Introduction

Heterogeneous catalysis holds great promise for practical applications in both the petrochemical and fine chemical industries, primarily due to its ease of catalyst separation and recovery1,2. However, it remains challenging when two or multiple reactants coexist in the system. To achieve satisfactory catalytic performance, dual-/multi- active sites catalysts have attracted great attention recently3,4. These catalysts possess multiple active sites that endow them with improved catalytic properties, enabling the activation of two or more reactants, which provide more efficient reaction pathway and accelerate the entire reaction, leading to the significantly enhanced activity and selectivity for the target products.

As one of the representative reactions that involves two reactants (CO2 and H2O) and multiple protonation steps, the electrocatalytic CO2 reduction reaction (CO2RR) to chemicals and fuels offers a promising route to store renewable energy and manage the global carbon balance5,6,7,8. Substantial efforts have been dedicated to developing dual-active sites catalysts to enhance CO2RR performance. Many reported catalysts consist of active sites responsible for CO2 activation (e.g., Au, Ag and Zn) and further hydrogenation or C-C coupling (Cu). It has been demonstrated that the intermediate migration (e.g., *CO) plays a crucial role on the selectivity of the reaction, which greatly influenced by the optimal distribution and distance between the dual active sites. Meanwhile, an increase in the *CO coverage over Cu sites has been shown to promote the formation of C2+ products9,10,11. In addition, considering H2O molecule serves as the proton source for hydrogenation step, the active site for H2O dissociation has been designed to be part of the dual-active sites catalysts12,13. The introduction of H2O dissociation sites, such as metal single atoms, sulfur, and oxygen vacancies, has been demonstrated to accelerate the rate of H2O dissociation into *H species, which are subsequently fed to CO2 conversion sites through *H spillover14,15,16. However, the existing findings have mainly focused on the influence of H2O dissociation in CO2-to-C1 product process (formate or methane), while it is of greater significance for the C2 products formation process. This is due to the commonly used alkaline electrolyte in CO2-to-C2 product, which slows down the H2O dissociation step, resulting in a sluggish reaction kinetic process15,17. Therefore, the constructing dual-active sites electrocatalyst to accelerate CO2 reduction and H2O dissociation respectively is a feasible strategy for achieving the desired electrochemical CO2-to-C2 products performance.

Cu is the most promising electrocatalyst for converting CO2 into C2 products, owing to its moderate adsorption capacity for key intermediates18,19,20. Various strategies have been investigated to improve the C2 selectivity and current density, such as the manipulation of crystal facets, morphology, particle size, and oxidation state21,22,23. However, most of these studies mainly focus on regulating the Cu structure for CO2 activation and C-C coupling, with limited reports on the impact of surface *H coverage through accelerated H2O dissociation. It is crucial to carefully control the coverage of *H species since an excessive amount promotes H2 production through the competing hydrogen evolution reaction (HER). Recently, single-atom catalysts, characterized by the uniform dispersion of isolated metal atoms on a substrate, have received significant attention in the field of HER due to their high metal utilization, tunable electronic structures and structural stability. Notably, the HER activity of single-atom Cu catalyst in alkaline electrolyte can be regulated through modifying coordinated environment and support type24,25,26. Therefore, designing of a dual-active sites catalyst incorporating co-loaded Cu nanoparticles (NPs) and single-atom Cu sites offers the potential for simultaneous CO2 conversion and H2O activation, favoring the high-performance for electroproduction of C2 products.

Herein, we have constructed a dual-active sites catalyst containing moderate content ratio (0.25) of atomic Cu sites to Cu NPs (M-Cu1/CuNP) for CO2 electroreduction to C2 products. The M-Cu1/CuNP exhibited a high C2 products Faradaic efficiency (FEC2) of 75.4% with corresponding a partial current density of C2 products (jC2) of 289.2 mA cm−2 at −0.6 V versus reversible hydrogen electrode (vs. RHE, all potentials are referenced to RHE), Moreover, the stable CO2-to-C2 products conversion with FEs of >70% can maintain at a constant current density of 400 mA cm−2 for a run time of 40 h. The experiments and density functional theory (DFT) calculations revealed that Cu NPs facilitated CO2 activation, *CO hydrogenation into *CHO, and C-C coupling into C2 products. The atomic Cu site accelerated H2O dissociation to provide *H, and the generated *H transferred to Cu NPs through N-doped carbon matrix and modulated the *H coverage on Cu NPs, which reduced the energy barrier for *CO to *CHO.

Results

Synthesis and characterization of M-Cu1/CuNP catalyst

The M-Cu1/CuNP catalyst was synthesized through a facile pyrolysis strategy by calcining Cu complexes in Ar flow and then reducing in H2 flow. The Cu complexes were obtained by mixing Cu(NO3)2·3H2O and guanidine thiocyanate with a 1:8 mole ratio in deionized water and subsequently evaporating the solvent. By varying the amount of Cu(NO3)2·3H2O, the catalysts containing the rich (R-Cu1/CuNP) and poor (P-Cu1/CuNP) amount of Cu1 site were also prepared using the same procedure. The morphology of the as-prepared catalyst was characterized by transmission electron microscopy (TEM). The dark-field TEM image (Fig. 1a) shows that obvious nanoparticles with the average size of ~4 nm are uniformly dispersed on carbon matrix in M-Cu1/CuNP. The corresponding high-resolution (HR) TEM image of the nanoparticle displays the distinct lattice fringes with an interplanar space of 0.208 nm, consistent with the (111) plane of metallic Cu (Fig. 1b), confirming the uniformly dispersed nanoparticles are Cu NPs. The dispersion of atomic Cu was then identified by the aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). As presented in Fig. 1c, the isolated bright dots can be recognized as well distributed Cu atoms in M-Cu1/CuNP, which have been highlighted by red circles. The energy-dispersive X-ray spectroscopy (EDS) elemental mapping images (Fig. 1d) reveal a uniform distribution of C, N and Cu, indicating that they are homogeneous over the entire architectures. Therefore, we can conclude that a dual-active sites catalyst containing atomically dispersed Cu sites adjacent to uniformly dispersed Cu NPs was successfully prepared. In addition, the TEM and aberration-corrected HAADF-STEM images of P-Cu1/CuNP and R-Cu1/CuNP are shown in Fig. S1 and Fig. S2, respectively. The TEM images showed that both P-Cu1/CuNP and R-Cu1/CuNP exhibited a similar average size of Cu NPs compared to M-Cu1/CuNP, while the mount of Cu NPs obviously decreased over R-Cu1/CuNP. Meanwhile, evident single atomic Cu can also be observed on HAADF-STEM images of P-Cu1/CuNP and R-Cu1/CuNP. The X-ray diffraction (XRD) patterns of P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP in Fig. 1e show that only the characteristic peaks of metallic Cu and carbon were observed, and the strong peak at 43.2° can be assigned to Cu(111) facet, which is consistent with the HRTEM results. The X-ray photoelectron spectroscopy (XPS) survey spectra show that the presence of Cu, N, C and O elements in P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP (Fig. S3), where the existence of O element can be attributed to the adsorbed oxygen on surface. The Cu content in P-Cu1/CuNP, M-Cu1/CuNP, and R-Cu1/CuNP was determined to be 7.4, 4.3, and 1.9 at%, respectively, and the corresponding N and C content are displayed in Table S1. The Cu 2p spectra (Fig. 1f) suggest that only Cu 2p3/2 peak attributed to Cu0/+ existed in the spectra, which moved to lower binding energy in the order of R-Cu1/CuNP (933.0 eV), M-Cu1/CuNP (932.7 eV) and P-Cu1/CuNP (932.5 eV), indicating a gradual increase in the valence state of Cu in the catalysts. The Cu LMM Auger spectra confirmed the coexistence of Cu+ and Cu0 in P-Cu1/CuNP M-Cu1/CuNP and R-Cu1/CuNP (Fig. S4). Considering that the valance state of Cu in Cu NP is 0, while that of atomically dispersed Cu site is Cuδ+, the content of Cu1 site in the catalyst changed in order of P-Cu1/CuNP < M-Cu1/CuNP < R-Cu1/CuNP. Meanwhile, the high resolution XPS N 1 s spectra of P-Cu1/CuNP, M-Cu1/CuNP, and R-Cu1/CuNP are displayed in Fig. S5. All spectra showed obvious Cu-N peak at around 399.2 eV, which suggests that the Cu1 sites were coordinated by N atoms and other peaks could be attributed to pyridinic N (~398.5 eV), pyrrolic N (~399.9 eV), and graphitic (~401.0 eV), respectively.

Fig. 1: Morphology and structure of the catalysts.
figure 1

a DF-TEM image (inset shows particle size distribution), b HRTEM image, c aberration-corrected HAADF-STEM image, and d EDS element mappings of M-Cu1/CuNP. e XRD patterns, f XPS spectra of Cu 2p orbits, g Cu K-edge XANES spectra, and h FT-EXAFS spectra of P-Cu1/CuNP, M-Cu1/CuNP and, R-Cu1/CuNP.

Full size image

To further determine the chemical state and local coordination environment, X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) measurements were conducted. The Cu K-edge XANES of R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP with the reference samples Cu foil and Cu phthalocyanine (Cu Pc) were presented in Fig. 1g, the adsorption edge of all catalysts located between that of Cu foil and Cu Pc, indicating the average Cu valence state of Cu in the catalysts laid between 0 and +2. Meanwhile, the adsorption edge shifted to higher energy in the sequence of P-Cu1/CuNP < M-Cu1/CuNP < R-Cu1/CuNP, confirming that the average Cu valence state increased with the increase in the Cu1/CuNP ratio. It is in accordance with the XPS results. The Fourier-transformed (FT) EXAFS spectra in the R-space of the three catalysts were processed and displayed in Fig. 1h, all three catalysts exhibited peaks at around 1.5 Å and 2.2 Å, attributed to Cu-N and Cu-Cu coordination, respectively. Therefore, both atomic Cu and metallic Cu NP coexisted in R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP. Notably, the intensity ratio of the Cu-N peak to the Cu-Cu peak increased in the order of P-Cu1/CuNP < M-Cu1/CuNP < R-Cu1/CuNP, indicating a gradual increase in the content ratio of atomic Cu to Cu NP. In order to obtain the local structure of Cu species and the content ratio of atomic Cu to Cu NP, the quantitative analysis by the least-squares EXAFS fittings was performed (Fig. S6). The percentages (P) of Cu-N4 (Cu1) and Cu NPs in total Cu species of catalysts were displayed in Table S2. Thus, the content ratio of Cu1 to CuNP of P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP were 0.05, 0.25 and 0.39, respectively, which is consistent with our catalyst design expectation.

Electrocatalytic CO2RR performance

The CO2RR performance of the as-prepared catalysts were first evaluated by linear sweep voltammetry (LSV) in 5 M KOH aqueous solution in a flow cell (Fig. S7). As shown in Fig. S8, all the catalysts exhibited significant increase in cathodic current density when the feed gas was change from N2 to CO2, indicating that CO2RR occurred over R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP. Meanwhile, M-Cu1/CuNP exhibited the most positive onset potential and highest cathodic current density in the presence of CO2 gas, and taking −0.8 V as an example, the current density reached up to 1093.0 mA cm−2 over M-Cu1/CuNP, which was roughly 2.5 and 1.7 times higher than that of R-Cu1/CuNP and P-Cu1/CuNP, respectively. Thus, the results of LSV experiments suggested that M-Cu1/CuNP has higher CO2 activity than R-Cu1/CuNP and P-Cu1/CuNP, which could be attributed to the moderate content ratio of Cu1 to CuNP.

The control potential electrolysis was then performed to analyze the reduction products and the catalyst loading was 1 mg cm−2 (Fig. S9). The gas-phase and liquid-phase products were analyzed by gas chromatography and 1H nuclear magnetic resonance spectroscopy, respectively. H2, CO, CH4, formate, C2H4, ethanol and acetate were formed. Figure 2a shows that the FE of C2 products (FEC2) of M-Cu1/CuNP exhibited a volcano-shaped dependence on the applied potential (Fig. S10, Tables S3–S5), and a maximum FEC2 could reach 75.4% at −0.6 V, which is much higher than that over R-Cu1/CuNP and P-Cu1/CuNP. The maximum FEC2 of P-Cu1/CuNP was 47.3% at −0.7 V, which closed to the performance of Cu nanoparticles reported in the literature22,27. Moreover, the M-Cu1/CuNP had a lower onset potential for C2 products formation. The FEC2 could reach 20.5% at −0.4 V over M-Cu1/CuNP, while C2 products cannot be detected over R-Cu1/CuNP and P-Cu1/CuNP under the same potential. The ratio of FEC2 to FEC1 could keep >4.5 from −0.6 to −0.9 V over M-Cu1/CuNP, while those of R-Cu1/CuNP and P-Cu1/CuNP were below 2 in the whole applied potential range (Fig. 2b). Furthermore, M-Cu1/CuNP was treated by sulfuric acid solution to completely remove Cu NPs, resulting in a catalyst denoted as Cu-N-C. The results of TEM, XRD, XPS and XAS indicated that only Cu single atoms existed in the Cu-N-C catalyst (Figs. S11, S12). H2 was the dominant product over Cu-N-C in the whole applied potentials range, suggesting that the atomic Cu sites mainly facilitated H2O dissociation (Fig. 2c and Table S6). The formation of CO and CH4 was also observed, in agreement with previous reports that Cu single atoms could generate H2, CO, and CH425,28,29. These results suggest a dual-active sites effect between Cu sites and Cu NPs, and the proper ratio of Cu1 to CuNP would obviously enhance the selectivity for C2 products.

Fig. 2: CO2 electroreduction performance.
figure 2

a The FEs of C1 products, C2 products, H2, and b C2 and C1 product selectivity ratio at different potentials over the as-prepared catalysts in 5 M KOH electrolyte. c The product FEs and total current density of Cu-N-C at different potentials in 5 M KOH electrolyte. d The C2 partial current density at different potentials over the as-prepared catalysts in 5 M KOH electrolyte. Values are means and error bars indicate s.d. (n = 3 replicates). e Long-term stability of M-Cu1/CuNP at a constant current density of 400 mA cm−2 in 5 M KOH electrolyte (the electrode was washed, then dried and the electrolyte was refreshed at intervals 5 h to address the issues of flooding and carbonation).

Full size image

From FEC2 and total current density (Fig. S13), the C2 products partial current density (jC2) of catalysts at different potentials was obtained and shown in Fig. 2d. It could reach 289.2 mA cm−2 over M-Cu1/CuNP at −0.6 V vs. RHE, exceed by up to 13.1-fold and 4.2-fold than that on R-Cu1/CuNP (22.1 mA cm−2) and P-Cu1/CuNP (69.3 mA cm−2), respectively. Besides, a high total current density of 1207.6 mA cm−2 was obtained at −0.9 V over M-Cu1/CuNP, delivering a FEC2 of 47.1% with a partial current density of 568.5 mA cm−2. We further normalized the jC2 on the basis of the electrochemical active surface area (ECSA), which was measured by the double-layer capacitance (Cdl) method (Fig. S14). Even after normalization, M-Cu1/CuNP still exhibited the highest normalized jC2 among the catalysts (Fig. S15), confirming that the intrinsic catalytic activity of M-Cu1/CuNP is higher than that of P-Cu1/CuNP and R-Cu1/CuNP. Moreover, we also increased the loading of P-Cu1/CuNP and R-Cu1/CuNP in accordance with the ECSA results, ensuring the catalytic activity were measured under similar ECSA condition. The results suggested that M-Cu1/CuNP maintained the largest C2 products selectivity and partial current density (Fig. S16), further confirming that the intrinsic catalytic activity of M-Cu1/CuNP exceeds that of both P-Cu1/CuNP and R-Cu1/CuNP. Compared with reported electrocatalysts for CO2RR to C2+ products in the literature, M-Cu1/CuNP exhibited higher CO2RR to C2+ products activity at low applied potential, especially in term of C2+ partial current density (Table S7).

The long-term stability test was conducted through chronopotentiometry for 40 h and the electrode was washed, then dried and the electrolyte was refreshed at intervals 5 h to address the issues of flooding and carbonation. Fig. 2e and Fig. S17 showed that no obvious change was observed over the potential and products selectivity, the high FE of C2 products could be hold over 70% on M-Cu1/CuNP during the electrolysis. The morphology, valence states and crystal structure of M-Cu1/CuNP after 40 h CO2RR experiment were characterized (Figs. S18, S19). TEM image suggested the size Cu NPs is similar to that before used and the aberration-corrected HAADF-STEM image confirmed that Cu1 site and Cu NPs still coexisted in M-Cu1/CuNP. The Cu LMM Auger spectra show that Cu+ and Cu0 coexisted in M-Cu1/CuNP. Meanwhile, no obvious difference was observed in N 1 s XPS spectra between before and after CO2RR experiment, and only the peaks corresponding to Cu were observed in XRD spectrum after CO2RR experiment. All these results indicated the excellent stability of the catalyst.

The kinetic isotope effect (KIE) of H2O/D2O (H/D) experiments were performed to get insights into the role of H2O dissociation in CO2-to-C2 products (Fig. 3a). When H2O was replaced by D2O in the electrolyte, the formation rate of the product (e.g., C2H4) decreased over P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP, and the level of decrease closely related to the content ratio of Cu1/CuNP. If the KIE value (defined as the ratio of C2H4 formation rates in H2O and D2O) closes to 1, H2O dissociation is not the rate-determining step over the catalyst. The KIE values for R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP were 1.12, 1.21 and 2.87, respectively, which indicated that H2O dissociation was accelerated gradually with an increasing Cu1/CuNP content ratio. These results confirmed that the atomic Cu sites were responsible for accelerating H2O dissociation and provided proton to adjacent Cu NPs, thus affecting CO2-to-C2 products. Meanwhile, the N-doped carbon matrix catalysts has been reported to favor the migration of proton30. In addition, the H2 formation rate in 5 M KOH H2O solution and D2O solution was presented in Fig. S20. An obvious decrease was observed over all the catalysts, while the decreasing degree followed the sequence of P-Cu1/CuNP > M-Cu1/CuNP > R-Cu1/CuNP, suggesting that the increase of Cu single atom content can reduce the influence of isotope effect on H2 formation rate.

Fig. 3: The mechanism analysis.
figure 3

a The kinetic isotope effect (KIE) of H2O/D2O on P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP at −0.6 V in flow cell with 5 M KOH electrolyte. b The electrochemical CO stripping voltammetry tests of P-Cu1/CuNP, M-Cu1/CuNP and R-Cu1/CuNP in 0.1 M Na2SO4 electrolyte. c In situ FT EXAFS spectra at Cu K-edge over M-Cu1/CuNP at various potentials in 5 M KOH electrolyte. In situ surface-enhanced Raman spectra recorded at different applied potentials for (d) P-Cu1/CuNP, (e) M-Cu1/CuNP and (f) R-Cu1/CuNP during CO2RR in 5 M KOH electrolyte.

Full size image

Furthermore, we studied the influence of electrolyte pH on CO2RR performance over R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP. Three different concentrations of KOH aqueous electrolytes, i.e., 0.1 M, 3 M and 5 M, were employed to adjust the pH environment at the electrode/electrolyte interface. The results in Fig. S21 show that the FE and formation rate of C2 products increased with the increasing pH value of electrolyte, suggesting that strong basic local environment favored the C2 products formation. Although CO2RR activity of M-Cu1/CuNP in low concentration KOH electrolytes was lower than that in 5 M KOH electrolyte, M-Cu1/CuNP exhibited higher FE and formation rate of C2 products than P-Cu1/CuNP and R-Cu1/CuNP in low concentration KOH electrolytes, which suggested that the cooperative effect of Cu NPs and atomic Cu sites still facilitated C2 products formation. The ratio of C2 products formation rate over M-Cu1/CuNP to P-Cu1/CuNP, i.e., RateM/RateP, in 5 M KOH (3.4) is higher than those in 3 M KOH (3.0) and 0.1 M KOH (2.3). This demonstrated that the role of atomic Cu sites in accelerating the H2O dissociation process was significantly more pronounced at higher pH values, leading to the enhanced C2 products formation, even if the dissociation of H2O in higher pH electrolyte was a sluggish step.

We then investigated the capacity adsorption and activation of CO2 and CO molecules on R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP via gas electro-response experiments in a self-designed gas adsorption electro-response device (Fig. S22)31. The results in Fig. S23 and Fig. S24 show that the CO2 and CO adsorption responses changed in the sequence of R-Cu1/CuNP < M-Cu1/CuNP < P-Cu1/CuNP, suggesting that the adsorption and activation of CO2 and CO were promoted as the ratio of Cu1 to CuNP decreased. On the other hand, the electrochemical CO stripping voltammetry tests of R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP were also performed to study the CO adsorption ability. Figure 3b shows a peak in the potential range of 0.8–0.9 V for all three catalysts. No peak was observed in the LSV curve without CO adsorption (Fig. S25). According to prior literature32,33,34, the peak at around 0.8–0.9 V can be attributed to the CO stripping peak. Interestingly, the CO stripping peak occurred at around 0.89 V for M-Cu1/CuNP and P-Cu1/CuNP, while it was around 0.82 V for R-Cu1/CuNP. The positive shift suggested that M-Cu1/CuNP and P-Cu1/CuNP had a stronger CO binding ability than R-Cu1/CuNP, which have higher Cu NP content. The results above indicated that the Cu NPs was beneficial for the adsorption and activation of CO2 and CO.

In situ X-ray absorption spectroscopy (XAS) experiments were performed on M-Cu1/CuNP to investigate the changes of Cu valence state and structure during CO2RR (Fig. S26). In the XANES spectra (Fig. S27), the Cu K-edge adsorption spectra did not show obvious difference under different potentials, indicating that the average Cu valence state kept stable during the reaction. Moreover, the peaks corresponding to Cu-N and Cu-Cu coordination still existed in FT EXAFS spectra (Fig. 3c), with no notable change was observed in peak intensity, suggesting the stability of the content ratio of Cu1 to CuNP. The results of in situ XAS experiments showed that the structure and content of Cu1 and CuNP remained stable during CO2RR.

In situ surface-enhanced Raman spectroscopy (SERS) was employed to reveal the interactions between Cu species and the reaction intermediates during CO2RR (Fig. S28)35,36,37,38. As shown in the Raman spectra under different potentials (Fig. 3d–f), the peaks assigned to the restricted rotation of adsorbed CO (298 cm−1, ν(Cu-CO)), Cu-CO stretching (365 cm−1, ν(Cu-CO)) and C-O stretching of atop *CO (2000–2100 cm−1, ν(CO)) were observed from −0.3 to −0.8 V over M-Cu1/CuNP. In contrast, the ν(Cu-CO) and ν(CO) peaks over P-Cu1/CuNP appeared from −0.4 V and −0.5 V, respectively. These observation could be related to the lower onset potential for C2 products formation over M-Cu1/CuNP. Only CO rotation peaks can be observed on R-Cu1/CuNP from −0.4 to −0.8 V, meaning poor CO2 reduction activity. In situ SERS was also conducted over Cu-N-C (Fig. S29), whereas no peaks associated with *CO intermediate were observed, suggesting that single atomic Cu was not conducive to the conversion of CO2 to adsorbed CO. The results of in situ SERS indicated that CO2-to-C2 products proceeded via CO intermediate process and a moderate content ratio of Cu1 to CuNP facilitated the conversion of CO2 to adsorbed CO.

Online differential electrochemical mass spectrometry (DEMS) was conducted to investigate reaction mechanisms (Fig. S30), which extracts volatile intermediates and products generated on the electrode surface into the mass spectrometer within milliseconds, utilizing the pressure difference in a vacuum as the driving force39,40. During five continuous cycles at −0.6 V vs. RHE, the m/z signal of 29 that correspond to CHO could be detected over M-Cu1/CuNP, while it was absent in P-Cu1/CuNP and R-Cu1/CuNP (Fig. S31). The results of online DEMS demonstrated that the crucial role of the proper Cu1 to CuNP content ratio in CHO formation. Meanwhile, in situ attenuated total reflection-surface-enhanced IR absorption spectroscopy (ATR-SEIRAS) spectra were further collected to trace the evolution of reaction intermediates during CO2RR from −0.2 to −0.9 V over R-Cu1/CuNP, M-Cu1/CuNP and P-Cu1/CuNP (Figs. S32, S33). The peak at around 2100 cm−1 can be attributed to electrogenerated CO adsorbed (*CO) on catalyst surface41,42, which did not appear until the applied potential reached −0.5 V on R-Cu1/CuNP and P-Cu1/CuNP. However, the ν(*CO) peak over M-Cu1/CuNP obviously existed from −0.3 to −0.9 V and the peak intensity was stronger than that of R-Cu1/CuNP and P-Cu1/CuNP, suggesting that M-Cu1/CuNP preferred to generate *CO intermediate. Meanwhile, a peak at 1748 cm−1 observed over M-Cu1/CuNP from −0.3 to −0.9 V can be ascribed to the *CHO intermediate12,38,39, and its change trend was similar to that of *CO intermediate, indicating that the *CHO was originated from the hydrogenation of *CO with the assistance of *H. Nevertheless, the *CHO peak was not observed on R-Cu1/CuNP and P-Cu1/CuNP, meaning that proper content ratio of Cu1 to CuNP is necessary for *CHO formation.

DFT calculations

DFT calculations were performed to gain insights into the CO2RR mechanism. According to the HRTEM and EXAFS fitting results, the Cu(111) facet and the Cu-N4 were used as the models of Cu NPs and atomic Cu sites, respectively. All the computational structure models and the detailed data are shown in Figs. S34–S40. We first calculated CO2 to adsorbed CO (*CO) process to verify if the catalyst is beneficial for CO2 activation, as *CO is a crucial intermediate for C2 products formation. The calculations revealed that CO2 could be converted to *CO via a *COOH intermediate. However, the energy barrier of CO2-to-*CO over the Cu(111) facet and Cu-N4 was found to be 0.44 and 1.40 eV, respectively (Fig. 4a and Fig. S35). This indicated that CO2 preferred to be activated on Cu(111) facet, which is accordance with the experiment results.

Fig. 4: The DFT calculations.
figure 4

a The energy barrier of CO2 was converted to *CO over Cu(111) and Cu-N4. b The free energy diagram for the CO2RR to describe the possible C–C coupling step from *CO over Cu(111). c Reaction energy of *CO hydrogenation to *CHO on Cu(111) as a function of *H coverage. d Proposed reaction mechanism for CO2RR to C2 products on the M-Cu1/CuNP.

Full size image

In general, the C-C coupling reaction was considered as the rate-determining step for CO2-to-C2+ products23,43. Although CO2 could be easily converted to *CO on Cu(111) facet, high uphill reaction energy was needed to form *O*CCO intermediate, meaning that the C-C coupling reaction through *CO dimerization is difficult. We further calculated the C-C coupling reaction through different C1 species (*CO, *CHO, *COH). As shown in Fig. 4b, the uphill reaction energy for the *CHO formation was 0.65 eV lower than that of the *COH formation (1.07 eV). This indicated that the *CO preferred to hydrogenate into *CHO rather than *COH, which is consistent with the ATR-SEIRAS results. Afterwards, both *CHO-*CO and *COH-*COH coupling were endoergic, while *CHO dimerization to form *OHCCHO* intermediate was exergonic. Therefore, the hydrogenation of *CO into *CHO then coupling of *CHO into *OHCCHO* is the most favorable routes in C-C coupling reaction.

Considering that the *CHO intermediate was originated from the hydrogenation of *CO with the assistance of *H, the influence of *H coverage on *CHO formation over Cu(111) facet were investigated. Our calculations revealed that that increasing the *H coverage on Cu(111) up to 1/4 decreased the reaction energy of *CO hydrogenation to *CHO (Fig. 4c and Fig. S41). However, a further increase of the *H coverage to 1/2 disfavored the formation of *CHO. These findings confirm that a moderate *H coverage facilitated the formation of *CHO, thus promoted the C-C coupling reaction and improved the CO2-to-C2 products efficiency. The moderate atomic Cu sites in M-Cu1/CuNP can accelerate the H2O dissociation process in alkaline electrolyte to provide the moderate *H coverage on Cu NP surface (Fig. 4d), therefore, the M-Cu1/CuNP exhibited excellent CO2RR to C2 products performance.

Discussion

In summary, a series of dual-active sites catalysts with co-loaded Cu NPs and atomic Cu sites on N-doped carbon matrix have been successfully synthesized and evaluated for CO2 electroreduction. Among them, M-Cu1/CuNP exhibited a C2 products FE of 75.4% with a corresponding partial current density of 289.2 mA cm−2 at −0.6 V, as well as remarkable long-term stability. The detailed study revealed that atomic Cu sites could promote H2O dissociation to provide *H, and the Cu NPs was beneficial for the adsorption and conversion of CO2. The in situ spectroscopic characterization showed that moderate content ratio of atomic Cu sites to Cu NPs facilitated CO2 conversion to adsorbed CO, resulting in M-Cu1/CuNP possessing a lower onset potential for C2 products. Furthermore, online DEMS and in situ ATR-SEIRAS revealed the presence of the intermediate *CHO over M-Cu1/CuNP, which was originated from the hydrogenation of *CO with the assistance of *H. DFT calculations demonstrated that the C-C coupling step was promoted through *CHO dimerization reaction on Cu NPs, and the moderate *H coverage facilitated the formation of *CHO. Therefore, the excellent catalytic performance of as-fabricated dual-active sites catalyst originated from the dual-active sites effect of atomic Cu sites and Cu NPs. The atomic Cu sites promoted H2O dissociation to provide *H, which in turn migrated to Cu NPs and facilitated *CO protonation to form *CHO by modulating the *H coverage on Cu NPs, leading to high activity for C2 products production. This work puts forward rational concept for promoting conversion CO2 to C2 products through modulate the adsorbed hydrogen coverage on Cu-based catalysts. We believe that it will inspire the design of more dual-/multi-active sites catalysts for multi-step reactions.

Methods

Materials

Copper nitrate hydrate (Cu(NO3)2·3H2O, purity > 99%), guanidine thiocyanate (C2H6N4S, purity > 99%), potassium hydroxide (KOH, purity > 85%), potassium bicarbonate (KHCO3, purity > 99.5%), sodium sulfate (Na2SO4, purity > 99%), 2, 2-dimethyl-2-silapentane-5-sulfonate (DSS, 99%), Deuterium oxide (D2O, purity > 99.9), gas diffusion electrode (YLS-30) with 10% PTFE and microporous layer, anion exchange membrane (FAA-3-PK-130) and Ni foam (purity > 99.8%, thickness 0.5 mmm) were purchased from Alfa Aesar China Co., Ltd. CO2 (99.999%), N2 (99.999%) and 10% H2/Ar were provided by Beijing Analytical Instrument Company. All materials were used directly without further purification and all the aqueous solutions were prepared by Milli Q water (18.2 MΩ cm, 298 K).

Catalysts characterization

The morphologies of as-synthesized catalysts were characterized by a JEOL JEM-2100F high-resolution transmission electron microscopy (HR-TEM). The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) characterization and corresponding energy-dispersive spectroscopy (EDS) were conducted on aberration-corrected JEM-ARM300F operated at 300 kV. Powder X-ray diffraction (XRD) analysis of the samples were performed on the X-ray diffraction (Model D/MAX2500, Rigaka) with Cu-Kα radiation and the scattering range of 2θ was from 5° to 90°, with a scanning rate of 5° min−1. X-ray photoelectron spectroscopy (XPS) analysis was performed on Thermo Fisher Scientific ESCA Lab 250Xi using 200 W monochromatic Al Kα (1486.6 eV) radiation under a pressure of 3 × 10−10 mbar, and the binding energy was referenced to the C 1 s peak at 284.4 eV. The X-ray absorption data at the Cu K-edge of the catalysts were recorded in transmission mode using ion chambers at the 4B9A beamline of Beijing Synchrotron Radiation Facility (BSRF) and the radiation was monochromatized by a Si (111) double-crystal monochromator. All collected spectra were analyzed using Athena and Artemis program within the IFEFFIT software packages.

Catalyst synthesis

50.8 mmol of guanidine thiocyanate and a quantity of Cu(NO3)2·3H2O were added to a flask containing 50 mL deionized water with vigorously stirring at 80 °C. After stirring for 4 h, the temperature was raised to 100 °C to evaporated water. The collected solid product was grounded evenly and pyrolyzed at 600 °C with 10 °C min−1 for 3.5 h under N2 atmosphere, heated at 400 °C with 2 °C min−1 for 2 h under 10% H2/Ar atmosphere. After cooled down to room temperature, the as-obtained solid was washed with ethanol and deionized water for several times and dried in vacuum at 60 °C overnight. Specifically, for R-Cu1/CuNP, M-Cu1/CuNP, and P-Cu1/CuNP, the amount of used Cu(NO3)2·3H2O was 3.81, 6.35 and 8.89 mmol, respectively. For comparison, Cu-N-C was prepared by treating M-Cu1/CuNP with acid treatment. The M-Cu1/CuNP was added into 50 mL 1 M sulfuric acid aqueous solution and heated at 80 °C for 48 h, then washed with deionized water several times and dried at 80 °C overnight.

Electrocatalytic CO2 reduction

To prepared gas diffusion electrode, 1 mg catalyst was suspended in 500 μL isopropanol with 10 μL Nafion D-521 dispersion (5 wt%) to form a homogeneous ink. Then the catalyst ink was spread onto the gas diffusion electrode of 0.5 × 2 cm2 in area by a micropipette to make sure the mass loading of the catalyst was 1 mg cm−2, and then dried under room temperature. All the electrochemical experiments were conducted on the electrochemical workstation (CHI 660E) equipped with a high current amplifier (CHI 680 C) and the CO2RR performance was investigated in flow cell (Fig. S7). The prepared gas diffusion electrode and Ni foam were used as the cathode and anode, respectively. An anion exchange membrane was used to separate the cathode and anode. Aqueous KOH solution (5 M) was used as the electrolyte solution and the electrolyte volume was 30 mL, and the electrolyte volume in the cell is 0.5 mL. The catholyte solution and anolyte solution were recirculated by two pumps with flow rates of 10 mL min−1 and 30 mL min−1, respectively. Meanwhile, CO2 gas was continuously supplied to the gas chamber by using a mass flow controller with a flow rate of 40 mL min−1. All the potentials were measured against a Hg/HgO reference electrode and converted to versus RHE with iR (80%) compensations, i represents the current obtained at corresponding potential, R is ohmic resistance of the cell measured by electrochemical workstation:

$${E}_{{RHE}}={E}_{{Hg}/{HgO}}+0.098+0.059\times {pH}-{iR}\times 80\%$$
(1)

The chronopotentiometry method was used for evaluating long-term activity stability of electrocatalyst in 5 M KOH electrolyte. In order to address the issues of flooding and carbonation accumulation of GDE, the CO2RR was interrupted every 5 h and the GDE were removed, washed with deionized water and followed by dryness under N2 atmosphere. Meanwhile, the electrolyte solution was refreshed for each interval.

Product analysis

The gaseous products were collected using a gas bag and quantified by gas chromatography (GC 7890B). A thermal conductivity detector (TCD) and a flame ionization detector (FID) were used to quantify H2, CO, and other alkane contents, respectively. The Faradaic efficiency (FE) of gaseous products was calculated by the equation:

$${FE}=z{nVF}/Q\times 100\%$$
(2)

Where z represents the number of electrons transferred for product formation, n is the volume concentration from GC, V is the total volume calculated by outlet flow rate, F is Faraday constant (96485 C/mol) and the Q is the amount of cumulative charge recorded by the electrochemical workstation.

The liquid product was analyzed by 1H NMR (Bruker Advance III 400 HD spectrometer) in deuteroxide. To accurately integrate the products in NMR analysis, the sodium 2, 2-dimethyl-2-silapentane-5-sulfonate (DSS) was the reference for ethanol and acetic acid, and phenol was the reference for formate. The FE of liquid products was calculated by the equation:

$${FE}={znVF}/Q\times 100\%$$
(3)

Where z represents the number of electrons transferred for product formation, C is the liquid concentration obtained from NMR, V is electrolyte volume, F is Faraday constant (96485 C/mol) and the Q is the amount of cumulative charge recorded by the electrochemical workstation.

Double layer capacitance measurements

The electrochemical active surface area is proportional to double layer capacitance, which measures the capacitive current associated with double-layer charging from the scan-rate dependence of cyclic voltammogram. The double layer capacitance was determined in a single-compartment electrolytic cell with 0.5 M KHCO3 aqueous solution as electrolyte. Ag/AgCl eletrode and graphite rod were used as reference electrode and counter electrode, respectively. The scan rates of cyclic voltammogram were 20, 40, 60, 80, 100 mV s−1.

Gas electro-response experiments

A self-designed gas adsorption electroresponse device (Fig. S19) was used to perform gas electro-response experiments. To prepared the electrode, 5 mg catalyst was suspended in 1 mL isopropanol to form a homogeneous ink. Then the catalyst ink was spread onto the Cu foam with area of 1 × 2 cm2 by a micropipette and then dried under room temperature. The as-prepared electrode was put into a sealed container and connected with electrochemical workstation though two electrode system. Before Ar, CO2 or CO gas was injected into the sealed container, the container was kept in vacuum state by a vacuum pump. Various potentials were applied on the electrode to observed the change of current curve as a function of time under different atmosphere. The adsorption of various gas on the electrode surface would induce the change of current response. Considering the catalyst loading and the size of the Cu foam used in each experiment is the same, the difference of current density under Ar and CO2 (or CO) atmosphere can reflect the adsorbed capacity of CO2 (or CO) on the catalyst surface.

CO stripping test

CO stripping experiments were conducted in a single-chamber electrolytic cell with three electrode system through linear sweep voltammetry method. To prepared working electrode, 1 mg catalyst was suspended in 500 μL isopropanol with 10 μL Nafion D-521 dispersion (5 wt%) to form a homogeneous ink. Then the catalyst ink was spread onto the carbon paper of 1 × 1 cm2 in area and then dried under room temperature. Ag/AgCl electrode and graphite rod were used as reference electrode and counter electrode, respectively. The electrolyte was 0.1 M Na2SO4 aqueous solution. The CO adsorption procedure was accomplished by polarizing the working electrode at +0.2 V and bubbling the electrolyte with CO for 10 min and subsequently with N2 for another 10 min. Then the linear sweep voltammetry was conducted from 0.5 to 1.2 V with scan rate of 10 mV s−1.

In situ XAS measurements

A custom-designed flow cell was used to conducted in situ XAS measurements (Fig. S23), the gas diffusion electrode loaded with catalyst (1 mg cm−2), Ni foam and Hg/HgO electrode were chosen as the working electrode, counter electrode and reference electrode, respectively. The 5 M KOH aqueous solution was used as electrolyte and were recirculated by pump with flow rates of 20 mL min−1. CO2 gas was continuously supplied to the gas chamber with a flow rate of 40 mL min−1. Data were recorded in fluorescence excitation mode using a Lytle detector. During the in situ XAS experiments, the spectrum was recorded following the application of a corresponding potential to the electrode for a duration of 600 s. Beamline of Beijing Synchrotron Radiation Facility (BSRF) and the radiation was monochromatized by a Si (111) double-crystal monochromator. All collected spectra were analyzed using Athena and Artemis program within the IFEFFIT software packages.

In situ Raman measurements

A flow cell with a quartz window by GaossUnion (Tianjin) Photoelectric Technology Company was used to carry out in situ Raman measurements using a Horiba LabRAM HR Evolution Raman microscope (Fig. S25). A 785 nm laser was used and signals were recorded using a 20 s integration and by averaging two scans. The gas diffusion electrode sprayed with the catalyst was used as working electrode and a graphite rod and a Hg/HgO electrode were used as counter and reference electrodes, respectively. The counter electrode was separated from the working electrode by anion exchange membrane. The 5 M KOH aqueous solution was used as electrolyte and were recirculated by pump with flow rates of 20 mL min−1. Meanwhile, CO2 gas was continuously supplied to the gas chamber with a flow rate of 40 mL min−1.

In situ ATR-SEIRAS measurements

The experiments were conducted in a modified electrochemical cell that integrated into a Nicolet 6700 FTIR spectrometer equipped with MCT detector cooled by liquid nitrogen (Fig. S27). The catalysts ink was dropped on a germanium ATR crystal deposited with Au film. A platinum wire and a Hg/HgO electrode were used as counter and reference electrodes. Each spectrum was collected with 32 times with a resolution of 4 cm−1. The CO2-saturated 3 M KOH aqueous solution was used as electrolyte to avoid that the germanium ATR crystal was damaged by strong basic electrolyte. The background spectrum was collected at the potential of +0.1 V vs RHE.

Online DEMS measurement

A custom-designed flow cell was used to conducted online DEMS measurements (Fig. S30). The gas diffusion electrode loaded with catalyst, platinum wire and Hg/HgO electrode were chosen as the working electrode, counter electrode, and reference electrode, respectively. The 5 M KOH aqueous solution was used as electrolyte and were recirculated by pump with flow rates of 20 mL min−1. CO2 gas was continuously supplied to the gas chamber with a flow rate of 40 mL min−1. A potentiostatic test at −0.6 V (vs. RHE) was performed in an alternating manner. Following the completion of each electrochemical test and the return of the mass signal to its baseline level, the subsequent cycle commenced, employing identical experimental conditions to minimize the possibility of inadvertent errors. The experiment was terminated after the completion of five cycles.

DFT calculations

Density function theory calculation were performed by using the CP2K package44. PBE functional45 with Grimme D3 correction46 was used to describe the system. Unrestricted Kohn-Sham DFT has been used as the electronic structure method in the framework of the Gaussian and plane waves method47,48. The Goedecker-Teter-Hutter (GTH) pseudopotentials49,50, DZVPMOLOPT-GTH basis sets were utilized to describe the molecules. A plane-wave energy cut-off of 500 Ry has been employed. The potential-dependence of reaction free energies in elementary steps involving proton-electron transfers was evaluated using the computational hydrogen electrode (CHE) approach51.

The Gibbs free energy is calculated using:

$$\Delta {G}_{{free}}=\Delta {E}_{{DFT}}-T\Delta S+{{{\rm{ZPE}}}}$$
(4)

Where ΔGfree is Gibbs free energy, ΔEDFT is energy calculate from DFT, S is entropy, T is temperature (300 K), and ZPE is zero-point energy.