Metal to non-metal sites of metallic sulfides switching products from CO to CH4 for photocatalytic CO2 reduction

The active center for the adsorption and activation of carbon dioxide plays a vital role in the conversion and product selectivity of photocatalytic CO2 reduction. Here, we find multiple metal sulfides CuInSnS4 octahedral nanocrystal with exposed (1 1 1) plane for the selectively photocatalytic CO2 reduction to methane. Still, the product is switched to carbon monoxide on the corresponding individual metal sulfides In2S3, SnS2, and Cu2S. Unlike the common metal or defects as active sites, the non-metal sulfur atom in CuInSnS4 is revealed to be the adsorption center for responding to the selectivity of CH4 products. The carbon atom of CO2 adsorbed on the electron-poor sulfur atom of CuInSnS4 is favorable for stabilizing the intermediates and thus promotes the conversion of CO2 to CH4. Both the activity and selectivity of CH4 products over the pristine CuInSnS4 nanocrystal can be further improved by the modification of with various co-catalysts to enhance the separation of the photogenerated charge carrier. This work provides a non-metal active site to determine the conversion and selectivity of photocatalytic CO2 reduction.

or Cd vacancy CdS with single Au atom deposition for CO 2 adsorption is different 30 .CO 2 prefers to physically adsorb on single Au atoms of Au/CdS 1−x and photoreduction into CO, while CO 2 is more likely to chemically bond on the Cd vacancies of Au/Cd 1−x S, resulting in a remarkable CO and CH 4 generation rate on Au/Cd 1−x S.He et al. synthesized a ZnIn 2 S 4 nanosheet photocatalyst with abundant Zn vacancies 31 , where CO 2 can be efficiently adsorbed on Zn vacancies to form CO 2 − species and highly selective photoreduction into CO.Yu et al. designed a Cu 3 SnS 4 photocatalyst with S vacancies to increase ratios of Cu (I/II) for CO 2 photoreduction 32 .The formed Cu (I) acts as adsorption sites for CO 2 , conducive to further hydrogenation of CO intermediate into CH 4 .Xie et al. showed that the defect-state CuIn 5 S 8 ultrathin nanosheets have low-coordination Cu and In sites for CO 2 adsorption to form highly stable Cu-C-O-In intermediates, which tend to obtain 100% CH 4 selectivity 33 .Xu et al. designed a Co-Ni-P NH/BP catalyst with bimetallic sites to form a highly stable Co-O-C-Ni intermediate for the selective photoreduction of CO 2 to CH 4 34 .However, the complex structures of defects on photocatalyst make it only a plausible correlation between defect structures and product selectivity.Some research shows that the adsorbed interaction between CO 2 and metal sites is relatively weak since the formed metal-C bonds are weaker than the highly stable C=O bonds in CO 2 .This leads to the easy cleaving of metal-C bonds during the reaction process, hindering the deep reduction of CO 2 into hydrocarbons 30 .Obviously, the non-metal sites on metal sulfide photocatalysts are very rarely considered the primary active center for the adsorption and activation of CO 2 .
Here, we have successfully prepared multiple metal sulfides, including CuInSnS 4 octahedral nanocrystal and corresponding individual metal sulfides In 2 S 3 , SnS 2 , and Cu 2 S, through a simple one-step hydrothermal method.The CuInSnS 4 nanocrystal is thermodynamically favorable to activate CO 2 and leads to a switch of main products from CO to CH 4 with a yield of 6.53 μL h −1 for the visible-light-driven CO 2 reduction with H 2 O vapor without the assistance of any noble metal cocatalysts.In contrast, individual metal sulfides can only produce CO.We reveal that different adsorption configurations of CO 2 on metal sulfides lead to different products in CO 2 photoreduction.The nonmetal sulfur atom in the prepared multiple metal sulfides CuInSnS 4 octahedron nanocrystal is thermodynamically favorable to activate CO 2 and leads to a switch of main products to CH 4 , as compared with the common individual metal sulfides In 2 S 3 , SnS 2 and Cu 2 S with metal center as active sites to form CO products.CO 2 is revealed to be adsorbed on the S atom center of CuInSnS 4 to form an S-C-O-In structural unit, which is more conducive to protonation and leads to the efficient photocatalytic yield of CH 4 .Thus, we provide an insight into the role of non-metal center of photocatalyst in determining the conversion and selectivity of photocatalytic CO 2 reduction.Although the pristine CuInSnS 4 only exhibits a yield of CH 4 evolution of 6.53 μL h −1 (corresponding to 5.83 μmol h −1 g −1 ), the activity and selectivity of CH 4 evolution on CuInSnS 4 can be significantly improved by modifying with cocatalysts such as Pt, CoO, NiO, and Co(OH) 2 .We believe that this knowledge can contribute to the development of more efficient and selective photocatalysts for CO 2 reduction in the future.

Characterization of CuInSnS 4 octahedral nanocrystal
The as-prepared CuInSnS 4 nanocrystals belong to the cubic spinel structure (JCPDS No. 29-0548), as revealed by the X-ray diffraction pattern (Fig. 1a), which were prepared through a simple one-step hydrothermal reaction.The hydrothermal temperature (160 °C, 180 °C, and 200 °C) did not have any evident impact on the crystalline and purity of CuInSnS 4 (Supplementary Fig. 1).Under similar hydrothermal processes, In 2 S 3 with a tetragonal crystal phase structure (JCPDS No. (Supplementary Fig. 2).The cubic spinel CuInSnS 4 crystal structure shows that either In or Sn atom is coordinated with six S atoms to form an octahedron structure, while Cu atom is formed by a [CuS 4 ] tetrahedral structural unit (Supplementary Fig. 3).For comparison, in In  S1.The atomic ratio of Cu:In:Sn is 1.06:1.00:1.00,which closely matches the theoretical value of CuInSnS 4 .Supplementary Fig. 7 shows the SEM images of In 2 S 3 , Cu 2 S, and SnS 2 samples.In 2 S 3 exhibits a morphology of microspheres self-assembled from nanosheets.Cu 2 S has the basic shape of nanoparticles, while SnS 2 displays the morphology of ultrathin nanosheets.Both scanning electron microscopy and transmission electron microscopy images of nanoscale microstructure confirm that the prepared CuInSnS 4 displays an octahedral nanocrystal structure with a size of about 30 nm (Supplementary Fig. 8 and Fig. 1c).
The high-resolution TEM image (Fig. 1d) shows the exposed facets of the octahedron with a lattice spacing of 0.601 nm, which is assigned to (1 1 1) facets of CuInSnS 4 .Selected area electron diffraction further verifies that the CuInSnS 4 sample not only exposes the (1 1 1) crystal plane but also has a single crystal structure (Fig. 1e).All the results confirm the successful preparation of CuInSnS 4 nanocrystal with high-quality exposed (1 1 1) crystal faces.The specific surface area of the CuInSnS 4 sample is about 24.1 m 2 g −1 , while that of the prepared In 2 S 3 , SnS 2 , and Cu 2 S are 26.8 m 2 g −1 , 37.6 m 2 g −1 , and 3.6 m 2 g −1 , respectively (Supplementary Fig. 9).As displayed in Fig. 1f-i, the uniform distribution of Sn, S, In, and Cu elements in CuInSnS 4 octahedral nanoparticles indicates that the catalyst is of high purity.
X-ray photoelectron spectroscopy was used to compare the electronic states of the obtained sample.The Cu2p 3/2 and Cu2p 1/2 binding energies of CuInSnS 4 sample are 932.07eV and 951.90 eV, respectively (Fig. 2a).This demonstrates that the valence state of Cu is +1 in the CuInSnS 4 sample 35,36 , which is also confirmed by the Cu LMM spectra (Supplementary Fig. 10).Notably, the Cu2p-binding energies of the CuInSnS 4 sample is identical to that of Cu 2 S. The binding energies of In3d 5/2 and In3d 3/2 in the CuInSnS 4 sample are 444.63eV and 452.18 eV, respectively.These values indicate that the valence state of In in the CuInSnS 4 sample is +3.Compared with In 2 S 3 , the In3d binding energy of CuInSnS 4 uniformly shifts toward the lower binding energy, as shown in Fig. 2b.This is attributed to the difference in the In coordinated environment between CuInSnS 4 and In 2 S 3 because the partial In atom in In 2 S 3 exists in the state of [InS 4 ] tetrahedron 37 .In the CuInSnS 4 sample, the Sn3d 5/2 and Sn3d 3/2 doublets are centered respectively at 486.30 eV and 494.70 eV, assigning to Sn 4+ valence state.Notably, the binding energy of Sn in CuInSnS 4 is slightly lower than that in the parent SnS 2 (Fig. 2c).The possible reason is that the Sn atoms are in different crystal structures 38 .Furthermore, the binding energies of S2p 3/2 and S2p 1/2 in the CuInSnS 4 sample are measured to be 161.45eV and 162.70 eV, respectively, which corresponds to the S 2− valence state.In the S2p XPS spectra, the binding energies of S atoms increase in the order of Cu 2 S < In 2 S 3 < SnS 2 < CuInSnS 4 , as shown in Fig. 2d.S atoms in CuInSnS 4 have the highest binding energy.This is interpreted by the fact that the average bond length (0.253 nm) between sulfur and metal atoms in CuInSnS 4 is slightly larger than in monometallic sulfide 4,37,39 .Moreover, the S atom in CuInSnS 4 has a higher binding energy than that of monometallic sulfides, indicating an electron-deficient state of the S atoms in CuInSnS 4 compared to monometallic sulfides.This electrondeficient state of the S atom in CuInSnS 4 can serve as the reaction site for CO 2 adsorption and activation.

Photocatalytic conversion of CO 2 and H 2 O vapor
The photocatalytic CO 2 reduction performance of the samples was evaluated in a customized sealed quartz glass vessel, in a gas-solid reaction system, with a small amount of water vapor in a CO 2 atmosphere, under the irradiation of a 300 W Xe lamp with a 420 nm cutoff wavelength filter (Supplementary Fig. 11).It is important to note that this reaction is a gas-solid phase reaction.The system contains only 50 μL of water, which evaporates into water vapor upon injection into the reactor.As a result, only gaseous products such as CH 4 , CO, and a small amount of H 2 are detected, while liquid products (such as CH 3 OH, HCHO, HCOOH, etc.) are not detected.Figure 3a shows the corresponding photocatalytic CO 2 performance of the samples under visible-light irradiation.For single metal sulfides In 2 S 3 , Cu 2 S, and SnS 2 , CO is the main product with a yield rate of less than 3.2 μL h −1 from photocatalytic CO 2 reduction, while the multi-electron transfer product CH 4 is hardly formed.Whereas, the CuInSnS 4 sample shows excellent photocatalytic CO 2 reduction performance, yielding CH 4 as the main product besides a slight amount of CO and H 2 evolution.The hydrothermal temperature has no obvious impact on the CO 2 reduction performance of CuInSnS 4 (Supplementary Fig. 12).The rate of CH 4 generation reaches 6.53 μL h −1 for the CuInSnS 4 sample.The selectivity of CH 4 is calculated to be 67.3% based on the contents of carboncontaining products.The significant difference in the product selectivity demonstrates the different mechanisms of CO 2 reduction or the different active sites between CuInSnS 4 and single metal sulfides.The controlled blank experiments under other conditions were investigated to confirm the occurrence of CO 2 reduction on CuInSnS 4 , as shown in Supplementary Fig. 13.The CH 4 product is not detected without either light irradiation or catalyst, proving that CO 2 reduction is a light-induced catalytic reaction on CuInSnS 4 .Meanwhile, without adding H 2 O into the reaction system, only a very few products are detected, indicating that H 2 O is also one of the essential reactants involved in the reaction.When N 2 is used instead of CO 2 for the reaction, only a small amount of CO is detected, directly proving that the source of CO and CH 4 products is CO 2 .The presence of small amounts of CO may be attributed to contamination from ambient air, reactors, and equipment, as we can see that small CO products is also detected without CuInSnS 4 photocatalysts 40 .Figure 3b shows the stability of the CuInSnS 4 sample for photocatalytic reduction of CO 2 .It is clear that the CuInSnS 4 sample presents good performance without noticeable activity decrement after three-cycle photocatalytic CO 2 reduction tests of a total of 27 h (9 h visible-light irradiation for each cycle).Neither crystal structural transformation nor absorption behavior changes are found in the XRD pattern and ultraviolet-visible diffuse reflectance spectra for the CuInSnS 4 sample after photocatalytic reaction (Supplementary Fig. 14).These results suggest that CuInSnS 4 possesses good stability during photocatalytic CO 2 reduction.Moreover, the XPS of the catalyst after the reaction shows that a weak photocorrosion phenomenon occurs in CuInSnS 4 41,42 (Supplementary Fig. 15).Specifically, the photogenerated holes or the active oxygen species oxidize the surface S 2− of the catalyst to SO 3 2− .The peaks with binding energies in the range of 168.26~170.26eV are assigned to the XPS peaks of SO 3 2− species (Supplementary Fig. 16).The photocatalytic performance of CuInSnS 4 sample is evaluated under different monochromatic light wavelengths in Fig. 3c.It is observed that as the wavelength of the incident light increases, the photocatalytic activity significantly decreases.However, the CuInSnS 4 nano-single crystal photocatalyst demonstrates a CH 4 generation rate of 0.69 μL h −1 and CO generation rate of 0.22 μL h −1 under the irradiation of 525 nm monochromatic light, which is surprising.The different monochromatic light tests indicate that CuInSnS 4 is an exceptional catalyst for CO 2 photoreduction under visible light.The 13 CO 2 isotope experiment further validates that CH 4 product is generated from the photoreduction of CO 2 molecules, where only 13 CH 4 is detected when the reaction is carried out in a 13 CO 2 atmosphere, as shown in Fig. 3d.Meanwhile, the 13 CO 2 isotope also confirms that the generated CO was indeed a product of CO 2 photoreduction.When the reaction is conducted in a 13 CO 2 atmosphere, the weak peak of 13 CO with a m/z = 29 was detected due to the low activity of CuInSnS 4 for CO evolution (Supplementary Fig. 17a).It is noteworthy that the mass spectrum peak at m/z = 28 corresponds to the N 2 molecule from air interference, as evident from the distinct N 2 peaks in chromatogram.Additionally, the 12 CO 2 experiment further evidences that the generated CO results from the photoreduction of CO 2 molecules with no peaks at m/z = 29 (Supplementary Fig. 17b).
Although the pristine CuInSnS 4 only exhibits a yield of CH 4 evolution of 6.53 μL h −1 (corresponding to 5.83 μmol h −1 g −1 ) with a selectivity of 67.3%, the activity and selectivity of CH 4 evolution on CuInSnS 4 can be improved by coupling with semiconductor photocatalysts or noble metals as cocatalysts.We evaluated the photocatalytic performance of CuInSnS 4 modified with Pt, CoO, NiO, Co(OH) 2 , and dual co-catalysts Pt and Co(OH) 2 , for the CO 2 reduction reaction.The composition and chemical states of Pt, CoO, NiO, Co(OH) 2 cocatalysts are well verified by XRD patterns and XPS spectra (Supplementary Figs.18-23).Table S2 lists a comparison of the photoreduction activity of CuInSnS 4 and co-catalyst-modified CuInSnS 4 photocatalysts, along with the common metal sulfide systems currently used for CO 2 photoreduction.Clearly, both the yield and selectivity of CH 4 evolution on CuInSnS 4 can be significantly improved by modifying with cocatalysts such as Pt, CoO, NiO, and Co(OH) 2 .The activity of CH 4 evolution on the modified CuInSnS 4 photocatalysts surpasses the majority of the reported photocatalysts for CO 2 reduction up to now.Particularly, the incorporation of Co(OH) 2 as a cocatalyst significantly enhances the CO 2 photoreduction activity of the CuInSnS 4 photocatalyst.As the Co(OH) 2 loading increases, the photoreduction activity of CO 2 exhibits a characteristic volcanic pattern.With 5%Co(OH) 2 loading onto CuInSnS 4 , the production rates for CH 4 and CO, respectively, reach 145.45 and 32.32 μmol h −1 g −1 , corresponding to a CH 4 selectivity of 81.8%.The generation rates of CH 4 and CO are 25 times and 13 times that of pure CuInSnS 4 , respectively.Furthermore, when CuInSnS 4 is modified with a dual co-catalyst of 5% Co(OH) 2 as an oxidation co-catalyst and 1%Pt as a reduction co-catalyst, CH 4 production reaches 195.60 μmol h −1 g −1 , along with 22.00 μmol h −1 g −1 of CO, and a CH 4 selectivity of 89.9%.Photoelectrochemical characterization was employed to assess the separation efficiency of photogenerated carriers on the modified CuInSnS 4 , as shown in Supplementary Fig. 24.Clearly, the CuInSnS 4 samples modified with the co-catalyst exhibit a higher photocurrent signal and a smaller electrochemical impedance radius as compared to the parent CuInSnS 4 sample.The photocurrent increases sequentially in the order of CuInSnS 4 < 5%Co(OH) 2 /CuInSnS 4 ≈ 1%Pt/CuInSnS 4 < 5%Co(OH) 2 / CuInSnS 4 /1%Pt, indicating that the modification of the dual co-catalyst improves photoelectric carrier separation and migration compared to the single co-catalyst.The NiO and CoO cocatalysts also improve the separation efficiency and migration rate of the photogenerated carriers of CuInSnS 4 photocatalyst.The decreasing order of electrochemical resistance radius is CuInSnS 4 > 1%Pt/CuInSnS 4 ≈ 5%Co(OH) 2 / CuInSnS 4 > 5%Co(OH) 2 /CuInSnS 4 /1%Pt, consistent with the photocatalytic activity trend.Additionally, both 10%NiO/CuInSnS 4 and 10% CoO/CuInSnS 4 exhibit smaller electrochemical resistance radii than pure CuInSnS 4 , confirming that the cocatalyst promotes the photogenerated charge separation and migration.Therefore, the modification of CuInSnS 4 with various cocatalysts to enhance the separation of photogenerated carriers is related to the activity and selectivity of CH 4 products.The apparent quantum yield is calculated by measuring the yield of CH 4 and CO in 5%Co(OH) 2 /CuInSnS 4 /1%Pt under monochromatic light at 400 nm.Supplementary Fig. 25 shows the spectrum and intensity of monochromatic light at 400 nm.Under 400 nm monochromatic light irradiation, the apparent quantum efficiencies for CH 4 and CO are 0.16% and 0.01%, respectively.Based on the above analysis, we believe that CuInSnS 4 nano-single crystal photocatalysts through further optimization design of the different contents and types of cocatalyst modification can be more efficient and selective for CO 2 reduction in the future.

Energy band and photoelectrochemical characterization
The band energy potential is a key determinant of the driving force of redox reactions.Therefore, we have studied the band structure of the catalyst through UV-vis DRS and XPS valence band spectroscopy.As shown in Fig. 4a and Supplementary Fig. 26, the optical absorption band edge of CuInSnS 4 is calculated to be 787.5 nm, which corresponds to a band gap energy of 1.57 eV.For comparison, the absorption band edges of single metal sulfide Cu 2 S, SnS 2 , and In 2 S 3 are 747.0nm, 552.7 nm, and 641.6 nm, corresponding to the band gap of 1.66 eV, 2.24 eV, and 1.93 eV, respectively.Moreover, the valence band potential of CuInSnS 4 is determined to be 0.50 V from the valence band XPS spectra (Supplementary Fig. 27), while the Cu 2 S, SnS 2 , and In 2 S 3 possess valence band potentials of 1.02 V, 1.70 V, and 1.93 V, respectively.By using the formula E CB = E g -E VB , we have determined that the conduction band potentials of CuInSnS 4 , Cu 2 S, SnS 2 , and In 2 S 3 are −1.15V, −1.10 V, −0.54 V, and −0.55 V, respectively.Based on the optical band gaps, we have obtained the electronic band energies relative to a normal hydrogen electrode (Fig. 4b), indicating that both CuInSnS 4 and single metal sulfides have the ability to reduce CO 2 to CH 4 and CO.Notably, CuInSnS 4 exhibits the highest reduction potential for the photogenerated electrons to reduce CO 2 .Additionally, CuInSnS 4 shows a significantly increased photocurrent density (0.014 mA cm −2 ) compared to the single metal sulfide under visiblelight irradiation (Fig. 4c), indicating a more efficient separation of the photoinduced charge in multi-metal sulfides.The lower interface resistance in the corresponding electrochemical impedance spectra (Fig. 4d) confirms the rapid transfer of photogenerated electrons in CuInSnS 4 .The efficient separation efficiency and migration rate of photogenerated carriers make polymetallic sulfides exhibit higher photocatalytic CO 2 reduction performance compared to monometallic sulfides.However, the higher migration and separation efficiency of charge carriers in CuInSnS 4 alone is not sufficient to explain the substantial difference in the product selectivity between CuInSnS 4 and single metal sulfides.

The in situ CO 2 adsorption FT-IR spectra and mechanism
To understand the CO 2 reduction process over CuInSnS 4 and single metal sulfides, in situ Fourier-transform infrared spectroscopy is used to compare the reaction intermediates on the catalyst surface.No macroscopic infrared absorption peaks for intermediates are found on Cu 2 S or SnS 2 , even under light irradiation, possibly due to their weak chemical interaction with CO 2 (Supplementary Fig. 28a, b).However, In 2 S 3 shows a significant activation effect on the CO 2 adsorbed on the surface under light irradiation (Fig. 5a).Notably, CO 2 can form chemical adsorption with In 2 S 3 even under dark conditions, as indicated by the infrared peak at 1150 cm −1 , which can be considered an O-S stretching vibration 43 , suggesting that the oxygen atom of CO 2 is bonded to the sulfur atom of In 2 S 3 .Upon light irradiation, some infrared peaks of the produced intermediates on the catalyst surface are observed.The infrared peak at 1225 cm −1 is attributed to the vibration of bidentate bicarbonate 44 , while the infrared peak at 1412 cm −1 is attributed to the vibration of monodentate bicarbonate 45 .Most importantly, the infrared peak at 1610 cm −1 is attributed to the *COOH group, which is generally regarded as the crucial intermediate for CO 2 reduction to CO 46 .Surprisingly, the polymetallic sulfide CuInSnS 4 exhibits strong chemisorption of CO 2 and strong physisorption of H 2 O (Fig. 5b).The prominent infrared peak observed at 1627 cm −1 is attributed to the physical adsorption of H 2 O.However, In 2 S 3 does not exhibit a noticeable infrared adsorption peak of H 2 O at 1627 cm −1 .This is because In 2 S 3 has a lower affinity towards water adsorption than CuInSnS 4 , as indicated by the high contact angle on In 2 S 3 surface (Supplementary Fig. 29).Moreover, a prominent infrared peak at 1117 cm −1 assigned to C-S stretching vibration is observed upon the CO 2 adsorption on CuInSnS 4 47-49  activation are significantly improved on CuInSnS 4 compared to In 2 S 3 .Furthermore, the adsorption state of CO 2 on CuInSnS 4 is different from that on In 2 S 3 , as the carbon atom of CO 2 is bonded to the sulfur atom of CuInSnS 4 .Under light conditions, CuInSnS 4 produces specific CO 2 -activated intermediates, as indicated by the infrared peaks at 1225 cm −1 and 1260 cm −1 , attributed to the vibration of bidentate bicarbonate and bidentate carbonate 50,51 , respectively.Additionally, the infrared peaks at 1100 cm −1 and 1160 cm −1 are attributed to the absorption peaks of *CHO and *CH 3 O, which are intermediates for the yield of CH 4 52,53   .Therefore, either the CO 2 adsorption state in darkness or the produced intermediates under light irradiation show that the sites for CO 2 adsorption and the CO 2 reduction approach differ between single-metal sulfide In 2 S 3 and multi-metal sulfide CuInSnS 4 .This may account for the different selectivity of products between In 2 S 3 and CuInSnS 4 .
It has been demonstrated that sulfur defect sites in multiple metal sulfides acted as an active center for the selective photoreduction of CO 2 to CH 4 .However, CuInSnS 4 nano-single crystal shows no significant sulfur defect signals compared to the strong defect signals in SnS 2 and In 2 S 3 , as shown in Supplementary Fig. 30.This indicates that the selectivity of CH 4 products on CuInSnS 4 is not related to sulfur defects.The mechanism of selective photocatalytic CO 2 reduction on CuInSnS 4 and In 2 S 3 photocatalysts is further theoretically studied.Firstly, we investigated the adsorption behavior of CO 2 on the surfaces of In 2 S 3 and CuInSnS 4 .The (0 0 1) crystal plane of the In 2 S 3 sample and the (1 1 1) crystal plane of the CuInSnS 4 sample were selected as models, and all atoms on the crystal planes were considered as potential sites for CO 2 adsorption activation (Supplementary Figs. 31  and 32).Based on CO 2 adsorption energy, the optimal adsorption models of CO 2 on In 2 S 3 and CuInSnS 4 photocatalyst surfaces are optimized (Supplementary Figs.33 and 34).S1).Different adsorption configurations may be the key to determining the direction of electron transfer and thus the selectivity of CO 2 reduction on CuInSnS On the In 2 S 3 surface, there is extensive charge depletion for the C-O 2 and S-O 2 bonds, which implies that these chemical bonds are weakened.In contrast, CO 2 exhibits a wide charge accumulation region on CuInSnS 4 , leading to the formation of a strong C-S and In-O chemical bond.This strong interaction is beneficial for the firm adsorption of CO 2 on the CuInSnS 4 surface, promoting the further deep reduction reaction.The Bader charges analysis further confirms that there is more charge transfer between the CuInSnS 4 surface and CO 2 molecules adsorbed on it.The surface of In 2 S 3 and CuInSnS 4 loses 0.28e and 0.32e, respectively, after CO 2 adsorption (Table S3).
The stable configuration of CO 2 adsorption on CuInSnS 4 determines its excellent CO 2 photoreduction activity and selectivity.Therefore, DFT calculations were further carried out to study the conversion pathway of CO 2 on the CuInSnS 4 photocatalyst surface, as shown in Fig. 6.In Fig. 6a In summary, a CuSnInS 4 nano-single-crystal photocatalyst with exposed (1 1 1) facets is successfully prepared by a simple one-step hydrothermal reaction.Under visible-light irradiation, the CuSnInS 4 nano-single crystal photocatalyzes the conversion of CO 2 and H 2 O into main products of CH 4 with a generation rate of 6.53 μL h −1 , significantly higher than that of single metal sulfides (In 2 S 3 , Cu 2 S, and SnS 2 ).The electron-poor center sulfur atom on the CuSnInS 4 (1 1 1) crystal plane acts as the site for CO 2 adsorption and activation, which leads to the activation of the two symmetrical C=O double bonds of CO 2 molecule to form a stable S-C-O-In transition state.This induces CH 4 generation via the conversion route of COOH*→HCOOH*→H 2 CO*→H 3 CO*→CH 4 *.However, the asymmetric activation of CO 2 by monometallic sulfides is more likely to result in the cleavage of individual C-O bonds in the CO 2 molecule, leading to the preferential photoreduction of CO 2 to CO.This work provides a distinctive understanding of catalysts for CO 2 adsorption and activation for the CO 2 selective conversion to help the conversion of CO 2 resources into high-value-added products.

Preparation of CuInSnS 4 nanocrystal
A simple one-step hydrothermal method was used to synthesize CuInSnS 4 nanocrystal photocatalyst with the cubic crystal structure.The detailed operation process is as follows.Firstly, 1 mmol of CuCl, 1 mmol of SnCI 4 •5H 2 O, and 1 mmol of InCI 3 •4H 2 O were added to 40 mL of deionized water to form a solution under vigorous stirring.Then, 5 mmol of TAA was dissolved in the above-mixed solution and stirred at room temperature for 30 min.Finally, the resulting mixed solution was transferred to a 50 mL Teflon-lined autoclave and sealed into a stainless steel tank for hydrothermal reaction.The hydrothermal temperature is controlled at 160, 180, and 200 °C for 24 hours.After the reaction, the product was collected and washed with deionized water, and dried under vacuum at 60 °C.The obtained samples were labeled CuInSnS 4 (160 °C), CuInSnS 4 (180 °C), and CuInSnS 4 (200 °C) according to the reaction temperature.The detailed preparation processes of CuInSnS 4 modified Pt, CoO, NiO, and Co(OH) 2 cocatalysts can be seen in supplementary information.

Characterization
X-ray diffractometer (D8 Advance, Bruker) was used to analyze the crystal structure of the catalyst.The XRD test range is 10°−80°, and the scan rate is 10°min −1 .Scanning electron microscopy (su8010, Hitachi) was used to observe the surface morphology of the catalyst.The element composition and ratio of the sample are detected by EDS.The apparent morphology and high-resolution TEM image of the catalyst were tested by transmission electron microscope (TEM, TECNAI G2F20, FEI Company).At the same time, SAED and element mapping images of the catalyst were obtained in the TEM measurement mode.A UV-VIS-NIR Spectrophotometer (Cary 500) was used to obtain the catalyst UV-VIS-NIR DRS, in which BaSO 4 was used as a standard sample for 100% light transmission.The Micromeritics 3020 M physical adsorption instrument was used to obtain the nitrogen adsorption and desorption curves of different catalysts.The specific surface area of each catalyst was calculated from the type of nitrogen adsorption and desorption curves.The catalyst and dried potassium bromide were evenly ground, and 20 mg was weighed and pressed into slices, then placed in a quartz infrared tube for a carbon dioxide adsorption infrared test.In situ infrared spectra measurements were performed using a Fouriertransform infrared spectrometer (Nicolet iS50 FT-IR Spectrometer) equipped with a mercury cadmium telluride detector (Supplementary Fig. 37).In situ infrared spectra were recorded by averaging 32 scans at a resolution of 4 cm −1 .To initiate the experiment, the catalyst was placed in a 250 mL quartz tube and compacted into a film.The tube was then subjected to vacuum treatment for 60 min.Subsequently, high-purity CO 2 gas was introduced, and the quartz tube was sealed.A liquid sampler was used to inject 60 μL of deionized water into the sealed quartz tube.The tube was heated with a hot blower to vaporize the deionized water.The quartz tube was positioned in the FT-IR spectrometer, ensuring that the incident light of the infrared spectrometer was perpendicular to the sample surface.A xenon lamp visible-light source was introduced to directly illuminate the sample surface.Infrared spectra were recorded after pretreating the catalyst in a vacuum, introducing CO 2 gas, and vaporizing deionized water, respectively.After the introduction of light, infrared spectra were recorded every 5 min.Gas chromatography-mass spectrometry (Agilent 7890B, Agilent 5977B MSD) was used to analyze 13 CH 4 and 13 CO.Electron paramagnetic resonance spectroscopic measurements were performed at room temperature using a Bruker A300 EPR spectrometer.
Photocatalytic performance 300 W xenon lamp (Microsolar 300, Beijing Perfectlight Technology Co., Ltd.) was equipped with a 420 nm cutoff wavelength filter as a light source that simulates visible light for photocatalytic CO 2 reduction tests.Firstly, 50 mg of catalyst was dispersed in 5 mL of deionized water and sonicated for 10 min.Then the catalyst was dropped into a watch glass with a diameter of 3 cm and dried at 80 °C.Subsequently, the dried catalyst was placed in a quartz reactor with a volume of ~250 cm 3 , and then high-purity CO 2 gas (99.999%) was introduced to replace all the air.The flow rate of carbon dioxide gas was ~100 mL/min and lasted for 1 h.Finally, 50 μL deionized water was injected into the quartz reactor from the rubber stopper through a gas chromatography liquid syringe (maximum range, 50 μL), and the reactor was heated with a hair dryer to evaporate the water into water vapor.As a result, water is present in the form of water vapor throughout the reaction.The reactor was placed under a xenon lamp for photocatalytic reaction, and the current of the xenon lamp was 16 A.After 1 h of illumination, 0.5 mL of gas was extracted from the reactor with a 1 mL gas chromatograph gas syringe and injected into the gas chromatograph for product analysis and detection.Among them, H 2 , O 2 , and N 2 are detected by a thermal conductivity detector.CH 4 was detected by the flame ionization detector.CO passes through the flame ionization detector after being transformed by the nickel reformer.The product was qualitatively and quantitatively analyzed by gas chromatography retention time and appearance standard curve method.

Theoretical calculations
The density functional theory calculations were performed using the VASP code with the projected augmented wave method 54 .Generalized gradient approximation in the scheme of Perdew-Bueke-Ernzerhof was used for the exchange-correlation functional 54 .The PBEsol exchangecorrelation function was adopted in the optimization calculations.Grimme's DFT-D3 scheme was used to describe the long-range vdW interactions 54 .The cutoff energy of the basis function was 420 eV.For the CuSnInS 4 (1 1 1) crystal plane, a 2 × 2 × 1 supercell with a four-layer slab was constructed, and only the top two layers were allowed to relax.A vacuum region of 12 Å was set above the surface for periodic boundary conditions, and dipole correction was also applied.Gammacentered 1 × 1 × 1 grid k-points were used for the interface.Geometry relaxation was performed until the energy, and atomic forces converged to be smaller than 10 −5 eV and 0.03 eV/Å.Charge transfers were calculated using the Bader charge analysis 54,55 .
The free energy of each reaction intermediate was determined by: G = E + ZPE-TS.The electronic energy was directly obtained from DFT calculations.The zero-point energy and entropy correction (TS, T = 298.15K) were computed from vibration analysis according to standard methods.The adsorption-free energy of the adsorbates can be calculated using: ΔGads = ΔEads + ΔZPE-TΔS, where ΔEads is the adsorption energy of the adsorbates, and ΔZPE and ΔS are the difference between ZPE and S, respectively.After the adsorption-free energies of the adsorbates are obtained, the reaction-free energies of CO 2 reduction reaction elementary steps can be determined correspondingly by using the computational hydrogen electrode model 40 .

Fig. 1 |
Fig. 1 | The characterization of crystal phase, composition, and morphology.a XRD patterns of several metal sulfides, b EDS spectra of CuInSnS 4 photocatalyst, c a TEM image of CuInSnS 4 sample, d an HRTEM image, and e an SAED pattern of the CuInSnS 4 sample, as well as f-i EDS elemental mapping images.
2 S 3 crystals, the In atom is present in [InS 4 ] tetrahedron and [InS 6 ] octahedron coordination (Supplementary Fig. 4).Cu and Sn atoms exist as [CuS 4 ] tetrahedron and [SnS 6 ] octahedron coordination in Cu 2 S and SnS 2 crystal structure, respectively (Supplementary Figs. 5, 6).The coordination structure of each metal in CuInSnS 4 coincides with the individual Cu 2 S, In 2 S 3 , and SnS 2 .The composition of CuInSnS 4 is analyzed by energy-dispersive spectroscopy, as shown in Fig. 1b.EDS offers that the Cu, Sn, In, and S atomic ratio is ~1:1:1:4, very close to the stoichiometric value of CuInSnS 4 compounds, indicating the high purity of CuInSnS 4 nanocrystal.Meantime, the ICP-MS test results of the metal element content of the nano-single crystal CuInSnS 4 sample are presented in Table

Fig. 3 |
Fig. 3 | Photocatalytic conversion performance of CO 2 and H 2 O. a The production rate of CH 4 , CO, and H 2 in various photocatalysts under visible-light conditions.b Photocatalytic CO 2 reduction stability test of CuInSnS 4 sample.c Photocatalytic performance of CuInSnS 4 sample under monochromatic light irradiation.d GC-MS spectra of 13 CH 4 generated from 13 CO 2 .

Fig. 5 |
Fig. 5 | In situ CO 2 adsorption FT-IR spectra and differential charge density map as well as free energy diagrams of CO 2 photoreduction to CH 4 .In situ FT-IR spectra of CO 2 adsorbed on a In 2 S 3 and b CuInSnS 4 .CO 2 adsorption configuration and differential charge density map of c In 2 S 3 and d CuSnInS 4 photocatalysts.In the differential charge density map, the yellow and blue regions indicate electron accumulation and depletion, respectively.

Fig. 4 |
Fig. 4 | Characterization of energy bands and optoelectronic properties.a UVvis absorption spectrum of various metal sulfides.b The optical band gap energy (E g ) of the corresponding CuInSnS 4 and various single metal sulfides.c Photocurrent response and d electrochemical impedance spectroscopy of the asprepared samples.
Figure 5c, d depict schematic diagrams of CO 2 stable adsorption configurations and the charge density difference of CO 2 on In 2 S 3 and CuInSnS 4 , respectively.The stable CO 2 adsorption configuration on In 2 S 3 is the C atom of CO 2 bonded to In atom with a bond length of 2.27 Å, while the O atom of the CO 2 molecule is bonded with the S atom with a bond length of 1.70 Å.The CO 2 adsorption model for the polymetallic sulfide CuInSnS 4 is the opposite.The unique C-S bond with a bond length of 1.90 Å is formed between the C atom of CO 2 and the surface S atom, and one O atom of CO 2 is bonded with In atom with a bond length of 2.23 Å.The various adsorption configurations of CO 2 on In 2 S 3 and CuInSnS 4 surfaces are attributed to the coordination environment and charge number of the surface S atoms.The surface S atom of In 2 S 3 is an electron-rich site with [SIn 3 ] coordination structure, while the S atom on CuInSnS 4 is an electron-poor center with [SInSnCu] coordination structure (Supplementary Fig. 35 and Table 4 and In 2 S 3 .Adsorption of CO 2 on the In 2 S 3 (0 0 1) crystal plane leads to inconsistent changes in two C=O lengths.The length of the C-O 2 bond is 1.43 Å, equal to the ordinary C-O (1.43 Å) single bond, while the length of C-O 1 is shortened to 1.19 Å, close to the length of C-O (1.12 Å) in a CO molecule.This asymmetric activation is more likely to cause the rupture of C-O 2 , thereby preferentially producing CO on In 2 S 3 .In the case of the CuInSnS 4 (1 1 1) crystal plane, both C-O bonds are similar in length, measuring 1.26 ± 0.02 Å.They are longer than the C-O bond (1.16 Å) in a free CO 2 molecule, indicating that the bond energy of two C=O bonds of the activated CO 2 is simultaneously weakened.The calculation of the charge density difference reveals the difference in electronic structure and electron flow resulting from the interaction of CO 2 with the surface atoms of In 2 S 3 and CuInSnS 4 .

Fig. 6 |
Fig. 6 | DFT calculations of the CO 2 conversion pathway.a Calculated adsorption configuration of CO 2 and reactive intermediates on CuSnInS 4 .b Gibbs free energy diagrams for CO 2 reduction to CH 4 on CuSnInS 4 .