Introduction

Solar-driven CO2 reduction to yield high value-added fuels or chemicals is considered as a highly attractive and promising approach to ameliorate global warming issues associated with enduring CO2 emissions1,2,3,4,5. However, efficient CO2 photoreduction with high selectivity and stability remains a critical issue because of the high dissociation energy of the C=O bond (ca. 750 kJ mol−1) and the kinetically sluggish multiple-electron-transfer reaction process6,7,8. The CO2 concentration in the exhaust gases discharged from industries, such as coal-fired power stations is relatively low, restricting practical implementation of CO2 photoreduction9,10,11. Designing an effective strategy to boost the low-concentration CO2 molecular activation and improve the CO2 conversion efficiency has become the key challenges in CO2 photoreduction.

Metal sulfides have tunable surficial electronic structures owing to their adjustable metal valence states, which have been extensively applied in photocatalytic CO2 reduction reaction12,13,14,15,16,17,18,19. Two-dimensional (2D) metal sulfides exhibit high specific surface areas, which shall benefit photocatalytic reactions by offering a large number of catalytically active sites. Moreover, the ultrathin architecture of these 2D materials can also significantly shorten the migration pathways of photogenerated charge carriers, thereby substantially reducing electron-hole recombination and improving efficiency of charge carrier separation. To further improve the performance of metal sulfide photocatalysts in photocatalysis, tuning electronic structure of metal sites by strain engineering provides an effective strategy20,21,22,23,24,25,26,27,28,29,30,31. Strain engineering of nanomaterials can be mainly achieved by engendering interfacial mismatch or deformation. These methods can induce long-range lattice strain but are not applicable for manipulating the strain environment around specific atomic sites in a short range32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47. On the other hand, modulating chemical environment around metal cations may provide an innovative strategy to adjust strain on the atomic scale (Fig. 1a).

Fig. 1: Structural characterizations.
figure 1

a Schematic illustration of atomically strained metal-site catalyst. b TEM image. c HAADF-STEM image and the corresponding elemental mapping images. d HRTEM image and e the corresponding strain mapping image along the εxx direction by GPA. f XANES and g FT-EXAFS spectra recorded in the In K-edge. h WT-EXAFS plots of In2S3, In2O3, strained In2S3 and oxidized In2S3. Source data for Fig. 1f,g are provided as a Source Data file.

Full size image

In this work, we developed a wet-chemistry method to introduce oxygen coordination and sulfur vacancy into In2S3, which induced a small-domain strain at the atomic scale (strained In2S3, Fig. 1a). The as-prepared atomically strained In2S3 exhibited approximately 9 times higher photocatalytic activity than pristine In2S3, towards converting CO2 to CO in ambient air under visible light irradiation. Based on a collection of in-situ spectroscopic measurements and density functional theory (DFT) calculations, the atomically strained In2S3 was found capable to promote CO2 adsorption/activation and at the same time boost photogenerated charge carrier separation, as a result, greatly enhancing CO2 photoreduction performance under full spectrum/visible light illumination in both pure CO2 and ambient air environment.

Results and discussion

Structural characterizations of layered In2S3 with atomically strained Indium sites

Oxygen coordination and sulfur vacancy were introduced into In2S3 via a facile one-step solvothermal reaction, in which the induced strain degree in In2S3 could be well controlled by the volumetric ratio of water and ethylene glycol used in the solvothermal reaction (Supplementary Fig. 1). X-ray diffraction (XRD) patterns (Supplementary Fig. 2) of the as-prepared pristine In2S3 and strained In2S3 belong to cubic-phase In2S3 (PDF #65-0459)46. The (311) diffraction peak of strained In2S3 displayed a shift to lower 2θ angles as compared to pristine In2S3, indicating possible presence of local structural disorder and strain in the strained In2S3. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images as displayed in Supplementary Fig. 3, Fig. 1b and Supplementary Fig. 4 showed that both of the pristine In2S3 and strained In2S3 exhibited flower-like morphology assembled from thin nanosheets. Elemental mapping images indicated homogeneous distribution of In, S and O elements (Fig. 1c) in strained In2S3. As illustrated in high-resolution TEM (HR-TEM) image (Fig. 1d and Supplementary Fig. 5), the strained In2S3 possessed moderate crystallinity with a fine lattice spacing of 0.325 nm, assignable to the (311) plane of cubic In2S3. Figure 1e and Supplementary Fig. 6 showed the strain maps obtained by geometric phase analysis (GPA). Figure 1e displayed the strain distribution around the lattice disorder defects (white circles in Fig. 1d), which showed widespread compression strain around 7%, with some regions reaching 10%. The marked arrows indicated the atomic-level distortion at the lattice, confirming the atomically strained surficial In sites33. Supplementary Fig. 7 showed the strain distribution around the lattice disorder areas of oxidized In2S3. The oxygen atoms partially replaced the In-S bonds in layered In2S3, which in turn creates significant tensile stress. For comparison, pristine In2S3 and In2O3 were also analyzed by GPA, and both of pristine In2S3 and In2O3 displayed good crystallinity with ultrathin sheet-like morphologies as presented in Supplementary Fig. 8-9. X-ray photoelectron spectroscopy (XPS) spectra (Supplementary Fig. 10) indicated decreased S 2p XPS peak and enhanced O 1 s XPS peak in strained In2S3 as compared to pristine In2S3, suggesting that strain creation was associated with In-O bond formation. The same conclusion can be drawn from the observed shift of S 2p and In 3 d XPS peaks to higher binding energies in strained In2S312. It is observed that the intensity of O 1 s XPS peak gradually reduced with increasing Ar+ sputtering time (Supplementary Fig. 11), corroborating that the O2− ions were introduced into the surface of strained In2S3 by replacing S2− ions in In-S bonds48,49.

The atomic and electronic structures of different In2S3 samples were further examined by X-ray absorption spectroscopy (XAS). All samples showed similar pre-edge peaks at about 27935 eV in the In K-edge X-ray absorption near-edge structure (XANEs) spectra, suggesting similar In atomic arrangements (Fig. 1f). As displayed in the Fourier-transformed (FT) In K-edge extended X-ray absorption fine structure (EXAFS) spectra (Fig. 1g and Supplementary Table 1), the strained In2S3 sample showed a reduced In-S bond length (1.96 Å) as compared to that of pristine In2S3 sample (1.99 Å). Moreover, wavelet transform (WT) EXAFS plots in both K space and R space (Fig. 1h and Supplementary Fig. 12-17) show different In-O and In-S coordination in strained In2S3 as compared to reference samples. An obviously stronger electron paramagnetic resonance (EPR) signal at g = 2.002 was observed over strained In2S3 than pristine In2S3 (Supplementary Fig. 18), indicating a higher S-vacancy content in strained In2S3. A similar EPR signal was reported in defective metal sulfide12. All of the above characterization results indicate presence of In-O bonds together with sulfur vacancies in strained In2S3.

The light absorption properties of the as-prepared In2S3 catalysts were studied by UV–vis absorption spectroscopy. As shown in Fig. 2a, compared with pristine In2S3, the strained In2S3 displayed enhanced light absorption ability. The bandgap of the strained In2S3 was calculated to be about 1.76 eV (Supplementary Fig. 19), which was slightly smaller than that of pristine In2S3 (1.82 eV). Combined with ultraviolet photoelectron spectroscopy (UPS) measurements (Supplementary Fig. 20), it was inferred that the conduction band (CB) minimum of pristine In2S3 and strained In2S3 was located at −0.33 and −0.41 eV, respectively, as illustrated in Fig. 2b. It is important to note that the downshifted CB edge position for strained In2S3 can still meet the thermodynamic demand for CO2 reduction50,51,52. As an important consideration for photocatalysis, the separation and transfer of photogenerated charge carriers were further investigated. As revealed in Fig. 2c, the strained In2S3 exhibited the lowest photoluminescence (PL) peak intensity relative to pristine In2S3, oxidized In2S3 and In2O3, indicating its lowest photoinduced charge carrier recombination rate. The same conclusion could be deduced from the surface photovoltage (SPV), transient photocurrent and electrochemical impedance spectroscopy measurements (Fig. 2d and Supplementary Fig. 21). Femtosecond transient absorption (fs-TA) spectroscopy was further performed to study the charge carrier dynamics. Figure 2e-f compared two-dimensional (2D) pseudocolor TA spectra for pristine In2S3 and strained In2S3, which displayed two distinctive absorption bands at 450-550 nm and 650-700 nm, respectively. Typical photobleaching peaks are observed in the pseudocolor plots of In2S3 (~450 nm) and strained In2S3 (~480 nm). These signals were assigned to the ground-state bleaching (GSB), reflecting the excited state relaxation. Figure 2g showed the fs-TA kinetic profiles at 475 nm for In2S3 and strained In2S3. With increasing probing time, an obvious decrease in TA intensity was detected, indicating a decrease of active photogenerated charges (Supplementary Fig. 22). The fs-TA spectra of In2S3 and strained In2S3 could be well fitted by a tri-exponential function and the fitting parameters are τ1 = 3.26 ps, τ2 = 30.06 ps, τ3 = 69.82 ps for In2S3, and τ1 = 0.21 ps, τ2 = 5.15 ps, τ3 = 48.35 ps for strained In2S3. The short-lived τ1 corresponded to electron trapping by electron trapping state (e-TS), while the long-lived τ2 and τ3 were related to electron transition and recombination, respectively. Kinetic fitting gave the decay time of the ground-state bleaching (GSB) signal at 42.85 and 25.05 ps for In2S3 and strained In2S3, respectively, based on a triexponential decay function. The shorter GSB decay time in strained In2S3 resulted from charge carrier transfer to the atomically strained sites. The above results provide concerted evidences that the strained In sites can promote rapid transfer of photogenerated electrons.

Fig. 2: Insights into charge transfer dynamics.
figure 2

a UV–vis diffuse reflectance spectroscopy (DRS) spectra and b band structure of pristine In2S3, strained In2S3 and In2O3. c Steady-state PL spectra and d SPV signals over pristine In2S3, strained In2S3, oxidized In2S3, In2O3 and In2S3-x. 2D transient absorption pseudocolor plots for e pristine In2S3 and f strained In2S3. g The decay signals of pristine In2S3 and strained In2S3. Source data for Fig. 2a,c,d,g are provided as a Source Data file.

Full size image

Photocatalytic CO2 reduction performance

The CO2 photoreduction performance of strained In2S3 was evaluated under full spectrum or visible light ( > 420 nm) irradiation in pure CO2 without presence of any sacrificial reagents. The amount of photocatalyst used was first optimized in the photocatalytic CO2 reduction reaction (Supplementary Fig. 23). Besides, we have conducted additional control experiments using pure Ar to replace CO2 as the feeding gas. No CO2 reduction products could be identified (Supplementary Fig. 24), eliminating the influence of carbon contamination during photoreaction. Figure 3a and Supplementary Fig. 25 show that the CO and CH4 yields over strained In2S3 can reach 397.2 and 44.0 μmol gcatalyst−1 (204.9 and 20.8 μmol gcatalyst−1) after 4 h of photoreaction under full spectrum (visible) light illumination. The yield of CO is ca. 8.5 higher than that over pristine In2S3. In the meanwhile, water was oxidized to evolve molecular oxygen (Supplementary Fig. 26). As displayed in Fig. 3b, the strained In2S3 could still maintain appreciable CO and CH4 yields under full spectrum and visible light irradiation (Supplementary Fig. 27) even if the reaction environment was switched from pure CO2 to atmospheric CO2 far superior to pristine In2S3 (Supplementary Fig. 28-29). The CO and CH4 products were indeed produced via CO2 photoreduction, verified by the 13CO2 labeling experiments (Supplementary Fig. 30), and the presence of molecular O2 had little influence on the CO2 photoreduction reaction (Supplementary Fig. 31). The apparent quantum efficiency (AQE) of strained In2S3 was estimated and the result was shown in Supplementary Fig. 32, which well resembled the UV–vis absorption spectrum, reaching a maximum AQE of around 0.45% at 400 nm. The surface temperature of the strained In2S3 under visible light irradiation was at around 70 °C, measured using a thermographic technique (Supplementary Fig. 33). The control experiments disclosed that the CO2 reduction reaction was completely terminated in dark at 70 °C or even at as high as 105 °C (Supplementary Fig. 34). Furthermore, when the photogenerated heat was removed by ice water, the strained In2S3 can still maintain nearly the same photocatalytic performance (Supplementary Fig. 34). These results reaffirmed the photoinduced CO2 reduction reaction over strained In2S3. Besides high photocatalytic activity, the strained In2S3 also displayed decent photocatalytic stability: no significant photocatalytic activity decay was noticed after ten consecutive photoreaction cycles (Fig. 3c and Supplementary Fig. 35). The structure of the strained In2S3 could be well maintained after ten cycles of photoreaction based on a collection of characterizations including XRD, TEM, EPR and XPS (Supplementary Fig. 36), indicating its excellent stability in photoreaction. Compared to the reported In-based photocatalysts in literature, the strained In2S3 exhibited an obviously better photocatalytic CO2 reduction performance in both pure and dilute CO2 environment (Supplementary Table 2-3).

Fig. 3: Photocatalytic CO2 reduction performance.
figure 3

Product yields in CO2 photoreduction reaction over various In2S3 photocatalysts conduced in a pure CO2 and b ambient air. Error bars represent the standard deviation of three independent measurements. c Cyclic photocatalytic CO2 reduction reaction conducted in ambient air for 40 hours. In-situ DRIFTS spectra recorded over strained In2S3 d under dark condition in pure CO2, e under dark condition in ambient air, f under full spectrum irradiation condition in pure CO2, g under visible light irradiation condition in pure CO2, and h under visible light irradiation condition in ambient air. i The in-situ FTIR spectra in the wavenumber range from 2020 to 2100 cm-1 recorded over strained In2S3 under full spectrum irradiation condition in pure CO2 (left) and under visible light irradiation condition in diluted CO2 (right). Source data for Fig. 3a-i are provided as a Source Data file.

Full size image

To understand the CO2 photoreduction reaction mechanism over strained In2S3, in-situ Fourier transform infrared spectroscopy (FTIR) measurements were performed to probe the reaction intermediates. Carbonate species were found existing on the surface of strained In2S3 in humid CO2 atmosphere under dark condition (Fig. 3d). The absorption peak at 1636, 1431 and 1287 cm−1 could be attributed to *CO2, *HCO3- and b-CO32−, respectively. The observation of carbonate species and *CO2 suggested that CO2 could be adsorbed and activated over strained In2S3 under dark condition. Lowering CO2 partial pressure reduced the intensities of infrared characteristic peaks (Fig. 3e). Compared with dark condition, several new absorption peaks at around 1576 cm−1, 1135 cm−1 and 1070 cm−1 appeared under full spectrum (Fig. 3f) or visible-light (Fig. 3g) irradiation and their intensities increased with increasing light irradiation time, due to increased coverage of *COOH (a crucial intermediate for CO2 reduction to CO or CH4), *CH3O and *CHO (the important intermediates for CO2 reduction to CH4), respectively. Besides, the *CO2 (1642 cm−1), *HCO3- (1415 and 1375 cm−1) and b-CO32− (1252 cm−1) absorption peaks also gradually strengthened with elongating light illumination time. Switching reaction environment from pure CO2 to ambient air reduced the coverage (Fig. 3h) and changed the adsorption strength of CO2 reduction intermediates. The *CO absorption peak at around 2076 cm−1 was clearly detected over strained In2S3 in pure CO2 atmosphere under both visible and full spectrum light irradiation (Fig. 3i left), suggesting relatively strong *CO adsorption over strained In2S3 surface, which should be beneficial for the subsequent steps of protonation towards CH4 formation. In ambient air environment, no obvious band near 2076 cm−1 was detected over strained In2S3, which explained the observed selective CO2 photoreduction to CO.

Insight of the increased photocatalytic activity

To gain further theoretical insights, density functional theory (DFT) calculations were conducted to explore CO2 adsorption and activation over strained In2S3. The strained In2S3 calculation model is shown in Fig. 4a, Supplementary Fig. 37 and Supplementary Data 1, 2 for details on the models, which was built based on a collection of characterization results. The CO2 adsorption energy on the In site of pristine In2S3 was calculated to be −1.25 eV (Fig. 4b). The CO2 adsorption energy over the In site was significantly more negative than that over the S and O site in strained In2S3, suggesting that formation of coordinatively unsaturated In sites is favorable for CO2 adsorption34. It can be observed in Fig. 4c that strained In2S3 showed a much better CO2 adsorption ability than pristine In2S3. In particular, the CO2 adsorption curve of strained In2S3 rose significantly faster in the middle pressure range. At P/P0 = 1.0, the CO2 adsorption capacities of pristine In2S3, strained In2S3, oxidized In2S3, In2O3 and In2S3-x were 0.66, 3.54, 2.57, 1.05, and 2.25 cm3 g−1, respectively. Moreover, an upshift of the p-band center from -0.72 to -0.60 eV was also observed in strained In2S3 (Fig. 4d and Supplementary Fig. 38). It is well established that the upshifted p-band center can push more antibonding states above the Fermi level, thus resulting in decreased occupation and stronger adsorbate bonding. The strained In site shifted the p-band center closer to the Fermi level and resulted in stronger surface bonding  between the active sites and CO2 molecules53.

Fig. 4: DFT study.
figure 4

a Schematic illustration showing the structural models for In2S3 and strained In2S3. b Calculated CO2 adsorption energy on pristine In2S3 and strained In2S3. c CO2 adsorption isotherms. d Calculated Bader charge and p-band-center. e CO2 TPD curves. f CO TPD curves. g Calculated free energy diagrams for CO2 reduction. h Calculated differential charge density of *COOH intermediate over pristine In2S3 and strained In2S3 surface. The insets a, g, and h show the DFT-optimized structures (In: purple, S: yellow, O: red, C: brown, and H: pink). Source data for b–f are provided as a Source Data file.

Full size image

Temperature-programmed desorption (TPD) of CO2 was performed to verify the enhanced CO2 adsorption over strained In2S3 (Fig. 4e). The CO2-TPD peak area for strained In2S3 was obviously larger than those for other In2S3 samples with a higher CO2 desorption temperature, indicating that the strained In2S3 indeed could enhance CO2 adsorption. Moreover, the CO TPD measurements indicated a concurrent stronger CO adsorption over strained In2S3, which was beneficial for *CO hydrogenation to form CH4 in CO2 photoreduction reaction (Fig. 4f). The reaction free energy and the structure configuration of each intermediate were simulated to illustrate the photocatalytic CO2 reduction reaction mechanism (Supplementary Fig. 39 and Supplementary Table 4). Of note, the *COOH formation energy on strained In2S3 (0.30 eV) was more negative than that on In2S3 (0.47 eV), indicating that the atomically strained In sites can lower the energy required for the key conversion of *CO2 to *COOH (Fig. 4g). Afterward, the downhill free energy profiles revealed that the formation of *CO from *COOH was more spontaneous over strained In2S3. Meanwhile, *CO protonation to *CHO on strained In2S3 had an energy barrier (0.48 eV) higher than that on pristine In2S3 (-0.05 eV), which can ensure a higher selectivity of CO generation on strained In2S3. As confirmed by differential charge densities over the *COOH-adsorbed structure models, higher concentrations of electrons were transferred to the near-surface region of strained In sites on strained In2S3 (Fig. 4h). Besides, water contact angle measurements (Supplementary Fig. 40) and DFT calculations (Supplementary Fig. 41) uncovered that the strained In2S3 could also tailor the rate-determining step for water reduction to boost proton generation (Supplementary Table 5).

Discussion

In summary, an atomically strained In2S3 photocatalyst has been rationally designed and successfully prepared by a wet-chemistry method, which exhibits high CO2 photoreduction performance in both pure CO2 and ambient air environment under full spectrum/visible light illumination. It is demonstrated that the strained In2S3 can significantly enhance photogenerated charge carrier separation and improve CO2 adsorption/activation as characterized by a collection of measurements and DFT calculations, as a result, greatly boosting CO2 photoreduction performance.

Methods

Materials

InCl3⋅4H2O (AR, ≥99%), thioacetamide (TAA, AR, ≥99%), In2O3 (AR, ≥99%) granular NaOH (AR, ≥99%) were purchased from Aladdin Reagent Co. Ltd. Ethylene glycol was purchased from Shanghai Chemical Reagent Co., Ltd. All chemical reagents were used without further purification and ultrapure water (18.25 MΩ cm−1) was used throughout the whole experiment.

Synthesis of In2S3

A total of 1.6 mmol of InCl3⋅4H2O was dissolved in 50 mL deionized water and stirred for 60 min. Then 2.4 mmol TAA was added and stirred for another 15 min. The obtained mixture was transferred into a 100 mL Teflon-lined stainless-steel autoclave, which was sealed and kept at 95 °C for 90 min. After the reaction, the autoclave was cooled down naturally, and the products were washed with deionized water and ethanol for several times and dried in ambient environment.

Synthesis of strained In2S3

A total of 1.6 mmol of InCl3⋅4H2O was dissolved in a mixed solvent containing 35 mL of deionized water and 15 mL of ethylene glycol and stirred for 60 min. Then, 2.4 mmol of TAA was added and stirred for another 15 min. The obtained mixture was transferred into a 100 mL Teflon-lined stainless-steel autoclave, which was sealed and kept at 95 °C for 90 min. After the reaction, the autoclave was cooled down naturally, and the products were washed with deionized water and ethanol for several times and dried in ambient environment.

Synthesis of oxidized In2S3

A total of 1.6 mmol of InCl3⋅4H2O was dissolved in 50 mL of ethylene glycol and stirred for 60 min. Then 2.4 mmol TAA was added and stirred for another 15 min. The obtained mixture was transferred into a 100 mL Teflon-lined stainless-steel autoclave, which was sealed and kept at 95 °C for 90 min. After the reaction, the autoclave was cooled down naturally, and the products were washed with deionized water and 1 M NaOH for several times and dried in ambient environment.

Synthesis of In2S3-x

The as-obtained In2S3 was rapidly heated at 300 °C for 2 min in Ar with 5% H2 and then cooled to room temperature, and the products were washed with deionized water and ethanol for several times and dried in vacuum oven overnight

Characterization

The structural information of the as-prepared samples was analyzed by powder X-ray diffraction (XRD) on a D/MAX-RB diffractometer with Cu Kα radiation. The morphologies of the samples were examined by transmission electron microscopy (AC-TEM, JEM-ARM200F) at 200 kV. X-ray photoelectron spectroscopy (XPS) spectra were recorded on a Thermo VG Multilab 2000 spectrometer with Al Kα source. Ultraviolet photoelectron spectroscopy (UPS) measurements were performed on a Thermo Fisher ESCALAB 250Xi instrument with the He I radiation (hν = 21.22 eV) source. EPR measurements were obtained using a Bruker model EMXPLUS 10/12 spectrometer. The optical properties of the samples were studied on a UV–vis-NIR spectrophotometer (UV-3600). The surface photovoltage (SPV) spectroscopy was performed on stable surface photovoltage spectrometer (PL-SPV/IPCE1000, Beijing Perfectlight). The equipment consists of a Xe lamp (CHF-XM-500 W), a monochromator (ZLolix SBP500) and a lock-in amplifier (SB830-DSP) with light chopper (SR540). The CO2/CO adsorption properties were studied by performing CO2-TPD and CO-TPD experiments on a Micromeritics AutoChem 2950HP. The CO2 adsorption-desorption isotherms were measured on a Micromeritics ASAP2460. The transient photocurrent responses and the electrochemical impedance spectroscopy (EIS) spectra of the samples were recorded on an electrochemical workstation (CHI660C Instruments, China) with a three-electrode configuration in 30 mL 0.5 M Na2SO4 (aq.), where an Ag/AgCl (saturating KCl) electrode, a Pt plate and an indium tin oxide (ITO) glass coated with the sample worked as the reference, counter and working electrode, respectively. X-ray absorption near edge structure (XANES) spectra at the In K-edge were collected at the TLS16A1 station in the National Synchrotron Radiation Research Center (NSRRC). In foil was used as the reference. Data reduction, data analysis and extended X-ray absorption fine structure (EXAFS) fitting were performed using the Athena and Artemis software packages. In-situ Fourier transform infrared spectroscopy (FTIR) measurement was performed on a Vertex 70 FTIR spectrometer (Bruker). The sample to be measured was first pretreated inside the chamber in Ar at 120 °C for 30 min to clean the sample’s surface. Afterwards, the background spectrum was then collected after the chamber’s temperature was lowered to room temperature. Next, the gas (CO2 or air 10 mL min−1, H2O/Ar 5 mL min−1) was introduced into the reactor. After reaching the adsorption equilibrium in 30 min, the background spectrum was collected again. Subsequently, the light source (Xenon lamp with fiber optics, PLS-SXE300) was turned on, and the spectrum was recorded every 5 min under full spectrum or visible light irradiation.

Photocatalytic CO2 reduction

Typically, 20 mg of photocatalyst was dispersed in 5 mL of deionized H2O under ultrasonication. The above suspension was then dropped onto a quartz disc with a diameter of 5 cm and dried at 70 °C to obtain the photocatalyst-coated quartz disc. Afterwards, the quartz disc was placed in a custom-made reaction vessel (250 mL) filled with 5 mL of deionized H2O on the bottom. Then, ~60 kPa of high purity CO2 (99.999%) or compressed air was used as the CO2 source. Light irradiation was provided by a PLS-SXE300D Xe lamp (Beijing Perfectlight) with a standard AM 1.5 G filter having an output light density of ~500 mW cm-2, and the visible light was obtained by adding a filter ( > 420 nm). Note that the distance from the lamp to the sample was ~10 cm. After adsorption equilibrium, the online gas-closed photocatalytic system (Labsolar-6A, Beijing Perfectlight) and gas chromatograph (GC-2014C, Shimadzu Corp., Japan) were used to quantify the gaseous products. The temperature of the gas-solid reactor was controlled at 298 ± 0.2 K by a recirculating cooling water system during light irradiation. The light intensity for CO2 photoreduction reaction was measured by an optical power meter (PL-MW2000, Beijing Perfectlight).

Computational details

All density functional theory (DFT) calculations were performed in Vienna Ab initio Simulation Package (VASP) within the generalized gradient approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) formulation. The projected augmented wave (PAW) potentials were chosen to describe the ionic cores and valence electrons were taken into account using a plane wave basis set with a kinetic energy cutoff at 450 eV. For the optimization of both geometry and lattice size, the Brillouin zone integration was performed with 2 × 2 × 1 Γ-centered k-point sampling. The self-consistent calculations applied a convergence energy threshold of 10-5 eV. The equilibrium geometries and lattice constants were optimized with maximum stress on each atom within 0.02 eV Å−1. In this work, a (3 × 3) supercell of In2S3 (111) and modified In2S3 (111) surface slabs were adopted, where the half bottom atomic layers were fixed for all calculations. To avoid interactions between periodic structures, ~15 Å vacuum space was inserted. The weak interaction was described by DFT + D3 method using empirical correction in Grimme’s scheme. Spin polarization method was adopted to describe the magnetic system. The triaxial strain ε of +3% was applied equally for all in-plane directions (A and B axes, compress strain), while a -9% strain was used in the perpendicular direction (C axis, tensile strain). The adsorption energy was calculated as: Eads = E(*adsorbent) - E(*) - E(adsorbent), where E(*adsorbent), E(*) and E(adsorbent) represent the total energy of * adsorbent, * and adsorbent molecule, respectively. The Gibbs free energy for each elementary step was calculated as: G = Eelec + EZPE – TS, in which Eelec is the electronic energy at 0 K calculated by DFT, EZPE is the zero-point energy term, and T is the absolute temperature (here 298.15 K), S is the entropy.