Electrocatalytic reduction of carbon dioxide (CO2RR) into commercial chemicals and high-valued fuels is as an effective strategy to achieve a carbon–neutral society [1,2,3,4,5,6,7,8,9,10]. However, the CO2RR process still faces great challenges due to sluggish reaction process and unfavorable competitive hydrogen evolution reaction, thus requiring highly efficient electrocatalysts to obtain desirable products at low overpotentials [11,12,13,14,15,16,17,18,19,20]. Though precious metals (e.g., Pd, Au and Ag) have been reported to be efficient in catalyzing the CO2-to-CO conversion process, the high cost, low stability and limited reserves hampered their practical application [21,22,23,24,25,26]. Hence, it is of great importance to develop non-precious metals electrocatalysts with low cost, low overpotential and high Faraday efficiency.

Recently, transition metal–nitrogeN-Carbon (M-N-C, M=Ni, Fe and Co) catalysts with atomically dispersed M-Nx sites have demonstrated attractive performance because of their unique electronic structures and near 100% atomic utilization [27,28,9,30,31,32,33]. Among them, Ni-N-C has been reported to display superior selectivity for CO2-to-CO conversion compared to other M-N-C materials [34,35,36,37]. In general, the reported Ni-N-C catalysts were often synthesized by directly pyrolyzing metal salts and NC precursors. For instance, Li et al. prepared Ni-N-C via a simple solid-phase pyrolysis of urea and Ni(NO3)2·6H2O [38]. The Ni-N-C catalysts could only deliver a CO partial current density (jCO) of –6 mA·cm−2 at –0.7 V (vs. RHE). Hou et al. reported single-atom NiN4 active sites supported by a carbon matrix catalyst, prepared by carbonized the mixture of nickel nitrate and ophenylenediamine, delivering a maximal jCO of –3.5 mA·cm−2 at –1.0 V (vs. RHE) [39]. Despite great efforts have been made, these Ni-N sites were often randomly dispersed in the carbon host with very limited exposure, resulting in a low utilization of the catalytic centers [40–42]. Therefore, the distribution of Ni-N-C and the nanostructure of the carbon support should be rationally designed to sufficiently expose the active sites, thus improving the catalytic performance toward CO production. The ordered porous structure with a specific surface area not only facilitates the exposure of the active site to the electrolyte, but also ensures the efficient exchange of protons, electrons, carbon dioxide and the target product. However, it is practically very challenging to synthesize porous structures while controlling the formation of Ni-N active sites.

Herein, we report a three-dimensional (3D) ordered porous nitrogen-doped carbon-supported Ni-Nx as effective electrocatalyst for CO2RR. As illustrated in Scheme 1, the composite catalyst was synthesized by first pyrolyzing a mixture containing polyvinyl pyrrolidone (PVP), nickel-phenanthroline complex and SiO2 templates, followed by the removal of SiO2. The porous structures could be easily regulated by tuning the PVP contents which was beneficial to the formation of the ordered honeycomb-like structure, promoting the mass transfer during the catalytic process. The obtained Ni/N-C-100 (Ni/N-C-X, where X refers to the different amount of PVP added) catalyst achieved a maximal FECO of 91.1% at − 0.8 V with a large jCO of –16.2 mA·cm−2. Furthermore, it exhibited a high FECO over 85% at a broad potential range (− 0.6 to − 0.9 V (vs. RHE)) and maintained high FECO above 98% after 20 h of reaction at − 0.8 V (vs. RHE). Density functional theory (DFT) calculations showed that the Ni-Nx sites could enhance CO2RR reaction dynamic process and suppress hydrogen evolution reaction, giving rise to enhanced activity and selectivity.

Scheme 1
scheme 1

Schematic illustration of synthesis of Ni/N-C-X

Full size image

The morphology of the as-prepared catalysts was first investigated by SEM. Without PVP, a broken frame-like structure was observed for Ni/N-C-0. With the addition of PVP, a honeycomb-like three-dimensional porous structure was formed, signifying that the PVP played a key role in maintaining the integrity of the 3D framework (Figs. 1a–c and S3). Figure 1d shows TEM image of Ni/N-C-100, confirming the interconnected porous structure. The diameter of the pores in Ni/N-C-100 and Ni/N-C-200 was measured to be approximately 180 nm, consistent with the size of silica templates. Moreover, the lattice fringe spacing in the Ni/N-C-100 was 0.34 nm (Fig. S4), corresponding to the (002) plane of graphitized carbon, which indicated high graphitization of the carbon support. Aberration-corrected HAADF-STEM images further illustrated that a large number of isolated bright spots and small patches were identified by red and yellow circles, indicating the presence of both single atoms and clusters in Ni/N-C-100 (Fig. 1e, f). Additionally, the energy-dispersive X-ray spectroscopy (EDX) mapping analysis revealed the homogeneous distribution of C, N, O and Ni in Ni/N-C-100 (Fig. 1g).

Fig. 1
figure 1

SEM images of a Ni/N-C-0, b Ni/N-C-100 and c Ni/N-C-200; d TEM image and e, f HAADF-STEM image of Ni/N-C-100; g EDS mapping images of Ni/N-C-100

Full size image

X-ray diffraction (XRD) patterns exhibited two broad diffraction peaks at approximately 23° and 43° (Fig. 2a), attributed to the (002) and (100) planes of graphitic carbon, respectively, which confirmed that no Ni-based crystalline phases were formed in all the three samples. As shown in the Raman spectra (Fig. 2b), two distinct peaks at about 1340 (D band) and 1580 cm−1 (G band) were observed [27, 36]. Notably, the intensity ratio of D band and G band (ID/IG) of Ni/N-C-X gradually decreased with the increase in PVP contents, indicating that the introduction of PVP could enhance the graphitization degree of the catalysts. The porous structures of the as-prepared samples were further analyzed by N2 adsorption–desorption isotherms. Typical type IV isotherms with some hysteresis were observed for all samples (Fig. 2c), indicating mesoporous structure in these catalysts. The corresponding pore size distributions exhibited a sharp peak centered at 0.5 nm (dV/dD represent pore area), demonstrating the presence of abundant micropore (Fig. 2d). The specific surface area (SBET) of Ni/N-C-0, Ni/N-C-100 and Ni/N-C-200 was measured to be 807.1, 912.9 and 726.7 m2·g−1, respectively, signifying that the PVP played a key role in maintaining the integrity of the 3D framework. Benefiting from the high specific surface area and 3D ordered porous structure, the exposure of Ni-Nx sites and mass transfer could be significantly improved for CO2RR.

Fig. 2
figure 2

a XRD patterns, b Raman spectra, c N2 adsorption–desorption isotherms and d pore size distribution of Ni/N-C-0, Ni/N-C-100, Ni/N-C-200

Full size image

X-ray photoelectron spectra (XPS) were further performed to study the elemental composition and chemical state of the as-prepared catalysts. XPS survey spectra of all the three samples showed the existence of C, O, N and Ni (Fig. S5). The high-resolution C 1s spectra of as-prepared catalysts were divided into C–C/C=C (~ 284.7 eV), C−N (~ 285.5 eV) and C−O (~ 288.5 eV), indicating that the N atoms were successfully doped into the carbon framework (Fig. 3a) [43]. Moreover, the binding energies of Ni 2p in Ni/N-C-0, Ni/N-C-100 and Ni/N-C-200 catalysts were reflected by the peak at 855.2 eV, in between metallic Ni0 (853. eV) and Ni2+ (855.7 eV), suggesting that the Ni atoms in these samples are likely to be at a low-valence state (Fig. 3b) [44]. The N 1s spectra of these samples could be fitted into pyridinic-N (398.5 eV), Ni-Nx (399.2 eV), pyrrolic-N (400.8 eV), graphitic-N (401.7 eV) and oxidized-N (404.1 eV). It was believed that the Ni-Nx bond could promote the proton transfer and facilitate the intermediate protonation during the CO2RR process (Fig. 3c) [3, 39, 45]. Furthermore, the content of each N species is illustrated in Fig. 3d, and the large amount of pyridinic-N in the as-prepared catalysts may favor the activation of CO2 molecules [46,47,48,49]. Furthermore, the presence of Ni-Nx indicates that Ni would directly bind to N rather than forming nanoparticles or clusters, indicating the absence of metallic Ni, which agreed with XRD results.

Fig. 3
figure 3

High-resolution XPS spectra of Ni/N-C-0, Ni/N-C-100, Ni/N-C-200: a C 1s, b Ni 2p, and c N 1s; d atomic content of pyridine–N, Ni-N, pyrrole-N, graphitic-N and oxide-N for three samples

Full size image

The CO2RR electrochemical performance of the three catalysts was first investigated by linear sweep voltammetry (LSV). As depicted in Figs. 4a and S6, all samples showed a higher current response in CO2-saturated 0.5 mol·L−1 KHCO3 electrolyte than in the Ar-saturated counterpart, suggesting that the Ni/N-C catalysts were active for CO2RR. Both Ni/N-C-0 and Ni/N-C-100 showed much larger j than Ni/N-C-200 at all applied potentials. Specifically, the j of Ni/N-C-100 reached as high as 40 mA·cm−2 at –1.0 V (vs. RHE), showing excellent catalytic activity. Next, controlled potential electrolysis at various potentials was carried out in an H-type cell with a three-electrode system to further evaluate the catalytic activity and selectivity in CO2RR. Nuclear magnetic resonance (NMR) spectroscopy and online gas chromatography (GC) were used to analyze the gaseous and liquid products, respectively, and only CO and H2 were detected as the gaseous products (Fig. S7). As depicted in Fig. 4b, the cathodic jtotal of Ni/N-C-100 gradually increased as the potential increased from –0.6 to –1.0 V (vs. RHE) and remained unchanged after 1800 s, reflecting enhanced electrochemical stability at these applied potentials (Fig. S8). The j could reach about 30 mA·cm−2 at –1.0 V (vs. RHE) during the long-term test for 30 min. By comparing the FECO of Ni/N-C-0, Ni/N-C-100, and Ni/N-C-200 catalysts at various potentials, it could be found that the FECO of Ni/N-C-100 remained above 85% ranged from –0.6 to –0.9 V (vs. RHE) (Figs. 4c and S9). Notably, the FECO of Ni/N-C-100 was up to 91.1% at –0.8 V (vs. RHE), while that of Ni/N-C-0 and Ni/N-C-200 was 90.1% and 86.9%, respectively. To get more insight of the active site, the KSCN poisoning experiment was conducted. As shown in Fig. S10, a significant decrease in CO Faradaic efficiency was observed for Ni/N-C-100, confirming the catalytic role of Ni-Nx active site. Moreover, the jCO of Ni/N-C-100 was calculated to be 14.34 mA·cm−2 at –0.8 V (vs. RHE), which was 3.7 and 3.1 times higher than those of Ni/N-C-0 (3.85 mA·cm−2) and Ni/N-C-200 (4.6 mA·cm−2), respectively (Fig. 4d), further confirming the advantage of the hierarchical porous structure.

Fig. 4
figure 4

a LSV curves for Ni/N-C-100 in 0.5 mol·L−1 CO2-saturated KHCO3 electrolyte; b chronoamperometric (i-t) curves of Ni/N-C-100 at different potentials; c FECO at different potentials; d jco for all samples at various potentials; e electrochemical double layer capacitance measurements for Ni/N-C-0, Ni/N-C-100, Ni/N-C-200; f long-term stability of Ni/N-C-100 at –0.8 V (vs. RHE)

Full size image

To further reveal the origin of the superior performance, the electrochemical active surface area (ECSA) of the as-prepared samples was measured. The double-layer capacitance (Cdl) was first determined by the cyclic voltammetry (CV) curves at different scan rates (Figs. 4e and S11), where the value of Cdl was proportional to the ECSA of the electrocatalyst. It is thus apparent that Ni/N-C-100 had a higher ECSA than the other two counterparts, resulting in the highest jCO. Additionally, ECSA-normalized current density of all the three samples is provided in Fig. S12. Notably, Ni/N-C-100 possessed a lower specific CO current density than Ni/N-C-0. The porous structure is beneficial to the exposure of active sites and the penetration of the electrolyte, resulting in a larger current density of Ni/N-C-100 in CO2RR. The long-term electrolysis of Ni/N-C-100 at the potential of –0.8 V (vs. RHE) (Fig. 4f) indicated that 90% of its initial FECO could be retained without any significant change in current density after 20 h of continuous operation, and its 3D ordered porous structure (Fig. S13) and XRD pattern (Fig. S14) demonstrated almost no change before and after the long-term test, reflecting its high robustness.

To gain a deeper understanding of the superior CO2RR electrochemical performance of Ni/N-C-100, DFT calculation was performed. Based on XRD and HAADF-STEM results, the modeling for cluster/Ni-N-C was built by arranging a Ni cluster onto the surface of Ni-N-C. For comparison, a Ni cluster and Ni-N-C model were also constructed (Fig. S15). As shown in Fig. 5a, the first step of CO2 to form COOH* intermediates was the rate-limiting step (RDS) for cluster/Ni-N-C and Ni-N-C, and the free-energy change (ΔG) for this step on cluster/Ni-N-C (0.99 eV) was significantly smaller than that on Ni-N-C (1.62 eV), indicating enhanced reaction dynamic process [27, 29]. Although the Ni cluster exhibited a lower free-energy change for the formation of *COOH intermediates (− 0.38 eV) than cluster/Ni-N-C (1.46 eV), the CO desorption energy on Ni cluster was much higher than that on cluster/Ni-N-C, which indicated that CO molecule would be trapped on the Ni cluster surface, resulting in a sluggish reaction dynamic process [33, 44]. Furthermore, cluster/Ni-N-C possessed a higher H* adsorption energy (0.81 eV) than pure Ni cluster (− 0.56 eV), demonstrating that HER view was well inhibited on cluster/Ni-N-C (Fig. S16). Moreover, the structural evolution of active site on cluster/Ni-N-C is shown in Fig. 5b, visually reflecting that CO2RR catalytic process. In addition, Fig. 5c exhibits the charge density difference of the adsorbed COOH* intermediates on cluster/Ni-N-C, Ni cluster and Ni-N-C. The electron transfer between adsorbed COOH* and cluster/Ni-N-C was much stronger than that with Ni-N-C, reflecting enhanced ability toward COOH* formation for cluster/Ni-N-C.

Fig. 5
figure 5

a Free-energy diagram of CO2 electroreduction to CO over cluster/Ni-N-C and Ni cluster; b structural evolution of active site for cluster/Ni-N-C in CO2RR process; c charge density difference of cluster/Ni-N-C, Ni cluster and Ni-N-C with COOH* adsorption from 3-dimensional

Full size image

In summary, utilizing SiO2 as the hard template and PVP as a structure directing agent, highly ordered honeycomb-like 3D porous carbon framework loaded with atomically dispersed Ni-Nx species was constructed. When used for CO2RR, the as-prepared Ni/N-C-100 catalyst exhibited a high FECO of 91.1% at –0.8 V (vs. RHE) with jCO of –16.2 mA·cm−2, and robust stability after a 20-h long-term electrolysis. DFT calculations have demonstrated that the Ni-Nx active sites were beneficial to the formation of CO key intermediates and suppress hydrogen evolution reaction, giving rise to enhanced activity and selectivity. This work developed a facile method for the preparation of high-performance Ni single-atom electrocatalysts, which can also be applied to a widespread range of electrochemical applications.