Sonochemical boron incorporation enhances activity and durability of ruthenium oxide for acidic water oxidation

Cavitation-mediated sonochemical synthesis of B-RuO₂

Cavitation erosion is defined as material degradation on a solid surface caused by the formation, growth and subsequent rupture of a large number of bubbles in a liquid³³. It has long been recognized as a damaging phenomenon in hydraulic systems (e.g., pump impellers, turbine blades)^34,35. We propose to repurpose this traditionally detrimental mechanism into a synthesis strategy: harnessing cavitation-induced microjetting to (1) selectively erode solid boron precursors and (2) generate reactive boron species for atomic-level dopant incorporation. This paradigm shift from destructive failure to constructive materials engineering presents a distinct synthesis pathway, though the fundamental mechanisms governing cavitation-enabled doping remain underexplored.

The acoustic cavitation effect induced by high-intensity ultrasonic irradiation (24 kHz, 1000 W cm⁻²) exhibits a dual mechanism of action: in the bulk liquid phase, cavitation bubbles collapse in a spherically symmetric manner, generating transient high-temperature (≈5000 °C) and high-pressure (~1000 atm) hotspots; while at the solid-liquid interface, cavitation bubbles exhibit asymmetric collapse characteristics, generating high-speed microjets (>100 m s⁻¹) and shockwaves^36,37. These combined mechanical-thermodynamic effects induce interfacial material fatigue damage. Leveraging this principle, we apply ultrasonic cavitation erosion to an ethylene glycol reduction system containing TiB₂@TiO₂ nanoparticles and Ru³⁺ ions: cavitation-induced microjets preferentially erode TiB₂@TiO₂ nanoparticles to release reactive boron species, while the transient thermal field significantly reduces the diffusion barrier for boron atoms within nascent metallic Ru lattices, enabling efficient solid-state boron doping kinetics. This coupling is difficult to achieve under conventional solvothermally or oil-bath conditions, where the same precursor chemistry instead leads to incomplete precursor conversion, phase separation, and severe Ru aggregation.

Scanning electron microscopy (SEM), transmission electron microscopy (TEM), high-angle annular dark-field-scanning transmission electron microscopy (HAADF-STEM), and energy-dispersive X-ray spectroscopy (EDS) elemental mapping were employed to investigate the synthetic process. As characterized in Supplementary Fig. 1, the pristine TiB₂@TiO₂ precursor exhibits a distinct core-shell architecture, where the TiB₂ core is encapsulated by a native TiO₂ passivation layer, evidenced by surface Ti/O enrichment. During the initial 120 s of ultrasonic irradiation (Supplementary Fig. 2a, b), these particles undergo progressive fragmentation, evolving into numerous irregular flake-like nanostructures. As shown in the elemental mapping (Supplementary Fig. 2c, d), the Ru and B species exhibit significant spatial aggregation, while Ti and O species are distinctively colocalized in separate regions, indicating an initial tendency toward phase separation between the newly formed B-Ru species and the TiO₂ support.

With extended ultrasonic processing (180 and 240 s), this structural reconstruction ultimately results in the formation of 2D microporous meshes comprising ultrafine B-Ru nanowires (diameter = ~2 nm) anchored on a residual TiO₂ layer (Supplementary Figs. 3 and 4a–d). The X-ray diffraction (XRD) analysis confirmed the emergence of metallic Ru phases and the disappearance of TiB₂ peaks, indicating the complete decomposition of the boron precursor (Fig. 2a). Supplementary Fig. 4e confirms the presence of a thin TiO₂ layer underlying the 2D B-Ru mesh architecture, with a lattice interspacing of 0.356 nm at the heterointerface corresponding to the (101) planes of anatase TiO₂. Supplementary Fig. 4f illustrates that the in-situ decomposition of TiB₂ generates residual TiO₂ overlayers that function as heterogeneous nucleation templates, providing sites for Ru nucleation and enabling the subsequent growth of B-Ru nanoparticles (B-Ru/TiO₂). The granular morphology of the B-Ru nanowires implies that the transient high-temperature/pressure conditions (which accelerate atomic diffusion) facilitate the oriented attachment of ultrafine B-Ru nanoparticle precursors into short nanowires, followed by their atomic fusion and interconnection into a 2D mesh-like architecture^38,39.

Elemental mapping of the TiB₂@TiO₂ following 240 s of ultrasonic treatment demonstrates a distinct phase separation: Ru and B are homogeneously co-distributed throughout the matrix, while Ti and O exhibit spatial co-localization in the substrate (Supplementary Fig. 5). This spatial decoupling between (Ru/B) and (Ti/O) phases demonstrates two concurrent processes: (1) preservation of residual TiO₂ domains from the decomposition of TiB₂@TiO₂, and (2) boron migration into the Ru lattice through transient thermal activation. Notably, the comparative experiments demonstrated that the erosion of TiB₂ nanoparticles occurs exclusively in the presence of Ru³⁺ ions (Supplementary Figs. 6 and 7), suggesting a synergistic effect between physical cavitation and chemical corrosion on the successful release of boron and its subsequent doping into the Ru lattice. In summary, ultrasonically induced microjets, combined with Ru³⁺-driven etching, initially crack and corrode the TiB₂@TiO₂ nanoparticles, liberating reactive boron species while preserving the residual TiO₂ scaffold, while subsequent bubble-collapse thermal shocks promote rapid boron diffusion into nascent metallic Ru domains. Consequently, after 240 s of ultrasonic treatment, we obtained the B-Ru/TiO₂ sample.

Subsequent air-annealing and hydrofluoric acid (HF) treatments transformed the B-Ru/TiO₂ meshes into pure rutile B-RuO₂ nanoflakes, while completely removing the TiO₂ layer. For comparison, a RuO₂ control sample (denoted as RuO₂) was prepared following the same procedure but without the addition of TiB₂ as the boron source. X-ray photoelectron spectroscopy (XPS) confirms the complete removal of titanium, showing only the Ru 3p peak at 462.6 eV with no detectable Ti 2p signal (Supplementary Fig. 8). XRD analysis confirms the crystalline structure of B-RuO₂, showing characteristic peaks consistent with the rutile phase of RuO₂ (Fig. 2a). STEM reveals a lattice spacing of approximately 0.319 nm for the (110) plane in B-RuO₂, indicating boron incorporation without significant structural distortion (Fig. 2b). The HAADF-STEM, TEM, and atomic force microscopy (AFM) images (Fig. 2c, Supplementary Figs. 9 and 10) reveal that the B-RuO₂ nanoflakes are composed of numerous interconnected nanosheets with a thickness of approximately 1.2 nm. In contrast, although the RuO₂ control also crystallizes in the rutile phase, TEM and elemental mapping show that it consists of severely aggregated particles rather than the interconnected ultrathin nanoflake network observed for B-RuO₂ (Supplementary Fig. 11). This comparison indicates that TiB₂ serves not only as the boron source, but also plays an essential role in regulating the transient nucleation and structural evolution of the Ru-based phase under cavitation conditions. Brunauer–Emmett–Teller specific surface area measurement (Supplementary Fig. 12 and Supplementary Table 1) confirms that the cavitation-induced decomposition generates hierarchically porous B-RuO₂ nanoflakes with a higher specific surface area (27.0 m² g⁻¹) than that of RuO₂ (11.9 m² g⁻¹), thereby increasing the exposure of active sites and enhancing catalytic efficiency. The synthesis of the optimized B-RuO₂ catalysts was carried out under conditions involving the addition of 35 mg of TiB₂ followed by annealing at a temperature of 350 °C for a duration of 3 h, resulting in the final B-RuO₂ catalysts with the highest OER activity and stability (Supplementary Fig. 13).

To directly verify the role of the sonochemical route, we carried out control experiments using the same precursor system (RuCl₃·xH₂O + TiB₂ in ethylene glycol), but replacing ultrasound with solvothermal treatment or oil-bath heating at 160 °C⁴⁰. Two representative durations were examined: 4 min, which matches the actual sonochemical reaction time, and 3 h, which is sufficiently long for substantial TiB₂ corrosion under conventional thermal conditions according to previous reports⁴⁰. At 4 min, neither solvothermal nor oil-bath treatment induced effective Ru reduction or TiB₂ corrosion; the precursor retained its original morphology, and only TiB₂ diffraction was observed (Supplementary Figs. 14 and 15). Even after 3 h, both conventional routes still failed to produce a homogeneous boron-doped Ru-based product. Instead, metallic Ru was clearly detected by XRD (Supplementary Fig. 16a, d), while TEM and elemental mapping revealed inhomogeneous Ru-containing structures, including TiO₂-supported shell-like Ru domains and large agglomerated Ru particles (Supplementary Fig. 16b, c, e, f). After the same subsequent annealing and HF treatment steps, residual metallic Ru diffraction peaks were still present (Supplementary Fig. 17a, d), indicating incomplete oxidation of the Ru phase. Correspondingly, elemental mapping showed that oxygen was mainly distributed at the particle surface (Supplementary Fig. 17b, c, e, f), consistent with insufficient oxidation of the large Ru aggregates. Furthermore, boron species are not observed in the B 1s XPS spectra of the RuO₂ derived from either solvothermal or oil bath treatments (Supplementary Fig. 18), indicating that these conventional routes failed to achieve boron incorporation into rutile RuO₂. As a result, the final catalysts exhibited much poorer acidic OER activity than the sonochemically synthesized B-RuO₂ (Supplementary Fig. 19). These results demonstrate that the cavitation-mediated microenvironment is not merely a faster heating process, but a qualitatively different reaction field that synchronizes TiB₂ corrosion, Ru³⁺ reduction, nucleation, and boron incorporation within a few minutes.

To further evaluate the practical potential of this sonochemical strategy, we carried out preliminary scale-up experiments by increasing the reaction volume from 7 mL (i.e., B-RuO₂) to 20, 50, and 100 mL, with the resulting products denoted as B-RuO₂–20 mL, B-RuO₂−50 mL, and B-RuO₂−100 mL, respectively, while correspondingly adjusting the sonication time (Supplementary Table 2). The scaled-up products retained highly consistent phase structure, morphology, and elemental distribution (Supplementary Fig. 20), indicating that the cavitation-mediated formation process is reproducible beyond the small-batch condition. Importantly, the 100 mL sample still delivered high activity together with similar stability behavior (Supplementary Fig. 21), demonstrating that the synthetic advantages of this route can be maintained upon scale-up. Quantitative analysis further showed that the Ru reduction efficiency remained above 99.70%, and that only a very small amount of Ru was lost during the subsequent HF treatment step (Supplementary Table 3). The practical mass yield increased markedly with increasing batch size (Supplementary Table 2), indicating that the lower yield at small scale mainly originates from handling losses during washing, centrifugation, and collection rather than from incomplete Ru conversion or substantial Ru dissolution. These results highlight that the present sonochemical route is not only rapid and structurally effective, but also chemically efficient and practically scalable.

The composition of B-RuO₂, particularly the boron content and distribution, was rigorously characterized by electron energy loss spectroscopy (EELS) (Fig. 2d), EDS elemental mapping (Fig. 2e), and electron probe X-ray micro-analyzer coupled with wavelength-dispersive X-ray spectroscopy (EPMA-WDS) (Fig. 2f) and Ar-ion sputtering XPS depth profiling (Supplementary Figs. 22 and 23). Because B-RuO₂ consists of ultrathin nanoflakes with a thickness of only ~1.2 nm (Supplementary Fig. 10), bulk doping in this system refers to boron distributed through the full nanoflake thickness, and XPS can probe a substantial fraction of the nanoflake interior. EDS mapping (Fig. 2e) shows clear co-localization of Ru, O, and B across the nanoflakes, while the B 1s signal at 191.2 eV remains clearly observable after sputtering to depths of 4 and 8 nm, confirming that boron is not restricted to the outermost surface (Supplementary Fig. 22). XPS depth-profile in the Ti 2p/Ru 3p region further shows only the Ru 3p_3/2 signal, without any detectable Ti-related contribution after HF treatment, even after sputtering to deeper regions (Supplementary Fig. 23), thereby ruling out residual or substitutional Ti in the final catalyst. EPMA-WDS, prioritized for its high sensitivity to light elements, confirmed a Ru/B atomic ratio of 91.6:8.4 (Supplementary Table 4), effectively circumventing the limitations of EDS (severe B Kα/Bremsstrahlung overlap) and XPS (poor B 1s signal-to-noise ratio), ensuring reliable quantification even at low boron concentrations. To note, the spatial intensity variation in the EPMA-WDS maps mainly reflects micrometer-scale differences in aggregate stacking rather than boron segregation (Fig. 2f). Collectively, these structural and compositional analyses demonstrate that the cavitation-mediated sonochemical strategy successfully overcomes thermodynamic barriers to achieve uniform, bulk doping of boron into the RuO₂ lattice.

Substitutional boron doping modulates electronic structures

To further identify the lattice occupancy of boron, we combined Ru K-edge wavelet-transformed extended X-ray absorption fine structure (WT-EXAFS) and solid-state ¹¹B magic-angle-spinning nuclear magnetic resonance (¹¹B MAS NMR). The Ru K-edge WT-EXAFS result provides site-specific evidence for substitutional incorporation (Fig. 3a, b). In B-RuO₂, the first-shell Ru-O scattering feature remains nearly identical to that of RuO₂ in both R space and k space, indicating that the basic RuO₆ framework of rutile RuO₂ is largely preserved. In contrast, the second-shell Ru-Ru scattering feature is distinctly modified, particularly around k ≈ 10 Å⁻¹ and R + ΔR ≈ 3 Å. This selective perturbation of the cation–cation correlation is consistent with partial replacement of Ru-O-Ru motifs by Ru-O-B motifs, because boron is a much lighter scatterer than ruthenium and therefore changes the second-shell backscattering behavior much more strongly than the first-shell Ru-O feature. If boron were incorporated primarily at interstitial sites, the Ru sublattice would remain largely intact, and such a distinct modification of the Ru-Ru second-shell correlation would be much less likely. Therefore, the WT-EXAFS result strongly favors substitution of boron at Ru lattice sites rather than simple interstitial incorporation.

The solid-state ¹¹B MAS NMR spectrum provides an independent boron-centered view of the local environment. B-RuO₂ exhibits a single very broad and clearly asymmetric resonance centered at 8.6 ppm, with a large full width at half maximum (FWHM) of approximately 82.9 ppm (Fig. 3c). By contrast, the Na₂B₄O₇ reference shows a sharp dominant resonance near 0 ppm together with two much weaker peaks at around 12 and 16 ppm, corresponding to well-defined borate environments (Supplementary Fig. 24). This comparison first excludes the possibility that the boron signal in B-RuO₂ originates from a crystalline borate-like impurity phase, because such phases would give narrow and well-resolved resonances rather than the broad envelope observed here. More importantly, the unusually large FWHM and pronounced asymmetry indicate that boron does not occupy a simple and well-defined local site, but instead resides in a strongly distorted and highly distributed set of B-O-Ru environments within the RuO₂ lattice⁴¹. This behavior is expected for aliovalent substitution of the much smaller B species onto Ru sites, which necessarily introduces strong local structural relaxation, local symmetry breaking, and a broad distribution of electric-field gradients. By contrast, interstitial boron would be expected to reside in a comparatively more regular local cavity environment and would generally produce a much narrower and more symmetric ¹¹B resonance^42,43. Together with the XPS depth profile results described above, these observations support that boron is incorporated predominantly through homogeneous substitution at Ru lattice sites rather than through interstitial doping.

To elucidate the influence of substitutional boron incorporation on the electronic configuration and local coordination environment of the catalyst, we conducted systematic investigations using XPS and X-ray absorption spectroscopy (XAS). In the Ru 3d and Ru 3p XPS spectra (Fig. 3d and Supplementary Fig. 25), B-RuO₂ exhibits binding energy peaks at 280.9 eV (Ru 3d_5/2) and 462.6 eV (Ru 3p_3/2), respectively. In comparison, RuO₂ shows the binding energy peaks at 281.0 eV (Ru 3d_5/2) and 462.9 eV (Ru 3p_3/2). These shifts suggest a slightly reduced average Ru oxidation state in B-RuO₂. This is further corroborated by Ru K-edge X-ray absorption near edge structure (XANES) spectra (Fig. 3e and Supplementary Fig. 26), where the energy of the absorption edge of B-RuO₂ is shifted by ~0.2 eV to lower energy than that of RuO₂. Using Ru foil, RuCl₃, and RuO₂ as references, the average Ru oxidation state of B-RuO₂ is semi-quantitatively estimated to be 3.91 (Fig. 3f), consistent with partial electron redistribution induced by boron incorporation²¹.

Fourier-transformed EXAFS (FT-EXAFS) analysis (Fig. 3g) at the Ru K-edge reveals two key differences in the B-RuO₂ compared to RuO₂. Firstly, the dominant peak at 1.48 Å, corresponding to the first-shell Ru-O coordination, shows a markedly attenuated intensity compared to that of RuO₂ (1.46 Å), suggesting a lower Ru-O coordination number^44,45. Secondly, this peak of the B-RuO₂ shows a radial distance shift of +0.02 Å, suggesting an elongation of the Ru-O bond length in B-RuO₂. This bond elongation is further supported by Raman spectroscopy. The E_g and A_1g peaks of Ru-O bond vibrations in B-RuO₂ exhibit red-shifts of 8.4 and 23.2 cm⁻¹, respectively, relative to the RuO₂ peaks at 509.5 cm⁻¹ (E_g) and 628.0 cm⁻¹ (A_1g) (Fig. 3h)^46,47. These results indicate that substitutional boron incorporation induces severe local structural distortion that affects the Ru-O bonding.

O K-edge soft X-ray absorption spectroscopy (SXAS) provides further insight into the bonding nature (Fig. 3i)⁴⁸. Direct comparison of the baseline-corrected spectra shows that both the t_2g and e_g pre-edge features decrease in B-RuO₂ relative to RuO₂, with a more pronounced suppression of the t_2g component⁴⁹. This result indicates that the O K-edge change cannot be interpreted simply by the t_2g/e_g, intensity ratio alone. Instead, it reflects a combined effect of reduced overall Ru 4d-O 2p hybridization associated with the elongated Ru-O bond, increased occupation of the lower-energy t_2g-derived states due to partial Ru reduction, and local symmetry lowering induced by substitutional boron incorporation^50,51. Thus, substitutional boron doping modulates both the local geometry and the electronic structure of Ru sites in rutile RuO₂, leading to a modified Ru-O bonding environment with reduced overall hybridization. We note that the slightly enhanced intensity above ~535 eV may include contributions from the overall oxygen-related absorption background and B-O related local coordination environments; therefore, this high-energy feature is not used as the primary basis for assigning reduced Ru 4d-O 2p hybridization. Mechanistically, the boron-induced weakening of Ru-O hybridization is expected to suppress the LOM pathway and thereby improve structural stability, while also favoring oxygen-intermediate adsorption along the AEM pathway, thus contributing to the enhanced intrinsic OER kinetics of B-RuO₂^13,49.

B-RuO₂ drives active and durable OER

The OER performance of B-RuO₂ was systematically evaluated in 0.1 M HClO₄ using a standard three-electrode setup. For comparison, RuO₂ and commercial RuO₂ (denoted as Com-RuO₂) were also tested under the same experimental conditions. Ninety percent iR corrected linear sweep voltammetry (LSV) curves (Supplementary Fig. 27a and Fig. 4a) reveal that B-RuO₂ delivers a current density of 10 mA cm⁻² at an overpotential (η) of 180 mV, significantly lower than RuO₂ (253 mV) and Com-RuO₂ (279 mV). This enhancement is further evidenced by the Tafel slope analysis (Supplementary Fig. 27b), where B-RuO₂ exhibits a much smaller slope (35.23 mV dec⁻¹) compared to RuO₂ (55.80 mV dec⁻¹) and Com-RuO₂ (57.50 mV dec⁻¹), indicating remarkably accelerated kinetics. In addition, electrochemical impedance spectroscopy (EIS) reveals a markedly lower charge transfer resistance (R_ct) for B-RuO₂ compared to RuO₂ and Com-RuO₂ (Supplementary Fig. 27c), demonstrating that boron doping effectively facilitates interfacial electron transfer kinetics.

To investigate the source of the intrinsic activity of B-RuO₂ catalysts, double-layer capacitance (C_dl) measurements were first performed to estimate the electrochemical surface area (ECSA) (Supplementary Fig. 28). B-RuO₂ displays the highest C_dl value (8.08 mF, Supplementary Fig. 29a), corresponding to an ECSA 5.1-fold larger than that of RuO₂. This increased surface area is attributed to the hierarchical porosity of the B-RuO₂ nanoflakes, which maximizes active site exposure. Critically, however, ECSA-normalized LSV curves (Supplementary Fig. 29b) reveal that B-RuO₂ still delivers the highest activity among the three samples, underscoring its enhanced intrinsic activity. To further quantify the density of electrochemically accessible redox-active surface sites, pulse voltammetry-based surface charge storage measurements were performed (Supplementary Fig. 30a–c)^52,53. B-RuO₂ exhibits a fitted capacitance of 55.7 mF, much higher than that of RuO₂ (8.4 mF) (Supplementary Fig. 30d). This 5.6-fold increase is in agreement with the 5.1-fold enhancement obtained from the C_dl-derived ECSA, confirming that B-RuO₂ exposes a much larger population of electrochemically accessible active sites. The turnover frequency (TOF) and mass activity (MA) were further calculated to quantify the active-site efficiency. At η = 200 mV, B-RuO₂ achieves a TOF of 149.02 h⁻¹ (vs. 6.00 h⁻¹ for RuO₂) and an MA of 142.39 A g_Ru⁻¹ (vs. 6.34 A  g_Ru⁻¹ for RuO₂) (Supplementary Fig. 31). These order-of-magnitude improvements highlight the critical role of boron in activating Ru sites. As summarized in the radar plot of six key performance metrics (Fig. 4b and Supplementary Table 5), B-RuO₂ demonstrates competitive performance.

To clarify whether the HF treatment affects the catalytic performance through fluorine incorporation, we carried out several control experiments (Supplementary Fig. 32). The intermediate B-RuO₂/TiO₂ sample before HF etching shows clearly inferior OER activity to the final B-RuO₂ (Supplementary Fig. 33a), confirming that HF treatment is necessary to fully expose the catalyst. However, B-RuO₂ samples prepared under different HF concentrations and etching times exhibit nearly identical activity (Supplementary Fig. 33b, c), and no detectable F 1s XPS signal is observed after etching (Supplementary Fig. 33d). Moreover, the same HF treatment does not improve the activity of the RuO₂ control sample (Supplementary Fig. 34). Therefore, the beneficial effect of HF treatment originates from removal of the inactive TiO₂ template rather than fluorine doping. Taken together, we conclude that the high activity arises not merely from increased active site density, but also from the boron-induced modification of the Ru-O bonding environment, evidenced by XPS, XANES, Raman, and O K-edge SXAS.

The stability of the B-RuO₂ catalyst in an acid environment is a crucial parameter for its practical applications. Galvanostatic chronopotentiometry (CP) at 10 mA cm⁻² (Fig. 4c) demonstrates long-term stability for B-RuO₂, with a negligible potential increase rate of 27.3 μV h⁻¹ over 3000 h. In contrast, both RuO₂ and Com-RuO₂ suffer rapid failure within dozens of hours. To quantitatively benchmark this stability, we calculated the S-number metric; the S-number for B-RuO₂ reaches 1.53 × 10⁷, which is on the same order of magnitude as that of crystalline IrO₂ (Supplementary Fig. 35)⁵⁴. A comprehensive comparison with recently reported doped RuO₂ OER catalysts (Supplementary Fig. 36a and Supplementary Table 6) positions B-RuO₂ as a competitive catalyst, addressing the activity-stability trade-off by achieving a low overpotential (η = 180 mV@10 mA cm⁻²) and long-term stability (>3000 h@10 mA cm⁻²). Notably, compared with previously reported B-doped RuO₂ prepared by conventional routes^55,56,57,58, the present sonochemical strategy combines competitive activity, longer durability, and a rapid 4 min synthesis step. This advantage is directly supported by our solvothermal- and oil-bath-derived control samples, which show much poorer OER activity than sonochemically synthesized B-RuO₂ (Supplementary Fig. 19).

To assess the practical applicability, the B-RuO₂ catalysts were evaluated in a PEM water electrolyzer at 60 °C using 0.1 M HClO₄ as the electrolyte (Supplementary Figs. 37 and 38a). The I–V polarization curve (Supplementary Fig. 38b) shows that the PEMWE achieves 1 A cm⁻² at 1.714 V, showing competitive performance relative to commercial benchmarks. Corroborating this high activity, the measured oxygen Faradaic efficiency remains near 100% across various current densities (Supplementary Fig. 39), confirming that the recorded current derives exclusively from water oxidation rather than catalyst corrosion. Furthermore, the E–t curve at 200 mA cm⁻² demonstrates high stability for B-RuO₂ with over 2000 h of continuous operation with a negligible voltage degradation rate of 20 μV h⁻¹ (Fig. 4d). This differs from the Com-RuO₂ benchmark, which fails rapidly within less than 20 h under identical conditions. More significantly, even under the industrial-grade current density of 1 A cm⁻², B-RuO₂ maintains robust stability for over 200 h, exhibiting a low decay rate of only 65.0 μV h⁻¹. This stability in a practical PEM electrolyzer is competitive among reported doped Ru-based systems, supporting the substantial stabilization effect of B doping (Supplementary Fig. 36b and Supplementary Table 7). Beyond performance metrics, the economic feasibility of the B-RuO₂-based PEMWE was assessed. The calculated energy consumption at 1 A cm⁻² is only 45.97 kW h kg⁻¹ H₂, which is lower than that of commercially reported PEMWEs (50.00–55.56 kW h kg⁻¹ H₂). The estimated hydrogen production cost of this PEMWE is $0.92 per kilogram of H₂, which is below the U.S. Department of Energy’s target of $1.00 per kilogram of H₂ by 2030. This further confirms the practical application potential of B-RuO₂ catalysts. Collectively, the combination of long-term durability, high energy efficiency, and low hydrogen production cost supports the practical application potential of B-RuO₂ catalysts for next-generation green hydrogen technologies.

Robust B-O bonding stabilizes Ru active sites

To elucidate the stabilizing mechanism of boron dopants, we first investigate the chemical state and structure of boron in the pristine B-RuO₂ using XPS, boron K-edge SXAS, and Raman spectroscopy. Prior to OER, the Raman spectrum of B-RuO₂ reveals a broad peak at 789.4 cm⁻¹ corresponding to B-O bond vibrations (Fig. 3h)⁵⁹. The B 1s XPS depth profiling corroborates this finding, showing a characteristic B-O bonding peak (B³⁺) at 191.2 eV (Supplementary Figs. 22 and 40a)⁶⁰. Furthermore, the B K-edge SXAS (Fig. 5a) spectrum shows a dominant feature at 193.9 eV, assigned to trigonal BO₃-like lattice B-O coordination, together with a broad high-energy envelope above ~196 eV that is consistent with mixed contributions from distorted BO₄-like and BO₃-like local coordination^61,62. The weaker low-energy feature near 191.8 eV is attributed to surface-segregated boron species (B-B dimers)^61,62. Together with the substitutional-doping evidence established above, this result indicates that boron is predominantly incorporated into a distorted lattice environment in pristine B-RuO₂. The pronounced size mismatch between B and Ru, together with the slightly larger B-O electronegativity difference relative to Ru-O, is expected to induce strong local relaxation and modify the B-O bonding configuration around the substitutional boron centers. Theoretically, these surface-segregated boron species possess high surface energy and undercoordinated configurations, rendering them thermodynamically metastable and susceptible to rapid oxidative leaching. In stark contrast, the lattice incorporated boron dopants are rigidly confined within the rutile lattice, where the strong bonding strength and covalency of the B-O bonds effectively anchor the lattice oxygen, creating a high kinetic barrier against structural collapse.

During the OER, time-resolved inductively coupled plasma mass spectrometry (ICP-MS) was employed to probe the dynamic dissolution behavior of boron. The dissolution profile reveals a rapid but limited surface-reconstruction process: an initial transient stage within the first 20 s (corresponding to a rapid boron dissolution rate of ~298.35 μg h⁻¹), followed by a passivated stage where boron dissolution becomes negligible (Fig. 5b and Supplementary Table 8). By contrast, Ru exhibits consistent, low-level quasi-linear dissolution kinetics (0.02 μg h⁻¹) throughout the entire operation. This behavior indicates that the dissolved boron mainly originates from a small population of labile surface-exposed species rather than from continuous loss of lattice-incorporated boron. To explicitly verify the depletion of surface boron, we employed NH₃ chemisorption Fourier-transformed infrared spectroscopy (FTIR), exploiting the selective Lewis acid-base interaction between NH₃ and boron sites^63,64. Control experiments on RuO₂ confirm negligible NH₃ adsorption at Ru/O sites (Supplementary Fig. 41a). For the NH₃ purged B-RuO₂ sample (denoted as B-RuO₂-NH₃ purge), pre-OER characterization reveals distinct spectral fingerprints at 1082.8 cm⁻¹ (B-N stretching) and 3133.7 cm⁻¹ (N-H stretching) (Fig. 5c). These spectral features match those observed in the B₂O₃ reference spectra (Supplementary Fig. 41b), confirming the presence of accessible boron sites on the catalyst surface. In contrast, post-OER analysis of the NH₃ purged B-RuO₂ (denoted as B-RuO₂-after OER-NH₃ purge) demonstrates a complete disappearance of these boron-associated vibrations. This supports that the outermost surface-exposed boron species are rapidly removed during the initial oxidative polarization, driving a surface reconstruction that generates a stable, boron-depleted outer layer which protects the underlying lattice from further corrosion.

Post-OER characterizations of the B-RuO₂ after 24-h chronopotentiometry at 10 mA cm⁻² (denoted as B-RuO₂-after OER) reveal that most lattice B-O bonding (from subsurface to bulk) is retained after this initial reconstruction. First, EDS mapping confirms that boron remains distributed throughout the bulk phase of B-RuO₂ after OER (Supplementary Fig. 42a). As expected, quantitative analysis by EPMA-WDS shows only a marginal decrease in the Ru/B atomic ratio from 91.6: 8.4 to 92.1: 7.9 after the OER (Supplementary Fig. 42b and Supplementary Table 4). This minor change aligns with the rapid but limited surface-reconstruction process, confirming that only the outermost surface-exposed boron is dissolved during the OER process. Crucially, spectroscopic analyses (B 1s XPS, Raman, and B K-edge SXAS) demonstrate that the strong covalent B-O bonds in the subsurface remain preserved. Specifically, B 1s XPS and Raman spectra of B-RuO₂-after OER maintain the characteristic B-O bonds (Supplementary Figs. 40b and 43)⁴⁷, confirming that subsurface B³⁺ substitution is retained. In the B K-edge SXAS spectra (Fig. 5a), the low-energy feature at 191.8 eV associated with the surface-segregated boron species is diminished after OER, whereas the dominant BO₃-like lattice B-O feature at 193.9 eV remains clearly observable. A direct comparison of the baseline-corrected spectra shows that the absolute intensity of this B-O-related feature decreases slightly after OER, indicating that a minor loss of subsurface lattice B-O species cannot be excluded. Nevertheless, the comparison still shows that a substantial majority of the subsurface lattice B-O species remain after OER. Specifically, the peak-to-edge-step ratio increases from 2.7 in B-RuO₂ to 4.6 in B-RuO₂-after OER, consistent with preferential removal of the surface B-B-related component while retaining a substantial majority of the subsurface lattice B-O species. These results indicate that the catalyst reaches a reconstructed steady-state surface supported by a boron-stabilized subsurface framework, which explains why the OER performance does not decay to the level of undoped RuO₂. This interpretation is further supported by the nearly overlapping first and tenth LSV curves (Supplementary Fig. 44a, b) and the minimal decrease in C_dl after OER, from 8.08 to 7.94 mF (Supplementary Fig. 44c, d), indicating that the catalyst rapidly reaches a stable reconstructed state without detectable performance decay. These results demonstrate that while the surface reconstructs, the lattice B-O bonding maintains structural integrity. This creates a functional “core-shell-like” architecture (Fig. 5d): a catalytically active, reconstructed surface stabilized by a robust subsurface framework anchored by strong B-O bonds.

Having identified this boron-stabilized subsurface configuration, we next employed a suite of characterization techniques to probe its operational stability. Combined XRD and TEM analyses confirmed the preservation of both the rutile crystal structure and nanosheet morphology (Supplementary Fig. 45). Ru K-edge FT-EXAFS analysis (Fig. 5e) demonstrated the stable behavior of the first-shell Ru-O coordination, as its corresponding peak remained unchanged at 1.49 Å after 24-h chronoamperometry tests across a range of potentials (1.3–1.7 V vs. Ag/AgCl, denoted as B-RuO₂-1.3/1.5/1.7 V-24 h) (Supplementary Fig. 46). This stability also extended to the second-shell Ru-Ru coordination, as validated by WT-EXAFS analysis (Supplementary Fig. 47).

With the structural integrity confirmed, we then assessed the electronic structure stability of the Ru active sites using XANES and XPS. Ru K-edge XANES spectra (Fig. 5f and Supplementary Fig. 48) show no shift in either the absorption edge energy or the first-derivative peaks throughout the testing period, demonstrating the stability of the Ru oxidation state under elevated potentials. Similarly, O 1s XPS spectra (Supplementary Fig. 49a) show that the lattice oxygen was well-preserved. This was further corroborated by O K-edge SXAS spectra (Supplementary Fig. 49b), which show nearly unattenuated t_2g and e_g peaks, confirming the preservation of Ru-O covalency. Collectively, these findings uncover the fundamental origin of the enhanced durability: the subsurface boron dopants function as robust “lattice anchors.” By forming robust covalent B-O bonds, they effectively suppress the LOM pathway, the primary trigger for catalyst dissolution, thereby preserving the integrity of the Ru active sites even under harsh oxidative conditions.

Investigation of suppressed lattice oxygen participation and enhanced intrinsic activity

To gain complementary insight into the suppressed lattice oxygen participation and enhanced intrinsic activity, we combined pH-dependent OER measurements, density functional theory (DFT) calculations, ¹⁸O isotope-labeled operando differential electrochemical mass spectroscopy (DEMS) and in situ electrochemical FTIR. We first analyzed the pH dependence of the OER kinetics by plotting log(j) at 1.42 V vs. RHE as a function of pH (Fig. 6a, b, and Supplementary Fig. 50). B-RuO₂ shows a much weaker pH dependence, with a fitted slope of only −0.176, whereas RuO₂ shows a significantly larger slope of −0.393. This weaker proton-activity dependence is consistent with suppressed lattice oxygen participation and a more AEM-dominant reaction pathway on B-RuO₂^49,65.

To gain theoretical insight, we performed DFT calculations on the B-RuO₂(110) surface. Surface Pourbaix diagram analysis suggests that the surface covered by 1 monolayer (ML) oxygen under OER-relevant potentials is thermodynamically favored (Supplementary Fig. 51a, b), consistent with previous studies on RuO₂-based materials and provides a reasonable reference state for further analysis^11,66,67. We further compared different configurations with varying boron positions in the sub-layer and used the energetically favorable structure illustrated in Supplementary Fig. 51c for electronic structure analysis. Projected density of states (PDOS) analysis reveals a profound modulation of the lattice oxygen electronic structure. As exemplified in Fig. 6c, the oxygen 2p band center (Ɛ_p) was downshifted away from the Fermi level (E_f), from −2.4 to −2.57 eV after boron doping. This downshift is consistent with reduced Ru 4d-O 2p hybridization, which may suppress the activation of the LOM pathway¹³ and thereby improve structural stability.

To experimentally examine this model-derived inference, we directly probed lattice oxygen participation using ¹⁸O isotope-labeled operando DEMS (Supplementary Fig. 52). ¹⁸O isotope labeling was performed on RuO₂ and B-RuO₂ catalysts via LSV in 0.1 M HClO₄ + H₂¹⁸O electrolyte (Supplementary Fig. 53). Mass spectrometry analysis of oxygen evolution products (Fig. 6d, e) reveals distinct m/z signals at 32, 34, and 36 for both ¹⁸O-labeled catalysts. B-RuO₂ significantly suppresses the formation of ³⁴O₂ and ³⁶O₂ products compared to RuO₂. Quantitative analysis of oxygen isotope ratios reveals that B-RuO₂ exhibits reduced lattice oxygen participation in the OER compared to RuO₂. Specifically, the ³⁴O₂/³²O₂ mass signal ratio for B-RuO₂ is 0.45%, compared to 0.71% for RuO₂, and the ³⁶O₂/³²O₂ mass signal ratio is 0.005% for B-RuO₂ versus 0.043 % for RuO₂ (Fig. 6f), indicating a lower contribution of lattice oxygen to OER in the boron-doped material. Collectively, these experimental results support that subsurface strong B-O bonding effectively anchors the surface lattice oxygen, thereby suppressing the LOM pathway and contributing to the high structural stability of B-RuO₂¹³.

We further employed in situ electrochemical FTIR spectroscopy to probe the adsorption behavior of the OER intermediates (Fig. 6g and Supplementary Fig. 54). The spectra exhibit two distinct peaks at ~1120 and ~1230 cm⁻¹, attributed to OOH*/OO* intermediate adsorption and monocrystalline Si substrate oxidation (Si-O-Si), respectively⁶⁸. Notably, as the applied potential increases, B-RuO₂ displays enhanced OOH*/OO* adsorption, whereas no such signal is observed in RuO₂. These in situ spectroscopic results further support that OER on B-RuO₂ proceeds more predominantly through the adsorbate evolution mechanism (AEM), with boron doping facilitating the formation of oxygenated intermediates.

To provide complementary insight into the enhanced intrinsic activity, we performed advanced microkinetic modeling⁶⁹. As depicted in Fig. 6h, the descriptor of B-RuO₂ lies closer to the volcano apex than that of RuO₂, consistent with the experimentally observed activity enhancement and suggesting faster HOO* formation kinetics, as proposed by Dickens et al.⁶⁹. Furthermore, PDOS analysis on the d_z² orbital^69,70 of surface Ru was conducted for both RuO₂ and B-RuO₂. Figure 6i shows a negligible electron density near the Fermi level for RuO₂. In contrast, B-RuO₂ exhibits a pronounced electron distribution at the same energy level, potentially enhancing its affinity for oxygen-containing intermediates during the OER process.