Selective electroreduction of carbon dioxide to methanol on copper selenide nanocatalysts

Production of methanol from electrochemical reduction of carbon dioxide is very attractive. However, achieving high Faradaic efficiency with high current density using facile prepared catalysts remains to be a challenge. Herein we report that copper selenide nanocatalysts have outstanding performance for electrochemical reduction of carbon dioxide to methanol, and the current density can be as high as 41.5 mA cm−2 with a Faradaic efficiency of 77.6% at a low overpotential of 285 mV. The copper and selenium in the catalysts cooperate very well for the formation of methanol. The current density is higher than those reported up to date with very high Faradaic efficiency for producing methanol. As far as we know, this is the first work for electrochemical reduction of carbon dioxide using copper selenide as the catalyst. While the conversion of CO2 to valuable, storable chemicals is attractive, there are few inexpensive and abundant catalysts that are also active and selective for liquid fuels. Here, the authors study copper selenide as a high-performing and efficient electrocatalyst for CO2 conversion to methanol.

E lectroreduction of carbon dioxide (CO 2 ) is a potential strategy to transform the intermittent sources of energy into high-energy chemicals, which can potentially reduce our dependence on fossil fuels and alleviate atmospheric pollution [1][2][3][4] . Among the products formed upon electrochemical CO 2 reduction, hydrocarbons, and alcohols with high energy density, like methanol, are compatible with existing infrastructures and can substitute for fossil fuels [5][6][7] . It is known that the electrocatalytic reduction of CO 2 to methanol requires intricate six-electron/ proton coupling steps and sluggish kinetics [8][9][10][11] . Therefore, the reaction usually suffers from low current density, poor selectivity and the large overpotential 9,[12][13][14] . As a result, rational design of highly active and robust electrocatalysts that could generate high current density and high selectivity is critical for large-scale application.
Metal and metal-based catalysts have been used for electroreduction of CO 2 to CO, hydrocarbons and alcohols 15,16 . To date, some electrocatalysts, such as precious metal and copper-based catalysts, have shown to be promising for electroreduction of CO 2 to methanol [9][10][11][12]14,[17][18][19][20][21][22] . Among these materials, Cu has been reported as the promising electrocatalyst that is active and selective for CO 2 reduction to hydrocarbons and alcohols. However, the activity and selectivity of bulk Cu for producing methanol are usually low 14,[18][19][20] . Metallic Pd or Pt, Pd-Cu and Mo complexes all have also been employed as catalysts for electrochemical synthesis of methanol 10,12,19,21,22 . In addition, Ru/Ti bimetallic oxide is another promising catalyst for production of methanol 23 . It was reported that Mo-Bi bimetallic chalcogenide (MoS 2 /Bi 2 S 3 ) could be used as electrocatalyst to promote the reaction 8 . Nevertheless, achieving high current density and Faradaic efficiency (FE) simultaneously for conversion of CO 2 to methanol remains to be a challenge, and only a few catalysts reported up to date could reach relatively high current density and selectivity, as shown in Supplementary Table 1. Therefore, designing efficient catalysts to enhance the activity and FE, and reduce the overpotential is very interesting from both scientific and practical viewpoints.
In recent years, nanoscale transition metal oxides (TMOs) and chalcogenides (TMCs) have attracted considerable attention, which have great potential of application in photo-electric devices, lithium-ion batteries, gas sensors, and electrocatalysis 24,25 . For electroreduction of CO 2 , it is known that metal-based oxides, and sulfides could have good performance 8,[26][27][28] . However, only few studies have reported the use of TMCs, such as metal selenides and tellurides for CO 2 electroreduction 29,30 . WSe 2 nanoflakes were reported as an efficient catalyst for CO 2 electroreduction to CO 29 . Density functional theory (DFT) calculation indicated that molybdenum sulfides and selenides were also possible catalysts for CO 2 electroreduction 30 , which showed that the intermediates COOH and CHO were more easily adsorbed on the S and Se atoms at the edges than the intermediate CO. Therefore, transition-metal selenides may be a class of promising catalysts for CO 2 electroreduction. Copper selenides are interesting materials that form nonstoichiometric (i.e., Cu 2−x Se) as well as stoichiometric (i.e., CuSe, Cu 2 Se, Cu 2 Se 3 ) phases 31,32 . They also have structure stability and compositiondependent optical/electrical properties 33 . In addition, they are low-cost materials compared with many other materials 25 , especially noble metals. The advantage of copper selenide is its multiple oxidation states and high electrical conductivity, which can deliver better electrochemical properties 32 . Moreover, the unsaturated Se atoms along its edges may enhance the number of exposed active sites, electrical conductivity and catalytic activity in CO 2 reduction 29,30 .
Herein, we report a facile solvothermal synthesis of Cu 2−x Se(y) nanocatalysts in diethylenetriamine (DETA, Supplementary Fig. 1 for the structure) and H 2 O binary solution, where y represents the volume ratio of DETA and water (V DETA /V H2O ), and the value of x is in the range of 0.3 to 0.4, depending on the atom ratio of Cu and Se in the catalysts. The properties of the catalysts, such as size and morphology, are solvent-dependent. The catalysts synthesized in the mixed solvent with V DETA /V H2O of 1:3 can convert CO 2 into methanol with a current density of 41.5 mA cm −2 at FE of 77.6%. The current density is higher than those reported up to date with very high methanol selectivity (Supplementary Table 1).

Results
Synthesis and characterization of Cu 2−x Se(y) nanocatalysts. The Cu 2−x Se(y) nanocatalysts were synthesized by solvent coordination molecular template method 34,35 , which is shown schematically in Supplementary Fig. 1. In the synthesis of the catalysts, the positively charged ammonium ions coordinated with Se to incorporate into the neighboring Cu 2−x Se(y) nanoparticles. The protonated amine molecules then act as a template, resulting in the new morphology of the Cu 2−x Se(y) nanocatalysts 34,35 .
It is clearly shown that the V DETA /V H2O affected the morphology of the Cu 2−x Se(y) nanocatalysts considerably ( Supplementary  Fig. 2). The nanoparticles tended to be granular with increasing content of water in the solvent. In addition, the Cu 1.63 Se(1/3) nanoparticles synthesized at the V DETA /V H2O of 1/3 had the smallest size. The detailed characterization results of Cu 1.63 Se(1/3) nanocatalysts are shown in Fig. 1. The images of scanning electron microscopy (SEM) (Fig. 1a) and transmission electron microscopy (TEM) (Fig. 1b) reveal that the size of the Cu 1.63 Se(1/3) nanoparticles was~50 nm, which is consistent with the results obtained from dynamic light scattering (DLS, inset of Fig. 1a). Elemental distribution mappings (inset images of Fig. 1b) analysis further indicated the uniform distribution of Cu (yellow) and Se (blue) atoms in the catalysts. Thermogravimetry (TG) curve is shown in Supplementary Fig. 3. The weight loss of DETA (if incorporated into the nanoparticles) should occur at 280-580°C 36 . It can be seen from the figure that there was no weight loss in the temperature range, indicating that there was no DETA in the catalysts. The high-resolution TEM (HR-TEM) image demonstrated the high crystallinity of nanoparticles (Fig. 1c). Furthermore, X-ray photoelectron spectroscopy (XPS) identified the valence states of Cu and Se in the nanocatalysts. The binding energies of Cu 2p 3/2 and Cu 2p 1/2 (Fig. 1d) were 932.1 and 952.0 eV respectively, and there was a peak at 916.9 eV in the Cu LMM spectrum (Fig. 1e), indicating the existence of Cu(I) [37][38][39] . In addition, the binding energies of Cu 2p 3/2 and Cu 2p 1/2 at 933.5 and 953.9 eV (Fig. 1d) along with the satellite lines can be assigned to Cu(II) 40 . The double peaks at 55.3 and 54.6 eV (Fig. 1f) in the XPS spectrum were the typical Se 3d binding energy for lattice Se 41 . The atomic ratio of Cu(I) to Cu(II) in the nanocrystal determined by XPS was 3.41. Meanwhile, the atomic ratio of Cu to Se determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES) was 1.64, which is consistent with the value determined by XPS (1.63). In terms of the molecular formula, nonstoichiometric Cu 2−x Se(y) could be considered as a mixture of stoichiometric Cu 2 Se and CuSe 31 and thus the molecular formula can be estimated as Cu  37,42 , which is in agreement with the HR-TEM result (Fig. 1c). Meanwhile, the difference of XRD patterns of the catalysts synthesized at various V DETA /V H2O ratios was not noticeable ( Supplementary Fig. 9). The results showed that the crystal structures of different samples were not changed with the V DETA /V H2O ratio notably. However, the size and morphology depended strongly on the composition of the solvents, which influenced the performances of CO 2 electroreduction.
Electrocatalytic performance of CO 2 reduction over Cu 2−x Se(y) nanocatalysts. The linear sweep voltammetry (LSV) study was conducted to investigate the performances of the Cu 2−x Se(y) nanocatalysts under the same conditions in [Bmim]PF 6 (30 wt %)/CH 3 CN/H 2 O (5 wt%) ternary electrolyte, and the results are shown in Fig. 2a. The results indicate that Cu 1.63 Se(1/3) exhibited a more positive onset potential of −1.815 V vs. Ag/Ag + than other Cu 2−x Se(y) nanocatalysts, suggesting that Cu 1.63 Se(1/3) was favorable to the binding of CO 2 . In addition, the current density over Cu 1.63 Se(1/3) reached a high value of about 40 mA cm −2 at −2.1 V vs. Ag/Ag + , which exhibits higher activity than other Cu 2 −x Se(y) nanocatalysts. The much higher current density of the CO 2 -saturated than the N 2 -saturated on the Cu 1.63 Se(1/3) (around −2.1 V vs. Ag/Ag + ) indicates the reduction of CO 2 .
Constant-potential electrolysis of CO 2 over different catalysts was performed in a typical H-type cell 8 . Liquid-phase and gasphase products were quantified by nuclear magnetic resonance spectroscopy ( 1 H NMR) and gas chromatography (GC), respectively. The current density and FE are displayed in Fig. 2b, c. It can be found that all Cu 2−x Se(y) nanocatalysts basically yielded a certain amount of methanol, HCOOH, CO, and H 2 with a combined FE of around 100%, and no other product was detected ( Supplementary Fig. 10). The Cu 1.63 Se(1/3) electrode had better performance than other Cu 2−x Se(y) nanocatalysts. The maximum FE occurred at −2.1 V vs. Ag/Ag + , and it could reach 77.6% with a current density of 41.5 mA cm −2 (Fig. 2b, c). The equilibrium (thermodynamic) potential for CH 3 OH was −1.815 V vs. Ag/Ag + , which was obtained by extrapolation of partial current density vs. potential curve to zero partial current density ( Supplementary   Fig. 11) [43][44][45] . Therefore, the overpotential for CO 2 electroreduction to methanol was 285 mV at −2.1 V vs. Ag/Ag + . This catalyst exhibits the highest current density with very high FE for producing methanol, as can be known in Supplementary Table 1. The cell voltage is an important factor for practical application, which depends mainly on the performances of the electrocatalysts. In this study, we calculated the cell voltage using the reported method 46,47 , and the cell voltage of our system was 2.67 V, which is in the range of reported values (2.2-3.7 V, Supplementary Table 3). In Supplementary Fig. 12, the FE for methanol production increased with the cell voltage to reach the maximum value of 77.6% at 2.67 V. We also calculated the energy efficiency (EE) for methanol production at different cell voltages using the reported method 48 , and the results are given in Supplementary Fig. 12. The EE exhibited a similar tendency to the FE of methanol with variation of the cell voltage. Furthermore, the highest EE was 61.7% at the optimized cell voltage of 2.67 V.
We also carried out the experiment using 13 CO 2 to replace CO 2 in the electrolysis over Cu 1.63 Se(1/3). The 1 H NMR spectra of the product indicated that only 13 CH 3 OH was produced (Supplementary Fig. 13), confirming that the product methanol originated from CO 2 . Meanwhile, the electrolyte after different electrolysis times was also tested by IR 8,49 in Supplementary  Fig. 14, which can further confirm formation of methanol. In addition, FE and current density did not vary during 25 h of operation ( Supplementary Fig. 15), which indicated exhibit longterm stability in the electrolysis. The composition and structure of the catalyst did not change after the reaction as characterized by XPS and XRD analysis (Supplementary Figs. 16 and 17), indicating the excellent stability of the Cu 1.63 Se(1/3).
The particle size and morphology of the materials can affect the number of exposed active sites. From Fig. 2d and Supplementary  Fig. 1, we can see that the size of the catalysts synthesized at V DETA /V H2O = 1/3 was smallest and the current density and FE were largest. More active sites were exposed with more unsaturated Se atoms on the smaller particles may be one of the main reasons for high current density and FE 29,30 .  Table 2 show the variation of overpotential with partial current density for methanol production over these catalysts. The resulting Tafel plots of various Cu 2−x Se(y) electrodes are linear in the overpotential (η) range from 0.035 to 0.285 V. The Tafel slope of Cu 1.63 Se(1/3) was smaller than other Cu 2−x Se(y) nanocatalysts, which leads to faster increment of CO 2 reduction rate with increasing overpotential 26,50 . Combining with efficient and stable electrocatalytic CO 2 conversion to methanol, the Cu 1.63 Se(1/3) nanocatalysts can be a promising catalyst in practical applications. The excellent activity of Cu 2−x Se(y) electrodes may also result partially from large electrochemical active surface area. According to the Randles-Sevcik equation, the current density at −1.85 V (vs. Ag/Ag + ) plotted against the square root slope of scan rate is shown in Supplementary Fig. 19. The double-layer capacitances (C dl ) of various catalysts were then calculated and illustrated in Supplementary Table 2. The obvious difference in C dl values (ranging from 0.00906 to 0.0183 Fcm −2 ) suggests that the Cu 1.63 Se(1/3) had the largest electrochemistry surface area, which is beneficial to the reaction.
Electrochemical impedance spectroscopy (EIS) also revealed an easier electron transfer process on the electrode interface. The Nyquist plot was conducted by running the experiment at an open circuit potential ( Supplementary Fig. 20), and the equivalent circuit R(C(R(Q(RW)))) ( Supplementary Fig. 21) is then used to fit the impedance data. The simulated results of charge transfer resistance (R ct ) are given in Supplementary   0 -1.2 -1.4 -1.6 -1.8 -2.0 -2.2 -2 Supplementary Fig. 23), which resulted partially from the difference of the interaction between CO 2 and the anions of the ILs 55 .
[Bmim]PF 6 exhibited higher current density and Faradaic efficiency for methanol among all the ILs used. The separation of the reaction mixture is crucial for practical application. Although this is out of the scope of this work, we would like to discuss this very briefly.  56 . It is very interesting to investigate the reasons for the outstanding performance of the Cu 1.63 Se(1/3) electrocatalyst in the electrocatalytic reduction of CO 2 to methanol. Therefore, we carried out a series of control experiments to investigate the crucial role of Se in the catalysts. CO 2 electroreduction with Cu, CuO, Cu 2 O, CuS, Cu 2 S, CuSe, and Cu 2 Se as catalysts were studied. Both current density and FE for methanol over Se-free catalysts were obvious lower ( Fig. 3 and Supplementary Fig. 24). The results suggest that the Cu and Se in the catalysts cooperated very well for the formation of methanol. In other words, the capacity of electroreduction of CO 2 to methanol was enhanced when O or S atom was replaced by Se atom in the catalysts. Moreover, when commercial CuSe or Cu 2 Se was utilized as the catalysts, both current density and FE were much lower than that over Cu 1.63 Se(1/3). The catalytic performance was also evaluated using the electrochemical active surface areas (ECSA) determined by reported method 57 . Results in Supplementary Fig. 25 and Supplementary Table 6 show that the formation rate of methanol over Cu 1.63 Se(1/3) was intrinsically higher than that on the other catalysts. Thus, on the basis of above results, we can deduce that Se in the catalysts is crucial for efficient CO 2 reduction to methanol.
We also carried out extended X-ray absorption fine structure spectroscopy (EXAFS) experiments to study Cu K-edge, which can disclose the local atomic arrangements of the catalysts. The Cu K-edge k 2 χ(k) oscillation curve for Cu 1.63 Se(1/3) was obviously different from that for CuSe and Cu 2 Se ( Supplementary  Fig. 26). It can be seen that the coordination number in Cu 1.63 Se (1/3) was smaller than that in CuSe and Cu 2 Se (Supplementary  Figs. 27-30 and Supplementary Table 7). Thus, there existed unsaturated Se atom in the Cu 1.63 Se(1/3), which may enhance the performance for CO 2 electroreduction.
To understand the reaction pathway for the formation of methanol, some control experiments were conducted in the presence of the possible reaction intermediates, such as formic acid, CO and formaldehyde (Supplementary Table 8). From the production rates of methanol, it can be seen that CO and formaldehyde clearly promoted the formation of methanol and thus they are possible intermediates in the formation of methanol. On the basis of the results above, we propose a possible reaction pathway over Cu 2−x Se(y) nanocatalysts (Fig. 4a). In the initial stage of the reduction, the electrolyte containing ionic liquids can enhance the concentration of CO 2 in electrolyte and transport of CO 2 to the catalyst surface to improve further transform of CO 2 into adsorbed CO 2 •−2,55 . The adsorbed CO 2 •− could bind with the active sites on the surface of catalysts and accelerate the formation of adsorbed-CO species, which was a crucial intermediate for facilitating methanol production as reported 20,58 . Furthermore, the appropriate Cu active sites over the catalysts can enhance the initial intermediate of absorbed-CO to accept electrons and protons to form absorbed-CHO and then reduced to methanol.
The density functional theory (DFT) calculations were also conducted on the multiple elementary reaction steps, and the results are shown in Fig. 4b 59 . The Cu 1.63 Se(1/3) catalyst also has a moderate binding energy for *CO among the three catalysts, which is beneficial for CO 2 transformation to more reduced products that require more than a two-electron reduction 20,58 . Based on the Brønsted-Evans-Polanyi (BEP) relationship 60,61 , the reaction barrier has a linear relationship to the reaction energy, and it can also be seen that the step of *CO reduction to *CHO was an endothermic and likely rate-limiting step since the highest energy potential (0.56 eV) is needed in this step. Compared with Cu 2 Se and CuSe, the free energy of *CHO over Cu 1.63 Se(1/3) catalyst is more negative, which may be mainly originated from the moderately strong binding energy for *CO intermediate. In addition, the C-Cu bond ( Supplementary Fig. 34) between Cu 1.63 Se(1/3)-CHO is 1.926 Å, which is shorter than those of Cu 2 Se-CHO (2.188 Å) and CuSe-CHO (2.002 Å), indicating that *CHO is easier to adsorb on the surface of the catalyst to accept electrons and protons to form *OCH 2 and *OCH 3 , and then is reduced to methanol. These results illustrate that the structure distortion of Cu 1.63 Se(1/3) was beneficial for CO 2 electroreduction to methanol.
In summary, a series of Cu 2−x Se(y) nanocatalysts were synthesized for selective electroreduction of CO 2 to methanol. The Cu and Se in the catalysts had excellent cooperative effect for catalyzing the reaction. The size and morphology were crucial for the performance of the catalysts, which could be controlled by the V DETA /V H2O ratios. The Cu 1.63 Se(1/3) nanocatalysts yielded outstanding current density of 41.5 mA cm −2 with FE of 77.6% at −2.1 V vs. Ag/Ag + . Moreover, the catalyst was also very stable in the reaction. Despite the catalytic system is far from industrial production, it is still very interesting that Cu 1.63 Se(1/3) nanocatalysts can yield highest current density up to date at very high Faradaic efficiency. We believe that some other transition metal selenides can be designed as efficient electrocatalysts for CO 2 reduction.  Synthetic procedures for Cu 2−x Se(y) nanocatalysts. The Cu 2-x Se(y) nanocatalysts were prepared according to the procedures reported by other researcher 35 . In a typical procedure, 1 mmol CuCl 2 ⋅2H 2 O, 1 mmol Na 2 SeO 3 , and 2 mL hydroxylamine were added into a mixed solvent (80 mL) including DETA and deionized water with a volume ratio of V DETA /V H2O = 1/3 (or 1/0, 3/1, 1/1, 1/5 and 0/1).

Methods
After stirring for half an hour at room temperature, the blue solution was then transferred into a Teflon-lined autoclave. Subsequently, the sealed Teflon-lined autoclave was maintained at 180°C for 15 h and then naturally cooled to room temperature. The resulting precipitates were obtained via the method of centrifugation and washed with distilled water and absolute ethanol five times, followed by drying at 80°C overnight in a vacuum oven.
Physicochemical characterization. The microstructures of the catalysts were characterized by scanning electron microscope (SEM, HITACHI S-4800) and transmission electron microscopy (TEM, JEOL JEM-2100F) equipped with energy dispersive spectrometer (EDS). Dynamic light scattering (DLS) measurements in ethanol were performed on a Zetasizer Nano instrument (Malvern Instruments, Worcestershire, UK). X-ray photoelectron spectroscopy (XPS) study was carried out on the Thermo Scientific ESCALab 250Xi using a 200 W Al-Kα radiation. In the analysis chamber, the base pressure was about 3 × 10 −10 mbar. Typically, the hydrocarbon C1s line at 284.8 eV from adventitious carbon was used for energy referencing. X-Ray diffraction (XRD) analysis of the samples were performed on a Rigaku D/max-2500 X-ray diffractometer with Cu-Kα radiation (y = 0.15406 nm) and the scan speed was 5°/min. The elemental contents of the catalysts were detected using inductively coupled plasma optical emission spectroscopy (ICP-AES, Vista-MPX). The thermogravimetric (TG) curve was obtained using Pyris1 TGA under N 2 atmosphere.
Electrode preparation. The electrode of Cu 2-x Se(y)/CP was prepared as follow 19 . 10 mg Cu 2-x Se(y) prepared above and 1 mg carbon black (Vulcan XC 72) were suspended in the solution with 3 mL acetone and 20 μL Nafion D-521 dispersion (5 wt%) via ultrasound. Then, 302 μL of the above solution was uniformly spread onto the CP (1 × 1 cm −2 ) surface assisted by a micropipette and then dried under room temperature. On each carbon paper, the loading of Cu 2−x Se(y) catalyst was 1.0 mg cm −2 . Before experiment, all the auxiliary electrodes were sonicated in acetone for 10 min and then washed with H 2 O and absolute ethanol, and then dried in N 2 atmosphere.
Electrochemical study. All the electrochemical experiments were conducted on the electrochemical workstation (CHI 6081E, Shanghai CH Instruments Co., China). Linear sweep voltammetric (LSV) scans were conducted in a single compartment cell with a three electrodes configuration, including a working electrode, a counter electrode (Pt gauzes), and a reference electrode (Ag/Ag + with 0.01 M AgNO 3 in 0. Product analysis. After electrolysis reaction, the gaseous products were collected using a gas bag and then analyzed by an Agilent 4890 gas chromatograph equipped with a TCD detector with helium as internal standard. The liquid products were analyzed by 1 H NMR measured on a Bruker Avance III 400 HD spectrometer in CD 3 CN with TMS as internal standard. The Faradaic efficiency of the products was calculated through GC and NMR analysis 8 .
Tafel analysis. The partial current densities for products under different potentials were measured and the equilibrium potential was obtained by extrapolation method. The overpotential was obtained from the difference between the equilibrium potential and the catalytic potential. Multiple electrolysis experiments were performed at each potential to obtain the current density versus overpotential data in the H-type electrolysis cell as described above. Tafel plots were constructed from these data.
Double-layer capacitance (C dl ) measurements. The electrochemical active surface area is proportional to C dl value. C dl was determined in H-type electrolysis cell by measuring the capacitive current associated with double-layer charging from the scan-rate dependence of cyclic voltammogram (CV). The CV ranged from −1.8 to −1.9 V vs. Ag/Ag + . The C dl was estimated by plotting the Δj (j a −j c ) at −1.85 V vs. Ag/Ag + against the scan rates, in which the j a and j c were the anodic and cathodic current density, respectively.
IR spectroscopy study. To further understand the process of CO 2 electroreduction, a Bruker Tensor 27 IR spectrometer was used to analyze the species produced in the electrolyte. In the experiment, 100 μL electrolyte after desired electrolysis time was dropped on CaF 2 disc window and then IR spectrum was obtained.
Electrochemical active surface areas (ECSA) measurement. The ECSA values of all electrodes were evaluated by cyclic voltammetry (CV) using the ferri-/ferrocyanide redox couple ([Fe(CN) 6 ] 3-/4-) as a probe 57 . The CV curves were obtained in a N 2 -saturated 5 mM K 4 Fe(CN) 6 /0.1 M KCl solution including a counter anode (platinum gauze), and a reference electrode (Ag/AgCl with saturated KCl). According to the Randles-Sevcik equation 57 , the values of ECSA were obtained.
The calculation of overpotential (η) and equilibrium potential (E 0 ). Overpotential (η) is the difference between the equilibrium potential and the actual potential for the transformation of the substrate CO 2 into the product methanol via  Fig. 4 Mechanism study of CO 2 reduction to methanol. a Proposed mechanism on Cu 2−x Se(y) electrode; b free energy diagrams on Cu 1.63 Se(1/3) electrode ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-08653-9 Eq. (1): Here, the E 0 CO2→methanol referred to the equilibrium potential for CO 2 transformation to CH 3 OH, which can be obtained by extrapolation method [43][44][45] . Taking the Cu 1.63 Se(1/3) electrode as example, stepped potential electrolysis experiments between −1.8 and −2.0 V were carried out and the electrolysis products were collected and characterized. The current densities for CH 3 OH at each potential are shown in Supplementary Fig. 11, and the potential at j CH3OH = 0 by extrapolation method is the equilibrium potential. Therefore, the overpotential can be obtained. The method to calculate the overpotential over other electrodes was similar.
The calculation of cell voltage. In this study, we calculated the cell voltage using the method reported 46,47 . It is mainly from the half reaction potentials for water oxidation, CO 2 reduction and the ohmic drop (E iR ) from electrolyte resistance (R s ). The R s stands for solution resistance which was determined by electrochemical impedance spectroscopy (EIS) at frequencies ranging from 10 −2 to 10 5 Hz and the I represents for amps of average current. Therefore, we calculated E iR by Eq. ( Cell efficiency. A characteristic cell energy efficiency (EE) of reduction CO 2 to methanol was obtained by Eq. (4) 48 : In the equation, ΔE°represents the difference between the standard half reaction potentials for water oxidation (1.23 V vs. RHE) and reduction CO 2 to methanol (−0.89 V vs. RHE).
Extended X-ray absorption fine structure (EXAFS) experimental details. The homogeneously mixed samples (20 mg) and graphite (100 mg) sample were pressed into circular slices with a diameter of 10 mm which was used for further EXAFS measurement under ambient condition. The EXAFS measurements were carried out on the beamline 1W1B station of Beijing Synchrotron Radiation Facility, P.R. China (BSRF). A water-cooled Si (111) double-crystal monochromator (DCM) was utilized to monochromatize the X-ray beam and the detuning was done by 10% to remove harmonics. The electron storage ring of BSRF was operated at 2.5 GeV with a maximum current of 250 mA. The EXAFS data of Cu K-edge was obtained in the energy range from −200 to 1000 eV in transmission mode. The EXAFS oscillations were then extracted according to standard procedures via the ATHENA module implemented in the IFEFFIT software packages. With a Fourier transform k-space range of 2.2-12.8 Å −1 , the quantitative curve-fittings were conducted in the Rspace according to the module ARTEMIS of IFEFFIT software packages. The phase shift Φ(k) and backscattering amplitude F(k) were calculated by FEFF 8.0 code.
Computational method. The free energies of CO 2 reduction states were performed using Vienna Ab-initio Simulation Package (VASP) 62 , taking advantage of the density functional theory (DFT) and the Projected Augmented Wave (PAW) method 63 . The revised Perdew-Burke-Ernzerhof (RPBE) functional was used to describe the exchange and correlation effects 64,65 . For all the geometry optimizations, the cutoff energy was set to be 450 eV. The (110) surface was chosen to represent the catalysis surface of CuSe; while the (220) surface was modeled to simulate the sites on Cu 1.63 Se(1/3) and Cu 2 Se. The Monkhorst-Pack grid 66 of 2 × 2 × 1, 3 × 4 × 1, and 3 × 4 × 1 were used to carry out the surface calculations on CuSe, Cu 1.63 Se(1/3), and Cu 2 Se, respectively.
The reduction of CO 2 to CH 3 OH could occur via a pathway involving six elementary steps 67 by Eqs. (5-10): where * denotes the active sites on the catalyst surface. Based on the above mechanism, the free energies of the corresponding intermediate states are important to identify the activity of a given material in catalyzing CO 2 reduction. The computational hydrogen electrode (CHE) model 68 proposed by Norskov et al. was used to calculate the free energies of CO 2 reduction intermediates, based on which the free energy of an adsorbed species is defined as Eq. (11): where ΔE ads is the electronic adsorption energy, ΔE ZPE stands for the zero point energy difference between adsorbed and gaseous species, TΔS ads denotes the corresponding entropy difference between these two states, and ∫C P dT is the enthalpy correction. The electronic binding energy is referenced as graphene for each C atom, ½ H 2 for each H atom, and (H 2 O − H 2 ) for each O atom, plus the energy of the clean slab. The corrections of zero point energy, entropy, and enthalpy of adsorbed can be found in Supplementary Table 9.
Code availability. All code supporting the findings of this study are available from the corresponding author on request.

Data availability
Data for Figs. 1 to 4, Supplementary Figures 1 to 34 and Supplementary Tables 1 to 9 can be found in the Source Data file.