Morphological and structural characterization. La-Ni bimetallic sites are incorporated into COF-5 colloid through a facile electrostatic-driven self-assembling 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). 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)42. The morphology of COF-5, as illustrated in Supplementary Figs. 3 and 4 displays a one-dimensional (1D) nanorods structure with a width of 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 (Fig. 2b). Importantly, as illustrated in Fig. 2c, the atomically dispersed La and Ni ions in COF-5 colloid are visible using aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM). X-ray photoelectron spectroscopy (XPS) results reveal that the valence states of the Ni and La ions are + 2 and + 3, respectively (Supplementary Fig. 6, and Supplementary Note 1). 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 Fig. 7). The formation of metal-nitrogen bonds is further corroborated by the slight bathochromic shift of the FTIR bands, as depicted in Supplementary Fig. 8 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. 9 and Supplementary Table 1). Notably, the COF-5 colloid in LaNi-Phen/COF-5 comprises a high specific surface area, open pore structure, and outstanding CO2 capture capability, ensuring good access to the La-Ni active sites by CO2 molecules in the photocatalytic process (Supplementary Fig. 10–13 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 (Supplementary Fig. 14–16 and Supplementary Note 4). 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.
The atomic structure and coordination environment of LaNi-Phen/COF-5 are explored further using X-ray absorption fine structure (XAFS) analysis. The X-ray absorption near-edge structure (XANES) spectrum is recorded to confirm the positive charge of La-Ni species due to their near-edge shoulder positioned between metal foils and metal oxides (Supplementary Fig. 17a-b). This is presumably attributed to the strong chelation effect of Phen and close interaction with COF-5 colloid. Furthermore, the dominant Ni-N and La-N peaks in the Fourier transform (FT) extended X-ray absorption fine structure (EXAFS) spectra of LaNi-Phen/COF-5 are near 1.57 and 1.72 Å, respectively (Fig. 2d, 2e), with no emergence of the characteristic peak of Ni-Ni (≈ 2.15 Å) and La-La bonding (≈ 3.94 Å). In other words, the atomic dispersion of La and Ni sites, as well as the absence of metal nanoparticles or clusters in LaNi-Phen/COF-5, are fully confirmed. 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. 2f). 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). To reinforce these results, wavelet transform (WT) analysis is performed to discriminate different atoms within one atomic shell by resolving backscattering wave function centers in energy space. When compared with metal foil and metal oxides, the WT plots of LaNi-Phen/COF-5 exhibit only one maximum intensity at ca. 4.0 Å−1 (Ni K-edge) and 7.0 Å−1 (La L3-edge), which are primarily attributed to Ni-N and La-N contributions, indicating that La-Ni ions in LaNi-Phen/COF-5 exist as isolated single atoms without the presence of metallic crystalline species (Figs. 2g, 2h). La-N and Ni-N scattering at the first shell over LaNi-Phen/COF-5 with coordinated structures of La-N4 and Ni-N2 are further confirmed by the EXAFS fitting analyses (Fig. 2i, Supplementary Fig. 18–20, and Supplementary Table 2), which is consistent with the results of DFT optimization (Supplementary Fig. 21). Free LaNi-Phen was assembled to demonstrate the involvement of the pore confinement effect of COF-5 colloids in facilitating the formation of the La-N4/Ni-N2 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 synergies because they are fully coordinated in the free state (Supplementary Fig. S21 and 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 synergistic effect of LaNi-Phen in COF-5 colloids.
CO 2 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. 3a and 3b). 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 demonstrate the strong synergistic effect of the self-assembled La-Ni diatomic sites. 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). Additionally, 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 outstanding CO2RR performance in this work can be attributed to COF-5’s intrinsic structural and electronic characteristics, as well as the synergistic effect of the self-assembled La-Ni diatomic sites. 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 BIH and H2O operate as electron sacrificial agents and proton sources, respectively (Fig. 3c, Supplementary Figs. 22 and 23). 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. As illustrated in Fig. 3d, the total ion chromatographic peak at ~ 7.45 min corresponds to CO, which generates three signals in the mass spectra (MS). 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. 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. 3e). Furthermore, the heterogeneity tests demonstrate the heterogeneous nature of the LaNi-Phen/COF-5 catalysts (Supplementary Fig. 24 and Supplementary Note 5). 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. 25–27). The catalyst’s microstructure was retained during the CO2RR process (Supplementary Figs. 28 and 29), 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 6). 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. 30)43. 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. 31 and Supplementary Table 6)44. Moreover, as evidenced by investigations of the transient photocurrent response and electrochemical impedance spectroscopy data, LaNi-Phen/COF-5 exhibits excellent charge separation efficiency of the photoinduced electron-hole pairs and low transfer resistance of the photogenerated charges (Supplementary Note 6 and Supplementary Figs. 32 and 33).
Reaction mechanism of CO 2 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. 4a), whereas no obvious CObridge band is observed when using single La or Ni sites on COF-5 (Fig. 4b)45,46. 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 synergistic effects discussed above as the source of the observed outstanding photoelectric performance of the LaNi-Phen/COF-5 catalyst. In-situ XAFS measurements are performed to dynamically monitor the oxidation state of active sites to better understand their function on the atomic scale47. 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 CO2. 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. 34). 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.
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. 4d and Supplementary Fig. 35)48. The subsequent CO2RR on the surface of LaNi-Phen/COF-5 catalysts relies on an increased CO2 adsorption capacity. As depicted in Fig. 4e, 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 surface49. Additionally, there is a significant increase in the intensities of the vibrational peaks at 1540 and 1511 cm− 1, assigned to the COOH* group50. During the CO2RR process, the CO* absorption band at 2,036 cm− 1 reveals the source of CO production in the photocatalytic system. 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)51. Based on these results, CO2*, COOH*, and CO* species should be significant intermediates influencing the photoreduction performance of the LaNi-Phen/COF-5 catalysts. 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. 36).
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. 4f, demonstrating that C and O in activated CO2* interact with the metallic Ni and La sites, respectively. The calculations demonstrate that the formation energy barrier of COOH* on LaNi-Phen/COF-5 is 0.74 eV, confirming that this process is the rate-limiting step. On the contrary, the formation energy barrier of the CO* intermediate is only 0.07 eV, implying that this process is thermodynamically favorable. Additionally, CO desorption is thermodynamically preferred over CHO* formation, with an energy barrier of 0.87 versus 1.14 eV, resulting in a promising photocatalytic performance with a high CO selectivity.
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. 4g). 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 synergistic photocatalysis in LaNi-Phen/COF-5, which combines photoexcited charge-directed transfer with active CO2 adsorption.
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. 4h), 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 novel 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.