Introduction

Photoelectrochemical (PEC) water oxidation to generate hydrogen peroxide (H2O2) as a clean and promising alternative to the traditional anthraquinone production process has received considerable attention1,2,3,4. Many suitable photoanodes, such as BiVO4, TiO2 and WO3, have been experimentally and theoretically explored for fulfilling the H2O2 evolution reaction3. Due to the thermodynamical inferiority of H2O2 production by two-electron (2e−) water oxidation reaction (WOR) (E0: 1.78 VRHE) relative to the O2 evolution by four-electron (4e−) WOR (E0: 1.23 VRHE)5, it is a vital issue to improve the reaction selectivity for H2O2 production on the photoanode surface. Among the strategies for enhancing the PEC reaction selectivity for H2O2, including exploring electrolytes6, coating surface layers7,8, defects regulation9 and modulating crystal facets10,11, the HCO3− anion in electrolytes is demonstrated to be indispensable for PEC H2O2 production from WOR6,12,13,14. However, the studies on the selectivity mechanisms and reaction dynamics of photoinduced holes lag behind the screening of catalyst materials and catalyst structure modulations.

PEC H2O2 evolution (2e− WOR) is first reported on FTO/WO3/BiVO4 photoanode with HCO3− as an effective electrolyte6. Since then, many researchers are devoted to revealing the reaction mechanism of PEC H2O2 evolution reaction. Several possible intermediates have been claimed during PEC H2O2 evolution reaction in HCO3− electrolyte with the characterization of surface species with spectroscopic techniques. On the one hand, hydroxyl radical (•OH) was detected and claimed to be an intermediate by electron paramagnetic resonance (EPR) on BiVO4 photoanode8 and fluorescence probe method on TiO2 photoanode15. This result is in line with the calculated potential-limiting step on corresponding anodes for 2e− and 4e− WOR16,17, in which the roles of HCO3− electrolyte are not discussed. On the other hand, the adsorbed species related to HCO3− and its derivatives were observed by in situ Raman, infrared and ex situ X-ray photoelectron spectroscopies on BiVO4 anode9,14. In the solution containing CO32− anion, the products of peroxymonocarbonate (HCO4−) and peroxycarbonate (C2O42−) were identified on Pt anode by in situ Raman and differential electrochemical mass spectroscopy18; and in the electrochemical H2O2 evolution reaction, the carbonate radical (CO3•−) species as intermediate was also claimed by EPR on FTO anode19. Furthermore, a clear dependence of H2O2 product on the local concentration of HCO3− at the ZnGa2O4 and FTO anodes surface was also observed by shear-force scanning electrochemical microscopy20. These HCO3− related species indicate the participation of HCO3− and CO32− in generating H2O2 from PEC water oxidation. However, without kinetic results, it is still impossible to confirm the participation way and kinetic roles of HCO3− in PEC H2O2 evolution (2e− WOR), due to the chemical equilibrium between H2O2 and HCO3−-related species21.

To date, HCO3−-related species/intermediates have been detected, but what are the kinetic roles of HCO3− for improved reaction selectivity of H2O2 by PEC water oxidation have not been well studied. In principle, the reaction selectivity is mainly determined by the relative components of 2e− WOR relative to that of 4e− WOR, in which charge dynamics and reaction kinetics should play crucial roles. Therefore, it is indispensable to understand the kinetics and its roles for PEC H2O2 evolution by time-resolved techniques, as reported in the study of the rate-determining step (RDS) and kinetics in oxygen evolution reaction (OER)11,22. However, these kinetic aspects are absent in studying the mechanisms of PEC H2O2 evolution reaction.

In this work, we studied the kinetic characteristics in PEC water oxidation on representative BiVO4 photoanodes, especially the roles of electrolyte anions. The (photo)electrochemical reactions were studied in three typical electrolytes, involving HCO3− to generate H2O2, phosphate buffer solution (PBS) for OER, and sulfite as a hole scavenger. The influence of electrolyte anions on charge dynamics and reaction kinetics is revealed. The first-order reaction kinetics in 2e− WOR and the same first hole-transfer kinetics as that in 4e− WOR illustrate that the HCO3− anion accelerates the consumption of the first-hole intermediates, resulting in that the improved selectivity of 2e− WOR and enhanced H2O2 production. The kinetic study advances the fundamental understanding of HCO3−-mediated water oxidation mechanism, and sheds light on mechanism study on reaction selectivity.

Results and discussion

PEC H2O2 evolution performance in HCO3 − electrolyte

The BiVO4 films were prepared by the metal-organic decomposition (MOD) method23. The BiVO4 films present nanoparticle aggregation (grain size ~200 nm), monoclinic crystal phase and direct band-gap (~2.6 eV) from atomic force microscope (AFM) image, X-ray diffusion (XRD) analysis and UV-vis absorption spectra (Fig. S1). To understand the underlying mechanisms of electrolytes on water oxidation products, the PEC performances were investigated in PBS, sodium bicarbonate solution (NaHCO3) and mixed solution of PBS and sodium sulfite (Na2SO3) by light/dark cyclic voltammograms (CV), chopped light linear sweep voltammetry (LSV), and chopped light open circuit potential (OCP), as shown in Fig. 1. The BiVO4 photoanode exhibits lower onset potential and higher photocurrent in NaHCO3 than that in PBS, regardless of illumination or dark. This variation trend is similar to the PEC performances in Na2SO3 relative to those in PBS. Moreover, the chopped LSV in both NaHCO3 and Na2SO3 in Fig. 1b display higher photocurrent without the transient spikes, which are obviously different from that in PBS with transient spikes. The improved PEC performances and disappearance of transient spikes in NaHCO3 demonstrate the fast hole transfer processes and reduced reverse recombination, similar to the function of the hole scavenger of Na2SO3 in photoelectrocatalysis24.

Fig. 1: PEC performances of BiVO4 photoanodes with AM 1.5 G illumination in different 0.5 M aqueous electrolytes.
figure 1

a CV in PBS (0.5 M), NaHCO3 (0.5 M) and Na2SO3 (0.5 M) solutions. Solid lines: illumination; dashed lines: dark; black arrow: negative shift of onset potential; scan rate: 50 mV/s. b Chopped light LSV in three solutions. c Calculated decay lifetimes of OCP. d Concentration of the H2O2 evolution under different applied potentials in PBS and NaHCO3 measured by UV-vis method with the cobalt-carbonate assay. The potential is not iR corrected.

Full size image

Then, the OCPs of BiVO4 photoanodes were investigated in the three electrolytes. In Fig. 1c, the OCP behaviors of BiVO4 in NaHCO3 are similar with that in PBS, with similar initial OCP decay lifetimes (tinitial: 20 ~ 40 ms) and identical ∆OCP (OCPdark − OCPlight ~ 150 mV) (Fig. S2 and Table S1). In contrast, there is a much slower initial OCP decay lifetime (~386 ms) and smaller ∆OCP ( ~ 90 mV) in Na2SO3. The fast OCP decay in NaHCO3 suggests that the reaction in NaHCO3 is quasi-reversible, which is different from the irreversible reaction of hole consumption in Na2SO3 (Figs. 1c and S2b). Combining the photocurrent and OCP results, the presence of HCO3− can change the reaction path of photoinduced holes, which is different from the reversible OER cycle in PBS and the irreversible sulfite oxidation in Na2SO3. According to the increased yield of H2O2 product in Fig. 1d and the decreased yield of O2 product in Fig. S2c, the changed reaction path can be assigned to H2O2 evolution reaction, in line with previous reports14,25. Moreover, the presence of an oxidation peak at ~2.57 VRHE in dark CV in NaHCO3 (Fig. 1a) also implies enhanced selectivity of H2O2 evolution reaction14. Thus, the changed reaction path for H2O2 formation should be the reason for the much higher PEC performances of BiVO4 in NaHCO3, which provides the basis for the mechanism investigation on PEC water oxidation to H2O2.

The first hole transfer processes revealed with transient absorption (TA) spectroscopy

Firstly, to understand the influence of electrolyte on the photoinduced charge dynamics in PEC H2O2 evolution from WOR, operando transient absorption (TA) spectra were carried out in three electrolytes (Fig. S3). Figure 2a compares the wavelength-dependent TA amplitudes at 10 μs. It’s found that the TA amplitudes in PBS and NaHCO3 are higher than those in Na2SO3, verifying the function of Na2SO3 as a hole scavenger to effectively capture photoinduced holes on the BiVO4 photoanode surface before 10 μs26. More importantly, BiVO4 in PBS and NaHCO3 display almost overlapped TA band shapes and magnitudes, suggesting the identical bulk transport and recombination in the pre-μs time scale.

Fig. 2: TA spectra of BiVO4 photoanode in three eletrolytes.
figure 2

Wavelength-dependent TA spectra at t = 10 μs (a) and t = 1 ms (b). c Kinetic traces and corresponding fitting curves at fixed probe wavelength (543.5 nm) and applied potential (1.4 VRHE). The potential is not iR corrected.

Full size image

Since water oxidation reaction usually occurs at timescale of ms−s27, the TA amplitudes at 1 ms in three electrolytes are compared in Fig. 2b. In PBS and NaHCO3, the TA spectra at 1 ms also show similar shape with a absorption peak at ~550 nm, while TA in Na2SO3 displays lower amplitude and different TA shape, further confirming that operando TA spectroscopy can monitor reaction species. Furthermore, at the same applied potential (1.4 VRHE) and probe wavelength (543.5 nm, highest TA amplitude), the kinetics in PBS and NaHCO3 display similar decay traces and lifetimes (~0.35 s) by fitting with power and exponential functions (Fig. 2c)28. Due to low hole density under irradiation of a nanosecond laser pulse, this TA species at 543.5 nm in PBS is previously assigned to a single-hole species corresponding to the first-order reaction27,29. Therefore, the identical TA results indicate that the first-hole transfer step or formation of one-hole intermediate during water oxidation on the BiVO4 surface may be the same in both the four-electron and the two-electron WOR processes.

The accumulated species under quasi-steady-state excitation with photoinduced absorption (PIA) spectroscopy

The apparent discrepancy in different PEC performances and similar TA results in HCO3− and PBS may originate from inconsistent charge density due to different illumination intensities. To realize quasi-steady-state excitation in PEC measurement for time-resolved characterization, we utilized photoinduced absorption (PIA) spectroscopy with a temporal resolution of ‘ms’ to study the quasi-steady-state kinetics (Fig. 3 and S4). Firstly, we compare the band shape of quasi-steady-state photohole absorption of the BiVO4 photoanode in three electrolytes. In Fig. 3a, the typical PIA spectrum peak at 543.5 nm in PBS is consistent with the above TA spectra in Fig. 2b, corresponding to the same first-hole intermediate for OER23. The PIA spectrum in NaHCO3 exhibits a broader peak relative to that in PBS, indicating a different reaction intermediate. More importantly, the quasi-steady-state PIA spectrum in NaHCO3 is also different from its TA spectrum at 1 ms (Fig. 2b). These results suggest that the accumulated species under quasi-steady-state should be the second-hole intermediate in 2e− WOR, which is different from the first-hole intermediate. When BiVO4 was measured in Na2SO3, the disappeared PIA spectrum implies the short-lived intermediates with sub-ms timescale in sulfite oxidation.

Fig. 3: Quasi-steady-state reaction kinetics of BiVO4 photoanode in different eletrolytes.
figure 3

a Wavelength-dependent PIA amplitudes on the equilibrium plateau in three electrolytes (illumination: 456 nm LED, 2.5 mW/cm2). b Compare the PIA amplitudes and photocurrents under illumination (456 nm LED, 4.82 mW/cm2) (PBS: 1.8 VRHE, NaHCO3: 1.8 VRHE, Na2SO3: 1.4 VRHE, the potential is not iR corrected.). c Relationship of quasi-steady-state PIA amplitudes at 543.5 nm against the light pulse intensity (PBS: 1.8 VRHE, NaHCO3: 1.8 VRHE, Na2SO3: 1.4 VRHE, the potential is not iR corrected.). The solid lines are only trend lines and are used to guide the eye. d Reaction rate versus surface hole density. Hole density of x-axis is obtained from the integrated charges in PMPS results (Figs. S8 and S11-S13), and photocurrent of y-axis is obtained from the steady-state photocurrent in PMPS results (Figs. S8 and S11–S13); half-solid symbol: AM 1.5 G (100 mW/cm2); blank symbol: 456 nm LED (2.5 mW/cm2); dash line: linear fit or turning point of first-order to three-order.

Full size image

Then, the PIA amplitudes in the three electrolytes are compared in Fig. 3a, b to evaluate the amount of the accumulated species. BiVO4 photoanode shows high absorption signal in PBS, medium absorption magnitude in NaHCO3 and almost zero absorption amplitude in Na2SO3. Moreover, the spectroelectrochemical absorption spectra with temporal resolution of ms−s also exhibit the same trend with that of PIA amplitudes in the three electrolytes (Fig. S5). The high and medium absorption magnitude in PBS and NaHCO3 indicate high and medium amount of the accumulated species, respectively, while there is almost no accumulated species in Na2SO3. The amount of the accumulated species (the absorption amplitudes) displays an inverse proportion to the photocurrent in the three electrolytes. This should be because of the different reaction rates of surface photoholes, rNa2SO3 > rNaHCO3 > rPBS. The much lower absorption amplitude in NaHCO3 relative to that in PBS should be due to the faster reaction kinetics, which can suppress accumulation of reaction intermediates in the potential region of (photo)electrocatalytic reaction.

The reaction kinetics identified by photo-multi-potential steps (PMPS) technique

To reveal the quasi-steady-state reaction kinetics, the light-intensity-dependent PIA spectroscopy were first used (Fig. S6), as usually applied in refs. 29,30 To avoid the interference of recombination under low potential, light-intensity-dependent PIA spectra were carried out at a fixed potential of 1.8 VRHE (Fig. 3c). The PIA amplitudes increase with the light intensity increasing in PBS and NaHCO3 (Fig. 3c). In Na2SO3, there is almost no PIA signal, resuling in the impossibility of the analysis of reaction kinetics. Then, the relation between photocurrent and photohole density is quantified by rate law analysis (J = k•[PIA]β, k is rate constant and β is reaction order) to reveal the RDS of chemical reaction (Fig. S6g). The reasonability of rate law analysis is based on that the illumination intensities only influence hole density at the valence band edge rather than hole energy, as illustrated by Mesa et al. 27 The reaction order of BiVO4 in NaHCO3 shows the second order, which is different from the third-order reaction kinetics in PBS. However, the second order is not fully reasonable due to the possibility of overlapped PIA signlas and the quite low signal-to-noise ratio of PIA in NaHCO3 (Fig. S6). Thus, an alternative technique is highly needed to obtain accurate photocurrent and hole density, and carry out the rate law analysis in NaHCO3 to obtain the reaction kinetics in 2e− WOR for H2O2 evolution.

Next, we used a photo-multi-potential steps (PMPS) technique to obtain quasi-steady-state photocurrent and surface photoinduced hole density31,32. The PMPS technique is based on the multi-potential steps (MPS) technique under light irradiation for PEC reactions (Supplementary Information Note 1 and Fig. S7), while MPS has been widely applied to collect and quantify the charge density in electrocatalysis31,33. The density of surface holes (Q) at a measured potential (photocurrent, J) can be quantified by integrating the recombination/back current with respect to low potential (≤ onset potential) (Fig. S7). With the PMPS results in PBS as an example, the reasonability of the PMPS technique is first confirmed experimentally (Figs. S8–S10). As shown in Fig. S10, the potential-dependent integrated charges (PMPS results in Fig. S8) exhibits a linear relationship with the potential-dependent hole density (PIA amplitudes in Fig. S9), demonstrating that the PMPS technique can obtain the same physical quantity of photoinduced holes as that in PIA. Thus, the reaction order can be analyzed by correlating the photocurrent with the surface hole density obtained from PMPS results under a series of applied potentials, in which the applied potentials only influence charge separation by increasing band bending of the space charge region rather than the position of the valence band edge22,34,35,36,37. The reaction orders (Fig. 3d) in PBS calculated from PMPS results are the first order at low hole density (low potential) and the third order at high hole density (high potential), in line with the reaction orders obtained from the above PIA results and demonstrating the reasonability of PMPS analysis. Moreover, the transition of reaction order under the excitation of AM 1.5 G: 100 mW/cm2 (Figs. 3d and S11) displays almost the same trend as that under the excitation of 456 nm LED: 2.5 mW/cm2, further confirming the reasonability of the PMPS technique.

Then, the reaction kinetics of the H2O2 evolution reaction in NaHCO3 was studied with the PMPS technique (Figs. S12 and S13). As shown in Fig. 3d, the linear correlation between the photocurrent with the surface hole density obtained from PMPS results demonstrates that the reaction order of BiVO4 in NaHCO3 is the first order (β = 0.92) under 456 nm LED excitation in the potential range from 0.4 to 2.0 VRHE (Fig. S12). To further confirm the first-order kinetics, the light sources of AM 1.5 G was applied in the PMPS experiments (Fig. S13). Amazingly, the reaction kinetics maintain at the first order in the whole studied range of hole density (10−3 − 102 h+ nm−2), while the transition of the reaction order for OER occurs at about 101 h+ nm−2 of the surface hole density from the first order to the third order. In the same range of hole density, the different reaction orders in PBS and NaHCO3 suggest that the kinetics of H2O2 evolution reaction is weakly dependent on the effect of light-induced valence band edge unpinning38, consistent with the electroneutral intermediates (such as, –OH, =O, –OOH) and chemical bond formation (O–O) as RDS, which are insensitive to the potential drop across the Helmholtz layer. The first order demonstrates that the RDS of H2O2 evolution on BiVO4 photoanode in NaHCO3 is the consumption processes of the single-hole intermediate. In combination with the improved PEC photocurrent and H2O2 evolution in NaHCO3 (Fig. S14), the reaction rate constants (k2e-WOR = J/Q ~ 1.8 s−1) and turnover frequency (TOF) (TOF2e-WOR = J/2Q ~ 0.90 s−1) for 2e− WOR are much larger than the reaction rate constants (k4e-WOR = J/Q ~ 0.060 s−1) and TOF (TOF4e-WOR = J/4Q ~ 0.015 s−1) for 4e− WOR in terms of J = k•[PIA]β, k is rate constant and β is the first-order reaction order. Therefore, the HCO3− anions considerably accelerate the consumption rates of the single-hole intermediate about 30-fold in rate constants or 60-fold in TOF.

When the reaction kinetics of sulfite oxidation reaction in Na2SO3 was also studied with PMPS technique (456 nm LED in Fig. S15 and AM 1.5 G in Fig. S16), the fast and irreversible reaction leads to high photocurrent but quite low integrated charges. Since the hole density in Na2SO3 solution is too low, the rate law analysis of sulfite oxidation reaction is no more discussed further.

To verify the university of the HCO3−-mediated water oxidation reaction kinetics, the reaction kinetics in NaHCO3 were further studied in other semiconductor photoanodes. As shown in Figs. S17–S21, the reaction kinetics on TiO2 and WO3 photoanodes in NaHCO3 solution exhibit the first order in the whole range of hole density, in good accordance with that on the BiVO4 photoanode. Therefore, the reaction order of the first order is universal for water oxidation in NaHCO3 solution to produce H2O2, where the HCO3− anions accelerate the consumption rates of the first-hole oxidation intermediates.

The accelarated charge transfer rates by the HCO3 − anions

To further illustrate the electrolyte-regulated water oxidation mechanisms, the charge transfer processes at the interface of BiVO4 and electrolytes were monitored by (photo)electrochemical impedance spectroscopy ((P)EIS) (Figs. S22–S24), in which the distribution of relaxation times (DRT) method is utilized to distinguish different processes39,40. By comparing potential-dependent DRT results at dark and illumination (Fig. S25), the apperance of fast time constants in three electrolytes is in good accordance with their onset potential of dark-/photo-current (Table S1), demonstrating that the time constants reflect the reaction processes directly. Specifically, we discuss the reaction kinetics with the time constants from DRT results under 1.2 VRHE potential in Fig. 4. In dark (Fig. 4a), BiVO4 in three electrolytes at an applied potential of 1.2 VRHE show similar and slow time constant (~5 s), in good accordance with the lacking of surface reactions. In Fig. 4b, BiVO4 in PBS under 1.2 VRHE displays two potential-dependent time constants: a slow time constant around 800 ms appearing at lower potential of around 0.6 VRHE, and a fast time constant around 100 ms appearing at the higher potential around 0.8 VRHE. These two time constants, corresponding to two charge transfer process, can be related to two of the four-step OER paths on BiVO4 in PBS. In terms of the transition of reaction order from the first to the third order in PBS, the slow and fast time constants may result from the first and third OER steps in the four-step OER paths. Importantly, in NaHCO3 and Na2SO3, there only appears one characteristic time constant (around ~tens of ms) under light irradiation at 1.2 VRHE (Fig. 4b), suggesting only one typical charge transfer process despite two-electron transfer of the overall reaction (2H2O + 2 h+ → H2O2 + 2H+; SO32− + 2 h+ + H2O → SO42−+ 2H+). This single time constant is in good accordance with the first-order reaction order for H2O2 production in NaHCO3, further indicating that the RDS of H2O2 evolution on the BiVO4 surface in NaHCO3 is the consumption processes of the single-hole intermediate. The time constant around 10 ms in NaHCO3 relative to 100 or 800 ms time constants in PBS demonstrate that the acceleration of charge transfer is about 10 to 100 fold.

Fig. 4: Electron transfer processes.
figure 4

DRT analysis of (P)EIS data under dark (a) and illumination (b) in three electrolytes (456 nm LED: 2.5 mW/cm2, applied potential: 1.2 VRHE, the potential is not iR corrected.).

Full size image

Comparing the time constant in NaHCO3 with those of PBS, the difference are quite obvious that only one time constant was obtained in NaHCO3 with no corresponding time constants for OER reaction. This is in good accordance with the first-order reaction kinetics in NaHCO3 in the studied hole density range. As shown in Fig. 3d, the reaction kinetics remain at the first order, even though the hole density in NaHCO3 is much larger than that in PBS when the reaction order is already the third order. In combination of the constant time constant and reaction kinetics in NaHCO3, it is quite reasonable that almost all the PEC reaction in NaHCO3 is the water oxidation to H2O2, and the low selectivity measured may be due to the decomposition of H2O2 to produce O2. Therefore, we can understand the reaction mechanism of water oxidation to H2O2 by PEC in NaHCO3.

Discussion on the kinetic origination of 2e− pathway in HCO3 −

According to the above PEC and spectroscopic results, the kinetic mechanisms in PEC water oxidation for H2O2 or O2 production are proposed as shown in Fig. 5, where HCO3− can facilitate selective H2O2 evolution with the first-order kinetics in a millisecond timescale. The first-order kinetics in H2O2 evolution is tentatively ascribed to the consumption process of the single-hole intermediate, which restricts selectivity and reaction rate. The same TA kinetics indicate that the first hole transfer step is the same in PBS and NaHCO3. The first hole transfer step can be ascribed to the proton-removing process of adsorbed H2O on the BiVO4 surface, as illustrated by theoretical calculation41,42,43. For the ideal surface of BiVO4, a single Bi cation with unsaturated coordination (Bi(3−δ)+) acts as the reaction site and can adsorb physically H2O molecular (Bi(3−δ)+…H2O)41,44. Therefore, the first hole transfer step is expressed as Eq. (1):

$${{{{\rm{Bi}}}}}^{\left(3-{{{\rm{\delta }}}}\right)+}\cdot \cdot \cdot {{{{\rm{H}}}}}_{2}{{{\rm{O}}}}+{{{{\rm{h}}}}}^{+}\to {{{{\rm{Bi}}}}}^{\left(4-{{{\rm{\delta }}}}\right)+}-{{{\rm{OH}}}}+{{{{\rm{H}}}}}^{+}$$
(1)
Fig. 5: Proposed mechanism.
figure 5

Water oxidation mechanisms on the BiVO4 surface in PBS (OER, right) or NaHCO3 (H2O2 evolution reaction, left). The lifetimes are the reciprocal of TOF values in Fig. S14.

Full size image

Following the formation of the single-hole species, the second hole transfer step plays a crucial role in determining the reaction selectivity of WOR. In PBS, the single-hole species display as rate-determining species at low hole density, resulting in the first-order reaction kinetics, whereas at high hole density the single-hole species receive further two holes to reach three-hole rate-determining species (Bi(4−δ)+–OOH) (the right part in Fig. 5)45. In NaHCO3, the consumption of the single-hole species (Bi(4−δ)+–OH) is regarded as RDS based on the first-order reaction kinetics. Comparing with the rate constant of the first-order kinetics in 2e− and 4e− water oxidation, there is an increasement of ~30-fold. Therefore, the improvement of the consumption rate of single-hole species in HCO3− is the intrinsic reason of achieving 100 % H2O2 evolution by water oxidation reaction.

Then, we tentatively discuss how the HCO3− anions accelerate the second hole transfer kinetics. In general, formation and rupture of a chemical bond is a slow process, like O-O bond formation in OER22,31. Because of the kinetically unfavorable coupling for two adjacent -OH groups44, the single Bi(4−δ)+–OH species can only interact with surrounding H2O molecule or HCO3− anions. Due to the applied positive potential in PEC WOR, BiVO4 surface is prone to attracting species with negative charges. Namely, HCO3− can be attracted and bond with Bi(4−δ)+–OH species. The adsorption of HCO3− has been confirmed by attenuated total reflection Fourier transform infrared and X-ray photoelectron spectroscopies on BiVO4 anode14. Thus, the 2e− WOR intermediates observed in PIA will form due to the adsorption of HCO3−, followed structure reorganization to form O-O bond, which is a chemical step and assigned to RDS (the possible species is shown in the left part of Fig. 5). The function of HCO3− anions may act as a promoter (catalyst) or reactant after the first hole transfer (the simplified charge transfer processes is shown in Fig. S26). Additionally, it is also possible that the second hole transfer leads to HCO4− evolution, along with deprotonation and returning to initial state (Bi(3−δ)+) for reaction sites. Then the HCO4− species reacts fast with H2O to produce H2O219,21. However, due to the limitation of TA and PIA spectroscopy, the identification of specific species is beyond the scope of this study.

Based on the above kinetics analysis, our study indicates that H2O2 evolution reaction in PEC water oxidation are catalytic reactions on catalyst surface, which are sequential reaction mechanisms of the first oxidation reaction of H2O molecular followed by the second oxidation reaction with HCO3– anion. This mechanism is totally different from the mechanism with outer-sphere electron transfer in CO2/carbonate-mediated H2O2 evolution reaction proposed by Fan et al.19. The difference may be due to the different bias used in PEC and electrochemical H2O2 evolution. In the electrocatalytic H2O2 evolution reaction, the high bias (larger than 2.0 VRHE) could be the driving force resulting in the outer-sphere electron transfer without reaction intermediates absorbed on the catalyst surface. Oppositely, in PEC H2O2 evolution reaction, the lower bias (lower than 2.0 VRHE) should drive the catalytic reaction with surface-adsorbed intermediates as the proposed mechanisms in Fig. 5. This mechanism of inner-sphere electron transfer is also in accord with the surface-property-sensitive H2O2 evolution efficiency in literatures8,16. It should be noted that the absorbed HCO4− is used to represent the possible reaction intermediates on the catalyst surface in the second step, which is mainly used to emphasize the participation of HCO3– anion in the second step. It is different from the generally mentioned free HCO4− in solution, where the free HCO4− may always exist due to the chemical equilibrium of HCO3− + H2O2 ⇌ HCO4− + H2O in bicarbonate-H2O2 solution.

This work provides inspirations about the RDS and the role of HCO3− anions during H2O2 evolution by PEC water oxidation. During water oxidation, HCO3− anions do not directly act on the first hole transfer step, but affect subsequent steps, including O-O bond formation, structure reorganization and the second hole transfer. Moreover, low concentration of NaHCO3 (0.5 M) have a great effect on reaction paths (selectivity) and reaction kinetics of photoinduced holes, highlighting the importance of interface regulation. From the view of reaction kinetics, the observed fast first-order kinetics in NaHCO3 solution suggests that the Faradaic efficiency of H2O2 evolution by water oxidation in bicarbonate-containing solution should be close to 100 %, inconsistent with the low product selectivity in practical measurement. The kinetics analysis implies that it may be not a right direction to work on the improvement of the H2O2 selectivity in PEC water oxidation to increase the production of H2O2. The key point should be how to prevent further oxidation of generated H2O2 in the PEC water oxidation system. This issue can be well supported by the general strategy of surface modifications for H2O2 production. That is, the surface modification by molybdenum doping7, phosphate group7 and SnOx8 can retard the hole-mediated H2O2 decomposition and improve H2O2 production (a detailed literature review about the functions of dopant or surface modification layer in promoting PEC H2O2 evolution reaction listed in Table S2). Moreover, the stabilizer like Na2SiO3 can be a choice to improving concentration of PEC H2O2 evolution46. On the other hand, the kinetic characteristics of PEC-generated H2O2 determines its possible application fields. Due to the fast generation kinetics of H2O2 under PEC conditions and its subsequent decomposition to O2, there are two main strategies for utilizing the produced H2O2. One strategy involves the on-site utilization of H2O2 to synthesize valuable organic compounds before it decomposes47,48,49, by coupling it with other fast reactions. The other strategy is to directly use the obtained low-concentration H2O2 for applications such as wastewater treatment, disinfection, and pulp bleaching, without long-distance transport and long-term storage.

In summary, charge dynamics and reaction kinetics in PEC H2O2 evolution on BiVO4 photoanode were studied to understand the determining roles of HCO3− from kinetic aspects. We have found that introducing HCO3− does not change the first-hole transfer process in PEC H2O2 (2e− WOR) in comparing with that in O2 (4e− WOR) evolution reaction, but considerably accelerate the second-hole transfer dynamics, resulting in different reaction kinetics. In PEC H2O2 evolution reaction in 2e− WOR, the consumption process of the first-hole intermediates is the RDS, exhibiting the first-order reaction kinetics. More importantly, HCO3− facilitate the second-hole transfer process (TOF2e-WOR: ~1 s−1), and suppress the accumulation of multi-hole intermediates in 4e− OER (TOF4e-WOR: 0.01 ~ 1 s−1), resulting in the improved selectivity for H2O2 from water oxidation. This work contributes to advancing our fundamental understanding of electrolyte-regulated H2O2 evolution reaction in photo(electro)chemical water oxidation, and inspires us to take advantage of reaction kinetics to control reaction selectivity.

Methods

Materials

Chemicals during synthesis and measurement were used without further purification.

Fabrication of BiVO4 photoanodes

BiVO4 photoanodes were fabricated by a modified metal-organic decomposition (MOD) method23. In detail, Bi(NO3)3•5H2O (0.4 mol·L−1, Alfa Aesar, 98 %) was dissolved in acetic acid (1.5 mL, Sinopharm, 99.5 %), and vanadiumoxy acetylacetonate (0.03 mol·L−1, ACROS, 99 %) was dissolved in acetylacetone (20 mL, Sinopharm, 99 %). Two solutions were stirred at 90 °C for 1 h, respectively. After dissolution, two solutions were mixed following further stirring at 50 °C for 1 h. Besides, FTO glass substrates (2 × 2 cm, Nippon Sheet Glass, 14 Ω·sq−1) were cleaned for 30 min by sonication in acetone, isopropanol, ethanol and deionized water, respectively. BiVO4 films were prepared by decomposing the spin-coating precursors (1000 rpm for 20 s for each layer). After each layer was deposited by spin-coating, the film was calcined at 450 °C for 10 min. This spin-coating/calcination procedure was repeated 8 times. After depositing the 8th layer, the coated FTO was calcined at 450 °C for 10 h to ensure formation of the monoclinic-scheelite structure.

Fabrication of TiO2 photoanodes

TiO2 photoanodes were prepared by a hydrothermal method50. Briefly, 25 mL of 37% hydrochloric acid (Sinopharm, 37%) was added in 30 mL of deionized water and stirred for 5 min. Subsequently, 1 mL of titanium tetraisopropanolate (Sinopharm, 99%) was added and vigorously stirred for 1 h. The solution was transferred into a 100 mL Teflon-lined stainless autoclave. The cleaned FTO substrate was immersed into the solution. The autoclave was then put into an oven and kept at 220 °C for 20 min. After naturally cooling down to ambient conditions, the film was removed from the autoclave and rinsed thoroughly. Lastly, the rutile TiO2 film was annealed at 400 °C for 1 h in a muffle furnace to remove the residual organics.

Fabrication of WO3 photoanodes

WO3 photoanodes were prepared through a chemical bath decomposition (CBD) method51. Briefly, 0.4 g Na2WO4·2H2O (Innochem, 98 %) and 0.15 g (NH4)2C2O4 (Innochem, 99 %) were dissolved in 33 mL deionized water, followed by adding 9 mL HCl with vigorous magnetic stirring to form a yellow suspension. After dropping 8 mL H2O2 and 30 mL ethanol for about 20 min stirring to form a transparent solution, the cleaned FTO glasses were immersed in the precursor solution, and the deposition process was carried out at 90 °C for 150 min. Then, the film was taken out, drying at ambient atmosphere and annealing in air at 600 °C for 2 h followed by annealing at 800 °C for 15 min, removing the crystalline water and filling potential oxygen vacancies.

(Photo)electrochemical measurements

(Photo)electrochemical properties were measured in a quartz cell (5 × 5 × 5 cm, 30 mL electrolyte) using a standard three-electrode configuration, including a Pt plate (1 × 2 cm) counter electrode, a reference electrode (Ag/AgCl/saturated KCl, correction by the other standard Ag/AgCl/saturated KCl electrode) and the BiVO4 work electrode (BiVO4 films were coated with epoxy remaining ~0.6 cm2 area) (photos are shown in Fig. S27). Phosphate buffer solution (PBS, 0.5 M K2HPO4 and KH2PO4, pH = 8.32 ± 0.02, Sinopharm, 99%) and NaHCO3 solution (0.5 M, pH = 8.27 ± 0.01, Innochem, 99%) were employed to generate oxygen gas and H2O2, respectively. Na2SO3 (0.5 M, pH = 8.32 ± 0.02, Innochem, 98%) was introduced into PBS (0.5 M) to act as hole sacrifice reagent. All electrolytes were freshly prepared prior to use and stored in sealed glass bottles. Applied potentials measured versus the Ag/AgCl reference electrode were converted to the reversible hydrogen electrode (RHE, the potential of H+/H2 is set as 0 V) using the Nernst equation: ERHE (V) = EAg/AgCl (V) + 0.0591·pH + E0Ag/AgCl23. Where ERHE is the applied potential vs. RHE; EAg/AgCl is the applied potential vs. Ag/AgCl, and E0Ag/AgCl is the standard potential of the Ag/AgCl reference electrode versus RHE (0.1981 V vs. RHE). All (photo)electrochemical measurements were performed at room temperature (~22 °C) using a CHI 760 C potentiostat controlled by its proprietary software.

H2O2 production and measurement

The production and accumulation of H2O2 were measured by chronoamperometry under irradiation of AM 1.5 G and in a quartz cell. To avoid cathodic reduction and further oxidation, the chronoamperometry was performed with 180 s and without stirring. The concentration of H2O2 was quantified by the UV−vis method with the cobalt−carbonate assay14. An aliquot of the photoelectrochemically produced H2O2 (2.0 mL) was added to a 2 M solution of KHCO3 and 1 mM CoSO4 (2.0 mL, Sinopharm, 98 %). After reacting for 30 min, the absorbance at 257 nm (200 − 400 nm) was collected on a spectrophotometer (Jasco V-650, Japan) with a scan rate of 400 nm/min. Although the use of cobalt−carbonate assays might overestimate the amount of H2O2 product due to the chemical equilibrium of HCO3− + H2O2 ⇌ HCO4− + H2O in bicarbonate-H2O2 solution21, it is reasonable using this method to verify that bicarbonate can influence water oxidation product from four-electron O2 to two-electron H2O2. The concentration of H2O2 product was measured only once and displayed in Fig. 1d.

O2 production and measurement

The O2 product was measured in a sealed PEC cell by a gas chromatograph (Shimadzu GC-8A) equipped with a thermal conductivity detector. The cumulative amounts of O2 evolved from the PEC cell were analyzed. The Faradic efficiency (FE) of O2 product was calculated from the equation: FE = 4 × the amounts of O2 / the amounts of passed charges × 100 %.

PMPS technique and its feasibility analysis

In the PMPS measurements, three potential steps were applied by the CHI 760 C potentiostat. The potential was kept at a low potential (depending on the onset potential of photocurrent, Elow are 0.3, 0.3 and 0.2 VRHE for PBS, NaHCO3 and Na2SO3, respectively) for 15 s, then switched and kept at a higher potential (Ehigh ≥ Elow) for 15 s before returning to the initial Elow for 15 s. The sampling of current was every 0.002 s for the acquisition time. The current was obtained at the end of the 15 s higher potential segment. The charge accumulated at the applied potential step was calculated by integrating the current pulse at the second low potential. By changing the applied high potential, the potential-dependent photo/dark-current and photo/dark-charge can be obtained. The difference of current/charge under illumination and dark is the absolutely photoinduced current/charge, which can be used for rate law analysis. The feasibility of rate law analysis using the PMPS technique is as follows.

Rate law analysis in traditional catalysis is the relation of reaction rate (r) and species concentration (n). In photocatalysis, photo(electro)catalysis and electrocatalysis, the electron/hole can be regarded as a reactant to participate in chemical reaction, and is also analyzed by rate law analysis27,31,33. The MPS technique has been widely applied to collect and quantify the charge density in electrocatalysis31,33. When the MPS technique is adopted to collect charge, there is an implicit assumption that holes induced by high potential (Ehigh) can be fast consumed by reverse electron transfer after turning low potential (Elow). For surface OER reaction, it is consistent between the temporal resolution of electrochemical workstation (~ms) and the timescale of O–O coupling (ms−s)22. In addition, reverse electron transfer from back contact to catalyst surface is a faster timescale (<ms)28,37. Moreover, lower hole density will participate in water oxidation by a slower path (single-hole RDS) than that with higher hole density (three-hole RDS). Therefore, the current curve after turning low potential (Elow) will have a trail with slower decay. Based on the above analysis of temporal resolution, the MPS technique is a powerful technique to quantify the charge density in electrocatalysis for OER and to be applied in rate law analysis.

Then, the PMPS technique was developed based on the MPS measurement. Under light irradiation, the difference of light and dark charge density should be extracted to exclude the effects of band bending in order to obtain a precise photoinduced hole density, since external potential is usually used to change band bending of space charge region. The dark charge density can be obtained by the MPS technique, while the light charge density can be measured with the MPS measurement under light irradiation. Finally, the photoinduced hole density can be obtained from the difference of light charge density and dark charge density.

Next, we analyze the reasonability of the measurement of the light charge density for the PEC reaction to demonstrate the ability of the PMPS technique to obtain the photoinduced hole density. In order to obtain accurate photo-/dark current in PEC, the high potential (Ehigh) need to be kept for several seconds resulting in steady-state photo-/dark current with an equilibrium of surface reaction, recombination and bulk transport. Taking the surface section as an illustration, although surface reaction, surface trap and surface recombination are concomitant, surface reaction is the only contribution to steady-state photo-/dark current. Surface trap and surface recombination are interior processes, in which charges are temporarily separated but distributed in the photoanode rather than the external circuit. When the high potential is turned to low potential, the separation will be destroyed, and the trap-related recombination process will not contribute to the current. Thus, the obtained light charge density from the current curve after turning on low potential (Elow) under light irradiation quantify the total charge density contributing to the photocurrent. The difference of light and dark charge density results from only the photoinduced hole density for PEC reactions, and it can be applied in rate law analysis. To clearly compare the microscopic parameters, the charge density (\(Q:\frac{C}{{{cm}}^{2}}\)) is converted to number density \(\left(\frac{Q}{e}:\frac{C}{{{cm}}^{2}}\cdot \frac{1}{1.6\cdot {10}^{-19}C}\right):\frac{C}{{{cm}}^{2}}\cdot \frac{1}{1.6\cdot {10}^{-19}C}=\frac{1}{{\left({10}^{7}\cdot {nm}\right)}^{2}\cdot 1.6\cdot {10}^{-19}C}=\frac{{10}^{5}}{1.6}{{nm}}^{-2}\).

(P)EIS and DRT analysis

(P)EIS of BiVO4 photoanode at different electrolytes was carried out using an electrochemical workstation (VersaSTAT MC) from 105 Hz to 10−1 Hz with a signal amplitude of 10 mV. The resistances of three electrolytes were obtained by extracting the resistance at high frequency region. The DRT analysis was performed using an open-source MATLAB code40. DRT can extract the time characteristics of a (photo)electrochemical system from (P)EIS measurements by a sum of infinitely series of parallel resistors and capacitors40. More importantly, DRT can avoid artificial choice for equivalent circuits.

TA spectroscopy

TA spectra were measured by home-built spectroscopic device34. In detail, the TA spectra were measured using a Nd: YAG laser system (EKSPLA NT 342B) with third harmonic generation (355 nm, 5 ns pulse duration, 0.9 Hz repetition rate) as the excitation source. The laser pulses were delivered to the sample via a liquid light guide, providing an incident pump energy of approximately 230 μJ at 355 nm. For probe light generation, a 100 W tungsten-halogen lamp (Bentham, IL 1) was coupled to a monochromator (Zolix Omni-λ 300). Optical density changes (ΔOD) were determined by detecting the transmitted light with a Si photodiode (Hamamatsu), followed by signal amplification and recording using both an oscilloscope (Tektronix, TDS 2012C) and a data acquisition card (National Instruments NI-6221). Each data point represents an average of approximately 300 laser shots. Operando TA experiments employed a three-electrode configuration controlled by a CHI 760 C potentiostat. The electrochemical cell contained either 0.5 M PBS (pH 8.32), 0.5 M NaHCO3 (pH 8.27), or a mixed solution of 0.5 M PBS with 0.5 M Na2SO3 (pH 8.33), using BiVO4 film as the photoanode, Pt as the counter electrode, and Ag/AgCl as the reference electrode. During measurements, chronoamperometry maintained a constant applied potential throughout the TA spectral acquisition.

PIA spectroscopy

Long-pulsed (≈ s) LED light source in place of nanosecond laser was used for generating photoinduced charges in the photoelectrode. The interval of light on/light off was controlled by a wavefunction generator (0.09 Hz and 25 % duty ratio), which simultaneously sent a trigger signal to oscilloscope (Tektronics TDS 2012c) and DAQ card (National Instruments NI USB-6211). For collecting PIA spectra of BiVO4 in three electrolytes, the illumination duration was ~ 2.5 s, followed by ~ 7.5 s in the dark. The data were averaged about 10 light pulse to avoid bubble. The PIA spectra were carried out by a three-electrode setup controlled by a CHI 760C potentiostat in the above three electrolytes. During collecting PIA spectra, a constant potential was controlled by chronoamperometry.

Spectroelectrochemical absorption spectroscopy

Combination of UV-vis absorption spectrometer and electrochemical workstation is to in-situ monitor the absorption change of BiVO4 film in three electrolytes, known as spectroelectrochemical absorption spectroscopy. The spectrum was collected after the electrode was stabilized ~ 20 s under constant bias using chronoamperometry (a CHI 760 C potentiostat). By changing applied potential, a series of spectra can be obtained.