Electrochemical carbon dioxide (CO₂) reduction reaction (CO₂RR) to value-added chemicals by renewable electric energy offers a promising strategy to address global warming and energy shortage issues [1-3]. Among the CO₂RR products, C₁ chemicals such as carbon monoxide (CO) benefit from their simple kinetics process, high selectivity as well as low separation cost. Thus, it has more positive economic benefits for industrialization compared with C₂ or C₃ counterpart [4, 5]. To date, NiN_xC_y single-atom catalysts (SACs) is one of most promising electrocatalysts for CO₂RR to CO due to its high catalytic selectivity as a result from unfavorable competing hydrogen evolution reaction (HER) [6-12]. Furthermore, previous research has shown that judiciously modified the architecture of SACs is of equal importance for CO₂RR as it involves the consumption of gas and water [13-15]. For example, several explorations demonstrated that incorporating single-atom moieties on self-supporting substrate to form self-supported SACs would increase the area of liquid-gas-solid interface and avoid the weak contacts between substrate and SACs in traditional powdery catalysts. Such structure would exhibit industrial-level current density and stability [16, 17].

Although the newly developed self-supported SACs integrated the advantages of SACs and self-supporting electrode to exhibit the prospect of industrialization, the current synthesis strategies for self-supported SACs are still limited by tedious synthetic steps, including the preparation and assemble process of SACs precursors [18, 19]. The top-down synthetic routes can directly convert the metal nanoparticles (NPs) into SACs at high temperature, which makes it possible to prepare self-supported SACs via one-pot method and greatly simplify the preparation process [20-22]. Previous studies employed the Ni NPs loaded on self-supporting substrate such as carbon nanofibers (CNF) as seeds to grow N-doped carbon nanotubes (NCNTs). Ni single-atom moieties has been trapped by the defects in NCNTs and then achieved simple and scalable fabrication of self-supported Ni SACs [23]. However, the residual Ni NPs are prone to be enclosed by NCNTs to form Ni@C core-shell structure.

Moreover, these concomitantly generated graphite encapsulated Ni (Ni@C) would serve as heterogeneous active site to suppress catalytic selectivity [24-27]. A few reports have been tried to remove the graphite encapsulated metal (M@C) NPs from SACs through high-temperature Cl₂ treatment, low-temperature NH₄Cl treatment or leaching combined with ball-milling [28-30]. Nevertheless, these harsh treatment processes inevitably damage the single-atom sites or self-supporting structures. In such case, achieving the simple and scalable preparation of self-supported SACs with only NiN_xC_y moieties active remains grand challenges through top-down synthetic route.

Intrinsically, HER activity of Ni NPs is originated from the proper *H intermediates binding strength induced by its electronic structure [31]. Hence, one can imagine that modulating the electronic structure of Ni@C may make them inactive to competing with HER, and then improve the catalytic selectivity without removing the M@C NPs. Practically, alloying Ni with a 3d transition metal (e.g., Fe, Co, Cu) has been found as an effective way for altering their electronic structure [32-34]. To this end, we firstly sought to find Ni based alloy (NiM) with insufficient HER activity by theoretical calculations. Modeling shows that alloying Ni with Cu would enhance free energy for *H captured by alloy and inhibit HER. Inspired by this, we employed the Ni₉Cu₁ alloy to replace pure Ni NPs as seeds to prepare self-supported Ni₉Cu₁ NPs/single-atom carbon nanofiber membrane catalyst (Ni₉Cu₁@NCNTs/CFM) through electrospinning combined top-down synthetic method. Electrochemical experiment described that Ni₉Cu₁@NCNTs/CFM presented an ultrahigh CO Faradic efficiency (FE) over 97% at −0.7 V vs. reversible hydrogen electrode (RHE). Benefit from the good mechanical properties and catalytic performance, Ni₉Cu₁@NCNTs/CFM was further demonstrated to be an efficient cathode catalyst for a high-performance Zn-CO₂ battery with a power density up to ~0.65 mW/cm².

Density functional theory (DFT) studies were firstly established to investigate two types of Ni based alloy (i.e., NiCu and NiCo). Since the doping metal atoms may also dispersed in NCNTs to generate NiN_xC_y moieties, the content of doping metal atoms was controlled at a low level to avoid forming abundant doping metal single-atom moieties on NCNTs and affecting the selectivity [20, 35, 36]. As shown in Figs. 1a, b and Fig. S1 in Supporting information, the models for the N-doped graphene encapsulated Ni, Ni₉Cu₁ and Ni₉Co₁ (Ni: M = 9:1) sites were constructed. We queried the projected density of states (PDOS) of Ni 3d, Cu 3d and Co 3d to investigate the effect of alloying on their d-band center. As shown in Fig. 1c and Fig. S2 in Supporting information, introducing Cu atoms into Ni NPs regulates the PDOS of obtained sample, inducing a left shift of PDOS of 3d orbitals and a deeper d-band center of Ni₉Cu₁@NC (−2.17 eV) compared to those on Ni@NC (1.21 eV), whereas the d-band center of Ni₉Co₁@NC shift right to −1.06 eV. On the basis of d-band center theory, the binding of the intermediates on Ni₉Cu₁@NC and Ni₉Co₁@NC may be weakened and enhanced, respectively [37]. In light of this, we then simulated the Gibbs free energy of key reaction path ways for HER in various catalysts. As shown in Fig. 1d, the adsorption free energy of hydrogen (ΔG_H^*) in Ni₉Cu₁@NC is calculated to be 1.11 eV, higher than that of Ni@NC (0.96 eV). On the other hand, Ni₉Co₁@NC exhibited the surprisingly lower ΔG_H^* of 0.11 eV for HER. Generally, ΔG_H^* was used to evaluate the electrocatalytic activity of HER. In principle, moderate hydrogen adsorption (ΔG_H^* = 0) would deliver the optimal HER activity [38]. Collectively, the introduction of Cu into Ni NPs may lead to a suppressed HER activity of alloy NPs, and then improve the CO selectivity of Ni₉Cu₁@NCNTs/CFM.

To experimentally realize the concept of DFT predictions, the Ni₉Cu₁@NCNTs/CFM was prepared as shown in Fig. 2a. Initially, a precursor solution containing certain amount of nickel acetoacetate, copper acetoacetate and polyacrylonitrile (PAN) were electrospun to Ni₉Cu₁@PAN pristine nanofibers. Subsequently, the resulting nanofibers were pre-oxidated in air and carbonized with enough melamine on upstream in Ar atmosphere successively. In this process, the NiCu oxide NPs supported on CNF were reduced to NiCu alloy NPs by C and N-contained gas molecules and then served as seeds to catalyze the growth of NCNTs, and form Ni₉Cu₁@NCNTs/CFM eventually. The morphology of the obtained catalysts was characterized by scanning electron microscope (SEM). As shown in Figs. 2b–d, Ni₉Cu₁@NCNTs/CFM comprises self-supporting CNF and densely covered NCNTs with a length of ca. hundreds of microns. This composite structure has good mechanical properties and can be directly used as self-supporting catalyst (Figs. S3 and S4 in Supporting information). Transmission electron microscopy (TEM) image in Fig. 2e shows NCNTs with a diameter about 30–50 nm. A closer high-resolution TEM image in Fig. 2f demonstrate that Ni₉Cu₁ alloy NPs are enclosed by multiple carbon layers in NCNTs, which is because these encapsulated metal NPs are stable and not easily removed by leaching in dilute acid. The measured lattice spacing of the alloy (0.206 nm) is slightly larger than that of Ni NPs, indicating the formation of Ni₉Cu₁ alloy. The corresponding mapping image in Fig. 2h demonstrated the homogenous distribution of Ni, Cu, N, C species across the tube wall of NCNTs without observable metal clusters or NPs, suggesting the existence of atomically dispersed metal atoms. An aberration-corrected high-angle annular dark-field (HAADF) STEM image in Fig. 2g exhibited isolated bright spots on the NCNTs, which confirmed this suggestion as prior literatures conclusively show that these bright spots correspond to isolated Ni atoms due to the positive correlation between brightness and atomic number [39, 40].

As comparisons, we also produced Ni@NCNTs/CFM and Ni₉Co₁@NCNTs/CFM, which were synthesized by changing Ni₉Cu₁ to Ni or Ni₉Co₁ (Figs. S5 and S6 in Supporting information). The above results demonstrated that introducing doping metal salt in the precursor successfully tuned the final composition of obtained encapsulated NPs. As discussed in introduction, this composition control may provide a solution to inactivate the encapsulated Ni NPs to enable the MN_xC_y moieties active only in self-supported Ni NPs/single-atom catalyst.

The crystal structure of the as-prepared catalysts was investigated by X-ray diffraction (XRD) spectra. As shown in Fig. 3a, all samples comprised mixed crystal patterns including Ni NPs and graphite, which further confirmed that all catalysts have similar structures. The chemical compositions and elemental states of catalysts were further analyzed by X-ray photoelectron spectroscopy (XPS). The survey-scan spectrum (Fig. S7 in Supporting information) reveals the existence of Ni, N, C and doping metal atoms in all samples. The undetectable signal of doping metal atoms implies their low concentration (Fig. S8 in Supporting information), and the specific contents were estimated by XPS. As shown in Table S1 in Supporting information, the content of doping metal (~0.2 at%) is much lower than that of Ni (~1.2 at%), which suggest that the effect of doping atoms can be ignored. The binding energies of Ni 2p_3/2 (Fig. S9 in Supporting information) located between metallic Ni⁰ (853.5 eV) and Ni²⁺ (855.8 eV), suggesting a low-valent state of Ni species in Ni₉M₁@NCNTs/CFM catalysts. In addition, the high-resolution XPS N 1s spectra (Fig. 3b) could deconvolute into pyridine N (398.8 eV), Ni/M-N (~399.8 eV), pyrrole (400.5 eV), graphitic (401.3 eV) and oxidized (403.0 eV) N species. From the above XPS results, we infer that the isolated metal atoms in Ni₉M₁@NCNTs/CFM catalysts is bonded with N atoms. To further verify the chemical identity of the Ni and Cu atoms in the catalysts, Ni and Cu K-edge X-ray absorption near edge structure (XANES) spectra were performed. As shown in Fig. 3c, the edge location of Ni₉Cu₁@NCNT/CFM is higher than Ni foil but lower than the Ni²⁺ atoms in NiO, which shows the oxidation of the Ni species, consistent with Ni XPS. Similarly, in the Cu XANES spectrum as shown in Fig. 3e, the edge location situated between Cu foil and CuO references can be attributed to the unsaturated valence state of Cu^δ+. Furthermore, the coordination environment of the Ni or Cu species in Ni₉Cu₁@NCNT/CFM was assessed by extended X-ray absorption fine structure (EXAFS). The peaks centered at 1.35 and 2.14 Å were observed in Fig. 3d, which can be attributed to disperse Ni-N_x sites and encapsulated metal NPs, respectively. Similarly, in the Cu EXAFS spectrum as shown in Fig. 3f, the peak around 1.12 and 2.21 Å have been considered to correspond to Cu-N coordination and Cu-Ni/Cu metal bond, respectively. This observation is also in accordance with the XPS results. Similar to the previous literatures, the accurate local coordination structure of Ni single-atom moieties may not be acquired by further fitting EXAFS data as the fitting process is limited by the strong peak of Ni-Ni [21-23]. However, the above results undoubtedly confirm the existence of single-atom moieties with M-N coordination in the catalysts.

To investigate the effect of alloying on the electrocatalytic activity of Ni@C NPs. Linear sweep voltammetry (LSV) was used to evaluate the CO₂RR and HER activity of all catalysts on a rotating ring-disk electrode. As shown in Fig. S13 (Supporting information), the current density recorded in CO₂-saturated KHCO₃ solution was higher than that in the N₂-saturated electrolyte, suggesting that all samples possess good CO₂RR activity. Note that Ni₉Cu₁@NCNTs/CFM exhibited a lower current density compared with others (Fig. S14 in Supporting information). This decline in total current density is likely the result of suppressed HER activity of Ni₉Cu₁@C NPs.

To verify this, chronoamperometric measurements were carried to assess CO selectivity. CO and H₂ was the only gaseous product and no liquid product can be detected (Fig. S15 in Supporting information). As shown in Fig. 4a, Ni₉Cu₁@NCNTs/CFM exhibited a higher CO FE with respect to the others. Specifically, when the potential was −0.7 V vs. RHE, the CO FE reached up to 97% and can be maintained above 90% in a wide potential window from −0.6 to −0.9 V. As to Ni₉Co₁@NCNTs/CFM, the CO FE declined obviously to ~75% from 85% compared to Ni@NCNTs/CFM. Furthermore, Ni₉Cu₁@NCNTs/CFM showed a similar j_CO but decreased j_H2 compared to others (Fig. 4b and Fig. S14 in Supporting information).

To explore the origin of the suppressed HER activity in Ni₉Cu₁@NCNTs/CFM, the activity of M@C NPs and single-atom moieties was separately evaluated. We firstly compared the current density of various catalysts in N₂-saturated electrolyte as the current density obtained under CO₂ free condition will directly reflect the HER performance of catalysts. As shown in Fig. 4c, unlike Ni₉Co₁@NCNTs/CFM, Ni₉Cu₁@NCNTs/CFM exhibited an inferior HER activity compared to Ni@NCNTs/CFM. NiN_xC_y moieties has been proved to be almost inactive to HER [6, 21], thus the declined HER activity can be attributed to the difference of M@C NPs. Since previous reports demonstrated that forming hetero-diatomic moieties may affect their catalytic activity compared with single-atom moieties counterpart [41, 42], the effect of introducing another single-atom moiety on CO₂RR performance was analyzed by ball milling the catalysts in HCl to remove M@C NPs. As shown in Fig. 4d, the similar Tafel slope observed in catalysts ground with HCl (H-Ni₉M₁@NCNTs/CFM) suggests that introducing doping atoms has a negligible effect on CO₂RR performance of single-atom moieties, which is further confirmed by similar CO FE in H-Ni₉M₁@NCNTs/CFM catalysts (Fig. 4e). This phenomenon was probably caused by the very low content of doping atoms. Collectively, these results jointly confirm that the enhanced CO selectivity in Ni₉Cu₁@NCNTs/CFM originate from suppressed HER in Ni₉Cu₁@C NPs, consistent with the DFT simulations. In addition, Ni₉Cu₁@NCNTs/CFM showed stable current density and almost invariable CO FE during a 20 h operation at −0.8 V, indicating the good stability (Fig. 4f).

In view of the highly efficient CO₂RR performance and self-s-upporting structure of Ni₉Cu₁@NCNTs/CFM, we directly used Ni₉Cu₁@NCNTs/CFM as cathode to assemble a liquid rechargeable Zn-CO₂ battery by using 6.0 mol/L KOH with 0.2 mol/L Zn(Ac)₂ as anodic electrolyte and 0.5 mol/L KHCO₃ cathodic electrolyte. The bipolar membrane was used to maintain the different pH of two compartments [43, 44]. As shown in Fig. 4g, CO₂RR occurs on the cathode and Zn was dissolved into the anode electrolyte in the discharge. During the charge process, oxygen evolution reaction occurred on the cathode, accompanied by Zn deposition on the anode. Fig. S19 in Supporting information shows the discharge and charge polarization curves for Zn-CO₂ battery, confirming the rechargeable feature of the battery. Besides, the peak power density reached ~0.65 mW/cm² at 2.25 mA/cm² (Fig. 4h), indicating that its practical application potential. Furthermore, the Zn-CO₂ battery was continuously recycled at a constant current density of 0.5 mA/cm² with delivered discharge voltage of 0.7 V and charge voltage of 2 V for 150 cycles, demonstrating the excellent durability of the Zn-CO₂ battery (Fig. 4i).

In summary, as predicted by the DFT calculations, incorporating Cu atoms into Ni NPs can tune its d-band center and theoretically suppress the HER activity. In light of this, we employed the Ni₉Cu₁ alloy on CNF as seeds to facilely prepare a self-supported Ni₉Cu₁ NPs/single-atom catalyst through electrospinning combined top-down synthetic method. The optimized Ni₉Cu₁@NCNTs/CFM catalyst presents an ultrahigh CO FE over 97% at −0.7 V. The subsequent comparative experiments revealed that this excellent CO selectivity originate from suppressed HER activity of encapsulated Ni₉Cu₁ NPs. Furthermore, Zn-CO₂ battery based on the Ni₉Cu₁@NCNTs/CFM cathode were devised to show a peak power density of ~0.65 mW/cm², and an excellent long-term operation stability. Our discoveries provide a guidance for facile and scalable preparation of self-supported Ni SACs with unitary single-atom active sites for many potential applications, including Zn-CO₂ battery and beyond.

All authors approved this submission and there are no conflicts to declare. We confirm that this work is original and has not been published, and is not being submitted to any other journals during this submission.

This work was financially supported by the National Natural Science Foundation of China (Nos. 51773226, 61701514) and the Natural Science Foundation of Hunan Province (No. 2018JJ3603).

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2021.10.063