Steering surface reconstruction of copper with electrolyte additives for CO2 electroreduction

Electrocatalytic CO2 reduction to value-added hydrocarbon products using metallic copper (Cu) catalysts is a potentially sustainable approach to facilitate carbon neutrality. However, Cu metal suffers from unavoidable and uncontrollable surface reconstruction during electrocatalysis, which can have either adverse or beneficial effects on its electrocatalytic performance. In a break from the current catalyst design path, we propose a strategy guiding the reconstruction process in a favorable direction to improve the performance. Typically, the controlled surface reconstruction is facilely realized using an electrolyte additive, ethylenediamine tetramethylenephosphonic acid, to substantially promote CO2 electroreduction to CH4 for commercial polycrystalline Cu. As a result, a stable CH4 Faradaic efficiency of 64% with a partial current density of 192 mA cm−2, thus enabling an impressive CO2-to-CH4 conversion rate of 0.25 µmol cm−2 s−1, is achieved in an alkaline flow cell. We believe our study will promote the exploration of electrochemical reconstruction and provide a promising route for the discovery of high-performance electrocatalysts.

T he continuously increasing carbon dioxide (CO 2 ) emissions caused by huge fossil fuel consumption lead to severe environmental changes associated with global warming. The CO 2 electroreduction reaction (CO 2 RR) is attractive as a sustainable effort to recycle CO 2 into valuable industrial feedstocks and fuels for reducing the greenhouse effect 1 . Among all the catalysts for the CO 2 RR, copper (Cu) metal is uniquely capable of producing multielectron transfer hydrocarbon products with appreciable activity 2,3 , however, its catalytic activity and selectivity for a certain product are still far from satisfactory 4,5 . Because of their weak cohesive energy and high surface mobility, metallic Cu catalysts usually undergo reconstruction involving atomic re-arrangement and compositional change during electrocatalysis [6][7][8][9] . Since the electrocatalytic properties have a noticeable dependence on the surface structure, the reconstruction behavior of Cu metal greatly impacts its CO 2 electroreduction performance [10][11][12] .
The reconstruction behavior of Cu is a double-edged sword for the CO 2 RR. Several studies have shown that the initial well-defined morphology and highly active sites tend to be lost during electrocatalysis because of surface reconstruction, resulting in catalytic performance degradation 13,14 . Nevertheless, surface reconstruction sometimes generates unique active sites with increased catalytic activity for specific products. Our previous work has shown that the reconstruction of a Cu complex under working conditions causes the formation of Cu nanoclusters with high selectivity for CH 4 9 . Kim et al. found that densely arranged Cu nanoparticles undergo reconstruction to in-situ form a disordered Cu nanostructure with a high selectivity for multicarbons 15 . Unfortunately, previous findings indicate that catalytic activity induced by restructuring is hardly accessible through ex-situ paths. Even if it could be, the welldesigned catalysts are still at the risk of oxidation or poisoning in air, or deactivation induced by the restructuring process 16,17 .
Steering the dynamic reconstruction process in a favorable direction may be a strategy for improving catalytic performance. However, due to the lack of mechanistic understanding, there is currently almost no effective way to control the reconstruction of Cu. Recently, Li et al. showed that the surface atomic migration is driven by the interplay of electric fields and adsorbed reaction intermediates for cathodic metal electrocatalysts 18 . Except for the reaction intermediates, we have shown that surface adsorbed/ decorated species can also be introduced artificially to facilitate surface reconstruction 12 . Selective surface-capping additives have been widely used to guide the preferential exposure of crystal faces of the products in chemical synthesis 19,20 , it is thus of great potential as a simple but effective way to control surface reconstruction of metallic Cu catalysts for a practical CO 2 RR technology.
Here, we show our effort on the electrolyte additive way to well guide the surface reconstruction of commercial polycrystalline Cu (poly-Cu), and the additive presented here is ethylenediamine tetramethylenephosphonic acid (EDTMPA) to help achieve an excellent CO 2 electroreduction performance. Experimental and theoretical studies show that EDTMPA molecules are preferentially adsorbed on Cu(110) during the CO 2 RR, which not only induces the selective generation of Cu(110) facets that have an inherently high *CO binding strength, but also forms a local environment that promotes proton transfer from water to the Cu(110) facets. Since the ratedetermining step of CH 4 formation is *CO + *H → *CHO 21 , the simultaneously increased *CO and proton supplies together with *CHO stabilization caused by EDTMPA result in an excellent electrocatalytic CO 2 reduction performance for CH 4 production. In an H-cell configuration, commercial poly-Cu exhibits a CO 2 -to-CH 4 Faradaic efficiency (FE) of 61% with a partial current density of 25 mA cm −2 at −1.0 V versus a reversible hydrogen electrode (RHE) in a 0.5 M potassium bicarbonate (KHCO 3 ) electrolyte with 8 ppm EDTMPA, greatly outperforming the EDTMPA-free case. Even in an alkaline flow-cell configuration, where the CO 2 -to-CH 4 selectivity is usually completely suppressed according to previous reports 22,23 , our strategy is still highly efficient in obtaining a nearly constant CH 4 FE of 64 ± 2% at an operating current density of 300 mA cm −2 over 6 h. Consequently, a CO 2 -to-CH 4 conversion rate of 0.25 µmol cm −2 s −1 is achieved for commercial Cu-based electrocatalysts.

Results
CO 2 electroreduction performance. Electrocatalysis was performed at −1.0 V versus RHE in CO 2 -saturated 0.5 M KHCO 3 electrolytes with/without EDTMPA using an H-cell configuration ( Fig. 1 and Supplementary Fig. 1 amount (8 ppm) of EDTMPA, a commercial poly-Cu electrode has a much higher average CH 4 FE and partial current density (61% and 26.6 mA cm −2 ) than those (30% and 7.8 mA cm −2 ) in the EDTMPA-free electrolyte (Fig. 1a, b) during one-hour tests. EDTMPA also greatly increases the catalytic stability of poly-Cu electrodes during continuous electrocatalysis. In the EDTMPAfree case, the poly-Cu electrode retained only 35% and 20% of the initial FE and partial current density of CH 4 , respectively, accompanied by gradually increased CO production (Fig. 1c, d).
In sharp contrast, with the help of EDTMPA, the poly-Cu electrode showed an impressively stable FE and partial current density of CH 4 and a negligible CO increase (Fig. 1c, d). Considering that surface roughness can affect the electrocatalytic selectivity and activity 24 , the roughness factors of the poly-Cu electrodes were estimated from the cyclic voltammetry (CV) measurements. As a result, both the poly-Cu electrodes become rough and have similar roughness factors of 1.61 and 1.25 after electrocatalysis in the electrolytes with and without EDTMPA, respectively ( Supplementary Fig. 2), which eliminates the influence of surface roughness on the increased CO 2 electroreduction performance for CH 4 production.
Generation of Cu(110) induced by EDTMPA. Scanning electron microscope (SEM) images confirm that the originally smooth surface of the commercial poly-Cu electrode became rough after electrocatalysis in both electrolytes with and without EDTMPA, however, the resulting surface morphologies are significantly different ( Fig. 2a and Supplementary Fig. 3). For the one tested in the EDTMPA-free electrolyte, its rough surface has some aggregated irregular nanoparticles ( Supplementary Fig. 3c, d). In sharp contrast, massive uniformly distributed polyhedral nanoparticles were observed for the one tested in the EDTMPA-added electrolyte ( Fig. 2a and Supplementary Fig. 3e, f), which is different from the resulting cubic nanoparticles enclosed by {100} facets in previous studies 13,25,26 .
Since the commercial poly-Cu electrodes are difficult to prepare for the transmission electron microscope (TEM) characterization, we used electrodeposited Cu TEM grids as poly-Cu electrodes for the CO 2 RR and probed their crystal structure before and after electrocatalysis. Representative TEM images of the Cu grid after electrocatalysis in the EDTMPAadded electrolyte show a lot of Cu nanocrystals with hexagonal and cubic outlines, in good agreement with the ideal projections of a rhombic dodecahedral model bounded by {110} facets from different directions 27 (Fig. 2b and Supplementary Fig. 4). The Cu rhombic dodecahedrons are further confirmed by the highresolution TEM (HRTEM) image and the corresponding selected area electron diffraction (SAED) pattern of the equilateral hexagonal projection shape of Cu nanocrystals along [111] 27 (Fig. 2c). However, irregular Cu nanoparticles without any preferential surface orientation are merely observed for the electrodeposited Cu TEM grids before and after electrocatalysis in the EDTMPA-free electrolyte (Supplementary Figs. 5,6).
Grazing incidence X-ray diffraction (GIXRD) patterns show no surface phase other than Cu, and that the intensity of the Cu(220) peak at 74.1 o (JCPDS No. 04-0836) increases significantly for the poly-Cu electrode tested in the EDTMPA-added electrolyte (Fig. 2d). Its (220) to (111) peak intensity ratio is 4× that of the one tested in the EDTMPA-free electrolyte, which indicates that more Cu(110) surface planes are formed during electrocatalysis in the EDTMPA-added electrolyte.
The surface structure of the poly-Cu electrodes was further probed using the hydroxide (OH − ) electroadsorption technique 28,29 . Compared to the poly-Cu electrode tested in the EDTMPA-free electrolyte, the one tested in the EDTMPA-added electrolyte shows a pronounced Cu(110) OH − adsorption peak at~0.4 V versus RHE in its linear sweep voltammetry profile (Fig. 2e), suggesting a high surface density of (110). All the above results confirm that the EDTMPA induces the in-situ generation of Cu(110) surfaces during the reconstruction process. Density functional theory (DFT) simulations were performed to understand the role of EDTMPA in the generation of Cu(110). To simplify the calculations, the −CH 4 O 3 P fragment of an EDTMPA molecule (C 6 H 20 N 2 O 12 P 4 ) was taken with a H atom to passivate the terminal C atom as a methanephosphonic acid (MPA, CH 5 O 3 P) molecule, which was used to represent the huge EDTMPA molecule (Supplementary Fig. 7). The DFT results show that the MPA molecule has a higher adsorption energy on Cu(110) (−1.30 eV) than on Cu(100) (−1.06 eV) and Cu(111) (−1.11 eV) ( Fig. 2f and Supplementary Figs. 8,9), indicating that EDTMPA molecules prefer to be adsorbed on Cu(110). The adsorbed EDTMPA molecules stabilize the intrinsically highenergy Cu(110), and thus (110) surfaces are preferentially generated on the poly-Cu electrode during electrocatalysis in the EDTMPA-added electrolyte.
Adsorption of EDTMPA molecules on Cu(110). First of all, the stability of EDTMPA molecules is examined by CV tests, which were performed on poly-Cu electrodes between 0.3 V and −1.7 V versus RHE in Ar-saturated 0.5 M KHCO 3 electrolytes without and with a high concentration (100 ppm) of EDTMPA. The CV curves in both electrolytes are nearly identical, which confirms that the EDTMPA molecule is quite stable in the wide electrochemical window ( Supplementary Fig. 10). The adsorption of EDTMPA was verified by X-ray photoelectron spectroscopy (XPS). Obvious N and P elements were detected for the poly-Cu electrode tested in the EDTMPA-added electrolyte, but no for the one soaked in the EDTMPA-added electrolyte for the same time at an open circuit potential ( Supplementary Fig. 11). It is indicated that EDTMPA has a potential-driven specific adsorption behavior on the Cu surface 30 , which was further confirmed by in-situ Raman measurement ( Supplementary Fig. 12 Table 1).
The coverage of EDTMPA molecules on Cu(110) was also investigated to rule out the possibility of blocking the active sites for the CO 2 RR. A complete EDTMPA molecule was structurally relaxed on Cu(110) surface with a 2×3 supercell, which represents the close-packed structure of EDTMPA molecules on Cu(110) ( Supplementary Fig. 13a, b). Obviously, an adsorbed EDTMPA molecule only occupies up to two Cu atoms per six ones, while other Cu atoms are available for the CO 2 RR (marked with green circles in Supplementary Fig. 13a, b). Furthermore, we explored the transfer process of a CO 2 molecule through the EDTMPA adsorption layer on Cu(110) (Supplementary Fig. 13c, d). DFT results show that it is an exothermic process, indicating that CO 2 molecules can approach the Cu(110) surface without obstacle. Therefore, the adsorbed EDTMPA molecules do not block the CO 2 RR.
High *CO coverage on Cu(110). Previous studies argued that Cu(111) rather than Cu(110) favors the CH 4 yield during the CO 2 RR 32,33 , whereas Cu(110) seems to promote CH 4 generation in our work. As is widely reported, the rate-determining step for CH 4 production is the reaction (Supplementary Table 2) 21,34,35 .
Therefore, a higher *CO coverage thermodynamically favors the reaction (Eq. 1) and thus should promote CH 4 production. Among the three low-indexed Cu single crystal surfaces (111, 110 and 100), (110) has the strongest *CO binding strength 36 . To probe the interaction between *CO and the reconstructed Cu(110) facets, in-situ attenuated total reflectance-surface enhanced infrared absorption spectroscopy (ATR-SEIRAS) was conducted during the cathodic scan of a poly-Cu-coated Si crystal in a CO 2saturated 0.5 M KHCO 3 electrolyte with and without EDTMPA from 0.1 to −1.1 V versus RHE. In the vibration region of the linearly-bonded CO absorption band from 2040 to 2080 cm −1 (ref. 37 ), the spectra in electrolytes with and without EDTMPA are quite different (Fig. 3a). What is first apparent is that the electrode tested with EDTMPA always has a much stronger intensity of the surface-bonded CO peak than the one tested without EDTMPA during the cathodic scanning, as is more obviously seen using the integrated area in Fig. 3b, which shows that the reconstructed Cu(110) facets have much more *CO adsorption sites. When the potential is more negative than −0.47 V versus RHE, the integrated area of the CO peaks begins to decrease, which might be caused by either *CO desorption to deliver CO gas or *CO protonation to form *CHO. Another significant feature of the spectra, in the presence of EDTMPA, is that the CO band position has an initial blue shift and a subsequent red shift during cathodic scanning, which is in stark contrast to the EDTMPA-free case where only a red shift occurs at very negative potentials. The red shift is ascribed to the weakened *CO binding energy caused by the increased charge donation due to the potential-dependent electrostatic field in the double layer 38 . The initial blue shift is attributed to the strong adsorption of CO and hence the high *CO coverage on the reconstructed Cu(110) surfaces in the presence of EDTMPA 39 .
In-situ Raman and CO temperature-programmed desorption (TPD) measurements were also conducted to confirm the strong interaction of *CO with the generated Cu(110) surfaces. In the Raman spectra, the peaks located at 280 and 360 cm −1 are assignable to restricted rotation of adsorbed CO (P1) and Cu −CO stretching (P2), respectively 40,41 (Fig. 3c). As the P2-to-P1 intensity ratio is directly proportionate to the increased surface *CO concentration 40 , the EDTMPA-added case has obviously higher P2-to-P1 ratios than the EDTMPA-free one, indicating a higher *CO coverage on the reconstructed Cu(110) surface in the EDTMPA-added case. The CO TPD curve of the poly-Cu tested in the EDTMPA-added electrolyte has a pronounced broad CO desorption peak compared to the one tested in the EDTMPA-free case (Fig. 3d), showing a strong *CO binding energy of the generated Cu(110) surfaces. All the above results confirm that the reconstructed Cu(110) surfaces induced by EDTMPA have a high *CO coverage and facilitate *CO stabilization, agreeing well with a previous study 36 .
Locally high proton-feeding environment formed by EDTMPA. Although a restructured Cu(110) surface has been shown to have a strong *CO binding ability, the resulting high *CO coverage usually limits the number of *H adsorption sites 42,43 , which is in favor of ethanol formation but adverse to the rate-determining step of CH 4 generation (Eq. 1) 44,45 . This is the main reason why Cu(110) has rarely been reported to have a high CH 4 FE. Therefore, in addition to the high *CO coverage, an adequate proton supply is indispensable to the high CH 4 conversion in our work [46][47][48] . Considering that EDTMPA is a phosphonic acid, its adsorption on Cu(110) might form a local environment that provides a high proton supply.
To explore the function of EDTMPA on this, the CO 2 electroreduction performance was measured in electrolytes containing different amounts of EDTMPA. As the amount of EDTMPA increased from 0 to 8 ppm, both the FEs and partial current densities of CH 4 increased while those of H 2 remained almost unchanged (Fig. 4a, b). A reasonable interpretation of this is that EDTMPA adsorbed on Cu(110) provides adequate protons to help the protonation of *CO to *CHO rather than the formation of H 2 due to the high *CO coverage on Cu(110). When the amount of EDTMPA was increased to 16 ppm, an excessive proton supply resulted in increased H 2 production (Fig. 4a, b).
DFT calculations were carried out to explore the role of the adsorbed EDTMPA in increasing the number of protons available for CH 4 production. On bare Cu(110), protons are difficult to obtain from H 2 O because of the high dissociation barrier (1.32 eV) 49 ( Fig. 4c and Supplementary Fig. 14a). However, more protons are available after contacting EDTMPA with Cu(110). In details, one H atom is transferred from *MPA to Cu(110) with an energy decrease of 0.05 eV and a moderate kinetic barrier of 0.73 eV (Fig. 4d and Supplementary Fig. 14b). Subsequently, the *MPA that loses one H (*MPA−H) captures one H atom from the adjacent H 2 O molecule to become an *MPA again ( Fig. 4d and Supplementary Fig. 15). Note that this process is barrier free with an energy decrease of 0.10 eV, implying that it is both kinetically and thermodynamically favorable. The above DFT results clearly confirm that the adsorbed EDTMPA serves as a proton-delivering medium that accelerates the dissociation of water and continuously provides abundant *H for the conversion of CO 2 to CH 4 .
Apart from its high ability to provide protons, the adsorbed EDTMPA also stabilizes the *CHO species through hydrogen bond, which is supported by DFT calculations (Fig. 4e and Supplementary Figs. 16, 17). After introducing *MPA on Cu(110), the *CHO binding energy is significantly increased by about 0.44 eV, and the free energy change from *CO to *CHO (Eq. 1) is reduced from 0.79 to 0.41 eV (Fig. 4e), which suggests that the linear-scaling relations between the two coupled intermediates (*CO and *CHO) are broken, thus improving the kinetics of CH 4 formation. The improved kinetics were experimentally demonstrated by Tafel analysis with a reduced Tafel slope from 144 to 87 mV dec −1 in the presence of EDTMPA (Fig. 4f).
Feasibility study in a flow cell. Since poor CO 2 mass transport through electrolytes significantly limits the current density of the CO 2 RR in an H-cell configuration, the scalability of the additivecontrolled reconstruction approach was demonstrated in a flow cell configuration for industrial use. The results show that a high CH 4 selectivity is still achieved with the EDTMPA additive in an alkaline flow cell, which is totally different from the previous reports of a high C 2 selectivity in a pure alkaline electrolyte 23 (Fig. 5a). In a 1 M KOH electrolyte with 13 mM EDTMPA, the poly-Cu gas diffusion electrode (GDE) showed a very stable CH 4 FE of 64 ± 2% with a partial current density of 192 ± 6 mA cm −2 for 6 h (Fig. 5b), and thus a high CO 2 -to-CH 4 conversion rate of 0.25 µmol cm −2 s −1 was achieved (Fig. 5c). The high CH 4 selectivity indicates that EDTMPA still creates a locally high proton-feeding environment on poly-Cu electrodes even in an alkaline electrolyte, although the conditions were quite different from those in the neutral H-cell. The GIXRD results further illustrate that the surface reconstruction toward Cu(110) is also  Supplementary Fig. 18), indicating that the additive-controlled reconstruction approach has general applicability.

Discussion
A controlled surface reconstruction strategy using an EDTMPA additive substantially improves the CO 2 electroreduction performance to CH 4 on a commercial poly-Cu electrode. The CO 2 -to-CH 4 conversion is promoted by the increased availability of both *H and *CO substrates and stabilization of the *CHO resultant in the rate-determining step of CH 4 production (Eq. 1), all of which are resulted from surface reconstruction during the CO 2 RR. The surface reconstruction involves not only the Cu surface atomic re-arrangement induced by EDTMPA to generate Cu(110) facets, but also the formation of EDTMPA adsorption layer on Cu surface under working potentials.
To illustrate the effects of additives on the surface reconstruction and electrocatalytic performance, we also investigated two analogues of EDTMPA, namely, methylenediphosphonic acid (MDPA) and ethylenediamine tetraacetic acid (EDTA), as electrolyte additives for the CO 2 RR. According to SEM and X-ray diffraction (XRD) results, MDPA has an effect on the generation of Cu(110) surface like EDTMPA (Supplementary Fig. 19a, c). However, the CH 4 FE suffers from rapid decay accompanied with increased H 2 FE ( Supplementary  Fig. 20a), and their partial current densities are both decreased during one-hour electrocatalysis in the MDPA-added electrolyte (Supplementary Fig. 20b). This may be largely attributed to the smaller space structure of MDPA molecules and the resulting higher coverage on Cu surface compared with EDTMPA, which results in an excessive proton supply and limited CO 2 transportation. A substantial stability of the CO 2 -to-CH 4 conversion is achieved after adding EDTA into the electrolyte, although the CH 4 FE (~50%) and partial current density (18 mA cm −2 ) are lower than those in the EDTMPA-added case ( Supplementary Figs. 1 and 20c, d). This can be attributed to the similar proton-feeding capability of the carboxyl groups in EDTA with the phosphate groups in EDTMPA. The inferior performance for CH 4 production is due to the inability of EDTA to induce the atomic re-arrangement of Cu surface to generate Cu(110) like EDTMPA and MDPA (Supplementary Fig. 19b, c). Therefore, it can be concluded that the additives should be rational selected according to the desired products, although the selecting principles are still under exploration. Incidentally, the fact that the electrolyte effects are noticeable at low concentrations highlights that researchers need to pay careful attention to the presence of additives (intentional) or contaminations (unintentional or unknown) in their experiments. In this work, EDTMPA is presented as a successful example to illustrate the feasibility of manipulating the surface reconstruction of Cu metal by using selective surface-capping additives to achieve an CO 2 electroreduction performance. Our results show that the additive-controlled surface reconstruction approach has potentially general applicability. Therefore, we believe that more additives will be targetedly developed to facilitate CO 2 electroreduction to other high-value products, and such a strategy will be extended to other chemical process requiring high-performance electrocatalysts.

Methods
Chemicals. KHCO 3 (ACS Reagent 99.7%) was purchased from Sigma Aldrich, and EDTMPA (chelometric titration ≥95.0%) was purchased from TCI. All chemicals were used as received without further purification. Deionized water (18.2 MΩ cm −1 ) was used throughout all the experiments.

Preparation of electrodes.
In the H-cell, a commercial Cu mesh (0.5*3.3 cm 2 ) was electropolished in phosphoric acid (85 wt%) at 3 V versus a Cu counter electrode for 180 s, followed by rinsing with deionized water. The cleaned Cu mesh electrode was dried under a N 2 gas flow. In the flow cell, the GDEs were prepared by sputtering 100 nm thick Cu catalysts (Cu target, 99.999%) onto an AvCarb GDS 3250 GDL using a magnetron sputtering system. . GIXRD were performed on a Rigaku D/Max 2500 PC diffractometer (Cu-Kα, λ = 1.54056 Å). Elemental analysis was conducted using XPS (Thermo Fisher Scientific K-Alpha + , Al K α radiation, hv = 1486.6 eV). In order to avoid oxidation before ex-situ characterization, all the Cu electrodes after electrocatalysis were immediately washed with N 2 -saturated deionized water, following by blow-dry with N 2 as soon as possible, and finally were carefully preserved in a glove box under Ar atmosphere.
CO TPD. CO TPD was carried out on a TPDRO apparatus (TP-5080, Tianjin Xianquan Co. Ltd). Prior to the test, Cu electrodes tested with and without EDTMPA were pretreated in a He stream at 150°C for 1 h with a flow rate of 50 mL min −1 to clean the surface. After cooling to 50°C under a He atmosphere, the samples were subjected to adsorption of CO for 3 h. The residual CO was removed by purging with He for another 0.5 h. The desorption of CO was then performed by heating at a rate of 10°C min −1 from 50 to 400°C, and the TPD signal was recorded using a thermal conductivity detector.
Electrochemical OH − adsorption. Electrochemical OH − adsorption was performed in a N 2 -saturated 1 M KOH electrolyte using linear sweep voltammetry at a sweep rate of 100 mV s −1 for the poly-Cu electrode. The potential ranged from −0.2 to 0.6 V versus RHE.
Roughness factor (R f ). The roughness factor was estimated from the ratio of the double-layer capacitance (C dl ) between the working electrode and its corresponding smooth Cu foil electrode (assuming that the average double-layer capacitance of a polished smooth Cu electrode is 67 μF cm −2 ) 50 , R f = C dl /67. In-situ Raman. In-situ Raman spectra were obtained on a Raman spectrometer (LabRAM HR spectrometer, Horiba) with a laser wavelength of 785 nm. The COadsorption was monitored using a homemade electrolyzer with a CO 2 -saturated 0.5 M KHCO 3 aqueous solution as the electrolyte.
In-situ ATR-SEIRAS. Cu-coated Si ATR crystal was used as the working electrode for in-situ ATR-SEIRAS analysis in a CO 2 -saturated 0.5 M KHCO 3 aqueous solution with and without EDTMPA. A Thermo Fisher Nicolet IS50 spectrometer equipped with a MCT detector and a Pike Technologies VeeMAX III ATR accessory was used to collect the spectra. All spectra were acquired with a resolution of 4 cm −1 by accumulating eight scans.
Electrochemical measurements in the H-cell configuration. Controlledpotential electrolysis was conducted in an H-cell system, which was separated by an anion exchange membrane. An electropolished commercial Cu mesh was used as the working electrode. A graphite rod and a KCl saturated Ag/AgCl electrode were used as the counter and reference electrodes, respectively. The potentials were controlled by an electrochemical working station (IVIUM). All potentials in this study were measured against the Ag/AgCl reference electrode and converted to the RHE reference scale by E (vs. RHE) = E (vs. Ag/AgCl) + 0.197 V + 0.0591 × pH. The electroreduction of CO 2 was conducted in CO 2 -saturated 0.5 M KHCO 3 solution without or with different amounts of EDTMPA additives at room temperature and atmospheric pressure. The electrolyte was purged with CO 2 for at least 30 min to remove residual air in the reservoir.
Electrochemical measurements in the flow cell configuration. The controlledcurrent electrolysis was performed in an electrochemical flow cell with a threeelectrode system using an electrochemical working station (PARSTAT3000A-DX). 30 mL of 1 M KOH aqueous solution with or without 13 mM EDTMPA additive was circulated through the cathode chamber at a constant rate of 6.4 ml min −1 by a peristaltic pump. An anolyte using 1 M KOH was introduced to the anode chamber by a diaphragm pump. An anion exchange membrane (Fumasep FAB-PK-130) was used to separate the cathode and anode chambers. For the CO 2 RR, gaseous CO 2 (99.999%) was passed through the gas chamber at the back side of the Cu GDE (the exposed geometric area was 1 × 1 cm 2 ) at a flow rate of 14 ml min −1 . The CO 2 outlet flowrates were recorded by a mass flow detector (Alicat) for accuracy, which was used for all subsequent efficiency calculations. A piece of Ni foam and a Ag/AgCl (saturated KCl) electrode were used as the counter and reference electrodes, respectively. The ohmic loss between the working and reference electrodes was evaluated by electrochemical impedance spectroscopy and 80% iR compensation was applied to correct the potentials manually.
During the reaction, gas phase products were quantified with gas chromatography (SHIMADZU GC-2014), equipped with a thermal conductivity detector and a flame ionization detector. Liquid products were quantified with a 400 MHz 1 H NMR spectrometer with water suppression.
The FE of the catalysts were calculated using FE = αnF/Q, where α is the number of electrons transferred (α = 8 for CH 4 and 2 for H 2 production), n is the number of moles for a given product, F is the Faraday constant (96,485 C mol −1 ), Q is all the charge passed throughout the electrocatalysis process (measured by calculating the curve area of current density versus time plot). CH 4 and H 2 mole fractions were calculated using GC calibration curve.
Computational methods and models. All DFT calculations were carried out using the Vienna Ab Initio Simulation Package (VASP) 51 . The exchange correlation energy was represented by the Perdew-Burke-Ernzerhof (PBE) functional (Supplementary Table 3) within the generalized gradient approximation (GGA) and the electron-ion interactions were described by the projector augmented wave (PAW) method 52,53 . The (111) and (100) surfaces were constructed using four atomic layers with a 6×6 supercell, while a 4 × 6 supercell (four atomic layers) was used to simulate the (110) surface unless otherwise stated. Only the top two layers were fully relaxed during geometry optimizations. A vacuum layer of 20 Å was set between the periodically repeated slabs along the z direction, and the Brillouin zone was sampled by a Monkhorst-Pack k-point mesh of 3 × 3 × 1. The plane-wave cutoff energy was set to 500 eV, and the convergence criteria for force and energy difference were 0.05 eV/Å and 10 −5 eV, respectively. Dipole correction was applied to correct potential spurious terms a caused by the asymmetry of the slabs 54,55 . To take into account the van der Waals (vdW) interactions, the empirical correction in Grimme's method (DFT + D3) was used 56 . The minimum energy paths and saddle points were conducted with the climbing image nudged elastic band (CI-NEB) method 57 .
The computational hydrogen electrode (CHE) model 58 was used to calculated the Gibbs reaction free energy change (ΔG) of each step. The chemical potential of the proton-electron pair in an aqueous solution is related to that of one-half of the chemical potential of an isolated hydrogen molecule. In this model, the ΔG value can be obtained by the formula: ΔG = ΔE + ΔZPE − TΔS, where ΔE is the reaction energy of the reactant and the product species adsorbed on the catalyst, which was directly obtained from DFT calculations; ΔZPE and ΔS are the changes in zero point energies and entropy at 298.15 K, which were calculated from the vibrational frequencies.
The adsorption energy (E ads ) of the MPA (CH 5 O 3 P, one leg of the EDTMPA molecule) on the (100), (110) and (111) surfaces of Cu was calculated based on the equation: E ads = E total − E substrate − E adsorbate , where E total , E substrate and E adsorbate are the total energies of the systems containing the substrate and adsorbate, the substrate, and the adsorbate, respectively. According to this definition, a more negative adsorption energy indicates a stronger adsorption.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.