The use of electrochemical CO2 reduction reaction (CO2RR) to synthesize carbon-based fuels in conjunction with renewable electricity represents a promising strategy to address current concerns regarding global energy shortages and environmental issues such as climate change 1,2,3. This process can potentially generate a wide variety of products 4,5,6,7,8, among which the liquid compounds are preferred 9 because of their high energy densities (e.g., -1366.8 kJ/mol for CH3CH2OH) and ease of storage when serving as substitutes for gasoline 10,11. However, due to the high energy barrier to C-C bond formation, which competes with the formation of C-H or C-O bonds 12,13,14, the generation of multi-carbon liquid products via the CO2RR is still very difficult. To date, Cu is the only electrocatalyst found able to generate Cn (n ≥ 2) liquid products from the CO2RR with appreciable reaction rates, due to the strong adsorption of the *CO intermediate on the Cu surface that promotes C-C bond formation via CO dimerization or CO-CHO coupling 5,12,15. Unfortunately, the formation of C-C bonds on Cu is very complicated and the pathway remains elusive 16, which have greatly limited the development of highly selective electrocatalyst to reduce CO2 to Cn (n ≥ 2) liquid products.
Herein, we report a Sn-based tandem catalyst consisting of SnS2 nanosheets that are able to produce formate intermediates and single Sn atoms anchored on a 3D carbon support via a local SnO3 cluster (Sn1-O3G) that generate *OCO(OH) bicarbonate intermediates, followed by C-C coupling to electrochemically reduce CO2 to ethanol. Based on this strategy, an ethanol selectivity of more than 70% was achieved over a wide potential range from − 0.6 to -1.1 V versus the reversible hydrogen electrode (RHE). This catalyst was found to maintain 97% of its initial activity after 100 hours of continuous reaction at a geometric current density as high as 17.8 mA·cm− 2. The C-C bond formation mechanism on this newly-developed tandem catalyst was elucidated by combining isotope (13CO2 and H13COOH) labelling experiments with extensive density functional theory (DFT) studies, providing a highly efficient C-C coupling pathway for ethanol synthesis via the CO2RR, with a low activation energy, high selectivity and good stability. These properties can be attributed to the dual active centers of Sn and O in Sn1-O3G, which are able to adsorb *CHO and *CO(OH) intermediates, respectively, thereby promoting the subsequent C-C coupling of these intermediates to form ethanol. It is anticipated that this work will lead to new applications for single-atom catalysts 17,18 with polar active centers in multi-step chemical reactions.
The Sn-based tandem catalyst (denoted as SnS2/Sn1-O3G hereafter) was synthesized by a solvothermal method, the details of which are provided in the Extended data Figs. 21–22. We first evaluated the electrochemical CO2RR performance of the SnS2/Sn1-O3G using chronoamperometry in a H-type cell filled with CO2-saturated 0.5 M KHCO3 solution (Extended data Figs. 23–28). Extended data Fig. 25a provides typical current profiles obtained from the SnS2/Sn1-O3G, biased at various potentials from − 0.6 to -1.1 V vs. RHE in a CO2-saturated 0.5 M KHCO3 solution. The gaseous and liquid products were analyzed by an on-line gas chromatography and a nuclear magnetic resonance (NMR) spectroscopy (Extended data Figs. 1–20). The results obtained from combination of geometric current density of the potentiostatic electrolysis of CO2 over SnS2/Sn1-O3G and NMR spectral analysis (Extended data Figs. 8–13) are presented in Figs. 1a and b. These data show that the catalyst reproducibly yielded ethanol with a Faradaic efficiency (FE) as high as 82.5% at -0.9 V (vs. RHE) and a geometric current density of 17.8 mA/cm2. More importantly, the FE associated with ethanol production was maintained above 70% over the potential window from − 0.6 to -1.1 V (vs. RHE). The yields of ethanol produced over SnS2/Sn1-O3G using isotopically labelled 13CO2 and 12CO2 at multiple points for different sampling intervals (2, 6 and 12 hours) are consistent quantitatively (Extended data Figs. 27, 28). This SnS2/Sn1-O3G catalyst was also extremely stable during CO2 reduction to ethanol, maintaining approximately 80% ethanol selectivity at -0.9 V (vs. RHE) and 97% of its initial activity after 100 hours of continuous reaction (Fig. 1c). It is worth noting that no other C2 products (such as ethylene or acetate) were detectable over the entire potential range applied during the CO2RR. Compared to state-of-the-art catalysts for electrochemical CO2 reduction to ethanol (Fig. 1d), our Sn-based tandem catalyst shows the best CO2-to-ethanol performance.
To study the superb CO2RR performance to ethanol, the as-synthesized catalyst was carefully characterized by scanning electron microscopy (SEM), transmission electron microscopy (TEM), sub-angström-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), X-ray diffraction (XRD), Raman spectroscopy, X-ray photoelectron spectroscopy (XPS) and X-ray absorption spectroscopy (XAS) to reveal the structural information. The SnS2/Sn1-O3G catalyst comprises of two components: vertically-aligned SnS2 nanosheets (Extended data Figs. 29a) grown on a 3D carbon support decorated with atomically-dispersed Sn atoms (Figs. 2d and h). XPS and inductively coupled plasma mass spectrometry (ICP-MS) results show negligible metal contaminations (other than Sn) in 3D carbon and SnS2/Sn1-O3G (Extended data Figs. 30,31 and Table 7). To probe the exact local structure and coordination environment of these single Sn atoms on the carbon support, the SnS2 nanosheets were removed by acid etching, as shown in Fig. 2c, confirmed by XPS and SEM analyses (Extended data Figs. 33–34). Figure 2d presents a HAADF-STEM image of the acid-etched SnS2/Sn1-O3G (that is, the Sn1-O3G) in which bright spots corresponding to metal atoms are clearly seen to be well dispersed across the carbon framework. Figure 2e provides enlarged images of the areas indicated in Fig. 2d and shows that each metal atom is connected to the carbon substrate via three bonds. The line profile in Extended data Fig. 35 demonstrates bright spots with sizes of approximately 0.2 nm while the electron energy loss spectroscopy (EELS) spectra as shown in Fig. 2f indicate that around 475 and 540 eV, corresponding to the Sn M edge and O K edge 19, the intensity from the bright spot is stronger than that from the dark area. The insert to Fig. 2d highlights the difference in the EELS spectra obtained from the two regions and confirms that bright spots in the HAADF-STEM image are Sn atoms, and that these atoms are in an oxygen-rich environment. The Sn content in Sn1-O3G is 0.8 wt.% determined by ICP-MS.
The chemical oxidation state and coordination environment of the single Sn atoms were further examined by XAS. Figure 2g shows the Sn K edge X-ray absorption near edge structure (XANES) spectra of Sn1-O3G and reference Sn foil (Sn0), Sn (II) porphyrin (Sn2+) and Sn (IV) porphyrin (Sn4+). By comparing the energy positions of the white lines and the rising edges, the oxidation state of the Sn atoms in the Sn1-O3G was determined to be + 4, in good agreement with the XPS and Mössbauer spectroscopy results (isomer shift and quadrupole splitting are 0.088 mm s− 1 and 0.22 mm s− 1, respectively, Extended data Fig. 33d) 20. Figure 2h shows the Fourier transform of the phase-uncorrected extended X-ray absorption fine structure (EXAFS) spectra. The peaks centered at 2.3 and 3.8 Å, corresponding to Sn-Sn interactions in Sn foil, are not present in the Sn1-O3G spectrum, suggesting a high degree of dispersion of the Sn atoms in Sn1-O3G, which is consistent with the HAADF-STEM results. The main peak in the Sn1-O3G spectrum appears at approximately 1.60 Å, which is shorter than those associated with the Sn-N bonds in Sn (II) porphyrin and Sn (IV) porphyrin. The wavelet contour plot of the Sn1-O3G (Fig. 2i) exhibits only one intensity maximum at approximately 5.2 Å−1, which does not contain a peak that can be indexed to Sn-Sn coordination, as is evident in the wavelet contour plot for the Sn foil in Extended data Fig. 36. Considering that the 3D carbon support has a high concentration of oxygen functional groups (as demonstrated by the XPS and XAS results in Extended data Figs. 38 and 39) but does not contain N, this peak can be assigned to either Sn-O or Sn-C bond. The first shell of the Fourier transformation of the EXAFS spectrum of Sn1-O3G was fitted with Sn-O or Sn-C back-scattering path (Extended data Fig. 40) and Extended data Table 6 summarizes the fitting results. A coordination number of approximately 3.7 was determined using Sn-O, but a value of 6.8 was obtained using Sn-C. Based on the low temperature synthesis employed and the HAADF-STEM as well as the EELS, XPS, XAS and Mössbauer spectroscopy results (Fig. 2d), the Sn atoms in Sn1-O3G are deduced to coordinate to O atoms rather than C. The different charge densities and coordination environments of the Sn and O atoms render the Sn atoms electron-deficient and the surrounding O atoms electron-rich, and the spin-density population and Bader charge analysis implies a + 4 oxidation state for Sn (Extended data Fig. 41). Combining our experimental results with theoretical calculations, it appears that the Sn atoms in Sn1-O3G were immobilized on the O-rich carbon support through binding with three O atoms to form a tridentate complex with an additional OH to stabilize the single Sn atoms, that is, Sn1-O3(OH)G, as shown in the insert to Fig. 2j. DFT calculation gives the formation energy of the Sn1-O3(OH)G site of -4.75 eV, suggesting thermodynamically stable structure of Sn1-O3(OH)G. Further XAS and HAADF-STEM investigations (Extended data Fig. 42) indicate that the O atoms coordinated Sn catalytic sites are robust in the CO2RR process.
To understand the role of each component in the SnS2/Sn1-O3G for the CO2RR, we synthesized SnS2 nanosheets using the same protocol as for the SnS2/Sn1-O3G but without including a carbon support (structural information regarding the resulting SnS2 is provided in Extended data Fig. 43). The SnS2 nanosheets were found to catalyze electrochemical CO2 reduction to give formate with an FE of approximately 60% at -0.9 V (vs. RHE) (Fig. 3a), in agreement with the previous report 21. In contrast, Sn1-O3G alone was only able to reduce CO2 to CO as the main CO2RR product (Fig. 3b). Interestingly, a mechanically-mixed combination of SnS2 nanosheets and Sn1-O3G was found to produce ethanol with an FE of approximately 15% at -0.9 V (vs. RHE) (Fig. 3c), while a mixture of SnS2 nanosheets with 3D carbon (without the single Sn atoms, Extended data Fig. 44) failed to produce ethanol. From these findings, we propose that the formation of ethanol over the SnS2/Sn1-O3G may result from C-C coupling between the two intermediates associated with formate and CO.
To verify our hypothesis, we intentionally introduced a small amount of HCOOH into the reaction system (in a Na2SO4 electrolyte instead of KHCO3) during the CO2RR over the Sn1-O3G. As shown in Fig. 3d, compared to the case without HCOOH (Fig. 3b), the cathodic current exhibits a clear decrease in the N2-saturated Na2SO4 electrolyte upon HCOOH addition (Extended data Fig. 45). A similar trend was also observed in the CO2-saturated electrolyte, suggesting that HCOOH affected the electrochemical reaction. In CO2-saturated electrolyte, the FECO was found to decrease with increase in the amount of HCOOH, while the FEethanol value increased. The inset in Fig. 3d summarizes the CO2RR performance of Sn1-O3G with addition of HCOOH in a CO2-saturated Na2SO4 electrolyte at various applied potentials. At an applied potential of -1.0 V (vs. RHE), adding HCOOH (50 µL) to the CO2-saturated electrolyte decreased the FECO (inset in Fig. 3b) from 50% to approximately 40%, while increasing FEethanol nearly 40-fold, from close to 0.2–7.4%. The partial geometric current density associated with ethanol formation also increased, from 0.002 to 0.9 mA/cm2. These results demonstrate the importance of HCOOH in C-C coupling to form ethanol on O atoms coordinated Sn catalytic sites. In contrast, unlike the result for the Sn1-O3G catalyst, the Sn1-N4G catalyst (with 4 nitrogen atoms coordinated Sn center) failed to produce any C2 products in CO2RR with addition of HCOOH (Extended data Figs. 46 and 47), indicating the critical roles of both the single atomic Sn center and the coordinated oxygen atoms in Sn1-O3G for the C-C coupling reaction. We also investigated the CO2RR products generated by the Sn1-O3G catalyst using gaseous CO and HCOOH as the reactants and found no C2 products, indicating the critical role of the adsorbed *CO(OH) intermediate (see below) formed via CO2 reduction on the Sn1-O3G site in the formation of ethanol.
To elucidate the mechanism of ethanol formation through the CO2RR, isotopically-labeled reactants (12CO2, 13CO2, H12COOH and H13COOH) were used to trace the pathway of the C atoms in the final C2 product formed on the Sn1-O3G catalyst. Figure 3e and Extended data Fig. 48 show that the methyl C (*CH3) in ethanol comes from HCOOH, while the C in CO2 goes to the methylene C (*CH2OH). Therefore, Fig. 3f schematically summarizes the CO2RR process to form ethanol taking place on the SnS2/Sn1-O3G. The comprehensive discussion for the ethanol formation process on the SnS2/Sn1-O3G tandem catalyst is provided from additional structural characterization and kinetics simulation (Extended data Figs. 49–53 and Tables 9–17), indicating that diffusion of HCOOH generated from the SnS2 component to the single atomic Sn sites on Sn1-O3G is the rate-determining step for ethanol formation.
Based on the above experimental results, we built the active site model of the catalyst. Extensive DFT calculations were performed using the computational hydrogen electrode (CHE) model 22,23,24,25 to map the CO2RR pathway and energetics during the formation of ethanol. The results show that, during the CO2 reduction process, the proton-coupled electron transfer (PCET, H+/e−) and dehydration of the Sn1-O3(OH)G pre-catalytic sites is exergonic and the resulting Sn1-O3G is of radical nature (going from species 0 to 1 in Fig. 4, Extended data Figs. 54 and 55). Two possible CO2 binding mechanisms were assessed: monodentate (2 − 1) and side-on (2) 26,27 (Extended data Fig. 56a). Thus, there are two possible pathways for CO2 reduction, one of which produces CO and the other produces ethanol. Free energy pathways for competing C1 and C2 products were systematically calculated as shown in Extended data Figs. 56–57 and Extended data Table 18. In the case of species 2, the O2SnO…C linkage is cleaved and a face-to-face cyclo-addition between Sn = O double bond and O = CO double bond leads to the CO2 activation, with a CO2 adsorption free energy of -0.75 eV. The activated *CO2 (marked in bold) can then have a PCET to form O2Sn-O-CO(OH) in 3, with exergonicity of 1.68 eV. This bicarbonate species *OCO(OH) is formed as the dominant intermediate in the hydrogenation process of *CO2 catalyzed by O2Sn = O. As a result of forming the strongly bound carbonate species, it hardly generates CO directly except from the recovered catalyst via 3 − 1 intermediate (as shown in Extended data Fig. 55b). The differential charge distribution analysis and Bader charges of 3 determined that the positive charge was higher on the Sn atom, such that bonding between Sn atom and HCOOH (externally added or in-situ produced over SnS2) was favored. With PCET, the carbonyl group in HCOOH* can easily dehydrate to generate Sn-CHO in intermediate 4, with a free energy barrier of 0.52 eV. This intermediate 4 features *CHO on Sn and *CO(OH) on the neighboring O (i.e., O2Sn-O-CO(OH) bicarbonate), which explains the activities of Sn and O atoms as the dual active centers in the Sn1-O3G. Following the drastic reduction of all carbonyl groups in 4, intermediate 5 is formed as the precursor for C-C coupling. The free energy barrier to C-C bond formation is 0.13 eV, which is much lower than those of Cu-based catalysts (> 1.0 eV) 13,28. Thus, intermediate 6, which contains a Sn-C-C-O quaternary ring, is obtained. As reduction proceeds, with exergonicity of 0.99 eV, the Sn-C bond breaks upon further PCET, then forms the *CH3CHOH intermediate (7), followed by PCET and O-C bond cleavage (with a free energy barrier of 0.87 eV), and finally releases CH3CH2OH from the active site. Owing to the flexibility of the bond between the active O atom of Sn1-O3 and the C of the support, the Sn1-O3G catalyst is distinguished by the dual active sites of Sn and O that are able to adsorb different C-based intermediates, providing a novel platform for C-C bond formation during the CO2RR to produce ethanol.
In summary, we have developed a Cu-free, Sn-based tandem catalyst to produce ethanol from the CO2RR, in which the dual active centers of Sn and O atoms on Sn1-O3G serve to adsorb different C-based intermediates, which effectively lowers the C-C coupling energy between *CO(OH) and *CHO. Our tandem catalyst enables a formal-bicarbonate coupling pathway, which not only provides a novel platform for C-C bond formation during ethanol synthesis and overcomes the restrictions of Cu-based catalysts, but also offers a unique strategy for manipulating CO2 reduction pathways toward desired products.