Sulfur‐Decorated Ni−N−C Catalyst for Electrocatalytic CO2 Reduction with Near 100 % CO Selectivity

Abstract Developing highly efficient electrocatalysts for electrochemical CO2 reduction (ECR) to value‐added products is important for CO2 conversion and utilization technologies. In this work, a sulfur‐doped Ni−N−C catalyst is fabricated through a facile ion‐adsorption and pyrolysis treatment. The resulting Ni−NS−C catalyst exhibits higher activity in ECR to CO than S‐free Ni−N−C, yielding a current density of 20.5 mA cm−2 under −0.80 V versus a reversible hydrogen electrode (vs. RHE) and a maximum CO faradaic efficiency of nearly 100 %. It also displays excellent stability with negligible activity decay after electrocatalysis for 19 h. A combination of experimental investigations and DFT calculations demonstrates that the high activity and selectivity of ECR to CO is due to a synergistic effect of the S and Ni−N X moieties. This work provides insights for the design and synthesis of nonmetal atom‐decorated M−N−C‐based ECR electrocatalysts.


Introduction
About 500 gigatons of carbon dioxide will be generated from the combustion of fossil fuels in the next half century. [1] Excessive CO 2 emissions have aroused environmental issues such as global warming and erratic weather. [2,3] To alleviate these adverse effects, developing efficient carbon capture, utilization and storage (CCUS) technologies is of paramount importance. In recent years, electrochemical CO 2 reduction (ECR), potentially powered by renewable electricity sources, has received extensive attention. ECR can convert CO 2 to valueadded products such as C 1 (e. g., CO, HCOOH, CH 4 ), C 2 (e. g., C 2 H 4 , C 2 H 5 OH), and C 3 (e. g., C 3 H 7 OH) under ambient temperature and pressure. [4][5][6][7] In particular, it has been shown that ECR to CO by a two-electron reduction process is the most promising for commercialization, owing to high selectivity and low energy requirement. [8,9] Moreover, CO as an important chemical feedstock plays a vital role in various industries. [10] Even though much progress has been made in ECR to CO, it is still far from commercial application for reasons such as unsatisfactory activity, poor stability, and high cost. In addition, the competing hydrogen evolution reaction (HER) severely suppresses the CO selectivity in aqueous solutions. [11] Therefore, it is desirable to exploit CO 2 -to-CO electrocatalysts with excellent activity and high selectivity.
Among CO 2 -to-CO electrocatalysts explored, nitrogendoped carbon-based material-supported transition metal atoms (MÀ NÀ C) exhibit great potential for CO generation. [12][13][14][15][16][17] Carbonbased materials are utilized, owing to good electrical conductivity, which benefits electron transportation. Among these materials, carbon black (CB) not only exhibits merits such as large surface area and confinement of metal atoms, but also facilitates CO 2 diffusion across the gas diffusion barrier and accelerate the reaction rate. [18][19][20] Therefore, CB serves as substrate that could cost-effectively improve the atomic dispersion of M and the activity of ECR to CO. The MÀ N X moieties in these catalysts are usually regarded as active center where X denotes the coordination number of metal atoms. In addition, the MÀ N X moiety with a planar coordination structure has been confirmed to stabilize the metal atoms effectively. [21] More interestingly, nonprecious metals, such as Fe, Co, Ni, or Mn, in MÀ N X system could achieve identical and even better performance than noble metal-based electrocatalysts. For instance, isolated FeÀ N 4 structure on carbon materials could lower the energy barrier for the formation of *COOH intermediate during ECR to CO, thus enhancing the CO selectivity with a faradaic efficiency [FE(CO)] up to 93 %. [22] Co atom dispersed on N-doped porous carbon sphere was found to have > 90 % FE(CO) under potentials between À 0.57 and À 0.88 V versus the reversible hydrogen electrode (vs. RHE). Spectroscopic studies and density functional theory (DFT) calculations revealed that CoÀ N 5 site serves as the active center for boosting CO 2 activation and facilitating the formation of *COOH as well as CO desorption. [23] NiÀ N 4 moiety confined in Ni porphyrin-based framework displayed maximum FE(CO) of 97 % at À 0.90 V (vs. RHE) and good stability. Theoretical simulation demonstrated that NiÀ N 4 center could decrease the energy barrier for *COOH intermediate generation. [24] MnÀ N 3 embedded into graphitic carbon nitride shown 98.8 % of FE(CO) under a low overpotential of 0.44 V, and the outstanding activity was attributed to MnÀ N 3 moiety promoting the formation of the key intermediate *COOH. [25] Therefore, it can be concluded that the superior CO 2 -to-CO performance on these non-precious metalbased MÀ NÀ C electrocatalysts can be mainly attributed to the proper interaction between MÀ N X center and intermediates.
Recently, studies disclosed that the intrinsic activity of MÀ NÀ C could be further enhanced by introducing other heteroatoms. For instance, Zhang et al. reported that extra chlorine atoms (Cl) coordinated MnÀ N 4 À C could change the electron transfer of Mn atom, boosting *COOH adsorption and CO desorption. [26] As a result, the electrocatalyst exhibited a FE(CO) of 97 % and a high current density of about 10 mA cm À 2 under a low overpotential of 0.49 V. Pan et al. reported that S atoms could lift the Fermi energy of Fe3d state and increase the charge density of Fe atoms. [27] The inherent electrocatalytic activity and selectivity for CO 2 -to-CO were modulated by S atoms, resulting in improved interaction strength between Fe atom and intermediates. This unique structure and electronic properties endowed the FeÀ NSÀ C electrocatalyst with a FE(CO) of 98 % under an overpotential of 0.49 V. Indeed, the local electronic structures of host atoms could be altered by doping heteroatoms, further influencing catalytic properties. [28][29][30] Thus, it can be anticipated that the incorporation of proper heteroatoms into carbon sites close to the MÀ N X center could tune the electronic structure of metal atoms, which would effectively promote the activation of CO 2 and modify the binding strength of intermediates during the ECR process. Specifically, incorporating S atom into host materials has been regarded as an effective strategy to improve the activity of various catalytic reactions. [31][32][33] NiÀ NÀ C electrocatalyst have been reported to exhibit good CO 2 -to-CO activity. However, there is limited study on S modified NiÀ NÀ C system to regulate the ECR activity for CO 2 -to-CO. In this study, we synthesized N,S-codoped CB incorporating Ni atoms by facile ion-adsorption and subsequent pyrolysis treatment. The NiÀ NSÀ C catalyst exhibited a very high conversion efficiency of 99.7 % to CO with a high total current density of 20.5 mA cm À 2 under À 0.80 V (vs. RHE), outperforming S-free NiÀ NÀ C electrocatalyst and other ECR catalysts reported in literature. It also displayed excellent stability without activity decay after electrocatalysis for 19 h. First-principles computation was carried out to investigate the effect of the S atom decoration on the electronic structures of Ni atoms, catalytic mechanisms, and activity toward ECR as well as HER.

Electrocatalysts characterization
The structures of NÀ C, NSÀ C, NiÀ NÀ C, and NiÀ NSÀ C catalysts were firstly characterized by X-ray diffraction (XRD). As shown in Figure 1a, the catalysts exhibit similar XRD patterns with two broad diffraction peaks at around 25.1°and 43.2°, corresponding to the (002) and (100) planes of carbon. It is worth noting that both peaks of NSÀ C and NiÀ NSÀ C catalysts show low crystallinity and slight right-shift compared with the NÀ C and NiÀ NÀ C catalysts, which can be explained by lattice contraction and the formation of C vacancy after the introduction of S atoms with larger radius. [34,35] Furthermore, no peaks assignable to metallic Ni or its compounds were observable.
Raman spectra of the four catalysts exhibit two vibrational bands located around 1343 cm À 1 (D band) and 1594 cm À 1 (G band; Figure 1b), which are ascribed to the defect and graphitic sp 2 carbon. [36] The intensity ratios of D and G band (I D /I G ) of NÀ C, NSÀ C, NiÀ NÀ C, and NiÀ NSÀ C catalysts were also calculated as marked in Figure 1b. It can be observed that the introduction of S atom into the NÀ C catalysts induces more defects, consistent with XRD study. However, incorporating Ni atoms decreases the I D /I G value, indicating that Ni atoms are embedded into C vacancies. Compared with the NiÀ NÀ C catalyst, the position of the G band of NiÀ NSÀ C catalysts shows a slight upshift owing to doping of nonmetal S atoms into carbon-based materials. [37][38][39] Transmission electron microscopy (TEM) was used to investigate the microscopic morphology of the NiÀ NSÀ C catalyst. Carbon nanospheres with a diameter of 50 nm were observed (Figure 2a). The high resolution transmission electron microscopy (HRTEM) images displays distorted short-range graphic stripes with winkle and interlaces (Figure 2b), indicating defective carbon structure. No distinct nanoparticles or clusters were observed, implying that the Ni atoms are likely to present in the form of single atoms. Moreover, energy-dispersive X-ray spectroscopy (EDX) elemental mapping images clearly demonstrate that Ni, N, S and C species distribute uniformly over the NiÀ NSÀ C catalyst ( Figure 2c). Furthermore, the aberrationcorrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) shows the well-dispersed single Ni atoms as bright spots for the NiÀ NÀ C and NiÀ NSÀ C catalysts, which are highlighted by red circles in Figure 2d and e.
The surface area and pore structure of the catalysts were determined by N 2 adsorption-desorption measurements. As depicted in Figure 3a, the isotherms of the four electrocatalysts displayed sharp adsorption under relative pressures greater than 0.40 accompanied by an obvious hysteresis loop, which is indicative of dominant mesopores and is further corroborated by the pore size distributions (Figure 3b). The specific surface area and pore volume are summarized in Table S1 (see the Supporting Information). It can be found that the total pore volume increased from 1.26 to 1.39 m 3 g À 1 after doping S to the NÀ C structure, and increased further after introducing Ni atoms. The catalysts show quite close but very high surface areas in the range of 1073 to 1275 m 2 g À 1 , which also increase with the doping of S and Ni atoms. The high surface area and pore volume are beneficial for the dispersion of Ni atoms and the access of reactants to the active centers. As expected, the pore size distribution curve also shows more larger pores with the introduction of S and Ni atoms ( Figure 3b).
X-ray photoelectron spectroscopy (XPS) was employed to characterize the chemical state and surface concentration of the elements. The high-resolution N1s spectra of the NiÀ NSÀ C catalyst can be fitted into four peaks cantered at 397.9, 399.7, 400.5 and 403.1 eV (Figure 4a), which can be assigned to pyridinic N (Pyri-N), pyrrolic (Pyrr-N), graphitic N (Grap-N) and oxidized N (Oxid-N), respectively. [40][41][42][43] The existence of these types of N could promote the electrocatalytic activity. [44] In addition, N atom concentrations are 4.43, 4.19, 4.11 and 4.18 at.% for NÀ C, NSÀ C, NiÀ NÀ C, and NiÀ NSÀ C (Table S1). High-resolution S2p spectra of NiÀ NSÀ C catalyst ( Figure 4b) at lower binding energy can be ascribed as CÀ SÀ C (2p 3/2 at 164.1 and 2p 1/2 at 165.3 eV) and CÀ SO x À C (167.6 and 168.8 eV). [34,45] The S content was estimated to be 0.37 and 0.42 at.% for the NSÀ C and NiÀ NSÀ C catalyst, respectively. It was also observed that the percentages of N in these two catalysts remain almost unchanged, implying that introducing S atoms had little effects on the bonding patterns of N atoms. In the high-resolution Ni2p spectrum (Figure 4c), the Ni2p 3/2 binding energies for NiÀ NSÀ C and NiÀ NÀ C catalyst are 855.90 and 855.66 eV, higher than that of Ni 0 (852.5-853.0 eV) but lower than that of Ni 2 + (856 eV), [46,47] indicating that Ni species are likely to keep as ionic Ni δ + (0 < δ < 2). Therefore, these results further demonstrated the existence of single Ni atoms on the surface of NiÀ NSÀ C and NiÀ NÀ C catalysts. [48] Moreover, the peak of Ni2p 3/2 shifted slightly towards higher binging energy after incorporating S atoms, indicating that S could influence the electronic structure of Ni. The Ni contents in NiÀ NÀ C and NiÀ NSÀ C catalysts are 0.50

Electrocatalytic activity test
The ECR performances of catalysts were evaluated in a membrane-separated two chambers H-type cell with a standard three-electrode system immersed in 0.5 m KHCO 3 electrolyte. Under CO 2 -saturated electrolyte, the Linear sweep voltammetry (LSV) of NÀ C and NSÀ C catalysts show small current density (Figure 5a), exhibiting low electrocatalytic activity. Conversely, the NiÀ NSÀ C and NiÀ NÀ C catalyst display large current density thus high electrocatalytic activity. In addition, the doping of S atoms could boost the current density for both NÀ C and NiÀ NÀ C, confirming the role of S atom in the activity enhancement. In the presence of CO 2 , the current density of NiÀ NSÀ C increases faster in comparison with the reaction under N 2 atmosphere (Figure 5b), suggesting the enhanced current density from ECR. The results demonstrated that introducing NiÀ N X moiety into carbon-based materials could indeed improve its electrocatalytic performance. Meanwhile, the highest current density observed for the NiÀ NSÀ C catalyst could be fairly attributed to the synergistic effect of NiÀ N X moiety and S doping.
To quantify the product selectivity, the ECR test was conducted in constant potentiostatic electrocatalysis under different potentials. H 2 and CO were detected as the only reduction products under the potential window of À 0.20 V to À 1.20 V (vs. RHE). No other gaseous products were detected by GC, because CO, as a two-electron product, is more easily generated than other four-, six-, or eight-electron products such as CH 3 OH and CH 4 , consistent with previous studies. [49,50] Moreover, no liquid products were produced, as evidenced by 1 H NMR analysis (see the Supporting Information, Figure S1). As shown in Figure 5c, the FE(CO) over the four catalysts showed first increase then decrease with the decrease of applied potential. However, the FE(CO) of NiÀ NÀ C and NiÀ NSÀ C catalysts are always higher than the NÀ C and NSÀ C catalysts under the same applied potential, indicating that NiÀ N X are the true active centers. Meanwhile, the NiÀ NSÀ C catalyst exhibits larger FE(CO) than NiÀ NÀ C over the entire potential range, demonstrating that S atom could effectively enhance the ECR performance of the NiÀ N X moiety. The competing HER performance for the four catalysts were also compared, as displayed in Figure 5d. The FE(H 2 ) of NiÀ NSÀ C catalyst is always lower than that of NiÀ NÀ C catalyst, indicating that the HER performance of the NiÀ NÀ C catalyst was suppressed after the doping of S atoms. In contrast, the FE(H 2 ) of NÀ C is lower than that of NSÀ C, suggesting that S atom could promote HER ability of NÀ C in the absence of Ni. Therefore, it can be deduced that more protons are involved in ECR compared to HER after the introduction of Ni and S atoms, and there is a synergistic effect between Ni and S. As a result, the NiÀ NSÀ C catalyst exhibits high FE(CO) over 90 % in a broad potential range of À 0.60 to À 1.10 V (vs. RHE), and the maximum FE(CO) is as high as 99.7 % at a potential of À 0.80 V (vs. RHE) with a total current density of 20.5 mA cm À 2 .
The increase of electrochemical active surface areas (ECSA) also contributes to the excellent activity towards ECR. ECSA can be directly estimated by measuring double layer (D-L) capacitance (Figure 5e and Figure S2). The NiÀ NSÀ C and NiÀ NÀ C catalysts exhibit 25.0 mF cm À 2 and 21.6 mF cm À 2 , which are larger than that of the NSÀ C (15.1 mF cm À 2 ) and NÀ C catalysts (10.0 mF cm À 2 ). Consequently, the corresponding ECSA for NiÀ NSÀ C, NiÀ NÀ C, NSÀ C and NÀ C catalysts were 625, 540, 377 and 250 cm 2 ECSA , demonstrating that Ni and S atoms could effectively increase the ESCA (Table S1). It can also be found that there is a positive correlation between ECSA and BET surface area (Figure 5f), validating that larger surface area could expose more active sites. CO 2 adsorption on catalyst surface plays an important role in ECR. Therefore, we conducted temperature-programmed CO 2 desorption (CO 2 -TPD) to investigate the effect of S dopant on their CO 2 adsorption ability. It turns out that NiÀ NSÀ C shows slightly stronger CO 2 adsorption than that of NiÀ NÀ C, which could boost the ECR performance ( Figure S3). To further elucidate the effect of S atom on the reaction kinetics, the Tafel slope and electrochemical impedance spectroscopy (EIS) were conducted.
As shown in Figure 6a, the NiÀ NSÀ C catalyst exhibits a lower Tafel slope of 182 mV dec À 1 than NiÀ NÀ C of 193 mV dec À 1 , suggesting that the introduction of S atom could improve the reaction kinetics. The Tafel slopes of these two samples are close to the theoretical value, revealing that CO 2 -to-CO on these two catalysts proceed via the same mechanism that CO 2 accepting proton-electron pairs to form *COOH intermediate is the potential determining step (PDS). [51] The EIS test was further conducted at À 0.80 V (vs. RHE) and the corresponding complex-plane plot of NiÀ NÀ C and NiÀ NSÀ C catalysts is shown in Figure 6b. The NiÀ NSÀ C catalyst exhibited a smaller charge transfer resistance (R ct ), demonstrating that incorporating S atom could accelerateelectron exchange between the catalysts and reactants. In addition, the electrochemical stability of NiÀ NSÀ C was evaluated by chronoamperometric electrolysis under the potential of À 0.80 V (vs. RHE) with the largest FE(CO). As shown in Figure 6c, there is a fast drop in current density during the initial 20 min, which can be attributed to unbalanced CO 2 adsorption on the catalyst. After that, the current density and FE (CO) is highly stable with negligible decay. The current density is still around 20.5 mA cm À 2 and the FE (CO) is about 98 % after electrocatalysis reaction for > 19 h. We also studied the structure and element information of the spent NiÀ NSÀ C catalyst by TEM, EDX mapping and XPS (Figures S4 and S5; summarized in Table S1). It can be found that there is no obvious change in morphology and element content after ECR reaction, demonstrating excellent stability of the NiÀ NSÀ C catalyst. Very recent reports on nonmetal-decorated MÀ NÀ C catalysts are summarized in Table S2. Remarkably, the NiÀ NSÀ C catalyst in this study outperformed most of the catalysts in terms of applied potential, FE(CO) and stability.

DFT simulations
DFT calculations were performed to gain insight into the electrocatalytic reaction mechanisms of ECR and HER on NiÀ NÀ C and NiÀ NSÀ C catalysts based on the CHE model. As Ni atoms are more likely to form stable NiÀ N 4 moiety in carbonbased materials, the NiÀ N 4 structure was created in carbon matrix as active center in NiÀ N 4 À C catalyst. [13,[52][53][54] It is worth noting that similar method for synthesizing FeÀ N 4 À C catalyst has been reported recently. [55] The good ECR activity of Nidecorated nitrogen-doped carbon catalysts was previously attributed to single Ni atoms instead of Ni nanoparticles. [48] XRD indicated that the introduction of S atoms possibly induced the formation of C atom vacancies because the radius of S atom is larger than that of C atom (Figure 1a). To exclude the possibility of S directly substituting C atom without breaking TMÀ N 4 , we also built a configuration as shown in Figure S6, where S directly substituted C atom but TMÀ N 4 was maintained. Compared to the pristine S free structure, the CO desorption ability will be quite weak in this structure because of high free energy change of 1.46 eV, whereas the HER ability was strengthened owing to low free energy of À 0.53 eV. These results are not consistent with the experimental data. Therefore, S atom embedded into C vacancy close to Ni site (NiÀ NSÀ C) as model was proposed to investigate the effect of S atoms on the ECR performance of the NiÀ N 4 À C catalyst (Figure 7a). The adsorption energy of CO 2 on the NiÀ NÀ C and NiÀ NSÀ C catalysts were calculated to be À 0.28 and À 0.39 eV, further demonstrat- ing that S dopant could promote CO 2 adsorption, consistent with the CO 2 -TPD results. The adsorbed *COOH, *CO and *H intermediates were considered in ECR and HER. As shown in Figure 7b, the Gibbs free energy change (ΔG) diagram of CO 2to-CO over these two catalysts suggests that the first protonelectron pair to generate *COOH is the PDS, consistent with the experimental results. The NiÀ NÀ C catalyst shows ΔG *COOH of 1.86 eV, which is higher than that of the NiÀ NSÀ C catalyst at 1.66 eV, confirming that S atom could effectively decrease the CO 2 activation barrier. This is also consistent with the experimental result that introducing S atoms could enhance FE(CO) under all applied potentials. The calculated adsorption energy of CO over the NiÀ NÀ C catalyst was reduced by 0.06 eV after the incorporation of S atom, demonstrating the increase of CO desorption ability. The HER activities over the two catalysts were compared by calculating the Gibbs free energy of *H (ΔG *H ; Figure 7c). Notably, the ΔG *H of NiÀ NSÀ C catalyst increases from 1.53 to 1.82 eV, showing that S atom could effectively suppress HER when Ni is present, which again agrees with the experimental observation that FE(H 2 ) of the NiÀ NÀ C catalyst decreased after doping S (Figure 5d).
It has been well established that the difference between the limiting potentials of ECR and HER [U L (CO 2 )À U L (H 2 )] could be a reasonable descriptor for ECR selectivity. A more positive U L (CO 2 )À U L (H 2 ) indicates a higher selectivity towards ECR. As displayed in Figure 7d, the NiÀ NSÀ C catalyst shows a positive value of U L (CO 2 )À U L (H 2 ), which is negative for the NiÀ NÀ C catalyst, demonstrating increased ECR selectivity after incorporating S atoms. Furthermore, the electronic structure of NiÀ NSÀ C and NiÀ NÀ C catalysts were compared. The d band center of NiÀ NÀ C catalyst moved towards the Fermi level after introducing the S atom, and the Ni3d state of NiÀ NSÀ C catalyst crossed the Fermi level (Figure 7e), thereby improving the intrinsic activity of 3d electrons. When the *COOH intermediate was adsorbed on the catalysts, Ni atom (NiÀ NSÀ C) lost smaller charge (0.89e À ; Figure 7f), accounting for moderate interaction. The results demonstrated that the synergistic effect of the Ni and S atom boosts the ECR performance where the Ni atoms sever as the active center and S atom plays a role in the modification of electronic properties.

Conclusion
We have developed a facile method for the synthesis of a Sand Ni-doped NÀ C system as an electrocatalyst for CO 2 reduction to CO. The NiÀ NSÀ C catalyst can selectively reduce CO 2 to CO with high a FE(CO) of over 90 % in a broad potential range of À 0.60 to À 1.10 V (vs. RHE). The maximum FE(CO) reaches 99.7 % at À 0.80 V (vs. RHE) with a total current density of 20.5 mA cm À 2 . It also exhibits excellent stability for 19 h electrocatalysis without apparent activity decay. Experimental results demonstrated that Ni atoms served as active sites for ECR to CO, whereas S atoms could increase its activity further. Moreover, theoretical calculations disclosed that doping S atoms could decrease and increase the free energy barrier for the formation of *COOH and *H, respectively. Meanwhile, S atoms could lift the d band center of Ni atoms and induce the Ni3d state crossing the Fermi energy level. As a result, the performance of ECR to CO on NiÀ NSÀ C was improved, whereas the hydrogen evolution reaction (HER) was suppressed, contributing to overall higher ECR selectivity. Therefore, the excellent ECR performance of the NiÀ NSÀ C catalyst can be attributed to the synergistic effect of the NiÀ N X moiety and S dopant.

Electrocatalyst preparation
The CB was firstly activated by concentrated HNO 3 solution to increase surface defects and oxygen-containing groups. Typically, 4 g CB was dispersed in 100 ml of HNO 3 solution followed by refluxing at 100°C for 8 h with vigorously stirring. Subsequently, the suspension was washed with DI water several times until neutral pH and separated by vacuum filtration. Activated CB was obtained after drying at 120°C in a vacuum oven for 24 h.
In a typical synthesis of Ni 2 + adsorbed on CB (Ni 2 + -CB), 1 g activated CB was dispersed in 400 mL DI water under sonication for 2 h. The Ni 2 + solution (3 mg mL À 1 ) was prepared by dissolving 240 mg Ni (NO 3 ) 2 · 6H 2 O in 80 mL DI water. Thereafter, 40 mL Ni 2 + solution was added dropwise into the CB solution and kept under vigorous stirring for 12 h. The products were collected by vacuum filtration. After drying at 120°C in a vacuum oven for 24 h, Ni-CB was obtained.
The NiÀ NSÀ C catalyst was synthesized according to a modified method by Zheng et al. [18] Typically, 0.5 g Ni-CB and with 6.0 g thiourea was mixed and grinded to obtain fine powders. The powder was then transferred into a covered crucible and heated to 800°C at a heating rate of 3°C min À 1 under 10 mL min À 1 Ar flow and kept at 800°C for 1 h. For comparison, the NiÀ NÀ C catalyst was synthesized by replacing thiourea with urea. Metal free NÀ C and NSÀ C catalysts were also prepared by replacing Ni-CB with CB.

Electrocatalysts characterization
The crystalline structures of the prepared samples were analyzed by XRD (Bruker-AXS Micro-diffractometer D8 ADVANCE) equipped with a Cu Kα radiation source (λ = 1.54 Å) with a scan rate of 3°min À 1 . The crystallinity was further examined by Raman spectroscopy (Renishaw, with a 532 nm excitation laser). The samples were focused with a × 50LWD objective lens and exposed to emission line for 10 s. Nitrogen adsorption-desorption isotherms were measured on Micromeritics Tristar II 3020 instrument at À 196°C. The specific surface area was estimated by the Brunauer-Emmett-Teller (BET) method, and the pore size distribution was obtained from the Barrett-Joyner-Halenda (BJH) desorption isotherm. The CO 2 -TPD profile was measured on Micromeritics Autochem II ASAP 2920.
XPS was conducted on a 1486.6 eV X-ray photoelectron spectrometer (ESCALAB Xi-type) using Al Kα source.

Electrochemical measurements
All electrocatalytic tests were conducted by a standard threeelectrode system in an H-type cell with two compartments at room temperature (25°C). Platinum plate, catalysts coated carbon paper (Toray, TGP-H060) and Ag/AgCl (3 m KCl) served as the counter electrode, working electrode and reference electrode, respectively. All potentials were controlled by AUTOLAB PGSTAT302N workstation and converted to the reversible hydrogen electrode (RHE) by E (vs. RHE) = E (vs. Ag/AgCl) + 0.197 V + 0.059 × pH. The pH of 0.5 m KHCO 3 saturated with CO 2 is 7.2. Each chamber contained 40 ml 0.5 m KHCO 3 . The catalysts ink was prepared by sonicating a mixture of 3.0 mg catalysts, 145 μL ethanol, 90 μL DI water and 65 μL 0.5 wt.% Nafion-D521 solution for 1 h. The working electrode was prepared by loading 60 μL catalysts ink onto a carbon paper (1 cm 2 ) with a mass loading of 0.6 mg cm À 2 , followed by drying under an infrared lamp for 30 min. A piece of Nafion-117 ionic exchange membrane was used to separate the anode chamber (counter electrode) and cathode (reference and working electrodes). High purity CO 2 was pumped into the cathode chamber at a flow rate of 20 mL min À 1 for 1 h before and during the electrolysis test. The 0.5 m KHCO 3 electrolyte in the cathode was stirred at 800 rmp during the test. LSV was firstly performed under N 2 and CO 2saturated 0.5 m KHCO 3 electrolyte at a scan rate of 10 mV s À 1 . The electrochemical double-layer capacitance (C dl ) was evaluated by the cyclic voltammetry (CV) method to estimate ECSA of the catalysts. The CV tests were conducted in 0.5 m KHCO 3 electrolyte with a potential from 0 to À 0.20 V (vs. Ag/AgCl) and various scan rates (10, 20, 40, 60 and 80 mV s À 1 ) to avoid the faradaic process. C dl can be obtained by plotting the anodic and cathodic current difference at À 0.10 V against the scan rate. The ECSA can be determined by Equation (1) [56] below: where C dl is the slope of the double-layer charging current against the scan rate, and the value of C s was chosen to be 40 μF cm À 2 cm À 2 ECSA . The gaseous products were analyzed by on-line gas chromatography (Agilent GC-7890B) equipped with a HayeSep Q column and a 5 A molecular sieve column. The possible liquid products were determined by nuclear magnetic resonance (NMR) spectroscopy. The NMR spectra were recorded on a 400 MHz Bruker NMR spectrometer in deuterium oxide (D 2 O) with one drop of dimethyl sulfoxide. 1 H NMR chemical shifts were recorded in D 2 O. The FE of gaseous products (H 2 and CO) under different potentials were calculated according to Equation (2): [23] (2) where V, v and I represent the gas flow rate, volumetric concentration of CO or H 2 , and steady-state cell current at different potentials, respectively.

Computational methods
All spin-polarized calculations, including structure relaxation, electronic structure and energy, were performed by Vienna ab initio simulation package (VASP) using the plane-wave basis. [57,58] The generalized gradient approximation (GGA) in the parametrization of Perdew-Burke-Ernzerhof (PBE) were employed to describe the electronic exchange and correction. [59,60] Projector augmented wave (PAW) pseudopotential was used to treat the interaction of core and valence electron. [61] The long-range van der Waals (vdW) interactions was considered by incorporating empirical correction method (DFT-D3). [62] The kinetic cut-off energy was set to 500 eV. kpoint sampling of 8 × 8 × 1 (structure relaxation) and 10 × 10 × 1 (electronic structure calculations) for all systems were tested to achieve convergence. The convergence criterion for energy and force was set to 1.0 × 10 À 5 eV and 1.0 × 10 À 2 eV Å À 1 , respectively.
The Gibbs free energy change (~G) of intermediates in each step for ECR and HER were calculated by the computational hydrogen electrode (CHE) model: [63] DG where ΔE, ΔE ZPE and ΔS are the differences in total energy obtained from DFT calculations, zero-point energy, and entropy between reactants and products, respectively. T is temperature (298.15 K).