Unassisted bipolar photoelectrochemical synthesis of H₂O₂

Abstract

Photoelectrochemical bipolar systems enable bias-free H₂O₂ synthesis by integrating light harvesting, water oxidation and oxygen reduction. However, poor selectivity due to competing 4e^- pathways, slow charge transfer, and inferior stability severely limit their practical application. Herein, we present a bias-free photoelectrochemical bipolar system for H₂O₂ production, which integrates a Zn-TiO₂/BVO/WO₃ photoanode for the 2e^- water oxidation reaction and a Ru:CuO/Nb-CBO photocathode for the 2e^- oxygen reduction reaction. The photoanode suppresses *OOH intermediates to favor H₂O₂ formation, achieving a Faradaic efficiency (FE) of 90.2% at 0.6 V_RHE. The photocathode stabilizes *OOH intermediates and delivers an FE of 91.9% under the same conditions while maintaining performance for over 50 h. Practical testing under natural sunlight with 36 cm² electrodes achieves an FE of 163.37% and a yield of 11.44 mmol L⁻¹. In situ generated H₂O₂ achieves over 99% bacterial inactivation within 60 min. This work provides an adaptive platform for solar-driven H₂O₂ synthesis and energy conversion.

PEC H₂O₂ production usually suffers from low selectivity and poor stability. This study shows a Zn-TiO₂/BVO/WO₃∥Ru:CuO/Nb-CBO bipolar system that enables efficient and stable H₂O₂ synthesis under sunlight, achieving a combined Faradaic efficiency of 163.37%.

Introduction

Hydrogen peroxide (H₂O₂) is a versatile and environmentally benign oxidant widely used in chemical synthesis, pulp bleaching, wastewater treatment, and disinfection^1–3. Current industrial-scale H₂O₂ production primarily relies on the energy-intensive anthraquinone process, which involves multiple steps, requires high-purity H₂ and organic solvents, and generates explosive H₂/O₂ mixtures^4–6. To address these issues, electrochemical systems based on the two-electron water oxidation reaction (2e^- WOR) at the anode and two-electron oxygen reduction reaction (2e^- ORR) at the cathode have emerged as a sustainable alternative^7,8. This strategy directly converts H₂O and O₂ into H₂O₂ without hazardous intermediates. However, the thermodynamic equilibrium potential for H₂O₂ generation via the electrochemical 2e^- WOR is above 1.76 V_RHE, while that for the 2e^- ORR is below 0.7 V_RHE^9,10. This significant potential gap inevitably requires external voltage to drive the process, thereby limiting the development of efficient devices.

Bias-free photoelectrochemical (PEC) systems offer a promising, green, and electricity-free route to sustainable synthesis of H₂O₂. Their practical development, however, is constrained by insufficient activity and selectivity at the low potentials required for system integration. For instance, a key challenge at the photoanode is the poor Faradaic efficiency (FE) and low current density, primarily because the key *OOH intermediate tends to undergo further oxidation via the competing 4e^- oxygen evolution reaction (OER)^11,12. Currently, most reported photoanodes exhibit less than 70% Faradaic efficiency at 0.6 V_RHE, with current densities below 2 mA cm⁻². Therefore, simultaneously achieving high activity, selectivity, and stability in bias-free PEC systems for H₂O₂ production remains a critical obstacle to their practical implementation.

Herein, we report a bias-free photoelectrochemical system for efficient H₂O₂ production, integrating a Zn-TiO₂/BiVO₄/WO₃ photoanode for 2e^- WOR and a Ru:CuO/Nb-CBO photocathode for 2e^- ORR. The photoanode employs a type‑II heterojunction to suppress the *OOH intermediate and thus the competing 4e^- OER, achieving 90.2% H₂O₂ FE at 0.6 V_RHE. The photocathode combines heterojunction engineering and Ru cocatalyst modification to enhance *OOH retention and charge transfer, delivering 91.9% FE at 0.6 V_RHE and 50 h of durability. Notably, practical testing under natural sunlight with 36 cm² electrodes achieves an FE of 163.37% and a yield of 11.44 mmol L⁻¹. In situ generated H₂O₂ solution achieves >99% bacterial inactivation within 60 min, demonstrating a practical route to solar‑driven H₂O₂ synthesis for energy conversion.

Results

Synthesis and characterization

Figure 1a illustrates the fabrication process of the photoanodes. First, WO₃ nanosheets were synthesized on fluorine-doped tin oxide substrates via a hydrothermal method, followed by the electrodeposition of BiOI films. Subsequently, a VO(acac)₂ solution was drop-cast onto the BiOI films and sintered to form BiVO₄/WO₃ (denoted BVO/WO₃). A TiO₂ passivation layer was then deposited by atomic layer deposition using tetrakis(dimethylamido)titanium and H₂O as precursors, followed by annealing to yield TiO₂/BVO/WO₃. Finally, the samples were immersed in Zn(NO₃)₂ and sintered in an Ar atmosphere to obtain Zn-TiO₂/BVO/WO₃ for the photoelectrochemical two-electron water oxidation reaction (2e^- WOR). X-ray diffraction (XRD) patterns confirm that all photoelectrodes exhibit diffraction peaks corresponding to monoclinic BiVO₄ (JCPDS No. 83-1699) (Supplementary Fig. 1)¹³. Peaks at 23.1°, 23.6°, 24.4°, and 34.2°, corresponding to the characteristic diffractions of WO₃ (JCPDS No. 43-1035), are observed in BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃. No crystalline phases of TiO₂ are detectable due to the amorphous nature of the TiO₂ layer.

Fig. 1. Synthesis and structural characterization.

a Schematic illustration for the synthesis of Zn-TiO₂/BVO/WO₃. b Top-view SEM image of Zn-TiO₂/BVO/WO₃. c, d HRTEM images of the interface (c) between BVO and TiO₂ and (d) between BVO and WO₃. e, f FFT patterns of (e) BVO and (f) WO₃. g EDX mapping images of Zn-TiO₂/BVO/WO₃. h Bi 4 f XPS spectra of BVO, BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃. i Ti 2p XPS spectra of TiO₂/BVO/WO₃ and Zn-TiO₂/BVO/WO₃. j Band alignments of TiO₂, BVO, and WO₃.

Scanning electron microscopy (SEM) images reveal that all samples exhibit a nanoworm-like morphology (Fig. 1b and Supplementary Fig. 2). Transmission electron microscopy (TEM) images indicate the well-defined interfacial structure of the multilayer heterojunction (Fig. 1c–f and Supplementary Fig. 3). Clear boundaries are observed at both the BiVO₄/WO₃ and TiO₂/BiVO₄ interfaces, confirming the construction of the tri-layer architecture. Lattice fringes with a spacing of 0.269 nm, corresponding to the (022) plane of WO₃, are resolved in the WO₃ nanosheet. Within the BiVO₄ layer, a distinct interplanar distance of 0.308 nm is observed, indexed to the (112) plane of BiVO₄¹⁴. The outermost TiO₂ layer exhibits an amorphous morphology, consistent with its role as a conformal passivation layer. Energy dispersive X-ray spectroscopy (EDX) elemental mappings across the cross-section show the well-defined multilayer architecture (Fig. 1g). The intense and continuous W signal evidences the bottom WO₃ layer, which shows a sharp boundary with the overlying V- and Bi-rich layer. In the intermediate region, the uniform distribution of V and Bi signals confirms the BiVO₄ layer and correlates with its characteristic porous structure, beneficial for light absorption and electrolyte infiltration. A homogeneous Ti signal overlying this region indicates a continuous and conformal TiO₂ layer. Notably, the Zn signal is exclusively confined to the TiO₂ layer, confirming successful doping into the TiO₂ matrix without diffusion into the underlying BiVO₄ or WO₃ layers. The sharp transitions between these elemental profiles demonstrate well-defined interfaces with minimal interdiffusion, validating the successful fabrication of the Zn-TiO₂/BiVO₄/WO₃ heterojunction¹⁵.

X-ray photoelectron spectroscopy (XPS) analysis reveals the electronic structure across the Zn-TiO₂/BiVO₄/WO₃ heterojunction (Fig. 1h, i, Supplementary Fig. 4, and Table 1). The formation of the BiVO₄/WO₃ induces a positive shift of 0.1 eV in the Bi 4 f and V 2p core levels. Deposition of the TiO₂ overlayer further increases the binding energies of Bi and V by an additional 0.1 eV, indicating that TiO₂ acts as a stronger electron acceptor and enhances the interfacial band bending. Zn doping into the TiO₂ lattice is corroborated by a 0.1 eV shift in the Ti 2p peaks. The cumulative positive shift of the BiVO₄ core levels maps a coherent electron-transfer pathway from WO₃ to BiVO₄ and then to the TiO₂ layer¹⁶. This stepwise electron redistribution establishes built-in electric fields, which are essential for efficient spatial charge separation and the consequent enhancement in photoelectrochemical performance. Ultraviolet photoelectron spectroscopy and ultraviolet-visible absorption spectroscopy analyses indicate the formation of a type-II band alignment within the TiO₂/BVO/WO₃ multilayer photoanode (Fig. 1j and Supplementary Fig. 5). This band alignment promotes the directional migration of photogenerated holes from the inner WO₃ layer, through the BVO layer to the outermost TiO₂ layer. Consequently, the efficiencies of hole extraction and the subsequent oxidation reaction at the photoanode surface are significantly enhanced.

PEC WOR Performance

We evaluated the PEC H₂O₂ production performance of BVO, BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃ photoanodes in 1 M NaHCO₃ (pH = 10) under AM 1.5 G illumination (100 mW cm⁻²). As shown in Fig. 2a, Zn-TiO₂/BVO/WO₃ demonstrates superior activity, attaining a photocurrent density of 2.01 mA cm⁻² at 0.6 V_RHE with a low onset potential of 0.31 V_RHE. In comparison, TiO₂/BVO/WO₃, BVO/WO₃, and BVO show lower current densities of 1.27, 0.48, and 0.29 mA cm⁻² at 0.6 V_RHE, respectively, along with positively shifted onset potentials (Supplementary Fig. 6a). The dark currents in the LSV curves follow a similar trend (Supplementary Fig. 6b). The integrated incident monochromatic photon-to-current conversion efficiency (IPCE) with respect to the AM 1.5 G spectrum yields solar photocurrent densities (J_SC) of 2.01 mA cm⁻² for Zn-TiO₂/BVO/WO₃, 1.25 mA cm⁻² for TiO₂/BVO/WO₃, 0.48 mA cm⁻² for BVO/WO₃, and 0.29 mA cm⁻² for BVO at 0.6 V_RHE (Supplementary Fig. 7a), in agreement with the experimentally measured values (Fig. 2a). Steady-state and time-resolved photoluminescence measurements reveal that Zn doping reduces PL emission and prolongs the carrier lifetime (Supplementary Fig. 7b, c), indicating suppressed charge recombination. H₂O₂ production was quantified using a Ce⁴⁺ colorimetric method (Supplementary Fig. 8)¹⁷. Zn-TiO₂/BVO/WO₃ maintains high Faradaic efficiencies (FEs) over a wide potential range from 0.6 to 1.6 V_RHE (Fig. 2b), notably displaying an outstanding FE of 90.2% at 0.6 V_RHE, which is approximately 1.2, 2.2, and 6.3 times higher than those of TiO₂/BVO/WO₃, BVO/WO₃, and BVO, respectively. In addition, the FE at 0.6 V_RHE is competitive with those of the reported BiVO₄-based photoanodes (Supplementary Fig. 9 and Supplementary Table 2). In Fig. 2c, Zn-TiO₂/BVO/WO₃ achieves the highest H₂O₂ yield of 1.71 μmol cm⁻²min⁻¹ at 1.6 V_RHE, surpassing the yields of TiO₂/BVO/WO₃ (1.23 μmol cm⁻²min⁻¹), BVO/WO₃ (0.57 μmol cm⁻²min⁻¹), and BVO (0.21 μmol cm⁻² min⁻¹). At 0.6 V_RHE, Zn-TiO₂/BVO/WO₃ releases the lowest amount of O₂ among all photoanodes as detected by gas chromatography (Fig. 2d), indicating the highest selectivity for H₂O₂ production.

Fig. 2. PEC H₂O₂ production performance of photoanodes.

a LSV curves of BVO, BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃ under AM 1.5 G illumination. b FEs of H₂O₂ at various applied potentials. c Yield rates of H₂O₂ at various applied potentials. d H₂O₂ selectivities of H₂O₂ and O₂ at 0.6 V_RHE. e Charge transfer (k_trans) and charge recombination (k_rec) rate constants from IMPS spectra. f Relative electrochemical surface areas for the linear relationship. g I-t curves at 0.6 V_RHE. h Cyclic H₂O₂ productions and FEs on Zn-TiO₂/BVO/WO₃ over 10 h cycles. All electrolytes were 1 M NaHCO₃ (pH = 10, adjusted by NaOH).

The transient photoresponse shows that Zn-TiO₂/BVO/WO₃ delivers the highest steady-state to initial current ratio (i_steady/i_initial) among all the photoanodes (Supplementary Fig. 10). To probe the charge carrier dynamics, intensity-modulated photocurrent spectroscopy (IMPS) measurements were performed¹⁸. The charge transport (k_trans) and recombination (k_rec) rate constants were deduced from the IMPS profiles (Supplementary Fig. 11). Zn-TiO₂/BVO/WO₃ exhibits the highest k_trans and the lowest k_rec among all samples (Fig. 2e). These kinetic parameters demonstrate accelerated charge transfer and suppressed charge recombination, consistent with the enhanced PEC performance. In addition, the charge collection efficiency, represented by k_trans/(k_trans+ k_rec), reveals that the synergistic effect of Zn doping and the heterojunctions enhances charge carrier transport. To better elucidate the mechanism of PEC H₂O₂ synthesis, electrochemical surface area (ECSA) measurements were conducted (Fig. 2f and Supplementary Fig. 12)¹⁹. Zn-TiO₂/BVO/WO₃ exhibits a specific capacitance of 19.75 μF cm⁻², twice as high as that of BVO (9.6 μF cm⁻²), confirming a great number of electrochemically active sites for rapid interfacial charge transfer. Consistently, Zn-TiO₂/BVO/WO₃ shows the lowest charge-transfer resistance in electrochemical impedance spectroscopy (EIS) measurements (Supplementary Fig. 13 and Table 3).

Long-term operation stability is a critical factor for assessing the practical viability of photoelectrodes. As shown in Fig. 2g, both Zn-TiO₂/BVO/WO₃ and TiO₂/BVO/WO₃ maintain stable photocurrent densities for over 10 h. In contrast, BVO exhibits significant performance degradation after 2 h. The post-stability characterization tests confirm that the Zn-TiO₂/BVO/WO₃ photoanode retains its pristine morphological and crystallographic characteristics (Supplementary Fig. 14a, b), underscoring its excellent chemical and mechanical durability against electrolyte corrosion and catalyst detachment. ICP-MS and XPS analyses confirm partial leaching of Zn and Ti from the photoanode after 10 h, while other elements (V, Bi, W) remain below 0.05 mg/L (Supplementary Fig. 14c–e and Table 4). Furthermore, long-term cycling stability was evaluated over 10 consecutive 1 h production cycles with electrolyte replacement after each cycle. The Zn-TiO₂/BVO/WO₃ photoelectrode exhibits stable performance, sustaining a photocurrent while maintaining an FE of over 90% and an average steady-state H₂O₂ production rate of 31.68 μmol cm⁻² h⁻¹ (Fig. 2h).

PEC ORR Performance

Copper bismuth oxide (CuBi₂O₄)-based photocathodes are promising due to their optimal bandgap (~ 1.6 eV) and favorable band alignment for the PEC 2e^- oxygen reduction reaction (ORR)^20–22. We investigate the production of H₂O₂ via the PEC 2e^- ORR pathway using a heterojunction photocathode composed of CuO and Nb-doped CuBi₂O₄ decorated with Ru cocatalyst (denoted as Ru:CuO/Nb-CBO). The XRD pattern confirms the crystalline structure of CuBi₂O₄ (JCPDS No. 42-0334) (Supplementary Fig. 15)²³. Peaks at 35.5° and 38.7° correspond to the (002) and (111) planes of CuO in the sample. SEM images reveal that Ru:CuO/Nb-CBO exhibits a nanosheet morphology (Fig. 3a and Supplementary Fig. 16). TEM analysis reveals a distinct interface between the Nb-CBO and CuO phases (Fig. 3b and Supplementary Fig. 17). The metallic Ru nanoparticles with diameters of approximately 10 nm are observed on the CuO surface (Fig. 3c). Lattice spacings of 0.318 nm and 0.252 nm are indexed to the (211) plane of CuBi₂O₄ and the (002) plane of CuO, respectively, while a lattice spacing of 0.234 nm is assigned to the (100) plane of Ru. TEM-EDX elemental mappings confirm the homogeneous distribution of Cu, Bi, O, Nb, and Ru (Fig. 3d).

Fig. 3. PEC H₂O₂ production performance of photocathodes.

a SEM image of Ru:CuO/Nb-CBO. b, c HRTEM images and d TEM-EDX elemental mappings of Ru:CuO/Nb-CBO. e LSV curves of CBO, Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO under AM 1.5 G illumination. f FEs of H₂O₂ at various applied potentials. g Yield rates of H₂O₂ at various applied potentials. h Selectivities of H₂O₂ at various applied potentials. i I-t curves at 0.6 V_RHE. All electrolytes were 1 M NaHCO₃ + 1 M NaOH (pH = 10).

The PEC H₂O₂ production performance via the 2e^- ORR pathway was evaluated for CBO, Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO photocathodes in O₂-saturated 1 M NaHCO₃ (pH = 10) under AM 1.5 G illumination (100 mW cm⁻²). All photocathodes exhibit low photocurrent densities in Ar-saturated electrolytes (Supplementary Fig. 18), while significantly enhanced photocurrents are observed in O₂-saturated electrolytes (Fig. 3e), confirming their activity for the ORR. Notably, Ru:CuO/Nb-CBO achieves the highest photocurrent density of − 2.14 mA cm⁻² at 0.6 V_RHE, outperforming CBO (− 0.4 mA cm⁻²), Nb-CBO (− 0.84 mA cm⁻²), and CuO/Nb-CBO (− 1.32 mA cm⁻²). The integrated photocurrent density derived from IPCE spectra for Ru:CuO/Nb-CBO is − 1.97 mA cm⁻² (Supplementary Fig. 19), slightly lower than the LSV value, likely due to the absence of convection during IPCE measurements. The concentration of generated H₂O₂ was also quantified using the Ce⁴⁺ colorimetric method (Supplementary Fig. 20). Within the investigated potential range, the pristine CBO exhibits a low average FE of ~ 22%, indicative of a dominant 4e^- ORR pathway. In contrast, Nb-CBO and CuO/Nb-CBO show significantly higher average FEs of ~58% and ~ 76%, respectively, demonstrating that Nb-doping and the formation of the CuO heterojunction promote the selective 2e^- ORR toward H₂O₂ production. The deposition of Ru nanoparticles further enhances the selectivity of Ru:CuO/Nb-CBO, achieving an outstanding average FE of 91% and a peak FE of ~ 91.93% at 0.6 V_RHE (Fig. 3f and Supplementary Table 5). Correspondingly, Ru:CuO/Nb-CBO delivers the highest H₂O₂ yield of 1.43 μmol cm⁻²min⁻¹ at 0.3 V_RHE (Fig. 3g), which remarkably surpasses those of CuO/Nb-CBO (0.71 μmol cm⁻²min⁻¹), Nb-CBO (0.26 μmol cm⁻²min⁻¹), and CBO (0.04 μmol cm⁻²min⁻¹) at the same potential. Moreover, Ru:CuO/Nb-CBO exhibits the highest selectivity within the potential range of 0.5-0.7 V_RHE, with an average value of ~ 95.5% (Fig. 3h).

From IMPS measurements (Supplementary Fig. 21), Ru:CuO/Nb-CBO exhibits the highest k_trans and the lowest k_rec across the tested potential range among all photocathodes (Supplementary Fig. 22). These kinetic parameters indicate accelerated charge transfer and suppressed charge recombination, consistent with the enhanced PEC performance. In the ECSA measurements, Ru:CuO/Nb-CBO has a specific capacitance of 5.73 μF cm⁻², which is 2.5-fold higher than that of CBO (2.27 μF cm⁻²) (Supplementary Fig. 23), confirming a greater number of electrochemically active sites for rapid interfacial charge transfer. Consistently, Ru:CuO/Nb-CBO exhibits the lowest charge-transfer resistance in EIS measurements among these samples (Supplementary Fig. 24 and Table 6). Importantly, the CuO layer serves as a protective coating that significantly improves the stability of the CBO-based photocathodes. Ru:CuO/Nb-CBO shows no notable current decay over 50 h of continuous operation during the 2e^- ORR (Fig. 3i). After stability testing, its structural integrity is maintained, with no observable morphological or crystallographic change (Supplementary Fig. 25a, b). XPS and ICP-MS results reveal selective leaching of Ru and Cu (Nb and Bi < 0.05 mg/L) (Supplementary Fig. 25c–e and Table 4).

Unbiased H₂O₂ production

Efficient and sustainable unbiased H₂O₂ production remains a significant challenge, primarily due to the insufficient catalytic activity and selectivity of bipolar photoelectrodes in PEC systems^24,25. Fig. 4a illustrates the schematic diagram of the unbiased system in this work. The core configuration consists of two functionally complementary photoelectrodes, where the Zn-TiO₂/BVO/WO₃ photoanode is responsible for the 2e^- WOR and the Ru:CuO/Nb-CBO photocathode mediates the 2e^- ORR. A mixed solution of 1 M NaHCO₃ (pH = 10) serves as the electrolyte, providing a stable chemical environment for the bias-free reaction. LSV measurements verify the compatibility of the paired electrodes, yielding an initial operating current density of ~ 2.08 mA cm⁻² at 0.608 V_RHE (Fig. 4b). When assembled with a proton exchange membrane (PEM), the system maintains a stable current density for over 10 h during periodic electrolyte refreshment (Fig. 4c). Scaling the photoanode and photocathode to an area of 2 cm² increases the current density by a factor of 1.9 while preserving stability over 10 h. Each cycle produces H₂O₂ at both electrodes, achieving an average FE of ~ 85% and a yield of ~ 61 µmol cm⁻² on each side (Fig. 4d, e).

Fig. 4. Unassisted H₂O₂ production.

a Schematic diagram of unbiased device of Ru:CuO/Nb-CBO ‖ Zn-TiO₂/BVO/WO₃ under illumination. b LSV curves for Ru:CuO/Nb-CBO photocathode and Zn-TiO₂/BVO/WO₃ photoanode in 1 M NaHCO₃ (pH = 10). c Chronoamperometry of unbiased Ru:CuO/Nb-CBO ‖ Zn-TiO₂/BVO/WO₃. The surface areas of the photoelectrodes are 1 cm² and 2 cm², respectively. d H₂O₂ FEs for five cycles. The surface areas of photoelectrodes are 1 cm². e H₂O₂ yields for five cycles. The surface areas of photoelectrodes are 1 cm². f Digital image of real outdoor solar tests. g J-V curves of Ru:CuO/Nb-CBO || Zn-TiO₂/BVO/WO₃ under real sun (36 and 20 cm² for both photoelectrodes). h H₂O₂ FEs and yields for 1 h test in membrane-free systems (The solvent volume is 100 mL). i Bacterial inactivation efficiencies of PEC-generated H₂O₂ against E. coli and S. aureus over time.

To evaluate the device’s adaptability for potential application, J-V curves were measured under simulated AM1.5 G sunlight using 2 cm² electrodes and under natural sunlight using 20 and 36 cm² electrodes (Fig. 4f, g). A 1 h continuous production test under natural sunlight confirms stable performance, yielding FEs of 165.38% (20 cm²) and 163.37% (36 cm²) with corresponding H₂O₂ concentrations of 7.64 and 11.44 mmol L⁻¹, respectively (Fig. 4h). Notably, the system delivers a cumulative FE of ~170% in a membrane-free system, which exceeds the theoretical upper limit of 100% for a single-electrode process because both the photoanode and photocathode simultaneously produce H₂O₂ using the same photogenerated charge flow^26,27. Stable photocurrent responses under both conditions demonstrate the practical adaptability of the device without requiring specialized light source regulation (Supplementary Fig. 26). Using the H₂O₂ produced at the achieved concentrations of 7.64 mmol L⁻¹, two pathogenic species are completely inactivated (>99%) within 60 min (Fig. 4i and Supplementary Fig. 27)²⁸. These results confirm the system’s capacity for rapid H₂O₂ generation and effective water disinfection without requiring external chemical input. Compared to previously reported bias-free PEC systems, our cell offers distinct advances (Supplementary Table 7). The successful operation under natural sunlight with large area photoelectrodes directly validates the practical potential of our system for decentralized solar-driven hydrogen peroxide synthesis.

Mechanism analysis

In situ attenuated total reflection Fourier transform infrared spectroscopy (ATR-SEIRAS) was performed on the photoanodes to track the evolution of surface species under progressively increasing applied bias (Fig. 5a and Supplementary Fig. 28). For BVO and BVO/WO₃, a marked enhancement of the characteristic peak at ~ 1650 cm⁻¹, which is attributed to the symmetric C-O vibration of HCO₃^-, is observed^29–32. Concurrently, the peak at ~ 1200 cm⁻¹ is associated with the O-O stretching vibration of *OOH species related to the 4e^- OER and H₂O₂ adsorption. The *OOH peak intensities on Zn-TiO₂/BVO/WO₃ and TiO₂/BVO/WO₃ are significantly weaker than those of BVO/WO₃ and BVO, and nearly vanish at 1.6 V_RHE on Zn-TiO₂/BVO/WO₃. This indicates that the TiO₂ heterojunction effectively suppresses the 4e^- OER pathway, and the induced Zn further inhibits the anodic decomposition of H₂O₂. It is noteworthy that a peak at ~1400 cm⁻¹ corresponding to CO₃^2- is consistently observed throughout the measurements³³, which can be attributed to the distinct equilibria between surface and bulk phases in bicarbonate systems, especially under an applied bias. The emergence of the CO₃^2- peak provides clear evidence of substantial H₂O₂ generation, a conclusion further corroborated by the corresponding FE measurements and the detected H₂O₂ production via the bicarbonate-mediated 2e^- WOR^34,35. In contrast, in bicarbonate-free electrolyte, only a weak *OOH peak appears, and the related H₂O₂ FE drops to 8.3% for Zn-TiO₂/BVO/WO₃ (Supplementary Fig. 28c, d), confirming that bicarbonate species occupy the active sites and prevent H₂O₂ oxidation.

Fig. 5. Reaction mechanism investigation.

a In situ ATR-SEIRAS spectra of adsorbed intermediates recorded on BVO/WO₃ and Zn-TiO₂/BVO/WO₃. b In situ Raman spectra for BVO/WO₃ and Zn-TiO₂/BVO/WO₃. c EPR spectra detecting hydroperoxyl (*OOH) radicals on BVO/WO₃ and Zn-TiO₂/BVO/WO₃ under light illumination. d Electrochemical online DEMS results over photoanodes for the WOR. e Schematic diagram of the photoanode reaction pathway. f In situ ATR-SEIRAS spectra of adsorbed intermediates recorded on Nb-CBO and Ru:CuO/Nb-CBO. g In situ Raman spectra for Nb-CBO and Ru:CuO/Nb-CBO. h EPR spectra illumination. i Schematic diagram of the photocathode reaction pathway.

In situ Raman measurements were performed to monitor intermediate dynamics under applied bias³⁶. As shown in Fig. 5b and Supplementary Fig. 29, for Zn-TiO₂/BVO/WO₃, the characteristic signal of the *OOH intermediate exhibits a progressive decrease with increasing potential from 0.6 to 1.6 V_RHE. TiO₂/BVO/WO₃ exhibits a weak but discernible *OOH peak. This decreasing trend of *OOH peak on Zn-TiO₂/BVO/WO₃ differs from those of BVO/WO₃ and BVO, in which the *OOH signal intensifies continuously and significantly with increasing potential, ultimately promoting oxygen evolution. The detection of a trace-level *OOH signal on Zn-TiO₂/BVO/WO₃ indicates that the Zn-induced TiO₂ passivation layer effectively stabilize the key intermediate of the 2e^- WOR pathway, thereby preventing its dissociation into *O species and suppressing the competing 4e^- OER.

Electron paramagnetic resonance (EPR) measurements further confirm the generation of the *OOH intermediate (Fig. 5c). Compared to BVO/WO₃, the characteristic *OOH signal is significantly weaker for Zn-TiO₂/BVO/WO₃. In the differential electrochemical mass spectrometry (DEMS) analysis (Fig. 5d), although O₂ evolution signals are detected from all four photoanodes, the signal intensity for Zn-TiO₂/BVO/WO₃ is nearly eliminated, confirming that the Zn-TiO₂ passivation layer strongly suppresses the competing 4e^- OER pathway³⁷. Fig. 5e illustrates the proposed reaction pathways of the photoanodes under applied bias in the electrolytes. BVO/WO₃ primarily follows the 4e^- OER pathway, in which *O combines with *OH to form the *OOH intermediate, leading to O₂ evolution. Zn-TiO₂/BVO/WO₃ relies on the Zn-induced TiO₂ passivation layer to suppress *OOH formation, favoring the 2e^- WOR pathway for H₂O₂ production.

In situ ATR-SEIRAS spectra reveal significant differences among the four photocathodes during the ORR (Fig. 5f and Supplementary Fig. 30). The ORR pathway involves two proton-coupled electron transfer steps, with *OOH as the key intermediate³⁸. For Nb-CBO and CBO, two characteristic peaks are detected with one being a gradually intensifying peak assigned to the *OH adsorption, and another being a weak peak corresponding to the *OOH stretching vibration, indicating a preferential tendency toward the 4e^- pathway. Conversely, Ru:CuO/Nb-CBO and CuO/Nb-CBO exhibit a drastically weakened *OH signal and a pronounced *OOH peak, the intensity of which increases as the potential is reduced from 0.8 to 0.3 V_RHE. This trend correlates with the enhanced H₂O₂ FEs, confirming that Ru modification and CuO heterojunction formation shift the reaction pathway from 4e^- ORR to 2e^- ORR by stabilizing *OOH and suppressing *OH formation.

In situ Raman measurements tracked intermediate vibrations under applied bias (Fig. 5g and Supplementary Fig. 31)³⁹. For Ru:CuO/Nb-CBO, two characteristic peaks are detected, with one assigned to the *OO^- stretching vibration derived from O₂ adsorption and subsequent electron transfer, and the other corresponding to *OOH stretching vibration. As the potential decreases from 0.8 to 0.3 V_RHE, the intensity of the *OO^- peak remains stable, whereas the *OOH peak increases markedly. This reflects efficient protonation of *OO^- to form *OOH, a critical step that promotes the 2e^- ORR pathway. In contrast, CBO, Nb-CBO, and CuO/Nb-CBO exhibit the *OO^- peak along with a weak *OOH signal. EPR measurements validate the formation of key radicals (Fig. 5h and Supplementary Fig. 32). Both Nb-CBO and Ru:CuO/Nb-CBO show a signal characteristic of *OO^-, but only Ru:CuO/Nb-CBO displays a clear *OOH signal. Although both materials can adsorb oxygen, Ru:CuO/Nb-CBO predominantly follows the 2e^- ORR pathway for H₂O₂ generation. Figure 5i illustrates the proposed reaction pathways of the photocathodes under applied bias in the electrolytes. Nb-CBO primarily proceeds via the 4e^- ORR pathway, in which the generated *OOH intermediate decomposes into *OH, ultimately yielding H₂O as the final product. Ru:CuO/Nb-CBO relies on the synergy between Ru nanoparticles and CuO to suppress *OOH decomposition, promoting O₂ and H₂O adsorption to form and stabilize the *OOH intermediate, thus favoring the 2e^- ORR pathway for H₂O₂ production.

DFT calculations

Density functional theory (DFT) calculations were performed to elucidate the fundamental mechanisms of PEC WOR on the photoanodes. Supplementary Figs. 33–35 and Supplementary Data 1 present the catalytic models, including BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃, whose crystal structures were constructed based on the experimentally determined configurations and systematically optimized. Considering the inherent limitations of the computational model in reproducing the practical catalytic environment, the DFT results serve as complementary evidence to the experimental data rather than standalone conclusions. Density of states (DOS) analysis reveals that the induced Zn works synergistically with TiO₂ to modulate the electronic structure of the heterojunction. Specifically, the d-band center of Ti for Zn-TiO₂/BVO/WO₃ shifts to − 1.88 eV compared with − 2.03 eV for TiO₂/BVO/WO₃ and − 2.56 eV for BVO/WO₃ (Fig. 6a). This upward shift strengthens the interaction between the photoanode surface and *HCO₃^-, thereby improving the selectivity toward the 2e^- WOR. By integrating in situ characterization results with DFT calculations, we clarify the 2e^- WOR and 4e^- OER pathways over the photoanodes and quantify the free energy of each intermediate in these processes. The 2e^- WOR pathway responsible for H₂O₂ generation employs *HCO₃^- as the key mediator, which adsorbs onto the active sites of Zn-TiO₂/BVO/WO₃ to form the critical *HCO₄^- intermediate, followed by proton-electron transfer to release H₂O₂. Zn-TiO₂/BVO/WO₃ exhibits the strongest adsorption affinity toward the critical *HCO₄^- intermediate (ΔG_*HCO4^-= − 0.17 eV) along the 2e^- WOR pathway, which is substantially lower than those of TiO₂/BVO/WO₃ (0.31 eV) and BVO/WO₃ (0.46 eV) (Fig. 6b), promoting efficient 2e^- PEC WOR. In contrast, the 4e^- OER pathway for O₂ production proceeds via sequential formation of *OH, *O, and *OOH intermediates (H₂O → *OH → *O → *OOH → O₂). For the *O intermediate in the 4e^- pathway, Zn-TiO₂/BVO/WO₃ shows a significantly higher free energy (ΔG_*O = 0.49 eV) compared with other samples (Fig. 6c), hindering the 4e^- OER process.

Fig. 6. DFT calculations.

a Total DOS of surface states for BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃ photoanodes. b 2e^- WOR and c 4e^- OER processes on BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃ photoanodes. d Total DOS of surface states for Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO photocathodes. e 2e^- ORR and f 4e^- OER processes on Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO photocathodes.

Using optimized models of Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO, we conducted additional DFT calculations to uncover the underlying reasons for the improved 2e^- ORR performance on the photocathodes (Supplementary Figs. 36-38)⁴⁰. DOS analysis shows that Ru:CuO/Nb-CBO exhibits a d-band center of − 1.59 eV, which is upshifted relative to CuO/Nb-CBO (− 1.71 eV) and Nb-CBO (− 2.07 eV) (Fig. 6d). This shift strengthens the adsorption of O₂ and key intermediates. Through a combination of in situ characterization and DFT calculations, we elucidate the 2e^- ORR and 4e^- ORR pathways on the Ru:CuO/Nb-CBO photocathode, and calculate the free energy profile of each intermediate involved. The 2e^- ORR pathway for H₂O₂ production involves O₂ adsorption on active sites to form the critical *OOH intermediate, which then undergoes electron transfer to release H₂O₂. Ru:CuO/Nb-CBO displays the most favorable adsorption affinity toward the critical *OOH intermediate (ΔG_*OOH = 0.07 eV) for the 2e^- ORR pathway, which is lower than those of Nb-CBO (0.41 eV) and CuO/Nb-CBO (0.22 eV) (Fig. 6e), thereby driving efficient 2e^- PEC ORR. Conversely, the 4e^- ORR pathway leading to H₂O proceeds via dissociation of *OOH into the critical *O intermediate, followed by sequential electron transfers to form *OH and finally H₂O. Ru:CuO/Nb-CBO exhibits the highest free energy for the critical *O intermediate (ΔG_*O = 1.55 eV) in the 4e^- pathway, which is higher than those of CuO/Nb-CBO (1.46 eV) and Nb-CBO (1.44 eV) (Fig. 6f). The above findings demonstrate that the Zn-TiO₂/BVO/WO₃ heterojunction for the photoanode, together with CuO/Nb-CBO heterojunction and Ru active sites for the photocathode, collectively optimizes the electronic structures of the catalytic systems. Simultaneously, the optimized photoanode and photocathode enhance the adsorption affinity of the critical intermediates involved in the 2e^- pathways while inhibiting the progression of the competing 4e^- pathways, enabling the PEC bias-free system to achieve high H₂O₂ activity and selectivity.

Discussion

This work presents a high-performance, bias-free photoelectrochemical bipolar platform for efficient and sustainable H₂O₂ synthesis. The integrated system combines a Zn-TiO₂/BVO/WO₃ photoanode, optimized for the two-electron water oxidation reaction, with a Ru:CuO/Nb-CBO photocathode tailored for the two-electron oxygen reduction reaction. Through rational design, key challenges such as low selectivity, sluggish kinetics, and poor stability have been effectively overcome. The photoanode achieves an FE of 90.2% for H₂O₂ at 0.6 V_RHE with stable operation for 10 h, while the photocathode delivers an FE of 91.9% and maintains exceptional performance for over 50 h. The system also demonstrates practical viability under natural sunlight, sustaining an FE of 163.37% with 36 cm² electrodes, while the in situ produced H₂O₂ solution enables over 99% bacterial inactivation. This study provides a versatile strategy for solar-driven, low-carbon H₂O₂ production and offers fundamental insights into designing high-performance PEC bipolar systems through targeted intermediate regulation and optimized energy level alignment.

Methods

Material preparation

Preparation of BVO, BVO/WO₃, TiO₂/BVO/WO₃, and Zn-TiO₂/BVO/WO₃ photoanodes

Initially, a precursor mixture was prepared by dissolving 1.25 g of tungstic acid (Aladdin, 99%) and 0.5 g of polyvinyl alcohol (Innochem, PVA-124) in 20 mL of hydrogen peroxide (Sinopharm, 30 wt%) under continuous stirring at 80 °C. This mixture was spin-coated onto an FTO substrate, and subsequently annealed at 500 °C for 2 h to produce a WO₃ seed layer. Next, another precursor solution was prepared by dissolving 0.625 g of tungstic acid in a mixed solvent of 10 mL of 30 wt% H₂O₂ and 30 mL of deionized water, with stirring maintained at 92 °C. A precursor mixture was then prepared by combining 3 mL of the aforementioned solution, 2.75 mL of deionized water, 0.25 mL of 37.5 wt% hydrochloric acid (Sinopharm, 36  ~  38 wt%), and 10 mL of ethylene glycol (Alfa,99%). This mixture, together with the FTO substrate bearing the WO₃ seed layer, was transferred into a 50 mL polytetrafluoroethylene-lined autoclave and heated at 180 °C for 3 h. After the reaction, the substrate was collected, thoroughly rinsed with deionized water, dried at 60 °C, and finally annealed in air at 500 °C for 1 h to obtain the WO₃ nanofilm. Subsequently, a nanoporous BiVO₄ layer was deposited on the WO₃/FTO substrate through electrodeposition using the following procedure: first, 50 mL of a 0.4 M potassium iodide (Aladdin, 99%) solution was adjusted to pH 1.7 with concentrated nitric acid (Sinopharm, GR), followed by the addition of 0.97 g of bismuth nitrate pentahydrate (Aladdin, 99%). Then, 20 mL of 0.23 M p-benzoquinone (Aladdin, 97%) solution in ethanol (Aladdin, 99.5%) was introduced. Using a three-electrode configuration, a BiOI film was electrodeposited onto the WO₃/FTO substrate at a working potential of −0.1 V (vs. Ag/AgCl) for 180 s. Afterwards, 0.15 mL of a 0.2 M vanadyl acetylacetonate (Aladdin, 98%) solution in dimethyl sulfoxide (Aladdin, 99%) was drop-cast onto the BiOI film. The sample was then annealed at 450 °C for 2 h to form BiVO₄. The as-prepared electrode was soaked in 1 M NaOH for 30 min to remove residual V₂O₅, followed by thorough rinsing with deionized water and drying, yielding the BVO/WO₃ photoanode. For comparison, a pure BVO photoanode was prepared directly on a bare FTO substrate. Next, a TiO₂ layer was deposited by atomic layer deposition (ALD) using titanium tetrakis(dimethylamino)titanium and water as the Ti and O precursors, respectively. The process consisted of 30 cycles at 120 °C. Subsequent annealing in an argon atmosphere at 200 °C for 30 min produced the TiO₂/BVO/WO₃ multilayer structure. Finally, the samples were immersed in a 5 mM Zn(NO₃)₂ (Aladdin, 99%) solution for 1 min and then sintered at 200 °C for 20 min under argon, resulting in the Zn-TiO₂/BVO/WO₃ photoanode.

Preparation of CBO, Nb-CBO, CuO/Nb-CBO, and Ru:CuO/Nb-CBO photocathodes

A 0.23 M p-benzoquinone solution in ethanol was stirred until fully dissolved, while a 0.4 M KI solution was prepared and its pH was adjusted to 1.7 using HNO₃. Then, 0.04 mol Bi(NO₃)₃·5H₂O was dissolved in this solution and blended with the p-benzoquinone solution. BiOI was electrodeposited onto FTO substrates at − 0.1 V vs. Ag/AgCl for 300 s. Next, Cu was electrodeposited onto the BiOI/FTO electrode in a 0.2 M Cu(CH₃COO)₂·H₂O (Aladdin, 98%) solution at − 0.3 V vs. Ag/AgCl for 100 s, followed by oxygen annealing at 450 °C for 3 h to form the CBO electrode. For Nb-doping, 3 mM ammonium niobamyl oxalate hydrate (Aladdin, 99.9%) was drop-cast onto the BiOI precursor and annealed under the same conditions. The CuO/Nb-CBO electrode was prepared by sputtering Cu onto Nb-CBO at 150 W for 7 min in an Ar atmosphere, followed by oxygen annealing at 450 °C for 100 min. Finally, Ru nanoparticles were deposited by immersing the electrode in Ru(acac)₂ (Aladdin, 97%) solutions with various concentrations, followed by oxygen annealing to form the Ru:CuO/Nb-CBO.

Photoelectrochemical measurements

Anodic photoelectrochemical (PEC) measurements were performed in a closed H-cell using a three-electrode system with Zn-TiO₂/BVO/WO₃ as the working electrode, Ag/AgCl as the reference electrode, and Pt foil as the counter electrode. A Nafion 117 membrane separated the working electrode from the counter electrode. The electrolyte consisted of 1 M NaHCO₃ (pH = 10, adjusted with NaOH). Measurements were performed for 1 h at various applied potentials. Before each independent test, the optical power was calibrated to align with the standard AM 1.5 G illumination (100 mW cm⁻²). The potentials were recored against the Ag/AgCl reference electrode and converted to the reversible hydrogen electrode (RHE) scale. Cathodic photoelectrochemical measurements were performed in a closed H-cell using a three-electrode system with Ru:CuO/Nb-CBO as the working electrode, Ag/AgCl as the reference electrode, and Pt foil as the counter electrode. The electrolyte consisted of 1 M NaHCO₃ and 1 M NaOH (pH = 10). All measurements were performed on three independent samples under controlled conditions. There is no IR correction in PEC work.

Product analysis

The evolution of H₂O₂ was detected using the Ce⁴⁺ colorimetric method. During irradiation, an aliquot (5 mL) of the sample was collected with a syringe and mixed with 4.5 mL of 1 M H₂SO₄ and 0.5 mL of 2 mM Ce(SO₄)₂ in 1 M H₂SO₄. The resulting solution was analyzed by UV-vis spectroscopy. The Faradaic efficiency (FE) of H₂O₂ can be calculated using the following equation:

$$\begin{array}{c}\hfill {\mathrm{\eta }}_{{\mathrm{H}}_2{\mathrm{O}}_2}=\frac{\mathrm{amount}\mathrm{of}\mathrm{generated}{\mathrm{H}}_2{\mathrm{O}}_2}{\mathrm{theoretical}\mathrm{of}{\mathrm{H}}_2{\mathrm{O}}_2}\times 100\%\hfill \\ \hfill =\frac{\mathrm{amount}\mathrm{of}\mathrm{generated}{\mathrm{H}}_2{\mathrm{O}}_2}{\mathrm{amount}\mathrm{of}\mathrm{generated}\mathrm{electrons}÷2}\times 100\%\hfill \end{array}$$

Definition of total FE for the bias‑free, membrane-free PEC system

In a closed circuit, the total charge passed through the external circuit (Q_total) equals the charge consumed by anodic oxidation (Q_anode) and that consumed by cathodic reduction (Q_cathode): Q_total= Q_anode= Q_cathode. Both the photoanode and the photocathode catalyze independent 2e^- pathways for H₂O₂ generation:

Anode (2e^-WOR):2H₂O→H₂O₂+2H⁺+2e^-

Cathode (2e^-ORR):O₂+2H⁺ + 2e^-→H₂O₂

The total FE of the tandem system is defined based on the total charge passing through the external circuit and the total electron demand for all H₂O₂ molecules produced at both electrodes. It is not a simple sum of the individual electrode FEs. The definition is:

$$\mathrm{Total}\mathrm{FE}=\frac{2\mathrm{F}\left({\mathrm{n}}_{\mathrm{anode}}+{\mathrm{n}}_{\mathrm{cathode}}\right)}{{\mathrm{Q}}_{\mathrm{total}}}\times 100\%$$

This definition avoids double-counting charge and conforms to charge conservation, because Q_total represents the only charge pathway driving both half-reactions simultaneously. The individual FE of each electrode is calculated based on its own half-reaction charge (Q_anode or Q_cathode), which is consistent with standard single-electrode FE definitions.

Bactericidal experiment

Escherichia coli (ATCC 25922) and Staphylococcus aureus (ATCC 25923) were purchased from a certified microbial resource center. Pre-cultivation was performed by inoculating the strains into sterile Luria-Bertani (LB) medium and incubating at 37 °C for 12 h to reach the exponential growth phase. The bactericidal treatment was carried out by mixing 1 mL of the pre-cultivated bacterial suspension with 9 mL of the PEC-generated electrolyte (adjusted from pH 10 to pH 7 using KH₂PO₄) at a volume ratio of 1:9, followed by incubation at 37 °C in the  dark. This dark incubation was designed to exclude potential photodynamic effects. Sampling was performed at 20, 40, and 60 min after mixing. For each sampling time, 100 μL aliquots of the mixture were serially diluted from 10⁰ to 10⁻⁴ using sterile 0.9% NaCl solution and then spread evenly on LB agar plates. The plates were inverted and incubated at 37 °C for 24 h. After incubation, colony-forming units (CFU) were counted to calculate the bacterial inactivation efficiency.

Computational details

All spin-polarized density functional theory (DFT) computations were carried out using the Vienna Ab-initio Simulation Package, with the projector-augmented wave method employed to represent electron-ion interactions^41,42. The exchange-correlation energy was treated within the generalized gradient approximation using the Perdew-Burke-Ernzerhof (PBE) parametrization, a functional that has been extensively validated for transition metal oxides, including BiVO₄, TiO₂, WO₃, and CuBi₂O₄-based systems⁴³.

For both the OER on photoanodes and the ORR toward H₂O₂ on photocathodes, the Gibbs free energy change (G) for each elementary step was determined via the computational hydrogen electrode (CHE) model, adapted to incorporate photoelectrochemical conditions.

To construct the reaction energy diagrams, the photogenerated voltage (V) was integrated into the DFT framework to simulate the actual photoelectrochemical environment. This photovoltage shifts the Fermi level of the catalyst, thereby modifying the activation barriers of individual reaction steps. Specifically, the effective overpotential under illumination (η_light) was corrected by the photogenerated voltage according to η_light= η_dark - V, where η_dark represents the overpotential in the absence of photovoltage. This correction accounts for the role of photogenerated carriers in facilitating the electrochemical reaction, leading to a reduction in the required overpotential. The magnitude of V was derived from the optical absorption properties of the catalyst, as determined by the DFT calculations.

Author contributions

L.M.W., L.W., and R.-T.G. proposed and supervised the project. H.C. performed the materials synthesis and PEC measurements, and wrote the original manuscript under the discussion of R.-T.G. L.M.W., L.W., and R.-T.G. revised the manuscript. All authors contributed to the discussion and manuscript review.

Peer review

Peer review information

Nature Communications thanks Ji-Wook Jang, Wooseok Yang, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

The work was supported by the National Science and Technology Major Project (2022YFA1205200), the National Natural Science Foundations of China (22269016, 22479083, 22405138), the Group Project of Developing Inner Mongolia through Talents from the Talents Work Leading Group under the CPC Inner Mongolia Autonomous Regional Committee (2025TYL03), the Natural Science Foundations of Inner Mongolia Autonomous Regional (2026ZD007), the Leading Talents in Science and Technology of Inner Mongolia Autonomous Regional (2026LJRC0002, 2026LJRC0004), and program for Young Talents of Science and Technology in Universities of Inner Mongolia Autonomous Region (Grant NJYT25001).

Data availability

The source data generated in this study are provided in the Source Data file. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-75533-4.