Synergistic electroreduction of carbon dioxide to carbon monoxide on bimetallic layered conjugated metal-organic frameworks

Highly effective electrocatalysts promoting CO2 reduction reaction (CO2RR) is extremely desirable to produce value-added chemicals/fuels while addressing current environmental challenges. Herein, we develop a layer-stacked, bimetallic two-dimensional conjugated metal-organic framework (2D c-MOF) with copper-phthalocyanine as ligand (CuN4) and zinc-bis(dihydroxy) complex (ZnO4) as linkage (PcCu-O8-Zn). The PcCu-O8-Zn exhibits high CO selectivity of 88%, turnover frequency of 0.39 s−1 and long-term durability (>10 h), surpassing thus by far reported MOF-based electrocatalysts. The molar H2/CO ratio (1:7 to 4:1) can be tuned by varying metal centers and applied potential, making 2D c-MOFs highly relevant for syngas industry applications. The contrast experiments combined with operando spectroelectrochemistry and theoretical calculation unveil a synergistic catalytic mechanism; ZnO4 complexes act as CO2RR catalytic sites while CuN4 centers promote the protonation of adsorbed CO2 during CO2RR. This work offers a strategy on developing bimetallic MOF electrocatalysts for synergistically catalyzing CO2RR toward syngas synthesis.

electrocatalytic CO 2 RR activity toward syngas synthesis; hereby one metal center will show high selectivity for CO 2 -to-CO conversion while the other metal center will be utilized for H 2 generation due to its low binding energy of CO and high proton generation rate.
Herein, a 2D c-MOF electrocatalyst with bimetallic centers is synthesized by solvothermal approach for electrocatalytic CO 2 RR. This 2D c-MOF consists of phthalocyaninato copper as the ligand and zinc-bis(dihydroxy) complex (ZnO 4 ) as the linkage, named as (PcCu-O 8 -Zn). The electrochemical measurements indicate that PcCu-O 8 -Zn exhibits highly selective catalytic activity for CO 2to-CO conversion (88%) and high turnover frequency (TOF) of 0.39 s −1 at −0.7 V vs. RHE and excellent stability. Syngas compositions with different molar H 2 /CO ratio (from 1:7 to 4:1) can be tuned via varying the metal centers (Cu and Zn) of ligand/ linkage as well as applied potentials. Operando X-ray absorption spectroscopy (XAS) and surface-enhanced infrared absorption (SEIRA) spectroelectrochemistry are utilized to probe the catalytic sites and the reaction process. The spectroscopic studies combined with contrast experiments and density functional theory (DFT) calculation reveal that ZnO 4 complexes in the linkages of PcCu-O 8 -Zn exhibit high catalytic activity for CO 2 -to-CO conversion, while CuN 4 complexes in the Pc macrocycles act as the synergetic component to promote the protonation process and hydrogen generation along with the CO 2 RR. Thus, the bimetallic active sites contribute to a synergistic effect on the CO 2 RR. Our work highlights the bimetallic MOF electrocatalyst for highly selective CO 2 RR.

Results
Material design and reaction energetics. Density functional theory calculations were firstly employed to optimize the electrocatalyst design by simulating the reaction energetics of CO 2 Tables 1-7). Typically, the electrochemical CO 2 -to-CO reduction steps include the first proton-coupled electron transfer to generate a carboxyl intermediate (*COOH), and subsequently the second charge transfer (one electron and one proton) for the formation of *CO intermediate, as well as the desorption of CO for the final CO product (Eqs. 1-3 in Supplementary Methods) 30 [5][6][7], as confirmed by Fouriertransform IR (FT-IR) spectroscopy and powder X-ray diffraction (XRD) measurements. The disappearance of the ligand OH signals (3300 and 630 cm −1 ) and the peak shift from 1288 cm −1 (C-OH) to 1270 cm −1 (C-O-Zn) in the FT-IR spectra ( Supplementary  Fig. 8) demonstrate the successful coordination of O to Zn atoms 34 . The XRD pattern (Fig. 2b) shows intense peaks at 5.0°, 7.1°and 10.1°, assignable to (100), (110) and (200) plane, respectively, which indicates the long-range order within the ab plane 35 . The broad peak at 27.3°originates from the weak longrange stacking along the c direction with a layer distance of 0.33 nm, which is a typical feature of layered MOFs 36 . Compared to the calculated structures, the observed XRD pattern of PcCu-O 8 -Zn is in a good agreement with the AA staggered stacking geometry. Scanning electron microscopy (SEM, Supplementary Fig. 9) images indicate aggregated nanosheets in the resulting MOF samples.
Transmission electron microscopy (TEM) images also present a mass of MOF nanosheets with an average size of 24 nm (Fig. 2c). The selected area electron diffraction pattern (SAED, inset image in Fig. 2c) and the high-resolution TEM (HR-TEM, Fig. 2d) images further manifest the crystalline structure of PcCu-O 8 -Zn based on a square lattice of 1.75 nm.
Element mapping images (Supplementary Fig. 10) disclose the homogenous distribution of Cu, Zn, C, N and O in the PcCu-O 8 -Zn sample. Furthermore, X-ray photoelectron spectroscopy (XPS) analysis also confirms the presence of Cu, Zn, C, N and O elements ( Supplementary Fig. 11). In the high-resolution Cu 2p spectrum, the set of peaks at 936.7 and 953.8 eV is assigned to Cu 2p 3/2 and Cu 2p 1/2 , respectively, which suggests one type of oxidized Cu (II) in the PcCu-O 8 -Zn 37 . The deconvolution of N1s spectra further verifies the coordination of Cu and N 38 . For the high-resolution scan of the Zn 2p region, the typical feature of Zn (II) is found 39 .
To further investigate the chemical state of the Cu and Zn atoms in the PcCu-O 8 -Zn sample, XAS and extended X-ray absorption fine structure (EXAFS) analyses were performed. The Cu K-edge X-ray absorption near-edge structure (XANES) spectra ( Supplementary Fig. 12) show that both PcCu-O 8 -Zn and the monomer PcCu-(OH) 8 exhibit a typical Cu(II) peak at 8985 eV (1s to 3d electron transition), which is similar to that of the reference copper(II) phthalocyanine (PcCu), thus confirming the presence of Cu-N in PcCu-O 8 -Zn 40,41 . Generally, two characteristic signals are observed in the Zn XANES spectra including the pre-edge peak at around 9660 eV and the main absorption peak at 9660−9680 eV, which correspond to the electron transition from 1s to 3d (typically found for the transition metal Zn) and the 1s to 4p electronic transition,   Low-pressure N 2 sorption was measured to evaluate the porous properties of PcCu-O 8 -Zn ( Supplementary Fig. 13). The Brunauer Emmett Teller surface area was measured to be 378 m 2 g −1 . The pore size distribution indicates its abundant micropores (1.4 nm) and mesopores (6 nm), which can be favorable for the mass transport during the catalytic process 28 .  Fig. 18), demonstrating its feasibility for CO 2 RR. To verify that the currents originate from the catalytic CO 2 RR, constant potential electrolysis was performed. The products were detected via gas chromatography (GC) and nuclear magnetic resonance (NMR) measurements.
The results indicate that only gaseous (H 2 and CO) products were generated at the applied potentials with total Faradaic efficiency of 99 ± 2.2% ( Supplementary Figs. 19-21). The resultant CO 2 RR catalytic performance including the maximum CO efficiency and the molar CO/H 2 ratio suggested strong dependence on the type of metal centers and applied potential (Fig. 3a, b and Supplementary Fig. 20). Among the synthesized 2D c-MOFs/CNT hybrids, the PcCu-O 8 -Zn/CNT sample yielded the highest partial current density for CO (j CO ) and the highest corresponding Faradic efficiency toward CO (FE co ) over the investigated potential range (Fig. 3a, b), indicating superiority of the ZnO 4 sites for selective conversion of CO 2 to CO over ZnN 4 , CuN 4 and CuO 4 centers, which is also supported by the DFT calculations ( Fig. 1). Notably, j CO for PcCu-O 8 -Zn/CNT showed a maximum value at −1.0 V vs. RHE, while H 2 generation (j H2 ) displayed a steady rise with the increased overpotential ( Supplementary  Fig. 20a). This observation can be attributed to the competitive reactivity between the CO 2 RR and HER as well as the limitation   Table 8). The molar ratio of the syngas CO/H 2 catalytically generated by the 2D c-MOFs could be additionally controlled via the applied potentials. As shown in Fig. 3d, the molar H 2 /CO ratio for the PcCu-O 8 -Zn/CNT system could be tuned from around 1:7 to 4:1 by increasing the applied potential from −0.4 to −1.2 V vs. RHE.
To elucidate the kinetics of these MOFs toward the catalytic CO 2 RR, Tafel slopes were derived (Supplementary Fig. 22a) Fig. 22b). Besides, the PcCu-O 8 -Zn/CNT system presents long-term catalytic durability. The high FE CO (86%) and current density were maintained over the course of 10 h of operation at −0.7 V vs. RHE (Fig. 3e). No obvious changes of morphology and structure ( Supplementary Fig. 23) were observed in SEM image, XRD pattern, Raman and FR-IR spectra of PcCu-O 8 -Zn/CNT after the CO 2 RR long-term testing, demonstrating the high stability of PcCu-O 8 -Zn/CNT during electrocatalytic CO 2 conversion.
Unveiling the active sites. Operando XAS measurement was employed to gain insight into the valence state and coordination structure of Cu and Zn in the PcCu-O 8 -Zn/CNT under the CO 2 RR turnover condition (Fig. 4a-d and Supplementary  Fig. 24). As shown in Fig. 4a, a typical pre-edge signal of Zn(0) at around 9660 eV is not observed in the Zn K-edge XANES spectra for all PcCu-O 8 -Zn/CNT samples 42,43 . This excludes the generation of metallic Zn in PcCu-O 8 -Zn/CNT electrocatalyst during the CO 2 RR process. Importantly, the main absorption peak at 9665 eV was not shifted in the Zn K-edge XANES spectra of PcCu-O 8 -Zn/CNT (Fig. 4a) as the applied potential was decreased to −0.4 (red circle dot curve) and −0.7 V (blue diamond dot curve) vs. RHE, respectively, and then increased back (indigo triangle dot curve) to the initial (black square dot curve) open circuit voltage (OCV). The results reveal that the oxidation state of Zn(II) in PcCu-O 8 -Zn/CNT was maintained throughout the catalytic process, which can be explained by the fact that the Zn(II) already has a full 3d electron shell 46 . In addition, the preedge peak at 8985 eV and the main absorption peak at 8998 eV in the Cu K-edge XANES spectra of PcCu-O 8 -Zn/CNT were not varied upon changing the applied potential, which indicates that the valence state of Cu(II) was not changed during the CO 2 RR process. Notably, the missing pre-edge peak at 8980 eV in Cu Kedge XANES spectra of all PcCu-O 8 -Zn/CNT samples further confirms that no metallic Cu was generated at the PcCu-O 8 -Zn/ CNT electrode under electrolysis condition (Fig. 4b).
To monitor the local coordination environment changes, in situ EXAFS measurements were performend. As the applied potential was performed for one cycle, the peak at 1.55 Å assigned as Zn-O bond length in PcCu-O 8 -Zn/CNT was not shifted (Fig. 4c). Meanwhile, the peak intensity presents a negligible decrease (black square dot and indigo triangle dot curves in Fig. 4c), which is possibly due to the interaction of the reaction intermediates and the ZnO 4 sites during the catalytic process, such as *H, *COOH, *CO and so on 43 . Therefore, the above in situ EXAFS results reveal no obvious change in Zn coordination number and bond length of Zn-O for PcCu-O 8 -Zn/CNT under the electrolysis condition. Furthermore, the characteristic signal of Zn−Zn bonding at 2.27 Å does not appear in the EXAFS spectra of all the PcCu-O 8 -Zn/CNT samples, again excluding the formation of metallic Zn or Zn cluster at PcCu-O 8 -Zn/CNT catalyst throughout CO 2 RR process. Regarding the CuN 4 complexes, no obvious change of the Cu-N coordination peak at 1.54 Å was detected in the Cu K-edge EXAFS spectra of PcCu-O 8 -Zn/CNT (Fig. 4d) upon performing the potential in one cycle. Additionally, no obvious signal of Cu −Cu bonds was observed at 2.23 Å, which demonstrates that no heavy backscattering atoms (Cu) are bound to Cu sites in all PcCu-O 8 -Zn samples. Therefore, the operando XAS results fully prove that the well-defined sites (ZnO 4 and CuN 4 ) act as stable catalytic centers during the CO 2 RR process, while no metals or metal clusters form via the reduction of high-valence metal centers.
Next, operando SEIRA spectroelectrochemistry was employed to elucidate the electrocatalytic mechanism of the 2D c-MOF catalysts. The 2D c-MOFs were evenly deposited as a closed film onto a nanostructured Au surface, which acted as IR signal amplifier. SEIRA spectra were recorded at different potentials covering a broad potential window. SEIRA difference spectra taken under turnover conditions were derived using the spectrum of the respective system at −0.6 V vs. Ag/AgCl (Fig. 4e, f). The SEIRA difference spectra of PcZn-O 8 -Cu/CNT and PcCu-O 8 -Zn/ CNT show distinct features that likely arise from their intrinsically different reactivities (Supplementary Fig. 25). Upon lowering the potential, a negative band at 2343 cm −1 assigned to dissolved CO 2 (g) was observed. This band was found to decrease with decreased potential indicating the consumption of CO 2 near the surface in the catalytic process 45 . Strong positive bands in the region of 1660-1640 cm −1 were observed in both cases and attributed to the changes of the interfacial H 2 O, which accumulated in the MOFs due to catalysis or increasing negative polarization of the electrode. The high-frequency bands above 1800 cm −1 typically arise from metal bound species. Specifically, the bands located in the higher frequency region at 1933 and 2071 cm −1 were assigned to CO bound to the CuN 4 and CuO 4 centers, respectively (Fig. 4e, f) 47 . The shift of the ν(CO) mode could arise from the different electronic properties of Cu metal in the N 4 and O 4 frame, respectively. In this respect, CuN 4 centers can stabilize the CO via π backbonding leading to drastically lowered ν(CO)s, while CO bound to Cu and oxide-derived Cu surfaces has been reported above 2000 cm −1,47,48 . The strong band centered at 1851 cm −1 for the PcZn-O 8 -Cu/CNT system matches the frequency for (isolated) Cu-H and is thus assigned to the Cu-H intermediate formed at the CuO 4 nodes in the HER cascade 47,49 . The particularly high intensity of this band suggests a dominating HER process over CO 2 RR at PcZn-O 8 -Cu/CNT in CO 2 -saturated solution. This interpretation is consistent with the electrocatalytic results ( Supplementary Fig. 20), revealing that the PcZn-O 8 -Cu/CNT system shows high selectivity for H 2 (>90%) over the complete potential range in CO 2 -saturated electrolyte. In contrast, Cu-H is not observed at the CuN 4 units of the PcCu-O 8 -Zn/CNT systems. This may be due to low accumulation of the Cu-H species during catalysis, which could result from the fast proton transfer kinetics at CuN 4 complexes to ZnO 4 sites and yield H 2 . Interestingly, no indication for CO binding to the ZnO 4 nodes was found due to its too low transient concentration to be observed with our current SEIRA spectro-electrochemical setup. This can be explained by the weak binding energy between ZnO 4 and CO, which could facilitate a quick deliberation of the product and thus suggests fast CO 2 RR kinetics at the ZnO 4 complexes in PcCu-O 8 -Zn/CNT.

Discussion
To obtain further insight into the reactivity of 2D c-MOFs towards HER and CO 2 RR, the calculated free energy profiles on M1O 4 site at U = 0.55 V were analyzed (Fig. 5a). For HER, the Gibbs free energy values of the key intermediates (*H) on M1O 4 units are positive, with a minimum barrier of 0.7 eV, and therefore expected to be kinetically prohibited. However, the free energy values of CO 2 RR at the same equilibrium potential are negative, which reveals that the CO 2 RR at M1O 4 site is thermodynamically downhill. It further verifies the favorable CO 2 RR process at M1O 4 complexes of 2D c-MOFs. Although the CuN 4 complexes show the lowest energy barriers for HER, PcCu-O 8 -Zn still exhibits the lowest free energy for the generation of rate-determining *COOH intermediate as compared to the other 2D c-MOFs during CO 2 RR catalysis. This establishes the synergetic effect of CuN 4 and ZnO 4 in enhancing the CO 2 RR activity. A proposed synergistic catalytic scheme is presented in Fig. 5b. CuN 4 complexes attract numerous electrons and H 2 O toward producing abundant protons, wherein protons are partially transformed into molecular H 2 and partially transferred to ZnO 4 complexes. Simultaneously, the adsorbed CO 2 on ZnO 4 complexes is reduced to *COOH by coupling with these protons/ electrons from the CuN 4 sites and electrode/electrolyte, and subsequently the resultant *COOH will be transformed into *CO intermediate by a further charge transfer step (one electron and one proton). The desorption of *CO results in the final CO product. As a result, the kinetics of CO 2 RR on ZnO 4 is greatly enhanced in PcCu-O 8 -Zn 2D c-MOF.
In summary, we have synthesized a layered 2D c-MOF (PcCu-O 8 -Zn) with bimetallic centers (ZnO 4 /CuN 4 ) capable of synergistic electroreduction of CO 2 to CO based on the theory-guided design. The electrocatalytic results indicated that PcCu-O 8 -Zn mixing with CNTs exhibited high CO 2 RR catalytic activity with high selectivity for CO conversion of 88%, TOF of 0.39 s −1 and long-term durability (>10 h), which is superior to the reported MOF-and Zn-based electrocatalysts. The molar H 2 /CO ratio could be rationally adjusted through varying the metal centers and applied catalytic potentials, beneficial for industrial applications. Theoretical calculation and the operando XAS and SEIRA analysis, as well as the control experiments suggested that the CO 2 RR takes place at the ZnO 4 units while the CuN 4 units promote the proton and electron transfer during the reaction process. Thus, the combination of ZnO 4 and CuN 4 complexes generates a synergetic effect, which contributes to the high CO 2 RR performance of PcCu-O 8 -Zn/CNT. Our work demonstrates the capability of bimetallic 2D c-MOFs as highly efficient electrocatalysts for promoting the CO 2 RR, which is of importance for conductive MOFs design and their electrocatalysis application and also sheds light on the development of high-performance bimetal-heteroatom doped carbon electrocatalysts.

Methods
Computational studies. The computational modeling of the reactants, intermediates and products, and reaction process involved in the reactions on 2D MOFs was performed by using DFT with the PBE exchange-correlation functional 50 , as implemented in the VASP code 51,52 . The total energies were converged within 10 −6 eV/cell. The cut-off energy for plane wave basis was set at 500 eV. The Brillouin zone of the supercells was sampled using 4 × 4 × 1 Monkhorst-Pack grid of k-points. All calculations have been performed using the spin-polarized setup. Dispersion interactions were taken into account as proposed by Grimme within the DFT-D2 scheme 53  Characterization. Powder XRD measurements were collected on a PW1820 powder diffractometer (Phillips) using Cu-Kα radiation (λ = 0.15418 nm, 40 kV, 30 mA). TEM images were obtained using a Cs-corrected TEM (Carl Zeiss Libra 200) operated at 200 kV. SEM was recorded on Zeiss Gemini S4 500. Raman spectra were collected with a Renishaw 2000 model confocal microscopy Raman spectrometer with a CCD detector and a holographic notch filter at ambient conditions. FT-IR tests were performed on a Bruker Optics ALPHA-E spectrometer equipped with Attenuated Total Reflectance (ATR) sample holder. The porosity was detected by nitrogen sorption using a micromeritics ASAP 2020 analyzer. XPS spectra were collected with an ESCALAB MK II X-ray photoelectron spectrometer using an Al Kα source. The rotating disk electrode (RDE) was performed on MSR electrode rotator (Pine Instrument Co.). The XAS and EXAFS data were collected at room temperature in transmission mode at beamline BL14W1 and BL15U1 of the Shanghai Synchrotron Radiation Facility (SSRF, China).
Electrode preparation. One milligram of catalyst was added into 100 μl of ethanol containing 10 μl of Nafion solution (5% in ethanol) and ultrasonically treated for 30 min. And the catalyst ink was drop-casted onto carbon paper.
Electrochemical test. Before testing, the Nafion membrane (115) was treated in H 2 O 2 solution (5%) and pure water for 1 h. And the carbon paper with loading catalyst, Pt mesh and Ag/AgCl are used as the working, counter and reference electrode. Firstly, the electrolyte in the cathodic compartment was degassed by bubbling with Ar for at least 30 min for removal of oxygen, and then purged continuously with CO 2 . CO 2 gas was delivered into the cathodic compartment at a rate of 30.00 sccm and was vented directly into the gas-sampling loop of a gas chromatograph. GC run was initiated every 20 min. All reference electrodes are converted to the RHE reference scale using E (vs. RHE) = E (vs. Ag/AgCl) + 0.197 V + 0.0591 V × pH.
The partial current densities of CO and H 2 production were calculated from the GC peak areas as follows: where V CO and V H2 are the volume concentration of CO and H 2 , respectively, P 0 is the standard atmospheric pressure (1.013 bar), T is the absolute temperature (273.15 K), F is Faradaic constant (96,485 C mol −1 ), and A is the electrode area (1 cm 2 ). Faradaic efficiencies for a given product were calculated by dividing these partial current densities by the total current density. The liquid products were analyzed by NMR spectroscopy, in which 0.5 ml of the electrolyte was mixed with 0.1 ml D 2 O and 0.05 μl dimethyl sulfoxide (DMSO), wherein DMSO was serviced as an internal standard. The one-dimensional 1 H spectrum was measured with water suppression using a pre-saturation method.
Operando XAS measurement. Operando XANES and EXAFS experiments were carried out at the BL14W1 beamline of the Shanghai Synchrotron Radiation Facility (SSRF). All data were collected in fluorescence mode under applied potential controlled by CHI electrochemical workstation. A custom-designed cell ( Supplementary Fig. 21) was used for the in situ XAS measurements, which was applied to the identical conditions as the real CO 2 RR testing. The X-ray energy was calibrated using a Cu metal foil and Zn metal foil.
Operando SEIRA spectro-electrochemistry. All measurements were conducted in aqueous CO 2 saturated 0.1 M KHCO 3 . An FT-IR spectrometer (Bruker IFSv66) equipped with a N 2 -cooled MCT detector was employed. The measurements were carried out in attenuated total reflection (ATR) mode in Kretschmann geometry using an Si prism as IR active waveguide. A thin and nano-scale rough Au layer was coated onto the prism for conductivity/contacting purposes prior to MOF deposition/drop-casting. Deposition of the Au film is described elsewhere 54 . MOF drop-casting followed procedures as described above. The Au layer acted as a signal amplifier giving rise to strong surface-enhancement of IR signals of compounds close to the Au surface. In this way, we achieve to record SEIRA spectra of the MOF layers close to the electrode surface, which should exhibit excellent electronic contact. For applying potentials, the MOF-coated prism was mounted into a customized three-electrode containing spectro-electrochemical cell as described elsewhere 54 . A hydrogen-flamed cleaned Pt wire and Ag/AgCl in 3 M KCl (DriRef, World Precision Instruments) acted as counter and reference electrode, respectively.