (15.6 mF cm^–2) and RuO₂ (8.9 mF cm^–2). Remarkably, the high turnover frequency (TOF) of F–RuO₂/FC further confirming that the incorporation of F in RuO₂ produced more active sites (Supplementary Fig. 58). In addition, durability is also a crucial factor in assessing their potential for practical application. Therefore, chronopotentiometry tests were performed in Supplementary Fig. 59, where the stability of RuO₂/C and RuO₂ can be observed to be similar, suggesting that the contribution of existing C towards stability is negligible. Importantly, F–RuO₂/FC can operate continuously for over 60 h with slightly dissolution of Ru, indicating that the partial substitution of O by F in RuO₂ can significantly improve stability even under harsh acidic conditions (Supplementary Fig. 60). Moreover, the flat polarization curve and fleeting stability prove that FC is inert for OER (Supplementary Fig. 61). It is noteworthy that the activity and stability are closely resemble that of RuO₂ when it is mechanically mixed with FC, confirming that the enhanced OER performance of F–RuO₂/FC is not caused by the FC substrate (Supplementary Figs. 62, 63). In addition, the lower OER performance demonstrate its sensitivity to the fluorine doping level (Supplementary Fig. 64). Specifically, the D_i remains largely unchanged at low F doping levels, while the TOF decreases significantly, suggesting that low F doping primarily affects the number of active sites rather than the D_i (Supplementary Figs. 51a, 65). Conversely, as the F doping levels increase, D_i increases while TOF decreases, demonstrating that excessive symmetry breaking degrades OER performance (Supplementary Fig. 51b). Similarly, the heat treatment time will also affect the OER activity and stability of F–RuO₂ (Supplementary Figs. 66–69). Furthermore, the sharply reduced activity and stability of RuO₂/(F)C prove that F is decisive for improving the OER performance of RuO₂ (Supplementary Figs. 70, 71). As predicted by the DFT, the RDS energy barrier of OER and the enthalpy change caused by O loss will undergo significant changes when O was replaced by S, Se, P, N, Cl, Br, suggesting that the symmetry–breaking induced by different anions result in varied performance (Supplementary Figs. 72, 73).

Fig. 4: Electrochemical activity studies.

a OER polarization curves with 95% iR correction of F–RuO₂/FC, RuO₂/C, and RuO₂ in 0.5 M H₂SO₄ (pH = 0.3 ± 0.01, scan rate = 5 mV s^–1), the resistances of F–RuO₂/FC, RuO₂/C, and RuO₂ are 5.4 ± 0.1 Ω, 5.1 ± 0.2 Ω, and 5.2 ± 0.1 Ω, respectively. b Tafel plots of F–RuO₂/FC, RuO₂/C, and RuO₂. c Histogram of overpotentials at 10 mA cm^–2 and Tafel slopes of F–RuO₂/FC, RuO₂/C, and RuO₂. (Error bars represent the standard deviations from three independent measurements) d Polarization curves of PEM electrolyzer with F–RuO₂/FC (2 mg_Ru cm^–2) as anode and Pt/C (0.5 mg_Pt cm^–2) as cathode at 80 °C under ambient pressure. e Chronopotentiometry tests of PEM electrolyzer using F–RuO₂/FC| |Pt/C as electrocatalysts. All measured in PEM electrolyzers were without iR correction. Source data for this figure are provided as a Source Data file.

Encouraged by the efficient OER performance, F–RuO₂/FC and Pt/C were utilized as the anode and cathode electrocatalysts, respectively, to evaluate their practical application potential in the PEMWE device. Specifically, to deliver current densities of 500, 1000, 2000, and 3000 mA cm^–2, F–RuO₂/FC| |PEM| |Pt/C cell requires voltage as low as 1.58, 1.70, 1.89 and 2.05 V, respectively, which are lower than those of RuO₂| |PEM| |Pt/C (1.61, 1.78, 2.10 and 2.42 V) and RuO₂/(F)C| |PEM| |Pt/C (1.70, 1.96, 2.44 and 2.86 V) (Fig. 4d and Supplementary Fig. 74a). Subsequently, the chronopotentiometry test was carried out to evaluate the stability of F–RuO₂/FC in the PEMWE device (Fig. 4e). Excitingly, the F–RuO₂/FC–based cell can operate stably for 2 months even at 100, 500 and 1000 mA cm^–2, far surpassing RuO₂| |PEM| |Pt/C and RuO₂/(F)C| |PEM| |Pt/C, indicating the promising potential for practical applications (Supplementary Fig. 74b, c). Notably, after long–term PEMWE testing, the morphology, chemical states, and crystal structure of F–RuO₂/FC remain largely unchanged, in sharp contrast to the severely degraded RuO₂/(F)C, proving long–lasting stability (Supplementary Figs. 75–77)^48,49. Apparently, F–RuO₂/FC present a larger current density and longer stability, exhibiting competitive performance among the recently reported PEM electrocatalysts (Table S13).

Mechanism studies

To investigate the underlying mechanism during the OER process, in situ characterizations were conducted in Fig. 5. Note of that the diffraction peak intensity does not decrease significantly as the potential range from 1.0 to 1.5 V by in situ electrochemical XRD testing, implying the structure of F–RuO₂/FC remain intact during a wide range of potential tests (Fig. 5a). More importantly, the well–preserved diffraction peak in the situ monitoring for up to 60 minutes suggests that the O in F–RuO₂/FC may not participate in the OER process, thus deviating from the traditional LOM route (Supplementary Fig. 78). Therefore, the dynamic evolution of the surface structure during the OER process is monitored by in situ electrochemical Raman spectroscopy (Fig. 5b). Clearly, the vibrational peaks corresponding to Ru–O bonds (E_g, A_1g, and B_2g) are evidently present across the entire potential window, suggesting that the structure of F–RuO₂/FC is not greatly damaged, consistent with the results of in situ XRD. Significantly, a Raman peak emerges gradually at ~731 cm^–1 with the applied potential increases, compared with RuO₂/C, attributed to the formation of Ru–OOH intermediates in the AEM pathway (Supplementary Fig. 79)⁵⁰. As shown in Fig. 5c, the corresponding contour plots exhibit significantly enhanced vibration signals of B_2g for Ru–OOH intermediates except for the predominant feature signals of Ru–O bonds, proving that the reaction intermediates in the AEM pathway are indeed captured.

Fig. 5: Mechanism explorations.

a In situ XRD patterns, b in situ Raman spectra and c corresponding contour plots, d in situ ATR–IR spectra of F–RuO₂/FC. e DEMS signals of ³²O₂, ³⁴O₂, and ³⁶O₂ from the gaseous products for ¹⁸O–labeled F–RuO₂/FC catalysts in H₂¹⁶O aqueous H₂SO₄ electrolyte during three times of LSV cycles. f Schematic illustration of the mechanism transformation from lattice oxygen mechanism (LOM) to adsorption evolution mechanism (AEM) caused by F–mediated symmetry breaking on RuO₂. Source data for this figure are provided as a Source Data file.

To further elucidate the actual OER route, in situ attenuated total reflectance infrared (ATR–IR) spectroscopy was carried out to detect the reaction intermediates. As shown in Supplementary Fig. 80, two absorption peaks attributed to the *OO intermediate formed during the LOM pathway are detected, indicating that RuO₂/C follow the conventional LOM pathway during the acidic OER process⁵¹. Conversely, in situ ATR–IR spectroscopy of F–RuO₂/FC showed a sharp characteristic peak appear at 1237 cm^–1, belonging to the stretching vibration of surface–adsorbed *OOH intermediates, evidencing the dominance of the AEM pathway (Fig. 5d)⁵². Obviously, the signals of *OOH intermediates gradually appear and are significantly enhanced with the increase of potential, indicating that the incorporation of F can force the reaction mechanism of RuO₂ from the traditional LOM to AEM, thus ensuring long–lasting stability even at high applied potential. Additionally, F–RuO₂/FC show clearly pH–independence, suggesting the significant suppression of the LOM pathway and the dominant role of the AEM pathway (Supplementary Fig. 81)^53,54. Furthermore, the reaction intermediate signals of *OOH in the AEM pathway and *OO in the LOM pathway were simultaneously detected during the OER process of N–RuO₂/(F)C and S–RuO₂/(F)C, but the signals of *OOH are disappeared across the entire potential window of P–RuO₂/(F)C, Se–RuO₂/(F)C, Cl–RuO₂/(F)C and Br–RuO₂/(F)C (Supplementary Fig. 82)⁵⁵. Subsequently, operando differential electrochemical mass spectrometry (DEMS) was performed on ¹⁸O isotope–labeled F–RuO₂/FC using linear sweep voltammograms (LSV) in H₂¹⁶O–0.5 M H₂SO₄ (Supplementary Fig. 83a). As shown in Fig. 5e, the mass signals of ³²O₂ and ³⁴O₂ were detected without ³⁶O₂ products, suggesting the non–participation of O_L in the OER process⁵⁶. The mass signals of ³²O₂ gradually increase and present a constant ratio (³⁴O₂:³²O₂) with the occurrence of OER, further demonstrating the dominant AEM pathway (Supplementary Fig. 83b, c). Therefore, it can be concluded that the LOM pathway that O_L is involved in the OER process, including a labeled ¹⁸O and an ambient O (¹⁶O) from the electrolyte, resulting in the instability surface of the RuO₂ and consequently generating ¹⁶O¹⁸O species (Fig. 5f). Based on the DFT computations and detailed experimental characterization, the F induced symmetry–breaking in octahedral configuration of RuO₂ improves the interaction between adsorbent and the electrocatalysts surface, alters the OER route to follow the AEM pathway, and ultimately enhancing the electrocatalytic performance.

In summary, a highly efficient acidic OER electrocatalyst has been predicted by DFT calculation, which uses F to replace partial O in RuO₂ to break the symmetry of the octahedral configuration and thus switch the reaction path. The improved O species adsorption and attenuated demetallation process of Ru improve the OER intrinsic activity and resistance to solvation. As a validation, to deliver the current density of 10, 50, and 100 mA cm^–2, only overpotentials of 192, 243, and 296 mV need to be provided for F–RuO₂/FC in acidic conditions. More importantly, the cell voltage of 1.58, 1.70, 1.89 and 2.05 V can drive current densities up to 500, 1000, 2000, and 3000 mA cm^–2, respectively, simultaneously continuously operating at 100, 500 and 1000 mA cm^–2 for 2 months in a PEMWE device. Detailed in situ experiments reveal that the F–induced symmetry–breaking ensures the structural stability of RuO₂ during the OER process and changes their traditional OER mechanism from LOM to AEM pathway, thus enhancing the stability. This study offers a fresh perspective for altering the reaction mechanism of OER electrocatalysts based on anion–induced symmetry–breaking to realize efficient performance.

Methods

Chemicals

Ruthenium (III) chloride (RuCl₃, 97%) was obtained from Shanghai Xihang New Materials Technology Co., Ltd. Fluorocarbon (FC) was purchased from Wenzhou Bangxinyuan Technology Co., Ltd. Melamine (C₃H₆N₆, 99%), selenium (Se, 99.99%), isopropyl alcohol (C₃H₈O, 99.5%), Nafion (5 wt%) and VC–X72 carbon were obtained from Meryer. Manganese dioxide (MnO₂, 99%), sulfur (S, 99.99%), ruthenium (IV) oxide (RuO₂, 99.9%), polyvinyl chloride ((C₂H₃Cl)_n, K–value 68–65) and sodium hypophosphite (NaPO₂H₂, 99%) were obtained from Aladdin. Hexabromobenzene (C₆Br₆, 99%) was obtained from Macklin. Pt/C (20 wt%) and Nafion 115 membrane (diameter: 3 cm, thickness: 127 μm) were obtained from Sinero. Ethanol (C₂H₆O, 99.7 wt%) was purchased from Anhui Ante Food Co., Ltd. Sulfuric acid (H₂SO₄, 95.0–98.0 wt%) was obtained from Zhejiang Zhongxing Chemical Reagent Co., Ltd. Reagents were used directly without further purification, and all experiments utilized deionized water (18 MΩ cm^–1).

Preparation of F–RuO₂/FC

10 mg of FC and 5.2 mg of RuCl₃ were added to ethanol and sonicated homogeneously at ambient temperature (25 °C). Subsequently, the homogeneous solution was dried at 80 °C to yield RuCl₃/FC. At last, the RuCl₃/FC was annealed at 400 °C for 4 h in an air atmosphere (denoted as F–RuO₂/FC). In the preparation of RuO₂/C, FC was replaced by VC–X72 carbon.

Preparation of RuO₂/(F)C

The synthesis condition of RuO₂/(F)C is the same as F–RuO₂/FC, except that the FC used for RuO₂/(F)C was pre–annealed in air at 400 °C for 12 h to remove F (denoted as (F)C).

Preparation of M–RuO₂/(F)C (M = N, S, P, Se, Cl and Br)

The M–RuO₂/(F)C electrocatalyst was synthesized by vapor transport method, where the anion sources and RuO₂/(F)C (mass ratio 1:2) were placed in a quartz tube as shown in Supplementary Fig. 84. To obtain M–RuO₂/(F)C, the RuO₂/(F)C was annealed at 400 °C for 1 h (heating rate: 5 °C min^–1), while the anion sources were simultaneously heated to their respective target temperatures. In addition, melamine, sulfur powder, sodium hypophosphite, selenium powder, a mixture of polyvinyl chloride and manganese dioxide, and hexabromobenzene were used as anion sources, releasing anions under heat treatment at the target temperatures specified in Table S14, respectively.

Preparation of electrolyte

13.8 mL of sulfuric acid was slowly added to deionized water and diluted to a final volume of 500 mL in a volumetric flask. After thorough mixing, the resulting homogeneous solution was the final 0.5 M H₂SO₄ electrolyte with a pH of 0.3 ± 0.01. The prepared electrolyte was stored in a sealed container under dry conditions at room temperature (25 °C).

Electrochemical measurements

A CHI1100D workstation (CHI Instruments, Inc., Shanghai) was employed to assess the OER activity of the electrocatalysts within a standard three–electrode system. Then 2.0 mg of catalyst was added to 200 μL of a solution comprising Nafion (5 μL) and isopropyl alcohol (195 μL), and the ink was prepared by ultrasonic at 25 °C. The content of Ru is 14.71 wt%, as determined by inductively coupled plasma atomic emission spectroscopy (ICP–AES) measurements. For the preparation of the working electrode (0.196 cm², GaossUnion), 10 μL of ink was dripped onto a glassy carbon electrode and dried at 25 °C. The loading amount of Ru on working electrode is 75 μg cm^–2, based on the calculation (10 μL × 10 μg μL^–1 × 14.71 wt%) / 0.196 cm². The counter electrode and reference electrode used in the OER tests were graphite rod and saturated calomel electrode (SCE). Therein, the calibration of the reference electrode was periodically performed in an H₂–saturated 0.5 M H₂SO₄ electrolyte using a three–electrode system with a Pt wire served as both the working and counter electrodes. All OER polarization curves (95% iR compensation) are averages of the multiple test results collected at a scan rate of 5 mV s^–1. The resistance of the electrolyte was determined using iR correction.

Structure Characterizations

A FEI Tecnai F20 instrument was employed to acquire the scanning electron microscope (SEM) images of the catalysts. The microscopic morphology was examined through transmission electron microscope (TEM) imaging using a HITACHI HT700 operated at 100 kV. The surface chemical composition of the catalysts was investigated by X–ray photoelectron spectroscopy (XPS), with data acquired using a Kratos Axis Supra instrument. A FEI Talos F200S microscope operating at 200 kV was employed to characterize of microstructure and elemental composition by acquiring high–resolution transmission electron microscope (HRTEM) and energy dispersive X–ray spectroscopy (EDS) data.

Calculation of the turnover frequency (TOF)

Based on the condition that each Ru atom on the electrode was considered as an active site, the TOF were calculated according to Eq. (2):

$${{\rm{TOF}}}=\frac{{{\rm{Number}}}\; {{\rm{of}}}\;{{{\rm{O}}}}_{2}\;{{\rm{per}}}\; {{\rm{second}}}}{{{\rm{Number}}}\; {{\rm{of}}}\; {{\rm{active}}}\; {{\rm{Ru}}}\; {{\rm{sites}}}}$$

(2)

The number of oxygen per second was obtained from the LSV curves based on Eq. (3):

$${{\rm{Number}}}\; {{\rm{of}}}\;{{{\rm{O}}}}_{2}\;{{\rm{per}}}\; {{\rm{second}}} =\left(|{{\rm{J}}}|{{\rm{A}}}\right)\left(\frac{1\;{{\rm{C}}}}{{{\rm{s}}}}\right)\left(\frac{1\;{{\rm{mol}}}}{96453.8\;{{\rm{C}}}}\right)\left(\frac{1\,{{\rm{mol}}}}{4\;{{\rm{e}}}}\right)\left(\frac{6.023\times {10}^{23}}{1\,{{\rm{mol}}}\;{{{\rm{O}}}}_{2}}\right)$$

(3)

where J is the geometric current.

The number of active Ru sites can be calculated following Eq. (4):

$${{\rm{Number}}}\; {{\rm{of}}}\; {{\rm{active}}}\; {{\rm{Ru}}}\; {{\rm{sites}}}=\frac{{m}_{{{\rm{cat}}}}\;\times\; {{\rm{Ru\; wt}}}\%}{101.07\;{{\rm{g}}}\;{{{\rm{mol}}}}^{-1}}\times 6.023\times {10}^{23}$$

(4)

Where m_cat is the mass of catalysts loaded on the work electrode.

In situ XRD tests

In situ XRD measurements were carried out in a representative three–electrode setup. The electrodes were assembled in a in situ electrolytic cell and then tested on the Bruker D8 ADVANCE X–ray with the potential range from 1.0 to 1.5 V versus reversible hydrogen electrode (vs. RHE). The voltage and current of the tubes used are 45 kV and 200 mA.

In situ ATR–IR tests

A professionally customized three–electrode reaction cell was used for the In situ ATR–IR measurements (Supplementary Fig. 85). ATR–IR spectra were collected using unpolarized IR radiation on a PerkinElmer Frontier FT–IR spectrometer that was equipped with a built–in mercury–cadmium–telluride (MCT) detector and operated at a resolution of 4 cm⁻¹. Following data acquisition, all spectral data were converted to absorbance units. This conversion was based on the formula -log (I/I₀), with I and I₀ denoting the reflected radiation intensity from the sample and reference spectra, respectively.

In situ Raman tests

A Renishaw InVia Raman spectrometer equipped with 50× objective and 532 nm laser was employed for in situ electrochemical Raman measurements. The experiments were conducted in a standard three–electrode setup, comprising a catalyst–loaded L–shaped Au working electrode, a Pt wire counter electrode and an Ag/AgCl reference electrode. The in situ Raman spectra were ultimately collected while applying a constant potential ranging from 1.0 to 1.7 V (vs. RHE) to the working electrode.

PEMWE tests

A Nafion 115 membrane (N115, DuPont, geometric area: 1 × 1 cm²) with a thickness of 127 μm was used to construct the membrane electrode assembly (MEA). F–RuO₂/FC and commercial Pt/C (20 wt%) were used as the anode and cathode, respectively. The N115 membrane was pretreated prior to MEA preparation with sequential 1 h treatments at 80 °C in 5 wt% H₂O₂ solution and 5 wt% dilute H₂SO₄, each interspersed with a 0.5 h deionized water rinse. The catalyst ink was formulated with a solution of isopropyl alcohol and deionized water in a 2:1 volume ratio serving as the dispersion medium, with Nafion ionomer (5 wt%) added (25 wt% relative to the catalyst) before ultrasonication. Then, the prepared ink was sprayed onto the both sides of Nafion 115 membrane to get the catalyst–coated membrane (CCM), achieving noble metal loadings of 2 mg cm^–2 for anode and 0.5 mg cm^–2 for cathode. Subsequently, the CCM was sandwiched between porous Pt–plated Ti foam gas diffusion layer (GDL) and assembled into the PEM cell. Last, 5 N m torque was applied on PEM cell for sealing. All performance evaluations, conducted without iR correction, were carried out in a PEM cell at 80 °C under ambient pressure, where the stability of F–RuO₂/FC was evaluated by chronopotentiometry at 100, 500 and 1000 mA cm^–2.

DFT calculations

DFT calculations were conducted using the Vienna ab initio simulation package (VASP), which employed the projector augmented wave (PAW) method to simulate the electron–ions interactions^57,58. The exchange–correction interactions were described by the generalized gradient approximation (GGA) of the Perdew–Burke–Ernzerhof (GGA–PBE)⁵⁹. The RuO₂(110) facet was selected because it exposes fully coordinated bridged Ru sites and coordination unsaturated sites with quintuple coordination that bind to triple–coordinated O for better identification of the active site^60,61,62. Additionally, this surface is known to exhibit the most favorable thermodynamic free energy during most synthesis processes, as established in prior studies⁶³. The cutoff energy was set as 400 eV during the structural optimization of all slab models. During the structural optimization calculations, the atoms in the bottom two layers were fixed and other layers of atoms were allowed to relax. The vacuum layer above the surface is at least 15 Å, and the calculation module was placed in a 6.2146 × 12.6698 × 21.7973 Å lattice. The k–points (Monkhorst–Pack mesh) during the structural optimization and electronic structure–involved calculations were set to 4 × 2 × 1 and 5 × 5 × 1, respectively.

The Gibbs free energy change during OER was calculated by ΔG = ΔE + ΔZPE – TΔS, where the ΔE is the energy change of intermediates adsorbed on catalysts surface based on the DFT calculations, ΔZPE and ΔS is the change in zero–point energy and entropy, respectively, T is room temperature (298.15 K). The free energy of (H⁺ + e^–) at standard conditions was assumed as the energy of 1/2 H₂. Namely, G(H⁺ + e^–) = 1/2 G(H₂).