Synthesis and microstructural analysesThe Aun/Au1-CMS were synthesized following a four-step procedure as schematically illustrated in Fig. 1a (experimental details can be found in the Supporting Information (SI)). In step 1, a homogeneous precursor solution containing volatile organic solvent (e.g., tetrahydrofuran [THF]) and metal precursors (molybdenyl acetylacetonate, MoO2(acac)2) was prepared, then poured into a ceramic boat containing NaF crystals and mixed uniformly. The polar organic solvent can evenly diffuse and evaporate on the salt crystal surfaces. After complete evaporation of the solvent in an oven at 70 °C, the sulfur powder was mixed. In step 2, MoS2 crystals were generated on the surface of salt crystals by thermal annealing the mixed powder. During the calcination process, the MoO2 group in the precursor readily forms MoO2X2 (X = F, Cl, Br, and I) molecules with halogens21. As the sulfurization bottleneck barrier of molybdenum oxyhalides increases with the increasing electronegativity of the halogen X, the rate and amount of MoO2F2 being gradually substituted by S and transformed into the MoS6 precursor molecule of MoS2 are significantly reduced22. Consequently, it is easy to form cracked and wrinkled MoS2 nanosheets. Its exposed abundant layered edges serve as ideal substrates for subsequent light-driven oriented anchoring of precious metals. In step 3, fluorine-free CMS can be obtained after removal of salt crystals via multiple deionized water washing and soaking. In step 4, simply put, by illuminating a mixed solution of chloroauric acid and CMS with visible light, Au NPs and SAs preferentially anchored onto the highly active edge sites. On the contrary, using high-energy band illumination will quickly reduce and agglomerate Au ions into particles, with almost no SAs.Fig. 1: Synthesis and structural characterizations of Aun/Au1-CMS.a Schematic illustration of the fabrication process of Aun/Au1-CMS. b TEM image, (c) the corresponding AFM image, (d) HRTEM image and SAED pattern for Aun/Au1-CMS. e, f Corresponding color-coded channel map of filtered atomic-resolution HAADF-STEM image, (g) HAADF-STEM image and corresponding EDS elemental mapping of Aun/Au1-CMS.Distinct morphologies of the MoS2 nanosheets synthesized on NAF and NaCl substrates were observed in scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images (Supplementary Fig. 1a–c), with the former being cracked and wrinkled (termed CMS), and the latter being flat and intact (termed FMS). Subsequently, by using light of different wavelengths, photoreduction of chloroauric acid was carried out to achieve deposition of Au satellite domains on the edges of the CMS host (see details in SI). Shorter wavelength light irradiation could reduce Au ions and induce aggregation more quickly while extending the illumination time facilitates the production of more Au SAs23,24. The catalysts prepared at λ ≥ 400 nm and λ ≥ 100 nm were named Aun/Au1-CMS and Aun-CMS, respectively (Fig. 1a and Supplementary Fig. 1e, f). Concurrently, Au1-CMS was synthesized as a reference compound for comparative studies (Supplementary Fig. 1i, j). Obviously, Au particles on FMS were much larger than CMS (termed Aun-FMS, Supplementary Fig. 1d). The final loading amount of Au in Aun/Au1-CMS and Aun-CMS was measured by ICP-OES to be 2.14 and 2.86 wt.%, respectively (Supplementary Table 1). The as-synthesized Aun/Au1-CMS exhibited a mean thickness of 5 nm (Fig. 1c). In the Raman spectrum (Supplementary Fig. 2), CMS, Aun/Au1-CMS, Au1-CMS, Aun-CMS and FMS have two strong peaks near 380 cm−1 and 410 cm−1, which belong to the E2g and A1g modes of the 2H phase structure of MoS2, respectively25. As can be seen from Aun/Au1-CMS in Supplementary Fig. 3, the powder X-ray diffraction pattern could be well indexed to 2H-MoS2 (JCPDS No. 37-1492) and cubic Au (JCPDS No. 04-0784). The high-resolution TEM (HRTEM, Fig. 1d) images exhibited lattice fringes of 0.63 nm and 0.23 nm, corresponding to MoS2 (002) and Au (111) planes, respectively. The selected area electron diffraction (SAED) image showed two sets of diffraction patterns, with the (100) or (110) and (111) spots still indexed to MoS2 and Au. The images demonstrated that Au species tend to be deposited on the layer edges of MoS2. Atomic-resolution HAADF-STEM evidenced the coexistence of Au SAs and NPs in CMS, with isolated or adjacent Au SAs highlighted with red and yellow circles (Fig. 1e, f and Supplementary Fig. 1g). These are attributed to Au because the atomic number of Mo (z = 42) is greater than that of S (z = 16) but lower than that of Au (z = 79). Thus as in Supplementary Fig. 1h, no Au SAs were observed in Aun-CMS. HAADF-STEM with EDX elemental mapping measurements of Aun/Au1-CMS (Fig. 1g) revealed that Au SAs were uniformly loaded on the layer edges of CMS, while Au NPs were concurrently embedded at its interlayer edges. The above results clearly demonstrate that the photophysical field of appropriate wavelength bands can drive selective loading of Au SAs and NPs onto CMS.The local structure and coordination environment of Au species in Aun/Au1-CMS were further explored by EPR, XPS and X-ray absorption spectroscopy (XAS). As shown in Supplementary Fig. 9, CMS had a broader peak at ~324 G than FMS, attributed to the magnetic interaction of the unpaired spin-polarized electrons in the adjacent Mo sites, suggesting the presence of low-coordination Mo-Mo sites at the layer edges of CMS whereas FMS is dominated by isolated S vacancies17,26. This was further confirmed by the enhanced signal of the bridging disulfide (S22–) ligand at the edge in XPS (Supplementary Fig. 10a)27. Combined with TEM results, it was confirmed that the boundaries of CMS are mainly arranged in an unsaturated Mo dot-matrix pattern. Determination of the chemical state by X-ray absorption near-edge structure (XANES) combined with XPS revealed that Mo species in CMS exhibit different unsaturated coordination environments, compared to Mo in FMS (Fig. 2a and Supplementary Fig. 10b). As can be seen in Fig. 2b, the Au 4f spectrum of Aun/Au1-CMS can be simulated with four peaks compared to Aun-CMS, implying the existence of different Au species (Au0, Auδ+). The species types and local coordination environment of Au were studied in depth using XAS. The Au L3-edge X-ray XANES of a standard Au foil and KAuCl4 are also included for comparison. The white line peak at 11,923 eV in the Au L3-edge XANES spectrum arises from the transfer of electrons from the occupied 2p to the unoccupied 5d orbitals28. Therefore, the intensity of the white line peak could verify the presence of Auδ+. The XANES spectrum of Aun-CMS and Aun-FMS were closer to the reference spectrum of Au foil, where metallic Au was supposed to be the dominant Au species. The slightly higher intensity of the white line peaks in Aun-CMS than in Aun-FMS was attributed to Au size shrinkage29. Aun/Au1-CMS showed a mediate intensity white line peak compared to Aun-CMS and Au1-CMS, which in combination with the XPS results confirmed the presence of Au cationic form (Fig. 2c). The EXAFS spectra for Au-foil, Aun-CMS and Aun-FMS, Fig. 2d, exhibited two peaks at ≈2.5 and ≈2.9 Å, which were assigned to typical Au−Au scattering. Au1-CMS exhibited a peak at ≈1.9 Å, which can be classified as the signal of Au−S. The EXAFS spectra Aun/Au1-CMS, apart from the Au−Au signals, presented two peaks at ≈2 Å and ≈1.6 Å. The former belongs to the Au−S bonds, and the latter is associated with small-sized Au NPs and trace amounts of oxidized gold30. Wavelet transform (WT) analysis allowed for high-resolution identification in both k-space and R-space, thus enhancing coordination data. From the WT contour plots of Au foil and Aun-CMS (Fig. 2e), it can be analyzed that the maximum intensity peak at 9 Å−1 in k-space corresponds to Au−Au configuration. Remarkably, the WT spectrum of Aun/Au1-CMS detected a moderate intensity peak at 2 Å in R-space, which can be referred to as Au−S configuration. Furthermore, the accurate coordination configuration of Aun/Au1-CMS was extracted using the multi-shell EXAFS best-fitting analysis (Supplementary Fig. 12 and Supplementary Table 2). The fitting result demonstrated that the peaks between 2.3 and 4 Å originate from Au−Au coordination, while the minor peak at 2 Å originated from the contribution of Au−S bonds. The coordination number of S atoms around one Au atom was estimated to be 4.00, and the average bond length is about 2.32 Å. Density functional theory (DFT) calculations were performed to model and optimize the structure of Aun/Au1-CMS. In Aun/Au1-CMS, the AuS4 coordination structure was formed on the unsaturated Mo dot-matrix of CMS (inset in Fig. 2d), which is consistent with the EXAFS fitting results.Fig. 2: Spectral characterization.a XANES spectra of Mo K-edge over CMS, Aun/Au1-CMS, Aun-CMS, FMS and Mo foil. b High-resolution Au 4f XPS spectra of Aun/Au1-CMS, Aun-CMS and Au1-CMS. c XANES and (d) corresponding EXAFS spectra for Au L3-edge for Au-foil, Aun/Au1-CMS, Aun-CMS, Au1-CMS, Aun-FMS, and KAuCl4. e Wavelet-transformed k2-weighted EXAFS spectra of samples. Au: gold, Mo: blue, S: pink.Photocatalytic CO2 reduction performanceArmed with the structural information of Aun/Au1-CMS, Au1-CMS and Aun-CMS, we evaluated their photocatalytic CO2 reduction (PCR) performance without adding sacrificial agents. The yields of H2, CO, and C2H5OH of Aun/Au1-CMS under Vis-NIR light irradiation were 4.1, 2 and 1.8 μmol g−1 h−1, respectively, while the yield of CH3COOH reached as high as 26.9 μmol g−1 h−1 (Fig. 3a). Aun/Au1-CMS also showed favorable PCR performance under NIR light irradiation, with a CH3COOH evolution rate of 8.2 μmol g−1 h−1 (Fig. 3b). These results could be verified by 1H NMR and GC. In the spectrum, singlets and triplets were observed near 2.08 and 1.17 ppm, respectively, which are the characteristic peaks of methyl hydrogen in CH3COOH and methylene hydrogen in C2H5OH in D2O (Supplementary Fig. 14a, b). Impressively, the selectivity of the CO2 to CH3COOH product reached 86.4% under Vis-NIR light and 95.1% for the C2 products, with a slight decrease in selectivity under NIR light (Fig. 3c). The yield and selectivity comparable to most state-of-the-art photocatalysts (Supplementary Table 3). Interestingly, compared to Aun/Au1-CMS, Aun-CMS generated C2H5OH as the main product, and the reaction rates under Vis-NIR and NIR light were 7.6 and 2.1 μmol g−1 h−1, respectively. Compared to Aun-CMS (Fig. 3a, b), Aun-FMS exhibited a significantly lower yield of C2H5OH and CH3COOH, suggesting that Mo-edge-rich MoS2 was more conducive to the production of C2. Notably, while Au1-CMS exhibited a lower yield for CH3COOH, its selectivity (60.6%) was remarkably higher than that of Aun-CMS (13.6%). This observation underscores the synergistic interplay between Au NPs and SAs as the key to achieving both high yield and high selectivity for acetic acid synthesis. In contrast, CMS predominantly produced H2 at a rate of only 3.9 μmol g−1 h−1, whereas FMS is nearly photocatalytically incapable toward CO2 reduction. Simultaneously, the half-reaction products of PCR could be monitored. As shown in Fig. 3a and Supplementary Fig. 13, Aun/Au1-CMS and Aun-CMS generated O2 at a rate of 50.2 μmol g−1 h−1 and 24.1 μmol g−1 h−1 under Vis-NIR, respectively. For comparison, the amount of O2 generated was only 1.4 μmol g−1 h−1 at CMS, which deviates from the theoretical amount of oxidation products. This implied that the presence of Au favored the water oxidation half-reaction.Fig. 3: CO2 photoreduction performance.Evolution rates of H2, CO, O2, ethanol, and acetate products from Aun/Au1-CMS in (a) Vis-NIR or (b) NIR. c Product selectivity (%) of photocatalytic CO2 reduction reaction from Aun/Au1-CMS, Aun-CMS, and Au1-CMS under different light irradiation. d Control experiments under different reaction conditions for CO2 reduction over the Aun/Au1-CMS (e) 13C Isotope labeling mass spectra of acetate was obtained from photocatalytic 13CO2 reduction over Aun/Au1-CMS. f Cycling tests of CO2 photoreduction to acetate of Aun/Au1-CMS.We tested samples in controlled experiments under different conditions (Fig. 3d). The results indicated that the catalyst and the light source are the primitives of the photocatalytic reaction and H2 and O2 were generated in stoichiometric ratios without CO2. To further confirm the carbon source of the products, 13CO2 isotope tracing experiments were performed during the reaction (Fig. 3e). The mass spectrum of the acetate fraction showed a set of peaks between m/z 40 and 63, consistent with the molecular ion of 13CH313COOH+ (m/z = 62) and the fragment ion of 13CH313COO+ (m/z = 61) and 13CH313CO+ (m/z = 45). The acetate fraction exhibited an isotope-induced (M + 2) mass transfer effect, demonstrating that the C in CH3COOH in the product is derived from CO2. As shown in Supplementary Fig. 14c, d, the isotope-labeled 13CH313CH2OH product had the 13CH313CH2OH+ (m/z = 48) molecular ion and fragment ions (13CH313CH2O+: m/z = 47, 13CH2OH+: m/z = 32, and 13CH3+: m/z = 16), and the isotope-labeled 13CO also had associated molecular and fragment ions (13CO m/z = 29, 13C+ m/z = 13, and 16O+ m/z = 16). In addition to the high acetate selectivity of Aun/Au1-CMS, the material was about 80.8% of its original catalytic activity after five consecutive runs of 25 h (Fig. 3f).Band structure and charge carrier dynamicsBy measuring and analyzing the optical properties and band structure of synthetic samples, underlying factors for the enhanced activity can be discovered. The UV-Vis-NIR DRS of CMS showed strong absorption across the entire spectrum, stronger than FMS (Fig. 4a), with Aun/Au1-CMS showing the strongest light absorption. The energy gaps (Eg) of CMS, Aun-CMS, Au1-CMS, and Aun/Au1-CMS were determined to be 1.30, 1.23, 1.28 and 1.25 eV, respectively. The electronic energy band structure of the sample can be determined using ultraviolet photoelectron spectroscopy (UPS) (Fig. 4b). The ionization potential [equivalent to the valence band energy (Ev)] can be calculated according to the formula: Eip = hν- (Ecutoff-Es) where the hv is the incident photon energy with 21.22 eV, Es and Ecutoff are the high-energy electron start edge and cutoff energy of secondary electrons, respectively31. The maximum positions (Ev) of the valence band (VB) of the Aun-CMS and Aun/Au1-CMS samples were located at 0.24 and 0.44 V versus NHE. Based on their bandgaps, the conduction band (CB) positions of the Aun-CMS and Aun/Au1-CMS samples could be elucidated to be −0.99 V and −0.81 V. The CB positions obtained from the Mott–Schottky measurements were close to the above results (Supplementary Fig. 16). As illustrated in Supplementary Fig. 17a, both Aun-CMS and Aun/Au1-CMS exhibited suitable band gap and band edge positions with sufficient driving force to reduce CO2 to various fuels.Fig. 4: Band structure analysis and charge-carrier dynamics.a UV-Vis-NIR DRS of Aun/Au1-CMS, Aun-CMS, Au1-CMS, CMS, and FMS. b UPS spectrum of Aun/Au1-CMS, Aun-CMS, Au1-CMS and CMS. c, d In situ XPS and (e) In situ EPR spectra of Aun/Au1-CMS. f R–S plots based on CV data of Aun/Au1-CMS, Aun-CMS, Au1-CMS and CMS. 2D pseudo-color plot of Vis-region transient absorption spectra under (g) Ar or (h) CO2 atmosphere pumped at 400 nm for Aun/Au1-CMS. i Schematic illustration of the charge transfer processes of the Aun/Au1-CMS in photocatalytic CO2 reduction reaction.To comprehensively reveal the charge-transfer mechanism between Au SAs, Au NPs, and CMS, in situ XPS and in situ EPR was studied. As shown in Fig. 4c, compared with the light-off condition, the Au0 4f peak of Aun/Au1-CMS shifted toward higher binding energy by 0.1 eV in the light-on condition, while the peak area ratio of Auδ+ 4f decreased (Supplementary Table 4). Simultaneously, the peaks of S 2p and Mo 3d both move toward higher binding energy (Fig. 4d and Supplementary Fig. 17b). Thus, the Au SAs work as electron acceptors during the PCR process. Briefly, electrons were excited from the valence band (S 2p orbital) to the conduction band (Mo 3d orbital) of MoS2 and then transferred to Au SAs, and hot electrons generated by Au NPs under photoexcitation could be rapidly transferred to Au SAs domains to enhance the reduction reaction30,32. Parallel to this, the signal of edge S at g = 2.004 rapidly weakened under short-time light irradiation is attributed to the transfer of electrons from unsaturated S, further indicating the existence of fast and intensive charge transfer at the layer edge of Aun/Au1-CMS (Fig. 4e)33.The separation efficiency of photogenerated carrier was judged by photoluminescence (PL) spectra (Supplementary Fig. 18a). CMS exhibited significantly weaker PL intensity than FMS, indicating enhanced extraction and reduced recombination of charge carriers due to the boundary-rich structure. The carrier separation and transfer behavior were further investigated by a series of photoelectric tests. Aun/Au1-CMS had the least charge transfer resistance and the highest photocurrent density of 1.8 μA cm−2, evidencing that it featured the greatest charge separation efficiency. As shown in Supplementary Fig. 19, compared with CMS and Aun-CMS, the CV curve of Aun/Au1-CMS under CO2 saturated atmosphere showed a significant increase in the current maximum, indicating that the improvement in electron transfer capacity was due to the synergistic effect of Au SAs and Au NPs34. Kinetic analysis of the electron transfer behavior of Aun/Au1-CMS, Au1-CMS, Aun-CMS and CMS was carried out using CV data based on the Randles-Sevcik (R-S) equation35. As shown in Fig. 4f, Aun/Au1-CMS possessed the highest diffusion coefficient which clarified that the introduction of Au SAs and Au NPs reinforced the electron transfer capacity.The growth of Au NPs and Au SAs domains on the layer edge of CMS altered the charge carrier dynamics, and we employed femtosecond transient absorption spectroscopy (TA) to explore charge transfer dynamics of Aun/Au1-CMS. As in Fig. 4g, Supplementary Figs. 22a and Fig. 24a, the negative signals (~440 nm) generated by the ground state exciton bleaching were observed in Aun/Au1-CMS, Au1-CMS and Aun-CMS36. In addition, a broad positive absorption signal could be clearly seen at around 450−650 nm, which is assigned to the holes of photoinduced absorption37. However, Aun/Au1-CMS exhibited a stronger absorption than Aun-CMS and Au1-CMS, and thus the former is more efficient in charge generation and separation. A similar positive absorption peak was observed in both Aun/Au1-CMS and Aun-CMS in the NIR region at around 850−1150 nm (Supplementary Figs. 21a and 23a). Notably, with the introduction of CO2, the positive signal of Aun/Au1-CMS is weaker than that of Aun-CMS, suggesting that CO2 is more prone to dissipate the photogenerated carriers produced by Aun/Au1-CMS (Fig. 4g, h and Supplementary Fig. 22a, b). To thoroughly understand the difference in charge transfer kinetics between Aun/Au1-CMS, Au1-CMS, and Aun-CMS, their kinetic differences were compared to decipher the electron transfer process of PCR (Supplementary Figs. 20c, 21e, 22e, 23e and 24e). The rise of the signal at 550 nm showed that one picosecond charge separation occurred in both Aun/Au1-CMS (Supplementary Fig. 20d), Aun-CMS (Supplementary Fig. 22f) and Au1-CMS (Supplementary Fig. 24f). The ΔA signal decayed slower at 550 nm for both samples, with the average lifetime of Aun/Au1-CMS (950 ps) being 1.6 times that of Au1-CMS (597 ps) and 2.5 times that of Aun-CMS (374 ps), and CMS having the shortest average lifetime (Supplementary Table 5). These findings suggest that Au SAs are more effective than Au NPs in prolonging the relaxation time of photogenerated electrons. The longer charge-separated state can also be partially attributed to steric hindrance of SAs and NPs, increasing the spatial distance for charge recombination38. Notably, the slower lifetimes were more favorable for semiconductor-adsorbate systems (typically in a time scale of a few ps to ns)37,39,40. In a CO2 atmosphere, the average lifetime of Aun/Au1-CMS at 550 nm was attenuated by 76.6%, which was higher than 65.8% of Au1-CMS and 56.1% of Aun-CMS. The average lifetime of Aun-CMS at 850 nm was extended by 240.4 ps, while the average lifetime of Aun/Au1-CMS remained shortened (Supplementary Table 6). These results suggest that Au SAs accelerate photogenerated electron injection into CO2, thereby enhancing catalytic activity.In the simplified model describing the PCR process using semiconductor-metal heterostructures (Fig. 4i), firstly, hot electrons generated by interband transitions in the 5d→6sp orbitals of Au NPs were transferred to the CMS (rate constant kET1)41, while at the same time photogenerated electrons from the CB of CMS were transferred around the Au SAs domain (rate constant kET2), which was corroborated by theoretical computation simulated electron redistribution and electrostatic potential (Supplementary Fig. 26). The Schottky barrier arising from the work function mismatch between Au NPs and CMS allowed the injection of hot electrons, generated by localized surface plasmon resonance excitation, across the interface and into the conduction band (CB) of the semiconductor, preventing the backflow of photogenerated electrons (Supplementary Fig. 27). The electron-hole recombination (rate constant ke-h1) of Au NPs was correlated with hole trapping (rate constant kHT1) and kET1. The electrons transferred to the Au SAs domain will subsequently reduce CO2 (rate constant kCO2R), which competed with the recombination of electrons and holes in the semiconductor region (rate constant ke-h2). For Aun/Au1-CMS, the large confinement potential of CMS could promote the transfer of holes and inhibit their photogenerated electrons from flowing back to Au NPs, thus strengthening the reduction potential of the Au SAs domain. The TA results indicated that Au SAs can promote CO2 hydrogenation to yield acetic acid by significantly increasing kCO2R.Mechanism of photoactivity enhancementBoth the optoelectronic properties and ultrafast TA spectroscopy measurements indicated that photogenerated electrons and holes can be effectively delocalized and reach Au SAs domain. Consequently, it was necessary to explore what was the core question of obtaining high yield and selectivity of Aun/Au1-CMS in PCR with the support of in situ experimental data and theoretical results. With the aid of In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), the behavior of reaction intermediates can be traced. Figure 5a, b shows a set of peaks that appear to enhance with irradiation time under visible light irradiation. The peak at 1245 cm–1 and 1510 cm–1 can be assigned to the stretching of absorbed b-CO32– (bidentate carbonates) and m-CO32– (monodentate carbonates), respectively42. The asymmetric stretching vibration peaks of *HCO3 was observed at 1661 cm−16. The infrared signals at 1213 cm−1 and 1548 cm−1 can be attributed to the COOH* group, which is a pivotal intermediate in the conversion of CO2 to other solar fuels4,6. The increase in the peak intensity of the absorption band at 2024 cm−1 confirmed the formation of *CO intermediates in the reaction43. A peak attributed to * OC−CO, a key intermediate for acetate synthesis, can be observed at 1347 cm−144. Additionally, at 1090 cm−1, the vibration of the *COH group gradually appeared, which was a potential precursor for the generation of C2H5OH45. The assignment of the peak at 1444 cm−1 to the COO stretching vibration of CH3COOH was direct evidence of acetate production7. Notably, the DRIFTS peak of Aun/Au1-CMS located in the 3500–3800 cm−1 region was found to be gradually shrinking, whereas the opposite was true for CMS (Supplementary Fig. 28), demonstrating that the presence of Au species was effective in accomplishing H2O oxidation. Additionally, in situ Raman spectroscopy was performed to analyze the C−C coupling of Aun/Au1-CMS with varying irradiation time providing further evidence (Fig. 5c and Supplementary Fig. 29). The peaks before 500 cm−1 were associated with the Aun/Au1-CMS catalyst and did not change significantly with radiation time. The band at around 899 cm−1 is designated to C−C−O stretching for CH3COOH formation46. This band was an indicator of the production of C2 products by C−C coupling. The bands at 1985 and 2075 cm–1 correspond to *CO species, which can once again verify the possibility of *CO coupling.Fig. 5: DFT calculations and mechanism of photoactivity enhancement.In situ DRIFTS spectra at (a) 1000–2200 cm−1 region for reaction intermediates and (b) 3450–3850 cm−1 region for -OH groups on Aun/Au1-CMS under light irradiation. c In situ Raman for reaction intermediates on Aun/Au1-CMS at different irradiation times. Free energy diagram for CO2 hydrogenation over (d) Aun/Au1-CMS-I and (e) Aun-CMS model. f Schematic illustration of adsorbed CO (5σ, 2Ï€*) orbital interactions with Au 5d and Mo 4d orbitals of Aun/Au1-CMS-I.The possible origins of the Au SAs domains for enhanced PCR activity and selective CH3COOH or C2H5OH production were explored by DFT calculations. We constructed three different configurations of active-site structures, including absence Au SAs (named Aun-CMS), isolated Au SAs (named Aun/Au1-CMS-I), and adjacent Au SAs (named Aun/Au1-CMS-A), on the base model of Au NPs anchored at the layer edges of CMS (Supplementary Fig. 30). The adsorption and activation of CO2 molecules were critical starting points for the subsequent proton-electron transfer process. When CO2 was adsorbed on the Aun/Au1-CMS-I surface, the interaction of Au−S4 and Mo−S4 units decreased the O1−C−O2 bond angle from 180° to 125.8° (Supplementary Fig. 31). In contrast, the bond angles formed by CO2 on the Aun-CMS and Aun/Au1-CMS-A surfaces via a pair of Mo−O and Au−O interaction were decreased to 134.7° and 179.6°, respectively. The spatial structure of *CO2 will also change from the V-shape of Aun-CMS to the quasi-L-shape of Aun/Au1-CMS-I. Besides, the C−O2 bond on Aun/Au1-CMS-I was all longer than the other structures. Unexpectedly, the Bader charge of adsorbed *CO2 changed from 0.545e over Aun/Au1-CMS-I surface to 0.738e on Aun/Au1-CMS-A surface, and the adsorption energy of the latter (–1.56 eV) was lower than that of the former (–0.18 eV), consistent with CO2-TPD data (Fig. 5d and Supplementary Fig. 32). An integrated crystal orbital Hamilton population (ICOHP) was completed, providing insight into what affects CO2 adsorption and activation. As shown in Supplementary Fig. 33, the ICOHP value of the Mo–O bond of Aun/Au1-CMS-I was 4.33, which is much larger than the Mo−O bond (1.97) of Aun/Au1-CMS-A. Higher ICOOP values for metal−O bonds indicated stronger interactions between the metal and O. The ICOHP value of C−O bond (12.07) of Aun/Au1-CMS-I was lower than that of Aun-CMS (15.10), which meant that Aun/Au1-CMS-I had higher thermodynamic feasibility of C=O bond dissociation. The electronic location function results of Aun/Au1-CMS-I indicated that asymmetric electron transport channels could be formed via Au-C and Mo−O bonds, whereas the adjacent Au SAs in Aun/Au1-CMS-A was almost incapable of forming effective electron transport channels with *CO2. These results revealed that introduction of isolated Au SAs, although weakening the adsorption of CO2, strengthens the O-affinity of Mo and the C-affinity of Au, thus weakening the C=O bond in favor of protonation and lowering the barrier to *COOH formation (Fig. 5d, e and Supplementary Fig. 33). The formation of C2 products depended on the occurrence of C−C coupling, and the reaction energy barriers for *CO + *CO → *OC−CO at heteroatomic sites in Aun/Au1-CMS-I was relatively low. Given that the coupling of (*CO + *CO) intermediates to form *OC−CO does not involve electron or proton transfer, it is essentially a thermal process. The underlying mechanism can be judged by studying the orbital coupling behavior between the metal active center and *CO intermediate. Supplementary Figs. 38–40 show the calculated projected DOS (PDOS) of the adsorbed CO (5σ and 2Ï€*) and metal-d orbitals (dxz, dxy, dyz, dz2, and dx2−y2) in these materials. In Aun/Au1-CMS-I, the Au-5dz2 and Mo-4dz2 orbitals resonate at −1.69 and 0.98 eV, and Au-5dx2−y2 and Mo-4dx2−y2 resonate at –6.21 and –3.54 eV, pointing to the existence of 5d-4d orbital coupling at the heteroatoms. In Aun/Au1-CMS-A, the Au-5d PDOS meaningfully overlap with the *CO molecular orbitals through (Au, dz2) – (CO 5σ) and (Au, dxz/dyz) – (CO 2Ï€*) interactions, whereas for Aun-CMS, the Mo-4d site exhibits significant orbital coupling only by (Mo, dxz/dx2−y2) – (CO 2Ï€*). For Aun/Au1-CMS-I, an analogous orbital coupling of the above two characteristics can be exhibited, and the (Au, dxz/dyz) – (CO 2Ï€*) interaction can be strengthened. As a result, the synergistic interaction of Au and Mo sites reduced the energy splitting levels of bonding and antibonding states (Fig. 5f and Supplementary Fig. 41), leading to weaker dipole-dipole interactions, increased the collision probability of two *CO intermediates, and ultimately promoted the coupling of neighboring *CO intermediates to form *OC−CO.Given the difference in selectivity of Aun/Au1-CMS and Aun-CMS for CH3COOH and C2H5OH, the key to this was found to be in the hydrogenation position of *CHCO by Gibbs free energy. Au−Mo sites tended to hydrogenate at the CH-terminal, whereas Mo−Mo sites had a lower energy barrier to hydrogenation at the CO-terminal. This might be due to the fact that the Au SAs alters the affinity of the adjacent Mo site for the O-containing intermediate, thus preferring hydrogenation at the CH-terminal. Additionally, the immediate hydrogenation of *OC to *COH at the Mo−Mo sites was also exothermic (*COH intermediates were observed in In situ DRIFTS spectra), which was another potential pathway for ethanol production. Interestingly, the adsorbed H* intermediates at the Au−Mo sites of Aun/Au1-CMS-I had a higher energy barrier than the Mo−Mo and Au−Au sites, suggesting that the hydrogen evolution side reaction was suppressed (Supplementary Fig. 42).By integrating the experimental data and theoretical calculations, the overall catalytic mechanism can be summarized as Fig. 6. First, under photoexcitation, hot electrons and holes generated by Au NPs (Aun) can be rapidly separated within ~1 ps and transport the photogenerated electrons to the layered edges of CMS. The Au SAs (Au1) domains grown on the layer edges can efficiently capture the electrons and prolong the photogenerated charge lifetime to facilitate subsequent CO2 reduction, while the holes of Aun are consumed by the water oxidation half-reaction. Simultaneously, the direct recombination of electrons and holes on CMS and Aun will dissipate part of the energy. Second, the isolated Au1 enhances the O-affinity of neighboring Mo, resulting in adsorbed CO2 exhibiting significant dihedral angle distortion and observing asymmetric electron transport channels. Furthermore, adjacent Au1 would excessively weaken CO2 adsorption and thus is not an effective active site. Third, the orbital coupling of Au 5d and Mo 4d reinforces the overlap with the CO molecular orbitals (5σ and 2Ï€*), leading to reduced energy splitting of the bonding and antibonding states, which favors generation of *OCCO intermediates. Most importantly, the synergistic effect of Au−Mo selectively converts *CHCO into *CH2CO, a critical step for acetate generation.Fig. 6: Photocatalytic CO2 reduction pathways at different active sites.Schematic illustration of the photocatalytic mechanism of different active sites on Aun/Au1-CMS.In conclusion, we successfully anchored Au NPs and SAs on the unsaturated Mo atoms of MoS2 through the salt template and photodeposition method. The synergistic effect of Au NPs and SAs exhibits excellent acetate generation performance in CO2 photoreduction and still have catalytic activity in the NIR region. The photogenerated charge dynamics show that the Au SAs significantly delays the recombination of electron-hole pairs and cooperates with adjacent Mo to create the channel for effectively injecting photogenerated electrons into CO2 molecules. Hot electrons are generated and transferred and holes are consumed on Au NPs. Theoretical calculations revealed that the Au−Mo dual sites favor *CO and *CO coupling as well as tend to *CHCO conversion to *CH2CO, Mo−Mo is unfavorable for C−C coupling, and the adjacent Au is difficult to adsorb CO2. Therefore, the orbital coupling of Au5d and Mo4d is thermodynamically more conducive to the generation of acetic acid. Our work provides valuable photocatalyst design concepts for converting CO2 into high-value C2 chemicals with specific targets.
