Asymmetric gradient orbital interaction of hetero-diatomic active sites for promoting C − C coupling

Diatomic-site catalysts (DACs) garner tremendous attention for selective CO2 photoreduction, especially in the thermodynamical and kinetical mechanism of CO2 to C2+ products. Herein, we first engineer a novel Zn-porphyrin/RuCu-pincer complex DAC (ZnPor-RuCuDAC). The heteronuclear ZnPor-RuCuDAC exhibits the best acetate selectivity (95.1%), while the homoatomic counterparts (ZnPor-Ru2DAC and ZnPor-Cu2DAC) present the best CO selectivity. In-situ spectroscopic measurements reveal that the heteronuclear Ru–Cu sites easily appear C1 intermediate coupling. The in-depth analyses confirm that due to the strong gradient orbital coupling of Ru4d–Cu3d resonance, two formed *CO intermediates of Ru–Cu heteroatom show a significantly weaker electrostatic repulsion for an asymmetric charge distribution, which result from a side-to-side absorption and narrow dihedral angle distortion. Moreover, the strongly overlapped Ru/Cu-d and CO molecular orbitals split into bonding and antibonding orbitals easily, resulting in decreasing energy splitting levels of C1 intermediates. These results collectively augment the collision probability of the two *CO intermediates on heteronuclear DACs. This work first provides a crucial perspective on the symmetry-forbidden coupling mechanism of C1 intermediates on diatomic sites.


Supplementary Information
S4 concentration as CO2 photoreduction experiment, were firstly dispersed on the DRIFTS accessory (two ZnSe windows and one SiO2 window), and then degassed at 120 °C for 6 h. Typical signals of various intermediates were captured subsequently after the introduction of the flowed CO2 and H2O vapors under dark and light irradiation (0 -80 min). To explore the influence of diatomic COF catalysts on the chemisorbed *CO, in-situ DRIFTS of CO gas adsorption were measured. In a typical process, the catalyst film was put in the accessory, which was sealed and purged with N2 for 20 min.
Then CO gas was switched into the system to collect the CO adsorption signal until equilibrium. After purging N2 again, the CO desorption signals were captured from 0 to 90 min. Therefore, the strongest peak is assigned to CO adsorption and desorption, and the peak intensity gradually decreased during N2 purging to remove the adsorbed CO as time extension.

Isotope labeling measurement
The isotope labeling measurement was carried out by using 13 CO2 gas (isotope purity, 99%) instead of pure 12 CO2 gas (Chemical purity, 99.999%) as the carbon source with the same photocatalytic process, and the reaction was conducted for 6 h. The photocatalytic species were separated by gas spectrometry columns into individual substances for detecting the products of 13 CO (HP-MOLESIEVE column, USA) and formate/acetate acids (HP-FFAP column, USA). As the separated substances emerge from the column, they flow into electron ionization equipment to ionize and fragment analyte molecules. The separated gas products were analyzed by mass spectrometry (JMS-K9, JEOL-GCQMS, USA). The helium was used as carrier gas with a flow rate of 0.8 ml L -1 . The mass-to-charge (m/z) ratio of mass scanning mode were set from 2 to 70.

Computational details
The first-principle calculation was performed using the Vienna Ab-Initio Simulation Package based on spin-polarized Perdew-Burke-Ernzerhof functional. Van der Waals correction of Grimme scheme (D2) was used to improve the description of the dispersion interaction between adsorbates and substrates. A vacuum thickness of over 15 Å was added in the z direction to avoid unphysical interactions between periodic images. In order to ensure accurate results, all calculations were conducted with a plane wave cutoff of 500 eV and a 2 × 2 × 1 Monkhorst-Pack k-point. In order to elucidate the structures of these COFs and calculate unit cell parameters, three types of possible 2D structure were generated for ZnPor-RuCuDAC (AA-eclipsed, ABstaggered, and slipped ABC-staggered stacking models) by using the densityfunctional tight-binding (DFTB) method. Pawley refinement was carried out by using Reflex package, which was a commercial-free software package for crystallographic structural analysis from PXRD pattern, implemented in Material Studio 2019 modeling version. Unit cell dimensions were manually resolved from the obtained PXRD pattern positions using the coordinates. Then Pawley refinement was carried out to optimize the lattice parameters iteratively until the RWP values converged and the observed overlay with refined profiles showed good agreements.
Moreover, the possible reduction path for these COFs catalytic systems could be as follows, where the asterisks denote active sites and the vertical arrows represent the produced gas in the intermediate reaction.
However, the parameters of individual building compartments from AC-ADF-STEM (Fig. 1g) are measured with a = b = 1.67 ± 0.1 nm (α = 90 ± 0.3°) in ZnPor-RuCuDAC COF. The 3D electron diffraction tomography of DPC images (Fig. 1h-j) reveal the spatial distances among adjacent Zn, Cu, and Ru are ca. 1.67 nm, which is equably atomic cross-distribution on the ZnPor-RuCuDAC substrate. These results confirm that the COF photocatalysts possibly present ABCD-staggered stacking structure rather than the simple AB-staggered configuration, which is evidenced by the negligible difference obtained in the Pawley refinement results.   The survey XPS ( Supplementary Fig. 13a) indicates that Zn, Ru, Cu, C, and N coexist in the precursors and/or these COFs. The high-resolution Zn2p XPS spectra ( Supplementary Fig. 13b) show that ZnPor has Zn(II) 2p3/2/2p1/2 binding energy (BE) peaks of 1021.81/1044.95 for Zn-N bond, which shows a positive shift for ZnPor-RuCuDAC (~0.18 eV), ZnPor-Ru2DAC (~0.22 eV) and ZnPor-Cu2DAC (~0.23 eV), ascribable to a decreased electron density of the ZnPor cores. The high-resolution Ru3p XPS spectrum ( Supplementary Fig. 13c)  Similarly, CuN3 monomer present BE peaks of Cu(II) 2p3/2/2p1/2 for Cu-N and Ru-Cl bond ( Supplementary Fig. 13d), which show a positive shift for ZnPor-RuCuDAC and ZnPor-Cu2DAC for an increased electron density of CuN3 cores. Contrastingly, as compared to CuN3 monomer, For comparison, metal-deficient COF counterparts of ZnPor-N3 COF (Ru-and Cu-free) and H2Por-RuCuDAC (Zn-free) were synthesized for reaching a comprehensive understanding of the chargetransfer mechanism between the ZnPor unit and dual-atom active sites, and the corresponding picosecond transient absorption spectroscopy (TA) was performed ( Supplementary Fig. 32). After being excited by a pump pulse with a wavelength of 400 nm, the TA spectra of ZnPor-N3 COF Supplementary Information S52 showed a pronounced negative peak at ca. 542 nm, which is assigned to ground-state bleach (GSB) and reflects the excited state relaxation. Except for the GSB peak, an extra positive absorption band at ca. 553-684 nm was observed in the TA spectra of H2Por-RuCuDAC COF at delay time, but it was not detected in the spectra of ZnPor-N3 COF. Furthermore, ZnPor-RuCuDAC exhibits a much broader and more intensive absorption band at 552-736 nm. This finding demonstrates that the fluctuant TA of H2Por-RuCuDAC and ZnPor-RuCuDAC is attributed to the charge transfer between ZnPor (Por) cores and Ru-Cu diatomic sites according to the energy transfer and electron diffusion procedure of MLCT, ILCT, and LMCT. This conjecture can be further confirmed by picosecond TRPL spectra (Supplementary Fig. 33 and Supplementary Table 12) of these COFs, where the time decays are fitted with a double exponential function (ΔA(t) = ΔA0 + A1e −t/τ1 + A2e −t/τ2 ), resulting in one component with a shorter lifetime (τ1, contributing radiative fluorescence quenching) and another component with a longer lifetime (τ2, reflecting nonradiative recombination). As seen, the ZnPor-RuCuDAC showed the fastest decay lifetime than ZnPor-N3 COF and H2Por-RuCuDAC, especially for τ2 lifetime with 49.6% percentage decay. The average fluorescence lifetimes (τave)