Regulating Electronic Structure of Transition Metal Single‐Atoms in COFs for Enhanced Photocatalytic CO₂ Reduction

ABSTRACT

The rational regulation of the electronic structure in single‐atom catalysts (SACs) is pivotal yet challenging for enhancing photocatalytic CO₂ reduction. Herein, we elaborately designed a series of M₁N₂ sites (M = Au, Pt, Pd, Ru, Mo) anchored on a vinylene‐linked covalent organic framework (sp²c‐COF) to construct M/COF SACs for gas‐solid CO₂ photoreduction. The M/COF SACs revealed a d‐orbital electronic configuration‐dependent activity, where the d‐band center exhibiting strong correlation with CO₂ adsorption energy (R² = 0.98). Notably, the Mo/COF catalysts delivered a superior CO rate of 294.43 µmol·g⁻¹·h⁻¹ with near‐unity selectivity under pure CO₂ and sustained 146.6 µmol·g⁻¹·h⁻¹ under simulated flue gas (15% CO₂). The superior activity originates from the synergistic interplay of its highest d‐band center (−0.314 eV) and strongest spin polarization among the series. This unique electronic structure, featuring abundant single‐atom states near the Fermi level and half‐occupied d orbitals, facilitates optimal σ‐donation (via d_z2) and π‐back‐donation (via d_xz/d_yz) for CO₂ activation, thereby significantly lowering the energy barriers for *COOH formation and *CO desorption. This work establishes a design principle for high‐performance SACs through the co‐modulation of d‐band configuration and spin polarization.

This work develops distorted M₁N₂ sites (M = Au, Pt, Pd, Ru, Mo) on a vinylene‐linked COF for CO₂ photoreduction. The severely distorted N─Mo─N configuration exhibits an upshifted d‐band center and strong spin polarization, synergistically enhancing CO₂ activation and driving stoichiometric CO and O₂ production, demonstrating an effective electronic‐structure co‐modulation strategy for efficient photocatalytic CO₂ reduction.

1. Introduction

Photocatalytic CO₂ reduction offers a solar‐driven route to convert a major greenhouse gas into value‐added chemicals, aligning directly with the mitigation of climate change and carbon‐neutrality goals [1, 2]. However, CO₂ is thermodynamically stable and kinetically inert. The reaction proceeds via multi‐step proton‐coupled electron transfer (PCET) and competing pathways, which together demand high electron density at well‐defined active sites and precise control over the intermediate energetics [3, 4]. Although versatile organic and inorganic semiconductors have been extensively explored, most state‐of‐the‐art photocatalysts still suffer from the sluggish charge transfer kinetics and insufficient driving force for CO₂ activation [5, 6].

Notably, the single‐atom catalysts (SACs) have emerged as powerful platforms for CO₂ photoreduction owing to their maximal atom utilization, distinctive electronic structure, and tunable coordination microenvironment [7, 8]. Particularly, Cu‐based SACs are attractive given their low cost, multivalent chemistry, flexible coordination, and compatibility with diverse supports [9, 10]. Although numerous studies report strong CO₂ adsorption on isolated Cu sites, the fully occupied 3d¹⁰ configuration intrinsically constrains its charge transfer with the front orbitals of CO₂ and thereby suppresses the molecular activation and catalytic conversion. To overcome these electronic‐structure limitations, two promising strategies can guide the fine design of metallic SACs for outstanding CO₂‐activation‐conversion capability by precisely modulating their d‐orbital configurations: (i) Heavier late‐transition metals below Cu in the periodic table (e.g., Pd, Pt, and Au). These metals possess larger atomic radii, inducing more diffuse d orbitals and weaker effective nuclear attraction to valence electrons, favoring electron migration into the CO₂ π* levels and strengthening back‐donation activation. (ii) Early‐transition metals (e.g., Ru, Mo). They offer partially filled, energetically accessible d states, enabling stronger synergistic σ‐donation/π‐back‐donation and potentially higher spin polarization for promoting CO₂ activation [11]. The d‑band center theory has established a fundamental framework for understanding and predicting the catalytic activity of transition metals, serving as a widely adopted descriptor for the adsorption strength of reaction intermediates [12]. Building upon this foundation, spin polarization is proposed as an additional tuning parameter that can further enhance catalytic performance through a synergistic mechanism. While the d‑band center primarily governs the thermodynamic binding energy of key intermediates, spin polarization provides complementary control over electron spin alignment and spin‑dependent charge transfer. Within this synergy, the d‑band center ensures favorable adsorption energetics, whereas spin polarization may facilitate CO₂ activation by lowering the electron pairing energy barrier and promoting spin‑allowed electron donation and back‑donation [13, 14]. Together, these two regulatory factors could enable a more comprehensive electronic optimization, potentially achieving both favorable adsorption thermodynamics and efficient spin‑selective charge transfer kinetics. It is therefore anticipated that the synergistic combination of d‑band center modulation and spin polarization may lead to superior CO₂ activation performance compared with modulating the d‑band center alone. Nevertheless, the general structure‐reactivity relationships across different metals, especially the synergistic influence of intrinsic d‐band position and spin state under identical coordination microenvironments on CO₂ activation and conversion, remain poorly understood and warrant systematic investigation.

Olefin‐linked, sp²‐carbon‐conjugated covalent organic frameworks (sp²c‐COFs) incorporating pyrene and 2,2′‐bipyridine units provide an ideal platform to construct SACs with an identical coordination configuration. Unlike imine‐linked COFs formed via reversible condensations, the C═C‐bonded frameworks offer superior chemical stability and enhanced crystallinity [15, 16]. Pyrene acts as a strong electron‐donating chromophore, broadening visible‐light harvesting and promoting exciton dissociation. Meanwhile, the 2,2′‐bipyridine motifs furnish a uniform bidentate chelation sites, enabling the precise anchoring of M₁N₂ single‐atom centers [17]. This system thus serves as a model platform to systematically probe the intrinsic influence of metallic d‐band position and spin state on CO₂ activation and conversion under an identical coordination structure.

Herein, a series of metallic single‐atom catalysts (M/COF; M = Au, Pt, Pd, Ru, Mo) with identical M₁N₂ coordination were synthesized via impregnation on a vinylene‐linked COF for gas‐solid CO₂ reduction under visible light (Scheme 1). Among them, Mo/COF delivered a CO rate of 294.43 µmol·g⁻¹·h⁻¹ with stoichiometric O₂ evolution (128.63 µmol·g⁻¹·h⁻¹) under pure CO₂, evidencing the robust activity and stability. The carrier dynamics in M/COF were clarified by serial characterizations such as temperature‐dependent photoluminescence, Hall measurements, and femtosecond transient absorption (fs‐TAS). The active sites and reaction pathways for CO₂ reduction and water oxidation were probed by in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) in combination with thermodynamic analysis. Furthermore, detailed density functional theory (DFT) calculations, like projected density of states (PDOS) analysis, were utilized to reveal the interplay between d‐band center and the spin state of the M₁N₂ sites, elucidating the underlying mechanism for the distinct CO₂ activation and photoreduction behaviors. The superior activity of Mo/COF was attributed to the optimized electronic configuration within its distorted linear N‐Mo‐N coordination geometry, which facilitates synergistic σ‐donation and π‐back‐donation between the Mo d‐orbitals (d_z2, d_xz/d_yz) and CO₂ molecular orbitals. This work highlights the decisive roles of d‐band position and spin state in CO₂ activation and provides a guiding framework for the rational design of high‐performance SACs for photocatalytic CO₂ conversion.

SCHEME 1.

(a) Synthetic scheme for sp²c‐COF and M/COF. (b) Schematic diagram of the bonding of CO₂ to the Mo/COF. (c) Enhanced photocatalytic simulated flue gas reduction mechanism.

2. Results and Discussion

2.1. Synthesis and Structural Characterization

The elaborately designed synthesis strategy of pristine sp²c‐COF and its metal‐loaded counterparts was schematically, where the pristine sp²c‐COF was obtained through an alkali‐catalyzed Knoevenagel condensation between 1,3,6,8‐tetrakis (4‐formylphenyl) pyrene (TFPPy) and 2,2′‐([2,2′‐bipyridine]‐5,5′‐diyl) diacetonitrile (BPYDAN). Various metal ions, e.g., Au, Pt, Pd, Ru, and Mo, were subsequently introduced into the sp²c‐COF framework by a post‐impregnation method, yielding the corresponding metal single‐atom catalysts (M/COF). The Powder X‐ray diffraction (PXRD) patterns of M/COF revealed the well‐defined diffraction peaks, consistent with pristine COF, at 2.70°, 4.74°, 5.46°, 7.32°, and 25.84°, corresponding to the (110), (020), (220), (040), and (001) planes, respectively, verifying their high crystallinity (Figure S1) [18]. The simulated pattern from the AA stacking model agrees well with both the Pawley refinement results and the experimental PXRD pattern of sp²c‑COF (Figure S2). The planar structures of the TFPPy and BPYDAN units promoted strong π‐π interactions between adjacent layers, accounting for the high crystallinity of both sp²c‐COF and M/COF materials. Noticeably, the characteristic C─H stretching peak of aldehyde in TFPPy at 2718 cm⁻¹ in the Fourier transform infrared (FTIR) spectrum disappeared after the Knoevenagel condensation, while an intensified C═C stretching vibration at 1600 cm⁻¹ was observed in all catalysts (Figure S3), evidencing the formation of alkene linkages [19]. Moreover, a distinct band at 2215 cm⁻¹, attributable to the C≡N stretching of the vinyl group, was present in both sp²c‐COF and M/COF, collectively verifying the successful formation of sp²c‐COF [20]. These findings were further supported by solid‐state ¹³C nuclear magnetic resonance (NMR) spectroscopy (Figure S4), where a peak at 107.34 ppm was assigned to the vinyl nitrile carbon [21]. Collectively, these results confirmed that the structural integrity of the sp²c‐COF framework was preserved after the incorporation of single metal atoms.

The morphology of the as‐synthesized catalysts was examined by scanning electron microscopy (SEM) and transmission electron microscopy (TEM), which showed that the fibrous structure of sp²c‐COF was preserved after metal incorporation (Figure S5), and no detectable metal nanoparticles or clusters were observed in M/COF (Figure 1a; Figure S6). Aberration‐corrected high‐angle annular dark‐field scanning transmission electron microscopy (HAADF‐STEM) revealed bright atomic‐scale spots, attributable to isolated metal atoms uniformly dispersed within the sp²c‐COF framework (Figure 1b; Figure S7). The energy dispersive X‐ray (EDX) mapping also confirmed the homogeneous distributions of C, N, and the respective metal species (Au, Pt, Pd, Ru, Mo) across the frameworks (Figure S6). The metal loadings of Au, Pt, Pd, Ru, and Mo were determined to be 0.30, 0.26, 0.27, 0.25, and 0.28 wt%, respectively, via inductively coupled plasma optical emission spectrometry (ICP‐OES) analysis. N₂ adsorption‐desorption measurements at 77 K showed reduced surface areas and pore volumes for M/COF compared with the pristine COF (Figure S8), owing to partial occupation of the internal cavities by single atoms (Table S1), consistent with previous reports on single‐atom coordination catalysts [22].

FIGURE 1.

(a) TEM and (b) HAADF‐TEM image of Mo/COF. (c) Mo 3d spectra of Mo/COF under dark and light illumination. (d) FT‐EXAFS spectra over Mo foil, MoO₃, and Mo/COF. (e) 3D wavelet transform EXAFS contour plots for Mo/COF. (f) Schematic model of Mo/COF.

The X‐ray photoelectron spectroscopy (XPS) was employed to probe the chemical composition and electronic states of M/COF (Figure S9; Figure 1c). Au 4f_5/2 (89.00 eV) and Au 4f_7/2 (85.50 eV) binding energies lay between Au³⁺ and Au⁰, indicating the presence of Au^α+ (0 < α< 3) [23, 24]. Similar features were observed for Pt, Pd, and Ru. Specifically, the Pt 4f spectrum of Pt/COF could be deconvoluted into two peaks at 76.40 and 73.30 eV, assigned to Pt 4f_5/2 and Pt 4f_7/2, respectively, and located between Pt⁰ and Pt²⁺, suggesting the presence of partially oxidized Pt^β+ (0 < β < 2) [25]. Pd 3d_5/2 and Pd 3d_3/2 peaks were detected at 343.20 and 337.70 eV, respectively [26, 27]. Ru 3p_1/2 and Ru 3p_3/2 peaks appeared at 484.20 and 463.80 eV, confirming Ru^γ+ species (0 < γ < 3) [28]. The Mo 3d spectrum of Mo/COF exhibited two peaks at 235.42 and 232.29 eV, corresponding to Mo 3d_3/2 and Mo 3d_5/2, respectively (Figure 1c) [29]. These results collectively attested that the metals existed in oxidized states rather than in metallic form. The N 1s spectrum of the pristine sp²c‐COF exhibited two peaks at 399.90 (cyano N) and 398.93 eV (pyridine N) [30, 31]. Upon the metal incorporation, the pyridine N peak shifted to higher binding energies by 0.29 (Au/COF), 0.14 (Pt/COF), 0.11 (Pd/COF), 0.15 (Ru/COF), and 0.18 eV (Mo/COF), signifying the formation of M─N bonds (Figure S9).

The coordination environment of Mo₁N₂ was investigated by X‐ray absorption fine structure (XAFS). As shown in Figure S10a the absorption edge of Mo/COF was close to that of MoO₃, indicating a valence state of approximately +6 [29]. Compared with MoO₃, the pre‐edge spectrum of Mo/COF shifted toward lower energies, which can be attributed to Mo─N coordination. Since the electronegativity of N was smaller than that of O, the lone pairs on N atoms could act as the Lewis basic sites to stabilize Mo atoms. The Fourier‐transformed k³‐weighted extended X‐ray absorption fine structure (FT‐EXAFS) spectrum of Mo/COF exhibited a peak near 1.88 Å, corresponding to Mo─N/C bonds (Figure 1d). Importantly, the absence of Mo‐Mo scattering paths at 2.43 Å ruled out the formation of metallic clusters or nanoparticles, confirming the exclusive presence of atomically dispersed Mo species [32]. To gain quantitative insight into the local coordination environment, detailed EXAFS curve fitting was performed (Figure S10b and Table S2). The fitting results showed that each Mo atom was coordinated by two N atoms with an average bond distance of 1.88 Å, forming a well‐defined Mo₁N₂ coordination structure. This configuration was further corroborated by wavelet transform (WT) analysis (Figure 1e), which offered improved resolution in both R‐ and k‐space. The WT contour map displayed a single intensity maximum at 5.5 Å⁻¹ (K‐space) and 1.88 Å (R‐space), characteristic of Mo₁N₂ coordination. In contrast, the typical WT signals of metallic Mo─Mo (8.34 Å⁻¹, 2.43 Å) and Mo─O (4.25 Å⁻¹, 3.03 Å) bonds were absent in Mo/COF (Figure S10c,d), providing definitive evidence for isolated Mo atoms without metallic or oxide phase contamination. Notably, while classical coordination chemistry dictates that highly unsaturated, low‐coordinate metal centers are thermodynamically unstable and prone to binding additional ligands (e.g., solvent molecules or residual precursor anions) to achieve electronic saturation [33], the comprehensive spectroscopic data unambiguously confirm a bare, the M₁N₂ configuration without extra axial coordination. The complete removal of precursor‐derived anions (such as Cl⁻) and physically adsorbed solvent molecules was ensured by exhaustive washing protocols and rigorous vacuum drying (60°C, 24 h) during synthesis, which is corroborated by the absence of heteroatom signals in XPS survey spectra and EDX mapping (Figures S6 and S11). Collectively, these results unambiguously confirm the successful formation of atomically dispersed Mo₁N₂ sites within the COF framework, with precise coordination geometry and distinct electronic structure (Figure 1f).

The photoelectrochemical properties of the as‐synthesized catalysts were systematically evaluated. As shown in Figure S12a, the UV–vis diffuse reflectance spectrum (UV–vis DRS) of the pristine sp²c‐COF displayed a characteristic absorption edge at 629 nm, corresponding to its intrinsic π‐π* transitions within the conjugated framework. The incorporation of single‐atom metal sites induced a pronounced bathochromic shift and narrowed the optical band gap owing to metal‐ligand charge transfer (MLCT) and strengthened π‐conjugation [34]. Tauc plots revealed progressively decreasing E_g values of 2.14 eV (COF), 2.09 (Au/COF), 2.07 (Pt/COF), 2.08 (Pd/COF), 2.08 (Ru/COF), and 2.05 (Mo/COF) (Figure S12b), with Mo/COF exhibiting the largest reduction, consistent with strong Mo‐N coordination. The conduction‐band (CB) potentials of COF, Au/COF, Pt/COF, Pd/COF, Ru/COF, and Mo/COF were determined to be −0.68, −0.90, −0.95, −0.99, −1.06, and −1.11 V versus NHE (Figure S13), respectively, via Mott–Schottky measurement, showing a gradual negative shift with metal incorporation [35]. Based upon E_g = E_VB—E_CB, the corresponding valence‐band (VB) positions were calculated as 1.46, 1.19, 1.12, 1.09, 1.02, and 0.94 V versus NHE (Figure S14). The increasingly negative CB (up to −1.11 V for Mo/COF) exceeded the CO₂/CO potential (−0.52 V vs. NHE at pH = 7) [36], while the VB remains sufficiently positive for water oxidation (+0.82 V vs. NHE), indicating thermodynamically favorable band alignment; together with enhanced light absorption, this rendered the M/COFs promising photocatalysts for CO₂ reduction to value‐added chemicals.

2.2. Charge Transfer Dynamics

The electron paramagnetic resonance (EPR) spectroscopy was utilized to probe the electronic states of pristine and metal‐incorporated COFs under dark conditions and visible‐light illumination (λ ≥ 420 nm). The pristine COF showed a strong signal at g = 2.005, attributing to the unpaired electron spin states in spin‐coupled electron pairs arising from the push‐pull units (Figure S15a) [37]. Under illumination, all M/COFs showed substantially increased EPR signals compared to their dark states. The most intense signal was observed for Mo/COF, which was consistent with its superior photoinduced charge separation efficiency (Figure S15b). Photoelectrochemical measurements further corroborated these findings: Mo/COF delivered the highest photocurrent under irradiation (Figure S16a), while electrochemical impedance spectroscopy (EIS) revealed a substantially smaller Nyquist semicircle for Mo/COF than other catalysts (Figure S16b), indicative of reduced charge‐transfer resistance and improved charge transport [38]. As demonstrated in Figure S17, the Mo/COF displayed the lowest photoluminescence intensity among all samples, manifesting the most effective suppression of radiative recombination. The time‐resolved photoluminescence (TRPL) was adopted to quantify the excited‐state lifetimes (Figure 2a; Figure S18). All M/COFs possessed longer average lifetimes (τ_avg) than the pristine sp²c‐COF (0.73 ns): 0.92 ns (Au/COF), 1.16 ns (Pt/COF), 2.92 ns (Pd/COF), 3.13 ns (Ru/COF). Notably, Mo/COF showed the longest τ_avg (3.72 ns), ∼5‐fold higher than the pristine COF, implying markedly suppressed charge recombination in the Mo/COF system. Furthermore, the exciton binding energy (E_b) was examined via the temperature‐dependent PL (10‐300 K) using Arrhenius fits. As the temperature increased, the PL intensity decreased owing to more thermally generated free carriers (Figure 2b; Figure S19). The Mo/COF presented a lower E_b (30.07 meV) than sp²c‐COF (37.34 meV), implying more efficient exciton dissociation into free carriers that can subsequently participate in photocatalytic reactions [39]. The carrier transport kinetics was further evaluated by Hall measurements (Figure 2c). Notably, Mo/COF showed obviously reduced resistivity (0.0154 Ω·cm⁻²) than the pristine sp²c‐COF (0.0380 Ω·cm⁻²), consistent with the EIS results. The carrier concentration and mobility (n = 1.18 × 10¹⁷ cm⁻³, µ = 0.641 cm²·V⁻¹·s⁻¹) of Mo/COF both exceeded those of COF (n = 4.05 × 10¹⁶ cm⁻³, µ = 0.406 cm²·V⁻¹·s⁻¹), underscoring that high‐valence Mo single atoms acted as electron‐accepting centers to facilitate charge migration. Furthermore, the light‐induced shifts in the binding energies of the N 1s (toward higher energy) and Mo 3d (toward lower energy) provided direct evidence for this charge‐transfer pathway (Figure 1c; Figure S9i).

FIGURE 2.

(a) TRPL spectra of COF and Mo/COF. (b) Temperature‐dependent PL and extracted exciton binding energy of Mo/COF. (c) Resistivity, charge mobility, and carrier concentration of COF and Mo/COF measured at 300 K based on the Hall effect. (d) The 3D fs‐TAs, (e) transient fs‐T As, (f) TAs kinetic plots, and typical fitting curves of Mo/COF. Calculated (g) HOMO and (h) LUMO distribution, and (i) distributions of the photo‐excited electrons and holes in Mo/COF.

To interrogate the underlying charge‐separation dynamics on the sub‐picosecond timescale, femtosecond transient absorption spectroscopy (fs‐TAs) was employed. As presented in Figure 2d and Figure S20a, sp²c‐COF and Mo/COF exhibited discernible 3D fs‐TAs of excited state absorption (ESA) or photoinduced absorption (PIA) signals across the full probe wavelength range (400–720 nm). In comparison to pure sp²c‐COF, Mo/COF manifested a pronounced negative absorption band in the vicinity of 500 nm, which was ascribed to its ground state bleaching (GSB) signal [40]. The new negative absorption band represented the trapping process of the hole in the surface trap. Besides, Figure 2e and Figure S20b illustrated the fs‐TAs attenuation at varying probe delays, spanning from 10 to 2000 ps. During this time period, the absorption bands of naked sp²c‐COF and Mo/COF gradually decreased at 700 and 580 nm, respectively, which could be assigned to the capture process of light‐induced electrons in the surface trap [41]. Following 300 ps, the carrier spectral properties of Mo/COF underwent a comparatively gradual alteration in comparison to those of pure sp²c‐COF, indicating that the recombination of trapped holes with deeply trapped electrons occurred at a relatively slow rate. The reduced mobility of electrons upon relaxing into deeper positions resulted in a decreased probability of recombination with trapped holes. TAs could be divided into three distinct components through a process of global fitting (Figure 2f; Figure S20c and Table S3). τ₁, τ₂, and τ₃ were ascribed to the cooling of hot electrons to the conduction band bottom of sp²c‐COF, the annihilation of band gap excitons, and the formation of a shallow electron trapping state, respectively [42]. Critically, Mo/COF exhibited a substantially elongated τ₃ (481.1 ps) compared with the pristine COF (152.5 ps), and its overall mean lifetime (324.8 ps) exceeded that of the parent framework (50.6 ps) by nearly one order of magnitude. Consequently, the prolonged, shallow electron capture and long‐lived carrier separation state afforded a superior opportunity for photocatalyzed CO₂ reduction reactions.

To investigate the intrinsic mechanism of the enhanced charge‐separation efficiency, the charge‐density distribution of the highest occupied (HOMO) and lowest unoccupied (LUMO) molecular orbitals were computed (Figure 2g,h; Figure S21). In the pristine sp²c‐COF, HOMO and LUMO largely overlapped and delocalized across the whole framework, resulting in the rapid recombination. Upon incorporation of single atoms, the orbitals became spatially separated: the HOMO resided mainly on the pyrene unit, while the LUMO was concentrated on the bipyridine unit and the coordinated metal center [43]. This spatial separation enabled directional transfer of photoexcited electrons to the LUMO, effectively suppressing recombination. Furthermore, the incorporation of metal atoms markedly reduced the work function of the pristine sp²c‐COF. Compared with sp²c‐COF (5.018 eV), Au/COF (3.963 eV), Pt/COF (3.782 eV), Pd/COF (3.716 eV), Ru/COF (3.828 eV), and Mo/COF (3.798 eV) all exhibited lower values, thereby facilitating rapid electron migration to surface reaction sites (Figure S22) [44]. The substantially larger excited‐state electron‐hole centroid distance of Mo/COF (2.63 Å) relative to sp²c‐COF (0.89 Å), Au/COF (1.06 Å), Pt/COF (1.78 Å), Pd/COF (1.81 Å), and Ru/COF (1.88 Å) further underscored its superior capability for promoting photogenerated charge separation and transport (Figure 2i; Figure S23) [45].

2.3. CO₂ Photoreduction Performance

The visible‐light‐driven CO₂ reduction was conducted without photosensitizers or sacrificial agents (Figure 3a; Table S4). The pristine sp²c‐COF afforded CO and H₂ yields of 35.8 and 4.1 µmol·g⁻¹·h⁻¹, respectively. In contrast, M/COF samples showed markedly enhanced activity, following the order Mo/COF (CO: 294.43 µmol·g⁻¹·h⁻¹; H₂: 6.8 µmol·g⁻¹·h⁻¹) > Ru/COF (CO: 202.50 µmol·g⁻¹·h⁻¹; H₂: 15.8 µmol·g⁻¹·h⁻¹) > Co/COF (CO: 189.55 µmol·g⁻¹·h⁻¹; H₂: 6.24 µmol·g⁻¹·h⁻¹) > Pd/COF (CO: 174.60 µmol·g⁻¹·h⁻¹; H₂: 13.7 µmol·g⁻¹·h⁻¹) > Ni/COF (CO: 121.36 µmol·g⁻¹·h⁻¹; H₂: 9.50 µmol·g⁻¹·h⁻¹) > Pt/COF (CO: 107.64 µmol·g⁻¹·h⁻¹; H₂: 9.2 µmol·g⁻¹·h⁻¹) > Au/COF (CO: 69.75 µmol·g⁻¹·h⁻¹; H₂: 8.0 µmol·g⁻¹·h⁻¹). The CO selectivity reached 98% for Mo/COF and 90% for Au/COF, both exceeding that of the pristine sp²c‐COF (85.65%) (Figure 3c). The electron‐consumption yield, defined as 2 × CO + 2 × H₂, also followed the trend Mo/COF > Ru/COF > Pt/COF > Pd/COF > Au/COF. Herein, Mo/COF achieved 601.06 µmol·g⁻¹·h⁻¹, representing a 7.2‐fold enhancement over that of the pristine COF. The apparent quantum efficiency of Mo/COF gradually decreases with increasing wavelength, reaching 1.85% at 420 nm (Table S5 and Figure S24). Notably, the Mo/COF exhibited superior CO evolution activity and selectivity compared with most reported COF‐based photocatalysts under comparable conditions, even including systems employing photosensitizers and sacrificial agents (Figure 3b; Table S6).

FIGURE 3.

(a) Photocatalytic CO₂ reduction activities of sp²c‐COF and M/COFs (λ ≥ 420 nm). (b) Performance comparison of Mo/COF with other COF‐based catalysts under pure CO₂ or simulated flue‐gas. (blue: with sacrificial agent and photosensitizer; orange: without sacrificial agent; red: without sacrificial agent or photosensitizer; green: dilute CO₂ atmosphere). (c) Calculated CO selectivity and electron consumption for sp²c‐COF and M/COFs. (d) Control experiments on Mo/COF under different conditions. (e) Mass spectra of products in photocatalytic ¹³CO₂ reduction over Mo/COF.

Under simulated flue‐gas conditions (15% CO₂), Mo/COF still maintained a high CO evolution rate of 146.6 µmol·g⁻¹·h⁻¹ (Figure S25), surpassing all the reported photocatalysts under comparable conditions (Figure 3b; Table S7). Control catalysts on alternative supports—Mo/C₃N₄ (CO: 60.77 µmol·g⁻¹·h⁻¹, 73.07%), Mo/COF (C═N) (25.36 µmol·g⁻¹·h⁻¹, 73.30%), and Mo/MIL‐41 (no detectable products)—showed markedly lower activity under identical irradiation (Figure S26), underscoring the structural advantages of the designed sp²c‐COF matrix for single‐atom catalysis. UV–vis DRS, photocurrent, and EIS measurements revealed that the sp²c‑COF offered two key structural advantages, namely a fully conjugated sp² framework for enhanced light absorption and an intrinsic donor‐acceptor structure for efficient charge separation (Figures S27 and S28). In addition, structural characterizations including N₂ physisorption, XPS, and EXAFS confirmed that the sp²c‑COF matrix possesses abundant bipyridine chelating sites within its open pores for single‑atom anchoring and mass transfer. The unique structural features of sp²c‑COF, including its conjugated framework and well‑defined catalytic sites, are likely the reasons why it exhibits advantages over the other carriers. ¹H NMR detected no liquid products, confirming CO as the main product (Figure S29). Blank tests showed that no CO was produced in the absence of catalysts or CO₂ (Figure 3d), verifying that CO arose from photocatalytic CO₂ reduction over Mo/COF. No carbonaceous products were detected in the absence of the Mo/COF catalyst with only the ionic liquid, indicating that the ionic liquid serves solely as an additive to promote CO₂ adsorption and local concentration near the active sites, thus boosting the overall efficiency. The quantitative O₂ evolution reached 85%–90% with no detectable oxidative by‐products, evidencing a stoichiometrically balanced CO₂/H₂O photocatalytic cycle (Figure S30). The series of photocatalytic CO₂ reduction chromatographic data for Mo/COF further corroborated that the O₂ generated in the system originated from the water oxidation process, ruling out the possibility of device leakage (Figures S31–S36). The isotopic labelling with ¹³CO₂ and H₂ ¹⁸O (Figure 3e) yielded MS signals at m/z = 29 (¹³CO) and m/z = 36 (¹⁸O₂), confirming that CO originated from CO₂ reduction while O₂ derived from water oxidation. Notably, the Mo/COF maintained its performance over ten consecutive runs with negligible loss (Figure S37). The density functional theory (DFT)‐calculated formation energies for metal anchoring followed Mo (−4.67 eV) > Ru (−4.13 eV) > Pd (−3.68 eV) > Pt (−3.39 eV) > Au (−2.96 eV), identifying Mo/COF as the most thermodynamically stable configuration. Post‐reaction PXRD and FTIR profiles were essentially unchanged from the fresh catalyst, further demonstrating the robust structural integrity (Figures S38 and S39).

In addition, the CO₂ activation process was evaluated through linear sweep voltammetry (LSV), where a greater current density was recorded for the CO₂‐saturated solution in comparison to the N₂‐saturated solution, indicating that the CO₂ molecules were sufficiently activated by M/COF (Figure S40). Notably, Mo/COF manifested the most robust current density in the CO₂‐saturated solution in comparison with Ru/COF, Pd/COF, Pt/COF, and Au/COF, thereby substantiating the pronounced activation capacity of Mo/COF toward CO₂. The CO₂ temperature‐programmed desorption (CO₂‐TPD) further corroborated these trends (Figure S41). In the 100°C–200°C region, Au/COF, Pt/COF, Pd/COF, Ru/COF, and Mo/COF exhibited larger desorption peak areas than the pristine COF, indicative of the stronger physisorption. Peaks at 200°C–300°C were assigned to chemisorbed CO₂, with Mo/COF showing the largest area, whereas the COF peak was negligible [46]. These results demonstrated the robust CO₂ chemisorption capacity of M/COFs and underscored the key role of single‐atom sites in promoting CO₂ adsorption and activation.

2.4. Insights Into CO₂ Photoreduction Mechanism

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was used to track key intermediates during CO₂ photoreduction over Mo/COF. Under vacuum and in the dark, the introduced high‐purity CO₂ led to a progressive increase of the CO₂ band at 2364 cm⁻¹ (Figure S42), evidencing the intense adsorption [47]. When comparing the CO₂ adsorption capacity, the Mo/COF demonstrated the strongest CO₂ band intensity, with the trend: Mo/COF > Ru/COF > Pd/COF > Pt/COF > Au/COF > COF (Figure 4a), verifying the exceptional CO₂ adsorption ability of Mo/COF. During illumination, the bands assignable to surface carbonaceous species—s‐HCO₃ ⁻ (1440 cm⁻¹), m‐CO₃ ²⁻ (1478 cm⁻¹), and HCO₃ ⁻ (1651 cm⁻¹)—increased (Figure 4b), alongside new peaks at 1502 and 1615 cm⁻¹ attributable to *COOH, a well‐known CO‐forming intermediate [48, 49, 50]. A *CO adsorption band at 2165 cm⁻¹ was also observed, revealing the elevated *CO coverage under irradiation [51]. The water oxidation process was further probed via in situ DRIFTS under pure H₂O vapor: bands associated with oxygenated intermediates—*OH (997 cm⁻¹), *OOH (1250, 1152, 834 cm⁻¹), and *O (1410 cm⁻¹)—grew with illumination time (Figure 4c), providing direct evidence for water oxidation on Mo/COF [52, 53]. On basis of these observations, a reasonable pathway was proposed: (i) CO₂ adsorption at Mo sites; (ii) adsorption of H₂O on pyrene units followed by stepwise dehydrogenation to generate oxygen species; (iii) protonation of activated CO₂ to form *COOH, which was further reduced to *CO; and (iv) CO desorption from the catalyst surface (Figure 4g).

FIGURE 4.

(a) The interactions between sp²c‐COF and M/COF with CO₂. In situ DRIFTS of Mo/COF for (b) CO₂ reduction and (c) water oxidation under visible‐light irradiation (λ ≥ 420 nm). (d) Gibbs free energy pathway for the formation of CO from CO₂ over sp²c‐COF and M/COF. (e) Relationship between the CO₂ adsorption energy and the d‐band centers of the M/COF catalysts. (f) Gibbs free energy profiles of sp²c‐COF and M/COF in the H₂O oxidation process. (g) Schematic diagram of proposed artificial photosynthetic CO₂ reduction mechanism for Mo/COF. (h) Schematic diagram of the bonding of CO₂ to the Mo/COF.

The thermodynamics of CO₂ reaction at different metal active sites were further studied by DFT calculation. The low CO₂ adsorption energy imposed a significant kinetic limitation on the CO₂RR in the pristine COF (−0.596 eV). Conversely, the adsorption of CO₂ by M/COF was markedly enhanced following the incorporation of the metal active sites. In general, CO₂ and metal atoms will bond in one of the four forms shown in Figure S43, namely η¹‐O, η¹‐C, η²‐C, O, and η²‐O, O, where the superscript number indicates the coordination number of CO₂ with the metal center [54, 55]. Taking Mo/COF as a representative, the CO₂ adsorption energy of different adsorption modes was calculated, and the optimal adsorption configuration was predicted to be η²‐C, O, as its adsorption energy (−2.585 eV) was markedly superior to η¹‐O (−0.366 eV), η¹‐C (−0.968 eV), and η²‐O, O, (−1.247 eV) configurations. Specifically, Mo/COF exhibited the highest CO₂ adsorption capacity, exceeding that of Ru/COF (−1.896 eV), Pd/COF (−1.496 eV), Pt/COF (−0.678 eV), and Au/COF (−0.181 eV). The formation energy barrier of *COOH in Mo/COF was 0.497 eV, which was markedly lower than that observed for COF (0.755 eV), Ru/COF (0.698 eV), Pd/COF (0.866 eV), Pt/COF (1.038 eV), and Au/COF (0.679 eV). In contrast, the formation of *OCHO by hydrogenation of *CO₂ was a process that required a significant input of energy (Mo/COF: 0.714 eV), which was consistent with the high selectivity of CO₂ for CO. Furthermore, the Mo/COF exhibited a considerable degree of free energy variation during the formation of *H, suggesting that the incorporation of Mo SAs inhibited the HER kinetics during CO₂RR (Figure S44). *COOH → *CO was an exothermic process. Moreover, due to the smaller energy barrier, CO molecules adsorbed in Mo/COF were more conducive to the formation of CO through *CO desorption (Figure 4d). In CO‐saturated solutions, the current density of Mo/COF was found to be the lowest in comparison to Ru/COF, Pd/COF, Pt/COF, and Au/COF, confirming the weak binding ability of Mo/COF to CO (Figure S45).

The H₂O oxidation pathway of these M‐COFs in the artificial photosynthetic reaction was further investigated. The HOMO‐LUMO analysis was conducted, which revealed that the holes were predominantly concentrated in the pyrene radicals. To provide further clarification regarding the active site of H₂O oxidation, the energy required for the initial step of *OH formation in site 1 and site 2 in the pyridyl was calculated to be 2.07 and 2.38 eV, respectively, indicating that the likely active site of H₂O oxidation was site 1 (Figure S46). Then, the free energy barrier of water oxidation on different catalysts was compared. The resultant findings evidenced that the free energy barrier of H₂O → *OH, the step determining the water oxidation rate of Mo/COF (2.07 eV), was significantly inferior to that of COF (2.73 eV), Au/COF (2.65 eV), Pt/COF (2.8 eV), Pd/COF (2.46 eV), and Ru/COF (2.28 eV), thus confirming the effectiveness of introducing a single Mo atom into COF to promote the oxidation activity of H₂O (Figure 4f).

To elucidate how different metal sites activate CO₂, we analyzed the projected density of states (PDOS) of the M₁N₂ centers and compared them with their PDOS after CO₂ adsorption. As illustrated in Figure S47, the gas‐phase CO₂ exhibited characteristic molecular orbitals near the Fermi level (FL): σ, π, lone pair (LP), σ*, and π*. The O‐centered LP (derived from O 2p) lied close to Fermi level and could donate electrons, whereas the antibonding π* level (mainly C 2p and O 2p) could accept electrons [56]. As depicted in Figure 5a, the Au PDOS in Au/COF was highly symmetric and largely degenerate, with most states lying below the Fermi level, indicating its electrons were confined to deep and less reactive levels. Likewise, the late transition metals Pt and Pd exhibited similarly symmetric PDOS with only modest state density near the FL (Figure 5b,c). By contrast, Ru and Mo in Ru/COF and Mo/COF displayed substantial state density near the FL (Figure 5d,e). In particular, Mo₁N₂ sites showed pronounced spin polarization, with asymmetric spin‐up and spin‐down channels straddling FL, imparting exceptionally high electronic activity in this region. Upon adsorption on Au/COF or Pt/COF, the CO₂ molecular orbitals remained essentially unchanged from the gas phase, implying negligible CO₂ activation (Figure 5f,g). In comparison, the adsorption on Pd/COF and Ru/COF induced the partial hybridization between CO₂ molecular orbitals and metal d states, indicative of moderate activation (Figure 5h,i). Strikingly, CO₂ coordinated on Mo/COF underwent an extensive orbital hybridization with Mo, pronounced level splitting, and a clear shift to deeper binding energies relative to gas‐phase CO₂, evidencing an intensive CO₂ activation (Figure 5j). Evidently, CO₂ activation across the M/COF series correlated closely with the metal d‐band center (ε_d). Specifically, according to the Hammer‐Nørskov d‐band model, when the metal d orbitals interact with the frontier orbitals of CO₂, particularly the π* antibonding orbitals, bonding and antibonding hybridized states are generated [57, 58]. As the ε_d shifts upward toward the Fermi level, the energy mismatch between the metal d states and the CO₂ π* orbitals decrease, thereby strengthening the orbital hybridization and facilitating stronger π‐back‐donation from the metal center to CO₂. Meanwhile, the hybridized antibonding states are shifted closer to or above the Fermi level and remain less occupied, reducing the energetic penalty associated with antibonding electron occupation. Mo/COF exhibited the shallowest d‐band center (ε_d = −0.314 eV), closer to FL than Au/COF (−2.638 eV), Pt/COF (−1.942 eV), Pd/COF (−1.430 eV), and Ru/COF (−0.845 eV). A strong linear relationship linked the d‐band center to CO₂ adsorption energy (E_ad): y = −1.044x – 2.864 (R² = 0.98) (Figure 4e). To test the universality of this correlation, additional M₁N₂ sites, including Co and Ni, were evaluated. While Co closely followed the established trend (ε_d = −1.215 eV, E_ad = −1.543 eV), Ni exhibited a moderate deviation of approximately 0.17 eV (ε_d = −1.524 eV, E_ad = −1.103 eV), which led to weaker CO₂ adsorption than predicted by the d‐band center alone (Table S4). This deviation highlighted the limitation of using the d‐band center as a single‐parameter thermodynamic descriptor for late transition metals. The electronic origins of this deviation, specifically related to orbital occupancy and energy alignment, were further elucidated in the following PDOS and orbital splitting analysis. Consistently, both electron transfer from M/COF to CO₂ and the CO₂ bond‐angle distortion scaled with the d‐band center: Mo/COF donated the most charge to CO₂ molecule (Mo/COF 0.753 e⁻ > Ru/COF 0.413 e⁻ > Pd/COF 0.249 e⁻ > Pt/COF 0.247 e⁻ > Au/COF 0.208 e⁻ > COF 0.188 e⁻) and induced the largest CO₂ bond‐angle bending (132° for Mo/COF vs 171° for Au/COF, 164° for Pt/COF, 158° for Pd/COF, and 144° for Ru/COF), further underscoring the superior activation capability of Mo/COF.

FIGURE 5.

PDOS plots of *CO₂, and central metals in (a,f) Au/COF, (b,g) Pt/COF, (c,h) Pd/COF, (d,i) Ru/COF, and (e,j) Mo/COF before and after interaction with CO₂. Schematics of d‐orbital electronic configurations of central metals in (k) Au/COF, (l) Pt/COF, (m) Pd/COF, (n) Ru/COF, and (o) Mo/COF.

Furthermore, given the two‐coordinate geometry (M₁N₂) of the active sites in M/COF, an approximately linear ML₂ crystal‐field model was adopted, with the relevant d orbitals grouped as d_z2, d_xz/d_yz, and d_x2‐y2/d_xy [59]. The d‐orbital splitting and occupancy were thoroughly analyzed from the integrated PDOS occupation numbers (Figure S48 and Table S8). In Au/COF, Pt/COF, Pd/COF, and Ru/COF, extensive d‐orbital electron pairing was observed (Figure 5k–n), and the associated pairing energy limited the electron activity. Noticeably, the Mo₁N₂ sites depicted significant spin polarization, with its d_xz, d_yz, d_x2‐y2, and d_xy orbitals being closely aligned in energy (Figure 5o). Three unpaired electrons were present, consistent with a magnetic moment of ∼3 µB. Furthermore, room‐temperature magnetization versus magnetic field (M‐H) measurements revealed a significantly enhanced magnetic response for Mo/COF (Figure S49). Specifically, Mo/COF exhibited a magnetization of approximately 0.035 emu·g⁻¹ at 4 T, which was approximately 18‐fold that of pristine sp²c‑COF (0.002 emu·g⁻¹). The linear and hysteresis‐free M‐H behavior indicated a dominant paramagnetic response, confirming that the atomically dispersed Mo sites possess significant spin polarization. In addition, the geometry optimization exhibited that the N─Mo─N angle (76.7°) deviated more strongly from the ideal linear configuration than N─Au─N (80.0°), N─Pt─N (83.1°), N─Pd─N (79.9°), and N─Ru─N (83.2°), which fundamentally contributed to the high‐spin state of Mo. Upon CO₂ adsorption, the Mo d_z2 orbital accepted the lone‐pair electrons from the O p_z orbital via σ bonding, while the Mo d_xz and d_yz orbitals donated electrons into the CO₂ π* orbital composed by p_x and p_y orbital of C and O through π back‐donation (Figure 4h). Overall, the reactivity of the d electrons was governed by the metal's d‐band characteristics. The shallowest d‐band center near FL and the strongest spin polarization of Mo₁N₂ induced the most abundant unpaired electrons, being free from pairing‐energy constraints. The unshackled single electrons in Mo/COF were most active, along with the half‐occupied d‐orbitals, could extensively activate CO₂ through synergistic σ‐donation/π‐back‐donation. To further evaluate the universality of this electronic‐structure‐dependent activity, additional M₁N₂ sites, including Co and Ni, were investigated. While Co closely followed the established linear trend between ε_d and E_ad, Ni showed a moderate deviation of approximately 0.17 eV, resulting in weaker CO₂ adsorption than predicted. Although the vacant d_z2 orbital of Ni (3d⁸) could facilitate σ lone‐pair donation, its substantially deeper d‐band center caused a pronounced energy mismatch with the CO₂ π* orbitals, thereby suppressing effective π‐back‐donation. Furthermore, the absence of high‐spin‐enabled orbital vacancies in Ni limited its ability to bypass pairing‐energy constraints (Figure S50). This comparative analysis revealed that the superior performance of Mo/COF stemmed from an optimal synergy: the shallowest ε_d provided the thermodynamic driving force, while high spin polarization furnished the essential kinetic freedom for efficient charge transfer. Furthermore, the introduction of the [Emim]BF₄ ionic liquid plays a pivotal synergistic role in the overall photocatalytic system. By selectively absorbing and enriching CO₂, the ionic liquid effectively mitigates kinetic mass‐transfer limitations, thereby substantially elevating the local reactant concentration in the immediate vicinity of the M₁N₂ active sites. Besides, it actively promotes CO₂ activation through a combined mechanism [60, 61]. The [Emim]⁺ cation polarizes the oxygen atoms of CO₂, while the interaction between the Mo center and the carbon atom further weakens the C─O bonds. This cooperative effect facilitates subsequent C─O bond cleavage and lowers the overall activation barrier for CO₂ reduction.

Obviously, the remarkable performance leap from pristine sp²c‐COF to Mo/COF is dictated by a fundamental structure‐performance synergy. The sp²‐carbon‐conjugated framework functions as more than a passive support; its pyrene chromophores serve as efficient antenna units for broadened visible‐light harvesting, while the fully conjugated backbone promotes initial exciton dissociation. Moreover, the rigid 2D pores and bidentate bipyridine motifs provide a well‐defined microenvironment that stabilizes the highly unsaturated Mo₁N₂ sites through strong metal‐support interactions and steric hindrance. While the pristine COF suffers from spatial overlap of HOMO and LUMO leading to rapid recombination, the incorporation of Mo atoms triggers a decisive orbital redistribution. The LUMO shifts to the Mo‐bipyridine sites while the HOMO remains on the pyrene units, establishing a directional intramolecular charge transfer pathway that prolongs the average carrier lifetime by nearly five‐fold. This allows energized electrons to be efficiently channeled to the d‐band‐optimized Mo centers, where the high spin polarization and abundant unpaired electrons facilitate intense CO₂ activation through synergistic σ‐donation and π‐back‐donation. Consequently, this multi‐level structural optimization ensures both high charge separation efficiency and low reaction energy barriers, boosting the CO₂ evolution.

3. Conclusions

In summary, various M₁N₂ single‐atom sites (Au, Pt, Pd, Ru, and Mo) were designed and anchored on a vinylene‐linked covalent organic framework via a facile impregnation protocol for gas‐solid photocatalytic CO₂ reduction under visible light. The optimal Mo/COF achieved a CO evolution rate of 294.43 µmol·g⁻¹·h⁻¹ with a stoichiometric O₂ release of 128.63 µmol·g⁻¹·h⁻¹ under pure CO₂ and still maintained a rate of 146.6 µmol·g⁻¹·h⁻¹ under 15% CO₂. The PDOS calculations revealed that the superior activity of Mo/COF was not solely a consequence of its shallow d‐band center (−0.314 eV), but originated from its synergistic interplay with strong spin polarization. Compared with modulating the d‑band center alone, this synergy breaks the electron pairing energy and generates abundant unpaired electrons near the Fermi level, enabling charge transfer free from pairing constraints. This specific electronic configuration generated abundant unpaired single electrons near the Fermi level, which exhibited enhanced activity for CO₂ activation owing to the absence of pairing‐energy constraints. Such synergy facilitated pronounced σ‐donation and π‐back‐donation, as quantitatively evidenced by the largest charge transfer (0.753 e⁻) and the most pronounced O═C═O bond‐angle distortion (132°) within the series. In situ DRIFTS and thermodynamic analyses further confirmed that the Mo₁N₂ sites facilitated *COOH formation, lowered the energy barrier for CO desorption, and promoted the water oxidation. This work elucidates how d‐band position and spin polarization in isostructural M₁N₂ SACs govern the CO₂ photoreduction process, providing a blueprint for the rational design of efficient photocatalysts.

4. Experimental Section

4.1. Synthesis of Pure COF and M/COF

TFPPy (60 mg), BPYDAN (45 mg), 1,4‐dioxane (2 mL), and 4 m KOH (0.2 mL) were placed in a 25 mL Pyrex tube and sonicated for 5 min. The mixture was then degassed by three freeze‐pump‐thaw cycles (liquid N₂), sealed under vacuum, and heated at 110°C for 72 h. The resulting precipitate was collected by centrifugation, washed with water and THF, and dried under vacuum at 110°C for 12 h to afford the COF. The Mo/COF catalyst was prepared by dispersing COF (25 mg) and Mo(CO)₆ (10 mg) in toluene (25 mL), heating the mixture at 110°C for 5 h under N₂, and isolating the product by washing with methanol and water, followed by drying under vacuum at 60°C for 24 h. The Au/COF, Pt/COF, Pd/COF, Ru/COF, Co/COF, and Ni/COF catalysts were obtained analogously by suspending COF (25 mg) in methanol solutions (15 mL) of PtCl₄, PdCl₂, RuCl₃, HAuCl₄, CoCl₂, and NiCl₂, respectively.

4.2. Photocatalytic CO₂ Reduction

For a typical photocatalytic CO₂ reduction reaction, a slurry was prepared by dispersing 5 mg of photocatalyst in 1 mL of H₂O, which was then coated onto a glass fiber. The glass fiber was transferred to a 40 mm × 1.5 mm glassware, dried at 60°C. The glassware was then placed in the reactor, and 200 µL of deionized water and 100 µL of ionic liquid (to enhance CO₂ adsorption) were injected. After CO₂ was introduced and the adsorption equilibrium was established in the dark, the reaction was initiated under simulated sunlight. To prevent leakage, strict sealing was performed, and Ar gas was purged before injection of CO₂ gas. The reactor was irradiated for 4 h with a 300 W Xe lamp (λ ≥ 420 nm) at an irradiance of 100 mW/cm², under vigorous stirring. The CO₂ reduction products were analyzed using a Shimadzu gas chromatograph (GC‐2014C) equipped with a flame ionization detector (FID) and a thermal conductivity detector (TCD). To determine the carbon source of the products, isotope‐labeling experiments with ¹³CO₂ were performed under identical conditions, and the resulting ¹³CO products were identified by gas chromatography‐mass spectrometry (GC‐MS, Agilent 7890B‐5977B) equipped with a CP‐Molecular Sieve 5A column (25.0 m × 0.32 mm × 30 µm). The analysis of ¹⁸O₂ was carried out using the same procedure as for ¹³CO.

The electron consumption rate for CO₂ reduction was evaluated based on the formula:

$${\mathrm{R}}_{\mathrm{electron}}=\left(2\mathrm{Yiel}{\mathrm{d}}_{\mathrm{CO}}+2\mathrm{Yiel}{\mathrm{d}}_{\mathrm{H}2}\right)/\left({\mathrm{m}}_{\mathrm{cat}}\times \mathrm{t}\right)$$

Funding

This work was supported by the National Natural Science Foundation of China (22272028), the Youth Talent Support Program of Fujian Province (00387077), the Industry‐University Research Collaboration Project of Fujian Province (2023H6005).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma73738‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.