Introduction

The redox mediator is ubiquitous and plays an important role in many energy conversions, such as NADPH/NADP+ redox pair in the light reaction of natural photosynthesis and water-splitting processes1,2,3. One of the keys in natural photosynthesis depends on the coordinated transfer of multiple electrons and protons, which is facilitated by the redox mediator. Leveraging this natural functionality, researchers have been investigating ways to replicate these processes in artificial systems. By unraveling the complex mechanisms of redox mediators, synthetic analogs can be developed to emulate these roles4,5. Such advancements open up possibilities for innovative energy solutions and sustainable technologies.

Mimicking the natural photosynthesis process, where electrons transfer through a Z-scheme pathway and energy storage and utilization occur under time-space decoupled circumstances, a novel artificial photosynthesis strategy has been developed6,7. In this approach, a redox mediator is employed to balance the solar energy storage subsystem with the hydrogen production subsystem. For instance, utilizing H3PMo12O40 as a redox mediator to facilitate charge transfer, the conceptualization of electron-coupled proton buffers (ECPBs) was pioneered by Cronin and Symes8. In the ECPBs system, solar energy is harvested through a photoelectrochemical subsystem and stored in reduced state H3PMo12O40. The stored energy was then utilized to produce hydrogen by coupling with an electrochemical subsystem. On this base, Wu et al. developed an aqueous Zn-polyoxometalate battery for decoupled alkaline water electrolysis for hydrogen production.9 In addition, Li and co-workers proposed the hydrogen farm project (HFP), which comprises solar energy capturing and hydrogen production subsystems integrated by a shuttle redox mediator, separating the H2 and O2 evolution in space for particulate photocatalytic water splitting10. Streb et al. covalently attached ruthenium-complex photosensitizers to Dawson-type polyoxometalates, which successfully serve as electron storage sites and hydrogen evolution catalysts.11 These strategies can effectively adapt the intermittent and unpredictable nature of solar irradiation through the synergistic approaches of solar energy and charge conversion for chemical fuels production. The key and bottleneck to mediate these strategies is to design stable and reversible redox mediates with suitable redox potential for efficient energy storage. The reversible redox mediators enable the efficient realization of time-space decoupling of the oxygen evolution reaction (OER) from the hydrogen evolution reaction (HER) in solar-to-hydrogen conversion, which fundamentally addresses the severe reverse reaction and the challenging hydrogen/oxygen separation issues in conventional overall water-splitting processes12,13.

Soluble redox mediators are undoubtedly the cornerstone of such a time-space decoupled solar-driven water splitting system14. The primary consideration for this system is ensuring that the redox mediator possesses suitable redox potential. This characteristic allows the gas evolution process of the entire system to run autonomously, without reliance on a single driving force. It also can effectively achieve a flexible transition between hydrogen production and hydrogen storage processes. Till now, various soluble inorganic redox mediators, such as Fe3+/Fe2+ (0.77 V), Fe(CN)63-/Fe(CN)64- (0.48 V), IO3-/I- (1.08 V), I3-/I- (0.55 V), VO2+/VO2+ (1.00 V), have been developed and widely used as electron mediators in Z-scheme photocatalytic water splitting15,16, dye-sensitized solar cell17,18 and redox flow battery etc19,20,21,22. However, most of these redox mediators facilitate only single-electron transfer, which greatly limits the solar energy storage capacity during the reaction23,24. In addition, metal-organic photocatalysts for water oxidation, represented by [Ru(bpy)3]2+ or cobalt complexes, have received widespread attention. [Ru(bpy)3]2+ exhibits good catalytic activity but typically requires the use of electron donors to compensate for the electrons consumed in the water reduction reaction, and its stability is limited25,26. Cobalt complexes mainly participate in water oxidation by absorbing photons or through the action of photo-generated electrons, but their water splitting rate is relatively low. Therefore, in practical applications, they are often used in combination or integrated with other catalysts to enhance their efficiency27,28. Therefore, it is crucial to develop efficient and reversible redox mediators with multi-electron characteristics that can be operated at suitable redox potentials and stable pH buffer ability. This will help to enhance the scalability and practical applicability of redox media in the system.

Polyoxometalates (POMs) are a class of poly-nuclear metal-oxide clusters, where the multiple metal centers can undergo redox transformations and form stable intermediate states, providing stable multi-electron transfer capabilities that enable efficient participation in reactions.29,30,31,32 Additionally, the structural tunability of POMs allows for adjustable redox potentials in electrochemical applications, which can be modulated by changes in the type of metal centers.33,34,35,36 The metal-oxygen bonds in the framework facilitate smooth proton transfer, and the electronic structure of the metal centers can quickly adapt to this process, thereby better supporting the cooperative proton-electron transfer (CPET) mechanism, which is crucial for long-term cycle stability in fast redox reactions, an important aspect of water splitting reactions.

Photocatalysts that can be directly used for overall water splitting are central to research in artificial photosynthesis and water splitting. An ideal photocatalyst needs to simultaneously possess broad-spectrum light absorption, efficient charge separation, a suitable bandgap structure for oxygen and hydrogen evolution, and good chemical stability. Although traditional semiconductors (e.g. oxide, nitrides, and sulfides materials) exhibit good photocatalytic properties, their limited light absorption range, high carrier recombination rates, and stability issues restrict their long-term application37,38,39. Emerging materials, such as metal-organic frameworks (MOFs), black phosphorus, and POMs, offer broader design possibilities and can notable enhance catalytic efficiency when combined with traditional semiconductor photocatalysts40,41,42. Among them, POMs with nanometer-level monodispersity provide a modular design strategy, enabling precise and rational optimization of the photocatalytic overall water splitting system’s performance, effectively filling the gap in the development of advanced redox mediators in current water splitting technologies.

In this work, based on the tunable characteristics of polyoxometalates, a screening process for transition metal substituted models was built upon high-throughput first-principles calculations. Vanadium was selected as the substituent atom for the mono- and tri-substitution structural derivatization of Dawson-type phosphotungstate. The structure-activity relationship accompanying these structural adjustments was explored from both kinetic and thermodynamic perspectives using density functional theory. Subsequently, [P2W17VO62]7- ({P2W17V}) and [P2W15V3O62]9- ({P2W15V3}) molecular clusters were screened to be the suitable redox mediators to establish a photocatalysis-electrolysis relay system for decoupling hydrogen and oxygen evolution under visible light. The vanadium-substituted {P2W15V3} polyoxometalates exhibit quick redox kinetics, multi-electron transfer property with strong pH buffer ability. Coupling with their multi-electron storage capacity and synergistic proton-electron transfer characteristics, this ensures a long-term stability of the reaction solution, which is essential for the photocatalytic water oxidation process. When coupling the photocatalytic O2 evolution using typical BiVO4 photocatalyst and electrolysis subsystems using {P2W15V3} as 3-electron redox mediator, a rapid hydrogen production can be achieved at a low voltage of 0.6 V, and a Faradaic efficiency exceeding 98.5% in this system was achieved while maintaining long-term stable operation.

Results and discussion

In an artificial solar-to-hydrogen system utilizing redox mediators, the redox potential of the redox mediator needs to match the conduction band energy level of the photocatalyst to achieve solar energy storage and utilization via the photocatalytic water oxidation reaction. Compared with WO3 and TiO2, BiVO4 photocatalyst is widely used in photocatalytic water oxidation process owing to its visible light response and great charge separation property43,44,45. Herein, well-defined decahedral BiVO4 was introduced for the redox mediator development (Fig. 1a). The monoclinic phase BiVO4 crystals were successfully synthesised as evidenced by Raman and X-ray diffraction (XRD) spectroscopic characterisation (Supplementary Figs. 1, 2). To accurately select a suitable redox potential, it is first necessary to determine the band gap and the corresponding conduction band position of BiVO4. Ultraviolet-visible diffuse reflectance spectroscopy (UV-Vis DRS) reveals that the band gap (Eg) of BiVO4 is 2.4 eV vs. RHE (Fig. 1b and Supplementary Fig. 4). Then, Mott-Schottky measurement was performed to confirm the conduction band position of BiVO446. By plotting the inverse capacitance squared (1/C²) against the potential (V) to create a Mott-Schottky plot, the conduction band (CB) position of BiVO4 is found to be 0.18 eV (Fig. 1c), and the corresponding valence band (VB) potential is 2.58 eV. Although the CB and VB positions of BiVO4 are theoretically close to the hydrogen evolution (below 0 V vs. RHE) and oxygen evolution (above 1.23 V vs. RHE) potentials for water splitting, they do not match perfectly. Moreover, although photoexcited electrons and holes are generated upon light irradiation, these photogenerated charge carriers tend to recombine in the absence of suitable redox mediators. As a result, BiVO4 alone is insufficient to achieve photocatalytic water splitting. When POM serves as the energy storage ions in the photocatalytic water oxidation process on BiVO4, the oxidized polyoxometalate (Ox-POM) effectively captures photoexcited electrons at the surface of BiVO4 to form proton-coupled reduced polyoxometalate (Re-POM[H]), while leaving photoexcited holes on the BiVO4 for the water oxidation reaction to produce O2 (Supplementary Fig. 3). This also requires that the redox potential of the redox mediator should be between 0.18 V and 1.23 V vs. SHE (pH = 0), from which the energy diagram of the photocatalytic system with BiVO4 as the photocatalyst was determined (Fig. 1d).

Fig. 1: Electronic band gap position of BiVO4 and the redox potential range for suitable redox mediators.
figure 1

a Scanning electron microscope (SEM) images of the BiVO4 samples and illustration shows a single BiVO4 crystal with (110) and (010) facets. b Ultraviolet-visible diffuse reflectance spectroscopy (UV-Vis DRS) of the BiVO4 sample. c Mott-Schottky plots (inverse capacitance squared (1/C2) versus potential (V)) of the BiVO4 samples. d Energy diagram of the photocatalytic system with the participation of POM-based redox mediators. CB stands for conduction band. VB stands for valence band. e- stands for photogenerated electron. h+ stands for photogenerated hole.

Full size image

As discussed above, the number of electrons and the redox potential of redox mediators are crucial to constructing such artificial solar-hydrogen conversion system which needs to match with the CB and valence band (VB) of BiVO4 photocatalyst. POMs exhibit interesting behavior with tailorable cluster structure to tune the redox behaviors. For example, the D3h α-P2W18O626− is a classic Dawson-type polyoxometalate, consisting of two {PO4} tetrahedra in the center and 18 {WO6} octahedra connected by edge-sharing or angle-sharing47. P2W18 O626− ({P2W18}) usually has a stable and reversible redox potential and has shown numerous interesting applications as aqueous redox mediator for energy storage and generation48,49. However, the initial one-electron reduction potential of {P2W18} is 0.28 V vs. SHE which is very close to the conduction band level of BiVO4 (Fig. 2a). This means it will not be sufficiently reduced during photocatalysis, thereby limiting the storage of more active hydrogen. Interestingly, the {P2W18} cluster can undergo vacancy trimming via molecular defect engineering under mildly alkaline conditions, resulting in different numbers of vacancies. The defect sites on the vacancy structure have a strong coordination activity for transition metals. Therefore, a series of redox-active transition metal elements can be introduced to substitute the vacancies in {P2W17}, thereby further adjusting the redox potential of {P2W17M}. Through the inorganic crystal structure database (ICSD), we searched for clipped Dawson-type POMs containing fourth-period transition metals to realize a high-throughput screening workflow (Fig. 2a). The frontier molecular orbitals of POMs substituted with a single active center were analyzed to determine the electron-donating capability and structural stability of the substituted POMs. Based on the band gap characteristics and conduction band position of the photocatalyst BiVO4, further molecular tailoring of the screened single active center-substituted POMs was performed, which aims at exploring mediators with multiple redox centers to achieve multi-electron storage.

Fig. 2: A theoretically high-throughput screening process of POMs-based redox mediators.
figure 2

a Schematic diagrams of ɑ-P2W18O626- single and triple vacancy molecular tailoring and the substitution based on a high-throughput screening process, and cyclic voltammetry curves of 5 mM {P2W18}, {P2W17V}, and {P2W15V3} in 1 M Na2SO4 (pH=3) at scan rate of 50 mV s−1. The yellow line represents the conduction band (CB) position of BiVO4. b The frontier molecular orbitals of {P2W18}, {P2W17}, {P2W15}, mono-substituted {P2W18} with transition metals (Ti, V, Mn, Fe, Co, Ni, Cu, and Zn), and {P2W15V3}. c The contribution of vanadium substitution to the LUMO of the {P2W15V3} structure. The calculation data is provided in Supplementary Data 1.

Full size image

The analysis of the frontier molecular orbitals of mono-substituted {P2W17M} according to Koopman’s theorem reveals that the lowest unoccupied molecular orbital (LUMO) will be tuned by the types of substituted transition metals (M = Ti, V, Mn, Fe, Co, Ni, Cu, and Zn)50 and vanadium-substituted POMs have the lowest LUMO energies (Fig. 2b). This means that V5+-substituted POMs are probably more conducive to electron injection than other substituted POMs, while ensuring their own structural integrity and stability (Supplementary Figs. 5-6 and Supplementary Discussion 1). Cyclic voltammetry (CV) shows that {P2W17V} exhibits a one-electron reversible redox peak of V4+/5+ at 0.65 V vs. SHE, which is consistent with the results of spin density analysis (Supplementary Fig. 7). This means that the structural adjustment strategy is successful. Unlike P2W18, in P2W17V, the stronger oxidant properties of V5+ lead to a more localized nature of the first d-electron. This redox potential is more ideal to match the CB of BiVO4 photocatalyst and the water splitting voltage window. However, the number of redox electrons is still limited to only one. In this regard, three vanadium fragments were further introduced into the triple vacancy {P2W15□3} to form {P2W15V3} molecular cluster in anticipation of more charge storage (Supplementary Figs. 8-13). Successfully, {P2W15V3} exhibits a 3e- redox process at similar redox potential with {P2W17V}. The introduction of more V atoms into the cluster vacancies slightly shifts the reduction potential of V towards lower potentials with a 1-electron redox potential at 0.56 V and a 2-electron redox potential at 0.41 V. This indicates that {P2W15V3} can achieve multi-electron storage at a lower potential, corresponding to greater solar energy storage and a lower hydrogen evolution potential in the photocatalytic system. Charge decomposition analysis (CDA) results show that with the continuous introduction of V atoms into the cluster vacancies, the contribution of fragments to the LUMO of POM gradually changes from {W-O} (WO) to {V-O} (VO) fragments, which may be the reason of the significant change in the CV curve of POM caused by V atoms (Fig. 2c and Supplementary Figs. 14-15)51. The introduction of V atoms also increases the amount of negative charge on the POM anion, reducing its ability to acquire electrons, which manifests itself in a rise in LUMO energy levels and a negative shift in the CV curve.

Furthermore, the electrochemical behaviors of {P2W18}, {P2W17V}, and {P2W15V3} as redox mediators were analyzed and compared in detail (Fig. 3a-c). The cyclic voltammetry (CV) curves of the three redox mediators indicate that {P2W15V3} exhibits a reversible multi-electron transfer characteristic compared to the single-electron reaction of {P2W17V} and {P2W18}, even over 100 cycles of repeated CV testing (Supplementary Fig. 16). According to the slope of the plot (j / ν0.5), the diffusion coefficient (D) of {P2W18}, {P2W17V}, and {P2W15V3} were calculated to be 3.43 × 10−6, 1.59 × 10−6, and 2.36 × 10−6 cm2 s−1, respectively (Supplementary Fig. 17). Meanwhile, based on the Butler-Volmer equation, the redox reaction rate constants (k0) for {P2W18}, {P2W17V}, and {P2W15V3} were further calculated to be 4.09 × 10−2, 2.79 × 10−2, and 3.40 × 10−2 cm s−1, respectively (Supplementary Fig. 18). Such a high electron exchange rate suggests that these clusters may facilitate faster transfer and utilization of photogenerated electrons during the photocatalytic reaction process, thereby reducing the recombination between photogenerated electrons and holes to some extent. The number of electrons in the redox process and coulombic efficiency were further evaluated by an electrochemical flow cell system (Fig. 3d-f and Supplementary Fig. 19). To be specific, the electron transfer numbers of {P2W18}, {P2W17V} and {P2W15V3} were calculated to be 0.91, 1.00 and 2.93 with high coulombic efficiencies of 96.05%, 99.63% and 99.66%, respectively. This result further demonstrates that substitution from single to multiple active sites can achieve a multiple increase in the number of electrons that can be stored per redox mediator molecule.

Fig. 3: Electrochemical analysis of POMs-based redox mediators.
figure 3

The cyclic voltammetry of 5 mM a, {P2W18}, b, {P2W17V} and c, {P2W15V3} in 1 M Na2SO4 (pH=3) at different scan rates (j represents current density); reduction and re-oxidation via bulk galvanostatic electrolysis of 5 mM d, {P2W18}, e, {P2W17V} and f, {P2W15V3} in 1 M Na2SO4 solution (25 mL, pH = 3, electrochemical cell resistance value: 125 ± 5 mΩ) at current density of 3 mA cm−2.

Full size image

The structural regulation of those clusters also brings about the changing of the corresponding proton-electron coupled transfer (PCET) behaviours. As further analyzed by the isodesmic proton-exchange reaction (IPER), the pKa results indicate that the electron and proton acquisition process of {P2W18}, {P2W17V}, and {P2W15V3} shifts from a sequential proton-electron transfer (SPET) process to a concerted proton-electron transfer (CPET) process52. Under different pH conditions, we measured the redox potentials of the redox mediators using CV, as the potential of the CPET process is typically highly pH-sensitive. The results show that a linear pH dependence of {P2W15V3} at pH = 2, 2.5, and 3 is observed, whereas {P2W18} and {P2W17V} did not exhibit any pH-dependent behavior even over a wider pH range (Supplementary Fig. 20). According to the surface electrostatic potentials analysis results of those three POMs, the introduction of V atoms increases the amount of negative charge of the POMs and enhances the attraction of H+ through electrostatic interactions (Fig. 4a). Furthermore, Marcus theory simulation results indicate that the introduction of more V atoms renders the valence states of POMs more negative and reduce the reaction barrier of V-substituted POMs to couple with protons in the reduction process, thereby facilitating the transition from SPET to CPET (Fig. 4b-d and Supplementary Tables 1-3). Based on the results of frontier molecular orbital and Marcus theory analysis, electrons transfer to the d-orbitals of V atoms, while protons are highly likely to bind with the bridging oxygens of {P2W15V3} (Supplementary Fig. 21). Conversely, during the reduction process of {P2W18} and {P2W17V}, only electron transfer occurs, while protons are predominantly dispersed within the solvent layer surrounding the POMs. Such changes of electron and proton acquisition process will greatly influence the pH buffer ability of those POMs redox mediators. As shown in Fig. 4b-d, the measurement of pH changes in the three media during the reduction process further confirms that the solutions of {P2W15V3} exhibit good buffering capacity due to their strong proton-coupled electronic properties. However, under the same conditions, the solution pH values of {P2W18} and {P2W17V} show an obvious decreasing trend at the beginning of the reaction. During the photocatalytic reaction on BiVO4, {P2W15V3} can mitigate the impact of pH changes on the surface of BiVO4 to ensure a long-term stability of photocatalytic water splitting.

Fig. 4: Proton-electron transfer mechanisms in POMs-based redox mediators.
figure 4

a The pKa values of proton transfer products of {P2W18}, {P2W17V} and {P2W15V3}, and the surface electrostatic potential of the three POMs-based redox pairs (The red area represents regions of high electrostatic potential, the blue area indicates regions of low electrostatic potential, and the green area denotes the transitional region). b-d The electron and proton transfer behaviors investigated via Marcus theory and the related pH changes during the electrolysis of {P2W18}, {P2W17V}, and {P2W15V3} solutions, respectively. The calculation data is provided in Supplementary Data 1.

Full size image

Above results demonstrated that the vanadium-substituted POMs exhibit advantages including suitable redox potentials, quick redox kinetics, multi-electron transfer property and strong pH buffer ability. Thus, a photocatalysis-electrolysis relay solar-to-hydrogen system was assembled (Fig. 5a and Supplementary Fig. 22), including a light-driven photocatalytic water oxidation reaction subsystem and a dark electrochemical hydrogen production subsystem. In the photocatalytic reaction, Ox-POM is reduced by photogenerated electrons to produce Re-POM[H], which undergoes an electrochemical reaction in the subsequent stage. This proton exchanging membrane (PEM) based electrolytic process involves the loss of electrons and release of H+ from Re-POM[H] at the graphite anode to complete the reversible transition to Ox-POM, while at the same time the free H+ rapidly couples with electrons at the cathode to form H2. The occurrence of the photocatalytic reaction and the extent to which it proceeds are prerequisites for the construction of the system.

Fig. 5: The demonstration and performance of photocatalysis-electrocatalysis relay water splitting.
figure 5

a Schematic diagram showing the role of POMs as redox mediators in the constructed photocatalysis-electrocatalysis relay system using BiVO4 photocatalyst. CB represents conduction band. VB represents valence band. ∆E represents potential difference between the redox medium and conduction band of semiconductor. Esolar represents storage of solar energy. PEM represents proton exchange membrane. Ox-POM represents oxidation states polyoxometalates. Re-POM[H] represents reducetion states polyoxometalates with coupled protons. b Investigation of photocatalytic O2 evolution and electrons transfer numbers of 1 mM {P2W17V} and {P2W15V3} in 1 M Na2SO4 solutions (pH = 3, 150 mL; BiVO4: 100 mg, stirring speed: 500 rpm) under 300 W Xe lamp with cut filter (λ ≧ 420 nm). c Long-term photocatalysis-electrocatalysis relay testing of 5 mM {P2W17V} and {P2W15V3} redox mediators in 1 M Na2SO4 solutions (pH = 3, 100 mL; BiVO4: 300 mg) and the related H2 collection over 12 hours (electrochemical cell resistance value: 125 ± 5 mΩ).

Full size image

Both redox mediators {P2W17V} and {P2W15V3} exhibit good solar energy storage properties in the presence of BiVO4, and the suitable reaction conditions were screened through a series of preliminary photocatalytic experiments (Supplementary Fig. 23). Figure 5b shows the time course of O2 evolution during the photocatalytic reaction stage under the corresponding conditions. The O2 evolution in the {P2W17V} solution reached equilibrium after 1 hour with a 1e- reduction, whereas in the {P2W15V3} solution, equilibrium was achieved after 4 hours, leading to an almost 3e- reduction under the same conditions, which is in good agreement with the 1-electron redox activity of {P2W17V} and 3-electron redox activity of {P2W15V3} as demonstrated above. Similarly, the initial oxidized state, 1e-, 2e-, and 3e- reduced states of {P2W15V3} were calibrated using UV-Vis spectroscopy, and quantitative analysis of the solution after the photocatalytic reaction confirmed that the reduction level of the redox mediator P2W15V3 reached 2.73e- (Supplementary Figs. 24, 25 and Supplementary Discussion 2). Further comparison of the initial activity reveals that {P2W17V} and {P2W15V3} exhibit higher reactivity compared to {P2W18}, indicating that the introduction of vanadium enhances the electron capture rate of the redox mediator during the photocatalytic water oxidation reaction (Supplementary Fig. 26). More importantly, compared to the single-electron transfer of {P2W18} and the relatively lower solar energy storage of {P2W17V}, {P2W15V3} demonstrates multi-electron storage characteristics and higher solar energy storage. Notably, no significant spontaneous oxidation was observed in these Re-POM solutions even after several months of storage, indicating long-term stability (Supplementary Fig. 27). To demonstrate the durability of the redox mediator {P2W15V3} for practical applications, we conducted ten cycles of photocatalytic reactions under simulated AM 1.5 G solar irradiation (100 mW cm−2). Except for minor sample loss during the transfer between cycles, there was no significant decrease in the gas production during the photocatalytic oxygen evolution process, indicating that the redox mediator exhibits good cycling stability (Supplementary Fig. 28). As shown in Fig. 5c, when using {P2W15V3} as redox mediator, a stable electrolysis current density of 1 mA cm−2 can be achieved at a low bias of 0.6 V. Its lower reduction potential allows it to convert more solar energy into hydrogen with minimal electrical energy input, making it more energy-efficient compared to conventional water electrolysis systems. The energy consumption for hydrogen production in this system was calculated, and it was found that the system operates at a current density of 10 mA cm−2 with an electricity consumption of only 1.58 kWh Nm−3 (H2), which is obviously lower than the theoretical limit of traditional water electrolysis (2.94 kWh Nm−3 (H2)) and other advanced water splitting electrolyzers (Supplementary Fig. 29). This voltage is over 10% less than that of {P2W17V} under the same current density, indicating that {P2W15V3} has a greater storage proportion of solar energy than {P2W17V}. Additionally, the theoretical and actual amounts of H2 produced in the {P2W17V} and {P2W15V3} solutions over 12 hours were continuously collected and recorded with Faradaic efficiencies exceeding 98.5 ± 0.3% (obtained from 3 independent measurements). Furthermore, the photocatalysis-electrolysis relay water splitting system using {P2W17V} and {P2W15V3} demonstrated good long-term operational stability. These outstanding performances can be attributed to the introduction of the transition metal V active center, which endows the redox mediator with multi-electron storage characteristics at suitable redox potentials. This facilitates more efficient storage and utilization of solar energy. The multi-electron storage characteristics of {P2W15V3} increase the charge capacity per unit volume by approximately three times, which is beneficial for improving the long-term efficiency and energy output of the system. Additionally, its unique pH buffering ability and high stability ensure the long-term stable operation of the system in practical applications. At the same time, this decoupled overall water splitting design prevents explosive gas mixtures and the formation of reactive oxygen species, leading to a purer hydrogen production process and effectively avoiding the need for purification53,54. Compared to advanced catalysts or systems capable of simultaneously achieving oxygen and hydrogen evolution (Supplementary Table 4), the decoupling mechanism of this system enables each half-reaction to proceed under optimized conditions independently, effectively reducing the occurrence of gas crossover and the electricity input to generate H2. It is noteworthy that the photocatalytic and electrocatalytic reaction processes can be divided into two independent stages as needed. Hydrogen energy can be stored in the form of protons in Re-POM[H] during the photocatalytic reaction and released as hydrogen gas through a rapid electrolysis process when needed (Supplementary Fig. 30). This makes Re-POM[H] a stable liquid hydrogen storage medium at ambient pressure.

In summary, based on the structural tuning of polyoxometalate clusters, a photocatalysis-electrolysis coupling system using POMs as redox mediators was established to realize solar-driven water splitting with decoupled H2 and O2 evolution in space. The substitution of V atoms into {P2W18} polyoxometalate leads to the transition of 1-electron redox active {P2W17V} with sequential proton-electron transfer behaviour into 3-electron redox active {P2W15V3} molecular cluster with concerted proton-electron transfer behaviour. The results demonstrate that adding an appropriate amount of V atoms to POMs can achieve more storage of photogenerated electrons and proton coupling, thereby achieving efficient utilization of solar energy. Using BiVO4 as a water oxidation photocatalyst, the POMs can achieve single or multiple reversible storage of protons and electrons during photocatalytic water oxidation under mild pH conditions. Finally, we built a photocatalysis-electrolysis coupling device, in which O2 and H2 are generated in an independent manner in space. The Faradaic efficiency of the electrocatalytic hydrogen evolution process is as high as 98.5% during stable operation for up to 12 hours, which means that the device is expected to achieve safe hydrogen production or storage in accordance with demand.

Methods

Materials

Sodium tungstate dihydrate (Na2WO4·2H2O, AR > 99.5%), Hydrochloric acid solution (HCl, AR, 36.0–38.0%), Phosphoric acid (H3PO4, AR > 85.0%), Potassium chloride (KCl, AR ≥ 99.5%), hydrogen peroxide (H2O2, AR, 30%), sulfuric acid (H2SO4, AR, 95.0–98.0%), nitric acid (HNO3, AR, 65.0–68.0%), Potassium bicarbonate (KHCO3, ACS, 99.7–100.5%), Sodium perchlorate monohydrate (NaClO4·H2O, AR ≥ 99.0%), Sodium chloride (NaCl, AR, ≥ 99.5%), Sodium hydroxide (NaOH, AR ≥ 96%), Sodium sulfate (Na2SO4, AR ≥ 99.0%)), Bismuth(III) nitrate pentahydrate (Bi(NO3)3·5H2O, AR ≥ 99.0%), Ammonium hydroxide aqueous solution (NH3·H2O, 25.0 - 28.0%) and Sodium carbonate (Na2CO3, AR ≥ 99.8%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Vanadium(V) trichloride oxide (VOCl3, 95%), Ammonium metavanadate (NH4VO3, AR, 99%) and Deuterium oxide (D2O, 99.9 atom % D) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Sodium metavanadate (NaVO3, 99%) was purchased from Energy Chemical. All chemicals were used without further purification, and ultrapure water (18.2 MΩ·cm−1 resistivity) was used in all experiments. All electrochemical components used in the experiments, including the glassy carbon electrode, graphite rod, and Ag/AgCl reference electrode (saturated KCl solution), were purchased from Tianjin Aida Hengsheng Technology Co., Ltd. (Tianjin, China). These electrodes were selected for their high quality and reliability to ensure the accuracy and reproducibility of the electrochemical measurements. The Nafion 212 membrane used in the experiments was purchased from DuPont (USA), with a standard thickness of 50 μm. The carbon paper, employed as the gas diffusion layer, was obtained from Shanghai Shengernuo Technology Co., Ltd. (China), with a thickness of 200 μm. The carbon felt, used as the electrode material, was sourced from Jiaxing Naco Technology Co., Ltd. (China), with a thickness of 2.5 mm. These materials were selected for their high quality and consistency to ensure the reliability of the experimental results. The membrane electrodes used in the article were Pt/C 0.5 g cm−2 and IrO2 0.5 g cm−2, both purchased from Amoy Island Hydrogen Technology. The size of the membrane electrode is 6 × 6 cm2 with a 2 × 2 cm2 catalyst spraying area in its center.

Synthesis of POMs

First, the yellowish crystal K6[ɑ-P2W18O62] (P2W18) was prepared using a high-yield (>90%) synthesis method with minimal by-products reported so far. Next, the deficient structures K10[ɑ-P2W17O61] (P2W17) and Na12[ɑ-P2W15O56] (P2W15) were obtained by adjusting the pH of the solution using KHCO3 and Na2CO3 solutions, respectively, with P2W18 as the raw material. After that, different molar ratios of NaVO3 were added to facilitate the assembly of structures with varying numbers of vanadium atom substitutions. Finally, K7P2W17VO62 (P2W17V) and K8HP2W15V3O62 (P2W15V3) were precipitated using KCl. All polyoxometalate crystals synthesized and used were purified by multiple recrystallizations. The synthesis procedures for the series of polyoxometalates are supported by the literature47,55,56,57,58,59.

Characterizations of POMs

The purified samples were characterised by thermogravimetric (TG) analysis, Fourier transform infrared (FT-IR), ultraviolet visible absorption spectroscopy (UV-Vis), NMR, and Raman spectrum. TG analysis was carried out by using the SDT Q600 analyser. Weigh 3-5 mg of POMs crystals and test under nitrogen atmosphere at a reaction temperature of 25 to 800 °C and a ramp rate of 10 °C/min. Fourier transform infrared spectrometer (Nicolet 380) and ultraviolet-visible spectrophotometer (UV-2100) were employed to determine the structure. FT-IR were measured using a Bruker Vertex 70 V spectrometer in an attenuated total reflectance setup under vacuum conditions. The POMs crystal particles were weighed and powdered in an agate mortar and subsequently weighed 3-5 mg as test samples. UV-Vis tested with V-780 UV-Visible Spectrophotometer. Measure 3.5 mL POMs aqueous solution (10 μmol L−1) into a quartz cuvette for testing. NMR spectra were collected by Bruker Avance III 500 MHz. 500 µL 25 mmol L−1 POMs solution was added into 50 µL D2O in the NMR tube and dispersed uniformly by ultrasonic. The Raman spectrum was collected with a confocal Raman microscope (Alpha-300, WITec) with an excitation wavelength of 532 nm. 2 drops of POMs solution (5 mmol L−1) was taken on a slide for testing.

Synthesis of BiVO4

First, 10.0 mmol of Bi(NO3)3·5H2O and 10.0 mmol of NH3VO3 are dissolved in 60 mL of 2.0 M HNO3 solution. Continuous stirring ensures complete dissolution of the precursors. Subsequently, ammonia solution (25–28 wt.%) is added dropwise under vigorous agitation to adjust the pH to 0.50. Stirring the mixture for 2 hours at room temperature yields light yellow BiVO4 slurries. These slurries are then transferred into 100 mL Teflon-lined stainless-steel autoclaves and subjected to hydrothermal treatment at 200 °C for 24 h. Following natural cooling to room temperature, yellow precipitates form in the autoclaves. Centrifugation separates the precipitates, which are then washed repeatedly with ultrapure water and dried at 80 °C overnight60.

Characterizations of BiVO4

To comprehensively characterize the properties of BiVO4, researchers employ multiple advanced analytical techniques. X-ray powder diffraction (XRD) analysis utilizes a Rigaku D/Max-2500/PC diffractometer with Cu-Kα radiation operating at 40 kV and 200 mA. The instrument records XRD patterns across a 10–60° range with a step size of 0.02° and a scan rate of 5°/min. A JASCO V-650 spectrophotometer equipped with an integrating sphere measures UV-visible (UV-vis) diffuse reflectance spectra. For Raman spectroscopy, a Renishaw spectrograph (2 cm−1 spectral resolution) applies a 532 nm solid-state laser as the excitation source, delivering approximately 0.3 mW power at the sample. Raman mapping adopts a 1 μm step size, with spectra acquired at room temperature. Additionally, a Quanta 200 FEG scanning electron microscope (SEM) analyzes sample morphologies and particle sizes. Mott-Schottky measurements use a 0.2 M Na2SO4 aqueous electrolyte under an AC potential of 1 kHz frequency and 0.01 V amplitude.

Calculation details

The first-principles computations were conducted by the Vienna ab initio simulation package (VASP) to determine the energies of the POMs and their electron-proton transfer products61. The projector augmented wave (PAW) was employed to model the electron-core interaction, and the Perdew-Burke-Ernzerhof (PBE) functional was utilized to determine the exchange and correlation energy within the generalized-gradient approximation (GGA). The cutoff energy applied for the plane-wave basis set is 450 eV. Electronic structures were self-consistently converged to energies within 1 × 10−5 eV and the convergence criterion of the frequency analysis was set to 1.0 × 10−7 eV. To describe the effect of static electricity and dispersion on the interaction between solute and solvent, VASPsol, which was integrated into VASP software, was used to perform the implicit solvent calculations and the dielectric constant of water was set to a relative permittivity of 78.462.

The relevant parameters of reorganization energy were calculated by Gaussian16 software and the polarizable continuum model (PCM) was applied to simulate the effect of the solvent (water)63. The single-point energies were calculated at the TPSS-D3/def2-TZVP level64,65. The analyses of LUMO, surface electrostatic potential, spin density and charge decomposition analysis were performed by Multiwfn software and visualized by VMD program66,67.

Marcus theory

Marcus theory can be employed to evaluate the rate of electron-proton transfer reactions, which can predict how kinetics are influenced by the reactants, solvents, electrode materials, and electrode adsorption layers68. Marcus theory postulates the following relationships among the activation energy (\(\Delta {G}^{\ne }\)), the change in Gibbs free energy (∆G), and the reorganization energy (λ, encompassing the energy needed for the conversion of reactants’ and solvents’ configurations into those of the products, composed of both internal reorganization energy (λ1) and solvent reorganization energy (λ2)):

$$\Delta {G}^{\ne }=\frac{\lambda }{4}{\left[1+\frac{\Delta G}{\lambda }\right]}^{2}$$
(1)

By the Computational Hydrogen Electrode (CHE) method and the Isodesmic Proton-Exchange Reaction (IPER) scheme with formic acid as the reference acid, we can analyze the thermodynamic property of proton transfer (PT) processes, electron transfer (ET) processes, and concerted proton-electron transfer (CPET) processes of POMs. By assessing the reorganization energy within the electron-proton transfer process, we can derive the corresponding activation energy, thereby offering intricate kinetic insights into the reaction.

For the internal reorganization energy, we employed the four-point method69. This method needs to obtain the energy in both the reactant state under the stable configuration of the product (EP[R]) and the product state under the stable configuration of the reactant (ER[P]). Additionally, it requires determining the energy of the reactant’s equilibrium configuration (ER[R]) and the energy of the product’s equilibrium configuration (EP[P]). The internal reorganization energy is defined as the follow70:

$${\lambda }_{1}=\frac{1}{2}(|{E}_{{{{\rm{P}}}}}\left[{{{\rm{R}}}}\right]-{E}_{{{{\rm{R}}}}}\left[{{{\rm{R}}}}\right]|+|{E}_{{{{\rm{R}}}}}\left[{{{\rm{P}}}}\right]-{E}_{{{{\rm{P}}}}}\left[{{{\rm{P}}}}\right]|)$$
(2)

Regarding the calculation of solvent reorganization energy, it necessitates consideration of both electron transfer (λ2,ET) and proton transfer (λ2,PT) scenarios. The specific calculation formulas are as follows:

$${\lambda }_{2,{ET}}=\,\frac{{e}_{0}^{2}}{4\pi {\varepsilon }_{0}}\left(\frac{1}{{\varepsilon }_{{{{\rm{op}}}}}}-\frac{1}{{\varepsilon }_{{{{\rm{s}}}}}}\right)\frac{1}{2a}$$
(3)
$${\lambda }_{2,{pT}}=\,\frac{1}{4\pi {\varepsilon }_{0}}\left(\frac{{\varepsilon }_{{{{\rm{s}}}}}-1}{{2\varepsilon }_{{{{\rm{s}}}}}+1}-\frac{{\varepsilon }_{{{{\rm{op}}}}}-1}{{2\varepsilon }_{{{{\rm{op}}}}}+1}\right)\frac{{({\mu }_{{{{\rm{R}}}}}-{\mu }_{{{{\rm{P}}}}})}^{2}}{{a}^{3}}$$
(4)

Here, ‘a’ represents the radius of the reactants. εop (2) and εs (78.4) denote the optical and static dielectric constants, respectively. μR and μP represent the dipole moments of the reactants and products.

The molecular orbitals (MOs) of POMs are seen as a linear combination of the MOs from various individual fragments, and quantifying the contribution of MOs from individual fragments to the MOs of POMs can facilitate our understanding of the redox behavior of POMs. Charge decomposition analysis (CDA) is a method based on fragment orbitals, which aims to decompose the charge transfer between molecular fragments into contributions from individual orbitals, and can be used to research how the LUMO of POMs were constructed from the MOs of fragments.

Electrochemical experiment

All electrochemical tests were performed using a Shanghai Chenhua electrochemical workstation (CHI760e). CV curves were tested in a customized four-neck electrolytic cell containing 5 mL of solution. A glassy carbon electrode (3 mm in diameter, electrode surface area: 7.065 cm2) was polished sequentially with 1 μm, 0.5 μm and 0.05 μm alumina slurry on a microcloth polishing pad to achieve a mirror-like surface, followed by thorough rinsing with deionized water and ethanol to remove any residual alumina particles. The polished glassy carbon electrode served as the working electrode, while a graphite rod electrode (5 mm in diameter) was used as the counter electrode. An Ag/AgCl (saturated KCl solution) reference electrode was employed, and its potential was calibrated using a standard reference electrode in a two-electrode configuration via open-circuit potential (OCPT) measurement to ensure accurate potential referencing. All potentials reported in this study are referenced to the calibrated Ag/AgCl electrode. The standard electrode potential of the Ag/AgCl electrode in saturated KCl is + 0.199 V (25 °C). The electrolyte solutions used in the experiments were freshly prepared prior to each test to ensure consistency and reliability. A 1 M H2SO4 solution was used to adjust the pH of the 1 M Na2SO4 solution to 3. The pH values of the electrolytes were precisely calibrated using a pH meter (model PHS-3E, purchased from Leici, Shanghai, China), which provided a measurement accuracy of ± 0.02 pH units. This rigorous pH calibration ensured the stability and reproducibility of the electrochemical measurements. Prior to testing, all solutions were degassed with Ar for 15 min. Rotating disk electrode (RDE) tests were conducted using a glassy carbon electrode with an active area of 0.196 cm2 as the working electrode, a graphite rod as the counter electrode. and an Ag/AgCl as the reference electrode at a scan rate of 50 mV s−1. The flow cell system was constructed for the galvanostatic charge-discharge test. The end plates on both sides of the flow cell are made of stainless steel and polytetrafluoroethylene, with a graphite bipolar plate in the middle. The internal resistance of the flow cell was measured to be 125 ± 5 mΩ, ensuring efficient charge transfer and minimal energy loss during the electrochemical processes. Tighten eight bolts (M6 × 60 mm fully threaded hex bolts) to a torque of 6 Nm to complete the assembly of the sample cell, and use polytetrafluoroethylene tubing to deliver the electrolyte into and out of the sample cell. A peristaltic pump circulates the liquid at a rate of 100 mL min−1 to minimize mass transfer overpotential. 25 mL of 5 mM POMs solution in 1 M Na2SO4 at pH = 3 was pumped and recycled between the cathode compartment of cell-a and the anode compartment of cell-b, while the anode compartment of cell-a and the cathode compartment of cell-b were filled with 1 M Na2SO4 (pH = 3). Linear Sweep Voltammetry (LSV) is tested by the two-electrode method at a scan rate of 10 mV s−1 (25 mL of 5 mM POMs solution in 1 M Na2SO4 at pH = 3).

Photocatalytic experiment

Photocatalytic water oxidation performance of BiVO4 photocatalysts was evaluated by both evacuation system (Beijing Perfectlight, Labsolar-6A) and Clark-type oxygen electrode. The photocatalytic O2 evolution reactions were carried out in a closed gas circulation and evacuation system using a 300-W Xe lamp (Ushio-CERMAX LX300) or simulated AM 1.5 G solar light (Class AAA Solar simulator, 100 mW cm−2) and optical cutoff filter (kenko, L42; 420 nm). The photocatalyst was dispersed in the POM solution in a Pyrex reaction cell and thoroughly degassed by evacuation in order to drive off the air inside. The amount of evolved O2 was determined by an on-line gas chromatograph (Shimadzu GC-8A, TCD, Ar carrier). The volume of the reaction solution is 150 mL, 1 mM {P2W17V} and {P2W15V3} are dissolved in 1 M Na2SO4 solution (pH = 3), and the mass of BiVO4 is 100 mg.

UV-Vis absorption spectrum

Firstly, a chronopotentiometry method was used to electrolyze a 1 mM P2W15V3 solution (pH = 3), injecting an equal number of electrons at proportional intervals to obtain P2W15V3 solutions with varying degrees of reduction. Next, UV-Vis spectroscopy was performed to collect absorption spectra of these P2W15V3 solutions in the range of 275–1000 nm, using a 1 cm quartz spectral cuvette (400 nm min−1, 200–1000 nm). After the photocatalytic reduction reaction, the P2W15V3 solution was removed, and the residual BiVO4 catalyst was filtered out to obtain a clarified reaction solution. Thus, the initial oxidized state, 1e-, 2e-, and 3e- reduced states of P2W15V3 were calibrated using UV-Vis spectroscopy. Under the same conditions, UV-Vis absorption spectra of the post-reaction solution were collected, and the absorbance at 479 nm were selected to construct a standard linear plot of absorbance versus the degree of reduction from 2e- to 3e-. The total number of reduction electrons in the solution was calculated using the standard curve.

Photocatalysis-electrolysis relay experiment

The construction of both sides of the electrolyzer is basically the same. Stainless steel end plates, PTFE plates, graphite bipolar plates, and fluid-collecting carbon felts were symmetrically arranged and separated by a Nafion membrane. The area of the bipolar plates was 6 × 6 cm2 (thickness 1 cm), the sinking plane was 2 × 2 cm2, and the internal flow-through was provided with a channel width of 0.16 cm, a channel depth of 0.2 cm, and an interchannel landing of 0.08 cm. The carbon papers were immersed in a nitric acid solution at 80 °C for 2 hours. The Nafion 212 membranes were pretreated with 5% H2O2 by weight at 80 °C for 30 minutes, then boiled with 1 M HNO3 at 80 °C for 30 minutes. The pretreated membranes were soaked overnight in 1 M H2SO4. The reaction region uses 300-W Xe lamp equipped with optical cutoff fer (λ > 420 nm, 1000 mW cm−2) as the light source, and shares the solution pool with the electrocatalytic device. POM and photocatalyst are homogeneously mixed in the photocatalytic reaction cell. The solution is transported and agitated by a peristaltic pump. A peristaltic pump circulates the liquid at a rate of 100 mL min−1 to minimize mass transfer overpotential. The hydrogen evolution reaction is coupled on the other side of the electrolytic cell, and the cell is connected with a gas collection device for gas quantitative analysis. The hydrogenolysis side was additionally equipped with a Nafion membrane, and filled with 1 M Na2SO4 (pH = 3). The volume of the reaction solution is 100 mL, 5 mM {P2W17V} and {P2W15V3} are dissolved in 1 M Na2SO4 solution (pH = 3), and the mass of BiVO4 is 300 mg. Electrolysis tests were carried out in a two-electrode cell with carbon paper electrodes (2 × 2 cm2) used as working electrodes on both sides. One side of the cell was circulating photocatalytic solution, while the hydrogen precipitation side was a 1 M Na2SO4 solution (pH = 3) equipped with a Pt-containing membrane electrode (Pt/C 0.5 g cm−2), and the generated H2 was real time collected.