## Introduction

Advanced oxidation processes (AOPs) are a recognized approach that can effectively treat various organic pollutions in aqueous solutions because the generated reactive oxygen species (ROSs) have strong oxidation capabilities for mineralizing the organic molecules1. Fenton reaction is one of the most representative AOPs which can rapidly generate hydroxyl radicals (·OH) with high concentration, therefore being widely adopted in wastewater treatment engineering2. However, the Fenton reaction is an irreversible process that needs the continuous supply of various chemical reagents, such as H2O2, iron salt, acid/alkali et al., which increases the costs of the whole wastewater treatment process by more than one time3. Heterogeneous photocatalysis over semiconductors can generate a range of ROSs by photoinduced redox processes, which are regarded as green and sustainable AOPs with affordable cost4. In the past several decades, numerous efforts have been devoted to photocatalytic degradation of organic pollution by developing or designing various photocatalysts and photocatalytic systems1,5. However, directly generating ROSs by photoinduced redox processes in aqueous solutions are profoundly restricted by rapid recombination of photoinduced electron/hole pairs, and high thermodynamic barriers of radical generation (e.g., ·O2 at −0.33 V and ·OH at 1.99 V vs. normal hydrogen electrode, NHE), while combined photocatalysis with Fenton reaction, so-called photo-Fenton reaction, also requires the supply of H2O26,7. Therefore, there is a great necessity to find indeed efficient, low-cost, and sustainable AOPs.

In recent years, it is found that bicarbonate (HCO3) can not only react with H2O2 to form HCO4 with high reactivity but also act as an auxiliary for water-oxidative H2O2 evolution under photoelectrochemical (PEC) condition, through a reversible reaction as follows8,9:

$${{{{{{{\rm{HCO}}}}}}}_{3}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\leftrightarrow {{{{{{{\rm{HCO}}}}}}}_{4}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(1)

The above reversible reaction offers a possibility of an in situ generation and decomposition of H2O2 in an aqueous solution containing HCO3 under solar light illumination10,11,12. In addition, the co-existence of HCO4 and H2O2 was reported to facilitate the formation of various ROSs (·O2, ·OH, CO3•− et al.) by Co, Cu or Mn ion activation13,14,15. The correlation between HCO3, H2O2, HCO4 and free radical species inspire us to design a self-cycled AOPs system with high efficiency and affordable cost16,17,18.

Herein, we propose a self-cycled Fenton-like system based on the deliberately designed artificial leaf, realizing the sustainable reaction system for wastewater treatment (Fig. 1a). First, the solar-motivated unassisted artificial leaf comprised of a SnO2-x/BiVO4/WO3 photoelectrode and a poly tetrafluoroethylene (PTFE)-modified Mo single-atom catalysts/mildly reduced graphene oxide-coated gas diffusion electrode (PTFE@Mo-SACs/mrG-GDE) cathode fulfills the efficient H2O2 production in a bicarbonate-containing electrolyte with a production rate of 0.77 μmol/(min cm2) under AM 1.5 G illumination, corresponding to a bias-free solar-to-hydrogen-peroxide efficiency (SHyE) of 1.46%. Second, the generated H2O2 can be immediately in situ activated into main ·OH, ·O2 and 1O2 et al. via the catalysis of Mn(II) species in the bicarbonate electrolyte, and the Mn(II) species are correspondingly oxidized to high-valent Mn(IV) species. Third, the Mn species can be recycled by reducing the Mn(IV) species into Mn(II) in the cathode (Fig. 1a). As a result, the self-cycling process with requirements for only water, oxygen, and sunlight demonstrates long-term stability for more than one month for the removal of various organic pollutants.

## Results

To demonstrate the self-cycled AOPs, the formation of ROSs and redox coupling of Mn(II)/Mn(IV) are first evidenced. As shown in Supplementary Fig. 1, the electron paramagnetic resonance (EPR) spectra of the co-existence of HCO3, H2O2, and Mn(II) demonstrate the remarkably enhanced signal intensities of ·OH, ·O2 and 1O2 compared to those of the HCO3/H2O2 or H2O2/Mn(II), indicating that multi-species of ROSs can be activated by the self-cycled AOPs system19. Then, both cyclic voltammetry (CV) tests and EPR spectroscopy were used to track the Mn species. As shown in Supplementary Fig. 2, electrochemical redox peaks relative to Mn(II)/Mn(III) and Mn(III)/Mn(IV) appear in the potential window between 0.98 and 1.58 V vs. reversible hydrogen electrode (RHE), the narrower potential window would be easily met by photoinduced charge. Further insight into the valence alternation of Mn species during the redox is revealed by EPR measurement. As shown in Fig. 1b, a typical Mn(II) signal with a signature six-line pattern centered around g~2 is observed in the initial 2 min (curve i)15. However, both the symmetry and amplitude of this signal gradually decreased in 10 min (curve ii), ascribing to the Mn(II) being oxidized when the H2O2 is in situ generated. Accordingly, a new signal around g~4 corresponding to Mn(IV) species is observed, confirming the Mn(II) being oxidized to Mn(IV)20. Furthermore, the Mn(II) signals are barely observed after the light off for 10 min (curve iii) due to its consumption during the activation of H2O2 as well as the blocking of the regeneration of Mn(II) caused by the suspending of counterpart reduction reaction in cathode21,22. Significantly, the oxidation/reduction quasi-equilibrium reappears after reopening the light (curve iv), indicting the regeneration of Mn(II) from the Mn(IV) reduction under the photoactivated cathode20. As a result, it is unambiguously confirmed that the Mn species can be recycled as a “redox shuttle” under the solar-activated self-cycled system. In order to further evaluate the vital role of the self-cycled system, control experiments were conducted in the H-cell that separates the cathode and Mn species. As shown in Supplementary Fig. 3a, under continuous light illumination, the signal of Mn(IV) only appears in the H-cell, while a stable quasi-equilibrium of Mn(II)/Mn(IV) in the one-compartment opened reactor (Supplementary Fig. 3b).

To realize the on-site H2O2 production, the PEC water-oxidative H2O2 evolution and electrochemical oxygen-reductive H2O2 reaction are simultaneously employed. According to our previous reports, a SnO2-x-coated BiVO4 (SnO2-x/BiVO4) photoanode can kinetically favor the 2e water oxidation reaction (2eWOR) for H2O2 evolution accompanied by slight water-oxidative ·OH generation by 1eWOR in a NaHCO3 electrolyte because the SnO2-x overlayer tunes the solid/liquid energetics23, and Mo-SACs/mrG, having the low thermodynamic barrier of 2e oxygen reduction reaction (2eORR), can present high activity and selectivity for electrochemical H2O2 generation in a NaHCO3 electrolyte24. Here, to drive the 2eORR by solar light only, a SnO2-x/BiVO4/WO3 photoanode is designed since the BiVO4/WO3 heterojunction is well-demonstrated to lower onset potential and increase photocurrent density25. Figure 2 shows a schematic diagram of the solar-driven H2O2 production device using a SnO2-x/BiVO4/WO3 photoanode and a Mo-SACs/mrG cathode. By connecting the water-oxidative generation of H2O2 at the photoanode with the O2-reductive generation of H2O2 at the cathode, the photovoltage and photocurrent of the coupled photochemical reaction are constrained to be equal at the thermodynamic level. The redox potential of the BiVO4/WO3 photoanode electrode is 0.34 V vs. RHE and 2.45 V vs. RHE26, which straddle the oxidation potential of the 2eWOR (1.77 V vs. RHE) and the reduction potential of the 2eORR (0.70 V vs. RHE), which means that charge carriers excited by solar light can have enough energy to activate the overall reaction27. The operating point for such a coupled photochemical reaction can be established at the intersection point of the independently measured J–V curves. As shown in Supplementary Fig. 4, the point of intersection corresponds to the operating current density, JOP. Considering the possible nonlinear correlation between the pH and redox potential, the SHyE should be calculated using H2O2 production rate (r(H2O2)) and the Gibbs free energy from O2 and H2O to H2O2 (ΔG(H2O2)) and sunlight power intensity (Psun)28:

$${{{{{\rm{SHyE}}}}}}=\frac{{{{{{\rm{r}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})\times \Delta {{{{{\rm{G}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2})}{{{{{{{\rm{P}}}}}}}_{{sun}}}$$
(2)

Either increasing the reaction kinetics (slope of the J–V curve) or cathodically shifting the onset potential of the photoanode or anodically shifting the onset potential of the cathode raises the intersection point of the J–V curve of the photoanode and cathode, corresponding to a considerable increase in the working current of the coupled photochemical reaction under bias-free conditions29, which is the key to assisting AOPs with desirable H2O2 concentration.

On the basis of the above discussions, both photoanode and cathode are rationally optimized to achieve high SHyE. By tailoring the WO3 texture (Supplementary Figs. 5 and  6) and BiVO4 content (Supplementary Figs. 7 and 8), the photocurrent density is first optimized to 5.63 mA/cm2 at 1.76 V vs. RHE with an onset potential of 0.34 V vs. RHE. Then, a similar SnO2-x overlayer approach was adopted to increase the Faradic efficiency (FE) of H2O2 production23, as schematically illustrated in Fig. 3a. The optimized SnO2-x/BiVO4/WO3 was characterized by a combination of techniques, which indicate a porous mesoscopic photoelectrode with a microscopic layer-by-layer component structure consisting of ~5 nm thick SnO2-x on a porous BiVO4/WO3 film (Fig. 3b–d and Supplementary Figs. 9 and 10). UV–vis diffuse reflectance spectra reveal that SnO2-x/BiVO4/WO3 has an absorption edge similar to that of BiVO4/WO3 but much greater than the absorption edge of WO3 (Supplementary Fig. 11), suggesting a negligible influence on the energy-level diagram of the SnO2-x overlayer on BiVO4/WO3. Electrochemical impedance spectroscopy (EIS) spectra demonstrate a large improvement in the charge carrier transport efficiency of SnO2-x/BiVO4/WO3 compared to BiVO4 (Fig. 3e). The PEC water-oxidative H2O2 generation performance measured in 0.4 M NaHCO3 under 100 mW/cm2 with AM 1.5 illumination shows that the SnO2-x overlayer slightly enhances the photocurrent density of BiVO4/WO3 (6.1 mA/cm2 at 1.76 V vs. RHE), while cathodically shifting the onset potential from 0.64 V vs. RHE of BiVO4 to 0.32 V vs. RHE (Fig. 3f). Significantly, the SnO2-x/BiVO4/WO3 photoanode exhibits a stable H2O2 FE greater than 84% at potentials ranging from 0.5 to 1.7 V vs. RHE, which is 2.4 and 2.9 times higher than those of BiVO4/WO3 and BiVO4, respectively (Fig. 3g). The potential-independent H2O2 FE is a notable characteristic feature of the PEC reaction because the rate constant and activation energy are less affected at the semiconductor/electrolyte interface30,31, which is important for achieving a high SHyE in the coupled photochemical reaction. Furthermore, potentiostat i–t tests reveal that the SnO2-x/BiVO4/WO3 photoanode is more stable than BiVO4/WO3 and BiVO4 (Supplementary Fig. 12) because the passivation layer inhibits the photocorrosion of BiVO432. As mentioned above, the higher selectivity of water-oxidative H2O2 generation together with high photocurrent density, low onset potential, and good stability are key factors in achieving high SHyE since the coupled photochemical reaction is mainly driven by the photoanode here.

It is essential to realize that the water-oxidative generation of H2O2 at the SnO2-x/BiVO4/WO3 photoanode accompanies the water-reductive generation of H2 at the Pt cathode, where the overpotential of the HER is almost perfectly neat. Compared to that of water splitting to form H2O2 and H2, the actual voltage required to drive the coupled photochemical reaction of H2O2 generation is greater and can be expressed by:

$${{{{{\rm{V}}}}}}=1.07\,{{{{{\rm{V}}}}}}+{{{{{\rm{{\eta }}}}}}}_{{{{{{\rm{WOR}}}}}}}+{{{{{{\rm{\eta }}}}}}}_{{{{{{\rm{ORR}}}}}}}+{{{{{{\rm{\eta }}}}}}}_{{{{{{\rm{R}}}}}}}$$
(3)

where ηWOR is the overpotential of the WOR, ηORR is the overpotential of the ORR and ηR is the inner resistance of the PEC cell. Therefore, decreasing ηORR (given by the Tafel behavior of ORR catalyst) is essential to achieve high H2O2 production rates and SHyE. First, Mo-SACs/mrG was carefully chosen as the cathode catalyst for oxygen-reductive H2O2 production24. Rotating ring-disk electrode (RRDE) tests were carried out for in situ quantification of the H2O2 catalytic activity of the Mo-SACs/mrG catalysts. To exclude the impact of the carbonaceous substrate, which is proposed to serve as a potential catalyst toward electrochemical 2eORR33, a control sample denoted mrG was prepared via the same fabrication process used for Mo-SACs/mrG but without a metal source. Distinct RRDE curves of Mo-SACs/mrG and mrG are presented in Fig. 4a. The Tafel slope and onset potential (at a disk current density of 1 mA/cm2) of Mo-SACs/mrG are 81 mV/dec and 0.61 V vs. RHE, respectively, while those of mrG are 190 mV/dec and 0.55 V vs. RHE, respectively (Fig. 4a and Supplementary Fig. 13). Furthermore, a much higher H2O2 selectivity was observed for Mo-SACs/mrG compared to mrG from 0.35 to 0.65 V vs. RHE (Fig. 4b). Therefore, Mo-SACs/mrG was concluded to enable optimization of the intersection point of the J–V curve of the coupled photoanode and cathode with ideal H2O2 FE.

Prior to physically coupling the cathode and photoanode, the charge balance between the cathode and photoanode should be optimized since the activity of the cathode is limited by gas diffusion and electron transfer34. Occasionally, a microporous gas diffusion layer (GDL) composed of PTFE and carbon-black powder is introduced between the active catalysts and carbon fiber to enhance the gas diffusion efficiency and boost the three-phase contact line (TPCL), on which the accessible active sites are functional. Nevertheless, a thick and insulated GDL will slow down electron transfer, resulting in serious ohmic loss and performance decay35. Therefore, increasing the TPCL without hindering electron transfer is desirable. Herein, an interfacial strategy is proposed to construct an oxygen local confinement structure on Mo-SACs/mrG nanosheets via PTFE nanoparticle decoration to increase the TPCL without inducing GDL (Fig. 4c and d). The PTFE@Mo-SACs/mrG-GDE cathode was characterized by SEM and HAADF-STEM, and the results are shown in Supplementary Fig. 14. The mrG nanosheets adhere tightly to the carbon fiber array, and PTFE nanoparticles with a size of ~150 nm are evenly distributed on the surface of mrG nanosheets, where single Mo atoms are the only metal species present (Supplementary Fig. 15). As illustrated in Supplementary Fig. 16, the quasi-nanoarray aerophilic area generated by the PTFE nanoparticle combined with Mo-SACs/mrG nanosheets is able to simultaneously contribute to compatible TPCL and significantly enhanced electron transfer compared to the conventional GDL electrode36,37. Further insights into the influence of PTFE on the electrocatalytic activity are obtained by altering the loading amount of the PTFE nanoparticles (Supplementary Fig. 17). Water contact angle tests reveal that the wettability of PTFE@Mo-SACs/mrG-GDE could be efficiently suppressed when the loading amount of PTFE increased (Supplementary Fig. 18), which is expected to accelerate gas diffusion. However, excessive PTFE loading deteriorates the conductivity of the electrode, as evidenced by EIS (Supplementary Fig. 19). Therefore, a delicate balance among the mass transfer of oxygen, the conductivity of the electrode and H2O2 formation efficiency is required for performance optimization. Ultimately, a 2:1 mass ratio of PTFE and Mo-SACs/mrG achieves the best performance, with a current density of 10 mA/cm2 at 0.51 V vs. RHE and a Tafel slope of 53.2 mV/dec (Supplementary Fig. 20). As a result, PTFE@Mo-SACs/mrG-GDE shows the highest current density in the voltage range from 0.65 to 0.1 V vs. RHE (Fig. 4c), with the smallest onset potential (0.51 V vs. RHE) and the largest H2O2 FE (82.1%) (Fig. 4e) compared to the Mo-SACs/mrG-GDL electrode (see “Methods” for synthesis) and PTFE@mrG-GDE (see “Methods” for synthesis), indicating that both the oxygen-confined structure and the effective Mo-SACs/mrG catalyst contribute to the outstanding electrochemical performance. Notably, the optimized PTFE@Mo-SACs/mrG-GDE cathode exhibits good electrochemical stability for the ORR, with 98.4% of the initial current maintained after 7 h (Supplementary Fig. 21).

After optimization of the H2O2 product selectivity, onset potential, and current density, the assembled SnO2-x/BiVO4/WO3||PTFE@Mo-SACs/mrG-GDE PEC cell for redox coupling of the 2eWOR and 2eORR has a theoretical intersection point at 0.61 V vs. RHE at a current density of 1.64 mA/cm2 (Fig. 5a). The PEC performance was first investigated in a three-electrode system with SnO2-x/BiVO4/WO3 and PTFE@Mo-SACs/mrG-GDE as the photoanode and cathode, respectively. The linear sweep voltammetry (LSV) scan reveals a current density of 5.59 mA/cm2 at 1.76 V vs. RHE with a total H2O2 FE of 151% calculated from both the photoanode and cathode (Supplementary Fig. 22). Next, controlled potential tests were carried out to evaluate the H2O2 FE of the PEC device in a two-electrode system to evaluate the monolithic cell performance, and the results demonstrate a steady FE greater than 150% under a wide applied voltage range from 0 to 0.9 V vs. cell (Fig. 5b). Specifically, a bias-free photocurrent density of 1.56 mA/cm2 with an H2O2 FE of 152% was achieved (Fig. 5c), corresponding to an unassisted H2O2 production rate of 0.77 μmol/(min cm2) and an SHyE of 1.46%. The difference from the theoretical value is attributed to the PEC limitation, such as the pH gradient, IR drop, and the polarization that occurs at the counter electrode in the two-electrode system (ηR in Eq. (2))34,38. As a result, an average solar-driven H2O2 production rate of 0.74 μmol/(min cm2) is achieved, and the H2O2 concentration reached 0.85 mM within one hour in a 50 mL reactor (Fig. 5d). A summary of the PEC and solar-driven H2O2 production exhibits the highest unassisted solar-driven H2O2 production rate of our SnO2-x/BiVO4/WO3||PTFE@Mo-SACs/mrG-GDE PEC device (Fig. 5e and Supplementary Table 1), surpassing other state-of-the-art PEC systems23,39,40,41,42,43,44,45,46,47,48,49. Potentiostatic i–t tests were carried out under bias-free conditions and showed good stability of the PEC device, with 89% of the current retained after 20 h of H2O2 generation (Supplementary Fig. 23).

Further on, detailed and systematic studies have been performed to evaluate the ROSs generation/interconversion as well as the reaction mechanism in our self-cycled Fenton-like system. First, H-cell configuration was taken to specifically investigate the roles of 1O2 and ·O2 during the H2O2 production in both cathode and anode since both of them are common intermediates during the photocatalytic ORR for H2O2 production51. For the cathode part, the scavenger experiments using p-benzoquinone (BQ, a sacrificial agent for ·O2) and 2, 2, 6, 6-tetramethylpiperidine (TEMP, a sacrificial agent for 1O2) over PTFE@Mo-SACs/mrG-GDE were performed. As shown in Supplementary Fig. 34a, b, the amount of the H2O2 is invariant with respect to the concentration of both BQ and TEMP, indicating that the quenching of either 1O2 or ·O2 has no effect on the H2O2 formation during the ORR in the cathode part52. Similar experiment results were also observed in the anode part (Supplementary Fig. 34c, d) that neither the quenching of 1O2 nor ·O2 has impact on the generation of H2O2 during the WOR in the anode part. Moreover, RRDE tests have shown that the main product of O2 reduction in the cathode is H2O2 (Fig. 4a). Above all, we believed that the 1O2 and ·O2 are converted from the in situ generated H2O2 in the self-cycled photo-like system.

To further investigate the role HCO3 in the generation of ROSs of the self-cycled photo-like system, KPi electrolyte with the same pH value (adjusted by KOH) is taken to replace NaHCO3 electrolyte. As shown in Supplementary Fig. 36a, barely degradation effect is observed in the absence of Mn2+, even if H2O2 is effectively generated in the reaction system (via 2eORR in cathode) (Supplementary Fig. 36b), which indicates KPi cannot activate H2O2 into ROSs as the HCO3 does (Fig. 6d and Supplementary Fig. S1). Furthermore, the degradation rate in KPi electrolyte slightly raised to 5.4% at 40 min when Mn2+ is added, which is still much lower than that in HCO3 electrolyte with Mn2+ (89.5% at 40 min). That means the effective activation of H2O2 is largely suppressed in the absence of HCO3. EPR technology was employed to further investigate the mechanism. Methanol and DMPO were added as capture agents to exclude the interface of ·OH for detecting ·O2H54. As shown in Supplementary Fig. 37a, no signals of ·O2H are detected in the system, indicating ·O2H not being generated. The signal intensities of ·O2 are unchanged after the addition of TEMP while being disappeared after the addition of TBA, indicating that the ·OH is essential for the generation of ·O2 while 1O2 is not55(Supplementary Fig. 37b). Besides, no obvious signal is observed in the detection of 1O2 (Supplementary Fig. 37c). Moreover, the addition of TEMP and BQ has no impact on ·OH (Supplementary Fig. 37d). Therefore, the mechanism for the ROSs generation could be given:

$${{{{{{{\rm{HCO}}}}}}}_{3}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\leftrightarrow {{{{{{{\rm{HCO}}}}}}}_{4}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(4)
$${{{{{{{\rm{HCO}}}}}}}_{4}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\to {{{{{{{\rm{HCO}}}}}}}_{3}}^{-}+\cdot \!{{{{{\rm{OH}}}}}}+{{{{{{{\rm{HO}}}}}}}_{2}}^{-}$$
(5)
$$\cdot {{{{{\rm{OH}}}}}}+{{{{{{{\rm{HO}}}}}}}_{2}}^{-}\to \cdot {{{{{{{\rm{O}}}}}}}_{2}}^{-}+{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}$$
(6)
$$\cdot {{{{{\rm{OH}}}}}}+\cdot\!{{{{{{{\rm{O}}}}}}}_{2}}^{-}\to {}^{1}{{{{{\rm{O}}}}}}_{2}+{{{{{{\rm{OH}}}}}}}^{-}$$
(7)

On the other hand, the addition of Mn2+ in HCO3 electrolyte could largely increase the degradation rate by improving the generation of both 1O2, ·O2 and ·OH (Fig. 6d and Supplementary Fig. S1). While the Mn2+ in KPi electrolyte shows a little effect of degradation (Supplementary Fig. 36a), indicating that the complex between Mn2+ and HCO3 plays the main role in the generation of the ROSs, which is also proved by the Fig. S1. EPR measurements were performed to further investigate the mechanism. Unlike in the HCO3 system, the addition of TBA has little impact on the generation of ·O2 and 1O2 during the EPR measurements, indicating that ·OH is not essential for generating them in the Mn2+(HCO3)n system (Supplementary Fig. 38a, b). However, the signal for 1O2 disappeared after the addition of BQ, indicating 1O2 come from ·O2 in the Mn2+(HCO3)n system (Supplementary Fig. 38a). Moreover, the addition of TEMP and BQ has no impact on ·OH (Supplementary Fig. 38c). Above all, the ROSs generation in the Mn2+−HCO3 system is given:

$${{{{{{\rm{Mn}}}}}}}^{2+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{n}}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\to {{{{{{\rm{Mn}}}}}}}^{4+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{m}}}}}}}+{{\cdot}} {{{{{\rm{OH}}}}}}$$
(8)
$${{{{{{\rm{Mn}}}}}}}^{4+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{m}}}}}}}+{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\to {{{\cdot}}} {{{{{{\rm{O}}}}}}}_{2}^{-}+{{{{{{\rm{Mn}}}}}}}^{2+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{n}}}}}}}$$
(9)
$${{{{{{\rm{Mn}}}}}}}^{4+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{m}}}}}}}+\cdot \!{{{{{{{\rm{O}}}}}}}_{2}}^{-}\to {}^{1}{{{{{\rm{O}}}}}}_{2}^{-}+{{{{{{\rm{Mn}}}}}}}^{2+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{n}}}}}}}$$
(10)

In the self-cycled system, the transformation of MnIV species into MnII species would be largely enhanced through Eq. (11) by the reduction of a cathode according to Fig. 1b and Supplementary Fig. S3.

$${{{{{{\rm{Mn}}}}}}}^{4+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{m}}}}}}}+{{{{{{\rm{e}}}}}}}_{\,({{{{{\rm{cathode}}}}}})}^{-}\to {{{{{{\rm{Mn}}}}}}}^{2+}{({{{{{{\rm{HCO}}}}}}}_{3})}_{{{{{{\rm{n}}}}}}}$$
(11)

## Discussion

In summary, we report an unassisted solar-driven self-cycled photo-Fenton-like system through in situ production and utilization of H2O2 using an artificial leaf. First, we confirm that the Mn(II)/(IV) redox can be effectively recycled in the PEC system with the production of various ROSs, including •OH, O2•− and 1O2 driven by H2O2 activation and photoinduced cathode reduction under bicarbonate media. Next, in order to fulfill the adequate H2O2 supplement, comprehensive guidance toward enhancing PEC-to-H2O2 efficiency was emphasized with kinetic optimization of both the photoanode and cathode. Dramatically improved stability and current density due to the enhanced carrier transport were fulfilled in the photoanode, and an interfacial strategy was proposed to construct an oxygen local confinement structure to simultaneously benefit both the electron-transfer and mass-transfer processes at the cathode. As a result, the high recorded unassisted H2O2 production rate of 0.77 μmol/(min cm2) and an SHyE of 1.46% were achieved by coupling the optimized photoanode and cathode. Especially, a 70-cm2 artificial leaf with 30 days of stability was successfully realized as an effective self-cycled photo-Fenton-like system for wastewater treatment with the consumption of only oxygen, water, and sunlight. Notably, some improvements in this PEC-driven wastewater treatment system are needed for industrial-scale applications, including enhanced solar utilization efficiency, continuous oxygen supplement, and tuning electrolyte components. Therefore, additional discussion for addressing the above issues was carried out to identify the limitations, which are expected to be addressed to achieve further breakthroughs (see Supplementray discussion). Overall, the design of effective solar-fuel device presented here is expected to inspire further sustainable application, while the successful demonstration of a one-pot self-cycled wastewater treatment system highlights the promising potential of artificial photosynthesis technology in decentralized environmental governance applications.

## Methods

### Synthesis of SnO2-x/BiVO4/WO3

The WO3 nanosheet array used as the template electrode was first synthesized by a modified hydrothermal method. Briefly, a precursor solution was prepared by slowly dropping 2 mL of 3.5 M HCl into 25 mL of 25 mM sodium tungstate dihydrate aqueous solution and allowing the reaction to occur for 1 min. Then, 25 mM ammonium oxalate was dissolved in the precursor solution and reacted for 15 min until the solution became completely transparent and colorless. Stirring was performed throughout the whole dissolution and reaction process. Then, a bare fluorine-doped tin oxide (FTO) substrates (2 × 2 cm2) was placed against the Teflon wall before the solution was poured, and the hydrothermal reaction was carried out in a sealed 40 mL Teflon autoclave at 110 °C for 60 min. The obtained WO3·(H2O)x nanoarray was washed with deionized (DI) water, dried, annealed at 500 °C in the air for 3 h, and cooled naturally to obtain the WO3 nanoarray. Next, the BiVO4 precursor solution was synthesized by dissolving 100 mM bismuth nitrate pentahydrate and 110 mM ammonium metavanadate in 20 mL of ethylene glycol. After that, a certain amount of BiVO4 precursor solution was dropped onto the WO3 photoanode (10 µL for 1 cm2), followed by annealing at 500 °C in the air for 2 h and cooling naturally. Then, the obtained sample was placed into a 0.1 M NaOH solution to remove excessive vanadium oxide and obtain the BiVO4/WO3 photoanode. Finally, the SnO2-x layer was deposited according to our previous work to obtain the SnO2-x/BiVO4/WO3 photoanode27.

### Synthesis of Mo-SACs/mrG and mrG

Graphene oxide (GO) was synthesized by oxidation of natural graphite flakes according to a modified Hummers’ method. Mo-SACs/mrG was synthesized using a one-step process. In a specific procedure for synthesizing Mo-SACs/mrG, a pristine GO suspension was washed with DMSO several times to obtain a GO/DMSO solution with a GO concentration of 0.5 mg mL−1. Then, 250 µL of 3 mg mL−1 MoCl5/DMSO was added to 30 mL of a 0.5 mg mL−1 GO/DMSO solution. The homogeneous solution was transferred into a 50 mL Teflon-lined autoclave, which was sealed and maintained at 135 °C for 12 h in an oil bath. Finally, the resulting Mo-SACs/mrG was freeze-dried to obtain a powder after removing the DMSO solvent. The mrG was synthesized as same as Mo-SACs/mrG but without adding metal elements.

### Synthesis of PTFE@Mo-SACs/mrG-GDE

The Mo-SACs/mrG ink was prepared by dissolving Mo-SACs/mrG powder in DI-water/ethanol/5 wt.% Nafion solution (volume ratio of 9:1:0.05) at a concentration of 1 mg/mL and stirring for 48 h. Then, the catalyst ink was obtained by adding a certain amount of poly tetrafluoroethylene (PTFE) aqueous solution (0.1 wt.%) to the Mo-SACs/mrG ink and stirring for another 48 h to obtain Mo-SACs/mrG nanosheets with evenly distributed PTFE nanoparticles at the required mass ratio (PTFE@Mo-SACs/mrG). After that, the catalyst ink was dropped on the Teflon-treated hydrophobic carbon fiber (TCF) with a loading amount of 0.5 mg/cm2, followed by natural drying to obtain PTFE@Mo-SACs/mrG-GDE.

### Synthesis of Mo-SACs/mrG-GDL electrode

Mo-SACs/mrG-GDL electrode was prepared by dropping Mo-SACs/mrG ink on the gas diffusion layer coated carbon fiber with a loading amount of 0.5 mg/cm2.

### Synthesis of PTFE@mrG-GDE

PTFE@mrG ink was prepared the same as the PTFE@Mo-SACs/mrG-GDE except for changing the Mo-SACs/mrG into mrG.

### Synthesis of the artificial leaf

For the wireless artificial leaf structure, an Ag paste-Cu sheet-Ag paste was used to connect the SnO2-x/BiVO4/WO3 photoanode to the PTFE@Mo-SACs/mrG-GDE cathode. Whole ohmic contacts were covered with epoxy for waterproofing.

### Material characterizations

Scanning electron microscope (SEM) images were recorded on a field-emission scanning electron microscope (FESEM, JSM-7800F, Japan). X-ray diffraction (XRD) patterns were obtained with a D500/5000 diffractometer in Bragg-Brentano geometry under Cu Kα radiation. High resolution transmission electron microscopy (HR-TEM) images were collected on a FEI Talos F200X JEOL electron microscope with energy-dispersive X-ray spectroscopy (EDS). High- angle annual dark field-scan transmission electron microscopy (HAADF-STEM) was performed on a JEOL JEM-ARM200F TEM/STEM (200 kV) with a spherical aberration corrector. The X-ray photoelectron spectroscopy (XPS) measurements were performed on a PHI-5000versaprobeIII. The spectra were calibrated by the reference of the C1s peak at 284.8 eV. Ultraviolet-visible (UV-vis) diffuse reflectance spectra were obtained via a UV–vis spectrometer (Shimadzu UV-3600 spectrophotometer) equipped with an integrating sphere. Electron paramagnetic resonance (EPR) measurements was carrier out by a Bruker EMXplus spectrometer. The microwave power was 6.325 mW, microwave frequency was 9.826 GHz, and temperature was 298 K.

### PEC measurements

There are two systems (a three-electrode system and a two-electrode system) for the photoelectrochemical (PEC) measurements to investigate the activity of photoanode/tandem device. Specifically, the three-electrode system was carried out using a photoanode as a working electrode, saturated Ag/AgCl as a reference electrode, and a carbon electrode or designed cathode as a counter electrode, among which the designed cathode was taken as a counter electrode only in the tandem device. The two-electrode system was carried out to evaluate the tandem device using a photoanode as a working electrode and a cathode as a counter electrode. All potentials were converted to the reversible hydrogen electrode (RHE) reference scale by the equation E (V vs. RHE) = E (V vs. Ag/AgCl) + 0.0591 × pH + 0.197. The illumination source was a 300 W Xe arc lamp with an AM 1.5G filter (100mW/cm2, FX 300, Beijing PerfectLight Co. Ltd). Electrochemical Impedance Spectroscopy (EIS) measurements were carried out with a AC perturbation of 10 mV applied over the frequency range of 10−2 Hz-106 Hz.

### ORR measurements

Electrochemical oxygen reduction reaction (ORR) tests were performed in a CS2350H potentiostat (Corrtest, Wuhan) with a three-electrode cell at room temperature. A RRDE assembly (AFE7R9GCPT, Pine Instruments) consisting of a glassy carbon rotating disc electrode (Φ = 5.0 mm) and a Pt ring (Φ = 8.0 mm) was used, with a theoretical collection efficiency of 37%. A glassy carbon electrode loaded with catalyst was used as the working electrode. The working electrodes were prepared by dispersing the catalyst powder in ethanol and 5% Nafion 117 solution (20 µL of Nafion for 1 mL of ethanol) to achieve a catalyst concentration of ~1 mg mL−1. After sonication for 60 min, 7 μL of the catalyst ink was drop-dried onto a glassy carbon disc (area: 0.196 cm2). The ORR activity and selectivity were investigated from polarization curves and RRDE measurements conducted in an oxygen-saturated electrolyte. The H2O2 selectivity was calculated using the following equation: η(%) = 200 × (IRing/N)/(IDisc + IRing/N). The electron-transfer number (n) at the disc electrode during the ORR was calculated using n = 4 − (2 × ηH2O2), where IRing is the ring current, IDisc is the disc current and N is the collection efficiency.

### H2O2 production efficiency evaluation and organic dye degradation measurements

A H-type quartz reactor with two-compartment was used to evaluate both the anodic and cathodic H2O2 production efficiencies of the artificial leaf. The two-compartment reactor contained a 0.4 M NaHCO3 solution in both the photoanode and cathode chambers separated by Nafion 117 membrane. The electrolyte solution in cathode chambers was first purged with pure O2 for 30 min before testing, and the pure O2 was continuously fed for stability testing. All current densities were normalized to the geometrical area of the photoanode. An integrated PEC system (PEC1000, Beijing Perfectlight Technology Co., Ltd.) with a solar simulator (AM 1.5G) was used.

The artificial leaf was placed in a single-compartment quartz reactor under AM 1.5 illumination with continuous O2 purging to carry out the dye degradation tests, the electrolyte was 15 mL of 0.4 M NaHCO3 solution containing 15 ppm NP and 0.04 ppm Mn(II) when the active area of the artificial leaf was 7 cm2. Furthermore, the wastewater was changed to 200 mL of 0.4 M NaHCO3 solution containing 5 ppm rhodamine B (Rh. B), 5 ppm methylene blue (MB), 5 ppm NP and 0.04 ppm Mn(II) when the active area of the artificial leaf was extended to 70 cm2. Sample aliquots was taken at regular intervals and tested quickly. The concentration of 4-nitrophenol (NP), methylene blue (MB) and rhodamine B (Rh. B) wereas measured by a Ultraviolet-visible (UV-vis) spectremeter (PE lambda 750 spectrophotometer).

### Quantification of H2O2

H2O2 evolution was detected using the N,N-diethyl-1,4-phenylene-diamine (DPD) method. A stock solution of DPD was prepared by dissolving 0.1 g of DPD in 10 mL of a 0.05 M H2SO4 solution. A peroxidase (POD) solution was prepared by dissolving 10 mg of POD in 10 mL of deionized water and was kept in a refrigerator until use. A potassium phosphate buffer solution was prepared by mixing 49.85 mL of deionized water, 43.85 mL of 1 M monobasic potassium phosphate, and 6.3 mL of 1 M potassium phosphate. Sample aliquots (2 mL) were collected by a syringe during irradiation and mixed with 0.4 mL of potassium phosphate buffer solution, 3 mL of water, 0.05 mL of DPD, and 0.05 mL of POD and were shaken for 90 s. The obtained solutions were analyzed by Ultraviolet-visible (UV–vis) spectremeter (PE lambda 750 spectrophotometer).