Excessive emissions of carbon dioxide (CO₂) caused by the ever-increasing consumption of fossil fuels have resulted in many environmental problems [1-8]. The conversion of photocatalytic CO₂ into value-added chemical products (CH₄, CO, C₂H₄, etc.) has attracted significant interest because it can utilize solar light and H₂O, that are abundant in nature, as energy and proton sources, respectively, without secondary pollution to the air environment. It is acknowledged that the prerequisite of improving the reduction performance of CO₂ photocatalysis is the establishment of suitable and high-performance photocatalysts. TiO₂ is one of the most widely utilized semiconductor materials in catalytic reactions owing to its environmental benignity, cost-effectiveness, and high physical and chemical stability [1,5,6,9]. More importantly, the conduction band (CB) of TiO₂ is relatively negative, which would improve the reduction ability of photogenerated electrons. However, the limited charge separation efficiency and optical absorption hinder the reduction performance of the CO₂ photocatalysis via TiO₂.

In the strategies of regulating activity of CO₂ photoreduction, constructing heterojunction with suitable semiconductors, photocatalysts with excellent performance can be obtained, which would facilitate charge separation by trapping photoelectrons [1,2,10-14]. Tungsten oxide (WO₃), has attracted much attention with relatively small bandgap (2.2–2.6 eV), low in toxicity as well as stable in acidic and oxidative conditions [10,14]. Importantly, WO₃ has the characteristic of band matching with TiO₂ [11,13], coupling the two semiconductors to build a heterojunction, which improves the charge separation by trapping the photoelectrons and extending the energy range of photoexcitation toward the visible range, seems to be an efficient strategy to promote the performance of CO₂ photoreduction. Generally, various heterostructures have been studied for photoreduction of CO₂ [1,10-14]. However, compare with conventional fuel production technologies, the efficiency of solar to fuel conversion is still very moderate. Thus, advanced strategies are required to design and fabricate a challenging heterojunction photocatalyst to improve efficiency of CO₂ conversion.

Metal co-catalysts have also been widely applied in many photochemical reactions because of their high activity and selectivity [15-24]. In particular, when tuned to the atomic scale, metal catalysts can maximize atomic utilization and accelerate the surface catalytic process; therefore, a number of atomically dispersed metals on semiconductors have emerged as effective platforms for establishing high-performance photocatalysts in recent years [4,6,8,19]. Compared with noble metals (Au, Rh, and Pt), the earth-abundant metal Cu is regarded as one of the most attractive metal co-catalysts for producing hydrocarbons from CO₂ due to the diversity of reduction products [4,8,20-22]. For example, Li et al. [4] developed a photoinduction strategy to achieve the formation of Cu single atoms (SA) on a UiO-66-NH₂ support. Notably, the introduction of Cu SAs not only significantly enhanced transfer and separation efficiency, but also acted as active centers to facilitate the conversion of CO₂ to CHO* and CO* intermediates, leading to excellent selectivity toward methanol and ethanol. Likewise, wang et al. [7] designed single-atom Cu modified polymeric carbon nitride (CN), in which the C-Cu-N₂ single-atom catalytic site activated CO₂ molecules and reduced the energy barrier toward photocatalytic CO₂ reduction, resulting simultaneous production of CO, CH₄ and CH₃OH. Inspired by these efforts, the improvement of CO₂ reduction activity on heterojunction could be achieved via loading Cu SA on its surface. In fact, there are several reports concerning single atomic multi-heterojunctions on photoreduction of CO₂. However, the mechanism of CO₂ conversion is not well elaborated, especially the effect of charge separation on the d-band of metal single atom. It is known that the interaction between metal d-band and CO₂ molecular orbitals is directly reflected in the adsorption energy, the position of d-band center also determined by the interact with CO₂ molecular orbitals of intermediates in catalytic behavior of photocatalyst [25-28]. Therefore, the d-band center has proven to be an extremely useful electronic descriptor to explain photocatalytic CO₂ reduction selectivity.

Hence, in this work, Cu-TiO₂/WO₃ heterojunction was synthesized by a simple chemisorption strategy to investigate the activity and selectivity of CO₂ photocatalytic reaction. In the process, Cu single atoms were firstly loaded on the surface of TiO₂ (Cu-TiO₂) via reduction method, and then coupled with WO₃ to synthesize a S-scheme heterojunction photocatalyst to ensure efficient electron utilization and chemical transformation. Under simulated solar light irradiation, the Cu-TiO₂/WO₃ photocatalyst displayed an obviously enhanced photoreduction activity and selectivity toward CH₄. Using first-principles density functional theory (DFT) calculations, we revealed the decisive influence of charge separation effect of heterojunction on interaction between Cu single atom and TiO₂ substrate, and the d-band position.

The morphologies and microstructures of as-prepared samples were shown in Fig. 1 and Fig. S1 (Supporting information). In the high-resolution transmission electron microscope (HRTEM) images (Fig. 1a), bare TiO₂ was uniform and the lattice fringes of 0.33 nm correspond to the (101) planes of anatase-TiO₂ [1,6]. And in the aberration-corrected high-angle annular dark-field scanning transmission microscopy (AC HAADF-STEM) images, isolated bright spots were observed on Cu-TiO₂ sample, which indicated that Cu were homogeneously diffused on the surface of TiO₂ nanoparticles in the form of atoms rather than clusters. After Cu-TiO₂ was contacted with WO₃, a clear boundary line (red) was observed (Fig. 1c), and the lattice fringes with distances of 0.33 nm and 0.37 nm were attributed to the (101) crystal plane of TiO₂ and (200) crystal plane of WO₃, respectively [1,6]. The result indicated that the Cu-TiO₂/WO₃ heterojunction was successfully obtained. Thereafter, according to the energy-dispersive X-ray (EDX) mapping shown in Figs. 1d and e, Cu-TiO₂/WO₃ displayed homogeneous distribution of Ti, O, Cu and W elements. Besides, according to the result of inductive coupled plasma emission spectra (ICP) shown in Table S1 (Supporting information), the loading amount of Cu single atom in Cu-TiO₂/WO₃ was just slightly loss, which may be caused by cleaning during in the process of preparation.

To explore the coordination environment and chemical states of Cu atoms, X-ray absorption fine structure (XAFS) was measured on Cu-TiO₂/WO₃. As shown in Fig. 2a, the peak of Cu K-edge X-ray absorption near edge structure (XANES) located at 8997 eV of Cu-TiO₂/WO₃ was similar to that of CuO, indicating the existence of Cu single atom with a positive 2 valence. Subsequently, k³-weighted Cu K-edge Fourier transform (FT) curve and wavelet transform (WT) analysis of Cu-TiO₂/WO₃ were shown in Fig. 2b and Fig. S2 (Supporting information), except for a main scattering peak at 1.47 Å attributed to the Cu-O coordination, the peak at 2.24 Å corresponding to Cu-Cu interaction in Cu-TiO₂/WO₃ was not detected, confirming the single atom dispersion of Cu. According to the fitting results of XANES shown in Table S2 (Supporting information), the Cu single atoms were coordinated with 4.0 ± 0.26 O atoms of TiO₂ with an average bond length of 1.922 ± 0.006 Å, which was slight shorter than Ti-O bond (1.94 Å) due to the lattice distortion and/or contraction in Cu-TiO₂/WO₃. Consequently, the tetrahedral oxygen-coordinated Cu-O₄ sites were anchored at the Ti sites, schematically shown in Fig. S3 (Supporting information).

Fig. 2c displayed the X-ray diffraction (XRD) patterns of samples. The spectra of TiO₂ agreed with simulated anatase structure (JCPDS No. 21–1272), and the introduction of Cu SA did not change TiO₂ crystallinity. For Cu-TiO₂/WO₃ heterojunction, besides the peaks of TiO₂, the peaks assigned to WO₃ (yellow line) were observed [1,14], indicating the successful construction of heterojunction. Next, as shown in Fig. S4a (Supporting information), X-ray photoelectron spectrum (XPS) of Cu-TiO₂/WO₃ verified the existence of Ti, O, Cu, and W. In Fig. S4b (Supporting information), the Ti 2p spectra of TiO₂ exhibited Ti 2p_1/2 and Ti 2p_3/2, respectively. A positive (0.21 eV) shift was observed in Cu-TiO₂, indicating a decreasing of electron density on Ti owing to the electron enrichment of Cu sites. For Cu-TiO₂/WO₃, a larger shift was happened (0.45 eV), suggesting that more electrons on Ti were transferred to Cu sites in the heterojunction. In Fig. 2d, the spectra of Cu 2p was divided into 2p_1/2 and 2p_3/2 on Cu-TiO₂ [8,21], respectively. While, the both peaks shifted to the lower binding energies (0.28 eV) in Cu-TiO₂/WO₃, indicating the electron density of Cu sites was increased. A similar tendency with Ti 2p appeared in W 3d XPS shown in Fig. S4c (Supporting information), which indicated that the electrons of W site of heterojunction also decreased. The results suggested that Cu-TiO₂/WO₃ heterojunction accelerated the charge migration behavior.

Ultraviolet-visible diffuse reflectance spectroscopy (UV–vis DRS) was applied to survey the optical property of samples. In Fig. S5a (Supporting information), comparing with Cu-TiO₂, the range of optical absorption over Cu-TiO₂/WO₃ further broaden due to the visible adsorption of WO₃, which was beneficial to CO₂ photoreduction. To reveal band structure of the heterojunction, the bandgap and Mott-Schottky plots were conducted and shown in Figs. S5b-e and Table S3 (Supporting information). According to the results, the valence bands (VB) of TiO₂ and WO₃ were 2.44, and 2.53 V, their conduct bands (CB) were −0.61 and 0.02 V [11,13], respectively. The movement of electrons in heterojunction photocatalyst can be described by the work functions (W) of semiconductors. W of TiO₂ and WO₃ were 4.69 and 5.05 eV [11,13]. When Cu-TiO₂/WO₃ heterojunction was formed, the free electrons would transfer from TiO₂ to WO₃ easily until the Fermi levels (Ef) tend to equilibrate at the interface. Concurrently, an internal electric field (IEF) was formed. When the heterojunction was exposed to light, photoelectrons were excited and transferred from VB to CB of WO₃ and TiO₂, respectively. Under the effect of IEF, the photoelectrons on CB of WO₃ easily migrated and combined with the holes on VB of TiO₂. This charge transfer behavior of Cu-TiO₂/WO₃ heterojunction was matched with the S-scheme mechanism [11,13], which was consistent with the XPS results. The illustration of charge transfer under simulated sunlight for efficient photocatalysis over Cu-TiO₂/WO₃ was shown in Fig. S6 (Supporting information). In this regard, efficient charge separation would be achieved. Finally, more photoinduced electrons would participate in the photocatalysis process. Comparing with E₀ (CO₂/CH₄ = −0.17 V vs. NHE, pH 0), Cu-TiO₂/WO₃ has the ability of photocatalytic reduction of CO₂ to CH₄.

To evaluate photocatalytic property of Cu-TiO₂/WO₃, CO₂ photoreduction experiment was performed under simulated sunlight irradiation with water as hole sacrificial reagents. To optimize the photoreduction performance, TiO₂ loaded with different amounts of Cu single atom and the heterojunction with different mass ratios of Cu-TiO₂ and WO₃ were also prepared. CO₂ photoreduction experiments of all samples were shown in Fig. 3a and Fig. S7 (Supporting information), both of pure TiO₂ and WO₃ converted CO₂ into CO, with formation rates of 3.28 and 1.08 µmol g⁻¹ h⁻¹, respectively. After Cu SA introduced, CH₄ become the main product of CO₂ photoreduction, and a small amount of CO and C₂H₄ were also observed. Moreover, comparing to 0.5 and 2.0 mL, 1.0 mL of Cu²⁺ solution was the optimal volume and Cu-TiO₂ exhibited the supreme photocatalytic performance with formation rates of 32.29 µmol g⁻¹ h⁻¹ for CH₄. The results suggested that the Cu single atom center may be a highly active site for CO₂ to CH₄. In addition, too small amount of Cu single atom would lead to insufficient active sites for photoreduction, and excessive ones would result in aggregation, which also affected the activity of CO₂ photoreduction. For Cu-TiO₂/WO₃ photocatalyst, it found that CH₄ was still the main product in CO₂ conversion. And the evolution rate of CH₄ reached 98.69 µmol g⁻¹ h⁻¹, which was much higher than those of pristine WO₃, TiO₂/WO₃ and Cu-TiO₂ samples. The results indicated that the formation rate of CH₄ were significantly improved due to the charge separation effect of Cu-TiO₂/WO₃ heterojunction.

To further verify the role played by Cu single atom and the effect of charge migration on Cu-TiO₂/WO₃ heterojunction, TiO₂/WO₃ sample was prepared in the absence of Cu single atom. In Fig. 3a, it found that TiO₂/WO₃ only could convert CO₂ into CO, with a rate of 6.13 µmol g⁻¹ h⁻¹, and no CH₄ was detected, suggesting that Cu site may be the key active center for the conversion of CO₂ to CH₄ on Cu-TiO₂/WO₃. Next, Cu-TiO₂ and WO₃ were mixed by simple mechanical stirring without interfacial chemical contact. This sample was named as Cu-TiO₂+WO₃ and its photoreduction result displayed in Fig. S7b (Supporting information) was just close to that of Cu-TiO₂, far lower than that of Cu-TiO₂/WO₃, indicating that the construction of Cu-TiO₂/WO₃ significantly improved the activity of Cu active sites due to the charge migration and separation, thereby accelerating the conversion of CO₂ to CH₄. In Table S4 (Supporting information), the selectivity of CH₄ was also calculated. For Cu-TiO₂/WO₃, the electron selectivity toward CH₄ was reached 88.5%, which were 15.1% higher than that of Cu-TiO₂, indicating that the charge separation effect of the heterojunction promoted the pathway reaction of CO₂ to CH₄. The above results indicated that the construction of Cu-TiO₂/WO₃ photocatalyst not only enhanced the activity of CO₂ photoreduction, but also increased the selectivity toward CH₄.

To further confirm the photoreduction of CO₂, ¹³C isotope labeling and blank control tests were performed. As shown in Fig. 3b, the overwhelming peaks of m/z = 17 (¹³CH₄), 29 (²⁹CO) and 30 (¹³C₂H₄) rooted in ¹³CO₂ sources indicated that all reduction products were originated from the ¹³CO₂ rather than other carbonaceous impurities. Blank control experiments were shown in Fig. S8 (Supporting information), where no reduction product was detected in N₂ or in the absence of photocatalysts, light irradiation, and H₂O, indicating that CO₂ gas, photocatalyst, light irradiation, and H₂O were the prerequisites for the photoreduction of CO₂. Subsequently, the stability of the heterojunction was investigated. As shown in Fig. S9 (Supporting information), the XRD structure of Cu-TiO₂/WO₃ did not change before and after photoreduction reaction. The result proved that Cu-TiO₂/WO₃ has a relatively stable structure.

Next, we studied charge-transfer properties to investigate the reason of the enhanced photocatalytic activity on Cu-TiO₂/WO₃. In Fig. S10a (Supporting information), the transient photocurrent of heterojunction was enhanced significantly compared with those of other samples, implying that more effective charge separation over Cu-TiO₂/WO₃. And in Fig. S10b (Supporting information) Cu-TiO₂/WO₃ showed a marginally lowest PL signal, indicating more effective retardation of the charge recombination [14]. Electrochemical impedance spectroscopy (EIS) can also reflects the charge-transfer behavior. As shown in Fig. S10c (Supporting information), Cu-TiO₂/WO₃ has the lowest charge-transfer resistance of all samples, suggesting that more efficient transmission of charge have been achieved in Cu-TiO₂/WO₃. According to the results, it demonstrated that the S-scheme heterojunction could promote the migration and separation of surface photogenerated charges, thus improving photocatalytic performance of CO₂, which provided a clear evidence for electron transfer within several components of Cu-TiO₂/WO₃. Time-resolved photoluminescence (TRPL) profiles of samples were measured to further reveal the efficient separation of photogenerated carriers. At the excitation wavelengths (E_W) of 425 nm, the fitted decay curves disclose the lifetime (τ) and percentage (Rel.%) of charge carriers. As shown in Fig. S10d and Table S5 (Supporting information), Cu-TiO₂/WO₃ has a lowest percentage of τ1-carriers than other samples, indicating that the charge recombination was further inhibited due to the charge migration [14]. The average carrier lifetime of Cu-TiO₂/WO₃ (τ_a = 8.86 ns) was higher than that Cu-TiO₂ (4.92 ns), which was attributed to efficient transfer and separation of charges over Cu-TiO₂/WO₃ heterojunction. Therefore, it was not surprising that the Cu-TiO₂/WO₃ composite sample exhibit improved photocatalytic CO₂ reduction performance.

To understand the reaction mechanism of CO₂ photoreduction over Cu-TiO₂/WO₃, in situ FTIR measurements were performed to investigate the intermediates in the process of CO₂ conversion. As shown in Fig. 3c, the peaks located at 1075, 1145 and 1341 cm⁻¹ were detected and ascribed to the CHO*, CH₃O*, and COOH* intermediates [29-32], respectively, which were the crucial groups for CO₂ conversion to CH₄. A new peak at 1203 cm⁻¹ was observed and ascribed to COCO* group [33], a crucial C-C coupling intermediate of C₂H₄ formation. Additionally, the peak at 1659 cm⁻¹ corresponded to the C=O stretching belonging to the CO* intermediate [34]. The in situ FTIR was agreed well with the results of CO₂ photoreduction test on Cu-TiO₂/WO₃. Based on the results, it speculated that Cu sites converted the adsorbed CO₂ to COOH*, CO*, CHO* and CH₃O* intermediates with the participation of protons and electrons, which in turn to produce CH₄. For weaker adsorbed groups [34], such as CO*, could be desorbed from the Cu active sites to generate CO gas. Meanwhile, partial CO* would also couple with the neighbor CO* to form COCO* intermediates, and finally form C₂H₄ with the assistance of multiple electrons and protons.

Next, DFT calculation was applied to examined the mechanism of CO₂ photoreduction on Cu-TiO₂/WO₃ photocatalyst. The model structure of the heterojunction was built by TiO₂ (101) and WO₃ (200) planes by replacing the Ti with Cu atoms. To further confirm the active site of the heterojunction, the adsorption energies of CO₂ on different metal sites were calculated and shown in Table S6 (Supporting information). According to the results, it proved that the Cu single atom site was the optimal active center in Cu-TiO₂ and Cu-TiO₂/WO₃ photocatalysts for CO₂ to CH₄. Next, Gibbs free energies (ΔG) were calculated in the premise of Cu single atom as active site, and shown in Figs. 4a and b and Table S7 (Supporting information). The ΔG of photoreduction of CO₂ to COOH*, CO*, CHO*, CH₃O* and CH₄ formation were displayed in Fig. 4a. Along with the hydrogen proton (H⁺) produced by water hydrolysis, the C bond of CO₂* was interacted with a pair of proton and electron to form COOH*. According to the calculated results, it found that the whole process of CO₂-to-CH₄ conversion on the Cu-TiO₂/WO₃ was more favorable in thermodynamics behavior than that on the Cu-TiO₂ surface. And the formation of COOH* intermediate was the rate-limiting step for both photocatalysts. For Cu-TiO₂/WO₃, an energy of 0.54 eV was needed to complete the step of CO₂ → COOH*. By contrast, ΔG of 0.79 eV (ΔG_COOH* - ΔG_CO) was essential in endothermic reaction of Cu-TiO₂, suggesting the structure of Cu-TiO₂/WO₃ could efficiently lower the formation energy of key COOH* intermediate. From the results, it suggested that more electrons were transferred to Cu active sites from TiO₂ due to the charge separation effect of the heterojunction, which helped to optimize the rate-limiting step and promote the conversion of CO₂-to-CH₄.

Typically, the adsorption and activation of CO₂ is also critical for photocatalytic reduction [29]. Thus, the molecule structure of CO₂ before and after adsorption on the catalyst surface was calculated. In Table S8 (Supporting information), comparing with Cu-TiO₂, the CO₂ adsorbed on Cu-TiO₂/WO₃ has a smaller C-O-C angle, and longer C=O bonds, resulting in more curved configuration of adsorbed CO₂ and easier cleavage of C=O bonds [35,36]. Furthermore, in Table S9 (Supporting information), the adsorption energy of CO₂ on heterojunction was lower than Cu-TiO₂, indicating CO₂ adsorption and activation were easier to occurred on its surface.

Electronic properties of the catalyst were further carried out and analyzed to figure out the underlying factors governing the significant activity of the heterojunction, Firstly, the feature of CO₂ adsorption was further evaluated through the partial density of states (PDOS) on Cu active site, and the results were shown in Fig. S11 (Supporting information). When the O atom of CO₂ molecule was adsorbed on the Cu-TiO₂/WO₃ surface, the peaks shifted to lower energy than that of Cu-TiO₂, suggesting that the orbital coupling of CO₂ molecule with Cu active site of Cu-TiO₂/WO₃ was more prone to carry out [37]. Next, the d-band center of Cu site was further explored to elaborate the adsorption behavior. In Fig. 4c and Table S10 (Supporting information), it suggested that the d-band centers of Cu in two catalysts depended on its local coordination environment, and was calculated to be −1.49 (Cu-TiO₂/WO₃) and −1.77 eV (Cu-TiO₂). Further analysis found that a more remarkable overlap between Cu 3d in Cu-TiO₂/WO₃ and O 2p of CO₂ at the identical position. Apparently, more substantial orbital and electron interactions of Cu sites and O were present than that of Cu in Cu-TiO₂, which uplifted the d-band center of Cu-TiO₂/WO₃ closer to the Fermi level. Such changes in the d-band centers of Cu were precisely correlated with the adsorption energy of key COOH* intermediates. In addition, the d-band of Ti was also calculated to elaborate the electron interaction between the Cu SA and the substrate TiO₂ in the heterojunction. As shown in Fig. 4d, comparing with Cu-TiO₂, the d-band of Ti become narrowed in Cu-TiO₂/WO₃, and the center of the d-band was also uplifted and closer to the fermi level, indicating more stronger interaction between the Cu SA sites and TiO₂ substrate.

The S-scheme heterojunction of coupling TiO₂ supported single Cu atoms with WO₃ provided an efficient method for charge separation and migration. The results showed that Cu single atom played a key role as active centers in CO₂ photoreduction over Cu-TiO₂/WO₃ photocatalyst, and the heterojunction significantly improved the mobility efficiency of surface photogenerated charge based on the unique advantages of S-scheme. Moreover, the charge separation effect of the heterojunction further promoted the adsorption and activation of CO₂ on Cu active sites. Importantly, the uplifted d-band centers of Cu and Ti of Cu-TiO₂/WO₃ photocatalyst enhanced the electron interaction between Cu single atom and TiO₂ substrate, and optimized the Gibbs free energies of CH₄ formation. The specific pattern of combining single-atom and heterojunction fully exploits the highly efficient charge separation to open up a new channel for the design of efficient photocatalyst systems.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was supported by the grants from the National Natural Science Foundation of China (Nos. 21872102 and 22172080) and the Tianjin "Project + Team" innovation team, 2020.

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2023.108607.