Room-temperature photosynthesis of propane from CO₂ with Cu single atoms on vacancy-rich TiO₂

Abstract

Photochemical conversion of CO₂ into high-value C₂₊ products is difficult to achieve due to the energetic and mechanistic challenges in forming multiple C-C bonds. Herein, an efficient photocatalyst for the conversion of CO₂ into C₃H₈ is prepared by implanting Cu single atoms on Ti_0.91O₂ atomically-thin single layers. Cu single atoms promote the formation of neighbouring oxygen vacancies (V_Os) in Ti_0.91O₂ matrix. These oxygen vacancies modulate the electronic coupling interaction between Cu atoms and adjacent Ti atoms to form a unique Cu-Ti-V_O unit in Ti_0.91O₂ matrix. A high electron-based selectivity of 64.8% for C₃H₈ (product-based selectivity of 32.4%), and 86.2% for total C₂₊ hydrocarbons (product-based selectivity of 50.2%) are achieved. Theoretical calculations suggest that Cu-Ti-V_O unit may stabilize the key *CHOCO and *CH₂OCOCO intermediates and reduce their energy levels, tuning both C₁-C₁ and C₁-C₂ couplings into thermodynamically-favourable exothermal processes. Tandem catalysis mechanism and potential reaction pathway are tentatively proposed for C₃H₈ formation, involving an overall (20e⁻ – 20H⁺) reduction and coupling of three CO₂ molecules at room temperature.

A photocatalyst for CO2 reduction to C3H8 is prepared by implanting Cu single atoms on vacancy rich TiO2 single layers. Key reaction intermediates, *CHOCO and *CH2OCOCO, are stabilized on the catalyst which promotes C-C bond formation.

Introduction

Using sunlight to generate fuels from CO₂ and water has the potential to reduce CO₂ emissions and facilitate the large-scale storage of renewable energy^1–5. Currently, light-driven reduction of CO₂ is mainly limited to two-electron-reduced CO and further reduced C₁ hydrocarbons such as methane (CH₄)^6,7 and C₂ products such as ethene (C₂H₄)⁸ and ethane (C₂H₆)⁹ in few cases. However, the formation of C₃ products by artificial photosynthesis is rare^10,11 and generally relies on a higher-order reaction pathway that requires the sequential formation of multiple C-C bonds¹², which involves the integration of two consecutive steps of CO₂-to-CO and CO-to-C₂₊ at different catalytic centres^13–16. These C-C couplings are usually challenging endothermic processes with huge uphill energy barriers owing to the high energy levels of the key *C₂ and *C₃ intermediates^17–20. These energy barriers result from the lack of effective catalytic centres that can stabilize these multicarbon intermediates^21,22.

Single-atom (SA) catalysts with maximum atom utilization efficiency and unique catalytic performance have emerged as an attractive frontier in heterogeneous catalysis^23,24. Atomically thin two-dimensional (2D) materials are suitable platforms to anchor metal SAs^25,26. These 2D single-layer (SL) materials can not only improve the activity of catalytic reactions²⁷, but also provide ideal models to gain atomic-level insights into real active sites and reaction mechanisms of catalytic processes through experimental and theoretical techniques^25,28.

Herein, we report that implanting Cu SAs in Ti_0.91O₂ atomic SLs allows the construction of a unique Cu-Ti-V_O/Ti_0.91O₂-SL photocatalyst for highly efficient and selective conversion of CO₂ into C₃H₈. The Cu SA promotes the neighboring Ti_0.91O₂ to generate oxygen vacancies (V_Os) and form a Cu-Ti-V_O unit in the Ti_0.91O₂ matrix. The Cu-Ti-V_O/Ti_0.91O₂-SL photocatalysis system efficiently converts CO₂ into C₂₊ products at room temperature with high electron-based selectivity of 64.8% for C₃H₈ (product-based selectivity of 32.4%) and 86.2% for overall C₂₊ hydrocarbons (product-based selectivity of 50.2%). As suggested by theoretical calculations, the Cu-Ti-V_O units may stabilize the key *CHOCO and *CH₂OCOCO intermediates and reduce their energy levels, tuning both C₁-C₁ and C₁-C₂ couplings into thermodynamically favorable exothermal processes. Based on the simulation results, tandem catalytic mechanism and reaction pathway for the production of C₂₊ hydrocarbons are proposed, involving an overall 20 e^—20 H⁺ reduction and two sequential C-C coupling processes of three CO₂ molecules (3 CO₂ + 20 e⁻ + 20 H⁺→ C₃H₈ + 6 H₂O). Our work provides an alternative paradigm for the photoconversion of CO₂ into multicarbon solar fuels, and represents a progressive step toward imitating natural photosynthesis.

Results

Formation and characterization of Cu-Ti-V_O units

Atomically thin layers of Ti_0.91O₂-SL were synthesized by exfoliating the layered protonic lepidocrocite-type titanate of H_0.7Ti_1.825O₄^29,30. The absence of peaks in the corresponding X-ray diffraction (XRD) pattern suggests the complete exfoliation into single layers³¹ (Supplementary Fig. 1). Cu SAs were then implanted in Ti_0.91O₂-SL to form Cu-Ti-V_O/Ti_0.91O₂-SL through a modified wet-chemical route, followed by a rapid thermal treatment (RTT) in an Ar atmosphere³² (see Methods for details). The amount of Cu loading was detected to be 0.29 wt% by inductively coupled plasma–optical emission spectrometry (ICP-OES). The XRD pattern of Cu-Ti-V_O/Ti_0.91O₂-SL was similar to that of pristine Ti_0.91O₂-SL (Supplementary Fig. 1). The thickness of Cu-Ti-V_O/Ti_0.91O₂-SL was found to be 0.85 nm using atomic force microscopy (AFM) (Supplementary Fig. 2), which corresponded closely to the theoretical monolayer thickness (0.75 nm)^30,33,34. Field emission-scanning electron microscopy (FE-SEM) and transmission electron microscopy (TEM) images show the ultrathin sheet-like morphology of Cu-Ti-V_O/Ti_0.91O₂-SL without discernible nanoparticles (NPs) (Fig. 1a–c), potentially implying the atomic-scale size of the Cu embedded in the Ti_0.91O₂ matrix. The isolated bright dots in atomic resolution aberration-corrected high angle annular dark-field-scanning transmission electron microscopy (AC HADDF-STEM) images (Fig. 1e) directly confirm the atomic dispersion of the Cu in the matrix. Energy dispersive X-ray spectroscopy (EDS) mapping (Fig. 1f–i) indicates that Cu is evenly distributed throughout the atomically thin Ti_0.91O₂ matrix.

Fig. 1. Morphological and structural characterization of Cu-Ti-V_O/Ti_0.91O₂-SL.

a FE-SEM and b, c TEM images, d, e AC HAADF-STEM images, and f–i EDS mapping.

Fourier transforms of the Cu K-edge extended X-ray absorption fine structure (EXAFS) spectra of Cu-Ti-V_O/Ti_0.91O₂-SL (Fig. 2a) contain a prominent peak at ~1.56 Å, assigned to Cu-O coordination in the first shell with a coordination number of 4 according to the EXAFS fitting analysis (Supplementary Fig. 3 and Supplementary Table 1). The characteristic metallic Cu-Cu bonding at ~2.24 Å is not observed, further validating the single-atom distribution of Cu. A minor scattering peak at ~2.42 Å is attributed to Cu-Ti coordination in the second shell based on the EXAFS fitting results^34,35. The presence of Cu-Ti coordination, originating from the strong electronic interaction between Cu and adjacent Ti atoms, suggests that Cu is present in the atomically dispersed Cu-Ti dual-metal coordination form³⁶.

Fig. 2. Electronic structure of Cu-Ti-V_O/Ti_0.91O₂-SL.

a Fourier transforms of EXAFS spectra at the Cu K-edge. b EPR spectra. Normalized XANES spectra at the c Cu K-edge and d Ti K-edge. e, f The atomic structure configuration (colour codes: light blue (Ti), blue (Cu), and red (O)). g PDOS and d-band centres of Cu 3d and Ti 3d orbitals. h, i COHP between Cu and adjacent Ti. j, k Charge density differences (yellow represents electron accumulation, and purple denotes electron depletion).

The normalized Cu K-edge X-ray absorption near-edge structure (XANES) spectra show that the near-edge absorption energy of Cu-Ti-V_O/Ti_0.91O₂-SL is higher than that of Cu foil and lies between the energies of Cu₂O and CuO (Fig. 2c), indicating that the average oxidation state of Cu is between +1 and +2, which is verified by Cu LMM Auger electron spectra (AES) and Cu 2p X-ray photoelectron spectra (XPS) as well as the theoretical results (Supplementary Fig. 4a, c). The near-edge absorption energy and white-line intensity of Ti K-edge XANES for Cu-Ti-V_O/Ti_0.91O₂-SL are higher than those for Ti_0.91O₂-SL (Fig. 2d), suggesting the presence of Ti species with lower electronic density, which is confirmed by the Ti^δ+ (δ > 4) species in Ti 2p XPS spectra^37,38 (Supplementary Fig. 4b). The formation of these electron-poor Ti centres and the partially oxidized Cu centres proves the electron donation to Cu SAs from the coordinated Ti atoms as a result of the strong electronic interaction in the Cu-Ti dual-metal coordination.

The O 1 s XPS spectrum of Cu-Ti-V_O/Ti_0.91O₂-SL displays a peak attributed to V_O at 531.6 eV^39,40 (Supplementary Fig. 5). The electron paramagnetic resonance (EPR) spectra also show a V_O signal at a g value of 2.003⁴⁰ (Fig. 2b), suggesting the formation of V_Os in Cu-Ti-V_O/Ti_0.91O₂-SL. In contrast, no V_O signal is observed for Ti_0.91O₂-SL (Fig. 2b and Supplementary Fig. 5). Density functional theory (DFT) computations reveal that the formation energy of V_O is 5.66 eV in Ti_0.91O₂-SL, while the value sharply decreases to 2.39 eV with Cu anchoring, implying that the introduction of Cu SAs can facilitate the formation of neighboring V_Os in the Ti_0.92O₂ matrix.

The strong coordination interaction of the anchored Cu SA with neighboring Ti atoms originates from the presence of V_Os. A control sample without V_Os (denoted as Cu-O/Ti_0.91O₂-SL) was synthesized through a similar RTT in an air atmosphere (see Methods, Fig. 2b, f and Supplementary Fig. 6 for details). The projected density of states (PDOS) of Cu-Ti-V_O/Ti_0.91O₂-SL (Fig. 2g) demonstrates that the d-band centre (ɛ_d) of Cu (−2.52 eV) is in good agreement with adjacent Ti (−2.58 eV), inducing a strong electronic coupling effect in the dual-metal sites³⁶. The crystal orbital Hamilton population (COHP) between Cu SAs and the closest Ti atoms was then calculated to quantitatively study the intensity of Cu-Ti interactions with and without V_Os. The less antibonding orbital populations and a more negative value of integrated-crystal orbital Hamilton population (ICOHP) for Cu-Ti-V_O/Ti_0.91O₂-SL than Cu-O/Ti_0.91O₂-SL prove that the Cu-Ti electronic interaction is much stronger when V_Os are present (Fig. 2h, i). Charge density differences and Bader charge analysis (Fig. 2j) suggest that 0.3 e⁻ directly transfers to Cu SAs from the neighboring Ti atoms in Cu-Ti-V_O/Ti_0.91O₂-SL, which causes notable electron accumulation at Cu sites and substantial electron depletion at Ti sites, revealing the asymmetric electron distribution at Cu-Ti coordination. In comparison, without V_Os, Cu-O/Ti_0.91O₂-SL displays no discernible electron perturbation at Ti sites, and electrons are localized within Cu-O coordination (Fig. 2k), demonstrating that Cu SAs share negligible interactions with Ti atoms and are in a relatively isolated single-metal form, confirmed by XPS and EPR (Fig. 2b and Supplementary Fig. 6). Therefore, both experimental and theoretical analyses indicate that Cu-Ti-V_O units are formed in the Ti_0.91O₂ matrix for Cu-Ti-V_O/Ti_0.91O₂-SL. In contrast, isolated Cu-O sites are formed in the Ti_0.91O₂ matrix without V_Os for Cu-O/Ti_0.91O₂-SL.

Photocatalytic CO₂ reduction performance

All of the photocatalytic CO₂ reduction metrics reported in this study were measured in CO₂-saturated acetonitrile aqueous solution (acetonitrile: water = 5:1 by volume) unless otherwise specified, and the testing details are described in the Methods section (Supplementary Fig. 7). Unexfoliated layered titanate (denoted as Ti_0.91O₂-B) mainly produces CO, with a formation rate of 7.0 μmol g⁻¹ h⁻¹ (Fig. 3b). Ti_0.91O₂-SL displays an enhanced CO production rate of 67.0 μmol g⁻¹ h⁻¹, indicating that the atomically thin 2D geometry is favorable for the improvement of CO₂ activity through the potential exposure of many rich active sites and a shortened charge-transfer distance from the interior to the surface^26,41. Both CO and CH₄ were detected on Cu-O/Ti_0.91O₂-SL with yields of 61.0 and 11.3 μmol g⁻¹ h⁻¹, respectively. Cu-Ti-V_O/Ti_0.91O₂-SL exhibit significant yields of C₂H₄ (7.6 μmol g⁻¹h⁻¹) and C₃H₈ (13.8 μmol g⁻¹h⁻¹), together with a small amount of CH₄ and trace C₂H₆/C₃H₆, in addition to CO (18.6 μmolg⁻¹h⁻¹) (Fig. 3a, b), showing a strong capability of C-C coupling. No H₂ production was detected (Supplementary Fig. 8), and O₂ was generated roughly stoichiometrically (Supplementary Fig. 9). The quantum efficiency (QE) was obtained 0.48%, 0.15%, and 0.06% at the wavelength of 385, 415, and 520 nm, respectively. A high electron-based selectivity of C₃H₈ of 64.8% was achieved (32.4% of the product-based selectivity), and that of total C₂₊ products was as high as 86.2% (50.2% of the product-based selectivity) (Fig. 3c and Table 1). The photocatalytic performance of Cu-Ti-V_O/Ti_0.91O₂-SL was also tested in pure water. The production rates of all carbon-based products decreased in pure water as compared with those (Supplementary Fig. 10). Moreover, the overall selectivity of C₂₊ products becomes lower in pure water, and the selectivity of C₃H₈ is less than C₂H₄ (Supplementary Fig. 11). H₂ is detected with a formation rate of ~2.2 μmol g⁻¹ h⁻¹ in pure water. The improved activity and selectivity are most likely due to the high CO₂ solubility in acetonitrile, which largely increases the local accessibility of CO₂, thus enhancing the contact between the catalyst and CO₂ molecules to facilitate C-C coupling. To the best of our knowledge, the efficiency and selectivity of C₃H₈ production in our work outperform most of the reported photocatalysts in acetonitrile medium, and also rank among the top of the reported C₂₊ production photocatalysts in non-acetonitrile medium (Supplementary Table 3).

Fig. 3. Photocatalytic CO₂ reduction performance in acetonitrile aqueous solution.

a Photocatalytic product evolution as a function of light irradiation times on Cu-Ti-V_O/Ti_0.91O₂-SL. b Product formation rates and c electron-based selectivity on Ti_0.91O₂-SL, Cu-O/Ti_0.91O₂-SL, and Cu-Ti-V_O/Ti_0.91O₂-SL. (Error bars indicate standard deviations.).

Table 1.

Selectivity of different products on Ti_0.91O₂-SL, Cu-O/Ti_0.91O₂-SL, and Cu-Ti-V_O/Ti_0.91O₂-SL.^a

Catalyst	Electron-based selectivity (%)				Product-based selectivity (%)
	CO	CH4	C2H4	C3H8	CO	CH4	C2H4	C3H8
Ti0.91O2-SL	94.7	5.3	Not detectable		98.6	1.4	Not detectable
Cu-O/Ti0.91O2-SL	57.4	42.6	Not detectable		84.4	15.6	Not detectable
Cu-Ti-VO/Ti0.91O2-SL	8.7	5.1	21.4	64.8	43.4	6.3	17.8	32.4

^a The details regarding the calculation of electron- and product-based selectivity of CO, CH₄, C₂H₄ and C₃H₈ are presented in the Methods section.

The substantially suppressed CO production and increased C₂₊ production on Cu-Ti-V_O/Ti_0.91O₂-SL, compared with pristine Ti_0.91O₂-SL, imply that the formation of C₂₊ products is potentially derived from the coupling depletion of the *CO intermediate. CO was used to substitute CO₂ as the starting reactant on Cu-Ti-V_O/Ti_0.91O₂-SL, and a considerable amount of C₃H₈ and C₂H₄ was indeed detected (Supplementary Fig. 12), further confirming *CO as an important intermediate for the present C₂₊ products. The distinctive CO₂ photoreduction activity and high selectivity of C₃H₈ and total C₂₊ products are still well maintained after three-cycle tests of 15 h in total (Supplementary Figs. 13 and 14), and the post-reaction characterizations of Cu-Ti-V_O/Ti_0.91O₂-SL display no obvious structural and morphological changes (Supplementary Figs. 15–17), demonstrating the excellent stability of the catalyst. Notably, with the decrease in the Cu loading amount, the C₂₊ yield declines as the lower Cu loading reasonably reduces the number of Cu-Ti-V_O units, restraining C-C coupling (Supplementary Figs. 18 and 19). Moreover, higher Cu loading results in Cu aggregation into nanoclusters (NCs) and nanoparticles (NPs), subsequently lowering the selectivity of C₂₊ products, which is probably due to the weaker coupling interaction between Cu and the Ti_0.91O₂ matrix as the metal particle size increases^42,43 (Supplementary Fig. 18). A series of control experiments were performed in the absence of illumination, the catalyst or CO₂, and no detectable CO or other hydrocarbon products were detected. The ¹³CO₂ isotope labeling experiment and the time profile of relative abundance of ¹³C labeled products confirm that the carbon source for CO and other hydrocarbon products originates from the input CO₂ gas⁴⁴ (Supplementary Figs. 20 and 21, and Supplementary Table 4).

The light utilization and the charge carrier dynamics of the as-prepared catalysts were studied. The absorption edge of single-layer Ti_0.91O₂-SL displays a blue shift compared with layered Ti_0.91O₂-B due to the quantum confinement effect of monolayer structure, while the light absorption is enhanced after implantation of Cu single atoms (Supplementary Fig. 22). Photoelectrochemical (PEC) measurements confirm the enhanced charge separation and migration efficiency of Cu-Ti-V_O/Ti_0.91O₂-SL, ascribed to the 2D atomically-thin structure, which effectively shortens the charge transfer distance from body to surface and lowers charge recombination possibility (Supplementary Fig. 23). The fast charge carrier dynamics of Cu-Ti-V_O/Ti_0.91O₂-SL is kinetically favorable for the multi-electron reactions of generating C₂₊ products. Moreover, CO₂ adsorption isotherms reveal that Cu-Ti-V_O/Ti_0.91O₂-SL possesses the highest CO₂ uptake capacity, which is a priority for CO₂ activation and reduction (Supplementary Fig. 24).

Theoretical calculations for CO₂ photoreduction mechanism

The substantial suppression of CO production and promotion of C₂₊ production by Cu-Ti-V_O/Ti_0.91O₂-SL, in sharp contrast to pristine Ti_0.91O₂-SL, has demonstrated the key role of Cu-Ti-V_O units in coupling *CO intermediates into C₂₊ products. In fact, our in situ diffuse reflectance Fourier transform infrared spectroscopy (DRIFTS) characterization using ¹²CO₂ and ¹³CO₂ has detected the formation of the key *COOH, *CO, *CHO, and *CHOCO intermediates on Cu-Ti-V_O/Ti_0.91O₂-SL (Fig. 4a, Supplementary Figs. 25 and 26, and Supplementary Note 1), indicating the coupling of *CO into *CHOCO intermediates. However, it is not feasible to experimentally establish a direct spatial correlation between the active sites and the evolution of reaction intermediates to allow for deeper understanding on CO₂ photoreduction mechanism on atomic level. For this reason, density functional theory (DFT) calculations with a computational hydrogen electrode (CHE) model^45,46 was employed in an attempt to describe one potential mechanism for the reaction of CO₂-to-C₂₊ products on Cu-Ti-V_O/Ti_0.91O₂-SL.

Fig. 4. Mechanism studies of CO₂ reduction on Cu-Ti-V_O/Ti_0.91O₂-SL.

a In situ DRIFTS spectra of the photocatalytic CO₂ reduction on Cu-Ti-V_O/Ti_0.91O₂-SL. Gibbs free energy diagrams of CO₂ reduction on b Ti_0.91O₂ matrix, c Cu-O site, and d Cu-Ti-V_O unit.

In the calculations, Cu-Ti-V_O/Ti_0.91O₂-SL are featured with two types of catalytic centres for CO₂ reduction: the Cu atom-free Ti_0.91O₂ matrix domain and the Cu-Ti-V_O unit domain. On Ti_0.91O₂ domain, CO₂ is reduced to *CO through *COOH intermediate⁴⁷ (Fig. 4b and Supplementary Fig. 27). DFT results suggest the easy desorption of *CO on Ti_0.91O₂ matrix rather than further hydrogenation or C-C coupling (Supplementary Note 2), consistent with the experimental observation of the dominant CO product on pristine Ti_0.91O₂-SL. On Cu-Ti-V_O unit domain, the absorbed CO₂ is firstly converted to *CHO through *COOH and *CO intermediates⁴⁸ (Supplementary Fig. 27). Then, the *CHO at Cu-Ti-V_O unit may couple with the CO diffusing from neighboring Ti_0.91O₂ domain to generate *CHOCO^49–51 with free energy change of −1.63 eV (see Supplementary Note 3 for more details about potential pathways). The following C₁-C₂ coupling (*CH₂OCO + *CO → *CH₂OCOCO) is also calculated to be a thermodynamically-favorable exergonic reaction (−0.13 eV) (Supplementary Note 3). As a contrast, it is worth noting that on the Cu-O site without Vos, C-C coupling processes are found to be challenging owing to the large uphill energy changes, while the hydrogenation of *CO into CH₄ is more preferred (Fig. 4c)^52,53. Meanwhile, for Cu-Ti-V_O unit, some of the *C₂ species will continue to hydrogenate through a series of proton-electron steps to form C₂H₄. The free energy change of the potential determining step (PDS) is calculated as 0.62 eV for the C₃H₈ formation pathway (Fig. 4d) and 0.90 eV for the C₂H₄ formation pathway (Supplementary Fig. 28), suggesting the easier formation of C₃H₈ than C₂H₄. This is in accordance with our photocatalytic experimental observation of the higher yield of C₃H₈ than C₂H₄ on Cu-Ti-V_O/Ti_0.91O₂-SL. The overall reaction pathway for the reduction of CO₂ to C₃H₈ and C₂H₄ over Cu-Ti-V_O/Ti_0.91O₂-SL is tentatively described in Supplementary Fig. 29. In summary, the above results suggest a tandem catalysis mechanism^{13–16,54–56}, where the Cu-free Ti_0.91O₂ matrix may be preferential to participate in the reduction of CO₂ to CO, and Cu-Ti-V_O unit is more beneficial to the exergonic C-C coupling to C₂₊ products (Supplementary Fig. 30).

The above result suggests favorable exergonic reactions for both C₁-C₁ and C₁-C₂ couplings on Cu-Ti-V_O unit, which is in sharp contrast with Cu-O sites and other previously-reported catalysts with challenging endergonic C-C couplings^17–20. Such downhill energy changes of C-C coupling on Cu-Ti-V_O unit are probably owing to the low energy levels of the *CHOCO and *CH₂OCOCO intermediates (Supplementary Fig. 31). A stable multiple-bonding configuration containing one Cu-C bond and two Ti-O bonds is built for the adsorption of *CHOCO on Cu-Ti-V_O unit (Supplementary Fig. 32). Moreover, a five-membered ring is formed between *CH₂OCOCO and Cu-Ti-V_O unit, largely alleviating the electron accumulation and relaxing the intermolecular and intramolecular electrostatic repulsion (Supplementary Fig. 33). The electronic and geometric effect of Cu-Ti-V_O unit may jointly stabilize these key *C₂₊ intermediates and lower their adsorption energy levels to promote C-C couplings.

Discussion

A unique Cu-Ti-V_O unit in atomically thin Ti_0.91O₂ monolayer nanosheets was developed for highly efficient and selective photoconversion of CO₂ to C₃H₈. Theoretical calculations suggested that the Cu-Ti-V_O unit, as a favorable reaction centre for both C₁-C₁ and C₁-C₂ couplings, can effectively facilitate the multistep photocatalytic reduction of CO₂ by reducing the energy levels of the key *CHOCO and *CH₂OCOCO intermediates via electronic and geometric effects. A tandem mechanism and possible reaction pathway are proposed for the conversion of CO₂ to C₃H₈. The results of our study may open an alternative avenue for designing and synthesizing tandem photocatalysts with dual-metal active sites and coordination vacancies that modulate the behavior of reaction intermediates for the production of multicarbon fuels driven by light.

Methods

Synthesis of catalysts

Cu-Ti-V_O/Ti_0.91O₂-SL: Single-layer Ti_0.91O₂ nanosheets were synthesized according to previous literature²⁹, and the specific procedure is presented below. The parent Cs compound, Cs_0.7Ti_1.825O_4, was obtained by repeating twice the heat treatment (800 °C, 20 h) for the mixture of 10 mmol Cs₂CO₃ and 53 mmol anatase TiO₂. The interlayer Cs ions were extracted by stirring 5 g of the as-prepared Cs_0.7Ti_1.825O₄ in 500 mL of 1 M HCl solution for 24 h. After four cycles of acid exchange, the solid was washed with DI water to remove excess acid and then dried in a freeze dryer. Layered protonic lepidocrocite-type titanate of H_0.7Ti_1.825O₄ was obtained.

0.4 g of the as-prepared layered H_0.7Ti_1.825O₄ was shaken with 100 mL of 0.08 M tetrabutylammonium (TBA) hydroxide aqueous solution for a week to produce stable colloidal suspensions of atomically thin Ti_0.91O₂ single layers. The colloidal suspensions were then dried in a freeze dryer.

The Cu-en precursor was prepared by mixing 30 mL of 0.025 g/L CuCl₂‧2H₂O aqueous solution with 360 μL of ethanediamine (en) at room temperature. Then, this Cu-en precursor was added to 7 g of the as-prepared colloidal suspensions of single-layer Ti_0.91O₂. After stirring for 5 h, the solid was filtrated, washed with deionized water, and dried in a freeze dryer to obtain a single-layer Cu-en/Ti_0.91O₂ sample. Then, the Cu-Ti-V_O/Ti_0.91O₂-SL sample was prepared by rapid thermal treatment (RTT) of single-layer Cu-en/Ti_0.91O₂ in an Ar atmosphere at 500 °C for 1 min³². Briefly, the powders were put into a quartz tube, which was then inserted into a tube furnace preheated to 500 °C. Under Ar flow, the powders were kept at that temperature for 1 min, and then the quartz tube was quickly removed and rapidly cooled to room temperature.

Cu-O/Ti_0.91O₂-SL: The sample was prepared by the same method as Cu-Ti-V_O/Ti_0.91O₂-SL, except that the RTT process was conducted in an air atmosphere.

Cu-Ti-V_O/Ti_0.91O₂-SL (lower): The sample was prepared by the same method as Cu-Ti-V_O/Ti_0.91O₂-SL, except that an aqueous solution of 0.0125 g/L CuCl₂‧2H₂O was used.

Cu NC/Ti_0.91O₂-SL: The sample was prepared by the same method as Cu-Ti-V_O/Ti_0.91O₂-SL, except that an aqueous solution of 0.1 g/L CuCl₂‧2H₂O was used.

Cu NP/Ti_0.91O₂-SL: The sample was prepared by the same method as Cu-Ti-V_O/Ti_0.91O₂-SL, except that an aqueous solution of 0.2 g/L CuCl₂‧2H₂O aqueous was used, and RTT was maintained at 10 min.

Ti_0.91O₂-B: The sample was prepared by treating layered protonic lepidocrocite-type titanate of H_0.7Ti_1.825O₄ with the same RTT process as used for Cu-Ti-V_O/Ti_0.91O₂-SL.

Characterization

XRD data were measured on an X-ray diffractometer (Rigaku Ultima III, Japan) by Cu-Kα radiation (λ = 0.154178 nm) at 40 kV and 40 mA with a scan rate of 5° min⁻¹. AFM analysis was performed on an MFP3D microscope (Asylum Research, MFP-3D-SA, USA) with a cantilever operating in tapping mode. The morphology was characterized by FE-SEM (FEI NOVA NANOSEM 230). TEM images were taken on an FEI Tecnai F20 TEM apparatus. Atomic-resolution STEM-HAADF images were obtained on a double spherical aberration-corrected STEM/TEM FEI Titan3 Cubed 60–300, and the samples were suspended on micron-scale carbon grids of Mo mesh. The Cu concentration was determined by ICP–OES with an Avio200 instrument. The XAFS spectra (Cu K-edge and Ti K-edge) were collected at the 1W1B station at the Beijing Synchrotron Radiation Facility (BSRF). The storage rings at BSRF were operated at 2.5 GeV with an average current of 250 mA. Using a Si (111) double-crystal monochromator, the data collection was conducted in transmission/fluorescence mode using an ionization chamber. All spectra were collected under ambient conditions. The EXAFS data were processed according to standard procedures using the ATHENA module implemented in the IFEFFIT software package. The k³-weighted EXAFS spectra were obtained by subtracting the postedge background from the overall absorption and then normalizing with respect to the edge jump step. Subsequently, k³-weighted χ(k) data of the Cu K-edge underwent Fourier transform to real (R) space using Hanning windows (dk = 1.0 Å⁻¹) to separate the EXAFS contributions from different coordination shells. To obtain the quantitative structural parameters around the central atoms, least squares curve parameter fitting was performed using the ARTEMIS module of IFEFFIT software packages. The chemical states of the samples were detected by XPS and AES, which were equipped with an ultrahigh vacuum Thermo Fisher Scientific electron spectrometer by using Al Kα radiation (1486.6 eV) as the X-ray source, and the binding energies were calibrated according to the C 1 s peak of adventitious carbon at 284.6 eV. The in situ DRIFTS spectra were obtained on a Bruker IFS 66 v FT spectrometer with Harrick diffuse reflectance with ZnSe and quartz windows at BL01B in NSRL, Hefei.

Measurements of photocatalytic CO₂ reduction

Photocatalytic CO₂ reduction experiments were performed in a Pyrex reaction vessel with a top irradiation window. Typically, 10 mg of photocatalyst powder was suspended in a 15 mL CO₂-saturated solution containing 12.5 mL acetonitrile and 2.5 mL H₂O in a 166 mL quartz reaction cell. Before illumination, the system was filled with CO₂ (purity >99.999%) to 1 atm. The reactor was then irradiated by a 300 W Xe lamp (PerfectLight, PLS-SXE300). During irradiation, 0.1 mL of gas was collected from the reaction headspace every hour, and the gaseous products were analysed by using gas chromatography (GC-2014C, Shimadzu Corp., Japan). The isotope experiment was conducted using ¹³CO₂ as feedstock, and the products were analysed using gas chromatography–mass spectrometry (7890 A and 5975 C, Agilent). The quantum efficiency was evaluated using a LED lamp (PerfectLight) with the wavelength of 385,415, or 520 nm as the light source.

The electron-based selectivity of C₃H₈ was calculated using (1):

$${\mathrm{Sel}}_{\mathrm{electron}}\left({\mathrm{C}}_3{\mathrm{H}}_8\right)=\left(\frac{n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)\times 20}{n\left(\mathrm{CO}\right)\times 2+n\left(\mathrm{C}{\mathrm{H}}_4\right)\times 8+n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)\times 12+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)\times 20}\right)\times 100\%$$ 1

The product-based selectivity of C₃H₈ was calculated using (2):

$${\mathrm{Sel}}_{\mathrm{product}}\left({\mathrm{C}}_3{\mathrm{H}}_8\right)=\left(\frac{n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)}{n\left(\mathrm{CO}\right)+n\left(\mathrm{C}{\mathrm{H}}_4\right)+n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)}\right)\times 100\%$$ 2

The electron-based selectivity of C₂₊ products was calculated using (3):

$${\mathrm{Sel}}_{\mathrm{electron}}\left({\mathrm{C}}_{2+}\right)=\left(\frac{n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)\times 12+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)\times 20}{n\left(\mathrm{CO}\right)\times 2+n\left(\mathrm{C}{\mathrm{H}}_4\right)\times 8+n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)\times 12+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)\times 20}\right)\times 100\%$$ 3

The product-based selectivity of C₂₊ products was calculated using (4):

$${\mathrm{Sel}}_{\mathrm{product}}\left({\mathrm{C}}_{2+}\right)=\left(\frac{n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)}{n\left(\mathrm{CO}\right)+n\left(\mathrm{C}{\mathrm{H}}_4\right)+n\left({\mathrm{C}}_2{\mathrm{H}}_4\right)+n\left({\mathrm{C}}_3{\mathrm{H}}_8\right)}\right)\times 100\%$$ 4

where n is the formation rate.

The quantum efficiency was calculated using (5):

$$\mathrm{QE}=\frac{N\left(\mathrm{CO}\right)\times 2+N\left(\mathrm{C}{\mathrm{H}}_4\right)\times 8+N\left({\mathrm{C}}_2{\mathrm{H}}_4\right)\times 12+N\left({\mathrm{C}}_3{\mathrm{H}}_8\right)\times 20}{N\left(\mathrm{photon}\right)}\times 100\%$$ 5

where N is the number of evolved gas molecules or incident photons.

Computational details

All calculations were performed based on density functional theory (DFT) through the Vienna ab initio Simulation Package (VASP)⁵⁷. Projector-augmented-wave (PAW) pseudopotentials⁵⁸ were used to treat the core electrons, while interactions between electrons were described by the Perdew-Burke-Ernzerhof (PBE)⁵⁹ exchange-correlation functional of the generalized gradient approximation (GGA). For all the calculations, the vacuum space in the z-direction was 20 Å to avoid potential interactions between periodic surfaces. The DFT-D3 method of Grimme⁶⁰ was applied to describe the van der Waals dispersion forces. DFT + U approach^61,62 with U = 3.5 eV was considered to evaluate the influence of strongly correlated d electrons on the calculated free energies. According to our previous report⁶³, the results of calculated free energies obtained from DFT + U and DFT are consistent. Therefore, regular DFT was employed in this work. The Monkhorst-Pack k-point grid with 3 × 3 × 1 mesh was applied until the maximal forces on each ion were smaller than 0.02 eV/Å. The convergence criterion of the energy was set to 10⁻⁴ eV, and a cut-off energy of 450 eV was used for the plane wave expansion. The Gibbs free energy change (ΔG) was defined as follows^45,64:

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }{E}_{\mathrm{ZPE}}-T\mathrm{\Delta }S$$ 6

where ΔE is the energy difference between the reactants and product obtained through DFT calculations. ΔE_ZPE and ΔS are the changes in the zero-point energies (ZPE) and entropy⁶⁵. T represents the temperature and was set as 298.15.

Statistics & reproducibility

No statistical method was used to predetermine the sample size. No data were excluded from the analyses. The experiments were not randomized.

Source data

Source data ^{(577.1KB, zip)}

Author contributions

Y.S, Y.Zho., Y.X, J.W, Q.S., and Z.Z conceived the idea and designed the present work. C.R. performed the DFT calculations. Y.S., X.X., Y.Zha., Yi.Y., and H.C. conducted the experiments. L.Z. performed the XANES and EXAFS measurements. R.L. and W.Z. performed the in situ DRIFTS test. Yo.Y. contributed to the ¹³CO₂ isotope labeling experiment.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work.

Data availability

The data that support the findings of this study are available within the paper and its supplementary information files or are available from the corresponding authors upon reasonable request. Source data are provided with this paper.

Code availability

All related codes regarding DFT simulations in this study are provided in the Source Data file.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-023-36778-5.