Introduction

Artificial solar-driven CO2 reduction to renewable fuels (e.g., CO1, HCOOH2, HCHO3, CH3OH4, and CH45,6) based on semiconductor-mediated photocatalysis is an intriguing strategy for producing carbon-neutral energy, while also mitigating emerging environmental crises7,8. Developing effective photocatalysts with high activity and selectivity has become one of the foremost challenges towards accomplishing this objective9. Atomically dispersed catalysts exhibiting maximum atom utilization and unrivaled photoelectric performance are considered the most promising candidates10,11. Recently, single-atom catalysts (SACs) with atomically distributed catalytic metal sites have demonstrated impressive catalytic performance in the selective CO2 reduction reaction (CO2RR)12,13,14. However, their limited light absorption capacity and production of numerous intermediates and by-products impede their further implementation. Moreover, the catalytic activity of SACs is limited by the lack of well-defined active sites, inadequate suppression of the competing hydrogen evolution reaction (HER), and weak interaction with the substrate materials15,16,17.

Generally, the CO2RR entails three primary processes: photoabsorption, carrier separation, and CO2 reduction18,19,20,21. Therefore, ideal photocatalysts should possess a sufficient optically active center to generate photoinduced charges as well as an efficient channel to transfer these carriers to a catalytically active center of high selectivity. In contrast to SACs with single catalytic sites, dual-atom catalysts (DACs) with bimetallic centers exhibit tremendous potential for the combination of atom-specific characteristics due to the optimized transfer of photogenerated charge carriers and more complex functionalities between adjacent active sites22,23. The latter can lead to the enhanced generation of photoinduced charges and improved catalytic activity while maintaining high atom utilization, stability, and properties, which are constrained by the nature of the atomic dispersion24,25,26,27,28,29. As an example, the collaborative coordination of CO at two sets of single-atom sites (Ni and Fe) anchored on nitrogenated carbon has recently been demonstrated to yield an efficient electrocatalyst for the CO2RR, which significantly reduced the reaction barriers for the formation of COOH* and desorption of CO, and resulted in a structural evolution into the CO-adsorbed moiety upon CO2 uptake30. Therefore, rationally developed bimetallic site photocatalysts can be used to effectively modulate the selectivity and activity of the CO2RR by controlling the reaction pathway31. However, substantial challenges persist in acquiring a comprehensive understanding of the precise roles of the specific single-atom sites and their interaction occurring at the atomic level during the dynamic photocatalytic CO2RR process.

Among possible scaffolds for the integration of the dual-atomic sites, crystalline porous network materials such as metal-organic frameworks (MOFs) and covalent organic frameworks (COFs) are renowned for their well-defined porosity, high specific surface area, and predetermined structure32,33,34,35,36. These features combined with the confining influence of the ordered pore structure and the abundance of surface binding groups produce a highly favorable environment for the anchoring of metal atoms37,38,39,40,41. On the other hand, there are only a few scarce in-depth studies elucidating the critical role of the substrates in the CO2RR process. From the perspective of DACs, a strong interaction between the catalyst and the substrate is desired to improve the catalytic stability. Furthermore, the interaction between the catalyst and the substrate modulates the processes of carrier separation and transfer and thus has a direct influence on the catalytic performance.

In this paper, we develop an electrostatically driven self-assembly assisted by Phen ligands strategy for incorporating atomically dispersed La-Ni sites (specifically, LaNi-Phen, where Phen= phenanthroline) into conjugated boronate-ester-linked COFs (COF-5 colloid, supplementary Fig. 1)42,43. The La-Ni coordination structure is designed to enable efficient production and transfer of photoinduced charges through the optically active (La site) and catalytically active (Ni site) centers. We demonstrate the coordination of the La and Ni sites and unravel the role of the substrate, fostering the effective transfer of the photogenerated carriers. As a result, the optimized LaNi-Phen/COF-5 photocatalyst without any additional photosensitizer exhibits a CO evolution activity of 605.8 μmol·g–1 h–1 with high selectivity (98.2%), corresponding to a remarkable 15.2-times activity improvement over pristine COF-5 colloid. Experimental characterization data and theoretical calculations confirm that the COF-5 colloid operates as an electron transfer channel through interactions with electrostatically self-assembled La-Phen and Ni-Phen in a mechanism that comprises photoelectron transfer from La-Phen to the COF-5 colloid and subsequent electron injection into Ni-Phen for the CO2RR process. This work provides a scalable strategy for producing ligand-assisted bimetallic self-assembled COF structures for efficient photocatalytic CO2 reduction.

Results

Morphological and structural characterization

La-Ni bimetallic sites are incorporated into COF-5 colloid through a facile electrostatically driven self-assembly assisted by Phen ligands process, in which the La and Ni atoms are captured by B atoms in COF-5 colloid and chelated by the Phen ligands (Fig. 1a), potentially facilitating CO2RR (Fig. 1b)44. Powder X-ray diffraction (PXRD) patterns demonstrate that the crystalline structure of COF-5 colloid remains unaltered throughout the self-assembly process (Supplementary Fig. 2)45. The morphology of COF-5, as illustrated in Supplementary Figs. 3 and 4 displays a one-dimensional (1D) nanorods structure with a width distributed around 70–80 nm. Further studies indicate that there is no discernible presence of discrete La-Ni nanoparticles in as-synthesized LaNi-Phen/COF-5 (Fig. 2a), but that the La and Ni ions are uniformly dispersed throughout the COF-5 colloid, with no sign of segregation or aggregation (Supplementary Fig. 5). Importantly, as illustrated in Fig. 2b, c, the atomically dispersed La and Ni ions marked with red box are visible using aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM), La-Ni distance of ~2.8 Å can be clear found, indicating the existence of La-Ni double-atomic sites46. X-ray photoelectron spectroscopy (XPS) results reveal that the valence states of the Ni and La ions are +2 and +3, respectively (Supplementary Figs. 6–7). In addition, the strong coordination of nitrogen atoms in Phen with Ni and La ions is confirmed by the energetic upshift of the N 1s peaks (Supplementary Note 1 and Supplementary Fig. 8)47,48. The FTIR data reveals that Phen can still be observed after the electrostatic self-assembly, as shown in Supplementary Fig. 9 and Supplementary Note 2. The specific elemental contents of La and Ni in LaNi-Phen/COF-5 are calculated to be 2.58 and 1.61 wt%, respectively (Supplementary Fig. 10 and Supplementary Table 1). Notably, the COF-5 colloid in LaNi-Phen/COF-5 comprises a high UV-vis specific surface area, open pore structure, and high CO2 capture capability, ensuring good access to the La-Ni active sites by CO2 molecules in the photocatalytic process (Supplementary Figs. 11–14 and Supplementary Note 3). Moreover, when compared to the host COF-5 colloids, the UV-vis absorption spectrum of LaNi-Phen/COF-5 exhibits a red-shift indicative of a more favorable energy level alignment for CO2 reduction (Fig. 2d, Supplementary Fig. 15). When comparing to Ni-Phen/COF-5, the modification of COF-5 colloid with La ions alone already leads to a significant red-shift of the absorption onset, implying that the introduction of the rare earth metal La as the optically active center results in enhanced light-harvesting properties and a high degree of electron delocalization in COF-5 colloid. In addition, the bandgap (Eg) of COF-5 and LaNi-Phen/COF-5 are estimated to be 3.12 and 2.86 eV by employing Tauc plots (Fig. 2d inset), respectively. The significant reduction of the bandgap facilitates the photogeneration of carriers and improves the light-harvesting properties of LaNi-Phen/COF-5. The Fermi energy level (EFermi) and the valence band maximum (VBM) of LaNi-Phen/COF-5 are determined by ultraviolet photoelectron spectroscopy (UPS)49. Based on the full scan UPS spectra of LaNi-Phen/COF-5 (Fig. 2e), we infer that the cut-off edge energy and Fermi edge energy are 15.93 eV and 0.76 eV, respectively (Fig. 2e inset). Accordingly, the Fermi energy level is calculated as 5.27 eV, while the top position of VB is determined to be −6.03 eV with respect to the vacuum level50. Combining with the bandgap (Eg = 2.86 eV) of LaNi-Phen/COF-5, the valence band maximum (VBM) and the conduction band minimum (CBM) of LaNi-Phen/COF-5 are calculated as 1.53 and −1.33 V versus NHE, respectively.

Fig. 1: Schematic illustration of the construction of the diatomic sites and CO2RR.
figure 1

a Self-assembly of LaNi-Phen into COF-5 colloid to create the diatomic site photocatalyst. b Schematic diagram of the photocatalytic CO2 reduction in LaNi-Phen/COF-5. Color code: carbon-gray, oxygen-red, boron-pink, hydrogen-white, nitrogen-blue, nickel-light blue, lanthanum-yellow. The same color scheme is applied in Figs. 3 and 6.

Full size image
Fig. 2: Visualization and spectroscopic characterization of LaNi-Phen/COF-5.
figure 2

a HR-TEM image, b atomic-resolution HAADF-STEM images of LaNi-Phen/COF-5. c measured distance of the representative La-Ni sites in panel b. d UV-vis diffuse reflectance spectra of COF-5 colloid, Ni-Phen/COF-5, La-Phen/COF-5 and LaNi-Phen/COF-5.(Inset: Tauc plots of LaNi-Phen/COF-5). e Full scan UPS spectra of LaNi-Phen/COF-5 (Inset: Cutoff edge (left) and Fermi edge (right) of LaNi-Phen/COF-5).

Full size image

The atomic structure and coordination environment of LaNi-Phen/COF-5 are explored further using X-ray absorption fine structure (XAFS) analysis. The position of the absorption edge of the K-edge X-ray absorption near-edge structure (XANES) spectra is closely related to the coordination environment of the metal atoms (Supplementary Fig. 16a, b). The near-edge absorption energies of LaNi-Phen/COF-5 is located between the metallic nickel and the metal oxide and the white line peak at 8350 eV is higher than that of metallic nickel, indicating that the Ni species is positively charged51,52,53. Furthermore, the dominant Ni-N and La-N peaks in the Fourier transform (FT) k2-weighted R-space extended X-ray absorption fine structure (EXAFS) spectra of LaNi-Phen/COF-5 are near 1.57 and 1.72 Å, respectively (Fig. 3a and b), suggesting the absence of the characteristic peaks of Ni-Ni (≈2.15 Å), La-Ni (≈2.80 Å) and La-La bonding (≈3.94 Å). Consistently, infrared (IR) spectroscopy analysis of the adsorbed CO on La-Phen/COF-5, Ni-Phen/COF-5, and LaNi-Phen/COF-5 reveals a set of CO absorption bands at 2117 cm−1, which should be assigned to the characteristic frequency for the C-O stretching of the linearly adsorbed CO on La-Ni ionic species (Fig. 3c). Furthermore, no obvious IR peak at 2092 cm−1 corresponding to La-Ni nanoparticles can be observed, confirming the absence of metal nanoparticles or clusters on LaNi-Phen/COF-5 catalysts, which further validates the AC-HAADF-STEM results (Fig. 2c). The intrinsic structure of La and Ni sites is further corroborated with the Fourier-transformed (FT) k3-weighted χ(k)-function of the extended EXAFS spectra in R space. EXAFS fitting analysis at Ni K-edge and La L3-edge show clear small peaks located at 2.90 and 2.88 Å, strongly demonstrating the existence of La-Ni coordination (Fig. 3d, e)46. In other words, the atomic dispersion of La and Ni sites as well as direct La-Ni bond are fully confirmed, which further reveals the absence of metal nanoparticles or clusters in LaNi-Phen/COF-5. According to fitting results in Supplementary Table 2, coordination number of La-N and Ni-N could be La-N4/Ni-N2 or La-N3/Ni-N3. Moreover, DFT calculations suggest that the formation energy of La-N3/Ni-N3 structure (–51.760 keV) is lower than that of La-N4/Ni-N2 structure (–51.747 keV), i.e., the La-N3/Ni-N3 structure is preferred (Supplementary Fig. 17). La-N and Ni-N scattering at the first shell over LaNi-Phen/COF-5 with the coordinated structures of La-N3 and Ni-N3 are further confirmed by the EXAFS fitting analyses (Supplementary Figs. 18–21, and Supplementary Table 2), which is consistent with the results of DFT optimization (Supplementary Fig. 22). Free LaNi-Phen is assembled to demonstrate the involvement of the pore confinement effect of COF-5 colloids in facilitating the formation of the La-N3/Ni-N3 coordination structure. The experimental results revealed that the bimetallic La-Ni center eventually self-assembles into a six-coordination structure (La-N6/Ni-N6) in the absence of COF-5 colloid, leading to undesired dual sites because they are fully coordinated in the free state (Supplementary Table 2). This result implies that the pore structure of COFs can significantly promote the precise modulation of the La-Ni diatomic structure, which is at the origin of the formation of bimetallic LaNi-Phen structure in COF-5 colloids.

Fig. 3: Structural characterization by X-ray absorption spectroscopy.
figure 3

a Ni K-edge extended X-ray absorption fine structure (R space plots), b La L3-edge extended X-ray absorption fine structure (R space plots). c DRIFTS spectra of adsorbed CO on La-Phen/COF-5, Ni-Phen/COF-5, LaNi-Phen/COF-5, and a reference material comprising metallic La and Ni. d Ni K-edge and e La L3-edge EXAFS (points) and curve-fit (line) for LaNi-Phen/COF-5 shown in R-space. The data are k3-weighted and not phase-corrected, inset in panel j is the schematic model of the structure of La-N3/Ni-N3.

Full size image

CO2 photoreduction activity of LaNi-Phen/COF-5 catalysts

The CO2 reduction reaction is conducted and described in the Methods section to evaluate the catalytic activity of LaNi-Phen/COF-5 under simulated solar irradiation. The CO2RR performance of LaNi-Phen/COF-5 with various La-Ni metal proportions is monitored and optimized systematically (Supplementary Table 3), yielding the highest catalytic activity of 608 μmol·g−1·h−1 (CO) and selectivity of 98.2% (CO over H2). Significantly, the catalytic activity of the COF-5 colloid (39.9 μmol·g−1·h−1), La-Phen/COF-5 (195.4 μmol·g−1·h−1), and Ni-Phen/COF-5 (224.4 μmol·g−1·h−1) is 15.2, 3.1, and 2.7 times lower, respectively (Fig. 4a, b). Moreover, the CO selectivity is substantially higher than that of COF-5 colloid (70.7%), La-Phen/COF-5 (95.9%), and Ni-Phen/COF-5 (91.2%). We also investigated the CO2RR performance of a physical mixture of La-Phen/COF-5 and Ni-Phen/COF-5 (denoted as mix-LaNi-Phen/COF-5), which exhibits significantly lower catalytic activity with a CO production rate of 115.9 μmol·g−1·h−1. These results unambiguously revealed that La and Ni sites served as predominant catalytic centers and optical centers, respectively, which is facilitated electron transfer and promoted light absorption. Remarkably, the aforementioned activity and selectivity of LaNi-Phen/COF-5 are significantly higher than those of previously reported photocatalysts without unstable noble metal photosensitizers (Supplementary Table 4). In addition, LaNi-Phen/COF-5 exhibits a higher total electron transfer (1,235.0 μmol g−1 h−1) than other catalysts (Supplementary Table 5). As will be shown in the following, the excellent CO2RR performance in this work could be attributed to COF-5’s intrinsic structural and electronic characteristics, which facilitates photogenerated carries transfer from La to catalytic center Ni. The foregoing results highlight the crucial role of COF-5 colloid and LaNi-Phen for achieving high performance, and control experiments conclusively confirm that it is an authentic CO2RR process driven by continuous photoexcitation, in which COF-induced confinement effect facilitates the transport of the photogenerated carriers, BIH and H2O operate as electron sacrificial agents and proton sources, respectively (Fig. 4c, Supplementary Figs. 23, 24 and 25). An isotope-labeled carbon dioxide (13CO2) photocatalytic reduction experiment based on LaNi-Phen/COF-5 is performed to investigate the origin of the CO2RR products. Three signals generate in the mass spectra (MS), where the total ion chromatographic peak at ~7.45 min corresponds to CO (Supplementary Fig. 26). The predominant MS signal at m/z = 29 corresponds to molecular ions of the (13CO+) peak of 13CO, whilst the others (13C+ at m/z = 13 and O+ at m/z = 16) originated from fragments of 13CO, demonstrating that CO2 gas is responsible for the formation of the carbon-related products (Fig. 4e). Cycling tests reveal that the CO2RR evolution rate and selectivity exhibit only negligible losses after at least 5 cycles and a total irradiation time of 15 h, demonstrating that the LaNi-Phen/COF-5 catalyst possesses excellent structural robustness and durability in the CO2RR (Fig. 4e). Furthermore, the heterogeneity tests demonstrate the heterogeneous nature of the LaNi-Phen/COF-5 catalysts (Supplementary Fig. 27 and Supplementary Note 4). Moreover, the XRD pattern, FTIR spectra, and XPS spectra of the recovered LaNi-Phen/COF-5 after 5 runs remain unchanged from the as-prepared sample (Supplementary Figs. 28–30). The catalyst’s microstructure was retained during the CO2RR process (Supplementary Figs. 31 and 32), demonstrating that La-Ni-Phen units are strongly maintained in the interior cavities of COF-5 colloid through the confinement effect of the pores. Photoelectrochemical measurements are performed to elucidate the separation and transfer capability of photogenerated carriers over LaNi-Phen/COF-5 (Supplementary Note 5). The steady-state photoluminescence (PL) emission intensity of LaNi-Phen/COF-5 is markedly more attenuated than that of pure COF-5 colloid, owing to improved exciton splitting and charge transfer (Supplementary Fig. 33)54. Noteworthy, when COF-5 colloid is only modified by La metal ions, a bathochromic shift (≈26 nm) of the PL peak is observed, suggesting that the modulation of COF-5 colloid’s optical properties can be attributed to La ions rather than Ni ions. Time-resolved photoluminescence (TR-PL) measurements illustrate that the carrier lifetimes of the photogenerated charges reduce from 4.75 ns (pure COF-5 colloid) to 0.53 ns (LaNi-Phen/COF-5), demonstrating that the lowering of the recombination of photogenerated charge carriers is due to the loading of COF-5 colloid with La and Ni ions (Supplementary Fig. 34 and Supplementary Table 6)55. In addition, the separation and the transfer of the photogenerated carriers in LaNi-Phen/COF-5 photocatalytic reduction system are investigated by photocurrent-time (I-t) response and electrochemical impedance spectroscopy (EIS) data. The EIS results show the biggest semicircular arc radius of the Nyquist plot in the COF-5 colloid (Supplementary Fig. 35), indicating the highest electron transfer resistance. As the construction of the bimetallic structure favors for charge transfer, LaNi-Phen/COF-5 has the smallest radius compared to La-Phen/COF-5 or Ni-Phen/COF-5, which is further confirmed by the increase in photocurrent density (Supplementary Fig. 36).

Fig. 4: Photocatalytic CO2 reduction performance.
figure 4

a, b Time-dependent CO (a) and H2 (b) evolution curves under UV-vis light irradiation (λ > 380 nm) within 5 h using COF-5 colloid (black spheres), La-Phen/COF-5 (green spheres), Ni-Phen/COF-5 (pink spheres), LaNi-Phen/COF-5 (red spheres), and mix-LaNi-Phen/COF-5 (blue spheres). c Control experiments of the photocatalytic CO2 reduction performance over LaNi-Phen/COF-5 under altered conditions (Conditions: 10 mM BIH, 50 mL MeCN, x mL H2O, 300 W Xe lamp, 298 K). d Mass spectra of 13CO (m/z = 29) produced in the photocatalytic reduction of 13CO2 over LaNi-Phen/COF-5. e CO and H2 production rates in cycling experiments over LaNi-Phen/COF-5.

Full size image

Reaction mechanism of CO2 photoreduction to CO

The adsorbed CO probe molecule on various catalysts is analyzed using in situ infrared spectroscopy to acquire a thorough understanding of the reaction mechanism of the CO2 photoreduction. The bands in the range of 1800 to 1900 cm−1 correspond to the C-O stretching vibrations of bridging CO (CObridge) adsorbed on the La-Ni dual-atom sites, and the intensity of CO absorption increases with irradiation time (Fig. 5a), whereas no obvious CObridge band is observed when using single La or Ni sites on COF-5 (Fig. 5b)56,57. These results illustrate unequivocally that only the construction of the La-Ni dual-atom catalyst provides efficient CO adsorption sites, enabling at the same time the efficient electron transfer between the La and Ni sites discussed above as the source of the observed excellent photoelectric performance for the LaNi-Phen/COF-5 catalyst58. In situ XAFS measurements are performed to dynamically monitor the oxidation state of active sites to better understand their function on the atomic scale59. The Ni K-edge XANES spectra of LaNi-Phen/COF-5 clearly illustrate that the white-line intensity is slightly enhanced in a CO2-saturated acetonitrile solution compared to that in an Ar-saturated acetonitrile solution, revealing the increase of the Ni oxidation state, which is generally ascribed to the spontaneous electron transfer from the active Ni center to the C 2p orbital of CO2 during the adsorption and activation of CO260. The intensity of the white line peak is closely related to the elemental chemistry state, as proven by the calculations of the Ni surface charge, which increases from −0.981 to −0.132 after CO2 adsorption (Supplementary Fig. 37). Interestingly, the white-line intensity of the Ni K-edge XANES spectra reduces slightly under the Xenon lamp irradiation and its position is between that of a CO2-saturated and an Ar-saturated acetonitrile solution, demonstrating the gradual recovery of the active Ni center’s oxidation state in the CO2RR process. Furthermore, an oxidation state variation is observed in the La L3-edge XANES spectra (Supplementary Fig. 38). Summarizing, the in situ XAFS analysis reveals that Ni atoms are the active centers for the CO2RR, while La atoms are not only the optically active center but also the catalytically active center for CO2 adsorption and activation.

Fig. 5: Detection of the reaction mechanism for the photoreduction of CO2 to CO.
figure 5

a In situ DRIFTS spectra of adsorbed CO at 1800–1900 cm−1 for detecting the effective active site over LaNi-Phen/COF-5. b DRIFTS spectra of adsorbed CO at 1800–1900 cm−1 over La-Phen/COF-5, Ni-Phen/COF-5, and LaNi-Phen/COF-5. c Ni K-edge XANES spectra of LaNi-Phen/COF-5 in CO2 photoreduction reduction reaction at room temperature and 1 atm of Ar or CO2-saturated aqueous solution (inset: the enlarged Ni K-edge XANES spectra). d In situ DRIFTS spectra of CO2 and H2O interaction with LaNi-Phen/COF-5 in the dark. e In situ DRIFTS spectra at 1350–2070 cm−1 for detecting the reaction intermediates in CO2 photoreduction and recording the adsorption and activation of CO2 over LaNi-Phen/COF-5 in the presence of H2O under subsequent light irradiation.

Full size image

The adsorbed surface species and CO2-derived intermediates in the CO2RR are dynamically monitored using in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). The time-resolved spectra of LaNi-Phen/COF-5 after introducing humid CO2 in the dark display the characteristic infrared peaks of active ∙CO2– intermediates (νs(O-C-O): 1634 cm−1), bidentate carbonate (b-CO32−, asymmetric CO3 stretching vibration (νas(CO3): 1537–1562 cm−1), monodentate carbonate (m-CO32−, νs(CO3): 1494–1508 cm−1), and bicarbonate HCO3− (σ(CHO): 1439–1462 cm−1). Moreover, the CO2 adsorption band (ν3(CO2)) is indicated by a peak at approximately 3595–3727 cm−1, and the intensities represent the CO2 adsorption process in the LaNi-Phen/COF-5 catalysts (Fig. 5d and Supplementary Fig. 39)61,62. The subsequent CO2RR on the surface of LaNi-Phen/COF-5 catalysts relies on an increased CO2 adsorption capacity. As depicted in Fig. 5e, the generation of CO2* at 1691 cm−1 with increasing light irradiation time implies activation of CO2 through the route of CO2 + e− → CO2*, which is primarily ascribed to the facile transfer of the photogenerated electron to CO2 molecules adsorbed on the LaNi-Phen/COF-5 surface63. The significant enhancements of the peak intensities at 1540 and 1511 cm−1 should be assigned to the COOH* group, which is known as an important intermediate for CO2 reduction64. This signifies that the bimetallic LaNi coordination facilitates the generation of abundant *COOH groups, leading to the effectively decreased activation barrier of CO2 transformation. Moreover, the characteristic peak of CO* absorption at 2036 cm−1 gradually increases with the prolonging of illumination time, which further reveals the origination of the CO product in the photocatalytic CO2RR process. Moreover, the CO2RR process detects the formation of monodentate carbonate (m-CO32−, νs(CO3): 1507 cm−1) and bicarbonate HCO3− (σ(CHO): 1439 and 1701 cm−1)65. Based on these results, CO2*, COOH*, and CO* species should be significant intermediates influencing the photoreduction performance of the LaNi-Phen/COF-5 catalysts66. The process of CO2RR originates from the adsorption of CO2 molecules, followed by the reaction with H+ and the photogenerated electrons to form the intermediate product (*COOH). The *COOH would further lead to the appearing of *CO, and finally CO is desorbed from the catalyst surface. Furthermore, under light irradiation, these intermediates efficiently participate in the CO2 conversion, which is accompanied by a gradual decrease in the intensities of the adsorbed CO2 molecules (Supplementary Fig. 40–41).

Density functional theory (DFT) calculations are carried out to unveil the critical role of LaNi-Phen in the selective photoreduction of CO2 to CO for further investigation of the CO2RR process over LaNi-Phen/COF-5. The CO2 molecule bends after interacting with LaNi-Phen as illustrated in Fig. 6a, demonstrating that C in activated CO2* interacts with the metallic Ni sites. The calculations demonstrate that the formation energy barrier of COOH* on LaNi-Phen/COF-5 is 0.65 eV, confirming that this process is the rate-limiting step. On the contrary, the formation energy barrier of the CO* intermediate is only 0.04 eV, implying that this process is thermodynamically favorable. In addition, CO desorption is thermodynamically preferred over CHO* formation, with an energy barrier of 0.72 versus 1.02 eV, resulting in a promising photocatalytic performance with a high CO selectivity.

Fig. 6: Analysis of the reaction mechanism for the photoreduction of CO2 to CO.
figure 6

a Reaction pathways for CO2 photoreduction on LaNi-Phen/COF-5 with corresponding geometry structures. b HOMO-LUMO charge-transfer transitions for LaNi-Phen. c Schematic energy level diagrams and possible reaction mechanism of LaNi-Phen/COF-5 with pH value of 7. d Schematic illustration of charge transfer in the photocatalytic CO2 reaction over LaNi-Phen/COF-5.

Full size image

The HOMO-LUMO charge-transfer transitions of LaNi-Phen are also analyzed to further understand the role of atomic La in the CO2RR enhancement mechanism (Fig. 6b). In particular, we demonstrate that the appropriate electronic characteristics of the La-Ni dual-atomic sites in LaNi-Phen/COF-5 are responsible for providing the electrons for the CO2 photoreduction. Among them, the HOMO energy level in LaNi-Phen is mainly located on Ni-Phen, whereas the LUMO level is on La-Phen, indicating that La-Phen in LaNi-Phen produces the necessary driving force for electron migration from the COF-5 colloid to the bimetallic La-Ni sites. The La atoms act as the optically active center and electron donor, continuously supplying photogenerated electrons to the LaNi-Phen/COF-5 system, while the COF-5 colloid acts as an electron bridge, directing to Ni atoms for the CO2 photoreduction. Moreover, La-Phen exhibits an electrophilic LUMO level after transferring the photogenerated electron to the COF-5 colloid, leading to the spontaneous replenishment of the excess electrons in Ni-Phen back to La-Phen. This method regulates the product selectivity while enabling closed-loop utilization of the photoinduced charges. The aforementioned result is fully compatible with the proposed efficient charge transfer-induced photocatalysis in LaNi-Phen/COF-5, which combines photoexcited charge-directed transfer with active CO2 adsorption.

The corresponding energy level structure and the hypothetic mechanism for the photocatalytic CO2 reduction with LaNi-Phen/COF-5 are illustrated in Fig. 6c. Notably, the CBM of LaNi-Phen/COF-5 (−1.33 V) is more negative than E0(CO2/CO) (−0.53 V) versus NHE. Electrons are photoexcited from the VBM to the CBM of LaNi-Phen/COF-5, which enables the reduction of adsorbed CO2 molecules to CO. Besides, for the oxidation half-reaction, the photoexcited holes in the VB of LaNi-Phen/COF-5 are consumed by the electrons provided by BIH, in which H2O is oxidized to O2 or H+. Combining the results of in situ characterization and theoretical calculations allows us to corroborate the predicted CO2RR mechanism of the LaNi-Phen/COF-5 diatomic photocatalyst (Fig. 6d), in which La atoms operate as optically active centers to promote the directional migration of photogenerated carriers and Ni atoms serve as functional catalytically active sites for the adsorption of activated CO2. The design advantage of the DACs system is the appropriate combination of light absorption and catalytic reaction processes on spatially close bimetallic centers supported on COF-5, which can efficiently overcome the longstanding problem of insufficient light absorption capacity for achieving high catalytic efficiency and CO-selectivity.

Discussion

We present a facile and scalable electrostatically driven self-assembly assisted by Phen ligands to develop a COF-supported photocatalyst comprising La-Ni dual-atom active centers, in which La atoms serve as light-harvesting centers and Ni atoms provide high catalytic activity for the CO2 photoreduction. The presented self-assembly process integrates both the La and Ni atoms into the COF-5 scaffold in single-site forms. The efficient charge transfer between the La-Ni double-atomic sites has been demonstrated using in situ characterizations and theoretical calculations. Specifically, photoelectric characterizations revealed that they accelerate the dynamic behavior of the photogenerated charge carriers, with Ni sites acting as the catalytic centers to promote CO2 photoreduction through COOH* formation and CO* dissociation and La sites favoring photogenerated electron transfer and long-lived charge separation. Therefore, LaNi-Phen/COF-5 exhibits a remarkable CO yield of 605.8 μmol g−1 h−1 and a high CO selectivity of 98.2% in the absence of any photosensitizer. This work highlights a approach for fabricating dual-atom photocatalysts with well-defined metallic centers inside porous COFs, in which the diatomic sites effectively combine light absorption and catalytic reaction cooperatively to achieve high CO2RR performance. The straightforward preparation of the described photocatalyst paves the way for the development of related systems with tailored characteristics for the desired catalytic application.

Methods

Materials

LaCl3·6H2O (99%), NiCl2·6H2O (99%), and 1,10-Phenanthroline (Phen) were purchased from Aladdin. 2,3,6,7,10,11-Hexahydroxytriphenylene (HHTP) and 1,4-phenylenediboronic acid (PBBA) were provided by Jilin Chinese Academy of Sciences-Yanshen Technology Co., Ltd. as the precursors for COF-5 colloid. Acetonitrile (≥99.5%), mesitylene (98%), and 1,4-dioxane (≥99.5%) were supplied by Aladdin. 1,3-Dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) was obtained from Bide Pharmatech Ltd. All reagents were of analytical grade and used without further purification.

Synthesis of COF-5 colloid

Colloidal COF-5 was synthesized by modifying the reported procedures67,68. Typically, HHTP (5 mM) and PBBA (7.5 mM) were dissolved in a mixture of acetonitrile: 1,4-dioxane: mesitylene (80:16:4 by volume; 20 mL) and sonicated for 3 min. The solution was subsequently filtered (0.45 μm PTFE) to remove insoluble particles. The solution was then heated to 90 °C without stirring for 24 h under atmospheric pressure. Finally, the resulting COF colloids were thoroughly washed three times with 10 mL acetone and resuspended with the aforementioned solvent mixture to obtain the final stable COF-5 colloidal suspension.

Synthesis of LaNi-Phen/COF-5

La and Ni ions were incorporated into the COF-5 colloid through adsorption and chelation with LaCl3·6H2O, NiCl2·6H2O, and 1,10-phenanthroline to synthesize LaNi-Phen/COF-5. The COF-5 colloid (3.3 mL, 10 mg) was dispersed in acetonitrile (5 mL) and stirred for half an hour to prepare the precursor solution. Afterward, NiCl2·6H2O (0.52 mg, 2.2 μmol) was added to the aforementioned solution, which was then stirred for 30 min under an N2 atmosphere. Then, Phen (18 mg, 0.1 mmol) was introduced to the precursor solution and stirred for 1 h to chelate Ni ions and prevent their aggregation. Finally, LaCl3·6H2O (1.2 mg, 3.3 μmol) was added to the above-mentioned Ni-ion-containing solution and stirred for 12 h at 25 °C under an N2 atmosphere. LaNi-Phen/COF-5 precipitates were collected through centrifugation and then dried for 24 h at 60 °C in a vacuum oven.

Synthesis of Ni-Phen/COF-5

Following a similar synthetic procedure to that of LaNi-Phen/COF-5, Ni-Phen/COF-5 catalyst was synthesized using 1.3 mg of NiCl2·6H2O (5.5 μmol), without the addition of any other metal salts.

Synthesis of La-Phen/COF-5

Following a similar synthetic procedure to that of LaNi-Phen/COF-5, La-Phen/COF-5 was synthesized using 1.9 mg of LaCl3·6H2O (5.5 μmol).

Synthesis of Mix-LaNi-Phen/COF-5

The physical mixture of LaNi-Phen/COF-5 (labeled as Mix-LaNi-Phen/COF-5) was prepared by physically mixing 10 mg of La-Phen/COF-5 and 10 mg of Ni-Phen/COF-5.

Characterization

Powder X-ray diffraction (XRD) patterns of the samples were recorded on a desktop X-ray diffractometer (MiniFlex600/600-C) with Cu Kα radiation (λ = 1.5418 Å) in the range of 2θ from 2° to 40°. The morphology and microstructure of the samples were examined using High-resolution Transmission Electron Microscope (HRTEM, JEM-2100F, JEOL, Japan) and an Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM, Titan Cube Themis G2 300). The X-ray absorption structures at the Ni K-edge and La L3-edge of the LaNi-Phen/COF-5 were acquired in fluorescence excitation mode using a Lytle detector at the BL14W1-XAFS beamline of the Shanghai Synchrotron Radiation Facility (SSRF). The photon flux of the BL14W1-XAFS beamline is 5 × 1012 photons.s−1 at 10 keV, with a Si(111) DCM and a beam size of 100 × 200 μm2. The k3-weighted χ(k) data in k-space was Fourier-transformed to R space using Hanning windows in the Athena software to separate the EXAFS contributions from different coordination shells. The wavelet transform (WT) of EXAFS spectra was calculated using the Hama Fortran program. The XAFS fitting data was obtained using Artemis software to corroborate the local atomic structure and coordination environment of La and Ni atoms in the LaNi-Phen/COF-5 catalyst. The elemental composition of the samples was evaluated using an Energy Dispersive Spectrometer (EDS). X-ray photoelectron spectroscopy (XPS) measurements were performed on a Thermo ESCALAB 250 spectrometer. The position of the C 1s line at 284.8 eV was utilized to correct all XPS spectra. The Fourier transform infrared (FTIR) spectra of the samples were acquired using a Nicolet 6700 spectrometer. The Brunauer–Emmett-Teller (BET) surface area of the samples at 77 K was determined by N2 adsorption and desorption isotherms using an ASAP 2460 system. The CO2 adsorption capacity of the samples was also tested by ASAP 2460 system at 298 K. The diffuse reflectance spectra (DRS) of the samples were acquired using a UV-vis spectrophotometer (Shimadzu, UV-2550, Japan). The energy band structure of the samples was determined by measuring ultraviolet photoelectron spectra (UPS, ESCALAB 250Xi, Thermo Fisher Scientific, USA). The photogenerated charge carrier separation and lifetime of the samples were recorded using Steady-state Photoluminescence (PL) spectra (LabRam HR, HORIBA Jobin Yvon, France). Time-resolved photoluminescence (TRPL) spectra were recorded on a fluorescence lifetime spectrophotometer (Spirit 1040-8-SHG, Newport, US) at an excitation wavelength of 365 nm. All of the electrochemical measurements were performed on a CHI-660e workstation (Shanghai Chenhua Instruments Co.), with Pt wire, Ag/AgCl (saturated KCl), and 0.5 M Na2SO4 solution functioning as the counter electrode, reference electrode, and electrolyte, respectively. Electrochemical impedance spectroscopy (EIS) was measured over the frequencies ranging from 0.1 Hz to 100 kHz with an amplitude of 5 mV. Photocurrent-time (I-t) curves with an interval of 60 s on/off switching were recorded on measured with an applied voltage of 0.2 V vs. Ag/AgCl.

Photocatalytic activities measurements

10 mg of photocatalyst powder and 10 mM of BIH were mixed in a 50 mL solution containing 48 mL of acetonitrile and 2 mL of H2O in a Pyrex glass reaction cell coupled to a CO2 reduction system. After the airtight system was completely evacuated using a vacuum pump (no O2 or N2 was detected by gas chromatography), ~80 kPa of high-purity CO2 (99.999%) gas was injected. After the adsorption equilibrium, a 300 W xenon lamp (~100 mW/cm2) was utilized as the light source to irradiate the photocatalytic cell, and the reaction system was kept at 10 °C by cooling water. Gas chromatography (GC-2030, Shimadzu Corp., Japan) equipped with different chromatographic columns was employed to analyze produced H2 and CO.

Isotope-labeling measurements

The carbon source for the isotope-labeling measurements was 13CO2 gas (Isotope purity, 99%, and chemical purity, 99.9%, Tokyo Gas Chemicals Co., Ltd.) instead of 12CO2 gas (Chemical purity, 99.999%, Showa Denko Gas Products Co., Ltd.). Typically, 10 mg of photocatalysts, 10 mM of BIH, 48 mL of acetonitrile, and 2 mL of water were loaded into the reaction cell. The 13CO2 photoreduction protocol was the same as described above, and the reduction products were further analyzed by gas chromatography-mass spectrometry (JMS-K9, JEOL-GCQMS, Japan and 6890N Network GC system, Agilent Technologies, USA) equipped with two different kinds of columns for detecting the products of 13CO (HP-MOLESIEVE, 30 m × 0.32 mm × 25 μm, Agilent Technologies, USA).

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements

In situ DRIFTS spectra were collected using an FT-IR spectrometer (Nicolet iS50 Thermo Scientific, USA) equipped with a mercury cadmium telluride (MCT) detector. Using an ultra-high vacuum pump to eliminate all of the gases in the reaction cell and adsorbed on the catalyst surface. The reaction cell was then filled with humid ultra-pure CO2 gas (99.999%) or CO (99.999%) for CO2 photoreduction or CO adsorption, respectively. Finally, the UV-vis light was turned on, and the in situ DRIFTS data were collected using a difference value of 0 min in light to avoid signals from the organic ligand in the catalysts.

First-principles-based computational details

We have systematically calculated structural relaxation and electronic characteristics within the framework of Density Functional Theory (DFT) formalism as implemented in Materials Studio (MS). The general gradient approximation (GGA) in the form of a Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional was employed for the self-consistent calculation. The energy cutoff utilized throughout the calculations was set at 700 eV, and the Brillouin zone was sampled using the 2 × 2 × 2 Monkhorst pack for the ionic relaxation of the system. The convergence thresholds for self-consistency and structural relaxation were set at 0.002 Hartree/Å for maximum force, 0.005 Å for maximum displacement, and 1.0 × 10−5 Hartree for energy change, respectively.