Dynamic restructuring of coordinatively unsaturated copper paddle wheel clusters drives CO2 reduction eciently

Coordinatively unsaturated metal sites in metal–organic frameworks (MOFs) hold promises for improving activity and selectivity of catalyst. As the building unit of many MOFs, copper paddle-wheel (CPW) is of great interest to function as activity sites for CO 2 reduction due to the ability of copper cations to attract and activate CO 2 molecules. However, little is known about the real structure of these active sites for their dynamic structural changes under realistic operating conditions. Here, we apply to design defect metal−organic framework HKUST-1 with coordinatively unsaturated copper paddle wheel through a facile “atomized trimesic acid” strategy, reveal its dynamic behaviour during electrochemical reconstruction and prove its superior CO 2 reduction activity. Through comprehensive analysis of various in situ spectroscopy characterizations, it is demonstrated that unsaturated copper paddle wheel clusters [CU-CPWC, Cu 2 (HCOO) 3 ] are nally formed during electrochemical reconstruction and act as the real active sites. Mechanistic studies based on density functional theory calculations reveal that the higher d band center in CU-CPWC compared with that in CPW accelerates the electron transport from copper atoms to CO 2 molecules, which can be exploited to improve the CO 2 reduction performance.


Background
Coordinatively unsaturated metal sites in metal-organic frameworks (MOFs) hold promises for improving activity and selectivity of catalyst. As the building unit of many MOFs, copper paddle-wheel (CPW) is of great interest to function as activity sites for CO 2 reduction due to the ability of copper cations to attract and activate CO 2 molecules. However, little is known about the real structure of these active sites for their dynamic structural changes under realistic operating conditions. Here, we apply to design defect metal−organic framework HKUST-1 with coordinatively unsaturated copper paddle wheel through a facile "atomized trimesic acid" strategy, reveal its dynamic behaviour during electrochemical reconstruction and prove its superior CO 2 reduction activity. Through comprehensive analysis of various in situ spectroscopy characterizations, it is demonstrated that unsaturated copper paddle wheel clusters [CU-CPWC, Cu 2 (HCOO) 3 ] are nally formed during electrochemical reconstruction and act as the real active sites. Mechanistic studies based on density functional theory calculations reveal that the higher d band center in CU-CPWC compared with that in CPW accelerates the electron transport from copper atoms to CO 2 molecules, which can be exploited to improve the CO 2 reduction performance.
Due to the constituents' geometry, size, and functionality can be exibly varied, metal-organic frameworks (MOFs) materials represent ideal platforms that can control the interplay between e ciency, activity and selectivity in a single catalytic system. [1][2][3][4] HKUST-1 is the prime example of MOFs based on copper paddle wheel (-Cu 2 (COO) 4 , metal node) and 1,3,5-benzene-tricarboxylic acid (H 3 BTC, organic link), which has the potential to be used in whole catalysis world . 5 Interest in the reactivity of metal copper and copper based MOFs stems from its unique electrochemistry CO 2 reduction selectivity toward higher-value hydrocarbons beyond CO. [6][7][8][9][10][11][12][13][14][15] Defect engineering in MOFs, such as missing linker defects, has attracted considerable research attention since it offers a means to construct coordinatively unsaturated metal sites, which is generally considered to be the active sites, and thus has important implications for improving the catalytic performance. [16][17][18][19][20][21][22][23] Therefore, constructing coordinatively unsaturated copper paddle wheel in HKSUT-1 may be served to facilitate the catalytic activity. So far, there has been rare report about it.
A critical question remaining to be addressed here is that little is known about the real structure of these active sites. Exploring the dynamic restructuring of electrocatalysts has been of interest for a long time, which is crucial to understand structure-reactivity correlations in many catalysis systems. [24][25][26][27][28][29] But it has proved challenging to relate local electronic and geometric structure of these open metal sites to catalytic function. The di culty arises from the dynamic change of electrocatalyst structure in local coordination under electrochemical reactive environments. 10,30 The changes in structure and surface composition control catalytic function. Especial challenge is identifying the real active sites of MOFs-based catalysts owing to the complexity of their structure, but it may lead to shift the catalyst design from trial-and-error paradigm to a rational approach.
Herein, we report a defect-engineered metal−organic framework HKUST-1 with coordinatively unsaturated copper paddle wheel, which is named "quasi-MOF (QMOF)", as pre-catalyst for electrocatalytic CO 2 reduction and reveal its dynamic behaviour during electrochemical reconstruction. Several unique aspects of this work are worth highlighting up in front: (1) defect metal−organic framework HKUST-1 with coordinatively unsaturated copper paddle wheel was synthesized through a facile "atomized trimesic acid" strategy; (2) unsaturated copper paddle wheel clusters [CU-CPWC, Cu 2 (HCOO) 3 ] were identi ed as real catalytic active site thanks to experimental results using a series of quasi-in situ and in situ characterization methods; (3) a boost in reaction activity for CO 2 reduction was observed owing to the structure with the superior CO 2 activation ability.

Results
Synthesis and structural characterization of QMOF. Previous investigation has indicated that the crystallization of the MOFs undergoes Ostwald ripening process, during which unsaturated metal sites (UMS) may be observed. 31 With this in mind, retarding the crystal ripening process can generate more UMS in MOFs, so as to obtain more open metal sites for electrocatalytic reaction. The synthetic procedures of catalysts are illustrated in Figure 1a and b, and the experimental details are summarized in Supplementary Information. Cu(OH) 2 nanowires (NWs) with smooth surface were rst grown on Cu foil based on a reported chemical oxidation method. 32 Subsequently, Cu(OH) 2 sample was reacted with  Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images con rmed the synthesis of a smooth-surface nanowire structure (Supplementary Figure S1 and S2). X-ray diffraction (XRD) patterns veri ed the presence of a crystalline Cu(OH) 2 phase (Figure 1c), which was further evidenced by Raman spectroscopy (Figure 1d). Note that no diffraction peaks for HKUST-1 were detected in QMOF (Figure 1c), suggesting an amorphous structure may be formed.
In order to investigate the surface structure of samples, we performed Raman and Fourier transform infrared (FT-IR) analysis. The high-frequency region ranging from 700~ 1200 cm -1 is dominated by bands associated with the organic part of HKUST-1 ( Figure 1d). [33][34][35] The band at 1006 cm −1 corresponds to C=C symmetric stretching of benzene ring. At 827 and 748 cm -1 , the out-of-plane ring (C-H) bending vibrations are recognized in QMOF and CMOF. 33,35 QMOF exhibits weaker bands in high-frequency region than CMOF, suggesting UMS of QMOF. In the low-frequency region, the doublet at 458 and 505 cm −1 is related to Cu-O stretching modes involving oxygen atoms of carboxylate bridges, which is consistent with previous reports, suggesting the metal−organic coordination in the samples. 7,35 Vibrational properties of the samples were further investigated by FT-IR spectroscopy. As shown in Figure 1e, the QMOF and CMOF exhibit the signi cant peaks ranging from 1100 to 1800 cm −1 compared with Cu(OH) 2 NWs.
Characteristic peaks assigned to vibrations of the carboxylate group in HKUST-1 are observed: 1370 and 1419 cm -1 corresponding to the symmetric vibrations of -COO groups and 1450, 1592 and 1646 cm -1 attributed to the asymmetric vibrations of -COO groups. 8,34,36 The peak at 1114 cm -1 is assigned to C-O vibration in the HKUST-1. In addition, the peaks at 732 and 762 cm −1 are observed in CMOF, which can be ascribed to the bonding of Cu with BTC linker molecules (Figure 1e). 7 A weak peak at 1545 cm −1 is also found in CMOF, which is due to water coordinated with Cu in the MOF. 34 Meanwhile, for the QMOF, two features of the FT-IR spectrum are worth noting. Firstly, notable shifts to lower wavenumber are observed for the peaks of asymmetric vibrations of -COO groups, suggesting an elongation of the associated bonds. Secondly, the peaks at 1545 cm -1 (water coordinated with Cu in the MOF), 1419 cm -1 (one of the symmetric vibrations of -COO groups) and 732 cm -1 (one of the Cu with BTC linker molecules) in QMOF are undetectable, which indicates that a defective HKUST-1 structure with UMS or distorted metal nodes may be formed in QMOF.
Notably, SEM and TEM images of QMOF reveal that the uniform nanowire structure with many small clusters on the surface (Figure 2a Moreover, the structural evolution of the Cu(OH) 2 NWs in the presence of the atomized H 3 BTC was explored. As shown in Supplementary Figure S3, the controllable structural evolution from QMOF clusters to complete MOF crystal (CMOF) is achieved through controlling the reaction time with "atomized H 3 BTC". In order to prove the superiority of "atomized H 3 BTC" method, common "immersed method" was performed, that is, Cu(OH) 2 NWs precursor was submersed into liquid H 3 BTC solution to synthesize HKUST-1. As a contrast, only CMOF is obtained by the ordinary "immersed method" even in a short time of 10 s (Supplementary Figure S3d). SEM image of CMOF reveals a typical octahedral HKUST-1 morphology (Supplementary Figure S3c), which is consistent with previous reports. [37][38][39] High magni cation SEM and TEM images identify Cu(OH) 2 NWs trace in CMOF (Supplementary Figure S4), further revealing the structural evolution from Cu(OH) 2 NWs to CMOF.
The surface hydrophilicity and CO 2 capture capability of the samples were characterized by contact angle and CO 2 -temperature programmed desorption (TPD) measurements (Supplementary Figure S5 and S6).
Contact angle measurements illustrated that Cu(OH) 2 surface is the most hydrophilic. The surface clusters of QMOF decrease the hydrophilicity of Cu(OH) 2 , but increase the CO 2 capture capability.
Electrochemical reconstruction of QMOF and CMOF. Attention was then paid to the structural changes of QMOF and CMOF during the electrochemical reconstruction process. X-ray photoelectron spectroscopy (XPS) tests under ex-situ (Figure 3a-c) and quasi-in situ (Figure 3d-e) conditions were rst performed to reveal the electronic con guration of each catalyst before and after electrochemical reconstruction.
The ex-situ XPS spectra of C 1s for QMOF and CMOF are shown in Figure 3a, which can be divided into four peaks, including sp 2 carbon in benzene ring (284 eV), sp 3 carbon (C-C, 284.8 eV), C-O (286 eV) and C=O (289 eV). [40][41][42][43] The metal node [-Cu 2 (COO) 4 ] defect of QMOF is further con rmed by the lower binding energy shift of C-O peak compared with that of CMOF. Note that an additional O 1s peak at 533.2 eV is clearly observed in QMOF, which is related to H 2 O molecule ( Figure 3b). 44,45 This result has been con rmed by FT-IR peak at 1545 cm −1 , corresponding to water coordinated with Cu in the MOF, as described above. O 1s spectra of QMOF and CMOF both display two peak regions at approximately 530.5 and 531.8 eV, which can be assigned to the Cu-O and -COO groups. 43 As expected, the QMOF sample exhibits the lower binding energy of Cu-O and -COO than CMOF for the defect of metal nodes [-Cu 2 (COO) 4 ] in QMOF. XPS data for the Cu 2p 3/2 exhibit Cu 2+ and Cu 1+ characteristic peak (Supplementary Figure S7). 41,46 Previous report suggests the conversion of Cu 2+ to Cu 1+ in CMOF is induced by the exposure of MOF to X-ray source and charge neutralizer during XPS data collection. 46 Note that Cu Auger (L 3 M 45 M 45 ) region shows that a reduced oxidation state for the Cu species in QMOF, which may be attributed to the lower coordinatively unsaturated Cu atoms ( Figure 3c). 47 Furthermore, the electronic structures and chemical bonding of Cu atoms in their local environment were investigated by X-ray absorption spectroscopy (XAS), including X-ray absorption near-edge structure (XANES) and extended X-ray absorption ne structure (EXAFS) spectroscopy. The QMOF sample exhibits lower-energy edge position as compared with CMOF, indicating a reduced oxidation state for the Cu species (Supplementary Figure S8), which is consistent with XPS auger spectrometry (Figure 3c). The dominant coordination peaks in EXAFS can be ascribed to Cu-O (EXAFS, Supplementary Figure S9).
Overall, a comprehensive study of ex-situ XPS, XAS, FT-IR, and Raman provides clear evidence that quasi MOF with UMS has been formed in QMOF sample.
Signi cant changes were observed in the XPS spectra for CU-CPWC and CPW under quasi-in situ condition, of which the detail is shown in Experimental Section. C 1s spectra in Figure 3d shows that sp 2 carbon in benzene ring (284 eV) is not detected in the samples, indicating the loss of organic coordinating ligands (linkers) during electrochemical reconstruction. But the peaks corresponding to C-O and C=O in the metal nodes [-Cu 2 (COO) 4 ] still show up (Figure 3d). Figure 3e displays that the -COO groups peaks (532.0 and 532.2 eV) in CU-CPWC and CPW have shifted towards higher binding energy compared with those in QMOF and CMOF, suggesting that the lost organic coordinating ligands may be replaced by hydrogen during electrochemical reconstruction. The peaks at 536.2 ~ 537.0 eV assigned to CO 2 molecule are also observed (Figure 3e), due to the superior CO 2 adsorption capacity. 44 The presence of Cu 0 and Cu 1+ characteristic peak indicates a reduced oxidation state for the Cu species (Figure 3f). Based on these ndings, it is suggested that [Cu 2 (HCOO) 4 ] cage and defective low-coordinated [Cu 2 (HCOO) 4-x ] (0 < x < 4) cage may be formed in CMOF and QMOF respectively after electrochemical reconstruction.
In order to further gain insight into the dynamic electronic structures and chemical state of the Cu active site during CO 2 RR, in situ XAS were investigated under real operating conditions at −0.4 V vs RHE in aqueous CO 2 -saturated 0.1 M KHCO 3 . Figure 4a presents the schematic of XAFS set-up and recording processes for the in situ spectro-electrochemistry experiment. Figure 4b shows the Cu K-edge XANES spectra with electron transitions from 1s to 4p states, including 1s → 4p z transition (A), and the 1s → The nethermost spectrum is the rst collected data. In situ Raman results show that the peak attributed to C-H and C=C peaks disappeared, indicating the loss of organic ligands during electrochemical reconstruction. In pursuit of a molecular-level understanding of electrochemical reconstruction for catalyst, advanced in situ ATR FT-IR was employed to dynamically monitor the structure evolution at the xed potential of -0.4 V (Supplementary Figure S11 and S12). As shown in Supplementary Figure  4g), which further con rms the increase in the Cu oxidation state compared with that in Ar-saturated KHCO 3 solution.
The SEM image of CPW shows that the octahedral morphology has collapsed into irregular nanoparticles after electrochemical reconstruction (Supplementary Figure S14). These changes indicate the organic links are removed from HKUST-1 frames during reconstruction, resulting in large cleavages and structural collapse. The morphological change is also shown for CU-CPWC, as revealed in Supplementary Figure  S15.
On the basis of these results from quasi-in situ XPS, in situ XAS, in situ Raman and in situ FTIR, it is concluded that the CU-CPWC sample surface has been modi ed by oxygen-bearing copper clusters based on coordinatively unsaturated [Cu 2 (HCOO )4-x ] cage. In order to verify whether the defect oxygen-bearing copper clusters promotes electrochemical CO 2 reduction, we carried out CO 2 RR assessments on target catalyst CU-CPWC and the control samples including Cu foil, ER-Cu(OH) 2 and CPW.
Electrochemical activity of catalysts in CO 2 reduction. We evaluated the electrochemical CO 2 reduction activity using a gas-tight H-cell system lled with 0.1 M KHCO 3 solution. Prior to CO 2 RR testing at a given potential, all of the samples were subjected to a potential of -0.4 V vs RHE in CO 2 -saturated KHCO 3 solution for activation. As indicated by the linear sweep voltammetry (LSV) curves shown in Figure 5a, CU-CPWC shows the lower overpotentials at xed current densities, suggesting the improved activity. The CO 2 reduction product distribution of the four catalysts was investigated under various potentials (Supplementary Figure S16). As shown in Figure 5b and c, the CU-CPWC exhibits dramatically increased CH 4 and C 2 H 4 Faradaic e ciency among the four catalysts, bene ting from the superior CO 2 capture and activation capability. Signi cantly, the total Faraday e ciency of CO 2 reduction to CH 4 and C 2 H 4 has reached approximately 80.5 % at -0.90 V.
Electrochemical impedance spectroscopy (EIS) tests were performed to investigate the kinetics of electrochemical processes. CU-CPWC exhibits the smaller Nyquist semicircle diameter than the control samples (Supplementary Figure S17), suggesting the fast charge-transfer kinetics in CRR. Cyclic voltammetry (CV) measurements were then performed to determine the double-layer capacitances (C dl ) (Supplementary Figure S18 and S19), which is related to electrochemical active surface area (ECSA). Not surprisingly, CU-CPWC exhibits the higher C dl than the control samples, indicating the larger ECSA, which is also an important contributor to the enhanced activity. In addition, LSV tests of the four samples in Arsaturated and CO 2 -saturated KHCO 3 electrolyte were performed as well (Supplementary Figure S20). The higher current density in CO 2 -saturated KHCO 3 electrolyte demonstrated that the carbon source for the evolved hydrocarbons comes from CO 2 , which provides a direct evidence for the reduction of CO 2 .
For clarifying the origin CO 2 RR mechanisms of CU-CPWC and CPW, the-rst principles-based density functional theory (DFT) calculations were performed. The CU-CPWC and CPW models have been built and optimized, the detail of which is shown in Supporting Information. The spin-orbit projected 3d density of states of CU-CPWC and CPW are shown in Figure 5d. From this gure, the CU-CPWC presents obvious spin properties with asymmetrical spin distribution around the Fermi level. The d band center of CU-CPWC and CPW is -1.16 and -1.96 eV respectively from the calculations, which indicates that the level change of d band center can be the intrinsic origins of the different CO 2 RR performance. 56 In addition, there are more states around the Fermi level for Cu-CPWC model with CO 2 adsorption (Figure 5e) than those of CU-CPW model as shown in Figure 5f. It means that the CU-CPWC model has better electron transport behavior than that of CPW model, which can help to accelerate the reaction process. The data shown in Supplementary Figure S21, further support the statement.
CU-CPWC catalyst also exhibits good stability for CH 4 and C 2 H 4 generation under continuous CO 2 electroreduction, as shown in Supplementary Figure S22. The Cu K-edge XANES spectra before and after stability testing reveal that the valence state shows negligible change after the constant-potential electrolysis (Supplementary Figure S23a), which is further con rmed by the rst-derivative spectra of XANES (Supplementary Figure S23b). In addition, FT-EXAFS spectra show that the distinct Cu-O peak is still present after stability testing, indicating the structure stability of the CU-CPWC (Supplementary Figure  S24).

Conclusions
Our results demonstrate that defect metal−organic framework HKUST-1 with coordinatively unsaturated copper paddle wheel, that is quasi-MOF clusters (QMOF), can be constructed through a facile "atomized trimesic acid" strategy. The restructuring mechanism based on hydrion substituting organic ligand during electrochemical reduction is revealed through a series of ex-situ, quasi-in situ and in situ characterization analysis, which indicates that unsaturated copper paddle wheel clusters Cu 2 (HCOO) 3  and helium (He, 99.99%) were used as received. All chemicals were used directly without further puri cation. Deionized water was used for the preparation of all aqueous solutions.
Synthesis of Cu(OH) 2 nanowires. In a typical synthesis, a uniform layer of blue Cu(OH) 2 nanowires was rstly prepared by chemical oxidation of Cu foil. Each Cu foil piece was initially washed with hydrochloric acid (36%) for 2 min to remove its native oxide layer, followed immediately by an ultrasonic cleaning in acetone, ethanol, and deionized water for 5 min sequentially. The cleaned Cu foil was immersed into 100 mL of an aqueous solution containing 5.2 g NaOH and 2.282 g (NH 4 ) 2 S 2 O 8 for 35 min at around 3 °C.
Then, the Cu foil was taken out of the solution, thoroughly rinsed with ultra-pure water, and dried with nitrogen. Thus, blue Cu(OH) 2 nanowires were obtained.
Meanwhile, the Cu foil with a layer of Cu(OH) 2 nanowires was reacted with atomized H 3 BTC solution for 60 s, obtaining QMOF. To investigate the evolution of the MOFs growth, Cu(OH) 2 nanowires was reacted with atomized H 3 BTC solution for 120, 240 and 480 s. When the reaction time was 480 s, CMOF sample was obtained.
Synthesis of complete-metal-organic frameworks HKUST-1 (CMOF) through "immersed method". For comparison, the Cu foil with a layer of Cu(OH) 2 nanowires was immersed into the mixed H 3 BTC solution for different time for the nal formation of CMOF.
(1) Fourier Transform Infrared Spectroscopy (FT-IR) and in situ Raman. FT-IR was carried out on Thermo Scienti c Nicolet iN10 FTIR Microscope (Thermo Nicolet Corporation, Madison, WI, USA). In situ Raman scattering spectra were recorded on a LabRAM HR JY-Evolution microscope using a 532 nm argon ion laser. A customized in situ cell was employed for collecting in situ Raman data under atmosphere pressure in CO 2 -saturated 0.1 M KHCO3 aqueous electrolyte. The laser power was set to 50 % and the acquisition time was 30 s.
(2) Temperature programmed reduction (TPR)-CO 2 . The TPR-CO 2 experiment was carried out on a BELCAT II (Microtrac BEL). The sample (0.5 g) was loaded into a quartz tube and swept with argon at 100°C for 120 min. After being cooled down to 40 °C, the sample was loaded with CO 2 at 50 °C for 180 min.
Subsequently, a temperature desorption was programmed at 5 °C/min. The CO 2 signal was detected using a thermal conductivity detector (TCD).
(3) Scanning electron microscopy (SEM) and transmission electron microscopy (TEM). SEM images were obtained on a JEOL JSM-6700 scanning electron microscope at 6 kV with the software package for automated electron tomography. Samples were prepared for TEM by dropping and drying ethanol dispersions of catalysts onto carbon-coated gold TEM grids at room temperature. TEM images were obtained by using a Titan G2 60-300 with an image corrector.
(4) X-ray photoelectron spectroscopy (XPS). For XPS, samples were thoroughly rinsed with deionized water and immersed into liquid nitrogen immediately after being removed from the electrochemical system. This effectively preserved the valence state and structure of the samples during the sample transfer. The samples were vacuum-dried and stored until the X-ray measurements were performed. XPS data were collected on an ESCALAB250Xi photoelectron spectrometer (Thermo Fisher Scienti c) using a monochromatic Al Kα X-ray beam (1,486.6 eV). All binding energies were referenced to the C 1s peak (284.6 eV).
(5) X-ray absorption spectroscopy (XAS) date collection and analysis. XAS spectra at the copper K edge were measured in uorescence mode at the X-ray absorption station (beamline 1W2B) of the Beijing Synchrotron Radiation Facility (BSRF) in China using a double crystal Si (111) monochromator. The storage ring was working at the energy of 2.5 GeV with an average electron current of 150 mA. The incident and output beam intensities were monitored and recorded using ionization chambers lled with pure nitrogen.
An in situ home-made cell was employed for collecting in situ XAFS data. As-prepared Cu(OH) 2 NWs, QMOF, and CMOF on Cu foil samples (1 × 5 cm) were put into ethanol-lled hermetic containers (8 mL) separately and ultrasonic stripped for 30 min to obtain suspension containing super cial Cu(OH) 2 NWs, QMOF, and CMOF powder, respectively. For each, 100 μL of catalyst suspension and 25 μL of 5 % Na on 117 solution were sonicated for 5 min. Each catalyst ink was then uniformly loaded onto a fresh piece of thin carbon paper (2 cm × 5 cm, thickness of 500 nm) and allowed to dry in vacuum. The Cu(OH) 2 NWs, QMOF, and CMOF powder on the thin carbon paper served as the working electrode, was in contact with a slip of copper tape, and was xed with epoxy glue to the exterior of the wall of the cell, over the 0.8 cm hole, with the OFC or OBC layer facing inward. A Pt foil (1 cm × 1 cm × 0.2 cm) and a Ag/AgCl electrode (KCl sat.) were used as the counter electrode and reference electrode, respectively.
The copper XAFS raw data processing was carried out by using ATHENA (Demeter 0.9.25, an interface to IFEFFIT) and ARTEMIS (Demeter 0.9.25) by the standard procedures. The k3-weighted χ(k) data in the kspace ranging from 2.0 to 12 Å −1 were Fourier-transformed to real (R) space using a Hann window (dk = 1.0 Å −1 ) to separate the EXAFS contributions from different coordination shells. To obtain the detailed structural parameters around the Cu atom in the as-prepared samples, quantitative curve ttings were carried out for the Fourier-transformed k3χ(k) in R-space using the ARTEMIS module of IFEFFIT. During curve ttings, the Debye-Waller factor (σ 2 ) and bond length (R) were treated as adjustable parameters.
No R factors yielded are larger than 0.01, indicating the good tting quality. The Morlet wavelet was chosen as the mother wavelet for the basic functions and the parameters (σ = 2, η = 5) were used to achieve better resolution in the wave vector k.
(6) In situ IR Measurements. For in situ IR measurements, as-prepared QMOF sample was put into ethanol-lled hermetic containers (8 mL) separately and ultrasonic stripped for 30 min to obtain suspension containing super cial powder. 100 μL of catalyst suspension and 25 μL of 5 % Na on 117 solution were sonicated for 5 min. Then 100 μL of QMOF ink solution was casted onto the Au lm coated Si prism. A homemade spectro-electrochemical cell ( Figure S4) with three-electrode system was employed for collecting in situ IR data. The QMOF powder on the Si prism served as the working electrode. A coiled Pt wire served as counter electrode and a Ag/AgCl electrode (KCl sat.) served as reference electrode. In situ IR measurements were performed on a Thermo Scienti c Nicolet iS50 FT-IR spectrometer with attenuated total re ection (ATR) con guration using a liquid nitrogen cooled MCT-A detector. All spectra were collected at a resolution of 4 cm −1 at a bottom-up incidence angle of ca. 67°. The FT-IR spectrometer was puri ed with high-purity argon for one hour. Before the test, the electrolyte was puri ed with high-purity CO 2 for 30 min. aqueous solution. All current densities were normalized to the geometrical area of the electrode.
The Faradaic e ciency can be calculated as follows: see formula 1 in the supplementary les.
where n i is the number of electrons required to produce an i molecule, which is 2, 2, 8, and 12 for CO, H 2 , CH 4 , and C 2 H 4 , respectively; F is the Faradaic constant (96,485.33289 C mol −1 ); P 0 is the atmospheric pressure (1.013 × 105 Pa); T 0 is the reaction temperature (298 K); R is the ideal gas constant (8.314 J mol -1 K -1 ); V o is the total headspace volume (typically 50 mL); V i (vol %) is the measured volume concentration of product i from the electrochemical cell at a given passing charge (typically 5 coulombs); Q total is the total amount of charge Q (in units of coulombs) passed through the sample (5 coulombs); and Q i is the amount of charge Q to produce species i from CO 2 .
DFT calculations. By using the Cambridge Serial Total Energy Package (CASTEP) in Materials Studio, all density functional theory (DFT) calculations were performed with the plane-wave pseudopotential method to apply rst-principles-based quantum mechanics simulations. In this work, the spin generalized gradient approximation (SGGA) within Perdew-Burke-Ernzerhof (PBE) form was setup as an exchangecorrelation function. A 3 x 3 x 1 Monkhorst-Pack k-points grid was used for k-points sampling, and 900 eV plane-wave expansion was setup for energy cut-offs. These parameters were carefully tested. The Cu-3d 10 4s 1 , O-2s 2 2p 4 , C-2s 2 2p 2 , and H-1s 1 were setup for valence states. A vacuum slab of about 15 Å was maintained in the super-cell con guration that was large enough for the calculations. The SCF tolerance for geometrical optimization and phonon calculations was less than 1e -6 eV·atom -1 .