A synergetic cocatalyst for conversion of carbon dioxide, sunlight and water into methanol

on a FEI Titan Cubed

The conversion of CO 2 into liquid fuels, using only sunlight and water, offers a promising path to carbon neutrality. An outstanding challenge is to achieve high efficiency and product selectivity. Here, we introduce a wireless photocatalytic architecture for conversion of CO 2 and water into methanol and oxygen. The catalytic material consists of semiconducting nanowires decorated with core-shell nanoparticles, with a copper-rhodium core and a chromium oxide shell. The Rh/CrOOH interface provides a unidirectional channel for proton reduction, enabling hydrogen spillover at the core-shell interface, as shown by density functional theory. The vectorial transfer of protons, electrons, and hydrogen atoms allows for switching the mechanism of CO 2 reduction from a proton-coupled electron transfer pathway in aqueous solution to hydrogenation of CO 2 with a record high solar-to-methanol efficiency of 0.29%. The reported findings demonstrate a highly efficient, stable, and scalable wireless system for synthesis of methanol from CO 2 that could provide a viable path towards carbon neutrality and environmental sustainability.
Solar-light conversion of CO2 into liquid fuels such as methanol and ethanol can provide a viable solution to renewable energy production and mitigation of greenhouse gases. 1,2,3 To date, current strategies for CO2 conversion rely mainly on thermocatalysis 4 and electrocatalysis 5 that consume a significant amount of thermal energy, or electricity. Solar light photocatalysis can offer a more sustainable approach. 6 The main outstanding challenges are to achieve: (1) high-efficiency of CO2 reduction, and (2) high product selectivity toward liquid fuels. Semiconductor materials with a large band gap can provide energetic carriers to reduce CO2, however, they unavoidably suffered from the limited light response range. 7,8 Furthermore, CO2 reduction toward CO is often more favorable than toward liquid fuels. 9 Recently, theoretical and experimental studies have suggested the importance of intermediates with different binding modes in controlling the efficiency, selectivity, and reaction rates during CO2 reduction. In fact, it has been suggested that *COOH is a key intermediate in the formation of CO and following hydrocarbon products, while HCOO* is a key intermediate in the formation of HCOOH. 10,11 Therefore, tuning the relative stability of reaction intermediates can favor a specific pathway and improve product selectivity during CO2 reduction. Here, we show that selective formation of methanol can be enhanced by loading a bimetallic cocatalyst on a semiconductor photocatalyst.
In nature, metalloenzymes like the photosynthetic ribulose bisphosphate carboxylase (RuBisCO) and formate dehydrogenases of methylotrophic yeast and bacteria can reduce CO2 with high selectivity toward specific products. For example, Mo-and W-dependent formate dehydrogenases (FDH) can reversibly catalyze CO2 reduction to formate (Scheme 1a). [12][13][14][15][16][17][18][19][20] Inspired by enzymatic CO2 reduction, we have developed a wireless architecture for selective CO2 reduction reaction to methanol. The photocatalytic system consists of InGaN/GaN nanowires loaded with co-catalytic Cu-Rh/CrOx core-shell nanoparticles (Scheme 1b), with bimetallic copper-rhodium cores and chromium oxide shells. Unique to this artificial photosynthetic system is the vectorial transfer of protons, electrons, and hydrogen atoms to Cu active sites, that can switch the mechanism from CO2 reduction by conventional proton-coupled electron transfer (PCET) via *COOH in aqueous solutions to hydrogenation of CO2 via HCOO*. High CO2 reduction activity is shown under visible light irradiation in a mild environment (1 atm CO2 atmosphere and pH 7 deionized water). Methanol is the major product, cogenerated with the dioxygen by-product in a 2:3 ratio. The system exhibits an average turnover frequency of 37.8 µmol·cm -2 ·h -1 , with the best measured activity being 49.8 µmol·cm -2 ·h -1 (298.5 mmol·g -1 ·h -1 ), which corresponds to a record-high solar-to-liquid fuel efficiency of 0.29% and 0.38%, respectively.

Design Principles
InGaN/GaN nanowires function as visible-light absorbers by separating charge carriers (Scheme 1b). The photogenerated electrons flow into the Cu-Rh/CrOx nanoparticles where protons are reduced and bind to the Rh surface as hydrogen atom adsorbates. Rh/CrOx core-shell nanoparticles are known to be efficient catalysts for H2 evolution by H2O reduction. [21][22][23] On our Cu-Rh/CrOx nanoparticles, the in situ generated hydrogen atoms spillover to the nearby Cu sites and react with CO2. Here, CO2 is hydrogenated leading to the selective production of CH3OH, in contrast to electrocatalytic CO2 reduction on Cu surfaces that typically yield a mixture of products. [24][25][26][27][28][29][30][31][32] We synthesized interfacial structures of GaN/Rh, GaN/Cu, Rh/Cu, Rh/CrOOH, and Cu/CrOOH to explore the possibility of forming an efficient photochemical architecture (Fig. 1a), considering that electrons transfer from Cu to Rh since the work function of Rh is larger than that of Cu. 33 We found that the Rh(111) facet matches the GaN(101 � 0) facet very well, with a lattice mismatch < 4% ( Fig. 1b and Table S4). The interaction between Cu and GaN, however, is not as favorable due to the larger lattice mismatch between the Cu(111) and GaN(101 � 0) surfaces (Table S4), yielding a weaker interaction (Supplementary Figure 11) and a large kinetic barrier for electron transfer from GaN to Cu. So, the electron transfer pathway from GaN to Rh determines the main electron transfer pathway since it is favored by both thermodynamics and kinetics.
Our density functional theory (DFT) analysis of Rh/CrOOH and Cu/CrOOH interfaces ( Fig. 1c and Supplementary Figure 12) shows that Rh is likely covered by a CrOOH shell, as suggested previously by Domen and coworkers. 33 Furthermore, we find that the CrOOH shell is unlikely to form on Cu due to the lattice mismatch (Table S5) Figure 12). The Rh/Cr and Rh/Cu interfaces allow for unidirectional proton transfer and reduction, followed by hydrogen spillover from Rh to Cu. The exposed Cu surface enables hydrogenation of CO2 by hydrogen adsorbates effectively inducing catalytic CO2 reduction. Therefore, our DFT computational analysis supports the viability of the catalytic architecture shown in Scheme 1b.
We analyzed CO2 reduction on Cu(111). We found that CO2 reduction can occur via two reaction pathways, including standard electrochemical reduction by proton-coupled electron transfer (PCET) using electrons from the electrode and protons from the bulk solution, and hydrogenation by surface adsorbed H atoms (reducing equivalents of H + + e − ). We modelled the reduction of CO2 with electrons by preparing a negatively charged Cu surface which serves as a model of the electrified Cu electrode. We analyzed binding of CO2 on the charged Cu surface. We found that binding of CO2 through the oxygen atoms is unfavorable, resulting in physisorption of CO2 with distances of ~3 Å from Cu to O in the CO2 molecule (Fig. 1d). However, binding of CO2 through C forms a *COO − species which can be readily protonated by protons from the bulk solution to form the *COOH intermediate, leading to formation of *CO and ultimately hydrocarbon products such as CH4, C2H4, etc. In fact, that is the favored mechanism for most electrochemical CO2 reductions on Cu electrodes which exhibits poor selectivity. In contrast, when CO2 is reduced by H atoms on the Cu surface, both the *COOH and HCOO* binding modes of CO2 are possible, with the HCOO* binding mode being more favorable. (Fig.   1d). Therefore, the photochemical architecture shown in Scheme 1b with vectorial transfer of protons, electrons, and hydrogen atoms to the Cu active site can achieve high selectivity of CO2 photoreduction to CH3OH. In the following section, we demonstrate that such a catalytic architecture can be experimentally realized on InGaN nanowires.

Results
Synthesis and characterization of photocatalysts. Single crystalline Mg-doped p-type InGaN/GaN nanowire arrays were grown on a silicon wafer utilizing molecular beam epitaxy (see Methods). The ratio of indium and gallium of InGaN was determined by X-ray diffraction (see Methods). The band gap of the InGaN/GaN nanowires was located at ~525nm. Co-catalysts Rh, Cu and Cr2O3 were then photo-deposited on the nanowire surface (see Methods). Fig. 2a shows InGaN/GaN nanowires grown on the silicon wafer with an average length of ~1 μm and width of ~100 nm. Fig. 2b shows the high angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) image of an InGaN/GaN nanowire after photo-deposition of co-catalysts (see Methods). The grey spots show that the co-catalytic nanoparticles decorated well the sidewalls of the InGaN/GaN nanowire. Fig. 2e shows the energy dispersive X-ray (EDS) elemental mapping of the material, showing 10 nm particles with a distinct Rh and Cu core and Cr2O3 shell. Note here that EDS mapping merely exhibits a 2-dimensional perspective, while the signals lay on the images with a 3dimensional depth. Likewise, Fig. 2c shows the deposited core-shell particles of Rh, Cu and CrOOH (Cr2O3•H2O) widely dispersed on the nanowire surface, also shown in Fig. 2b. Typically, the CrOOH shell shares a thickness of around 2 nm for most of the observed particles, corresponding to 3-4 layers of CrOOH (Supplementary Figure 12). The core could vary from 2 to 20 nm due to the randomness in the photodeposition. As predicted by our DFT calculations ( Fig. 2d and Supplementary Figure 12), Cr2O3•H2O selfassembles during the reaction in-situ and forms a film of CrOOH on the Rh surface. 33 The average loaded amount of Rh, Cu and Cr is 0.067 μmol/cm 2 , 0.21μmol/cm 2 and 0.0062 μmol/cm 2 , respectively, estimated through inductively coupled plasma-atomic emission spectroscopy (ICP-AES). The chemical states of each photo-deposited cocatalyst were analyzed by X-ray photoelectron spectroscopy (XPS) before and after the photocatalytic reaction. The Rh 3d5/2 spectrum exhibits four peaks which correspond to the metallic and the oxidized state of Rh (Fig. 2f) indicating a strong interaction between Rh and CrOx. It is further noticed that the ratio of intensities between the oxide and metal peaks increases after the photocatalytic reaction. Fig. 2g shows the Cu 2p3/2 XPS spectra taken before and after the photocatalysis reaction. Together with the location of copper LMM peak (Supplementary Figure 2) at 918.69 eV (Kinetic energy), these results show no chemical shift relative to metallic copper, suggesting a negligible amount of Cu2O. Therefore, there is no significant interaction, or synergetic effect between chromium oxide and copper likely due to the lattice mismatch and weaker interactions between Cu and CrOOH. It is further noticed that both copper and chromium show no chemical shift before and after the reaction (Fig. 2g, Fig 2h). The XPS spectrum is thus consistent with our DFT calculations ( Fig. 2d) showing that Rh atoms on the surface are directly interacting with both Cu atoms and CrOOH layers, while Cu atoms have no significant interaction with CrOOH. The Cu surface is thus exposed to the solution favoring adsorption of CO2.
Photocatalytic reaction experiments and results. The photocatalytic performance of RhCuCr2O3/InGaN was characterized using a solar simulator, with water and CO2 as the feedstocks (see Methods). Therefore, sunlight was the only energy input to this wireless photocatalytic reaction. Figure 3a shows the rate of generation of CH3OH, H2, CH4, and O2 using InGaN with (or without) the incorporation of Cu and Rh/CrOOH co-catalysts, averaged over three cycles (see Methods). We find that InGaN nanowires exhibit a negligible photocatalytic activity. Incorporation of Rh/CrOOH enhances the photocatalytic activity, yielding H2 and O2 as the major products due to water splitting, as well as trace amounts of CH3OH. Incorporation of Cu on the InGaN nanowire arrays generates a small amount of CH3OH and trace amounts of CH4 and O2. InGaN nanowires with both Cu and Rh/CrOOH exhibit a dramatic enhancement in the production rate and selectivity for CH3OH. Generation of CH3OH and O2 maintains a stoichiometric ratio of 2:3 only when loading both Cu and Rh/CrOOH.
The analysis of the photocatalytic performance of InGaN nanowire arrays with either Rh/Cu or Cu/CrOOH provided further insights on the reaction mechanisms and confirmed the superior performance of the Rh/Cu/CrOOH@GaInN photocatalysts. Fig. 3a shows the performance of Rh/Cu and Cr/Cu deposited on InGaN nanowires. Significant loss of catalytic activity is observed for InGaN nanowires with only Rh/Cu without CrOOH, or only Cu/CrOOH without Rh. Those results show that the highest turnover frequency and selectivity of methanol generation is obtained when Rh, Cu, and CrOOH are simultaneously incorporated into the InGaN nanowires. The simultaneous incorporation of Rh, Cu and CrOOH also leads to high turnover numbers with stable production of methanol over many cycles.
An important observation is that CrOOH does not function as a protective shell to preserve the methanol production efficiency in the case of Cu/CrOOH, even though it significantly enhances the overall water splitting performance in the case of Rh/CrOOH. These differences are likely due to the lattice mismatch between CrOOH (2.98 Å) and copper (2.56 Å), which prevents formation of the Cu/CrOOH core-shell structures observed for Rh with CrOOH (Supplementary Figure 12). In fact, significant differences in composition of the surface were confirmed by STEM characterization and the EDS signals spectrum of Cu/CrOOH of co-deposited samples (Supplementary Figure 6). Our data thus provides unambiguous evidence for a synergistic effect of Rh, Cu, and CrOOH in promoting CO2 reduction to methanol. Significantly, the production of methanol and oxygen remains nearly stoichiometric when using the Rh/Cu/CrOOH@GaInN photocatalysts. The activity of methanol and oxygen production remains nearly constant during three cycles of measurements (Fig. 3c), showing negligible decomposition or poisoning of the photocatalyst, commonly reported for other systems developed for photo-driven CO2 reduction.
Next, we have studied the effect of different loading amounts of Cu and Rh on the photocatalytic performance for reduction of CO2 to CH3OH (see Methods). Fig. 3b shows that a relatively low rate of CH3OH and low selectivity is observed when either Rh, or Cu dominates, presumably due to the lack of proper formation of Rh/CrOOH/Cu core-shell structures under those extreme deposition conditions. In addition, we observe a high production rate (49.7 µmol·cm -2 ·h -1 ) and high selectivity (>80%) for a wide range of Rh/Cu precursor ratios (e.g., from 10:2 to 2:10), consistent with formation of robust Rh/CrOOH/Cu core-shell structures as confirmed by STEM characterization.
Our analysis of products source tracing and GC/NMR measurements utilizing 13 CO2 and H2 18 O as feedstocks (see Methods) confirms that C in the product CH3OH originated from carbon dioxide. Furthermore, our analysis of the gas products from gas chromatography-mass spectrometry (GC-MS) (Fig. 3d) shows a clear m/z = 36 peak due to 18 O2, confirming that the by-product dioxygen was generated from water, consistent with the reaction mechanism predicted by our DFT calculations. Furthermore, our analysis of the liquid products by nuclear magnetic resonance (NMR) (Fig. 3e) shows a clear split peak due to the extra neutron in 13 C atom, supporting that 13 CO2 is the C source of methanol.
The change of attenuated total reflectance (AT R ) signals at CO2/H2O/RCC@GaInN interface reveals the gain or loss of species. As shown in Fig. 4a, during 60 min irradiation, the IR negative peak at 2355 cm -1 , attributed to the antisymmetric stretching vibration of O=C=O in CO2, 34 sharply decreased with irradiation time due to CO2 reduction. Meanwhile, an increasing IR peak at 1198 cm -2 , attributed to the stretching vibration of C-O bond in C-O-H, was also observed, which was close to the reported C-OH stretching band on Cu, 35 further demonstrating the reduction of CO2.
We have further performed long-term stability studies of the Rh/Cu/CrOOH@InGaN photocatalysts. The measurements for the stability test were performed on separately prepared samples for a total of twelve cycles (Fig. 4b) (see Methods). The results show a decrease in performance along the evolution of multi-cycles, although the productivity remained at a fairly high level >15 µmol·cm -2 ·h -1 . The activity loss is not due to degradation of the InGaN/GaN nanowire light absorber but rather related to mechanical loss of coated metal particles during rinsing, as confirmed by a detailed STEM characterization of the nanowire after the reaction (Fig. 4c). Between each cycle of the reaction, the sample was rinsed with deionized water to remove traces of methanol that might otherwise affect the analysis of performance. The rinsing process, however, unavoidably removed some number of co-catalysts and nanowires from the surface. In fact, the loadings of both Rh and Cu were significantly reduced by rinsing as characterized by ICP-AES. In future studies, we will explore enhancing the mechanical stability of the catalyst by coating the surface of the co-catalyst nanoparticles with a thin layer of Al2O3 or TiO2.
Studies of reaction mechanisms. Our computational analysis of CO2 reduction on the Cu(111) surface has been performed at the DFT level. The work functions of Cu and Rh and the favorable contact between Rh(111) and GaN(101 � 0) surfaces, favors a mechanism where photoelectrons transfer to the Rh nanoparticles and reduce protons on the Rh surface to form surface bound hydrogen atoms. Once the hydrogen atom adsorbates saturate the coverage of the Rh(111) surface, they diffuse to nearby Cu surface sites since the diffusion barriers for hydrogen atoms on transition metal surfaces are small (~ 0.1 eV). 34 The calculated adsorption energy of hydrogen on bare Rh(111) and Cu(111) surfaces (i.e., at a low coverage of θ = 1/12) is −0.64 and −0.34 eV, respectively ( Table S6). The obtained binding energy on Rh (111) is in fairly good agreement with the experimental value (-0.80 eV) for low coverage of hydrogen on Rh(111), 35 while the calculated adsorption energy on Cu(111) is consistent with previously reported computational results of -0.3 to -0.4 eV on Cu(111). 34,[36][37] Clearly, hydrogen adsorbs more strongly on bare Rh(111) than on Cu(111). However, in our photocatalytic system, the Rh(111) surface is covered by CrOOH. So, we have analyzed the adsorption energy of hydrogen at the Rh(111)/CrOOH(001) interface. Adding more CrOOH layers move the hydrogen adsorption energy at the Rh(111)/CrOOH(001) interface more positive, making the spillover of H from the Rh(111)/CrOOH(001) interface to Cu(111) energetically favorable ( Table S6). As already demonstrated in previous experimental study, 33 the CrOOH shell allows selective permeation of protons from bulk solution to the Rh/CrOOH interface. The work function difference between Rh and Cu ensures the preferred transfer of photo-generated electrons to Rh, providing a unidirectional channel for proton reduction. Furthermore, the interface modulates the hydrogen adsorption energy on the Rh(111) surface, enabling spillover of hydrogen atoms generated in situ at the interface to nearby Cu sites. Hydrogen on Cu serves as a reductant for hydrogenation of CO2 (Fig. 5a). Fig. 1c (also see SI, Supplementary Figure 13) shows that the CrOOH shell on top of the Rh core allows for proton transfer from the bulk solution to the Rh(111) surface. At the same time, the CrOOH film prevents other species (e.g., O2) to approach the Ru(111) surfaces and consume the reductive H atoms generated in situ. The resulting design ensures effectiveness of the Rh/CrOx core-shell structure for photoreduction of protons. 33 The surface H atoms generated in situ at the Rh(111)/CrOOH(001) interface can either generate H2 or diffuse to the Cu surface to hydrogenate CO2. The H spillover from the Rh/CrOOH interface to the Cu(111) surface is thermodynamically allowed (Fig. 5a). On the Cu surface, hydrogen atoms reduce CO2 generating the formate intermediate bound to the surface through its O atoms (bi-HCOO*). As shown in Fig. 1d, the bi-HCOO* intermediate leads to high selectivity of CO2 reduction to CH3OH.
The Bader charge analysis (Table S7) suggests that the surface bound H atom is better described as a hydride, as indicated by its negative charge, and is able to reduce CO2 forming the negatively charged HCOO − . There are two possible pathways to reduce the bi-HCOO* with a surface H atom.  [39][40] Here, we find that the free energy profiles for reduction of bi-HCOO* to *HCOOH and *OCH2O* are indeed very similar without solvation (Supplementary Figure 14). Upon considering the solvation effect, however, *HCOOH is greatly stabilized ( Table S8). Fig. 5b shows that the pathway via *HCOOH is thermodynamically more favorable. It is noteworthy that the *OCH2O* intermediate, which has been suggested to be an important intermediate in gas-phase CO2 hydrogenation over Cu catalysts, 40 is similar to the *OCH2O* intermediate observed upon CH4 oxidation, 41 suggesting its central role in both CH4 oxidation and CO2 reduction in the gas phase. In aqueous CO2 reduction, however, *OCH2O* is not as important as HCOOH* due to its poor water solubility (Table S8).
Further reduction of the HCOOH* intermediate forms the *OCH2OH* intermediate, which has been suggested to be important in CO2 hydrogenation. [36][37][38] Further reduction of *OCH2OH* can generate both *OCH3 + *OH, or *O=CH2 + *OH2, where the former pathway is thermodynamically favored (Fig. 5b). The *OCH3 intermediate can be further reduced to CH3OH*, leaving the Cu surface to complete the catalytic cycle.
The mechanism shown in Fig. 5 involves reduction of CO2 by H atoms adsorbed on the Cu surface, a process equivalent to proton coupled electron transfer in electrochemical CO2 reduction. H spillover from the Rh/CrOOH interface to the Cu surface is energetically favorable. The reduction of bi-dentate HCOO* to HCOOH* is a thermodynamically 'uphill' process, so the bi-dentate bound HCOO* should be a detectable intermediate by in situ spectroscopy. Reduction of CO2 according to this mechanism thus involves hydrogenation by in situ generated H atom adsorbates, leading to high selectivity for CH3OH production. Consistently, our experiments show CH3OH production with no significant amounts of any kind of by-product (e.g., CO, HCOOH, H2CO, or CH4). Furthermore, no enhancement of CH3OH production is observed in the presence of H2 consistent with reducing equivalents being generated in situ.
It is important to compare our reaction mechanism, with an industrial process for CH3OH production 39 based on hydrogenation of CO2, catalyzed by metallic surfaces such as Cu(111), a process that requires thermal cleavage of the H-H bond at 500 K. In contrast, our catalytic system generates H atoms in situ at room temperature on the Rh/CrOOH surface, allowing for CO2 reduction to CH3OH under ambient temperature and pressure. Fig. 5b shows that upon formation of the HCOOH* intermediate, the subsequent reaction steps are quite favorable. Therefore, formation of HCOO* represents the mechanistic bottleneck of CO2 reduction on the Cu surface, particularly during heterogeneous CO2 hydrogenation in the gas phase ( Supplementary   Figure 14) due to the large free energy cost of generating *OCH2O*, or *HCOOH intermediates. Therefore, it is unsurprising that the industrial production of methanol from mixtures of CO/CO2/H2 (synthesis gas) over a Cu/ZnO/Al2O3 catalyst requires typical reaction conditions of 230-280 °C and 50-120 atm. 39 Remarkably, we find that H atom adsorbates on Cu(111) can reduce CO2 in aqueous solution, facilitating the conversion of HCOO* to HCOOH*, and thus making CH3OH production possible under ambient temperature and pressure (Supplementary Figure 15). In addition, the overall reaction can be driven by the in situ generated H atoms at the Rh/CrOOH interface through H spillover (Supplementary Figure 16).

Discussion
We have introduced novel structures of photocatalytic systems that can switch the mechanism of CO2 reduction from conventional PCET of electrocatalytic systems on Cu surfaces to hydrogenation of CO2. For catalysts consisting of Rh/CrOOH and Cu, the dominant pathway is hydrogenation of CO2 by in situ generated H adsorbates, leading to highly selective production of CH3OH via the HCOO* intermediate. With only Cu serving as a cocatalyst for proton reduction, both pathways are open, including PCET via *COOH and hydrogenation of CO2 via HCOO*, leading to low selectivity with production of both H2, CH3OH and CH4. The PCET pathway becomes predominant for hydrocarbon generation when applying a sufficiently negative bias potential to the Cu electrode. However, the resulting reaction still exhibits poor selectivity.
It is important to compare the performance of our catalytic systems to the catalytic cofactors of natural enzymes that reduce CO2 with high selectivity. For example, molybdoenzyme formate dehydrogenases 39 catalyze the reduction of CO2 to formate, while carbon monoxide dehydrogenases reversibly reduce CO2 to CO. Remarkably, these two types of enzymes catalyze the two alternative pathways of CO2 reduction observed on Cu electrodes with catalytic nanowires, including hydrogenation of CO2 via the HCOO* intermediate and the PCET through the *COOH reaction intermediate.   . 6e and 6f show that intermediates with CO2 bound to metal centers through O atoms in both the enzyme active sites and the Cu surface led to hydride transfer and formation of formate. The main difference when comparing the binding modes of NO2 − and HCOO − is that NO2 − (and presuming HCOO − ) binds to [Mo]formate dehydrogenase as a mono-dentate O binding mode ligand, while HCOO − binds to the Cu(111) surface and exhibits a bi-dentate O binding motif. Beyond these small differences in binding modes, the comparison shows that the selectivity of CO2 reduction for one or the other reaction pathway can be tuned by tuning the specific binding mode of CO2 to the active site in either the catalytic cofactor in the natural enzyme or the synthetic catalysts. In our catalytic system, selectivity toward CH3OH is achieved by introducing bi-metallic active sites that play specific functional roles during the photochemical CO2 reduction.
PCET and hydrogen atom transfer are two different ways to deliver a reducing equivalent to CO2 during the reduction reaction. Beyond CO2 reduction, the PCET mechanism is ubiquitous in electrochemical, photoelectrochemical, and photochemical reactions in aqueous solutions. In contrast, hydrogen-atom transfer is more important in gas phase reactions, and in non-polar solutions. Our Rh/Cu/Cr system, however, operates in aqueous solutions and suppresses PCET on the Cu surface by enabling the unidirectional flow of photoelectrons from GaN to Rh. The shell of CrOOH allows for directional proton transfer from the bulk solution to the Rh surfaces. The two directional processes enable the generation of reducing equivalents at the Rh/CrOOH interface leading to hydrogenation of CO2 upon hydrogen spillover from Rh to Cu. The use of two metals with different work functions is crucial for establishing directionality of electron transfer while CrOOH has OH groups between layers (Supplementary Figure 13) that enable effective proton translocation through a Grotthus-type mechanism, as commonly observed in biological systems. 41 The CrOOH film also serves as an effective membrane that allows proton transfer from the bulk solution to the electrode surface while blocking other species (e.g., O2) that could compete with proton during the reduction reaction. Therefore, directional proton and electron transfer pathways in our Rh/Cu/Cr system mimic analogous directional processes for proton and electron transfer in enzymes (Scheme 1), providing a biomimetic system and inspiration for development of catalysts materials for selective chemical transformation.

Conclusion
We have shown that catalytic materials based on Rh/Cu/CrOOH nanoparticles dispersed on InGaN/GaN surfaces enable in-situ generation of hydrogen atoms on the Rh surface that spillover Cu where CO2 is selectively reduced by hydrogenation to methanol with high production rate and selectivity. A record-high activity of 37.38 µmol·cm -2 ·h -1 and a maximum solar-to-fuel conversion efficiency of 0.29% is achieved. The reported results demonstrate an unrivaled artificial photocatalytic configuration for unitary production of methanol and hydrogen from visible light using CO2 and water as the only chemical inputs.

Method
Synthesis of InGaN/GaN nanowires. InGaN/GaN nanowire arrays were grown on a silicon wafer using a plasma-assisted MBE system. The catalyst-free nanowire growth was carried out under a nitrogen-rich environment to promote formation of N-terminated surfaces protected against photo corrosion and oxidation. The nanowires, consisting of multiple segments of InGaN/GaN, were doped with a p-type magnesium dopant. Ga and In beam equivalent pressures (BEPs) were set at ~7.35 × 10 −8 Torr and ~7× 10 -10 Torr, respectively, and the growth temperature of InGaN was ~765 °C. 21,22 Cocatalyst loading. Rh, Cr 2 O 3 , and Cu co-catalyst nanoparticles were loaded on InGaN/GaN nanowires by photo deposition. The InGaN/GaN wafer was firstly stabilized on a Teflon holder. Then the holder was transferred to a reaction chamber containing 50 mL of 20vol% methanol aqueous solution. 5 μL of 0.2 mol L -1 Na 3 RhCl 6 (Sigma-Aldrich), 5 μL of 0.2 mol L -1 K 2 CrO 4 (Sigma-Aldrich), and 5 μL of 0.2 mol L -1 CuCl 2 ·2H 2 O (Sigma-Aldrich) were added to the solution. The chamber was covered by a quartz cover and vacuumized. The chamber was irradiated under 300 W Xe lamp (Cermax, PE300BUV) for 30 min 23 . Finally, the obtained photocatalyst wafer was soaked in 70 o C deionized water under vacuum for 10min and dried at 80 o C in vacuum.
Carbon dioxide reduction. The photo-reduction of CO 2 was performed in a glass chamber (diameter, 80mm; volume, 400 mL) sealed with a top quartz window under 300 W xenon lamp illumination. Both AM1.5G filter and 400nm long pass filters were inserted between the lamp and the reaction chamber. Prior to illumination, 50 mL of distilled water was purged with high-purity CO 2 (PurityPlus, 99.8%) for 15 minutes and then poured into the chamber. Subsequently, the chamber was vacuumized to further remove any gas in the water. The chamber was then filled with high-purity CO 2 (99.8%), and purged with CO 2 for 10 minutes.
Before illumination, the chamber was placed in a water bath bed to stabilize the temperature. After 10 minutes, the lamp was mounted on the reaction chamber to start illumination. During the stability test, at an interval of 1 hour, the sample was taken out of the reaction chamber, exposed to air, rinsed with deionized water, and dried in CO 2 flow to get rid of the methanol adsorbed on the surface. The chamber was also thoroughly cleaned. The next cycle of the reaction was then conducted to test the device stability. The CO 2 RR to methanol is shown below, Following this equation, the reaction Gibbs free energy (ΔG) was predicted to be 638.73 kJ•mol −1 (or 6.62 eV per molecule). 42,43 E CH 3 OH = ∑ n photon, i × 6×numbers of generated CH 3 OH n photon, i where n photon, i is the number of photons per interval I, with a wavelength of λ i .

STF =
Methanol production rate �mmol s −1 �×638730 J mol −1 Light intensity (W cm −2 ) × Wafer area (cm 2 ) The light intensity was calibrated with a Newport power meter light detector with 9 cm (distance from the device to the lamp lower end), shaded by aluminum foil with a 1 cm window to receive irradiance. Two wavelength filters (AM1.5G and 400nm long pass) were applied to simulate the visible light spectrum in solar light with intensity around 3 W•cm -2 . We note that the theoretical maximum efficiency for methanol generation is predicted to be 35.2%. 42 Calculation of turnover number (TON) and turnover frequency (TOF). The TON for photocatalytic CO 2 reduction reaction was calculated, as follows:   Variations of methanol selectivity (left axis) vs. co-catalyst deposition conditions. The production rates of methanol and hydrogen are shown in the right axis. Here 20:2 means 20μL of Rh precursor and 2μL Cu precursor were used in the photoadaptation process. c) Methanol production under optimized co-catalyst conditions during twelve cycles of measurements. Each cycle lasted for one hour. After each cycle, the solution was removed and refilled, and the sample was taken out of the reaction chamber and rinsed by deionized water to remove the potentially attached methanol molecules. Experimental conditions: 50 mL deionized water, CO 2 , 300 W Xenon-lamp, 3.5 W/cm 2 , AM1.5G + 400nm long pass filter. d) GC-MS spectrum of 18-oxygen generated during the CO 2 reduction reaction with 18-O labelled water. e) NMR H-spectrum of 12 C methanol generated during the CO 2 reduction reaction with CO 2 . NMR H-spectrum of 13 C methanol generated during the CO 2 reduction reaction with 13-C labelled CO 2 .