Electrochemical CO2 reduction reaction (CO2RR), as a promising strategy for CO2 mitigation and transformation, has been extensively studied over the past few decades1,2,3,4,5. Liquid products, especially CH3OH, possess significant advantages because of their high energy density and ease of storage6,7. However, few catalysts are capable to convert CO2 to products other than CO or HCOOH with high activity and selectivity. Electrochemical CO reduction reaction (CORR) emerges as an encouraging approach to produce other products8,9,10,11,12. Among the reported electrocatalysts, Cu is the mostly used one for CORR, but its selectivity towards a single product is poor because of the complicated chemical states of Cu surface13,14,15. Till now, developing efficient electrocatalysts for highly active and selective CORR remains challenging yet urgent.

Single atom catalysts (SACs) could act as a promising platform to selectively catalyze CORR due to the absence of consecutive active sites16,17,18. However, few SACs have been reported to reduce CO to generate liquid products because SACs usually show weak adsorption towards *CO, which is detrimental to the subsequent hydrogenation reaction. Moreover, the real structure of experimentally synthesized SACs is rather complex that contains a large variety of single-atom centers with different coordination environments, which greatly complicates the understanding of the structure-performance relationship. On the other hand, molecular catalysts with precisely defined coordination structures offer a model system to probe the reaction mechanism and reveal the catalyst structure-performance relationship19,20,21,22,23. Cobalt phthalocyanine (CoPc) molecular catalyst showed the capability to electrochemically reducing CO to CH3OH, however, the current density and the CH3OH selectivity are still unsatisfactory24. The spatial structure of binuclear cobalt phthalocyanine is very different from that of mononuclear cobalt phthalocyanine, which is anticipated to result in distinct crystal field and thus different catalytic performance, making binuclear cobalt phthalocyanine an interesting candidate to be investigated in CORR.

In this work, two model catalysts were constructed by anchoring cobalt phthalocyanine and binuclear cobalt phthalocyanine (M-CoPc and B-CoPc) on nitrogen-doped carbon support. It is found that the B-CoPc can be transferred from low-spin state (LS, 1/2) to high-spin state (HS, 3/2) after thermal treatment, and the HS B-CoPc is able to catalyze CORR to methanol much more effectively than LS M-CoPc and B-CoPc. The CH3OH partial current density reaches about 35 mA/cm2 at −0.8 V (vs. RHE) with a methanol Faradaic efficiency (FE) of 50%. Operando attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) measurements and density functional theory (DFT) calculations disclose that the high spin Co2+ in HS B-CoPc can enable electron accumulation on *CO that will greatly weaken the C–O bonding, promoting the hydrogenation of CORR intermediates (*CO/*CHxO), which lead to boosted CORR activity and selectivity.


Structure characterization

CoPc (M-CoPc) and binuclear CoPc (B-CoPc) were first anchored onto nitrogen-doped carbon (NC) via π-π interaction to obtain the model SACs (M-CoPc-RT and B-CoPc-RT), which were then thermally treated at 400 °C in an Ar atmosphere to enhance the structure stability of the composites (named as M-CoPc-400 and B-CoPc-400, respectively, Fig. 1a). Figure S1a, b and Figure S2a, b show mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectra of M-CoPc and B-CoPc. Both M-CoPc-400 and B-CoPc-400 display three-dimensional, ultrathin sheet-like structures with all elements uniformly distributed as shown in the scanning electron microscope (SEM), transmission electron microscope (TEM) images, and energy-dispersive X-ray spectroscopy (EDX) elemental mappings (Figs. S1, S2). Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was further used to directly observe the atomically dispersed Co atoms, which appear as bright spots as shown in Fig. 1e. The structure evolution of M-CoPc-RT and B-CoPc-RT during thermal treatment was investigated by in situ Raman spectroscopy. As shown in Fig. 1b, c, the Raman spectra of M-CoPc-RT and B-CoPc-RT exhibit identical characteristic features25. X-ray diffraction (XRD) reveals similar structures of M-CoPc and B-CoPc with only carbon diffraction peaks (Fig. 1d). Moreover, all characteristics of CoPc were preserved during the thermal treatment, indicating that the structure of CoPc did not change significantly before and after the thermal treatment.

Fig. 1: Structural characterization.
figure 1

a Schematic of the synthesis process for M-CoPc-RT/400 and B-CoPc-RT/400. In situ Raman spectra of b M-CoPc-RT and c B-CoPc-RT recorded during the thermal treatment process. d X-ray diffraction patterns. e The high magnification HAADF-STEM image of (i) M-CoPc-400 and (ii) B-CoPc-400 (scale bar, 2 nm). f Co K-edge XANES spectra before and after annealing. g The corresponding Fourier transformation (FT)-EXAFS spectra.

The valence states of Co in M-CoPc-400 and B-CoPc-400 were examined by X-ray photoelectron spectroscopy (XPS). Figure S3 shows the C 1s, N 1s, and Co 2p XPS spectra. The Co 2p XPS peaks for M-CoPc-RT and B-CoPc-RT are obviously different in peak shape and binding energy (Fig. S3c), indicating different interactions between M-CoPc-RT and B-CoPc-RT with NC support. In addition, the binding energy of Co 2p for M-CoPc-400 and B-CoPc-400 are similar to that for M-CoPc-RT and B-CoPc-RT, matching well with the in situ Raman spectroscopy results, suggesting minimal structural change of CoPc during thermal treatment.

X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) characterizations were employed to further explore the chemical states and local coordination structure of the Co atom. Figure 1f shows the Co K-edge XANES spectra of M-CoPc-RT/400 and B-CoPc-RT/400. The Co K-edge XANES spectrum of M-CoPc-RT/400 is similar to that of CoPc, indicating that the M-CoPc-RT/400 possesses the same Co oxidation state and D4h symmetry as CoPc26. However, the intensity of peak B (1s→4pz transition) which is a fingerprint of square-planar metal-N4 structure is slightly weaker in B-CoPc-RT/400, suggesting a distorted D4h symmetry of the Co atom27. Furthermore, the B-Co-RT/400 displays a remarkably larger intensity ratio of peak C to peak D (IC/ID) than M-CoPc-400 (peak C and peak D can be assigned to the 1s→4px,y transitions and multiple scattering processes, respectively). Increased IC/ID had been shown to be beneficial towards promoting electrochemical reactions28,29. In addition, C and N K-edge XANES spectra also indicate that M-CoPc-400 and B-CoPc-400 compose of the similar species (Fig. S4), in agreement with the XPS results. Figure 1g displays the Fourier-transformed EXAFS spectra. The peak at 1.55 Å for M-CoPc-RT/400 results from the scattering of the first Co-N shell. Notably, the first shell Co-N distance of B-CoPc-400 is slightly longer than that of B-CoPc-RT, which may be caused by molecular distortion induced by heat treatment30. Table S1 summarizes the EXAFS fitting results (coordination number and length of Co-N bond, Fig. S5).

Electrochemical CORR performance

CORR over M-CoPc-RT/400 and B-CoPc-RT/400 were assessed by cyclic voltammetry (CV) in an Ar-saturated or CO-saturated 0.5 M KOH solution. As shown in Figure S6c, d, CV curves of M-CoPc-400 and B-CoPc-400 display the clear redox peak at about −0.51 V (vs. RHE) in Ar-saturated 0.5 M KOH solution, corresponding to Co(II) reduction to Co(I)31. Compared to that in Ar-saturated 0.5 M KOH solution, the onset potential of the redox peak in CO-saturated 0.5 M KOH solution clearly shifted to lower cathodic potentials, suggesting that CO activation took place on the Co(II) center of M-CoPc-400 and B-CoPc-400, instead of Co(I).

Besides, the two thermally treated catalysts display much higher current densities than the untreated ones. The cathodic redox peak current of M-CoPc-400 and B-CoPc-400 exhibit a linear dependence on CV scan rate (Fig. S7a, b), suggesting that the CORR on M-CoPc-400 and B-CoPc-400 is a surface reaction-controlled process. In addition, the linear sweep voltammetry (LSV) curves (Fig. 2b) further show that the B-CoPc-400 exhibits a much better CORR performance than M-CoPc-400. Figure 2c compares the potential-dependent product distribution of CORR over M-CoPc-400 and B-CoPc-400. The main product on M-CoPc-400 is H2, with a Faradaic efficiency (FE) over 80% in the potential range from −0.5 V to −1.0 V (vs. RHE). The maximum FE for CH3OH over M-CoPc-400 is ~19% at −0.7 V (vs. RHE), matching well with the early work21.

Fig. 2: Catalytic performance.
figure 2

a Schematic diagram showing the MEA for CORR. b LSV curves acquired in Ar and CO-saturated 0.5 M KOH solution on a carbon paper at a scan rate of 5 mV/s. c Potential-dependent product selectivity for CORR catalyzed by M-CoPc-400 and B-CoPc-400. d Stability of M-CoPc-400 recorded at −0.7 V (vs. RHE). The measurement was performed at the condition of 1 atm CO and room temperature in CO-saturated 0.5 M KOH.

On the contrary, CH3OH becomes the dominate product over B-CoPc-400 in the potential range of −0.5 V to −0.8 V (vs. RHE), illustrating that the B-CoPc-400 could catalyze CORR much more effectively. 1H 1D NMR spectra of products and DMSO at different potentials (vs. RHE) for B-CoPc-400 and M-CoPc-400 are shown in Fig. S8a, b. The CH3OH partial current density can reach about 35 mA/cm2 at −0.8 V (vs. RHE) with a FE of 50%, and the maximum CH3OH FE of 57% appears at −0.6 V (vs. RHE) over B-CoPc-400. The yields of CH3OH produced over B-CoPc-400 using isotopically labeled 13CO and 12CO at multiple points for different potentials (−0.5 V, −0.6 V, −0.7 V vs. RHE) are consistent quantitatively (Fig. S9). Table S2 displays FE of all the products during CORR. The calculated CH3OH turnover frequency (TOF) over B-CoPc-400 and M-CoPc-400 are displayed in Fig. S10, suggesting that the loading amount of Co plays a negligible role in affecting the electrochemical performance. Table S3 compares the CORR performance between our prepared catalyst and those reported in the literature.

In addition, the influence of treatment temperature and types of support on catalytic performance were also investigated, and the results are summarized in Fig. S11, showing that the treatment temperature has a significant effect on the catalytic performance, while the influence of support is insignificant. Moreover, the B-CoPc-400 could produce CH3OH with a FE of 53% at −0.7 V (vs. RHE) with negligible current loss for 10 h, corroborating its excellent catalytic durability (Fig. 2d). Figure S12 shows the HRTEM, HAADF-STEM and XAS results after CORR stability test, corroborating the robust stability of B-CoPc-400.

Mechanistic insights into proton-feeding

To understand the different CORR catalytic activity as well as the CORR mechanism over M-CoPc-400 and B-CoPc-400, operando ATR-SEIRAS was conducted to probe the reaction intermediates. Figure 3a, b shows the operando ATR-SEIRAS spectra of CORR over B-CoPc-400 and M-CoPc-400. Three peaks at ~1870 cm−1, ~1650 cm−1, and ~1400 cm−1 are clearly identified, which can be assigned to *CO, H2O, and *CHO, respectively32,33,34. Notably, both M-CoPc-400 and B-CoPc-400 could adsorb CO at 0 V (vs. RHE). The *CO vibrational frequency on M-CoPc-400 is slightly lower than that on B-CoPc-400 (1911 and 1951 cm−1 at 0 V vs. RHE), indicating a slightly lower binding strength of B-CoPc-400 towards CO. With increasing cathodic potential from 0 to −0.9 V (vs. RHE) over M-CoPc-400 and B-CoPc-400, *CO vibrational frequency red-shifted with a decrease in peak intensity due to the Stark effect, and the intensity of the *CO stretching peak over B-CoPc-400 decreased faster with increasing cathodic potential as compared to that over M-CoPc-400 (Fig. 3e), indicating faster consumption of *CO over B-CoPc-400. When D2O was used instead of H2O, the position of CHO moved by 25 cm−1 towards smaller wavenumbers due to isotopic redshift caused by the mass effect, confirming that the peak at ~1400 cm−1 can be ascribed to *CHO (Fig. S14). Compared to M-CoPc-400, the peak intensity of *CHO is always lower on B-CoPc-400 in the entire potential range (from 0 to −0.9 V vs. RHE), indicating lower coverage of *CHO species on the B-CoPc-400 surface, which may result from the fast hydrogenation of *CHO to the final product. It is worth to note that the C–H bands between 3000 and 2800 cm−1 originating from different vibrational modes in −CH3 and CH2 over B-CoPc-400 are inverted peaks, different from the positive peaks over M-CoPc-400 (Fig. 3c, d). Generally, inverted peaks in ATR-SEIRAS spectrum indicate consumption of adsorbed species as compared to the condition of background. Therefore, the inverted C–H bands over B-CoPc-400 suggest that under applied cathodic potential, the coverage of C–H species decreased as compared to the OCP condition, and the increasing of cathodic bias promoted the C–H species desorption from the B-CoPc-400. Time-dependent infrared spectroscopy was carried out to explore the stability of intermediates during the CORR over B-CoPc-400. *CO, *CHO, and *C-Hx signals could remain intact throughout the measurement, highlighting the excellent durability of B-CoPc-400 (Figs. S15 and S16).

Fig. 3: Operando ATR-SEIRAS and KIE measurements.
figure 3

Operando ATR-SEIRAS spectra of CO reduction over a, c B-CoPc-400 and b, d M-CoPc-400 in CO-saturated 0.5 M KOH. The spectra were collected at constant potentials with 0.1 V interval in the cathodic direction from OCP to −0.9 V (vs. RHE). The operando ATR-SEIRAS spectra recorded in Ar-saturated 0.5 M KOH are shown in Fig. S13. e Potential dependent CO stretching peak intensity. f KIE of H/D in CORR to CH3OH at −0.6 V (vs. RHE) over B-CoPc-400 and M-CoPc-400.

To further examine the relation between kinetics of CORR and proton-feeding, the kinetic isotope effect (KIE) of H/D was measured using H2O and D2O as the proton sources, which was performed at −0.6 V (vs. RHE)30. When D2O was used to replace H2O in 0.5 M KOH or 0.5 M K2SO4 electrolyte, the partial current density of methanol was significantly reduced for M-CoPc-400. As shown in Fig. 3f, compared to B-CoPc-400, M-CoPc-400 displays a much higher KIE value of 2.10 and 2.13 in KOH and K2SO4 solution, respectively, indicating that the CORR over M-CoPc-400 was limited by proton transfer, consistent with the operando ATR-SEIRAS results. From the operando ATR-SEIRAS and KIE results, it can be deduced that the CORR over M-CoPc-400 and B-CoPc-400 was rate-limited by the *CO/*CHxO hydrogenation and methanol desorption step, respectively.

Mechanistic insights into spin structure

To dig out the origin of fast rate of hydrogenation over B-CoPc-400 in CORR to methanol, Co L2,3-edge XAS measurements were conducted to differentiate the electronic structure between M-CoPc-400 and B-CoPc-400. Transition metal L2,3-edge XAS spectra are sensitive to the change of oxidation state, spin state and/or orbital. Figure 4a, d shows the experimental and simulated Co L2,3-edge XAS spectra of the samples, which can be divided into L2 and L3 regions, corresponding to the transition from 2p1/2 and 2p3/2 levels to unoccupied 3d orbitals, respectively. As shown in Fig. 4a and Fig. S17a, b, the Co L2,3-edge XAS spectrum of M-CoPc-RT and M-CoPc-400 are similar, which display the same characteristic features as that of CoPc molecule35,36. Three features appear in the energy region of Co L3-edge, denoted as A1, A2 and A3 as shown in Fig. 4a, d. The peak A1 can be assigned to the transition from 2p3/2 to the 3dz2 orbitals, while the peaks A2 and A3 are originated from the transition to 3dx2-y2 orbitals37. The Co L2,3-edge XAS spectrum of M-CoPc-RT/400 were simulated using the configuration interaction cluster model that includes the full atomic Multiplet theory and the hybridization with the ligands38. The Co 3d to N 2p transfer integrals were estimated for the various Co-N coordinations according to Harrison’s prescription39. Calculations with the parameters (10Dq = 2.5 eV, Ds = 0.42 eV, Dt = 0.25 eV, Δ = 6.0 eV for M-CoPc-RT and 10Dq = 2.0 eV, Ds = 0.45 eV, Dt =  0.17 eV, Δ = 6.0 eV for M-CoPc-400) give rise to the theoretical spectra well consistent with the experimental results, as plotted in Fig. 4a. The ground states of Co2+ in both M-CoPc-RT and M-CoPc-400 are in 2A1g (dx2-y20dz21dxy2dxz,yz4) symmetry with a total spin of S = 1/2. Figure 4b shows the 3d orbital diagrams of M-CoPc-RT and M-CoPc-400, constructed by the simulation parameters of Co L2,3-edge XAS. Figure 4d shows the Co L2,3-edge XAS spectrum of B-CoPc-RT, which is similar to that of M-CoPc-RT and M-CoPc-400, implying that the Co in B-CoPc-RT has the same ground state (dx2-y20dz21dxy2dxz,yz4) with that of M-CoPc-RT/400, which is also in the low-spin state.

Fig. 4: Analysis of electronic structure.
figure 4

a, d The experimental and simulated cobalt L2,3-edge XAS spectra of M-CoPc-RT/400 and B-CoPc-RT/400, respectively. b, e 3d orbital diagrams of M-CoPc-RT/400 and B-CoPc-RT/400, respectively. c, f Schematic showing the effect of thermal treatment on the coordination environment of M-CoPc-RT/400 and B-CoPc-RT/400, respectively.

While, after heat treatment, the line shape of the L2,3-edge of Co2+ in B-CoPc-400 changed dramatically: L2,3-edge shifted to lower binding energies and the contribution of transition from 2p to 3dz2 orbitals significantly decreased, similar to the high-spin Co2+ in YBaCo3AlO7 with Td symmetry40. The theoretical spectra (Fig. 4d) were also calculated with the parameters (10Dq = 1.3 eV, Ds = 0.36 eV, Dt = 0.10 eV, Δ = 6.0 eV for B-CoPc-RT and 10Dq = 0.1 eV, Ds = 0.09 eV, Dt = −0.01 eV, Δ = 6.0 eV for B-CoPc-400), and the results (Fig. 4d and Fig. S17c, d) reveal that the Co in B-CoPc-RT was still at the ground state in 2A1g symmetry (dx2-y20dz21dxy2dxz,yz4), however, with severe distortion of coordination environment around Co after heat treatment, the 10Dq and Δeg reduced from 1.3 eV and 1.85 eV for B-CoPc-RT to 0.1 eV and 0.3 eV for B-CoPc-400 (Fig. S18b). As a result, the ground state of electron configuration changed from low-spin (b1g (dx2-y20), a1g (dz21), eg (dxz2, dyz2), b2g (dxy2)) to high-spin ((b1g (dx2-y21), eg (dxz1, dyz1), a1g (dz22), b2g (dxy2)) as shown in Fig. S18b and Fig. 4e, corresponding to a symmetry change from planar D4h to distorted D4h symmetry of Co2+. As shown in Fig. 4e, the transition from LS to HS of the Co center, accompanied with a decreased a1g (dz22) and an increased eg (dxz1, dyz1) orbital level would affect the adsorption of CO molecule on the Co center as well as the charge distribution in *CO.

Electron paramagnetic resonance (EPR) measurements were further conducted to obtain information on the unpaired electron of Co in the catalyst (Fig. S19). The EPR spectrum of M-CoPc-RT and B-CoPc-RT show a strong signal at g = 2.71, which is a typical characteristic signal of low-spin Co2+, suggesting that unpaired electron exists in M-CoPc-RT and M-CoPc-RT. In addition, this signal still exists when M-CoPc-RT is annealed at 400 °C to obtain M-CoPc-400, indicating that Co in M-CoPc-400 is still at the low-spin state. However, when B-CoPc-RT is annealed at 400 °C to obtain B-CoPc-400, the signal at g = 2.71 reduces, and at the same time, a new signal appears at g = 2.00, which is a typical high-spin Co2+ signal. The EPR results suggest that the 400 °C treatment of B-CoPc-RT changes its spin structure from low spin to high spin, in agreement with the L2,3-edge analysis.

Theoretical calculation

DFT calculation was employed to study the influence of structure distortion of CoPc on spin state of Co2+. As shown in Fig. S20a, it is found that for planar CoPc molecule, the low spin state of CoPc (LS-Co2+) is more stable than the high spin state (HS-Co2+). However, if a distortion is applied (rotation degree from 0o to 30o), the energy of HS-Co significantly decreases, both HS and LS states become energetically favorable, which indicates that the CoPc can be easily toggled back and forth between the two states.

To further clarify the effect of Co spin state on CORR performance, DFT calculation was performed to compare the reaction free energy of CORR on CoPc and distorted CoPc. Figure S20b displays the optimized structure of LS-Co2+ and HS-Co2+ with different rotation angles. Taking CoPc with rotation angles of 0° and 15° as a representative, there are two possible mechanisms of CORR to methanol over CoPc with rotation angles of 15° (LS-Co2+-(15°) and HS-Co2+-(15°)). As shown in Fig. 5a, compared to HS-Co2+-15°, the rate determining step of CORR over LS-Co2+-(15°) is the desorption of *CH3OH corresponding to ΔG of 0.96 eV, 0.53 eV higher than that over HS-Co2+-15°, indicating energetically favorable H-Co2+-(15°) catalytic center for CORR to methanol. CO adsorption on LS-Co2+-0° is slightly energetically more favorable than that on LS-Co2+-15° and HS-Co2+-15°, while the CORR over LS-Co2+-0° is limited by the step of *CO and *OCH2 hydrogenation, which become energetically favorable over HS-Co2+-15°. The rate-determining step (RDS) of CORR to CH3OH changes from hydrogenation of *OCH2 on LS-Co2+-0° to desorption of *CH3OH on HS-Co2+-15°, corresponding to an energy requirement decrease by 0.32 eV (from 0.75 to 0.43 eV), in accordance with the operando ATR-SEIRAS and KIE results. Moreover, the latter corresponds to physical desorption without electron transfer, which is more likely to occur and can be achieved by many thermodynamic methods such as heating and stirring. The charge density difference calculations and corresponding charge transfer analysis (Fig. 5b) show that the net charge transfer (nCT) from CO to Co2+ is 0.23e and 0.12e for LS-Co2+-0° and HS-Co2+-15°, respectively, and similar charge density distribution trend is observed for *CH2O intermediate (Fig. S21). The nCT for CO adsorbed on Co site is the consonant contributions of both the σ and π components. Based on theoretical study41, CO adsorption on Co site of LS/HS-CoPc is a two-step process, beginning with donation of electrons from the 5σ orbital (HOMO: the highest occupied molecular orbital) of CO to the 3dz2 orbital of Co, followed by back electron donation from the Co dxz/dyz orbitals to the 2π* orbital (LUMO: the lowest unoccupied molecular orbital) of CO.

Fig. 5: CORR mechanism.
figure 5

a Free energy diagram of CORR over LS-Co2+-(0 and 15°) and HS-Co2+-(15°). b The charge density distributions of LS-Co2+-CO and HS-Co2+-CO. Blue, yellow, pink and gray balls represent N, C, Co and O atoms, respectively, and the faint yellow and cyan regions refer to the increased and decreased charge density. Interactions between CO molecular frontier orbitals (5σ and 2π*) and the 3d orbital of c LS-Co2+ and d HS-Co2+ site. Dot in black of c and d is the electron from electrode under cathodic bias. e Schematic of σ and π-donation bonds between CO and 3d orbital of LS-Co2+ and HS-Co2+. The size of the arrow indicates the binding strength of 3dz2-5σ (in red color) and 3dxz/dyz-2π* (in green color), which also indicates the magnitude of the electron migration.

The interactions between CO molecular frontier orbitals (5σ and 2π*) and the 3d orbitals of LS-Co2+-0° and HS-Co2+-15° are displayed in Fig. 5c and d, respectively. As compared to LS-Co2+-0°, 3dz2 orbital of HS-Co2+-15° is fully occupied, resulting in weakened CO adsorption over HS-Co2+, consistent with the DFT calculation. On the other hand, compared to LS-Co2+-0°, the two unpaired electrons in 3dxz/dyz orbitals of HS-Co2+-15° are more active than those in 3dxz/dyz orbitals of LS-Co2+-0°, which can easily back-donate to the 2π* of CO (p orbital of C and O) as formation of HS-Co2+-15°-CO. Moreover, due to anti-bonding feature of the 3dxz/dyz-2π* orbital, the 3dxz/dyz-2π* bonding will weaken the C–O bond in *CO, which can facilitate the following step of *CO hydrogenation. Figure 5c, d shows the schematic of σ and π-donation bonds between CO and 3d orbitals of LS-Co2+-0° and HS-Co2+-15°; the weaker 3dz2-5σ and more electron transfer from Co to *CO via π back-donation for HS-Co2+-15° lead to smaller nCT from CO to Co site, enabling more electron accumulation on the 2π* orbital of *CO, which will effectively promote the hydrogenation of CO reduction intermediates ca. *CO (similar to *OCH2) in CORR. Once *CO formation on Co sites occurs, under cathodic bias, the electrons from the electrode will fill into the highest unpaired orbitals of Co sites, 3dz2 and 3dxz/dyz for LS-Co2+-0° and HS-Co2+-15°, respectively. In the case of LS-Co2+-0°-CO, the electrons from the electrode will fill in the anti-bonding orbital of 3dz2-5σ bond, resulting in weakened *CO adsorption strength, unfavorable for *CO reduction. While in the case of HS-Co2+-15°-CO, the electrons from the electrode will fill in the bonding orbital of 3dxz/dyz-2π* bond, weakening the C–O bond in *CO, which can effectively promote *CO hydrogenation. In addition, DFT calculations also show that H+ adsorption on Co2+ of HS-Co2+-15° requires an energy of 0.68 eV, much larger than that over LS-Co2+-0° (0.44 eV) (Fig. S22), which shall further benefit CORR against hydrogen evolution reaction (HER).


To sum up, L-edge XAS analysis on a model single-Co-atom catalyst verified the transformation of single atomic cobalt center from LS (S = 1/2) to HS (S = 3/2) after thermal treatment, which greatly promoted CORR performance. Combining operando ATR-SEIRAS, KIE experiments and DFT calculations, it showed that the change of spin state of single atomic cobalt center (II) from LS (S = 1/2) to HS (S = 3/2) resulted in a change of the CORR RDS from *CO/*CHxO hydrogenation to methanol desorption, corresponding to a decrease of the RDS energy barrier by 0.32 eV from 0.75 to 0.43 eV. Our work indicates that the chemical bonding of CO molecule on single atom catalyst determined by the specific orbital characters of d-bands is critical for the hydrogenation process in CORR, which provides valuable information for understanding the electrochemical catalytic reaction at the molecular level.



Cobalt phthalocyanine (M-CoPc) and binuclear cobalt phthalocyanine (D-CoPc) were purchased from Sigma-Aldrich and used directly without further purification. Melamine and L-alanine were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd., Ethanol, dimethylformamide, and hydrochloric acid were obtained from Alfa Aesar (China) Chemicals Co., Ltd. NafionTM 115 solution (5 wt.% in lower aliphatic alcohols and water) was purchased from Sigma-Aldrich and used directly. Carbon black (XFP05 7440-74-0) was purchased from XFNANO Materials Tech Co., Ltd. Deionized water from Millipore Q water purification system was used to prepare all solutions.

Synthesis of nitrogen-doped carbon (NC)

In a typical synthesis, a mixture of melamine (C3H6N6) (8 g) and L-alanine (C3H7NO2) (1.5 g) was first ground into a homogeneous precursor using zirconia ball in a nylon jar. Subsequently, the fine powder mixture underwent a two-stage pyrolysis and carbonization process (first stage: from 25 to 600 °C at a ramping rate of 3 °C/min, maintain at 600 °C for 2 h; second stage: from 600 to 900 °C at a ramping rate of 2 °C/min, maintain at 900 °C for 1 h) in a tubular furnace (Carbolite, UK) in argon atmosphere. After cooling down to room temperature, the product was successively leached at 80 °C in 1 M HCl for 12 h. Afterwards, the sample was heated again at 850 °C in Ar for 1 h to recover the crystallinity of carbon.

Synthesis of M-CoPc-RT and B-CoPc-RT

In a typical experiment, 100 mg of NC was dispersed in 100 mL of DMF under sonication for 1 h, followed by adding 5 mg of M-CoPc dispersed in DMF. After 30 min of sonication, the mixture was stirred for 48 h at room temperature. The precipitate was then collected by vacuum filtration and sufficiently washed by DMF and ethanol until colorless. Finally, the precipitate was lyophilized to yield the M-CoPc-RT. B-CoPc-RT was prepared using the same method by replacing M-CoPc with B-CoPc.

Synthesis of M-CoPc-400 and B-CoPc-400

The as-prepared M-CoPc-RT/B-CoPc-RT was thermally treated at 400 °C for 2 h in a tubular furnace in argon atmosphere to prepare M-CoPc-400/B-CoPc-400.


X-ray diffraction (XRD) patterns were recorded on a Bruker D8 diffractometer in the 2θ range from 10 to 80° using Cu Kα radiation (λ = 1.5406 Å). Nitrogen adsorption-desorption isotherms were measured at 77 K on a Micromeritics Tristar 2420 analyzer (USA). Before measurements, the samples were degassed in vacuum at 190 °C for 10 h. The Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods were used to calculate the specific surface area, pore volume, and pore size distribution, respectively. Field-emission scanning electron microscopy (FE-SEM) images were taken on a ZEISS Merlin Compact SEM equipped with energy-dispersive X-ray spectroscopy (INCAPentalFETx3 Oxford EDS). Sub-ångström-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) characterization was performed on a JEOL JEMARM200F STEM/TEM with a guaranteed resolution of 0.08 nm. Chemical states and composition of the samples were analyzed by X-ray photoelectron spectroscopy (XPS) on an ESCALAB 250 photoelectron spectrometer (Thermo Fisher Scientific) using a monochromatic Al Kα X-ray beam (1486.6 eV). All binding energies were referenced to the C 1s peak (284.6 eV). Inductively coupled plasma atomic emission spectroscopy (ICP-AES) measurements were conducted on Atomscan Advantage, Thermo Jarrell Ash, USA.

Electrochemical measurements

All electrochemical measurements remain consistent with our previous methods20,28. All electrochemical measurements were carried out at ambient temperature and pressure on a rotating disc electrode system (Pine Inc.) by a CHI 760e potentiostat. A three-electrode cell configuration was employed with a working electrode of glassy carbon rotating disc electrode (RDE) of 5 mm diameter, a counter electrode of graphite rod (5 mm diameter), and a saturated calomel reference electrode (SCE). The saturated calomel reference electrode (\({E}_{{RHE}}={E}_{{SCE}}+0.0591\times {pH}+0.244\)) was calibrated before the electrochemical measurement (Figure S23). Figure S24 shows the ohmic drop during CORR. To prepare the catalyst ink, 5 mg of catalyst and 25 μL of 5 wt.% NafionTM solution (DuPont) were introduced into 975 μL of water-isopropanol solution with equal volume of water and isopropanol and sonicated for 3 h. An aliquot of 6 μL of the catalyst ink was applied onto a glassy carbon RDE and allowed to dry naturally in air, giving a catalyst loading of 0.15 mg/cm2. A CO-saturated electrolyte was prepared by purging CO (99.99%) into 0.5 M KOH aqueous solution for 20 min, and a flow (100 cc/min) of CO was maintained over the electrolyte throughout the electrochemical measurements. For comparison, CV measurements were also performed in Ar (99.99%)-saturated electrolyte. All potentials were calculated with respect to the reversible hydrogen electrode (RHE) scale according to the Nernst equation. The pH was determined by a pH meter (S220 SevenCompact™ pH/Ion). The linear sweep voltammograms (LSV) were collected at a scan rate of 5 mV/s. All LSV data were corrected for the Ohmic drop. All current densities were normalized to the geometric area of the electrode.

Electrochemical flow cell for CO reduction

Electrochemical CO reduction was carried out in a flow cell. The windows for electrolysis were set to 2 cm × 2 cm. Each chamber has an inlet and an outlet for electrolyte, and an RHE reference electrode was placed in the catholyte chamber. The catalyst ink was prepared by mixing 10 mg of catalyst, 10 mL of ethanol, and 200 μL of a Nafion perfluorinated resin solution. Then, catalysts were air-brushed onto 2 × 2 cm2 gas diffusion layer (Fuel Cell Store) electrodes with mass loading of 1.0 mg/cm2, and used as the cathode. A 2 × 2 cm2 Ni foam was used as a counter electrode for oxygen evolution reaction. An anion exchange membrane (Dioxide Materials) was used to separate the cathode and anode chambers. 0.5 M KOH solution was used as the electrolytes. The catholyte and anolyte were cycled at a flow rate of 80 mL min−1 by using a peristaltic pump. The gas inlet and outlet on the cathode side were linked to a CO gas-flow meter (5 sccm) and a GC, respectively. The gaseous products (i.e., H2 and CH4) were quantified by a gas chromatography (GC-2014, SHIMADZU) equipped with a flame ionization detector (FID) for CH4 and a thermal conductivity detector (TCD) for H2 quantification. Ultra-pure helium (He, 99.9999%) was used as the carrier gas. The flow rate of CO was controlled at 5 sccm/min at the inlet of electrochemical cell by a standard series mass flow controller (Alicat Scientific mc-5 sccm) and measured by a flow meter (Thermo Scientific GFM Pro) at the exit of the electrochemical cell. GC was calibrated using standard gas mixtures under standard conditions (1 atm and 298 K). The liquid product (CH3OH) was characterized by hydrogen nuclear magnetic resonance (1H NMR) after the entire electrochemical measurement on a Bruker Avance III 600 MHz nuclear magnetic resonance spectrometer (Germany) with dimethyl sulfoxide (DMSO) as the internal standard.

Faradaic efficiency (FE) quantification

Based on the definition of Faradaic efficiency:


where i represents CH3OH, H2, or CH4.

\({Q}_{i}\) and \({Q}_{{total}}\) can be obtained from the following equations:

$${{FE}}_{i}={Z}_{i}\times F\times {N}_{i}$$
$${Q}_{{total}}=I\times t$$

Based on the GC data and ideal gas law:

$${N}_{i}={N}_{{total}}\times {V}_{i}$$
$${N}_{{total}}=\frac{{P}_{0}\times {V}_{0}}{R\times {T}_{0}}$$
$${V}_{0}=G\times t$$

where G is the volumetric flow rate.

Thus, Eq. (1) can be written as:

$${{FE}}_{i}=\frac{{z}_{i}\times {v}_{i}\times G\times F\times {P}_{0}}{I\times R\times {T}_{0}\times 60000}$$

where Ni: mole of product i in the GC sampling loop; Ntotal: mole of all gases in the GC sampling loop,V0: volume of the GC sampling loop,t: time for gas to fill the GC sampling loop, I: the average current in a period (t) of electrocatalysis:

$$I=\frac{{\int }_{0}^{t}I\left(t\right){dt}}{t}$$

Vi: the volume ratio of product i in the GC sampling loop, Zi: number of electrons required to produce an i molecule, which is 4, 2, and 6 for CH3OH, H2, and CH4, respectively, F: Faradaic constant (96485 C/mol),P0: atmospheric pressure 1.013 × 105 Pa, T: reaction temperature at 298 K,R: ideal gas constant, 8.314 J·mol·K–1. Turnover frequency (TOF) of methanol was calculated as follows:

$${{{\rm{TOF}}}}=\frac{{I}_{{product}}/{NF}}{{m}_{{cat}}\times w/{M}_{{metal}}}\times 3600$$

Iproduct: partial current for methanol, N: number of electrons required to produce a methanol molecule, which is 4, F: Faradaic constant (96485 C/mol), mcat: mass of catalyst on the electrode, g, w: metal loading in the catalyst based on ICP-AES results.

X-ray absorption spectroscopy

X-ray absorption spectroscopy (XAS) was collected by employing synchrotron radiation light source at BL14W1 beamline of Shanghai Synchrotron Radiation Facility (SSRF) at room temperature. Energy calibration was performed with a Co foil standard by shifting all spectra to a glitch in the incident intensity. Fluorescence spectra were recorded using a seven-element Ge solid state detector. The XAFS data were recorded in fluorescence excitation mode using a Lytle detector, and the spectra of all references were collected in the transmission mode. All samples were pelletized as disks of 13 mm diameter and 1 mm thickness using boron nitride as the binder. The acquired XAFS data were analyzed by Athena and Artemis software, according to the standard procedures. The k3-weighted EXAFS spectra were obtained by subtracting the post-edge background from the overall absorption and then normalized with respect to the edge-jump step. Subsequently, χ (k) data of Co K-edge in the k-space from 2.7 to 11.1 Å−1 were Fourier transformed to real (R) space using a hanning windows (dk = 1.0 Å−1) to separate the EXAFS contributions from different coordination shells. To obtain the quantitative structural parameters around Co atoms, least-squares curve parameter fitting was performed. Effective scattering amplitudes and phase-shifts for the M-CoPc-RT, M-CoPc-400, B-CoPc-RT, and B-CoPc-400 were calculated using the ab initio code FEFF 8.242. The amplitude reduction factor S02 was obtained by fitting Co foil and Co3O4. In the subsequent fitting of M-CoPc-RT, M-CoPc-400, B-CoPc-RT and B-CoPc-400, the coordination number N, interatomic distance R, Debye-Waller factor σ2 and the edge-energy shift ΔE0 were allowed to run freely.

Simulation method of Co L-edge

The simulation of Co L2,3-edge XAS spectra were performed based on the configuration interaction cluster model38, Hartree-Fock estimates of the radial part of the Coulomb interaction in terms of Slater integrals Fk and Gk and the spin-orbit coupling parameters ζ(3d) and ζ(2p) using the atomic theory developed by Cowan43. The reduction of 70% for Slater integrals was applied to account for the over-estimation of electron-electron repulsion found in the calculations of free ion. The hybridization of Co 3d orbit and N 2p orbit was treated using a charge-transfer model, in which 3dn and 3dn+1L configuration were involved. The energy difference of these configurations is defined using the charge-transfer energy Δ and the 3d-3d Coulomb interaction Udd (6.5 eV). The ground states φ can be written as φ = αi|3dn〉 + βi|3dn+1L〉.


The attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) experiments were performed on a Nicolet iS50 FT-IR spectrometer equipped with a MCT detector cooled with liquid nitrogen and PIKE VeeMAX III variable angle ATR sampling accessory.

Theoretical methods and computational details

All DFT calculations were performed in the standard Gaussian 16 software package44. Following which, the calculation of geometry optimization was carried out by using the PBE-D3(BJ)45,46 (including London-dispersion correction) function with the 6–31 G(d) basis set for all atoms. The vibrational frequency calculation was subsequently carried out at the same level as geometry optimization for two purposes: (1) characterizing the nature of the stationary point that all frequencies of local minimum is positive; (2) obtaining the thermodynamic quantities of the studied species at 298 K and 0.1 MPa such as relative Gibbs free-energies. In addition, the single-point energy of studied stationary point was calculated by using the PBE-D3(BJ) function with the 6–311 G(d,p) basis set for all atoms.