Well-defined diatomic catalysis for photosynthesis of C₂H₄ from CO₂

Abstract

Owing to the specific electronic-redistribution and spatial proximity, diatomic catalysts (DACs) have been identified as principal interest for efficient photoconversion of CO₂ into C₂H₄. However, the predominant bottom-up strategy for DACs synthesis has critically constrained the development of highly ordered DACs due to the random distribution of heteronuclear atoms, which hinders the optimization of catalytic performance and the exploration of actual reaction mechanism. Here, an up-bottom ion-cutting architecture is proposed to fabricate the well-defined DACs, and the superior spatial proximity of CuAu diatomics (DAs) decorated TiO₂ (CuAu-DAs-TiO₂) is successfully constructed due to the compact heteroatomic spacing (2-3 Å). Owing to the profoundly low C-C coupling energy barrier of CuAu-DAs-TiO₂, a considerable C₂H₄ production with superior sustainability is achieved. Our discovery inspires a novel up-bottom strategy for the fabrication of well-defined DACs to motivate optimization of catalytic performance and distinct deduction of heteroatom synergistically catalytic mechanism.

Introduction

Photocatalytic carbon dioxide (CO₂) reduction, a mimicking natural photosynthesis, is identified as an ideal technology to reduce the CO₂ level by the conversion of CO₂ into fuels or industrial feedstocks with the utilization of solar energy^1,2,3. Among all the photoreduction CO₂ products, ethylene (C₂H₄) is considered as high-value species due to their high energy densities and commercial prices in chemical industry^{4,5,6,7,8,9,10,11,12,13}. Bimetallic solid-photocatalysts, consisted of ordered stagger of heteronuclear metal atoms, can strikingly reduce C-C coupling energy barrier via weakening dipole-dipole repulsion of the neighboring adsorbed C-based intermediates, which has been regarded as the dominator in photosynthesis of C₂H₄^{14,15,16,17,18,19,20,21}. Nevertheless, the compact atom stacking during the synthesis of bimetallic solid catalysts inevitably contributes to the decrease of active sites^{22,23,24,25,26,27}, critically restricting the efficient C₂H₄ generation. Therefore, its deep desirability to accurately optimize bimetallic solid catalysts at the atomic scale to break the bottleneck of inherent scaling-catalytic relationship in photosynthesis of C₂H₄.

Single-atom catalysts (SACs), a well-defined mononuclear metal sites, which have attracted huge attention for their potential to overcome the disadvantages of previously developed solid catalysts²⁷. Conspicuously, heteronuclear DACs can not only maintain the maximized atom utilization but also modulate the reaction kinetics and even the reaction pathways by involving two metal atoms with cooperative modification of their steric and electronic properties^{26,27,28,29,30,31}. So far, all reported heteronuclear DACs fabrication have concentrated on bottom-up synthetic strategies, such as organometallic compounds pyrolysis^{26,32,33,34,35}, metal complexation^36,37,38, metal coprecipitation^{39,40,41,42,43}, in situ photoreduction^44,45, and physical desorption^46,47. Because of the repellency between the bare metal atoms under such prevailing bottom-up synthesis methods, disordered heteronuclear sites distribution have been an insurmountable barrier, which directly restrained the optimization of catalytic performance and distinct deduction of catalytic mechanisms⁴⁸. Consequently, it is of high priority to exploit novel strategy for well-defined heteronuclear DACs fabrication to establish a high-efficiency photosynthesis of C₂H₄ system.

Here, an up-bottom ion-cutting engineering for the synthesis of atomic-level catalysts was initially proposed. Depending on the vectored etching of Cu in a CuAu alloy (isolated Au atoms in the Cu lattice), the CuAu-DAs with subnanometer heteroatomic spacing supported by commercial TiO₂ (CuAu-DAs-TiO₂) were successfully fabricated. The C₂H₄ production of CuAu-DAs-TiO₂ proceeded at a remarkable rate of 568.8 μmol·g⁻¹·h⁻¹ without any sacrificial agent, which is superior to recent reported works in photoconversion CO₂ and H₂O into C₂H₄ (Supplementary Table 1), and no apparent catalyst deactivation was observed during the 120-h photocatalytic stability test. In such CuAu-DAs structures, Cu single atoms (Cu-SAs) are mainly responsible for the high-efficiency *CO generation rather than C-C coupling centers^{16,17,49,50,51,52}, while Au single atoms (Au-SAs) serve as *CO coupling centers to rapidly consume the *CO arising from Cu-SAs according to photocatalytic CO₂ reduction, which synergistically promotes the high efficiency and sustainability of photoconversion of CO₂ into C₂H₄⁵³.

Results

Catalyst synthesis and structural characterization

Compared to the widespread bottom-up synthetic strategies of SACs and DACs^{32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,50,51,52,53,54,55,56,57,58}, the up-bottom ion-cutting architecture controlled by the vectored etching of specific element contents in the alloy may provide novel insight for the adjustable design of SACs and DACs. Figure 1a illustrates the ion-cutting architecture fabrication of Cu nanoclusters (Cu-NCs) decorated Au-SAs (CuAu-NCSAs), CuAu-DAs, and Au-SAs/Cu-SAs supported by TiO₂ via adjusting the vectored etching time of the Cu₅Au₁ alloy. Transmission electron microscopy (TEM) and aberration-correction high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) were carried out to acquire a more spatially resolved structural configuration during the different catalyst synthesis processes. In Supplementary Figs. 1 and 2, the commercial TiO₂ displays the irregular nanoparticle morphology with the ca. 13.6 nm of particle size, and the EDS mapping analysis also verifies the uniform Ti and O elements distribution. Compared to commercial TiO₂, the regular and larger size of Cu₅Au₁ alloy was observed in Supplementary Fig. 3a, and the compactly connected heterointerface was observed between Cu₅Au₁ and TiO₂ (Supplementary Fig. 3b). In addition, the overlapping Cu and Au elements distribution were also exhibited in EDS mapping images, indicating the successful construction of CuAu alloy (Supplementary Fig. 3c–g). As shown in Fig. 1b and Supplementary Fig. 4, the ordered stagger of Cu/Au atoms stacked Cu₅Au₁ alloy nanoparticles are successfully loaded on bulk commercial TiO₂, and the isolated Au single atom in Cu lattice provided the precondition for CuAu-DAs formation. Moreover, the 4.43 and 1.71 Å of Au-Au and Cu-Au atomic distance further verify the isolated Au SAs surrounded by Cu atoms rather than the Au atoms in Cu₅Au₁ alloy (Supplementary Fig. 4a–c). After 1 h of vectored etching of Cu atoms (Fig. 1c and Supplementary Fig. 5a–d), slightly destroyed lattice and few Cu vacancies in the Cu₅Au₁ alloy were observed due to the destruction of the ordered arrangement of Cu-Au atoms. The ablative Cu₅Au₁ alloy and few CuAu-DAs were obtained after 3 h of etching (Fig. 1d and Supplementary Fig. 6a, b), indicating that the constant loss of Cu atoms could promote the collapse of the Cu₅Au₁ alloy framework and the redistribution of adjacent Cu-Au atoms on the TiO₂ surface. Moreover, the distinguished ordered lattice of TiO₂ and disordered lattice of Cu₅Au₁ were simultaneously observed in HRTEM and the corresponding pseudo-color images, which further implied the successfully vectored etching of such Cu₅Au₁ alloy (Supplementary Fig. 6c, d). As shown in Fig. 1e, the CuAu-DAs are uniformly dispersed on the surface of the TiO₂ (211) plane (lattice spacing: 0.165 nm) in E₇-Cu₅Au₁-TiO₂. In addition, the magnified images (Fig. 1f₁-f₄) and acquired AC-HAADF-STEM image intensity profile (Supplementary Fig. 7) clearly verified subnanometer distances (~2–3 Å) between the Cu-SAs (dark) and Au-SAs (bright) according to the different atomic mass of the corresponding elements⁵⁹, suggesting that the twin CuAu-DAs was successfully constructed after 7 h of vectored etching of the Cu₅Au₁ alloy. Moreover, when the vectored etching time was prolonged to 11 h (Fig. 1g and Supplementary Fig. 8), a few Cu-SAs and dominated Au-SAs were observed in the AC-HAADF-STEM image, indicating that the dynamic vectored etching process promoted the further dissociation of CuAu-DAs structure due to the consistent decrease of Cu atoms. In Supplementary Fig. 9, Cu-SAs modified TiO₂ was also constructed by 7 h of vectored etching of Cu-TiO₂, suggesting the huge potential of such up-bottom ion-cutting technology for atomic-level catalysts design.

Fig. 1: Morphological characterization of the CuAu-based TiO₂ composites.

a Dynamic schematic illustration from isolated Au-SAs in the Cu lattice to CuAu-DAs during the vectored etching process (green sphere: Cu⁰; orange sphere: Au⁰; purple sphere: Fe³⁺). AC-HAADF-STEM images of Cu₅Au₁-TiO₂ (b), E₁-Cu₅Au₁-TiO₂ (c), E₃-Cu₅Au₁-TiO₂ (d), E₇-Cu₅Au₁-TiO₂ (e), and E₁₁-Cu₅Au₁-TiO₂ (g). f Magnified images of the corresponding areas in image (e). Scale bars of f₁-f₄: 1 nm.

To further investigate the atomic-scale configuration of Cu and Au in E₇-Cu₅Au₁-TiO₂, Cu K-edge and Au L-edge X-ray absorption near-edge structure (XANES) spectroscopy was performed (Fig. 2a, b). Figure 2a illustrates the Cu K-edge XANES spectra for E₇-Cu₅Au₁-TiO₂ with Cu foil benchmarks as reference, and the absorption edge position of E₇-Cu₅Au₁-TiO₂ is more positive than that of Cu foil, indicating that partial Cu could directly connect with the lattice O of TiO₂. In Fig. 2b, the white line peak of E₇-Cu₅Au₁-TiO₂ exhibits Au characteristic features, which is similar to those of the reference Au foil, indicating the presence of Au⁰. As shown in Fig. 2c, the local coordination around the Cu-O shell and Cu-M shell (Cu-Cu and Cu-Au) was determined by the k³-weighted Fourier transform of the extended X-ray absorption fine structure (FT-EXAFS) spectrum. A predominant peak at ~1.49 Å of E₇-Cu₅Au₁-TiO₂ is assigned to the Cu-O coordination. Furthermore, the other obvious characteristic peak at ~2.53 Å is observed in the E₇-Cu₅Au₁-TiO₂ spectrum but not in the Cu foil spectrum, implying the possible formation of Cu-Au coordination. The local peak at ~1.52 Å in the Au L-edge spectrum of CuAu-DAs-TiO₂ (Fig. 2d) is close to that of Au-O, indicating the formation of Au-O coordination. An apparent path at ~2.49 Å is observed in the Au L-edge spectrum of E₇-Cu₅Au₁-TiO₂ (Fig. 2d), which is close to the values of Cu-Au and Au-Au (Fig. 2c). The Cu foil is used to calculate the standard amplitude reduction factor (S₀² = 0.845, Supplementary Table 2), and the Cu K-edge EXAFS analysis of E₇-Cu₅Au₁-TiO₂ in R spaces is exhibited in Supplementary Fig. 10. The EXAFS spectrum of E₇-Cu₅Au₁-TiO₂ is analyzed by using two backscattering paths (Cu-O and Cu-Au). The best-fitting results exhibited that the coordination number of the O and Au in the first coordination sphere of E₇-Cu₅Au₁-TiO₂ is fitted to be ≈3.3 and ≈1.2 at distances of 1.93 and 2.91 Å, respectively, implying the Cu-SAs is merely adjacent with single Au atom. Therefore, there is no doubt that the existence of CuAu-DAs structure under such vectored etching process, while it is still recognized indeed small existence of pure Au phase due to the shortage of such solvothermal method for the fabrication of highly ordered CuAu alloy. Moreover, the concurrent detection of Cu-M and Au-M distances is in the range of the observed distribution of twin dual atoms in atomic-resolution STEM imaging (Supplementary Fig. 7), further demonstrating that the Cu-Au bond originated from the initial CuAu alloy rather than from Cu-SAs and Au-SAs rearrangement during the etching process. The Cu K-edge and Au L-edge EXAFS oscillations are also analyzed by the wavelet transform (WT) method to further confirm the presence of the Cu-Au path. No WT maxima at 7.5 Å⁻¹ (Cu-Cu bond) and 8.3 Å⁻¹ (Au-Au bond) were observed in the spectra of E₇-Cu₅Au₁-TiO₂ (Fig. 2e, f), indicating the SAs-structure configuration of most Cu and Au sites. The WT maxima at 5.7 Å⁻¹ and 13 Å⁻¹ in the spectra of E₇-Cu₅Au₁-TiO₂ correspond to the Cu-O and Cu-Au bonds, respectively (Fig. 2e). Similarly, the WT maxima at 4.9 Å⁻¹ and 7.7 Å⁻¹ are attributed to the Au-O and Au-Cu/Au-Au bonds (Fig. 2f), respectively, consistent with the FT-EXAFS results (Fig. 2c, d). Consequently, the adjacent Cu and Au coexisted as twin diatomic centers and connected with the lattice O of TiO₂ to form O-Cu-Au-O.

Fig. 2: Structural characterization of the CuAu-DAs.

XANES analysis of E₇-Cu₅Au₁-TiO₂ and reference samples at the Cu K-edge (a) and Au L-edge (b). Corresponding k³-weighted FT-EXAFS spectra in the R space for E₇-Cu₅Au₁-TiO₂ and references at the Cu K-edge (c) and Au L-edge (d). e Cu K-edge WT-EXAFS spectra of Cu foil and E₇-Cu₅Au₁-TiO₂. f Au L-edge WT-EXAFS spectra of Au foil and E₇-Cu₅Au₁-TiO₂.

As shown in Supplementary Fig. 11a, the Cu 2p X-ray photoelectron spectroscopy (XPS) binding energy peaks at approximately 932.23 and 952.03 eV are attributed to Cu⁰/Cu⁺⁶⁰. The Auger electron spectrum (Cu LMM) was obtained to further distinguish between Cu⁰ and Cu⁺ in Cu-TiO₂ and Cu₅Au₁-TiO₂. As shown in Supplementary Fig. 12, the characteristic kinetic energy peaks at 918.06 and 921.82 eV⁶¹ have corresponded to the electron state of Cu⁰ in Cu-TiO₂ and Cu₅Au₁-TiO₂. The presence of weak peaks located at 934.70 and 954.40 eV confirms the presence of trace Cu²⁺ in Cu-TiO₂ and Cu₅Au₁-TiO₂. Notably, no Cu⁰ or Cu²⁺ were detected in E₇-Cu₅Au₁-TiO₂ because Cu was below the detection limits after the cooperative Fe³⁺ and H⁺ etching reaction of Cu⁰ and Cu²⁺, consistent with the rare residual Cu-SAs observed in AC-HAADF-STEM (Fig. 1e). In Supplementary Fig. 11b, the two Au 4f peaks of Au-TiO₂ located at 82.99 (Au 4f_7/2) and 86.70 eV (Au 4f_5/2) are attributed to zero valence Au. Compared to those of Au-TiO₂, the Au 4f_7/2 and Au 4f_5/2 characteristic peaks of Au₅Cu₁-TiO₂ are positively shifted to 83.18 and 86.90 eV, indicating the impeded electron transfer from TiO₂ to Au due to the surrounding Cu barrier. After the etching process, the Au 4f binding energy peaks of E₇-Cu₅Au₁-TiO₂ are more negative than those of Au-TiO₂ and Cu₅Au₁-TiO₂, illustrating that more TiO₂ local electrons are transferred to Au due to the decreased amounts of Cu, further suggesting the direct concatenation between Au and Cu. The most negative Au 4f binding energies of E₇-Cu₅Au₁-TiO₂ suggest the high-concentration electron density of Au, which inevitably benefits the multiple electron reaction conduction on the Au sites. Consequently, large amounts of photogenerated electrons may be transferred from TiO₂ to the CuAu-DAs due to the low Femi level of metallic Cu and Au^62,63, which may cause both Cu-SAs and Au-SAs to be the main centers for CO₂ adsorption-activation and C-C coupling. Significantly, the Ti 2p spectra of E₇-Cu₅Au₁-TiO₂ exhibits peaks of Ti 2p_3/2 and Ti 2p_1/2 at more negative binding energies compared to those of Cu-TiO₂, Au-TiO₂, and Cu₅Au₁-TiO₂ (Supplementary Fig. 11c), further indicating that the Cu-SAs and Au-SAs are more beneficial for photogenerated electron migration from TiO₂ to the CuAu-DAs under irradiation. All the O 1s spectra of the as-prepared samples show two typical peaks, which are assigned to the O-Ti bond of TiO₂ and the O–H bond of surface adsorbed OH groups⁶² (Supplementary Fig. 11d). X-ray diffraction (XRD) was conducted to further analyze the crystal surface information of these TiO₂-based samples. Compared to anatase TiO₂ (JCPDS No. 71-1166), the as-prepared Au-TiO₂ and Cu₅Au₁-TiO₂ composites gives rise to (200), (220), and (311) characteristic peaks of Au (JCPDS 65-2870), while no characteristic peak of Cu was observed in Cu₅Au₁-TiO₂ and Cu-TiO₂ (Supplementary Fig. 13), indicating the rapid growth of pure Au phase and the restrained ordered growth of Cu in the CuAu structure due to the different standard electrode potential (E⁰) of Au ([AuCl₄]^-/Au⁰, E⁰ = +0.93 V) and Cu (Cu²⁺/Cu⁰, E⁰ = +0.34 V)⁶⁴. Notably, the HRTEM and EDS mapping images exhibited the small-sized Cu nanoparticle (ca. 7.9 nm) decorated TiO₂ (Supplementary Fig. 14a–d), and the EDS mapping also showed the uniformly dispersed Cu elements (Supplementary Figs. 14 and 15), which indicated that Cu existed as the small-sized scale rather than extended growth into large-sized structure (Supplementary Fig. 14e–h). There has been reported that the low ordering degree of CuAu alloy merely display the Au characteristic peaks due to the dominated Miller indexes of Au phase⁶⁵, which is corresponded to the observation of weak Au characteristic peaks in our work. Although the AC-HAADF-STEM and EDS mapping display the Cu atoms surrounded isolated Au single atoms (Supplementary Fig. 4d–g), the predominated 0.233 nm of lattice spacing was observed in the AC-HAADF-STEM of Cu₅Au₁-TiO₂, which is ascribed to the (111) crystal facet of Au rather than Cu and CuAu characteristic lattice planes. Therefore, the apparent characteristic crystal facet of Au that displayed in AC-HAADF-STEM images and XRD reflections could be ascribed to the dominated Miller indexes of Au phase (Supplementary Figs. 4d and 13). According to the dynamic XRD patterns of Cu₅Au₁-TiO₂ after different vectored etching time (Supplementary Fig. 16a, b), no obvious enhanced Au characteristic peak intensity of E_t-Cu₅Au₁-TiO₂ was observed compared to that of Cu₅Au₁-TiO₂ after the vectored etching process, suggesting that the vectored etching processes of Cu could not promote isolated Au-SAs rearrangement to form Au lattice plane. Inductively coupled plasma-atomic emission spectroscopy (ICP‒AES) was conducted to evaluate the dynamic Cu/Au molar ratios during the etching process. In Supplementary Fig. 17, with increasing etching time, the Cu/Au molar ratio value of E_t-Cu₅Au₁-TiO₂ decreases deeply, while the molar ratio values of H_t-Cu₅Au₁-TiO₂ decrease slightly, indicating the presence of large amounts of Cu⁰ rather than Cu⁺/Cu²⁺ and the successful construction of an adjustable CuAu atomic-level nanostructure. Owing to the different redox capacity-related standard electrode potential between Fe³⁺ (Fe³⁺ + e^- → Fe²⁺, E⁰ = +0.77 V) and Cu⁰ (Cu²⁺ + 2e^- → Cu⁰, E⁰ = +0.34 V), the Fe³⁺ could spontaneously be reduced by Cu⁰ (Fe³⁺ + Cu⁰ → Fe²⁺ + Cu²⁺)⁶⁴. However, due to the insufficient oxidized capacity of Fe³⁺ for Au⁰ (Au³⁺ + 3e^- → Au⁰, E⁰ = +1.52 V) oxidation⁶⁴, the contents of Au keep constant during the whole etching process (Supplementary Table 3), which seriously restrains the dissolution of Au⁰ and promote the redistribution of Au atoms on TiO₂. Localized surface plasmon resonance (LSPR) is often related to the metal shape and the dielectric constant of the surrounding medium⁶⁶. Therefore, the connection between atomic interface engineering and the light absorption of the as-prepared samples was also checked by UV‒vis‒NIR diffuse reflection spectroscopy (DRS). As shown in Supplementary Fig. 18a, an obvious redshift is observed over metal-decorated TiO₂ compared to pure TiO₂, implying that Cu and Au could effectively enhance visible light absorption and produce more photogenerated carriers. In Supplementary Fig. 19a, E_t-Cu₅Au₁-TiO₂ shows strikingly increased Au (~520 nm) and Cu (~730 nm) LSPR response peaks compared to Cu₅Au₁-TiO₂, indicating that the regularly variational nanoscale of Au and Cu can benefit the formation of higher electron density centers upon irradiation. However, with the persistent etching of Cu, the LSPR response of Cu almost disappeared due to the consistently decreased content of Cu. Moreover, H_t-Cu₅Au₁-TiO₂ show much weaker Au LSPR response than E_t-Cu₅Au₁-TiO₂ (Supplementary Fig. 19b, c), and the negligible Cu LSPR response intensity of H_t-Cu₅Au₁-TiO₂ confirms that the H⁺ etching process eliminated only a small amount of surficial oxidized Cu to a certain extent, which implied that E_t-Cu₅Au₁-TiO₂ could induce a higher photogenerated carrier density on CuAu sites due to the stronger Cu and Au LSPR response intensity.

To reveal the charge carrier dynamics on CuAu-DAs modified TiO₂, photoluminescence (PL) spectroscopy and time-resolved photoluminescence (TRPL) spectroscopy were carried out. As shown in Supplementary Fig. 20a, the intensity of the emission peaks of Cu-TiO₂, Cu₅Au₁-TiO₂, and E₇-Cu₅Au₁-TiO₂ at ~425 nm decrease considerably compared to that of pure TiO₂, indicating that the decoration of metallic cocatalysts on TiO₂ can improve the charge separation and transfer. Notably, the spectrum of E₇-Cu₅Au₁-TiO₂ displays a weaker emission peak than that of Cu₅Au₁-TiO₂, indicating the lower photogenerated charge recombination of E₇-Cu₅Au₁-TiO₂ due to the feedthrough charge transfer channel between TiO₂ and CuAu-DAs. It is reported that the contributions of τ₁ and τ₂ are more related to charge transfer, and the PL decay is more dominated by τ₃^67,68,69. In Supplementary Fig. 20b and Table 4, the CuAu-DAs modified TiO₂ exhibits shorter τ₁, τ₂, and τ₃ compared to TiO₂, illustrating the compact interaction and suppressed charge recombination between CuAu-DAs and TiO₂. Moreover, the shortest τ₃ lifetime of CuAu-DAs modified TiO₂ represents the fastest decay in CuAu-DAs-TiO₂, which is ascribed to the fact that the direct connection between CuAu-DAs and TiO₂ is more beneficial for convenient photogenerated charge transfer from TiO₂ to CuAu-DAs rather than recombination in the Cu and Cu₅Au₁ bulk. Transient photocurrent and electrochemical impedance spectroscopy (EIS) measurements were also conducted to further reveal the efficiency of photogenerated charge separation and transportation of the as-synthesized samples. In Supplementary Fig. 20c, d, the highest photocurrent density and smallest arc radius of E₇-Cu₅Au₁-TiO₂ suggest that the CuAu-DAs exhibits much higher charge separation efficiency and faster interfacial charge transportation than TiO₂, Cu-TiO₂, and Cu₅Au₁-TiO₂. Therefore, the superior photoreduction CO₂ performance of CuAu-DAs-TiO₂ could also be attributed to the excellent photoabsorption and high-efficiency charge separation.

Photocatalytic performance toward CO₂ photoreduction

A suitable band edge position of TiO₂ is a prerequisite for the successful photoconversion of CO₂ gas and H₂O vapor into CO, CH₄, C₂H₄, C₂H₆, and O₂ upon 320–780 nm irradiation (Supplementary Fig. 18e). To uncover the influence of different CuAu atomic interfaces on the CO₂ photoreduction reaction, photocatalytic CO₂ reduction tests were carried out on different CuAu nanostructure-decorated TiO₂ samples. At the same loading amounts, Cu-TiO₂ and Au-TiO₂ achieved much higher CO and CH₄ production rates than pure TiO₂ (Fig. 3a and Supplementary Fig. 21a, b) due to the increased number of active sites and the improved charge transfer efficiency⁷⁰. As shown in Fig. 3a and Supplementary Fig. 21a, b, Cu₅Au₁-TiO₂ performs a higher CH₄ production rate (20.4 µmol·g⁻¹·h⁻¹) than Cu-TiO₂ and Au-TiO₂, which is attributed to the fact that the Au matrix often functions as an electron sink, while the adjacent surficial Cu serves as CO₂ activation centers and proton transfer stations^70,71, and the intermediate further combines with the surrounding abundant electrons originating from the Au matrix to promote CH₄ production through the carbene pathway^62,72,73. The actual active sites merely exist on the exposed alloy surface rather than in the interior of the bulk, resulting in the catalytic reaction typically occurring on the unsaturated active sites of the CuAu alloy surface. Therefore, photocatalytic CO₂ reduction tests of the etched samples were also performed to further investigate the influence of CuAu interfacial engineering on the photoreduction of CO₂. In Supplementary Fig. 22, a possible schematic illustration is shown to describe the dynamic change in the CuAu structural configuration during the constant vectored etching process and define a possible correlation with the photocatalytic performance. With increasing vectored etching time, a peculiar dual volcanic relationship of the hydrocarbon production rate is observed in the E_t-Cu₅Au₁-TiO₂ series (Fig. 3b). After 1 h of vectored etching of Cu₅Au₁-TiO₂, E₁-Cu₅Au₁-TiO₂ reaches the highest CH₄ production (68.2 µmol·g⁻¹·h⁻¹) and electron-based selectivity (84.7%) among the series of E_t-Cu₅Au₁-TiO₂ (Fig. 3b and Supplementary Fig. 21c, d), which is ascribed to the increased number of unsaturated Au and Cu active sites (Supplementary Fig. 22). Nevertheless, the CH₄ production rate continuously decrease due to the constant decrease in the number of Cu atom active sites after a longer vectored etching time. Notably, when the etching time is prolonged to 7 h (Fig. 3b, c and Supplementary Fig. 21c, d), the C₂H₄ and C₂H₆ production rates of E₇-Cu₅Au₁-TiO₂ reach 71.6 and 8.5 µmol·g⁻¹·h⁻¹, respectively, and the electron-based selectivity reached 68.3% (C₂H₄) and 9.4% (C₂H₆). The dramatically increased C₂H₄ production rate of E₇-Cu₅Au₁-TiO₂ is 305 and 73 times higher than that of TiO₂ and Cu₅Au₁-TiO₂, respectively, which is ascribed to the superior synergistically enhanced C-C coupling over CuAu-DAs. With the further etching of Cu-SAs, the C₂H₄ production rate of E₉-Cu₅Au₁-TiO₂ and E₁₁-Cu₅Au₁-TiO₂ constantly decrease due to the constantly decreasing amounts of Cu-SAs, implying that the mere presence of Au-SAs (Fig. 1g) could not satisfy the high-efficiency C-C coupling and C₂H₄ production. Moreover, the optimized amount of E₇-Cu₅Au₁-TiO₂ photocatalysts was verified to be 20 mg with a highest C₂H₄ production rate based on the excellent mass transfer and superior light utilization (Supplementary Fig. 23), and the apparent decrease of C-based products production rate was observed with the addition of superfluous photocatalysts, which could be ascribed to the impeded light transmission⁷⁴. Compared to Au-SAs modified TiO₂, Cu-SAs (Supplementary Fig. 9) modified TiO₂ (E₇-Cu-TiO₂) exhibits dramatically increased CO and CH₄ generation, while negligible C₂ generation is also observed on E₇-Cu-TiO₂, indicating that the Cu-SAs is beneficial for high-efficiency *CO production (Supplementary Fig. 24), consistent with recent research results on Cu-SAs based photocatalytic CO₂ reduction^{75,76,77,78,79}. Therefore, we suspect that only the specific existence of adjacent Cu-SAs and Au-SAs can efficiently and synergistically favor C-C coupling and C₂H₄ production. To further confirm that the CuAu-DAs structure can promote C₂H₄ generation, a photoreduction CO₂ test was also conducted on E_t-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ (Supplementary Fig. 25) constructed by adding additional Cu content to the Cu₅Au₁ alloy. The dual volcanic relationship of the hydrocarbon production rate is still observed on E_t-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ based on the transformation of the CuAu structure. Furthermore, E₁₁-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ exhibits the maximum C₂H₄ production rate after 11 h of vectored etching (Supplementary Fig. 25a), and the CO production also rapidly increase with enhanced C₂H₄ generation, similar to the behavior of E₇-Cu₅Au₁-TiO₂, indicating that the *CO species is a critical intermediate for C-C coupling, as reported^6,14,53. As shown in the ICP‒AES results (Supplementary Fig. 26 and Table 5), the Cu/Au molar ratios of both E₇-Cu₅Au₁-TiO₂ and E₁₁-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ are similar, further confirming that the dominated CuAu DAs-structure dramatically actuate C₂H₄ production. Moreover, photoreduction CO₂ tests were also conducted on H_t-Cu₅Au₁-TiO₂ to further identify the effects of different CuAu structures. Although a superior CO production rate (61.8 µmol·g⁻¹·h⁻¹) is observed on H₁₁-Cu₅Au₁-TiO₂ (Supplementary Fig. 27), the C₂ production is still nonideal, emphasizing that only Au-SAs, rather than low-coordination Au sites, could rapidly convert the high-concentration CO into C₂ products. Furthermore, a superhigh CH₄ (152.6 µmol·g⁻¹·h⁻¹) production rate is observed for H₁₁-Cu₅Au₁-TiO₂, indicating that low-coordination Au sites in the Cu lattice are more beneficial for converting *CO into CH₄ than *CO coupling on Au-SAs. Consequently, in such CuAu-DAs modified TiO₂ photocatalytic systems, the Cu-SAs guarantee rapid *CO generation and high-concentration coverage, and the adjacent Au-SAs further promote migration and coupling of the generated *CO. The source of photoreduced CO₂ products was investigated by using isotope labeling ¹³CO₂ and H₂¹⁸O as the reactant under irradiation, and the products were analyzed by gas chromatography‒mass spectrometry (GC‒MS). The GC‒MS peak sequences of ¹³CO, ¹³CH₄, ¹³C₂H₄, and ¹³C₂H₆ are shown in Supplementary Fig. 28, and the peaks at m/z = 29, m/z = 17, m/z = 30, and m/z = 32 are assigned to ¹³CO, ¹³CH₄, ¹³C₂H₄, and ¹³C₂H₆, indicating that the carbon source of CO and hydrocarbons is indeed derived from the input CO₂ gas. Overall CO₂ photoreduction is divided into two major half reaction steps, the CO₂ reduction and H₂O oxidation^12,80, and the detection of ¹⁶O¹⁸O and ¹⁸O₂ species verifies that O₂ originates from H₂O oxidation in the photocatalytic CO₂ reduction in (Supplementary Fig. 28). Notably, in Supplementary Fig. 29, O₂ evolution related to holes consumption of the TiO₂ based composites are also stoichiometrically approximate to products of photogenerated electrons reduction, which indicates the simultaneous CO₂ reduction and H₂O oxidation behaviors. The CO₂ photoreduction experiment was also taken under no existence of H₂O to further figure out the influence of H₂O species (Supplementary Fig. 30). No C-based product was detected in the absence of H₂O, indicating the significance of photogenerated holes consumption in the overall CO₂ photoreduction. Therefore, both the investigation of H₂O oxidation and CO₂ reduction are crucial and directive for the development of photocatalysis. Furthermore, a negligible amount of CO production was detected on E₇-Cu₅Au₁-TiO₂ and E₇-Cu₅Au₁-Al₂O₃ upon 420 nm irradiation (Supplementary Fig. 31a), indicating the inappreciable CuAu LSPR effect and the dominant TiO₂ electron donator in photocatalytic CO₂ reduction (Supplementary Fig. 31b). Consequently, the same DAs-modified strategy was conducted on the widespread reported photocatalysts of carbon nitride (C₃N₄) to further verify the universality of such optimized strategy, and the vectored etching Cu₅Au₁ modified C₃N₄ and TiO₂ also exhibit the efficient C₂H₄ production, which sufficiently prove the universality of such CuAu-DAs modification for the optimization of C-C coupling reaction (Supplementary Figs. 32–35).

Fig. 3: Photocatalytic performance of CuAu-based TiO₂ composites.

CH₄, C₂H₄, C₂H₆, CO, and H₂ production rates of the as-prepared photocatalysts (a) and E_t-Cu₅Au₁-TiO₂ (b). c Electron-based selectivity of photocatalytic CO₂ conversion over E_t-Cu₅Au₁-TiO₂. d Long-term photocatalytic stability test of E₇-Cu₅Au₁-TiO₂. e C₂H₄ production rate of E₇-Cu₅Au₁-TiO₂ with the reaction time in comparison with recent reports during the closed glass photoreduction of CO₂ with a H₂O gas-circulation system without any sacrificial agents. f CO-TPD profile of the as-prepared samples.

Cycling tests were performed to analyze the relationship between activity and stability during the photoreduction of CO₂ into C₂H₄ (Fig. 3d). Based on the limitation of catalyst stability, the majority of photocatalytic CO₂ reduction into C₂H₄ are maintained for only a few hours and accompany with C₂H₄ production stagflation due to the poor structural stability and surface poisoning effect on the photocatalyst (Fig. 3e)^{4,5,6,7,8,9,10,11,12,13,57}. As shown in Fig. 3e, regardless of the low-efficiency or high-efficiency C₂H₄ production rate on different photocatalysts, all exhibited limited reaction time due to the deactivation of photocatalysts. Interestingly, in this work, the yield of each product increases linearly during three cycles of 72 h irradiation, as shown in Fig. 3d, and the AC-HAADF-STEM image of E₇-Cu₅Au₁-TiO₂ displays an unchanged CuAu-DAs structural configuration, as shown in Supplementary Fig. 36, indicating the superior structural stability of E₇-Cu₅Au₁-TiO₂ under such high-efficiency C₂H₄ production in the static system. Moreover, compared to the traditional static system, the intermittent flow system is considered a better choice for the improvement of photocatalytic CO₂ reduction due to the high-efficiency transport of mass^{81,82,83,84,85,86}. Therefore, the intermittent flow system is also adopted to further analyze the stability of photocatalysts under higher efficiency CO₂ conversion circumstances. Notably, in such an intermittent flow system, the C₂H₄ production rate reaches 568.8 µmol·g⁻¹·h⁻¹ (Supplementary Fig. 37a) after 24 h of irradiation, which is superior to those reported in recent works (Fig. 3e). After 5 days of irradiation, C₂H₄ production maintains a rate of 483.2 µmol·g⁻¹·h⁻¹, indicating the superhigh stability of CuAu-DAs catalysts even during such high-efficiency C₂H₄ conversion. In addition, a series of long-term photocatalytic stability tests were carried out on different metallic structure-modified TiO₂ to determine the reason for the superior stability of such specific CuAu-DAs, and CO temperature-programmed desorption (CO-TPD) was also conducted to further analyze the connection between the catalyst structure and CO affinity. As shown in Fig. 3f, Cu-TiO₂ shows the strongest CO adsorption strength centered at ~120.1 °C in the weak absorption area compared to other catalysts⁷, indicating that CO desorption is most difficult on this catalyst surface, which may suppress fresh CO₂ adsorption and conversion on Cu-TiO₂ during the photocatalytic CO₂ reduction process. As shown in Supplementary Fig. 38, Cu-TiO₂ also exhibits nonlinear CO generation, which is ascribed to the partially deactivated Cu sites induced by strong CO absorption. Furthermore, an apparent color change (yellow to black) of Cu-TiO₂ after 24 h of photocatalytic CO₂ reduction is noticed (Supplementary Fig. 38), and the black samples became yellow after 100 °C annealing in a vacuum-treated system (Supplementary Fig. 39), further indicating that the color change could be induced by intermediate adsorption during photocatalytic CO₂ reduction. Furthermore, the possible absorbed isotopically labeled ¹³C intermediates of Cu-TiO₂ after 24 h photocatalytic CO₂ reduction and subsequent 100 °C annealing in a vacuum-treated system were analyzed by GC‒MS. The GC‒MS peak sequences of ¹³CO are shown in Supplementary Fig. 40, and the peak at m/z = 29 is assigned to ¹³CO, indicating the occurrence of CO poisoning on Cu-TiO₂, consistent with the results of the CO-TPD analysis and photocatalytic stability test. However, the CO adsorption strength (109.8 °C) of E₇-Cu-TiO₂ is obviously lower than that of Cu-TiO₂ (Fig. 3f), indicating that the contractible Cu size (nanoparticle to single atom) partially alleviated the CO poisoning effect. Although the CO production of E₇-Cu-TiO₂ increases drastically compared to that of Cu-TiO₂ due to the decreased Cu coordination number, E₇-Cu-TiO₂ still exhibits nonlinear CO generation and a slight color change (white to pale yellow, Supplementary Fig. 41) due to its slight surface CO poisoning (Supplementary Fig. 42), implying that low-coordination-number Cu inevitably suffers CO poisoning. Notably, the CO adsorption strength of Cu₅Au₁-TiO₂ is lower than that of Cu-TiO₂ (Fig. 3f), confirming that the introduction of Au in the Cu lattice weakens the CO adsorption of Cu, as reported⁵³. The linearly increased CH₄ yield (Supplementary Fig. 43) and weakened CO adsorption (Fig. 3f) of Cu₅Au₁-TiO₂ compared to Cu-TiO₂ suggest that the isolated Au in the Cu lattice efficiently promote *CO conversion to CH₄ on the Au sites and suppressed *CO accumulation, which benefits the resistance to catalyst deactivation. However, Cu₅Au₁-TiO₂ also displays a nonlinearly increased CO yield induced by weak CO poisoning due to the large amount of Cu in the Cu₅Au₁ alloy (Supplementary Figs. 43 and 44), which further implies that the addition of Au not only weakens CO adsorption on Cu sites but also serves as a *CO turnover site to rapidly consume the *CO arising from Cu to alleviate the CO poisoning effect on Cu sites. Moreover, the CO-TPD analysis indicates that E₇-Cu₅Au₁-TiO₂ exhibits the highest CO adsorption capacity and lowest CO desorption temperature at ~100.0 °C, as shown in Fig. 3f, implying the enormous capacity for CO coverage and the most resistant CO poisoning on CuAu-DAs, which is beneficial for the high efficiency and stability of *CO coupling. Based on the weakest CO chemical adsorption strength of E₇-Cu₅Au₁-TiO₂, no ¹³CO of E₇-Cu₅Au₁-TiO₂ was detected by GC‒MS after a 24 h stability test and 100 °C annealing in a vacuum-treated system (Supplementary Fig. 45). Meanwhile, there is also no special characteristic peaks of C-based residual surficial absorbates was observed in FTIR spectrum after CO₂ photoreduction over E₇-Cu₅Au₁-TiO₂ (Supplementary Fig. 46), suggesting the inexistence of residual C-based intermediate and the superior stability of such CuAu-DAs structure. Owing to the synergistic effect of CuAu-DAs heteronuclear sites, rapid *CO coupling and weakened CO adsorption are simultaneously realized, ensuring superior catalytic sustainability even under such high-efficiency C₂H₄ production.

Mechanism of the photocatalytic performance

Time-dependent in situ diffuse reflectance Fourier transform infrared spectroscopy (DRIFTS) was employed to elucidate the reaction intermediates and concrete evidence of the reaction mechanism under 355 nm laser irradiation for 15 min (Fig. 4a, b). Humid CO₂ was carried into the reaction chamber until equilibrium was reached, and different infrared adsorption characteristic peaks of E₇-Cu₅Au₁-TiO₂ were gradually observed when the photocatalyst was subjected to constant irradiation. As shown in Fig. 4b, monodentate carbonate (m-CO₃^2- at 1283 cm⁻¹) and bidentate carbonate (b-CO₃^2- at 1688 cm⁻¹) are generated from the co-adsorption of CO₂ and H₂O on the surface of E₇-Cu₅Au₁-TiO₂^{57,87,88,89,90}. The peaks at 1670 and 1646 cm⁻¹ are attributed to the vibrations of *CO₂⁻ and *COOH groups^7,91,92, respectively. Moreover, the characteristic peak at 1660 cm⁻¹ is assigned to H₂O decomposition signals⁹³, and the constantly increased broad IR bands at 3200–3400 cm⁻¹ are corresponded to the vibration of *OH groups generated from water dissociation under simulated irradiation^4,6,94. The different characteristic peaks at approximately 1947 and 2235 cm⁻¹ are assigned to *CO intermediates^4,7,95, including Cu-CO and Au-CO on E₇-Cu₅Au₁-TiO₂ (Fig. 4a). Moreover, the contact angles of Cu₅Au₁-TiO₂ and E₇-Cu₅Au₁-TiO₂ are 9° and 7° (Supplementary Fig. 47), respectively, indicating the better surface hydrophilicity of E₇-Cu₅Au₁-TiO₂, which further demonstrates that the low-coordinated Cu and Au atoms are beneficial for H₂O adsorption and the generation of protons to further facilitate the protonation reaction. Coincidently, asymmetric vibration of *OCCO is observed at ~1531 cm⁻¹ (Fig. 4b), providing significant evidence for the C₂ evolution pathway, indicating that the C₂ products arise from the coupling of *CO intermediates^4,94,96,97. Moreover, the unique C-C coupling intermediates *C = C (3080 cm⁻¹)⁹⁶, *OCCOH (1579 cm⁻¹)⁶, *OCCHOH (1307 cm⁻¹)⁴, and *C₂H₄ (1447 cm⁻¹)⁶³ were also detected spectroscopically, further indicating the complexly cascaded multiple electron and proton reaction for the ultimate C₂H₄ formation. The vibration frequency of surface-bound *CHO is observed at approximately 1732 and 1710 cm⁻¹ on E₇-Cu₅Au₁-TiO₂ due to the one-electron and one-proton reduced reaction of the *CO intermediate^63,96,98. Additionally, the peaks at approximately 2993, 2944, and 2881 cm⁻¹ are attributed to the C–H symmetric stretching vibrations of methylene (Fig. 4a), facilitating the evolution of CH₄ or C₂H₄ products⁴. The characteristic spectral peaks at approximately 1372, 1474, and 2965 cm⁻¹ are attributed to *CH₂, *CH₂, and *CH₃ intermediates, respectively^57,99, demonstrating that these intermediates could be the source of hydrocarbons. Gibbs free energy theoretical calculations were conducted to elucidate the correlation between the specific nanostructure, electronic properties, and catalytic performance. By combining these results with the time-dependent in situ DRIFTS results, these possible reaction pathways were further proposed by Gibbs free energy calculations. In Fig. 4c, the formations of *CO₂, *COOH, and *CO on E₇-Cu₅Au₁-TiO₂ are constantly exothermic and spontaneous processes, indicating that the synergistic effect between Cu-SAs and Au-SAs could significantly promote CO₂ adsorption and activation to produce large amounts of *CO intermediates, which is beneficial for *CO coupling under such high concentrations of *CO. The hydrogenation and desorption of *CO to *CHO and CO species require 1.30 and 1.85 eV of energy expenditure, respectively, while *CO coupling to *OCCO consume only 0.54 eV of energy input. Because the energy barrier of *CO coupling is much lower than that of *CO hydrogenation and desorption, the *OCCO intermediates are confirmed to form preferentially during the *CO transformation process. Furthermore, under the constant hydrogenation process of *OCCO, the C₂H₄ formation paths are theoretically proposed as described in the following formulas: *OCCO ⟶ *OCCOH ⟶ *OCCHOH ⟶ *OCCH ⟶ *OCCH₂ ⟶ *OHCHCH₂ ⟶ *CHCH₂ ⟶ C₂H₄. Therefore, the reaction mechanism of photoreduction CO₂ into C₂H₄ is determined in detail on the basis of the time-dependent in situ DRIFTS experiments and density functional theory (DFT) simulations.

Fig. 4: Schematics of C₂H₄ generation during photocatalytic CO₂ reduction.

a In situ DRIFTS detection of E₇-Cu₅Au₁-TiO₂. b Corresponding magnifying area of the in situ DRIFTS spectra in a. c The free energy diagram of CO₂ conversion over the CuAu-DAs-TiO₂ photocatalyst, together with the atomic structures of the reaction intermediates.

Mechanism for the resistance of catalyst deactivation

The enhanced CO poisoning resistance of CuAu hybrid catalysts during photocatalytic CO₂ reduction tests compared to that of pure Cu was considered. DFT calculations of Cu-NCs, Cu-SAs, CuAu alloy, and CuAu-DAs modified TiO₂ models were performed to further analyze the relation between C-based intermediate conversion efficiency and different nanostructures (Fig. 5a). During the *CO generation process, there is almost no energy input for *CO₂, *COOH, and *CO formation in these models (Fig. 5a), which implies rapid *CO formation due to these exothermic and spontaneous processes. Nevertheless, all the conversions from *CO to *CHO, *OCCO, and CO molecules in these models must overcome an enormous energy barrier, indicating the rate-determined significance of *CO conversion during the CO₂ reduction process. During three possible *CO conversion routes (*CO ⟶ CO, *CO ⟶ *CHO, *CO ⟶ *OCCO), the optimized energetically favorable *CO coupling modes could not be formed on Cu-NCs-TiO₂ (C-C distance, 3.68 Å), Cu-SAs-TiO₂ (C-C distance, 2.92 Å), and CuAu-alloy-TiO₂ (C-C distance, 3.76 Å) due to the weak interaction between adjacent absorbed *CO intermediates (Fig. 5b). In Supplementary Fig. 48, CuAu-DAs-TiO₂ (−1.98 eV) presents the highest *CO adsorption energies compared to CuAu-alloy-TiO₂ (−1.44 eV), Cu-NCs-TiO₂ (−1.55 eV), and Cu-SAs-TiO₂ (−1.79 eV), ensuring the compact *CO interaction and high-concentration *CO coverage¹⁰⁰ (Fig. 5b, C-C distance, 1.40 Å), which contributes to the successful construction of the *OCCO intermediate on CuAu-DAs-TiO₂ due to the optimized surface adsorption configurations resulting from the cooperative modification of the steric and electronic properties of CuAu-DAs^{26,28,29,30,31}. Compared to *CO desorption to CO, *CO is preferentially protonated due to the lower energy barrier and energy input for *CHO production on Cu-NCs-TiO₂ (1.15 eV), Cu-SAs-TiO₂ (1.41 eV), and CuAu-alloy-TiO₂ (0.21 eV). Although the transformation of reaction routes could be an efficient way to resist *CO species accumulation, the energy input for *CO conversion to *CHO on Cu-NCs-TiO₂ and Cu-SAs-TiO₂ was still enormous, which also dramatically limited the rapid *CO transformation and led to surface *CO accumulation. Affected by the introduction of Au, the energy input of *CO conversion sharply decreased due to the synergistic cooperation between Cu and Au sites, suggesting that the suppressed *CO poisoning effect could be ascribed to the rapid *CO conversion on CuAu sites^101,102. Moreover, the energy input for *CO conversion on CuAu-DAs-TiO₂ (*CO to *OCCO, 0.54 eV) was much lower than that on Cu-SAs-TiO₂ (*CO to *CHO, 1.41 eV) and Cu-NCs-TiO₂ (*CO to *CHO, 1.15 eV), which suggested that the extremely low C-C coupling energy barrier-induced rapid *CO consumption could be the key factor for the sharp *CO poisoning resistance on CuAu-DAs-TiO₂ (Fig. 5a, c). Owing to the synergistic cooperation between the adjacent Cu-SAs and Au-SAs, simultaneous and high-efficiency *CO formation and *CO conversion could be realized based on the reconstituted surface reactant intermediate adsorption configurations and reduced *OCCO coupling reaction energy barrier, which efficaciously overcomes the activity-stability seesaw effect in photoreduction of CO₂ with H₂O into C₂H₄ (Fig. 5c).

Fig. 5: Illustration of the model mechanisms.

a The free energy diagram of CO₂ reduction to CO, *CHO, and *OCCO over different modeled surfaces. b The most energetically favorable *OCCO adsorption configurations. c Schematic illustration of the CO poisoning pathway during the photocatalytic CO₂ reduction reaction on different models.

Discussion

In summary, CuAu-DAs modified commercial TiO₂ was successfully synthesized by an up-bottom atomic synthetic process involving the vectored etching of Cu atoms in a CuAu alloy. According to the activity tests, CO-TPD, and DFT calculation results of the CuAu-DAs structure, Cu-SAs are beneficial for high-efficiency *CO production, while Au-SAs not only moderate the *CO binding strength of the composites but also strikingly increase the *CO coupling efficiency and suppress catalyst deactivation under high-concentration *CO conditions. Owing to such synergistically catalytic effect of heteronuclear DAs, CuAu-DAs-TiO₂ exhibited a superhigh rate of 568.8 μmol·g⁻¹·h⁻¹ in an intermittent flow system, and the negligible CO poisoning of CuAu-DAs-TiO₂ was observed during the 120-h photocatalytic stability test. Herein, our discovery not only provides a novel insight into the adjustable synthesis of atomic-level catalysts but also provides a new technique to optimize the selectivity, activity, and stability of photocatalysts based on reconstituted surface adsorption configurations of reactant intermediates and reduced reaction barriers.

Methods

Synthesis of Cu₅Au₁-TiO₂

Two hundred milligrams of commercial TiO₂, 105 mg of PVP, 120 mg of L-ascorbic acid, and 300 mg of KBr were dispersed in 8 mL of deionized water, sonicated for 10 min in a 50 mL round-bottom flask, and transferred into an 80 °C oil bath under constant stirring. After 10 minutes of heating, 0.083 mmol of CuCl₂·2H₂O and 0.017 mmol of HAuCl₄·4H₂O were dissolved in 4 mL of H₂O, and the mixed CuCl₂·2H₂O and HAuCl₄·4H₂O solution was rapidly transferred into the round-bottom flask. Finally, the solution was kept at 80 °C for 3 h. The sample was washed with deionized water three times and absolute ethanol three times to remove impurities. The obtained samples were denoted as Cu₅Au₁-TiO₂, which could also be simplified to C₅A₁.

Synthesis of E_t-Cu₅Au₁-TiO₂

FeCl₃·6H₂O (0.06 mmol) was dispersed in 10 mL of 0.1 M HCl, and gaseous Ar was bubbled through the solution for 60 min to eliminate dissolved O₂. Subsequently, 100 mg of Cu₅Au₁-TiO₂ was dissolved into the as-prepared solution. The suspension solution was reacted in an oil bath at 50 °C for different etching reaction time (0.5, 1, 3, 5, 7, 9, and 11 h). The as-prepared samples were washed with 0.1 M HCl and deionized water to remove impurities. The obtained samples were denoted as E_t-Cu₅Au₁-TiO₂ (simplified as E_t-C₅A₁), where t = 0.5, 1, 3, 5, 7, 9, and 11 corresponded to the etching time.

Synthesis of E_t-${{{\mathbf{Cu}}}}_{{{\mathbf{7}}}}^{{{\mathbf{N}}}}$Au₁-TiO₂

E_t-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ was prepared by the same approach as E_t-Cu₅Au₁-TiO₂, except that the amounts of CuCl₂·2H₂O and FeCl₃·6H₂O were 0.119 mmol and 0.084 mmol, respectively. The obtained samples were denoted as E_t-${{{\mathrm{Cu}}}}_{7}^{{{\rm{N}}}}$Au₁-TiO₂ (simplified as E_t-${{{\mathrm{C}}}}_{7}^{{{\rm{N}}}}$A₁), where t = 0.5, 1, 3, 5, 7, 9, 11, and 13 correspond to the etching time.

Synthesis of H_t-Cu₅Au₁-TiO₂

H_t-Cu₅Au₁-TiO₂ was prepared by the same approach as E_t-Cu₅Au₁-TiO₂ without the addition of FeCl₃·6H₂O. The obtained samples were denoted as H_t-Cu₅Au₁-TiO₂ (simplified as H_t-C₅A₁), where t = 0.5, 1, 3, 5, 7, 9, 11, and 13 correspond to the etching time.

Synthesis of Cu-TiO₂

Cu-TiO₂ was prepared by the same approach as Cu₅Au₁-TiO₂ without the addition of HAuCl₄·4H₂O, and the amount of CuCl₂·2H₂O was 0.1 mmol. The obtained sample was denoted as Cu-TiO₂ (C-T).

Synthesis of E_t-Cu-TiO₂

E_t-Cu-TiO₂ was prepared by the same approach as E_t-Cu₅Au₁-TiO₂. The obtained samples were denoted as E_t-Cu-TiO₂ (simplified as E_t-C-T), where t = 0.5, 1, 3, 5, 7, 9, and 11 correspond to the etching time.

Synthesis of Au-TiO₂

Au-TiO₂ was prepared by the same approach as Cu₅Au₁-TiO₂ without the addition of CuCl₂·2H₂O, and the amount of HAuCl₄·4H₂O was 0.1 mmol. The obtained sample was denoted as Au-TiO₂ (simplified as A-T).

Synthesis of E₇-Cu₅Au₁-Al₂O₃

Cu₅Au₁-Al₂O₃ was prepared by the same approach as E₇-Cu₅Au₁-TiO₂. The obtained sample was denoted as E₇-Cu₅Au₁-Al₂O₃.

Synthesis of C₃N₄

C₃N₄ was prepared by polymerization method. 10 g of melamine powder was dispersed into 50 mL of deionized aqueous solution and kept in an ultrasonic bath for 30 min. The mixture was stirred in water bath at 50 °C, and then dried in oven at 60 °C. The as-prepared powder was grounded, and then transferred into ceramic crucible calcined at 300 °C for 1 h, 400 °C for 1 h, and 550 °C for 4 h with 2.5 °C·min⁻¹ heating rate in a muffle furnace. After natural cooling to room temperature, the samples were washed by DI water for three times to remove impurity.

Synthesis of Cu₅Au₁-C₃N₄

Cu₅Au₁-C₃N₄ was prepared by the same approach as Cu₅Au₁-TiO₂, except that the TiO₂ was replaced by C₃N₄.

Synthesis of E_t-Cu₅Au₁-C₃N₄

E_t-Cu₅Au₁-C₃N₄ was prepared by the same approach as E_t-Cu₅Au₁-TiO₂, except that the Cu₅Au₁-TiO₂ was replaced by Cu₅Au₁-C₃N₄.

Characterizations

TEM and EDS mapping analysis (Talos F200X G2 200KV) were applied to confirm the morphologies of the materials. AC-HAADF-STEM was conducted on Titan Cubed Themis G2300 and JEM-ARM200F. XAFS of the Cu K-edge and Au L-edge were measured at the BL14W1 station of the Shanghai Synchrotron Radiation Facility (SSRF, China), and Cu foil and Au foil were used as the reference samples. XAFS data were obtained by means of Athena and Artemis software according to standard procedures. ICP‒AES was carried out on a Varian VISTA-MPX instrument. The crystal structure was characterized by powder XRD using Cu Kα radiation (λ = 0.15406 nm) on a Bruker D8 Advance X-ray diffractometer. The compositions of the catalysts were analyzed by XPS (Thermo Scientific K-Alpha). UV‒vis‒NIR DRS was performed on a Shimadzu UV-2450 spectrophotometer. PL and TRPL analyses were conducted by means of an Edinburgh Instruments instrument (FLS-980). TPD tests were performed on an AutoChem1 II 2920 with a thermal conductivity detector. The surface hydrophilicity of the as-prepared samples was determined by contact angle measurement (CA, KSV CM200, Finland).

Photoelectrochemical tests

Photocurrent and EIS tests were conducted by a CHI660B electrochemical analyzer with a standard three-electrode system, where Ag/AgCl and Pt wire were used as reference and counter electrodes, respectively, and the samples spin-coated onto fluorine-doped tin oxide (FTO) glasses served as working electrodes. During the photocurrent and EIS tests, 0.2 mol·L^‒1 Na₂SO₄ solution and 0.1 mol·L^‒1 KCl solution containing 5 mmol·L⁻¹ K₄[Fe(CN)₆]/K₃[Fe(CN)₆] were applied as electrolytes. Moreover, the Mott-Schottky test was carried out in a 0.5 mol·L^‒1 Na₂SO₄ solution.

Photocatalytic activity tests in different reaction systems

(a) Static system: 20 mg of photocatalyst and 5 mL of deionized water were dispersed on a 45 mm diameter quartz glass under sonication and then placed at 60 °C in a vacuum oven for 3 h to dry. Photoreduction CO₂ reaction tests were conducted on a closed glass gas-circulation system (Labsolar-6A, Beijing Perfectlight Technology Co., Ltd) with the addition of 2 mL deionized water, and a quartz tray was used to separate water and the quartz glass coated by photocatalysts. The temperature was adjusted to 25 °C using circulating water. (b) intermittent flow system: 2.0 mg of photocatalyst was dispersed on the surface of the microporous membrane with a radius of 2.35 cm by filtration and then sealed in the intermittent flow system with the addition of 0.5 mL deionized water. Before irradiation, the gas-circulation system was vacuum-treated for 15 min and then filled with high-purity CO₂ (99.99%). The reactor filled with high-purity CO₂ was vacuum-treated again and then filled with high-purity CO₂ to reach 90 kPa. The photocatalyst was placed 10 cm away from a 300 W Xe lamp (Microsolar 300, 320–780 nm, 250 mW·cm⁻², Beijing Perfectlight Technology Co., Ltd). Gas products were detected by means of a gas chromatograph (GC9790II, FULI INSTRUMENTS) equipped with a flame ionization detector (FID) and thermal conductivity detector (TCD). The produced gases were calibrated with a standard gas mixture, and the identity was determined by the retention time. ¹³CO₂ isotope labeling experiments were conducted under the same conditions.

¹³CO₂ isotope labeling experiments

The gas products of ¹³CO₂ isotope labeling experiments were conducted on Agilent 7890B-5977B GC with MS detector. The HP–5MS type column was adopted to analyze the C-based products. The heating programming was started at 40 °C and maintained for 5 min. The temperature of system was increasing from 40 °C to 290 °C with a heating rate of 20 °C·min⁻¹, and the system kept 290 °C for 10 min. Helium was used as a carrier gas in GC with a flow rate of 1 mL·min⁻¹. The temperature of gas sampling valve is 280 °C under non-diversion mode. Electron impact ionization with 70 eV voltage under the full scan of 2-350 u. The matching degree of CO, CH₄, C₂H₄, and C₂H₆ relied on the database are about 93, 96, 92, and 93%, respectively, which could be regarded as the reference for source of C-based products.

In situ DRIFTS experiments

In situ DRIFTS experiments were conducted on a Nicolet iS10 (Thermo) machine, and the photocatalysts were sealed in the reaction chamber with a quartz window. CO₂ and H₂O were carried into the reaction chamber by N₂ flow until equilibrium. After taking the equilibrium system before reaction as the blank background, IR signals were collected under 355 nm laser irradiation (3W) through the quartz glass window.

XAFS date analysis

Data reduction, data analysis, and EXAFS fitting were performed and analyzed with the Athena and Artemis programs of the Demeter data analysis packages¹⁰³ that utilizes the FEFF6 program¹⁰⁴ to fit the EXAFS data. The energy calibration of the sample was conducted through a standard Cu foil, which as a reference was simultaneously measured. A linear function was subtracted from the pre-edge region, then the edge jump was normalized using Athena software. The k³-weighted χ(k) data were Fourier transformed after applying a Kaiser-Bessel window function (Δk = 1.0). For EXAFS modeling, the global amplitude EXAFS (CN, R, σ² and ΔE₀) were obtained by nonlinear fitting, with least-squares refinement, of the EXAFS equation to the Fourier-transformed data in R-space, using Artemis software, EXAFS of the Cu foil is fitted and the obtained amplitude reduction factor S₀² value (0.845) was set in the EXAFS analysis to determine the coordination numbers (CNs) in the Cu-O and Cu-Au scattering path in sample.

Computational methods

DFT models were constructed based on the results of AC-HAADF-STEM and XRD, and the (101) crystal plane of TiO₂ was selected to construct the computational models. All DFT calculations were performed by the Vienna Ab Initio Simulation Package (VASP). The Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) was utilized to address the exchange-correlation interactions. All slab models were applied with a 15 Å vacuum layer to prevent interactions between slabs. A 2 × 2 × 1 k-point was sampled in the Brillouin zone. A plane-wave basis expansion with a 400 eV energy cutoff was adopted to optimize the geometric structures, and the electronic forces were converged to 1 × 10⁻⁵ and 0.03 eV·Å⁻¹, respectively. The Gibbs free energies were calculated at 298.15 K by means of the formula G = E_DFT − TS + E_ZPE, where E_DFT, TS, and E_ZPE represent the electronic energy of each step, entropy contribution, and zero-point energy, respectively.