Near-unity CO2-to-ethylene photoconversion over low coordination single-atom catalysts

TL;DR

A new catalyst efficiently converts CO2 into ethylene using atomic-level coordination engineering, achieving near-complete selectivity.

Contribution

A low-coordination manganese single-atom catalyst is developed for high-efficiency CO2-to-ethylene conversion.

Findings

Manganese single-atom catalyst in zinc sulfide achieves 99.1% ethylene selectivity.

Sulfur vacancies at coordination sites enhance CO adsorption and C–C coupling efficiency.

Abstract

Photocatalytic conversion of carbon dioxide to value-added chemicals, particularly multi-carbon products, offers a promising route toward carbon-neutral cycles. However, achieving high activity and selectivity remains extremely challenging due to the instability of key reaction intermediates and limited C–C coupling efficiency. Herein, we report a low-coordination manganese single-atom catalyst embedded in zinc sulfide (Mn1–ZnSv) that enables efficient and selective CO2-to-C2+ conversion. In-situ spectroscopic analyses and density functional theory calculations reveal that sulfur vacancies are created at the Mn single-atom coordination sites and induce the formation of coordination-unsaturated Mn-S2 configuration. The asymmetric coordination environment of Mn modulates local charge distribution, strengthens *CO adsorption, and promotes *CO and *CHO coupling to form the *COCHO intermediate for efficient C–C coupling. As a result, the Mn1–ZnSv catalyst achieved 99.1% selectivity for ethylene with a formation rate of 76.6 μmol g-1 h-1. This study highlights the critical role of atomic-level coordination engineering in advancing photocatalytic CO2-to-C2+ conversion.

This work develops an atomic-level coordination engineering strategy that strengthens key intermediate adsorption to promote C–C coupling, enabling highly selective photocatalytic CO2 conversion to ethylene.

Figures

Schematic diagram showing regulatory effect of coordination environment on *CO adsorption.a On the pristine ZnS, the weak adsorption of *CO intermediates reduce the surface covering and then yields C_1_ product. b The low-coordinated Mn single-atom sites enhance the adsorption of *CO intermediates via charge modulation, facilitating asymmetric C–C coupling.

Theoretical guidance.a Schematic of calculation models for pristine ZnS, Mn_1_–ZnS, and Mn_1_–ZnS_v_ (the atoms marked with numbers represent metal atoms in different coordination states). b The free energy diagram for CO_2_ reduction reaction to display the conversion of CO_2_-to-*CO over ZnS, Mn_1_–ZnS and Mn_1_–ZnS_v_. c Illustration of the interaction between the 5*σ* and 2*π of *CO and metal surfaces. d Adsorption energies of *CO on Mn_1_–ZnS and Mn_1_–ZnS_v_. e, f Differential charge density and Bader charge of *CO adsorbed on Mn_1_–ZnS_v_ (e) and Mn_1_–ZnS (f) (the isosurface is 0.005 e

Structural characterization.a, b TEM image (a), STEM image and the corresponding EDS elemental mappings (b) of Mn_1_–ZnS. Inset: An enlarged HRTEM of the dotted line area. c ADF-STEM image of Mn_1_–ZnS_v_ with corresponding intensity profile of the representative region (red dotted line), and the isolated single atoms highlighted with yellow circles. d Normalized Mn K-edge XANES spectra of Mn_1_–ZnS, Mn_1_–ZnS_v_, and the related reference samples. e Fourier-transform (FT) spectra of EXAFS oscillations for Mn_1_–ZnS and Mn_1_–ZnS_v_. f, g EXAFS spectrum fitting result of Mn_1_–ZnS (f) and Mn_1

In situ spectra analysis.a, b In situ DRIFTS for CO_2_ photoreduction reaction over Mn_1_–ZnS (a) and Mn_1_–ZnS_v_ (b). c, d CO-adsorbed DRIFTS for Mn_1_–ZnS (c) and Mn_1_–ZnS_v_ (d). e, f Reaction pathway schemes for ZnS and Mn_1_–ZnS_v_ interface. Source data are provided as a Source Data file.

Photocatalytic CO_2_ reduction performance.a Product yields of products (HCOOH, CO, C_2_H_4_, and CH_4_) and the selectivity of C_2_H_4_ product over ZnS, Mn_1_–ZnS, and Mn_1_–ZnS_v_ catalysts (*T* = 298 K; irradiation time = 4 h; solvent: H_2_O, 5 mL: amount of catalyst: 0.2 *g*). b The evolution of C_2_H_4_ product during photocatalytic CO_2_ reduction over time (*n* = 3 independent experiments, data are presented as mean values ± SD). c The products yield over ZnS, Mn_1_–ZnS and Mn_1_–ZnS_v_ catalysts using CO_2_ and CO as reactant, respectively. d The AQE of C_2_H_4_ evolution catalyzed by

Full text

Introduction

Solar-driven conversion of carbon dioxide (CO_2_) into multi-carbon (C_2+) solar fuels has garnered significant attention as a promising green solution to address global energy demands and mitigate the impacts of climate change^1–3^. Although recent advances have demonstrated the potential of photocatalytic systems for CO_2-to-C_2+_ conversion^4^, achieving high selectivity remains a critical challenge, because the formation of C_2+_ products relies heavily on the stability of carbon-based intermediates at the catalytic active sites to enable efficient C–C coupling^5–7^.

Among various developed photocatalysts for CO_2_ reduction reaction (CO_2_RR), transition metal sulfides, such as CdS, Bi_2_S_3_, CulnSnS_4_, have attracted particular interest due to their tunable d-orbitals^8–10^. However, these materials often exhibit weak adsorption of C_1_ intermediates, e.g., *CO, leading to premature desorption^11–13^ and preferential formation of C_1_ products, such as CO, CH_4_, and HCOOH, rather than the desired C_2+_ compounds (Fig. 1a).Fig. 1. Schematic diagram showing regulatory effect of coordination environment on *CO adsorption.a On the pristine ZnS, the weak adsorption of *CO intermediates reduce the surface covering and then yields C_1_ product. b The low-coordinated Mn single-atom sites enhance the adsorption of *CO intermediates via charge modulation, facilitating asymmetric C–C coupling.

To address this limitation, several effective strategies, including heterojunction construction, crystal facet engineering, doping, and defect engineering, have been explored to enhance C_2+_ selectivity by stabilizing carbon-based intermediates^14–17^. For instance, in MoS_x_/Fe_2_O_3_ system, d-p orbital hybridization at Mo-Fe sites reduced electrostatic repulsion between *CO and *COH intermediates, promoting ethylene (C_2_H_4_) formation^18^. Co-O-Fe triatomic sites in partially oxidized FeCoS_2_ increased local charge density, enriching C_1_ intermediates and directing the reaction pathway toward C_2_ products^19^. Incorporation of Co into NiS_2_ tailored the coordination number and oxide state of Ni sites, facilitating asymmetric C–C coupling between *CO species in atop and bridge adsorption configuration, achieving 75% selectivity for C_2_H_4_^20^. Despite these advances, achieving high activity and selectivity for photocatalytic CO_2_-to-C_2+_ conversion remains an unmet goal.

Single-atom catalysts (SACs), with atomically dispersed metal centers and well-defined coordination environments, provide a powerful platform for tuning intermediate adsorption and enhancing CO_2_RR activity and selectivity^18,21–24^. Particularly, low-coordination SACs can break geometric symmetry and modulate the local electronic structure, strengthening the adsorption of carbon-based intermediates and facilitating C–C coupling^25,26^. Therefore, integrating low-coordination SACs into binary sulfides holds great potential for stabilizing *CO and enabling efficient photocatalytic CO_2_-to-C_2+_ conversion (Fig. 1b).

Herein, we develop a low-coordination Mn single-atom embedded zinc sulfide (Mn_1_–ZnS_v_) via a microwave irradiation-induced targeted defect engineering strategy for enhancing CO_2_-to-C_2_H_4_ photoreduction. Structural analyses reveal that Mn atoms are atomically dispersed and coordinated with a reduced number of sulfur atoms, forming Mn–Zn coupled sites with modulated electronic structures. Density functional theory (DFT) calculations and in-situ spectroscopic studies demonstrate that these low-coordination Mn sites effectively stabilize C_2_H_4_-related intermediates and promote asymmetric *CO–CHO coupling into *COCHO, a key intermediate for C–C bond formation. As a result, Mn_1_–ZnS_v_ exhibits attractive photocatalytic performance for CO_2_ reduction, achieving a C_2_H_4_ production rate of 76.6 μmol g^−1^ h^−1^ with nearly 100% selectivity, significantly emphasizing the capability of regulating the coordination environment of sing-atom catalytic sites for photoconversion CO_2_ into C_2+_ products.

Results

Enhanced C–C coupling via low-coordination Mn sites

To assess CO formation during the CO_2_RR, we calculated the free energy profiles for CO_2_-to-CO conversion on ZnS, Mn_1_–ZnS, and Mn_1_–ZnS_v_ (Fig. 2a, Supplementary Figs. 1–3, and Supplementary Data 1). Among these, Mn_1_–ZnS_v_ exhibits the lowest energy barrier (0.56 eV), significantly lower than that of ZnS and Mn_1_–ZnS (Fig. 2b), indicating that low-coordination structure of Mn single-atom is favorable for CO_2_ activation and CO generation.Fig. 2. Theoretical guidance.a Schematic of calculation models for pristine ZnS, Mn_1_–ZnS, and Mn_1_–ZnS_v_ (the atoms marked with numbers represent metal atoms in different coordination states). b The free energy diagram for CO_2_ reduction reaction to display the conversion of CO_2_-to-CO over ZnS, Mn_1_–ZnS and Mn_1_–ZnS_v_. c Illustration of the interaction between the 5σ and 2π* of *CO and metal surfaces. d Adsorption energies of CO on Mn_1_–ZnS and Mn_1_–ZnS_v_. e, f Differential charge density and Bader charge of CO adsorbed on Mn_1_–ZnS_v_ (e) and Mn_1_–ZnS (f) (the isosurface is 0.005 e/Å^3^). Blue and purple represent charge gaining and losing, respectively. g, h The partial density of states and the d-band center of the Mn atom for Mn_1_–ZnS and Mn_1_–ZnS_v_, ε_d_ represents the d-band center. i The free minimum-energy diagram for *CO hydrogenation and subsequent C–C coupling on Mn_1_–ZnS_v_ (111) facet. Source data are provided as a Source Data file.

To investigate the behavioral of CO intermediates on different coordination structures, the CO adsorption energy and protonation energy barriers of Mn_1_–ZnS and Mn_1_–ZnS_v_ were compared. The interaction between the Mn 3d orbital and the 2π** (LUMO) and 5σ* (HOMO) orbitals of *CO underpins the modulation of adsorption behavior (Fig. 2c). The low-coordination Mn site in Mn_1_–ZnS_v_ strengthens the Mn-CO π backbonding, resulting in a significantly enhanced *CO adsorption energy of −1.36 eV, much stronger than that of coordination-saturated Mn_1_–ZnS (Fig. 2d). Furthermore, the incorporation of low-coordination Mn single atoms ingeniously balances the competitive adsorption of *CO intermediates and *H species (Supplementary Fig. 4). The appropriate binding inhibits premature *CO desorption and enables its participation in subsequent protonation steps. It is worth noting that the energy barrier for *CO hydrogenation on Mn_1_–ZnS_v_ is reduced to 0.75 eV, compared to 1.25 eV on Mn_1_–ZnS, confirming the regulatory role of the low-coordination structure in steering the *CO reaction pathway (Supplementary Fig. 5).

To probe the origin of the stronger CO adsorption, we investigated the electronic structure differences between Mn_1_–ZnS and Mn_1_–ZnS_v_. Differential charge density analysis reveals a significant electron transfer (1.08 |e|) from the Mn single atom to the adsorbed CO in Mn_1_–ZnS_v_ (Fig. 2e), compared to just 0.32 |e| in Mn_1_–ZnS (Fig. 2f), highlighting the enhanced electronic interaction in the low-coordination structure. Furthermore, in Mn_1_–ZnS_v_ (coordination number, CN = 2), the Mn 3d orbital center shifts upward relative to Mn_1_–ZnS (Fig. 2g), resulting in a higher d-band center (Fig. 2h). This upshift implies an increased population of anti-bonding states, favoring stronger adsorption of reaction intermediates and facilitating the key C–C coupling process.

To validate this mechanistic insight, we calculated the free energy barriers for C–C coupling through different C_1_ species (*CO, *CHO, *COH) on Mn_1_–ZnS_v_ (Fig. 2i). The results suggested that direct dimerization of two *CO species into *COCO is energetically unfavorable (ΔG = 1.91 eV). Instead, *CO preferentially undergoes hydrogenation to *CHO, which subsequently couples with another *CO to form the *COCHO intermediate via an asymmetric C–C bond formation pathway (Supplementary Fig. 6). This pathway is thermodynamically more favorable and constitutes the lowest-energy pathway for CO_2_-to-C_2_H_4_ conversion (Supplementary Figs. 7 and 8 and Supplementary Table 1). Thus, low-coordination Mn single-atom sites in Mn_1_–ZnS_v_ significantly enhance *CO adsorption and facilitates the asymmetric coupling of *CO and *CHO, ultimately leading to the selective generation of C_2_H_4_. The theoretical screening of coordination structures serves as a guide for establishing trends in *CO adsorption and the subsequent reaction pathways.

Catalyst synthesis and characterization

Guided by the DFT-screened Mn_1_–ZnS_v_ structure, a Mn single-atom substituted ZnS catalyst with targeted sulfur vacancies (S_v_) was successfully synthesized via a microwave irradiation-induced targeted defect strategy (Supplementary Fig. 9), followed by H_2_O_2_ etching under an Ar atmosphere. Inductively coupled plasma optical emission spectrometry (ICP-OES) confirmed a Mn content of 0.6 wt% (Supplementary Table 2). X-ray diffraction (XRD) patterns (Supplementary Fig. 10) confirmed the formation of cubic-phase ZnS (PDF #05-0566)^27^ with an average crystallite size of 41.3 ± 0.2 nm, as estimated using the Scherrer equation (Supplementary Fig. 11).

High-resolution transmission electron microscopy (HRTEM, Fig. 3a) revealed distinct lattice fringes with an interplanar spacing of 0.312 nm, corresponding to the (111) plane of ZnS. Scanning transmission electron microscopy–energy-dispersive X-ray spectroscopy (STEM-EDS) elemental mapping images (Fig. 3b) and the STEM-EDS line scanning profiles (Supplementary Fig. 12) corroborated the even distribution of Zn, Mn, and S throughout the sample, with no observable aggregation of Mn into large clusters or particles. Aberration-corrected annular dark-field STEM (ADF-STEM, Fig. 3c) further confirmed the atomic dispersion of Mn, with isolated single atoms highlighted by yellow circles and localized intensity dips in the corresponding line-scan profile across a representative region.Fig. 3. Structural characterization.a, b TEM image (a), STEM image and the corresponding EDS elemental mappings (b) of Mn_1_–ZnS. Inset: An enlarged HRTEM of the dotted line area. c ADF-STEM image of Mn_1_–ZnS_v_ with corresponding intensity profile of the representative region (red dotted line), and the isolated single atoms highlighted with yellow circles. d Normalized Mn K-edge XANES spectra of Mn_1_–ZnS, Mn_1_–ZnS_v_, and the related reference samples. e Fourier-transform (FT) spectra of EXAFS oscillations for Mn_1_–ZnS and Mn_1_–ZnS_v_. f, g EXAFS spectrum fitting result of Mn_1_–ZnS (f) and Mn_1_–ZnS_v_ (g) in R space. (The inserted image shows the atomic structure model of Mn_1_–ZnS and Mn_1_–ZnS_v_, where the balls in blue, pink, and yellow represent Zn, Mn, and S atoms). h, i Wavelet transform (WT) contour plots of Mn foil (h) and Mn_1_–ZnS_v_ (i), respectively. Source data are provided as a Source Data file.

To confirm the formation of S_v_, electron paramagnetic resonance (EPR) spectroscopy was performed (Supplementary Fig. 13). A clear signal at g = 2.005 indicated the presence of S_v_ in Mn_1_–ZnS_v_^28^. Additionally, EDS point analysis (Supplementary Fig. 14) showed an increased metal-to-sulfur atomic ratio, further supporting the generation of S_v_, which increased with prolonged H_2_O_2_ etching time. Quantitative analysis from ICP-OES and EPR^29,30^ (Supplementary Figs. 15 and 16 and Supplementary Table 2) revealed an isolated S_v_ concentration of 1.54% for Mn_1_–ZnS_v_, closely matching the theoretically identified optimal configuration by DFT (Supplementary Fig. 17 and Supplementary Data 1).

To probe the local coordination environment of Zn and Mn atoms, X-ray absorption spectroscopy was conducted. X-ray absorption near-edge structure (XANES) spectra (Supplementary Fig. 18a) showed that Zn maintained an oxidation state of approximately +2.0. The Mn K-edge positions in both Mn_1_–ZnS and Mn_1_–ZnS_v_ were located between those of metallic Mn and MnO, indicating an average Mn valence between 0 and +2 (Fig. 3d). Notably, Mn in Mn_1_–ZnS_v_ exhibited a lower oxidation state than in Mn_1_–ZnS, attributed to the low-coordination environment induced by S_v_ formation.

Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) analysis (Fig. 3e, Supplementary Fig. 18, and Supplementary Table 3) displayed distinct peaks corresponding to Zn–S and Zn–Zn coordination, while a prominent peak at 1.96 Å was assigned to Mn–S bonding. The absence of Mn–Mn and Mn–O coordination in both Mn_1_–ZnS and Mn_1_–ZnS_v_ indicates that Mn atoms are atomically dispersed and coordinated with neighboring S atoms, consistent with ADF-STEM results^31^.

EXAFS fitting revealed that the first-shell Mn–S coordination number was 3.7 for Mn_1_–ZnS and 1.9 for Mn_1_–ZnS_v_ (Fig. 3f, g and Supplementary Table 3). These values are in excellent agreement with DFT-optimized structures, confirming that Mn atoms in Mn_1_–ZnS_v_ are coordinated with only two sulfur atoms (insets in Fig. 3f, g). No Mn–Mn path in second-shell coordination (Supplementary Fig. 19) and wavelet transform (WT) analysis further confirmed the atomic dispersion and low-coordination environment of Mn in Mn_1_–ZnS_v_ (Fig. 3h, i).

To investigate the effect of low-coordination Mn single-atoms on light absorption efficiency, UV–Vis diffuse reflectance spectra and X-ray spectroscopy (XPS) valence band spectroscopy were conducted. The optical absorption band edge of Mn_1_–ZnS_v_ is observed at 428 nm, corresponding to a smaller band gap of 2.88 eV (Supplementary Fig. 20). Ultraviolet photoelectron spectroscopy measurements revealed work functions (Φ) of 6.13 and 5.60 eV for Mn_1_–ZnS and Mn_1_–ZnS_v_, respectively (Supplementary Fig. 21). The valence band maxima for Mn_1_–ZnS and Mn_1_–ZnS_v_ were located at 1.00 and 0.55 eV below the Fermi level (Supplementary Fig. 22), and the electronic band structure versus Normal Hydrogen Electrode could be elucidated (Supplementary Fig. 23), which is thermodynamically favorable for the CO_2_-to-C_2_H_4_ conversion (Supplementary Fig. 24). These results are well-consistent with those obtained from Mott–Schottky plots (Supplementary Fig. 25), clearly suggesting the ability for CO_2_-to-C_2_H_4_ conversion and O_2_ evolution. Upon irradiation, photogenerated electrons are transferred from S (electron donor) to Mn (electron trap), and then rapidly injected into the adsorbed CO_2_ molecules through the electron transfer channel (Supplementary Fig. 26). Moreover, improved charge-transfer dynamics induced by low-coordination Mn sites were recorded in Supplementary Figs. 27–29.

To further elucidate intermediate adsorption and reaction pathways in photocatalytic CO_2_ reduction, in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed (Fig. 4a, b and Supplementary Table 4). Under simulated solar irradiation, Mn_1_–ZnS exhibited characteristic signals at 1618, 1655, and 1695 cm^−^^1^ (*COOH species)^26,32,33^, 2078 cm^−^^1^ (*CO)^34^, 1122 and 1415 cm^−^^1^ (*CHO)^33,35^, and a key band at 1315 cm^−^^1^ corresponding to *COCHO (Fig. 4a)^32^, a critical intermediate in C–C coupling toward C_2_H_4_^36,37^.Fig. 4. In situ spectra analysis.a, b In situ DRIFTS for CO_2_ photoreduction reaction over Mn_1_–ZnS (a) and Mn_1_–ZnS_v_ (b). c, d CO-adsorbed DRIFTS for Mn_1_–ZnS (c) and Mn_1_–ZnS_v_ (d). e, f Reaction pathway schemes for ZnS and Mn_1_–ZnS_v_ interface. Source data are provided as a Source Data file.

In contrast, Mn_1_–ZnS_v_ exhibited a stronger *CO signal at 2078 cm^−^^1^ (Fig. 4b and Supplementary Fig. 30), indicating enhanced stabilization and coverage of *CO intermediate. The progressive intensification of the 1315 cm^−^^1^ band over time confirmed efficient *CO–*CHO coupling and *COCHO formation, which was further verified by the ^13^CO labeling experiment (Supplementary Fig. 31). Additional bands at 1442, 1462, and 1364 cm⁻^1^ were attributed to *C_2_H_4_ and *CH_2_ intermediates^32,38^, confirming the high activity and selectivity of Mn_1_–ZnS_v_ toward C_2_H_4_ production.

To further probe CO adsorption behavior, CO-adsorbed DRIFTS was performed on both catalysts (Fig. 4c, d). Prominent peaks at ~2084 and 2036 cm^−^^1^ were assigned to linearly bonded CO (CO_L_) on Mn single atoms. Mn_1_–ZnS_v_ displayed significantly higher peak intensities than Mn_1_–ZnS, indicating a stronger CO binding capability, consistent with DFT predictions. Based on the in situ DRIFTS data and theoretical insights, a photocatalytic reaction pathway for CO_2_ conversion on Mn_1_–ZnS_v_ was proposed (Fig. 4e, f), highlighting the critical role of low-coordination Mn single-atom sites in stabilizing reaction intermediates, enabling *CO–*CHO coupling, and ultimately promoting selective C_2_H_4_ production.

Evaluating catalyst performance for CO2 photoreduction

The CO_2_ photoreduction performance of Mn_1_–ZnS_v_ catalysts was assessed under visible light irradiation (λ ≥ 380 nm) without the use of photosensitizers or sacrificial agents (Fig. 5, Supplementary Figs. 32–34, and Supplementary Table 5). As shown in Fig. 5a, pristine ZnS primarily produced CO at a rate of 64.6 μmol g^−^^1^ h^−^^1^, with a low C_2_H_4_ selectivity of only 5.6%. Compared to pristine ZnS, ZnS with sulfur vacancies (ZnS_v_) exhibits negligible enhancement in C_2_H_4_ yield and selectivity (7.3%) (Supplementary Figs. 35 and 36). Upon incorporation of saturated-coordination Mn single atoms (Mn_1_–ZnS), both activity and selectivity improved significantly, achieving a C_2_H_4_ formation rate of 47.5 μmol g^−^^1^ h^−^^1^ and a selectivity of 74.5%. Concurrently, CO evolution decreased to 28.4 μmol g^−^^1^ h^−^^1^ and the competing hydrogen evolution reaction was markedly suppressed (Supplementary Fig. 37), likely due to the *CO-stabilizing effect of Mn sites that promote its protonation and subsequent C–C coupling.Fig. 5. Photocatalytic CO_2_ reduction performance.a Product yields of products (HCOOH, CO, C_2_H_4_, and CH_4_) and the selectivity of C_2_H_4_ product over ZnS, Mn_1_–ZnS, and Mn_1_–ZnS_v_ catalysts (T = 298 K; irradiation time = 4 h; solvent: H_2_O, 5 mL: amount of catalyst: 0.2 g). b The evolution of C_2_H_4_ product during photocatalytic CO_2_ reduction over time (n = 3 independent experiments, data are presented as mean values ± SD). c The products yield over ZnS, Mn_1_–ZnS and Mn_1_–ZnS_v_ catalysts using CO_2_ and CO as reactant, respectively. d The AQE of C_2_H_4_ evolution catalyzed by Mn_1_–ZnS_v_ catalyst and the correlation between the related absorption spectra and absorption wavelengths (n = 3 independent experiments, data are presented as mean values ± SD). e The control experiments of photocatalytic CO_2_ reduction performance over Mn_1_–ZnS_v_ under altered conditions. f Cycling tests for CO_2_ photoreduction to C_2_H_4_ over Mn_1_–ZnS_v_ photocatalyst under the same as reaction conditions. Source data are provided as a Source Data file.

Notably, the low-coordination Mn_1_–ZnS_v_ catalyst delivered a significantly enhanced C_2_H_4_ formation rate of 76.6 μmol g^−^^1^ h^−^^1^, nearly 58.9 times higher than that of pristine ZnS, with a selectivity of 99.1% for C_2_H_4_ production (Fig. 5b and Supplementary Table 6). Meanwhile, CO formation further decreased to 4.2 μmol g^−^^1^ h^−^^1^ (Supplementary Fig. 37) and no liquid products were detected, underscoring the pivotal role of Mn coordination environment in modulating *CO adsorption and steering the reaction pathway toward selective C_2+_ hydrocarbon formation.

To further validate the role of low-coordination Mn SAs in promoting C–C coupling, a CO substitution experiment was carried out. This resulted in a C_2_H_4_ yield of 198.8 μmol g^−^^1^ (Fig. 5c), confirming that *CO is a crucial intermediate for C–C coupling and effectively activating and stabilizing by the low-coordination Mn sites. The apparent quantum efficiency (AQE) for CO_2_ photoreduction over Mn_1_–ZnS_v_ reached 8.1% at 420 nm (Fig. 5d). Control experiments (Fig. 5e) demonstrated that C_2_H_4_ formation occurred only under light irradiation in the presence of both CO_2_ and H_2_O, excluding contributions from dark reactions or carbon-containing contaminants^39^. Isotope-labeling experiments using ^13^CO_2_ and D_2_O (Supplementary Fig. 38) further confirmed CO_2_ and H_2_O as the carbon and proton sources, respectively, for C_2_H_4_ formation. Considering that the activation of H_2_O into active hydrogen species may be an important factor that affecting the turnover-limiting step during the photocatalytic CO_2_RR to C_2_H_4_, the KIE of H/D value over the Mn_1_–ZnS_v_ catalyst is close to 1 (Supplementary Fig. 39), further confirming the *CHO formation as the rate-determining step suggested by DFT results.

The photocatalytic stability of Mn_1_–ZnS_v_ was evaluated over 50 consecutive cycles (200 h), during which no significant decline in activity or selectivity was observed (Fig. 5f). Post-reaction characterizations, including TEM, XRD, and XPS (Supplementary Fig. 40), confirmed the preservation of the structural and electronic integrity of the catalyst. No characterized signals of Mn–Mn bond over Mn_1_–ZnS_v_ after long-term reaction further proved the great stability of the coordination environment of Mn atoms (Supplementary Fig. 41), ensuring effective photocatalytic performance. Furthermore, the well-balanced charge consumption between the oxidation and reduction half-reactions (RO/R ≈ 1) minimizes charge accumulation and effectively suppresses photocorrosion (Supplementary Table 5). Notably, under sacrificial-agent-free conditions, a comparative analysis of photocatalytic performance between our study and previous reported catalysts^17–19,32,40–50^ is displayed in Supplementary Table 6. Both the competitive C_2_H_4_ yield and selectivity of Mn_1_–ZnS_v_ underscores its potential as a highly efficient and robust CO_2_-to-C_2_H_4_ conversion.

To evaluate the general applicability of low-coordination transition metal single atoms in ZnS matrices, analogous catalysts incorporating Fe, Co, Ni, and Cu were synthesized and tested (Supplementary Fig. 42). All M_1_–ZnS_v_ (M = Fe, Co, Ni, Cu) variants exhibited substantially higher C_2_H_4_ yields than both pristine ZnS and their saturated-coordination M_1_–ZnS counterparts, indicating a certain enhancement in photocatalytic performance induced by low-coordination metal centers (Supplementary Fig. 43).

To elucidate the electronic origin of this enhancement, projected density of states analyses were conducted for the 5σ and 2π orbitals of adsorbed CO and the metal 3d orbitals (3dz², 3dxz, 3dyz) of the M_1_–ZnS_v_ systems (Supplementary Fig. 44). Among all the systems, Mn_1_–ZnS_v_ exhibits the highest d-band center at −1.85 eV, closer to the Fermi level. This upshift results in decreased population of antibonding d–5σ and bonding d–2*π** states, thereby enhancing *CO adsorption.

Furthermore, integrated crystal orbital Hamilton population values for M–*CO bonds exhibited a clear linear decrease from Cu_1_–ZnS_v_ to Mn_1_–ZnS_v_ (Supplementary Fig. 45), further validating the stronger *CO binding affinity of Mn sites. In situ DRIFTS spectra (Supplementary Fig. 46) and free energy of reaction pathway (Supplementary Fig. 47) presented that Mn_1_–ZnS_v_ exhibits significantly stronger *CO_L_ absorption peaks and comparatively favorable energetics for *CO protonation and coupling than that of Cu_1_–ZnS_v_, which is consistent with the results from the catalytic performance. These findings conclusively demonstrate that low-coordination transition metal single atoms, particularly Mn, enhance *CO stabilization and facilitate the critical C–C coupling step, thereby driving selective C_2_ product formation in CO_2_ photoreduction.

Discussion

In summary, we developed a Mn_1_–ZnS_v_ photocatalyst with a low-coordination Mn SA active site for efficiently selective CO_2_ photoreduction to C_2_H_4_. The detailed studies revealed that low-coordination Mn single-atom active sites break the geometric symmetry of traditional SACs, resulting in anisotropic charge redistribution upon *CO adsorption. This not only enhances the adsorption of *CO intermediates, but also provides new sites for further C–C coupling. Both DFT calculations and experimental evidences confirmed the obtained Mn_1_–ZnS_v_ enhanced *CO adsorption and promoted the coupling between *CO and *CHO intermediates, ultimately driving the formation of C_2_H_4_. As a result, the optimal Mn_1_–ZnS_v_ photocatalyst achieved a C_2_H_4_ selectivity of 99.1% with a formation rate of 76.6 μmol g^−1^ h^−1^, and demonstrated long-term stability over 200 h. We anticipate that this strategy for tailoring low-coordination single-atom active sites will provide new insights into the rational design of highly efficient and selective photocatalysts for converting carbon dioxide into multi-carbon products.

Methods

Chemical reagents

Zinc acetate (Zn(OAc)2), Manganese acetate (Mn(OAc)2), Copper acetate (Cu(OAc)2), Iron acetate (Fe(OAc)3), Nickel acetate (Ni(OAc)2), Cobalt acetate (Co(OAc)2), thiourea, ethylene glycol were purchased from Aladdin Reagent Co., Ltd., Shanghai. All chemicals are of analytical grade and used without further purification. The deionized water was supplied by a Millipore system (Outlet water resistivity > 18 MΩ cm).

Sample preparation

To synthesize ZnS catalyst, typically 6 mmol of Zn(OAc)2 and 12 mmol of thiourea were added to 100 mL of ethylene glycol. After continuous agitation for 30 min, the mixture was moved to a 250 ml three-necked flask with round bottom, and placed in a microwave reactor with a reflux unit. It was heated from room temperature to 150 °C with a heating rate of 25 °C min^−1^ and maintained for 10 min. After the reaction cooled to room temperature naturally, the resulting powder was washed with deionized water and ethanol for three times and dried under vacuum at 80 °C for 6 h. The obtained samples were named as ZnS.

To synthesize Mn_1_–ZnS catalyst, 6 mmol of Zn(OAc)2 and 12 mmol of thiourea were added to 100 mL of ethylene glycol. After continuous agitation for 30 min, the different mass ratios of Mn(OAc)2 (0.2%, 0.6%, 1%) was added to the suspension and continued stirring for 10 min. The mixture was moved to a 250 ml three-necked flask with round bottom, and placed in a microwave reactor with a reflux unit. It was heated from room temperature to 150 °C with a heating rate of 25 °C min^−1^ and maintained in Ar flow for 10 min. After the reaction cooled to room temperature naturally, the resulting powder was washed with deionized water and ethanol for three times and dried under vacuum at 80 °C for 6 h. The obtained samples were named as Mn_1_–ZnS-0.2, Mn_1_–ZnS-0.6 and Mn_1_–ZnS-1.0, respectively.

To synthesize Mn_1_–ZnS_v_, 100 mg Mn_1_–ZnS was immersed in H_2_O_2_ solution (0.05 mol L^−1^) under the protection of Ar in microwave reaction treated for 5, 10, and 15 s, respectively, all the obtained samples were carefully washed and dried using the method described above. The resulting samples were labeled as Mn_1_–ZnS_v_ with different S-vacancy concentrations.

The preparation process for ZnS_v_ is the same as that of Mn_1_–ZnS_v_, except that the precursor Mn salt was not added. The synthesis steps for Fe_1_–ZnS_v_, Co_1_–ZnS_v_, Ni_1_–ZnS_v_, and Cu_1_–ZnS_v_ are the same as Mn_1_–ZnS_v_, except that the precursor salt was replaced with the corresponding acetate. Unless otherwise specified, the precursor transition metal salt loading for the ZnS sample was 0.6%.

Characterization

The phase structure of the photocatalysts was performed on a powder X-ray diffractometer (XRD, Shimadzu XRD 6000). The morphology was observed via field emission scanning electron microscopy (FEI Quanta 200 F), transmission electron microscopy (TEM, JEOL JEM 2100). STEM-ADF images and EDS mapping were obtained by Hitachi HF5000, working at an accelerating voltage of 200 kV. To interpret the light absorption characteristics of the catalysts, UV–Vis DRS were detected on an UV–vis spectrophotometer (Shimadzu UV-2600) over a range of 200–800 nm. To study on the chemical bonds or functional information, in situ DRIFTS were examined using the IR Affinity-1 FTIR spectrometer. The Zn, Cu, and Mn K-edge X-ray adsorption spectra were acquired from 4B9A beamline in Beijing Synchrotron Radiation Facility (BSRF). Isotopic labelling experiments were performed on an Agilent 5977B GC/MS system, using ^13^CO_2_ and D_2_O as reactants under the photocatalytic CO_2_ reduction conditions. The GC-MS spectra reflect the relationship between the m/z of all ion fragments and their relative abundance. Photoluminescence spectra were analyzed on FluoroMax+ spectrophotometer (Hitachi, Japan) with excitation wavelength at 310 nm, and the receiving fluorescence range was from 400 to 700 nm. X-ray photoelectron spectroscopy (XPS) and valence band-XPS tests were conducted on the PerkinElmer PHI-1600 ESCA spectrometer with a Mg Kα X-ray source. All the calibration for the binding energies were based on the C 1 s at 284.8 eV^51^. Carbon dioxide-temperature programmed desorption (CO_2_-TPD) was recorded on a BSD-C200 chemisorption system (BSD Instrument).

In-situ DRIFT spectra measurement

In-situ DRIFT spectra were obtained from a Thermo Scientific Nicolet iS50 FTIR spectrometer. The compressed catalysts were placed in an diffuse reflectance cell (Harrick) equipped with a mixed atmosphere of CO_2_ and H_2_O vapor for in situ experiments. In the pretreatment stage, the system was progressively heated from room temperature to 200 °C under continuous N_2_ purging to remove surface impurities. After cooling to room temperature, the data were collected at a flow rate of 50 mL min^−1^ in a mixed atmosphere. The infrared spectra obtained in the dark were used as background data. Subsequently, time-dependent IR spectra were collected under illumination for 30 min.

In situ CO-adsorbed DRIFT spectroscopy measurement

In situ CO-adsorbed DRIFT spectroscopy measurement was carried out on a Nicolet 6700 instrument equipped with a mercury cadmium telluride detector. 10 mg catalyst was placed in an infrared diffuse reflection high-temperature reaction cell with a ZnSe window (Pike Technologies), pretreated in 10% H_2_/N_2_ flow at 300 °C for 30 min, then cooled to room temperature. This was followed by the introduction of CO flow for 30 min and subsequent evacuation. The CO-adsorbed FT-IR spectra were recorded at an apart of 2 min.

Photoelectrochemical measurements

Photoelectrochemical (PEC) measurements were carried out in the three-electrode system by utilizing an electrochemical workstation (CHI660E). The prepared coated catalysts serve as the working electrode, with an effective surface area of 1 cm^2^. Pt net and Ag/AgCl electrode were employed as the counter electrode and reference electrode, respectively. The Na_2_SO_4_ solution (0.1 mol L^−1^) was acted as electrolyte (pH = 7.02 ± 0.05). In order to prepare the working electrode, 25 mg catalyst and 10 μL Nafion were dispersed in 384 μL deionized water and 96 μL ethanol, then followed by ultrasonication for 25 min. The evenly mixed ink (5 μL) was dropped onto the surface of the glass-carbon electrode, and dried naturally at room temperature. A 50-s light‑on/light‑off cycle was applied to record the transient photocurrent response. The Mott–Schottky measurements were executed with the different frequencies (100, 500, and 1000 Hz), and electrochemical impedance spectra was performed, traversing a frequency range from 0.1 to 100000 Hz. Resistance (R) was measured to be 0.60 ± 0.02 Ω at the open circuit potential.

Tests on photocatalytic CO2 reduction with H2O

The photocatalytic CO_2_ reduction tests were performed in a sealed reaction system by employing the Labsolar 6 A reactor (Beijing Perfectlight Technology Co., Ltd.). 0.2 g of photocatalysts was dispersed in 5 mL of deionized water within a glass dish (diameter is ~4.0 cm). The mixed sample was then placed inside a reactor (volume of ~370 mL) of the photoreaction system, at a distance of 15.0 cm from the light source. The reaction system was subjected to three cycles of vacuuming and refilling with high-purity CO_2_, achieving a final CO_2_ pressure of 80.0 kPa in the reactor. And a cooling circulating water device was utilized to maintain the reaction system a constant temperature at 25 ± 1.5 °C. To simulate the natural photosynthesis, a 300 W xenon lamp (~100 mW cm^−2^) equipped with a 380 nm external filter served as light source. The light intensity of catalyst surface was measured by a ThorLabs PM100D Power with a photodiode sensor, with the average value determined from multiple regions of the watch glass^52^. After starting the light source and the gas chromatograph automatic sampling device, the sample was analyzed every 30 min, and the reaction test was carried out for 4 h. The amount of C_2_H_4_, CH_4_, CO, H_2_, and O_2_ products evolved was analyzed online by a GC-9560 gas chromatograph (HuaAiSePu Company) using the external standard calibration method. For the liquid product, the amount of HCOOH was determined by a proton nuclear magnetic resonance (^1^H NMR) analysis, taking DMSO as an internal standard. No other products were detected above the detection limit of instrument.

The product selectivity for CO_2_ photoreduction to C_2_H_4_ was calculated according to the following equations:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{S}}}}_{{{{{\rm{C}}}}_{2}{{\rm{H}}}}_{4}}\left(\%\right)=\frac{12n\left({C}_{2}{H}_{4}\right)}{12n\left({C}_{2}{H}_{4}\right)+2n\left({CO}\right)+8n\left({{CH}}_{4}\right)+\,2n\left({CHOOH}\right)}\times 100\%$$\end{document}

where n(product) is the molar amount of product generated.

To quantify the efficiency of photocatalytic light-energy conversion, the apparent quantum efficiency (AQE) was measured. The incident photon numbers were determined using monochromatic light sources at specified wavelengths (λ = 400, 420, 450, 500, 520, and 600 nm). AQE was defined as the ratio of consumed photons to incident photons, and according to the following equations:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{AQE}}}}_{{{{\rm{C}}}}_{2}{{{\rm{H}}}}_{4}}= \frac{12\times {{\rm{Number}}}\; {{\rm{of}}}\; {{\rm{reacted}}}\; {{\rm{electrons}}}}{{{\rm{Number}}}\; {{\rm{of}}}\; {{\rm{incident}}}\; {{\rm{photons}}}}\times 100\,\%\\= \,\frac{{12\times n}_{{{C}_{2}H}_{4}}\times {N}_{A}}{S\times I\times t\times \frac{\lambda }{h\times c}}\times 100\,\%\,$$\end{document}

where S is the irradiated area (cm^2^), I is the irradiation light intensity (W cm^−2^), t is the irradiation time (s), λ is the equivalent wavelength (m), h is Planck’s constant (6.626 × 10^−34^ J s^−1^), c is the speed of light (2.998 × 10^8^ m s^−1^) and NA is Avogadro’s number (6.022 × 10^23^ J mol^−1^). nC2H4 is the amount of C_2_H_4_ produced in 4 h.

The incident photon-to-current efficiency is defined as the ratio of the incident monochromatic photons converted to collected electrons and can be calculated by using Eq. (3):

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{IPCE}}}=\frac{\frac{{hc}}{\lambda }\times j}{{Ie}}$$\end{document}

where j is the photocurrent density (A cm^−2^), λ is the equivalent wavelength (m), and I is the irradiation light intensity (W cm^−2^).

Density functional theory (DFT) calculations

DFT calculations were carried out in the framework of the Vienna ab initio Simulation Package (VASP). The exchange and correlation energies were established by the Perdew–Burke–Ernzerhof functional with spin-polarized generalized gradient approximation^53^. The core electrons were described by the projected augmented wave pseudopotentials. For geometry optimization, a cutoff energy of 450 eV was used to expand the wave functions, and the Brillouin zone was sampled with 2 × 2 × 1 k-points. All the structures and energy were allowed to relax below 0.05 eV Å^−1^. The DFT-D3 method of Grimme et al. was employed to involve the van der Waals interaction^54^. The calculations of charge density difference were employed to analyze the movement and distribution of the charge. For the elementary reaction barriers, the transition states were determined by the climbing image nudged elastic band method and were confirmed by further frequency calculations showing one and only one imaginary frequency.

According to the XRD results, a 5 × 5 slab model composed of three layers along the (111) direction was constructed with the space group F-43m model of ZnS carrier. A 20.0 Å vacuum region between the slabs was constructed to avoid the interlayer interaction. The doped ZnS with other transition metal elements was simulated by replacing one of the Zn atoms. The S-vacancy was simulated by removing S atoms off, and the theoretical value of S_v_ is 1.33%. Considering the difference in the performance of as-prepared catalysts, ICP quantified vacancy concentrations, specific configuration of sulfur vacancies was adopted to model the surface. The atomic coordinates of the DFT models are provided as a Supplementary Data 1 file. The formation energy of sulfur vacancy \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(\Delta {{{{\rm{E}}}}_{{{\rm{V}}}}}_{{{\rm{S}}}}\right)$$\end{document} was calculated as follows:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {{{{\rm{E}}}}_{{{\rm{V}}}}}_{{{\rm{S}}}}={{{\rm{E}}}}_{{{\rm{slab}}}-{{{\rm{V}}}}_{{{\rm{S}}}}}-{{{\rm{E}}}}_{{{\rm{slab}}}}-{{\rm{E}}}_{{\rm{S}}}$$\end{document}

in which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mbox{E}}}_{{\mbox{slab}}-{{\mbox{V}}}_{{\mbox{S}}}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mbox{E}}}_{{\mbox{slab}}}$$\end{document} represent the energy of the slab after and before V_s_. The adsorption energy E_ads_ was calculated by a standard formula:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{E}}}_{{\rm{ads}}}}={{{\rm{E}}}_{{{\rm{catalyst}}}{\mbox{-}}}} * {-} {{{\rm{E}}}_{{\rm{catalyst}}}}{-}{{\rm{E}}} * $$\end{document}

in which Ecatalyst - * and Ecatalyst is the total energy of the slab with and without intermediates.

The Gibbs free energy change (ΔG) for the reaction was calculated by

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {{\rm{G}}}=\Delta {{\rm{E}}}+\Delta {{{\rm{E}}}}_{{{\rm{ZPE}}}}-{{\rm{T}}}\triangle {{\rm{S}}}$$\end{document}

in which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\triangle E$$\end{document} is reaction energy obtained from DFT calculations, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\triangle {{\mbox{E}}}_{{\mbox{ZPE}}}$$\end{document} represent the change of the zero-point energy, T∆S is the entropic contribution (T was set to be 300 K), which were obtained from the vibrational frequency calculations through the VASPKIT code. The free energy of the proton–electron pair is equal to half the free energy of the hydrogen molecule according to the computational hydrogen electrode method^55^.

Ab initio molecular dynamics simulations are carried out by CP2K package and the QUICKSTEP module with fully explicit solvent water molecules^56^. Godecker-Teter-Hutter (GTH) pseudopotentials were adopted to model the core electrons, and double-ζ valence single polarization (DZVP)-molecularly optimized (MOLOPT)-short-ranged-GTH basis set was employed to obtain optimized structure^57,58^. Valence electrons were expanded in an orthonormal plane-wave basis using a cutoff energy of 400 Ry, along with Grimme D3 dispersion correction^59^. The self-consistent field convergence criterion is set to 10^−5^ Ry for the total energies, and the simulation temperature (300 K) was controlled through a Nosé-Hoover thermostats with time step of 1.0 fs. DFT offers valuable insights for screening coordination structures and exploring reaction pathways, yet it remains inherent limitations in simulating practical catalytic conditions (e.g., actual illumination, solvation effects). Thus, the computationally identified coordination structures should serve as qualitative guides for catalyst design and mechanistic investigation.

Supplementary information

Supplementary Infomation Description of Additional Supplementary File Supplementary Data 1 Transparent Peer Review file

Source data

Source Data