Preparation and characterization of Ni–N/C. The Ni–N/C catalyst was prepared by the thermal conversion of zeolitic imidazolate framework (ZIF)-8, with some modifications to previous work.34 ZIF-8 was synthesized through the self-assembly of Zn2+ and imidazolate with a particle size of ~100 nm (Supplementary Fig. 1); Zn was replaced with Ni by ion-exchange. The resulting light-green powder was then carbonized at 1000 °C to produce a final product, Ni–N/C. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of Ni–N/C displayed rather contracted morphologies compared to ZIF-8 after high-temperature treatment (Supplementary Fig. 2).35 From high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of Ni–N/C (Fig. 1b and Supplementary Figs. 2c and d), bright dots can be observed on the support, indicating the generation of atomically dispersed Ni sites embedded on the N-doped carbon matrix without any agglomerated metal particles (Supplementary Fig. 3). Extended X-ray absorption fine structure (EXAFS) analysis confirms the structural information seen in the microscopy images (Fig. 1c). The Ni K-edge k3-weighted extended EXAFS spectrum of Ni foil shows a major scattering peak at ~2.0 Å at reduced distances, which describes the Ni–Ni bond of metallic Ni. The EXAFS spectrum of Ni–N/C exhibits a Ni–N scattering peak at ~1.5 Å at reduced distances without the presence of an Ni–Ni scattering peak, from which the isolation of Ni atoms bound to the N-doped carbon support can be concluded.
The Ni 2p X-ray photoelectron spectroscopy (XPS) of Ni–N/C shows a peak around 855 eV (Fig. 1d), corresponding to Ni(II) phthalocyanine (NiPc), and indicates the ionic character of Ni in Ni–N/C. X-ray absorption near edge structure (XANES) spectra of Ni–N/C and NiPc also show similar edge positions (Fig. 1e), confirming the similar oxidation state (Ni2+). In addition, the pre- and post-edge peaks in XANES spectra display distorted Ni–N/C coordination environments. The peak labeled A is usually attributed to square-planar structures such as NiPc, absent in Ni–N/C.32,36 The higher intensity of peak B compared to that of peak C in the XANES spectrum of Ni–N/C can be explained by the distortion of D4h symmetry, owing to the displacement of the metal centre,37 and is potentially induced by high-temperature pyrolysis.
Electrochemical cCO2RR reactivity of catalysts. The catalytic properties of Ni–N/C and cAg for the cCO2RR were investigated in a CO2-absorbed 5 M MEA aqueous solution. The electrocatalytic conversion was performed in an H-cell after purging with Ar for 10 min to remove unabsorbed CO2 in the MEA solution. Ni–N/C was found to exhibit superior performance compared to cAg both in the jCO and product selectivity. Ni–N/C has a maximum jCO of −3.5 mA cm−2, three times larger than cAg (−0.9 mA cm−2, Fig. 1f). Furthermore, it shows a maximum CO F. E. (63.2%) of approximately double that of cAg (30.1%) with suppressed HER selectivity (Fig. 1g and Supplementary Fig. 4). Notably, the CO F. E. of Ni–N/C is the highest value ever reported for the cCO2RR in CO2-adsorbed pure MEA electrolyte without CO2 flow (Supplementary Table 1). In addition, the on-set potential of Ni–N/C (−0.45 V vs. RHE) is lower than that of cAg (−0.60 V vs. RHE), which demonstrates the excellent catalytic activity of Ni–N/C for the cCO2RR to CO.
To examine the effect of metal centers on SACs, a number of different metal-based SACs were prepared. Mn-, Fe-, Co-, and Cu-based SACs can be obtained by controlling metal precursors during the ion-exchange step (Supplementary Fig. 5). In contrast with Ni, other metal SACs show almost zero jCO with high H2 selectivity in the cCO2RR (Supplementary Fig. 6). On N-doped carbon (N/C) prepared without ion-exchange, the HER only takes place in high overpotential regions. These results demonstrate that the Ni center plays a key role in the selective cCO2RR to CO. The changes in the oxidation state and structure of Ni–N/C during the cCO2RR were further investigated through an in situ XANES analysis. From the in situ XANES spectra, significant changes in the structure of Ni–N/C during the cCO2RR were not observed (Supplementary Fig. 7), illustrating the catalytic importance of the distorted coordination structure of single atom Ni catalyst, which allows for excellent catalytic performance in the cCO2RR without electrolyte-modulating additives.
Practical applicability test using a zero-gap membrane electrode assembly. To evaluate the practical applicability of the cCO2RR in high current density, an electrochemical characterization was additionally performed with a zero-gap membrane electrode assembly for the cCO2RR in CO2-captured 5 M MEA. BPM was used to drive the regeneration reaction of carbamate (RNHCOO− + H+ → RNH2 + CO2), to increase the local concentration of CO2 near the cathode and thus enhance conversion activity (Fig. 2a). At −50 mA cm−2, the CO F. E. of Ni–N/C was 64.9% (Fig. 2b and Supplementary Fig. 8), which is approximately 2.5 times higher than that of cAg (25.5%), showing a similar trend to H-cell experiments and demonstrating the superior catalytic performance of Ni–N/C for the cCO2RR in high current density. When an anion exchange membrane (AEM), the alkaline properties of which favor gas-fed CO2 reduction reaction, is used, significantly lowered cCO2RR selectivity has been observed for both Ni–N/C and cAg (Fig. 2b and Supplementary Fig. 8). This could be because AEM is inferior to proton transfer toward the cathode side. From these results, we propose that building a high concentration of liberated CO2 is essential for the cCO2RR at high current density of membrane electrode assembly, consistent with previous reports in bicarbonate solutions.18 The activity difference in the usage of BPM with AEM provides information on the initial stage of the reaction pathway of cCO2RR, which will be discussed in the following section.
The cCO2RR was then performed in 5 M MEA using Ni–N/C in the membrane electrode assembly with chronopotentiometry at −50 mA cm−2 for 10 h. Since the reaction was carried out under Ar, the chemical equilibrium of adsorption/desorption of CO2 in MEA (Equation (2)) continuously moves to favor desorption of CO2. Hence, the number of carbamate anions and the desorption rate of CO2 were gradually decreased over time, which caused a progressive drop in the CO F. E. after 10 h (Supplementary Fig. 9). Nevertheless, after the MEA solution was refreshed by CO2 adsorption, the CO F. E. was recovered to ~50% under Ar conditions. This result demonstrates stable Ni–N/C catalyst operation for the cCO2RR, and the reduction in CO selectivity is caused by the diminished amount of adsorbed CO2 in the capturing medium. An important point to note here is the stability of the capturing media. To identify electrochemical degradation or modification of MEA, analysis of the MEA solution before and after the stability test was carried out by 1H nuclear magnetic resonance (NMR) spectroscopy (Supplementary Fig. 10).38 No additional 1H NMR peaks were observed after the stability test, indicating the stable operation of MEA during the cCO2RR without electrochemical modifications. This result demonstrates that the cCO2RR is a promising catalytic system that could even resolve the degradation problems of capturing media in conventional CCUS processes.
Reaction pathway of the cCO2RR. Next, we designed experiments to understand the pathways of the cCO2RR and investigate the effects of different catalysts. First, two scenarios can be considered at the initial stage: i) the active site of the catalyst directly attacks the carbamate anion, or ii) CO2 is released from the MEA followed by electrocatalytic conversion by similar pathways to conventional CO2RR (Fig. 3a). A combination of these two scenarios is also possible. Sargent et al. argued for the direct reduction of the MEA–CO2 adduct (carbamate) through the detection of the ethanolammonium cation on the electrode surface using in situ surface-enhanced Raman spectroscopy,26 whereas Goetheer et al. considered the liberated CO2 to be the reactant in temperature-controlled experiments.27 In conventional CO2RR-to-CO, the reaction is initiated by electron injection into CO2, resulting in the unstable CO2− intermediate, which is typically regarded as the rate-determining step (RDS).39,40 The stabilization of this intermediate on the catalyst surface is the key to facilitating the CO2RR. In the cCO2RR, if the direct reduction of carbamate takes place in the initial stage of the reaction as in scenario (i) and the electron transfer step to carbamate is the RDS of the reaction, the CO production reaction rate with respect to carbamate concentration could help identify the initial stage. Thus, in CO2-absorbed 1–5 M MEA solutions, the electrochemical conversion rate of captured CO2 was measured with both Ni–N/C and cAg. The carbamate concentration of the respective MEA solution was determined by quantitative analysis of NMR data (Supplementary Fig. 11) using internal standards. The jCO for both catalysts is not significantly affected by the concentration of carbamate (Supplementary Fig. 12). Hence, the reaction is considered zeroth-order with respect to carbamate (Fig. 3b), i.e., the reaction rate is independent of the concentration of carbamate for both catalysts. From these results, we speculated that the reactant was in fact the released CO2 from carbamate rather than the carbamate itself.
To regulate the amount of CO2 released in the H-cell, the reaction temperature was controlled. Since the regeneration reaction is endothermic,41 the rate of the reaction can be accelerated by increasing the reaction temperature. As the reaction temperature is increased from 5 °C to 40 °C, the limiting current density (jlim.) increases for both catalysts (Ni–N/C: −0.52 mA cm−2 (5 °C) → −1.85 mA cm−2 (RT) → −3.5 mA cm−2 (40 °C); cAg: 0 (5 °C) → −0.39 mA cm−2 (RT) → −0.91 mA cm−2 (40 °C), Figs. 3c and d). If the direct reduction of carbamate occurs in the cCO2RR, the mass transport limit in such low current density is difficult to detect. Furthermore, the amount of carbamate in the electrolyte is reduced as the reaction temperature is increased, and thus even if the jlim. originates from the mass diffusion of carbamate, the jlim. has to be decreased. On the other hand, the increased amount of liberated CO2 can improve jlim., solidifying the proposition of liberated CO2 being the reactant. The sharp increase in reaction rate by the BPM being used in the membrane electrode assembly experiment is consistent with this scenario. It should be noted that the decrease in jlim. at 60 °C may originate from the drop in CO selectivity due to H2 being rapidly generated at high temperatures (Supplementary Fig. 13). Therefore, for the effective electrochemical conversion of captured CO2, the optimal temperature should be found, which involves the trade-off between the amount of liberated CO2 and the reaction rate of HER.
From the experimental results discussed, it can be concluded that the superior performance of Ni–N/C cannot be attributed to the different initial reactions of cAg and Ni–N/C for the cCO2RR. The cCO2RR pathway appears to be comparable to the conventional CO2RR, but with a reduced concentration of CO2 in the cCO2RR. Thus, the high performance of Ni–N/C must originate from the high tolerance of low CO2 concentrations in the CO2RR because of its superior CO2RR intrinsic activity and high activation energy for the HER, as we previously reported with low-concentration CO2 gas experiments.33
Universal reactivity of Ni–N/C for the cCO2RR. Subsequently, we investigate how Ni–N/C exhibits high reactivity in the presence of bulky ammonium cations for the cCO2RR. In order to stabilize the CO2− intermediate in the CO2RR, cations with high surface charge density on the cathode are essential.29 Thus, a smaller effective size of alkali metal cation, which creates a higher surface charge density on the cathode, is beneficial to the performance of the CO2RR.42–47 However, amine-based capturing media inevitably generate bulky ammonium cations in the adsorption of CO2, which hampers the stabilization of the CO2− intermediate. However, the effect of bulky cations in the cCO2RR has not yet been investigated. To understand the effects of bulky cations, we carried out the cCO2RR in systems wherein the bulkiness of the capturing molecule can be controlled. The electrocatalytic performances were examined in six different CO2 capturing media (1 M of potassium bicarbonate, MEA, 3-amino-1-propanol, 2-(methylamino)ethanol, 2-amino-2-methyl-1-propanol, and diethanolamine), (Figs. 4a–4f and Supplementary Fig. 14). The effect of the properties of absorbents on activity trends is also discussed below, with the aim of gaining insights into the activity for the cCO2RR (Figs. 4g and h).
Both jCO and CO F. E. were found to decrease with increasing bulkiness of capturing media for both Ni–N/C and cAg. One noticeable difference between Ni–N/C and cAg was activity sensitivity depending on cation type. The bulkiness of the cation has a significant impact on cAg, showing a dramatic loss in jCO as the effective size of the cation increases. The jCO of cAg fell as low as zero in diethanolamine, a secondary amine (Fig. 4h). In contrast, the jCO of Ni–N/C is less affected by the effective cation size (Fig. 4g), and the CO F. E. of Ni–N/C was maintained at ~ 50%, irrespective of the electrolyte type (Figs. 4a–4f). From these results, we hypothesized that the marginal cation effect is manifested on Ni–N/C, and the weak cation sensitivity allows the selective cCO2RR in universal absorbent media. We note, however, that the different types of amine electrolytes not only change the size of the cation but also alter the amount of desorbed CO2 and viscosity of the solvent, which can render the decoupling of the cation effect difficult.
Weak cation sensitivity of Ni–N/C for the CO2RR. To exclusively identify the difference in cation sensitivity of catalysts, we measured the activity and selectivity of the CO2RR in CO2-saturated 0.05 M carbonate electrolytes containing different alkali metal cations (Li+, Na+, K+, and Cs+). Similar to the amine solutions, the jCO of Ni–N/C and cAg decreased as the effective cation radius increased from Cs+ to Li+ (Figs. 5a and b), but the cation sensitivity was clearly different in Ni–N/C compared to cAg. The maximum CO F. E. of Ni–N/C (~100 %) was maintained irrespective of the electrolyte used, whereas that of cAg was reduced to ~85 % in 0.05 M Na2CO3 and ~75 % in 0.05 M Li2CO3 (Figs. 5c and d). The normalized jCO by jCO in the presence of Cs+ is higher in Ni–N/C than that in cAg, and the slope of the normalized value with respect to the effective cation radius of Ni–N/C is less steep than that of cAg (Fig. 5e). These results demonstrate that the cation sensitivity can be varied depending on the type of catalyst and that Ni–N/C shows weaker cation sensitivity compared with cAg.
The accumulated cation density at the EDL can be determined by the size of the cation, the applied potential, and the PZC of the electrode.48 The weak cation sensitivity of Ni–N/C may arise from its relatively high PZC. In the presence of any cation, Ni–N/C exhibits a ~0.6 V higher PZC compared to cAg (Supplementary Fig. 15). This positive-shifted PZC can create an increased surface charge density for the same applied potential in the reduction reaction, which in turn may be prone to stabilizing the CO2− intermediate, thus mitigating the cation effect. The positive-shifted PZC of Ni–N/C, therefore, allows for the effective accumulation of cations to stabilize the CO2− intermediate, even in the presence of large cations such as Li+ or ammonium, without additives. The marginal cation effect on Ni–N/C permits the selective cCO2RR in universal capturing media.