Introduction

The unanticipated consequences of rising CO₂ emissions and climate change are looming on a global scale.¹ Among the various approaches to carbon capture and utilization, direct photochemical fixation of CO₂ with water as an electron donor is one of the most promising global challenges.² Since Lehn et al. discovered the photochemical reduction of CO₂ into CO catalyzed by Re^I complexes as one of the critical milestones in modern artificial photosynthesis,³ much effort has been devoted to developing a benchmark reaction and improving reactivity, and intensive studies on the reaction mechanism.⁴ Following the Re-complexes, various other metal-complexes, such as Ru^II-,⁵ Ir^III-⁶ Cu^I-,⁷ Co^II-,⁸ Ni^II-,⁹ Mn^I-,¹⁰ and Fe^II-complexes,¹¹ have been reported to electrochemically/photochemically catalyze CO₂ reduction. Because of their superior properties, such as photo- and thermal stability, high reduction ability, and extremely low absorption in the visible region without interfering with light harvesting by a redox photosensitizer, trans(Cl)-[Ru(diimine)(CO)₂Cl₂]-type catalysts have been widely used in various photocatalytic systems for CO₂ reduction.^{5b, 12} Many photocatalytic CO₂ reduction systems, consisting of metal-complex catalysts and redox photosensitizers that initiate a single photochemical electron transfer from the reductant to the catalyst, exhibit high photocatalytic properties such as high quantum yield, durability, and product selectivity.^{2b, 5b, 13} Due to the low oxidation power of the redox photosensitizers in the excited state, these systems require sacrificial electron donors such as amines and ascorbic acid. The incorporation of water molecules into these systems is another challenging issue owing to the limitations of using sacrificial electron donors in molecular catalyst systems for CO₂ reduction. Pioneering reports on hybrid systems combining molecular photocatalysts (MCs) for CO₂ reduction with semiconductors for water oxidation,^{2b, 14} as well as fully semiconductor systems for both terminal ends,^{2c, 15} have appeared. High energy conversion efficiencies have also been reported for electrochemical molecular catalyst systems coupled with photovoltaic devices.¹⁶ These reports on photochemical CO₂ fixation with water as an electron donor, however, have focused on the evolution of dioxygen as a water oxidation reaction on metal oxides. Conversely, photochemical water oxidation to generate O₂ by MCs has long been hampered by the bottleneck, “Photon-flux-density problem of sunlight,”¹⁷ caused by rarefied solar radiation. To reach a higher oxidation state (+4), water oxidation requires four-electron oxidation via sequential four-photon processes. The rarefied photon flux density of sunlight forces MCs to wait for the next photon's arrival in by far the longer timescale up to the second order than the inherent one of molecules within ≈μs, which inevitably leads to unexpected transformation/decomposition of MCs during the long waiting period to suffer many collisional processes with impurities and solvent molecules and to lose water oxidation activity. However, another mode of water oxidation, such as a two-electron process forming H₂O₂ induced by one-photon excitation of MCs, has been developed,^17c which does not require waiting for the next photon to avoid the “Photon-flux-density problem.”¹⁷ Hydrogen peroxide is one of the most useful chemicals and would be far superior to O₂ as a water oxidation product due to higher energy storage (1.77 V for H₂/H₂O₂ vs. 1.23 V for 2H₂/O₂) and easier separation of the products (H₂, CO(gas)/H₂O₂(liquid)) than H₂, CO(gas)/O₂(gas).^17c Coupling CO₂ fixation at the reduction terminal with H₂O₂ formation at the oxidation terminal is one of the most promising targets, leading to more practical and useful artificial photosynthesis.^{17c, 18} We attempted to construct a Z-scheme-type artificial photosynthetic system consisting of a photoanode with a molecular photocatalyst for two-electron water oxidation into H₂O₂ and a photocathode with a molecular photocatalyst for CO₂ reductive fixation. A novel molecular catalyst photocathode for CO₂ reduction, a Ru-based molecular photosensitizer, and catalyst units immobilized on NiO particles using a polypyrrole layer, which exhibits stable reactivity even after extended visible-light irradiation, have recently been developed.¹⁹ Using this molecular catalyst photocathode (MC-photocathode), we have fabricated a new “Molecular catalyst photoanode (MC-photoanode)” capable of driving two-electron water oxidation to form H₂O₂ for the construction of a Z-scheme type artificial photosynthetic system. Crucial points in designing the MC-photoanode would be 1) tuning the redox character with the MC-photocathode (Ru-based photosensitizer+catalyst/NiO) to drive the Z-scheme system using only visible light as energy source, i.e., under the “Bias free” condition, 2) minimizing the unfavorable desorption of the molecular catalyst from the photoanode substrate during the reaction, 3) maximizing the oxidation ability of forming H₂O₂ from water, 4) exhibiting reactions of CO₂ fixation and H₂O₂ formation under nearly neutral conditions (pH 5–9), and 5) easy preparation with most available elements. Based on these design guidelines, we chose water-insoluble [Sn^IVTPyP(O⁻)₂]²⁻ (trans-dioxo-coordinated 5, 10, 15, 20-tetra(4-pyridyl)porphyrinatetin (IV): SnTPyP) with doubly deprotonated axial ligands (−O⁻) under neutral conditions as a novel visible-light-absorbing molecular catalyst for two-electron water oxidation. Another water-insoluble molecular catalyst is [Al^IIITPyP(OH)₂]⁻ (trans-dihydroxy-coordinated 5, 10, 15, 20-tetra(4-pyridyl) porphyrinate aluminum (III): AlTPyP) with hydroxy axial ligands (−OH) was also used as a reference molecular catalyst. These metalloporphyrins have sufficient hole potentials to drive the two-electron oxidation of water to H₂O₂ following one-electron oxidation of the porphyrin ring via photochemical or electrochemical initiation.²⁰

An n-type semiconductor (SnO₂ nanoparticles) was then selected as the substrate of the MC photoanode to fabricate Sn^IV- or Al^III-tetrapyridylporphyrin (SnTPyP or AlTPyP) decorated tin oxide particles (SnTPyP/SnO₂ or AlTPyP/SnO₂) through covalent bond formation between the axial ligands of SnTPyP or AlTPyP and the surface OH groups on SnO₂, which allowed exergonic electron flow to the NiO substrate of the MC-photocathode under bias-free conditions.

In this study, we report the first successful artificial photosynthesis construction that can drive CO₂ reduction on (Ru-complex photosensitizer+catalyst)/NiO as a dye-sensitized MC photocathode and H₂O₂ formation from water on SnTPyP/SnO₂ or AlTPyP/SnO₂ as MC photoanodes using only visible light as an energy source without any external bias. In terms of solar energy conversion, this reaction produces H₂O₂ from water (equations. 2 and 4) and can accumulate significantly more energy than conventional systems that produce O₂ (equations. 1 and 3.

Results and Discussion

In designing photoanodes for water oxidation, we developed two types of MC photoanodes, that is, SnTPyP or AlTPyP (MTPyP, Figure 1), both of which can photocatalyze the two-electron oxidation of water to form H₂O₂, on a SnO₂ n-type semiconductor electrode (SnTPyP/SnO₂ or AlTPyP/SnO₂), which was fabricated through covalent adsorption between the axial ligands of MTPyP and the surface OH group of SnO₂ (Figure 1A, see experimental (SI)). Cyclic voltammetry of SnTPyP or AlTPyP in CH₃CN/H₂O (8/2, v/v) with 0.1 M TBAPF₆ as the supporting electrolyte at neutral pH revealed catalytic oxidation waves at 1.56 V vs SHE for SnTPyP^20c and 1.26 V vs SHE for AlTPyP.^20b Because the excited energy of these metalloporphyrins is ≈2.1 eV²¹ in their singlet excited states, visible light irradiation of the photoanode (SnTPyP or AlTPyP adsorbed on the surface of SnO₂ (E_{conduction band}≈−0.41 V vs SHE at pH 7)²² would surely induce an electron injection from the excited MTPyP into the conduction band of SnO₂ (ΔG=≈−0.13 eV for SnTPyP and ΔG=≈−0.43 eV for AlTPyP) (Figure 1C). Because the electrochemically formed one-electron oxidized species of SnTPyP and AlTPyP drive water oxidation into H₂O₂,²³ visible light irradiation of the MC photoanode in water was investigated to see if an anodic photocurrent and the formation of H₂O₂ as the water oxidation product was induced. In a three-component cell with a SnTPyP/SnO₂ or AlTPyP/SnO₂ photoelectrode as the working electrode, Ag/AgCl (saturated KCl) as the reference electrode, and a Pt coil as the counter electrode, photoelectrochemical experiments for water oxidation on the MC photoanode were performed. The linear sweep voltammograms (LSV) for SnTPyP/SnO₂ and AlTPyP/SnO₂ photoelectrodes revealed distinct photo-responsive anodic currents: Anodic currents began to appear at ≈−0.05 V vs Ag/AgCl in both cases and became stable at the more positive potential >≈+0.1 V vs Ag/AgCl for SnTPyP/SnO₂ (≈40 μA: Figure 2A) and AlTPyP (≈15 μA: Figure 2 b), respectively, following visible light irradiation (λ=420 nm, 0.8 mW cm⁻²).

In designing artificial photosynthesis consisting of molecular photocatalysts that perform CO₂ reduction by utilizing electrons from water, we attempted to combine the MC photoanodes described above with a dye-sensitized molecular photocathode (MC photocathode) consisting of [Ru(diimine)₃]²⁺-type complexes as light-harvesting units and a Ru (diimine)(CO)₂Cl₂ type complex as a catalyst unit, both of which were covalently connected and fixed in and on a polypyrrole layer on the NiO electrode (RuCAT-RuC₂-PolyPyr-PRu/NiO) (Figure 1B).¹⁹ The MC-photocathode demonstrated its photocathodic response (λ>460 nm, 28.2 mW cm⁻²) starting from a relatively positive potential (E=+0.3 V vs Ag/AgCl, red lines in Figure 2) in CO₂-bubbled aqueous solution containing 50 mM NaHCO₃ as the electrolyte (pH=6.6).¹⁹ Since the conduction band of SnO₂ (≈−0.41 V vs SHE at pH 7)^22a is situated above the valence band of NiO (+0.2 V vs SHE at pH 7),²⁴ an exergonic electron flow from the conduction band of SnO₂ (injected electrons from the excited SnTPyP or AlTPyP) to the valence band of NiO (injected holes from the excited Ru photosensitizer units) would be surely expected when the MC-photoanodes are coupled with the MC-photocathode (Figure 1C). Both the anodic and cathodic photocurrents were observed to well overlap at the potential region, −0.05–+0.2 V vs Ag/AgCl, indicating that full-cell device consisting of photoanodes and photocathodes would be promisingly working for electron transport from H₂O to CO₂ without any external bias potential applied (Figure 2).

The formation of H₂O₂ as a water-oxidation product on the MC photoanode was investigated prior to the construction of the full-cell device (Figure S1). A colorimetric method with Ti^IV-tetrapyridylporphyrin (TiTPyP) as the sensor was used for the SnTPyP/SnO₂ photoanode (Figure S2).^{23a, 25} H₂O₂ was detected in the half-cell reaction at +0.1 V vs Ag/AgCl in water (([NaHCO₃]=50 mM) for 2-h irradiation (λ=420 nm, 0.8 mW cm⁻²) in the amount of 32 nmol (Turnover Number (TON)=4.6). When the gas phase in the half cell reaction system was checked using a gas chromatograph equipped with a thermal conductivity detector and nitrogen as the carrier gas, O₂ was not detected after 2 h of light irradiation. These findings suggest that H₂O₂ was the only oxidation product of water on the SnTPyP/SnO₂ photoanode in the photoelectrochemical system, as observed in the AlTCPP/TiO₂ photochemical system.²⁶ A similar photocatalytic reaction using AlTPyP/SnO₂ as the MC-photoanode instead of SnTPyP/SnO₂ showed a selective formation of H₂O₂ in the amount of 15 nmol (TON=3.1) in the half-cell experiment for 2 h (Figure S1B, λ=420 nm, 0.8 mW cm⁻²), though the reactivity was less than SnTPyP/SnO₂ as observed in the LSV (Figure 2). The lower reactivity of AlTPyP/SnO₂ was explained by the following theoretical DFT calculations. The one-electron oxidized form of [SnTPyP(O⁻)₂]²⁻ with deprotonated axial ligands (−O⁻) in a doublet electronic spin state has its electron spin exclusively localized on the axial oxygen atom as oxyl radical (−O⋅) that serves as a key intermediate in H₂O₂ formation, whereas the electron spin of the one-electron oxidized form of [AlTPyP(OH)₂]⁻ having protonated axial ligands (−OH) mostly delocalized over the porphyrin ring that would be less reactive against H₂O₂ formation (Figure S3). The higher oxidation potential of SnTPyP (1.56 V vs SHE) compared to AlTPyP (1.26 V vs SHE) also explains the differences in reactivities for two-electron oxidation of water into H₂O₂. Both factors would simultaneously control the reactivity.

To confirm the origin of the oxygen atoms in the produced H₂O₂, SnTPyP/SnO₂ was subjected to isotope labeling experiments with H₂¹⁸O (H₂¹⁸O/H₂¹⁶O=1/9, v/v). GC-MS is used to estimate the uptake of ¹⁸O in the forms of ¹⁶O–¹⁸O and ¹⁸O–¹⁸O in O₂ via an enzymatic reaction with catalase to convert H₂O₂ into O₂. As shown in Figure 3, treatment of the reaction mixture with the enzyme catalase resulted in an increase in ¹⁶O–¹⁸O and ¹⁸O–¹⁸O, but a decrease in ¹⁶O–¹⁶O (Figure 3), indicating that the oxygen atom of H₂O₂ originated from H₂O. These results confirm that irradiating the MC photoanode (SnTPyP/SnO₂) to visible-light induces selective two-electron oxidation of water to produce H₂O₂. The results of isotope labelled experiments clearly suggest the produced H₂O₂ originates from the water molecules. The electrolyte species bicarbonate (HCO₃⁻) ions might provide a promotive path for the formation of H₂O₂ through peroxo-carbonate (HCO₄⁻) intermediate which prevents the further oxidation of peroxides into molecular oxygen in the catalytic cycle.^{25b, 27}

Based on the results of the half-cell reactions, we can anticipate that simultaneous visible-light irradiation of both electrodes induces electron flow from SnTPyP or AlTPyP in the MC photoanode to the Ru catalyst in the MC photoanode without the use of an external bias (Figure 1C). A complete cell was built, with a SnTPyP/SnO₂ MC-photoanode and a RuCAT-RuC₂-PolyPyr-PRu/NiO MC-photocathode connected and set up in each compartment separated by a Nafion® membrane. Under the bubbling of CO₂ (pH 6.6), the two compartments were filled with an aqueous solution containing 50 mM NaHCO₃ as the electrolyte. The good overlap of the LSV curves between the SnTPyP/SnO₂ MC photoanode and the RuCAT-RuC₂-PolyPyr-PRu/NiO MC photocathode (Figure 2) indicates that these two pairs of photoelectrodes can be used as a fully molecular artificial Z-scheme system with no external bias. First, the requirement for simultaneous excitation of both the MC photoanode and photocathode was tested by maintaining one of the photoelectrodes in the dark. When one electrode was irradiated with visible light, whereas the other was kept in the dark, only a small number of electrons flowed between the photoelectrodes, as shown in Figure 4A. Under simultaneous excitation with monochromatic light at 420 nm for the MC photoanode (2.7 mW cm⁻²) and 480 nm for the MC-photocathode (3.5 mW cm⁻²), a rather stable photocurrent of 7 μA was observed.

The photoelectrochemical CO₂ reduction was then tested in the zero-resistance ammeter mode, when the wavelengths of light irradiation were chosen for the MC-photoanode (SnTPyP/SnO₂) as λ=420 nm (1.32 mW cm⁻²) and for the MC-photocathode (RuCAT-RuC₂-PolyPyr-PRu/NiO) as 460 nm<λ<650 nm (27.8 mW cm⁻²) to maintain the catalytic activities of photoelectrodes. The MC-photocathode (RuCAT-RuC₂-PolyPyr-PRu/NiO) remained unchanged without desorption of the catalyst, while the MC-photoanode (SnTPyP/SnO₂) suffered a slight desorption of SnTPyP without decomposition (Figure S4). The degree of desorption of SnTPyP was much smaller than that in the half-cell reaction where the pH of the reaction mixture might decrease to cause the desorption because the pH should be maintained in the full-cell reaction where OH⁻ is consumed on the MC-photoanode and H⁺ is consumed on the MC-photocathode. For more than 5 h, a relatively stable photocurrent was observed (Figure 4C). After 5 h of light irradiation, the photocathode produced primarily HCOOH (310 nmol), with minor amounts of CO (86 nmol) and H₂ (48 nmol) also detected. In the photoanode compartment, H₂O₂ (330 nmol) was only detected in the solution, while O₂ was not detected in the gas phase even after 5 h of irradiation (Figure 4B). On the photocathodic and photoanodic sides, the Faradaic efficiencies of the reduction and oxidation reactions were calculated to be 95.4 and 72.7 %, respectively. We demonstrate that the firstly reported fully molecular two-electron approach for simultaneous production of H₂O₂ and CO/HCOOH has comparable Faradaic efficiencies and energy conversion efficiencies with the reported Z-scheme systems integrated four-electron oxidation of water by using semiconductor based photoanodes (Table S1). The turnover number (TONs) was 17.8 for the reduction products and 47.3 for H₂O₂ based on the catalyst units used on each photoelectrode (Table 1).

Table 1. Summary of photoelectrochemical CO₂ reduction by RuCAT-RuC₂-PolyPyr-PRu/NiO photocathode in the full cell system using MTPyP/SnO₂ photoanode in CO₂ purged 50 mM NaHCO₃ solution (pH 6.6) (Photocathode: λ_ex=460–650 nm, 27.8 mW cm⁻², photoanode: λ_ex=420 nm, 1.32 mW cm⁻²).

• [a] Turnover number. [b] Faradaic efficiency of H₂O₂ formation. [c] Faradaic efficiency of overall reduction reactions.

These findings indicate that the molecular units on both photoelectrodes are catalytically cycled during the photoinduced reactions of selective two-electron oxidation of water to H₂O₂ and CO₂ reduction in an aqueous solution without any applied bias. The efficiency of photo- to chemical-energy conversion for the full cell reaction for CO₂ reduction coupled with H₂O₂ generation from water was discovered to be ≈1.2×10⁻²%.

Conclusion

The selective two-electron oxidation of water to H₂O₂ was demonstrated by one-photon excitation of SnTPyP or AlTPyP molecules adsorbed on SnO₂ as the MC photoanode. These MC photoanodes were successfully combined with an MC photocathode composed of polypyrrole layers immobilizing a supramolecular photocatalyst (RuCAT) fixed on NiO to fabricate the first example of molecular Z-scheme-type full-cell devices capable of inducing both photocatalytic reactions of two-electron water oxidation into H₂O₂ and CO₂ reduction to HCOOH/CO upon visible light irradiation. The photoelectrochemical full-cell combination of SnTPyP/SnO₂ and RuCAT-RuC₂-PolyPyr-PRu/NiO, in particular, exhibits good photocatalytic activity.

Supporting Information

Experimental details and additional data for supporting the findings in the manuscript have included in the Supporting Information.

Conflict of interest

The authors declare no conflict of interest.