Selection of a single atom for surface charge modulation to promote CO₂ activation and stabilize the *COOH intermediate for solar fuels

Abstract

Modulating the surface charge distribution is a key strategy for enhancing photocatalytic CO₂ reduction as it effectively activates the inert CO₂ molecule into a reactive, bent configuration. However, the rational design of such catalysts is fundamentally challenged by the lack of a systematic, predictive guideline for selecting optimally anchored single-metal atoms. We address this critical gap by establishing a novel predictive metal selection principle based on the synergistic effect of the valence electron count and electronegativity of a single atom. This principle successfully guided the selection of transition metals (Cr, Fe, and Ni) anchored onto WO₃ nanoplates via wet impregnation, ensuring selective surface modification while preserving the bulk properties. The Fe-anchored WO₃ catalyst, identified by our principle, demonstrated the most substantial surface electron density localization. This rational optimization led to a remarkable three-fold enhancement in the PC-CO₂RR efficiency, selectively yielding CO as the primary product. Comprehensive experimental and theoretical analyses confirmed that this localized charge accumulation promotes significantly stronger CO₂ chemisorption and critically stabilizes the key intermediate, *COOH. Specifically, the reaction became thermodynamically spontaneous with ΔG(*COOH) = −1.73 eV. Our findings establish a strong correlation between the localized surface charge density and CO₂ activation, providing a fundamental and generalizable guideline for the rational design of high-performance catalytic materials for energy and sustainability applications.

---

1 Introduction

Global warming has been a concern for several decades, leading to serious environmental problems, such as rising mean surface temperature and sea levels and climate change.¹ In particular, carbon dioxide (CO₂) is one of the most threatening greenhouse gases due to its massive emissions from human activities, which may cause serious or irreversible damage.² As one of the alternative solutions, the photocatalytic CO₂ reduction reaction (PC-CO₂RR) has attracted research attention as it can convert CO₂ into valuable fuels (such as CO and CH₄) using sustainable solar energy.^3–6 However, the PC-CO₂RR is limited by its low conversion efficiency due to the chemically inert nature of the linear configuration of the CO₂ molecule.⁷ According to the literature, the adsorption of CO₂ on surface active sites is a critical step and plays a significant role in influencing the reduction reaction pathways and ultimately determining the productivity and selectivity of PC processes.^3,8 While intrinsic defects, grain boundaries, and edges are recognized as active sites for the CO₂RR, activating CO₂ on the inert basal plane or the surface atomic structure of the catalyst is more advantageous.^9–11 Therefore, redesigning the surface of a catalyst to enhance the product yield is essential for achieving high conversion efficiency.

In the realm of photocatalysis, the arrangement of atoms on the catalyst surface is momentous due to its substantial impact on the surface electron density distribution for catalytic activity and selectivity.^12–14 For instance, charge-enriched dual metal site catalysts, e.g., V_S–CuIn₅S₈, alter the CO₂ adsorption mode from weak M–C or M–O bonds to stronger M–C–O–M bonds. This alteration allows CO₂ to remain on the catalyst surface for a longer duration, awaiting the subsequent protonation step, all the way to yield CH₄.¹² Generally, the localized surface electron density enables CO₂ adsorption through orbital hybridization between C or O and surface atoms.^13,15,16 Several studies have focused on surface engineering to modify the surface atomic structure to enhance the efficiency of the PC-CO₂RR.^13,15–17 For example, Zhou and coworkers manipulated charge accumulation by anchoring Au single atoms and regulating the vacancy types in the host material. As a result, CO₂ interaction with the catalyst surface changed from physical to chemical adsorption.¹³ The chemisorption of CO₂ on the catalyst surface leads to electron transfer from the catalyst to the antibonding orbitals of CO₂, which weakens the C=O bonds and facilitates subsequent chemical transformations.¹⁸ This structural change converts CO₂ to a nucleophilic species, leading to the easier activation of CO₂ for photocatalytic conversion.^18–20

To achieve surface charge accumulation, doping engineering is the most effective method.^21–24 However, conventional doping methods, such as one-pot synthesis, may not be ideal because they can alter the entire crystal structure of the catalyst, which may potentially introduce detrimental defects and cause significant carrier recombination.^25–27 In contrast, the wet impregnation approach enables selective modification of surface properties by anchoring dopants on the host semiconducting material without affecting its bulk characteristics.^25,28 Hence, anchored dopants can function as supplementary catalytically active sites directly at the surface-ambient interface.^28–30

Currently, single-atom catalysts (SACs) have garnered immense interest due to their exceptional ability to significantly boost catalytic performance even at minimal metal loadings. Beyond merely enhancing physical properties, such as light absorption and charge transfer and separation, and introducing new active sites,^31,32 recent studies have highlighted that these improvements fundamentally stem from the precise control of electronic structures.³³ SACs enable the strategic regulation of orbital electron occupancy in adsorbates, thereby stabilizing specific key intermediates.^34,35 Consequently, research has begun to focus on dual or multi-single-atom systems to steer product selectivity from C1 to C2 products.^36,37 By manipulating surface charge populations, it is possible to finely tune adsorption strength and intermediate stability, which is a vital consideration for the future development of multicarbon fuels.

However, a significant challenge remains: the selection of the optimal transition metal atom for anchoring lacks a systematic, predictable guideline. Existing studies often rely on trial-and-error, limiting the efficient, rational design of advanced catalysts. This limitation is particularly evident in the sparse exploration of spin effects on the occupancy of antibonding orbitals within key intermediates. Such electronic regulation is essential for optimizing the electronic acceptance–donation interaction between the catalyst and reactants, as recent studies suggest that spin-polarized states can significantly stabilize intermediate species through d-orbital rearrangement.^34,35 Addressing this crucial gap, our study establishes a clear selection mechanism. We propose that the key to maximizing electron enrichment on the model catalyst surface lies in the synergistic effect of the number of valence electrons of the metal and its electronegativity. This principle allows us to predictively select single transition metal atoms that can induce the strongest localized charge accumulation, optimizing CO₂ activation. Ultimately, the electronegativity and valence electron count can serve as robust descriptors to predict and guide the selection of suitable transition metals, thereby laying a foundational framework for the rational design of high-performance catalytic materials.

Inspired by the above statements, we designed a series of transition metal-anchored WO₃ catalysts to enhance CO₂ adsorption and promote CO₂ reduction. Notably, WO₃ nanoplates were selected as the model catalyst due to their earth-abundance, economical cost, nontoxicity, and photostability.^38,39 The wet impregnation method was employed to selectively introduce fourth-period transition metals (Cr, Fe, and Ni) as anchored dopants on the surface of WO₃. These elements were selected due to their excellent catalytic performance and their capability to effectively modulate surface charge density.^21,40,41 Characterization techniques, including X-ray absorption near-edge structure (XANES), X-ray photoelectron spectroscopy (XPS), in situ Fourier transform infrared (FTIR) spectroscopy, and density functional theory (DFT) calculations, revealed modulation of the charge density distribution and *COOH key intermediate stabilization; asterisks stand for the active site when transition metals are anchored on the surface. The enhancements in CO₂ adsorption and activation were also proven through theoretical calculations and the experimental PC-CO₂RR. Ultimately, a strong correlation between charge accumulation, CO₂ adsorption, and activation was observed, providing valuable insights for the design of next-generation catalysts.

2 Experimental

2.1 Chemicals

All chemicals and reagents were of analytical grade and used directly without further purification. Tungsten(VI) chloride (WCl₆, ≥99.9%) was purchased from Sigma-Aldrich. Oxalic acid (H₂C₂O₄, anhydrous, 98%) was purchased from Alfa-Aesar. Chromium(III) nitrate nonahydrate (Cr(NO₃)₃·9H₂O, 99%), iron(III) nitrate nonahydrate (Fe(NO₃)₃·6H₂O, 99%), and nickel nitrate hexahydrate (Ni(NO₃)₂·6H₂O, 99%) were obtained from Sigma-Aldrich. Ethyl alcohol (99.5%) was bought from Shimakyu's Pure Chemicals (Osaka, Japan).

2.2 Preparation of catalysts

2.2.1 Synthesis of WO₃. WO₃ was prepared via the solvothermal method following a literature report.⁴² First, 0.4 g of WCl₆ was dissolved in 80 mL of ethanol. Subsequently, 4 g of oxalic acid was added to the solution, and the mixture was stirred until the solution became transparent. Then, the solution was transferred to a sealed 100 mL Teflon-lined stainless-steel autoclave and placed in an oven at 100 °C for 24 h. After the heating process, the autoclave was cooled naturally to room temperature. The WO₃ powders were then collected by centrifugation and washed three times with deionized (DI) water and absolute ethanol. Subsequently, the collected products were subjected to freeze-drying to prevent particle aggregation. Finally, the products were annealed at 165 °C under atmospheric conditions to enhance their crystallinity.

2.2.2 Synthesis of SA-M/WO₃ (M = Cr, Fe, and Ni). SA-M/WO₃ denotes single-atom-anchored WO₃, where M represents the transition metal that is anchored onto the WO₃ surface. SA-Fe/WO₃ was synthesized using the wet impregnation method.^43,44 Initially, 1 mmol Fe(NO₃)₃·9H₂O and 20 mL of ethanol were added to a flask. The solution was then bubbled with high-purity argon for 30 minutes to remove any dissolved gas. Then, 80 mg of the WO₃ powder was added to the solution and stirred to evenly disperse the powder. The solution was heated to 80 °C in an oil bath to facilitate the anchoring of Fe ions onto the surface of WO₃. For SA-Cr/WO₃ and SA-Ni/WO₃, the procedures were the same, with only the metal precursors changed to Cr(NO₃)₃·9H₂O and Ni(NO₃)₂·6H₂O, respectively.

2.3 Material characterization

The crystal structure of the catalysts was analyzed using a powder X-ray diffractometer (XRD, Bruker, D2 Phaser) with Cu Kα radiation (λ = 0.1540 nm). The catalyst morphologies were studied using scanning electron microscopy (SEM, JEOL, JSM-6500), high-resolution transmission electron microscopy (HRTEM, JEOL JEM-2100F), and scanning transmission electron microscopy (STEM, JEOL, JEM-ARM300F2), operated at 200 kV. The surface areas of catalysts were studied by Brunauer–Emmett–Teller (BET) analysis (BELSORB MAX II). CO₂ adsorption behavior was determined using temperature-programmed desorption (TPD) measurements (AutoChem II 2920 Micrometrics). Light absorption measurements were performed by ultraviolet-visible-near infrared (UV-vis-NIR) spectroscopy (JASCO, V-670) with an integrated sphere using the BaSO₄ powder as the standard reference. Photoluminescence (PL) spectra were measured using a custom-built confocal fiber microscope integrated with a HORIBA iHR550 system, equipped with a Syncerity BI-NIR CCD camera (wavelength range: 400 to 1075 nm) and a diode laser (405 nm) as the excitation source. Time-resolved photoluminescence (TRPL) signals were collected using a photomultiplier detector (PicoQuant, PMA192, wavelength range: 230 to 920 nm). XPS was measured by an X-ray photoelectron spectroscopy system (ULVAC-PHI, PHI 5000 VersaProbe III) using monochromated Al Kα (1486.6 eV) X-ray as the excitation source. The XPS signals were calibrated to the C 1s peak at 284.6 eV. The elemental composition was measured by inductively coupled plasma optical emission spectroscopy (ICP-OES, Agilent 710-ES). X-ray absorption spectroscopy (XAS) data were acquired at the National Synchrotron Radiation Research Center (NSRRC, Hsinchu, Taiwan). The W L₃-edge was measured in the transmission mode at NSRRC TLS BL01 C1. The Cr K-edge, Fe K-edge and Ni K-edge spectra were collected in the fluorescence mode at NSRRC TPS BL44 A. The reaction intermediates formed during the photoreduction process were monitored using in situ FTIR spectroscopy (Bruker Tensor 27).^23,45,46

2.4 Photocatalytic CO₂ reduction reaction

A batch-type photocatalytic CO₂ reduction experiment was conducted to investigate the performance of the photocatalyst. Prior to the photocatalytic CO₂ reduction test, 50 mg of the photocatalyst was placed in a homemade gas-tight quartz reactor with a 30 mL volume and evacuated to remove any air. The reactor was then filled with high-purity CO₂ (99.9999%), which passed through a water bubbler at a constant flow rate of 100 sccm for 30 minutes under atmospheric pressure. A 150 W AM1.5 solar simulator (Xe lamp) with an intensity of 100 mW cm⁻² was used as the light source. After 4 h of light exposure, the photocatalytic products were analyzed by gas chromatography (GC, HP 6890 GC-HID) using a Shin-Carbon ST column for product separation. For the isotope test, ¹²CO₂ was replaced by ¹³CO₂, and the light exposure time was extended to 24 h. The photocatalytic products were analyzed by mass spectrometry (Agilent 7890 CB). The apparent quantum efficiency (AQE) is defined by the ratio of the effective electrons utilized in solar fuel production to the total photons incident on the catalyst. The AQE was calculated using the following equation:where α represents the number of electrons involved in solar fuel formation (α_CO = 2, α_CH4 = 8, and α_H2 = 2), Y is the yield of solar fuel products, N is the Avogadro number, θ is the photon flux, T is the irradiation time, and S is the irradiation area. To calculate the absorbed photon quantum efficiency (APQE), the total photons incident on the catalyst are replaced by the total photons absorbed (λ < λ_g) by the catalyst. In the case of SA-Fe–WO₃, N = 6.022 × 10²³ mol⁻¹, θ = 2.58 × 10¹⁷ s⁻¹ cm⁻² (integrated photon flux from 300 to 1100 nm), T = 4 h, and S = 4 cm². For APQE calculations, θ(λ < λ_g) = 8.86 × 10¹⁵ s⁻¹ cm⁻² (λ_g = 413 nm).

2.5 Theoretical calculations

Spin-polarized DFT calculations were performed using the Vienna ab initio simulation package (VASP).⁴⁷ The projector-augmented wave (PAW) method was employed to describe core-valence interactions.⁴⁸ The exchange–correlation functional was treated within the generalized gradient approximation (GGA) using the Perdew–Burke–Ernzerhof (PBE) scheme.⁴⁹ A 2 × 2 × 4 supercell of WO₃ (cubic phase, space group Pm3̄m) was constructed as the computational model. The plane-wave cutoff energy was set to 600 eV for all models. Valence electron configurations were 5d⁵6s¹ for W, 2s²2p⁴ for O, 3d⁵4s¹ for Cr, 3d⁷4s¹ for Fe, and 3d⁹4s¹ for Ni. A k-point convergence test was performed, leading to the selection of a 4 × 4 × 1 k-point mesh, as shown in Fig. S1. Electronic and ionic relaxations were carried out with convergence criteria of 10⁻⁵ eV for energy and 0.005 eV Å⁻¹ for forces, respectively. To eliminate the interaction between periodic images, a vacuum space (>20 Å) was added in the z-direction.

The formation energy (ΔE_f) was calculated as follows:

where E(anchored WO₃) and E(WO₃) represent the electronic energies of the WO₃ supercell with and without the metal dopant, respectively. n_i denotes the number of atoms added (n_i > 0) or removed (n_i < 0) relative to the WO₃ supercell, and µ_i is the chemical potential of the atom i, obtained from its pure metal phase. The CO₂ adsorption energy was calculated by taking the electronic energy of the system with CO₂ adsorbed and subtracting the combined electronic energy of the pristine surface and the energy of an isolated CO₂ molecule.

The Gibbs free energy was calculated by adding thermal corrections to the electronic energy, shown in eqn (3) (ref. 12 and 50–52) as follows:

G = E_DFT + ZPE + H_(0⃑T) − TS, (3)

where E_DFT is the electronic energy. ZPE, H_(0⃑T), and S represent the zero-point energy, enthalpy contribution, and entropy, respectively, obtained from vibrational frequency calculations. Only the vibrations of the active sites and adsorbates were considered, while the remaining catalyst atoms were assumed negligible and fixed at their optimized positions. T denotes the temperature, which was set to 300 K in all calculations. In the reaction pathway diagram, the free energy changes (ΔG) of intermediates were referenced to the free energies of a gas-phase CO₂ molecule, the clean surface, and half of the free energy of an H₂ molecule multiplied by the number of added hydrogen atoms.¹²

3 Results and discussion

To investigate the impact of the surface atomic structure and electron density distribution on the PC-CO₂RR, wet impregnation methods were employed to anchor the desired transition metals onto the surface of WO₃. This method enables the modification of the WO₃ surface while maintaining its crystalline structure and preserving its bulk properties. Hence, it hinders the PC-CO₂RR performance from being affected by other factors.

XRD and TEM analyses were conducted to confirm the crystal structure of the as-synthesized WO₃ and SA-M–WO₃ (M = Cr, Fe and Ni). As depicted in Fig. 1(a), the XRD patterns reveal that all the diffraction peaks of the as-synthesized WO₃ are consistent with those of cubic-phase WO₃. The TEM images illustrate a cubic nanoplate morphology for WO₃ with an average diameter of 50 nm and thicknesses ranging from 5 to 8 nm, which is beneficial for interlayer charge transfer (Fig. 1(b and c)). After the wet impregnation process, the XRD patterns of SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃ are nearly identical to those of the as-synthesized WO₃. This indicates that no metal or metal oxide clusters are formed, and no metals are intercalated during the surface modification process. TEM and SEM images (Fig. S2 and S3) show no significant morphological changes. To evaluate surface area variations, BET analysis was performed, which revealed similar surface areas across all samples (Fig. S4). The results exhibit a typical type II isothermal curve with a type H3 hysteresis loop, characteristic of plate-like particle aggregation.⁵³ Additionally, UV-vis measurements were conducted to confirm light absorption properties, showing an absorption edge near 450 nm for all samples (Fig. S5).⁵⁴ Power-dependent PL analyses were also carried out to examine the light emission properties (Fig. S6). The results show that the intensity ratios of emission peaks E_A and E_B under different excitation powers (0.25 to 15.25 mW) are similar, implying that all the modified samples have optical properties similar to the pristine WO₃. Based on the above-mentioned observations, it can be concluded that the surface modification introduces new catalytically active sites without affecting the electronic structure, accessible surface area, or morphology of the WO₃ support. To confirm the presence of anchored metals and their distributions, HRTEM, high-angle annular dark-field STEM, and energy-dispersive X-ray spectroscopy (EDS) elemental mapping images were recorded. As shown in Fig. 1(c), the interplanar spacing of SA-Fe–WO₃ nanoplates is 3.9 Å, corresponding to the (100) and (010) planes of cubic WO₃.^55,56 The observation of these mutually perpendicular lattice planes within the nanoplate confirms that the primary exposed surface (the basal plane) is the (001) facet. No additional metals or metal oxides are observed in the HRTEM images. Meanwhile, EDS elemental mapping images show that W, O and Fe are uniformly distributed over the WO₃ surface (Fig. 1(d)), confirming the presence of the atomically dispersed Fe on the WO₃ nanoplates. We observed the same for SA-Cr–WO₃ and SA-Ni–WO₃ (Fig. S7). Additionally, ICP-OES was performed to precisely determine the elemental composition of the materials (Table S1). The results indicate that the metal doping levels (i.e., M/(W + O)) are approximately 1% for all samples.

Fig. 1 Material characterizations. (a) XRD patterns of WO₃, SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃. (b and c) HRTEM images of SA-Fe–WO₃ under different magnifications. (d) HAADF and EDS elemental mapping images of SA-Fe–WO₃.

To identify the chemical information and coordination environment of loaded metals (Cr, Fe and Ni) in the SA-M–WO₃ samples, XAS was employed to measure the K-edge of Cr, Fe and Ni, respectively (Fig. 2(a–c)). The X-ray absorption near-edge spectroscopy (XANES) spectra of all the Cr, Fe, and Ni K-edges exhibit noticeably different spectral curves compared with the standard metal foil reference. The prominent white line intensity in the Cr, Fe, and Ni K-edge spectra indicates that the loaded metals are in oxidized states rather than metallic states, implying the absence of metal clusters on the surface. However, their XANES spectra also slightly deviate from their typical metal oxide spectra. To further confirm the exact coordination environment of the loaded metal, the Fourier transforms of the Cr, Fe and Ni K-edge k⁴-weighted extended X-ray absorption fine structures (EXAFS) were processed to acquire bond length information in the real (R) space. As shown in Fig. 2(d), the EXAFS spectra of SA-Cr–WO₃, SA-Fe–WO₃ and SA-Ni–WO₃ exhibit main peaks at 1.51, 1.52 and 1.57 Å, respectively, which are close to the bond length of the M–O bond in the first shell.^57–59 EXAFS fittings were also processed with the model of an anchored metal bonded to oxygen, which agrees well with the experimental results. The details of EXAFS fitting parameters are shown in Table S2. Moreover, no other extra peaks are found in the range from 2 to 4 Å, indicating that there are no M–M bonds or second shells of M–O–M bonds. This provides evidence that the loaded metals are atomically dispersed, isolated and bonded on the WO₃ surface.

Fig. 2 Structural characterization of transition metal-anchored WO₃ by XAS. XANES spectra (fluorescence mode) of (a) SA-Cr–WO₃, (b) SA-Fe–WO₃, and (c) SA-Ni–WO₃. (d) FT-EXAFS fitting results of transition metal-anchored WO₃. The black line represents the fitting curve, while the brown line represents the standard metal reference.

XPS spectra were recorded to investigate the interaction between the loaded metals and WO₃, study the chemical compositions, and gain insights into the charge transfer behavior. The core-level spectra of the loaded metals were measured to confirm their successful incorporation onto the WO₃ surface (Fig. S8). The peak positions of these metals correspond to their oxide states, indicating bonding with surface oxygen, which is consistent with the XAS results. To further examine charge transfer behavior, the core levels of tungsten and oxygen were analyzed. As depicted in Fig. 3(a), the doublet peaks of W 4f in pristine WO₃ are observed at 37.5 eV (W 4f_5/2) and 35.3 eV (W 4f_7/2). In its O 1s spectra, pristine WO₃ also exhibits two peaks located at 530.7 eV and 532.5 eV, which are attributed to lattice oxygen and –OH, respectively.⁶⁰ It should be noted that the ratio of O to W matches the stoichiometric ratio (Table S3). After metal loading, both W 4f and O 1s peaks shift toward lower binding energies, indicating electron transfer from the loaded metals to WO₃. The shift in XPS signals follows the trend: SA-Fe–WO₃ > SA-Ni–WO₃ > SA–SA-Cr–WO₃ > WO₃, exhibiting a volcano plot behavior. Particularly, SA-Fe/WO₃ exhibits the most significant shift for W 4f from 35.3 to 35.2 eV and for O 1s from 530.7 to 530.5 eV, implying the strongest interaction between Fe atoms and WO₃, with more electrons transferred from Fe atoms to WO₃. Notably, O 1s displays a larger change because it is in the first shell of the doped metals. In addition, XAS measurements were conducted to elucidate the charge transfer behavior within the materials. The intensity of the white line in the L-edge spectrum is an indicator of d-orbital occupancy, i.e., a lower edge intensity indicates higher d-orbital occupancy. The normalized W L₃-edge spectra are presented in Fig. 3(b). It can be observed that the intensity of the W L₃-edge decreases while the metals are loaded onto the WO₃, indicating a decrease in empty orbitals. It implies that introducing metals leads to charge transfer between loaded metals and WO₃. While the loaded metals provide excess electrons to fill the d-orbitals, the excited electrons are not allowed to process bound-state transitions (2p to 5d). Among all the samples, it is observed that SA-Fe–WO₃ exhibits the greatest decrease in white line intensity, indicating that the charge density transfer between Fe and WO₃ is the highest, while that between Cr and WO₃ is the lowest. This result is also consistent with the XPS results. Based on the initial material design principles, doped metal atoms with a higher number of valence electrons are more likely to donate electrons to the substrate material. Meanwhile, the electronegativity of the element influences the degree of electron transfer, as shown in Fig. 3(c and d). Hence, it can be concluded that Fe can be considered the single atom with the most electron density transfer to WO₃.

Fig. 3 Charge transfer analysis with different metal (Cr, Fe, and Ni) modifications. (a) W 4f and O 1s spectra of WO₃, SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃. (b) XANES spectra (transmission mode) at the W L₃-edge for WO₃, Cr–WO₃, Fe–WO₃, and Ni–WO₃. (c and d) Relationship between the charge transfer, valence electrons and electronegativity of W and O elements, respectively. (Electron transfer is calculated from DFT).

To study the impact of different metal loadings on the CO₂RR performance, we conducted a series of PC tests. For these tests, the materials were sealed in a batch-type quartz reactor filled with CO₂ and water vapor, followed by light irradiation for 4 h. As shown in Fig. 4(a), the main products for pristine WO₃ are CO, H₂, and CH₄, with yields of 4.5, 0.97, and 0.2 µmol g⁻¹, respectively. Doping with Cr, Fe, and Ni shows improvements in the product yields. Specifically, with Cr doping, the yields of CO, H₂, and CH₄ increase to 7, 1.3, and 0.7 µmol g⁻¹, respectively. Fe doping results in the highest increase in the yield among all three samples, with CO, H₂, and CH₄ yields reaching 12.1, 3.1, and 0.6 µmol g⁻¹, respectively, which are approximately three times higher than those of pristine WO₃. However, Ni doping leads to a modest increase in the yield, with CO, H₂, and CH₄ production falling to 8.9, 1.9, and 0.6 µmol g⁻¹, respectively. This volcano-like behavior is analogous to the degree of charge transfer observed (Fig. 4(a)). When the charge transfer within the material increases, the photocatalytic CO₂ reduction activity is also enhanced. This similar trend suggests a correlation between charge transfer and photocatalytic CO₂ reduction activity.

Fig. 4 Photocatalytic CO₂ reduction. (a) Average product yields of WO₃, SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃ for an irradiation time of 4 h. W 4f and O 1s peak shifts from the XPS results are also included for the comparison. (b) Cycling stability test for SA-Fe–WO₃.

To determine the optimal catalyst configuration, a series of SA-Fe–WO₃ samples with varying Fe concentrations was systematically evaluated. The results indicate that the optimal loading amount is 1.0%, and the CO₂ reduction efficiency begins to decrease as the concentration is further increased to 1.5% (Fig. S9). This performance decline is attributed to the fact that high loading amounts facilitate the formation of metal clusters, which leads to the loss of the unique charge transfer behavior characteristic of isolated single atoms. Supporting this conclusion, XPS results reveal that the 1.5% sample does not exhibit the significant binding energy shifts found at lower loadings, suggesting a lack of effective electronic interactions between the Fe species and the WO₃ support (Fig. S10).

To test the stability of the materials during the PC-CO₂RR experiment, the samples were recycled after each cycle of photocatalytic CO₂ reduction and then subjected to the same PC test. The results indicate that the samples maintain good stability after five cycles (∼20 h). In addition, a blank test experiment was conducted to confirm that the CO₂RR products were indeed converted from CO₂ and not from surface contamination (Fig. S11). Isotope tracer analyses were also performed to understand the carbon source of products CO and CH₄ (Fig. S12). The overall AQE and APQE reach approximately 0.007% and 0.21%, respectively, which is in the range of the highest productivity plants that have a typical efficiency of about 1% annually.⁶¹

A superior CO₂RR performance is often accompanied by electron localization in the material, which is beneficial to CO₂ adsorption and activation.^13,15,16 Our results show a potential correlation between the charge transfer and CO₂RR activity. To address it in detail, we performed DFT calculations to investigate the impact of anchored Cr, Mn, Fe, Co and Ni atoms on the redistribution of the surface charge density and CO₂ adsorption over the WO₃ surface. According to the XRD and TEM results, the cubic WO₃ surface was cleaved along the (001) direction of bulk WO₃. A 2 × 2 × 4 supercell was then constructed as the computational model. Based on the XAS results, the doped metal coordinates with oxygen atoms, suggesting two possible anchoring configurations for Cr, Fe, and Ni: (I) the metal is positioned directly on an oxygen site (“M on O”) or (II) the metal is surrounded by four oxygen atoms (“M anchored”). Formation energy calculations (Fig. S13) indicate that Cr, Fe, and Ni preferentially adopt the configuration anchored between four oxygen atoms (Fig. S14). The charge density differences were calculated to explore the redistribution of the charge density on the WO₃ surface. As displayed in Fig. 5, it is observed that electron accumulation occurs near the W and O atoms adjacent to the doped metals, i.e., redistributing surface charge density or creating localized electron-rich regions, while electron depletion occurs at the doped metals. To further quantify the amount of transferred electrons from the anchored metals, Bader charge analysis was conducted.⁶² The results clearly show that anchored Cr, Mn, Fe, Co and Ni can transfer 0.44, 1.21, 1.03, 0.84 and 0.73 e⁻ to the neighboring atoms, respectively, thereby enhancing the surface electron density. Specifically, the electron population of the neighboring W atoms increases from 3.45 e⁻ in pristine WO₃ to 3.47, 3.59, 3.58, 3.53 and 3.52 e⁻ for SA-Cr, SA-Mn, SA-Fe, SA-Co and SA-Ni, respectively. A similar trend is observed for the neighboring O atoms, where the electron population rises from 6.99 to 7.07, 7.14, 7.10, 7.11, and 7.08 e⁻, respectively. This quantitative trend, summarized in Fig. S15 and Table S4, aligns perfectly with our XPS and XANES findings. The charge localization induced by Fe can be attributed to the ideal balance between its valence electron count and electronegativity (see Fig. 3(c and d)). While Cr has a relatively low electronegativity, its limited number of valence electrons (3d⁵4s¹) restricts the total amount of transferable charge. Conversely, although Ni possesses the highest valence electron count (3d⁹4s¹), its high electronegativity exerts a stronger pull on its electrons, thereby hindering effective transfer to the WO₃ substrate. Mn (3d⁵4s²) and Fe (3d⁷4s¹) represent a synergistic optimum, providing a sufficient valence electron supply while maintaining a moderate electronegativity, which maximizes the localized surface charge density to promote CO₂ activation.

Fig. 5 Charge density distribution analyses of (a) SA-Cr–WO₃, (b) SA-Fe–WO₃ and (c) SA-Ni–WO₃ for charge densities with interface values of 0.003 e⁻ Bohr⁻³. Yellow and cyan areas denote electron accumulation and depletion, respectively. Top: top view. Bottom: side view.

In addition, the next question would be “how does the position of the metal dopant influence the charge density distribution?”. To address this question, we calculated the charge density distribution with the metals placed at various positions, as illustrated in Fig. S15. The results reveal minimal charge transfer when the metal substitutes at W sites. Furthermore, when the dopants occupy interstitial sites within the WO₃ lattice, the surface W atoms exhibit a tendency to lose electrons. This indicates that the surface assumes an electron-deficient character. Consequently, this suggests that only surface-anchored metals can effectively induce localized charge accumulation on the catalyst. However, substitutional and interstitial site doping are thermodynamically more stable than anchored doping, as shown in Fig. S16. Therefore, to promote metal anchoring on the catalyst surface and enhance kinetic feasibility, we employed a wet impregnation approach instead of in situ doping methods. Furthermore, considering that XPS analysis revealed the presence of surface hydroxyl groups (–OH), we constructed hydroxylated surface models (OH–WO₃ and SA-Fe–OH–WO₃) to evaluate the impact of –OH groups on the electronic environment. As summarized in Table S5, the introduction of surface –OH groups does not significantly alter the charge density of the surrounding W and O atoms. The consistent charge distribution between the pristine and hydroxylated surfaces confirms that the predictive selection principle remains robust under realistic, hydrated reaction conditions.

To elucidate the role of the anchored metals in modulating CO₂ adsorption, we constructed CO₂ adsorption models to calculate the adsorption energy and associated charge transfer. Initially, several possible adsorption structures were optimized to identify the most stable CO₂ adsorption configurations. Their adsorption models and adsorption energies are shown in Fig. S17. The most stable configurations for CO₂ adsorbed on WO₃, SA-Cr–WO₃, SA-Mn–WO₃, SA-Fe–WO₃, SA-Co–WO₃, and SA-Ni–WO₃ are displayed in Fig. 6(a). The results indicate that CO₂ molecules can spontaneously adsorb onto all systems, as evidenced by negative adsorption energies. The transition metal-anchored samples exhibit stronger CO₂ adsorption than the pristine WO₃, as reflected by their more negative adsorption energies. The adsorption energy of CO₂ basically follows the trend of charge transfer: SA-Co–WO₃ ≈ SA-Mn–WO₃ > SA-Fe–WO₃ ≈ SA-Ni–WO₃ > SA-Cr–WO₃ > WO₃, suggesting that the increased localization of electrons enhances CO₂ adsorption. Specifically, CO₂ exhibits weaker interactions on pristine WO₃, with an adsorption energy of −0.33 eV, while SA-Cr–WO₃, SA-Mn–WO₃, SA-Fe–WO₃, SA-Co–WO₃ and SA-Ni–WO₃ show stronger interactions (mixed configuration) with adsorption energies of −0.41, −1.26, −1.01, −1.29 and −1.08 eV, respectively. In addition, SA-Fe–WO₃ exhibits another chemisorption mode of CO₂ (oxygen-down configuration) with an adsorption energy of −0.48 eV. Therefore, it is expected that the SA-Fe–WO₃ system has a higher possibility of adsorbing CO₂ for activation. Additionally, the configuration of adsorbed CO₂ serves as a crucial descriptor for high-performance catalysts, as the bond length and angle of the CO₂ molecule reflect its activation state.⁶³ In the pristine WO₃ system, CO₂ preferentially adsorbs on W and O sites, forming O–W and C–O bonds, respectively. However, the charge transfer from WO₃ to the CO₂ molecule is merely 0.01 electrons, and the CO₂ molecule remains in a nearly linear configuration (178.1°), characteristic of weak physical adsorption. In contrast, in SA-Cr–WO₃, SA-Mn–WO₃, SA-Fe–WO₃, SA-Co–WO₃ and SA-Ni–WO₃ systems, CO₂ preferentially coordinates with W and the anchored metal sites, forming O–W and C–M bonds, respectively, which induce CO₂ bending. The charge transfer from SA-Cr–WO₃, SA-Mn–WO₃, SA-Fe–WO₃, SA-Co–WO₃ and SA-Ni–WO₃ to the CO₂ molecule increases significantly, reaching 0.52, 0.71, 0.69, 0.64 and 0.53 electrons, respectively. Notably, in the oxygen-down configuration of SA-Fe–WO₃, charge transfer can reach up to 0.74 eV. As a result, the C=O bond length elongates from 1.18 Å to 1.26 Å, and the CO₂ molecular angle (∠O–C–O) reduces from 180° to 132.1°. More details are presented in Table S6. These results indicate that charge density localization enhances CO₂ adsorption and activation. To account for more realistic reaction conditions, the influence of surface hydroxylation on CO₂ adsorption was also systematically investigated. As summarized in Table S7, the adsorption energy of CO₂ on OH–WO₃ exhibits only a minor shift from −0.33 eV to −0.40 eV compared to the pristine surface. For the SA-Fe–OH–WO₃ system, although the presence of surface –OH groups leads to a slight decrease in adsorption strength (from −1.01 eV to −0.78 eV), the adsorption energy remains significantly more favorable than that of both pristine and hydroxylated WO₃ supports. These findings confirm that the Fe single-atom sites maintain their superior CO₂ activation capability even in a hydroxylated environment.

Fig. 6 CO₂ adsorption model and desorption behavior. (a) CO₂ adsorption models on WO₃, SA-Cr–WO₃, SA-Mn–WO₃, SA-Fe–WO₃, SA-Co–WO₃ and SA-Ni–WO₃. (b) CO₂-TPD profiles of the transition metal-anchored WO₃.

Furthermore, we confirmed the enhancement in CO₂ adsorption through TPD measurements. As shown in Fig. 6(b), the CO₂-TPD profile of the pristine WO₃ system exhibits a desorption peak near 50 °C, indicating the occurrence of physical adsorption. After anchoring Cr, Fe, and Ni, the desorption peaks shift to higher temperatures, with the SA-Fe–WO₃ system showing a prominent peak near 110 °C. This shift suggests the stronger chemisorption of CO₂ due to interactions with the anchored metal and tungsten. Additionally, a significant increase in the peak area is observed, which can be attributed to the two kinds of chemisorbed CO₂ species in the SA-Fe–WO₃ system. These experimental findings agree well with our simulation results. These results clearly reveal that the anchored metal modulates the surface charge density distribution of WO₃ to promote CO₂ activation. In situ FTIR spectroscopy was employed to elucidate the mechanistic pathway of the metal-anchored system during photocatalytic CO₂ reduction. In situ FTIR spectra show that several new characteristic peaks appear after purging CO₂ and H₂O vapor (Fig. 7(a), S18 and S19). The peak at 1695 cm⁻¹ is assigned to the surface-adsorbed CO₂ (*CO₂), while the peaks at 1634 and 3421 cm⁻¹ are attributed to the surface-adsorbed H₂O (H₂O*).^23,45,46 These results confirm that the essential reactants for the CO₂RR are successfully adsorbed on the catalyst surface prior to light irradiation. The peak intensities of *CO₂ and H₂O* increase during gas purging (Fig. 7(b)). Upon light irradiation, the intensities of the *CO₂ and H₂O* peaks decrease, signifying the consumption of the adsorbed species during the CO₂RR (Fig. 7(c), S18 and S19). This consumption is observed to be time-dependent, as the peak intensities continuously decrease over the irradiation period (Fig. 7(d)). The decreases in the peak intensities of *CO₂ and H₂O* obtained with the SA-Fe–WO₃ sample are 20% and 25% higher than those obtained with the pristine WO₃, highlighting that the enhanced reactivity of the SA-Fe–WO₃ sample could result from Fe anchoring. Notably, a new peak emerges at 1516 cm⁻¹ after light irradiation, corresponding to the *COOH intermediate (Fig. 7(c)).^23,64,65 As noted in these references, this peak position is consistent with prior studies on metal oxide photocatalysts, where the characteristic bands of the *COOH species are typically observed in the 1500–1560 cm⁻¹ region. For example, similar intermediate features have been reported at approximately 1536 cm⁻¹ on Bi₂WO₆ surfaces.²³ The peak intensity of this intermediate is 40% greater on SA-Fe–WO₃ than on pristine WO₃, indicating that Fe anchoring enhances the formation of this key intermediate (Fig. 7(d)). This observation is consistent with our DFT calculations and TPD results. These results collectively demonstrate that anchoring Fe atoms significantly enhances CO₂ activation and facilitates the subsequent formation of key intermediates in the photocatalytic pathway.

Fig. 7 In situ FTIR measurements. (a) In situ FTIR spectra of SA-Fe–WO₃ and (b) variation in IR peak intensities associated with *CO₂ and H₂O*, recorded in CO₂ and H₂O atmospheres under dark conditions. (c) In situ FTIR spectra of SA-Fe–WO₃ and (d) time-dependent variation in IR peak intensities associated with *CO₂, H₂O*, and the *COOH intermediate, recorded in H₂O and CO₂ atmospheres under light irradiation.

To further elucidate the origin of the enhanced photocatalytic activity in metal-anchored systems, Gibbs free energy calculations were conducted for the possible reaction pathways, guided by in situ FTIR observations of key intermediates. The detailed results are summarized in Fig. 8 and Tables S8–S13. Metal anchoring is found to significantly enhance CO₂ adsorption to form *CO₂, decreasing ΔG(*CO₂) from 0.16 eV for pristine WO₃ to 0.05 eV for SA-Cr–WO₃ and to negative values for SA-Fe–WO₃ (−0.53 eV) and SA-Ni–WO₃ (−0.54 eV), indicating a thermodynamically spontaneous process. This can be attributed to the stronger localized charge density introduced by the anchored metals, which effectively promotes CO₂ adsorption. In the subsequent protonation step, the formation of *COOH is identified as the possible rate-determining step (RDS) on pristine WO₃, with a high ΔG(*COOH) of 1.5 eV. In contrast, all metal-anchored systems render *COOH formation exothermic (ΔG < 0), making the reaction thermodynamically favorable. Among them, the Fe-anchored system exhibits the most pronounced effect, with ΔG(*COOH) = −1.73 eV, consistent with the experimentally observed superior photocatalytic activity. SA-Cr–WO₃ and SA-Ni–WO₃ also facilitate COOH* formation, with ΔG(*COOH) of −0.47 eV and −0.55 eV, respectively. These results highlight that the stronger localized charge density not only enhances CO₂ adsorption but also stabilizes the first hydrogenated CO₂ intermediate. Notably, unlike other systems, SA-Fe–WO₃ spontaneously cleaves the C–OH bond in COOH*, forming *CO and *OH. Compared with the subsequent step of converting *COOH to *CO + H₂O, it can be inferred that the ease of *CO formation follows the order of SA-Fe–WO₃ > SA-Ni–WO₃ > SA-Cr–WO₃. SA-Ni–WO₃ exhibits a more favorable ΔG(*CO + H₂O) of −0.83 eV than SA-Cr–WO₃ (−0.03 eV), suggesting that the transformation of *COOH to CO is comparatively less favorable on SA-Cr–WO₃. This explains why the CO₂ conversion efficiency of SA-Ni–WO₃ is higher than that of SA-Cr–WO₃.

Fig. 8 Gibbs free energy calculation. Free energy diagrams for the photoreduction of CO₂ to CO over (a) pristine WO₃, (b) SA-Cr–WO₃, (c) SA-Fe–WO₃, and (d) SA-Ni–WO₃. (e) Key intermediate states involved in the CO₂-to-CO photoreduction pathway on WO₃, SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃. (f) Schematic of the charge transfer over different single atoms and adsorbed CO₂.

To further quantify the impact of charge accumulation on regulating the adsorption of *COOH, the adsorption energy and associated charge transfer were calculated and are summarized in Table S14. While *COOH adsorption is energetically unfavorable on pristine WO₃ with an adsorption energy of 0.51 eV, the introduction of single-atom sites dramatically stabilizes the intermediate, with the adsorption energy reaching −1.35 eV and −1.32 eV for SA-Cr–WO₃ and SA-Ni–WO₃, respectively. Structural parameters reveal that the O–C–OH bond angle of the adsorbed species on SACs decreases significantly to 115°, indicating a high degree of activation from the linear CO₂ geometry. Furthermore, Bader charge analysis confirms a substantial electron transfer to the *COOH moiety, reaching 1.62 electrons and 1.49 electrons for SA-Cr–WO₃ and SA-Ni–WO₃, respectively, which are significantly higher than the 0.47 e⁻ observed for pristine WO₃. This localized electron enrichment at the active site, driven by the synergistic interaction between the single atom and the WO₃ support, effectively weakens the C–OH bond and promotes COOH activation.

To further elucidate the product selectivity of CO, *COH and *CHO are also processed because both are critical intermediates for CH₄ formation. The Gibbs free energies (ΔG) for the formation of *COH on SA-Cr–WO₃ (1.28 eV), SA-Fe–WO₃ (0.65 eV), and SA-Ni–WO₃ (1.04 eV) are significantly higher than those for *CO formation. Similarly, the ΔG values for the formation of *CHO on SA-Cr–WO₃ (−0.21 eV) and SA-Ni–WO₃ (0.38 eV) are also higher than the energies associated with the *CO species formation. These elevated energy barriers indicate that further protonation of adsorbed *CO is thermodynamically unfavorable on the SA-M–WO₃ surfaces. Notably, on the SA-Fe–WO₃ catalyst, the CO–H bond in the potential *CHO intermediates undergoes spontaneous cleavage, which effectively terminates the pathway toward the formation of CH₄. These findings suggest that while the localized charge successfully stabilizes the initial *COOH intermediate, the specific electronic environment of the single-atom sites does not support the subsequent hydrogenation steps required for CH₄ production. Consequently, the catalysts exhibit high selectivity for CO desorption over further reduction because the energy required to reach the CH₄ intermediate is significantly higher.

The superior performance of SA-Fe–WO₃ is attributed to the maximal electron transfer from Fe to the WO₃ framework, as evidenced by the most significant shifts in XPS binding energies and W L₃-edge XANES intensities. Bader charge analysis confirms that Fe induces the highest localized charge accumulation at neighboring atoms compared to Cr and Ni, which stems from the optimal balance between its valence electron count and electronegativity. This enhanced surface charge localization promotes CO₂ activation and significantly stabilizes the key *COOH intermediate by lowering its formation energy. Collectively, these synergistic electronic factors lower the overall reaction barrier, leading to the observed threefold enhancement in photocatalytic CO₂ reduction efficiency.

4 Conclusions

In summary, single-atom-anchored WO₃ photocatalysts (SA-Cr–WO₃, SA-Fe–WO₃, and SA-Ni–WO₃) are successfully synthesized via the wet impregnation method, with the uniform dispersion of metals confirmed by EXAFS analysis. The anchored metals induce a redistribution of charge density, as supported by experiments (XPS and XAS) and DFT calculations. We observed that the anchored metals donate electrons to surrounding atoms, leading to localized charge density accumulation. This results in enhanced CO₂ molecule chemisorption, as demonstrated by TPD and DFT analyses. Moreover, Gibbs free energy calculations and in situ FTIR spectroscopy reveal that *COOH is the key intermediate, and metal anchoring lowers its activation barrier, rendering the reaction thermodynamically spontaneous. Stronger charge-density localization not only enhances CO₂ chemisorption but also promotes CO₂ activation. Consequently, the photocatalytic CO₂ reduction reaction performance is improved by three times with a yield of 12.1 µmol g⁻¹ per 4 h (or APQE = 0.21%) for the CO product. These findings illustrate a strong correlation between charge density localization, CO₂ adsorption, and CO₂ activation and provide valuable insights for designing improved catalyst surfaces.

Author contributions

C.-Y. H. contributed to the idea, investigation, methodology, experiment, visualization, simulation and writing of the original draft. Y.-R. L. and M. H. assisted the simulation. M. E. A. performed the BET experiments. Y.-F. H. contributed to the conceptualization. T.-Y. L., K. L. and H.-L. W. performed the in situ FTIR experiments. K.-H. C. and L.-C. C. acquired the funding. M. Q., K.-H. C. and L.-C. C. supervised the project and reviewed and edited the manuscript.

Conflicts of interest

The authors declare no competing financial interests.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information: Fig. S1–S19 and Tables S1–S14. See DOI: https://doi.org/10.1039/d5ta10187e.

Acknowledgements

This research was financially supported by the National Science and Technology Council (NSTC) in Taiwan under the Academic Summit Project (NSTC 112-2639-M-002-005-ASP, NSTC 113-2639-M-002-004-ASP and NSTC 114-2639-M-002-003-ASP), NSTC 111-2112-M-131-002-MY3 and NSTC 111-2112-M-019-006-MY3, as well as the Center of Atomic Initiative for New Materials (AI-Mat), the National Taiwan University [111 L9008, 112 L9008, 113 L9008 and 114 L9008], from the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) of Taiwan. M.Q. acknowledges support from the National Science and Technology Council (NSTC) Taiwan [grant number: 114-2112-M-003-019-MY2]. In addition, we would like to thank Ms. C.-Y. Chien of the Ministry of Science and Technology (the National Taiwan University) for the assistance in the TEM experiments. The technical support from the Advanced Materials Characterization Lab, the Institute of Atomic and Molecular Sciences (IAMS), Academia Sinica, is also gratefully acknowledged. We also acknowledge the National Synchrotron Radiation Research Center (NSRRC) TLS 01C1 and TPS 44A for the assistance in XAS measurements. We are grateful to the Computer and Information Networking Center, the National Taiwan University for the support of the high-performance computing facilities.

References

1. Q. Yi, W. Li, J. Feng and K. Xie, Chem. Soc. Rev., 2015, 44, 5409–5445.
2. S. Solomon, G.-K. Plattner, R. Knutti and P. Friedlingstein, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 1704–1709.
3. J. Fu, K. Jiang, X. Qiu, J. Yu and M. Liu, Mater. Today, 2020, 32, 222–243.
4. M. Aresta, A. Dibenedetto and A. Angelini, Chem. Rev., 2014, 114, 1709–1742.
5. S. Fang, M. Rahaman, J. Bharti, E. Reisner, M. Robert, G. A. Ozin and Y. H. Hu, Nat. Rev. Methods Primers, 2023, 3, 61.
6. P.-P. Huang, M. Qorbani, Y.-T. Hung, Y.-R. Lai, A. Sabbah, M.-F. Tseng, C.-Y. Huang, S. Koodathil, S. Kholimatussadiah, M. K. Hussien, T.-H. Feng, Y.-H. Liu, H. Wang, J.-W. Lin, C.-H. Wang, C.-I. Wu, M. Hayashi, K.-H. Chen and L.-C. Chen, ACS Nano, 2026, 20, 2211–2224.
7. Z. Sun, T. Ma, H. Tao, Q. Fan and B. Han, Chem, 2017, 3, 560–587.
8. B. Han, X. Ou, Z. Deng, Y. Song, C. Tian, H. Deng, Y. J. Xu and Z. Lin, Angew. Chem., Int. Ed., 2018, 57, 16811–16815.
9. W. Yao, B. Wu and Y. Liu, ACS Nano, 2020, 14, 9320–9346.
10. M. Qorbani, A. Sabbah, Y.-R. Lai, S. Kholimatussadiah, S. Quadir, C.-Y. Huang, I. Shown, Y.-F. Huang, M. Hayashi, K.-H. Chen and L.-C. Chen, Nat. Commun., 2022, 13, 1256.
11. W. Jiang, H. Loh, B. Q. L. Low, H. Zhu, J. Low, J. Z. X. Heng, K. Y. Tang, Z. Li, X. J. Loh and E. Ye, Appl. Catal., B, 2023, 321, 122079.
12. X. Li, Y. Sun, J. Xu, Y. Shao, J. Wu, X. Xu, Y. Pan, H. Ju, J. Zhu and Y. Xie, Nat. Energy, 2019, 4, 690–699.
13. Y. Cao, L. Guo, M. Dan, D. E. Doronkin, C. Han, Z. Rao, Y. Liu, J. Meng, Z. Huang and K. Zheng, Nat. Commun., 2021, 12, 1675.
14. J. Ji, R. Li, H. Zhang, Y. Duan, Q. Liu, H. Wang and Z. Shen, Appl. Catal., B, 2023, 321, 122020.
15. Y. Wang, C. Ban, J. Meng, J. Ma, H. Zou, Y. Feng, J. Ding, Y. Duan, L. Gan and X. Zhou, Sep. Purif. Technol., 2023, 312, 123379.
16. H. M. Ngo, U. Pal, Y. S. Kang and K. M. Ok, ACS Omega, 2023, 8, 8876–8884.
17. X. Jiao, X. Li, X. Jin, Y. Sun, J. Xu, L. Liang, H. Ju, J. Zhu, Y. Pan, W. Yan, Y. Lin and Y. Xie, J. Am. Chem. Soc., 2017, 139, 18044–18051.
18. A. Ñlvarez, M. Borges, J. J. Corral-Pérez, J. G. Olcina, L. Hu, D. Cornu, R. Huang, D. Stoian and A. Urakawa, ChemPhysChem, 2017, 18, 3135–3141.
19. M. Pérez-Jiménez, H. Corona, F. De La Cruz-Martínez and J. Campos, Chem.–Eur. J., 2023, 29, e202301428.
20. A. M. Appel, J. E. Bercaw, A. B. Bocarsly, H. Dobbek, D. L. Dubois, M. Dupuis, J. G. Ferry, E. Fujita, R. Hille, P. J. Kenis, C. A. Kerfeld, R. H. Morris, C. H. F. Peden, A. R. Portis, S. W. Ragsdale, T. B. Rauchfuss, J. N. H. Reek, L. C. Seefeldt, R. K. Thauer and G. L. Waldrop, Chem. Rev., 2013, 113, 6621–6658.
21. X. Jin, Y. Xu, X. Zhou, C. Lv, Q. Huang, G. Chen, H. Xie, T. Ge, J. Cao, J. Zhan and L. Ye, ACS Mater. Lett., 2021, 3, 364–371.
22. M. K. Hussien, A. Sabbah, M. Qorbani, R. Putikam, S. Kholimatussadiah, D. L. M. Tzou, M. H. Elsayed, Y. J. Lu, Y. Y. Wang, X. H. Lee, T.-Y. Lin, N. Q. Thang, H.-L. Wu, S.-C. Haw, K. C.-W. Wu, M.-C. Lin, K.-H. Chen and L.-C. Chen, Small, 2024, 2400724.
23. N. Q. Thang, A. Sabbah, R. Putikam, C. Y. Huang, T. Y. Lin, M. K. Hussien, H. L. Wu, M. C. Lin, C. H. Lee, K. H. Chen and L.-C. Chen, Adv. Funct. Mater., 2025, 2423751.
24. T. T. Mamo, M. Qorbani, A. G. Hailemariam, R. Putikam, C.-M. Chu, T.-R. Ko, A. Sabbah, C.-Y. Huang, S. Kholimatussadiah, T. Billo, M. K. Hussien, S.-Y. Chang, M.-C. Lin, W.-Y. Woon, H.-L. Wu, K.-T. Wong, L.-C. Chen and K.-H. Chen, Nano Energy, 2024, 128, 109863.
25. S.-M. Chang and W.-S. Liu, Appl. Catal., B, 2011, 101, 333–342.
26. H. Pan, M. Sun, X. Wang, M. Zhang, M. Murugananthan and Y. Zhang, Appl. Catal., B, 2022, 307, 121174.
27. J. Li, M. Zhang, Z. Guan, Q. Li, C. He and J. Yang, Appl. Catal., B, 2017, 206, 300–307.
28. Y. Pan, X. Wang, W. Zhang, L. Tang, Z. Mu, C. Liu, B. Tian, M. Fei, Y. Sun and H. Su, Nat. Commun., 2022, 13, 3063.
29. Z. Zhang, C. Feng, D. Wang, S. Zhou, R. Wang, S. Hu, H. Li, M. Zuo, Y. Kong and J. Bao, Nat. Commun., 2022, 13, 2473.
30. Y. Zhang, J. Zhao, H. Wang, B. Xiao, W. Zhang, X. Zhao, T. Lv, M. Thangamuthu, J. Zhang and Y. Guo, Nat. Commun., 2022, 13, 58.
31. W. Tu, Y. Xu, J. Wang, B. Zhang, T. Zhou, S. Yin, S. Wu, C. Li, Y. Huang, Y. Zhou, Z. Zou, J. Robertson, M. Kraft and R. Xu, ACS Sustain. Chem. Eng., 2017, 5, 7260–7268.
32. Y. An, H. L. Li, Y. Y. Liu, B. B. Huang, Q. L. Sun, Y. Dai, X. Y. Qin and X. Y. Zhang, J. Solid State Chem., 2016, 233, 194–198.
33. J. Yang, W. H. Li, K. Xu, S. Tan, D. Wang and Y. Li, Angew Chem. Int. Ed. Engl., 2022, 61, e202200366.
34. G. Wei, Z. Mao, L. Liu, T. Hao, L. Zhu, S. Xu, X. Wang and S. Tang, ACS Appl. Mater. Interfaces, 2024, 16, 52233–52243.
35. Y. Q. Zhang, D. Wang, G. P. Wei, B. L. Li, Z. C. Mao, S. M. Xu, S. B. Tang, J. Jiang, Z. Y. Li, X. J. Wang and X. Xu, JACS Au, 2024, 4, 1509–1520.
36. B. Hu, Z. Li, B. Wang, L. Chen, X. Wang, X. Hu, Z. Bai, Y. Li, G. Chen, X. Luo and S.-F. Yin, Appl. Catal., B, 2025, 371, 125196.
37. H. Shi, H. Wang, Y. Zhou, J. Li, P. Zhai, X. Li, G. G. Gurzadyan, J. Hou, H. Yang and X. Guo, Angew. Chem., 2022, 134, e202208904.
38. P. Dong, G. Hou, X. Xi, R. Shao and F. Dong, Environ. Sci.: Nano, 2017, 4, 539–557.
39. V. Dutta, S. Sharma, P. Raizada, V. K. Thakur, A. a. P. Khan, V. Saini, A. M. Asiri and P. Singh, J. Environ. Chem. Eng., 2021, 9, 105018.
40. X. Zhang, H. Su, P. Cui, Y. Cao, Z. Teng, Q. Zhang, Y. Wang, Y. Feng, R. Feng and J. Hou, Nat. Commun., 2023, 14, 7115.
41. S. H. Talib, S. Hussain, S. Baskaran, Z. Lu and J. Li, ACS Catal., 2020, 10, 11951–11961.
42. L. Liang, X. Li, Y. Sun, Y. Tan, X. Jiao, H. Ju, Z. Qi, J. Zhu and Y. Xie, Joule, 2018, 2, 1004–1016.
43. J. Guo, H. Liu, D. Li, J. Wang, X. Djitcheu, D. He and Q. Zhang, RSC Adv., 2022, 12, 9373–9394.
44. H. Zhang, Y. Wang, S. Zuo, W. Zhou, J. Zhang and X. W. D. Lou, J. Am. Chem. Soc., 2021, 143, 2173–2177.
45. A. Sabbah, I. Shown, M. Qorbani, F.-Y. Fu, T.-Y. Lin, H.-L. Wu, P.-W. Chung, C.-I. Wu, S. R. M. Santiago, J.-L. Shen, K.-H. Chen and L.-C. Chen, Nano Energy, 2022, 93, 106809.
46. F.-Y. Fu, C.-C. Fan, M. Qorbani, C.-Y. Huang, P.-C. Kuo, J.-S. Hwang, G.-J. Shu, S.-M. Chang, H.-L. Wu, C.-I. Wu, K.-H. Chen and L.-C. Chen, Sustainable Energy Fuels, 2022, 6, 4418–4428.
47. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169.
48. G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 1758.
49. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865.
50. J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard and H. Jonsson, J. Phys. Chem. B, 2004, 108, 17886–17892.
51. L. I. Bendavid and E. A. Carter, J. Phys. Chem. C, 2013, 117, 26048–26059.
52. V. Wang, N. Xu, J.-C. Liu, G. Tang and W.-T. Geng, Comput. Phys. Commun., 2021, 267, 108033.
53. M. Thommes, K. Kaneko, A. V. Neimark, J. P. Olivier, F. Rodriguez-Reinoso, J. Rouquerol and K. S. Sing, Pure Appl. Chem., 2015, 87, 1051–1069.
54. J. Klein, L. Kampermann, B. Mockenhaupt, M. Behrens, J. Strunk and G. Bacher, Adv. Funct. Mater., 2023, 33, 2304523.
55. H. Yang, H. Sun, Q. Li, P. Li, K. Song, B. Song and L. Wang, Vacuum, 2019, 164, 411–420.
56. M. Gillet, K. Aguir, C. Lemire, E. Gillet and K. Schierbaum, Thin Solid Films, 2004, 467, 239–246.
57. A. Boubnov, A. Roppertz, M. D. Kundrat, S. Mangold, B. Reznik, C. R. Jacob, S. Kureti and J.-D. Grunwaldt, Appl. Surf. Sci., 2016, 386, 234–246.
58. K. Pandya, W. O'grady, D. Corrigan, J. Mcbreen and R. Hoffman, J. Phys. Chem., 1990, 94, 21–26.
59. C. Zhong, Y. Xu, X. Wu, S. Yin, X. Zhang, L. Zhou and H. You, Adv. Mater., 2024, 36, 2309500.
60. T. J. Frankcombe and Y. Liu, Chem. Mater., 2023, 35, 5468–5474.
61. D. K. Dogutan and D. G. Nocera, Accounts Chem. Res., 2019, 52, 3143–3148.
62. W. Tang, E. Sanville and G. Henkelman, J. Phys.: Condens. Matter, 2009, 21, 084204.
63. B. M. Abraham, O. Piqué, M. A. Khan, F. Viñes, F. Illas and J. K. Singh, ACS Appl. Mater. Interfaces, 2023, 15, 30117–30126.
64. X. Li, Y. Li, P. Li, D. Yuan, Y. Fu, J. Huang, Y. Fang, F. Wang, M. Anpo and L. Tan, Mol. Catal., 2025, 585, 115356.
65. J. D. Yi, R. Xie, Z. L. Xie, G. L. Chai, T. F. Liu, R. P. Chen, Y. B. Huang and R. Cao, Angew. Chem., 2020, 132, 23849–23856.

---