Introduction

Hydrogen peroxide (H2O2) as one of the 100 most important chemicals in the world is widely used for various chemical and biological processes, including chemical and pharmaceutical production, environmental protection, and energy storage and conversion1,2. Currently, large-scale infrastructures and tedious multi-step processes are used for industrial production of H2O2 at a high temperature and pressure, thereby cost-intensive and rich with sewage by-product3,4,5,6. Thus, cost-effective and eco-friendly production of H2O2 is highly demanded7,8,9. In this context, the electrosynthesis of H2O2 through oxygen reduction reaction (ORR) via a two-electron pathway has recently attracted significant attention10,11,12,13,14. In particular, considerable effort has been made to enhance the activity and selectivity of electrocatalysts for the electrochemical production of H2O215,16. However, the sluggish reactivity and low selectivity are still two obstacles for efficient electrosynthesis of H2O2 via the two-electron pathway of oxygen reduction.

Owing to their low coordination environment and high catalytic performance, carbon-based single-metal-atom catalysts (SACs) have recently attracted much attention for electrocatalysis17,18,19,20,21. In most SACs, however, single atoms are anchored on an N-doped carbon support to prevent migration of the metal atoms22,23, which usually lead to the four-electron pathway toward ORR24,25. Therefore, they are not ideal for electrosynthesis of H2O2, which requires a two-electron pathway toward ORR (i.e., 2e ORR)26,27,28,29,30. Compared to N-doped carbon materials, several oxygen functional groups, including COOH and C-O-C, in oxidized carbon materials have been proven to act as the defect-type less active sites for ORR via the two-electron pathway31. So, oxidized carbon materials can show good activity and selectivity for electrosynthesis of H2O2 by reducing O2 via a 2e ORR31. These studies prompt us to explore possible synergistic effects by anchoring SACs on oxidized carbon supports to boost the electrochemical production of H2O2 via 2e ORR.

Herein, we report an innovative approach to a class of SACs with single Co atoms anchored on graphene oxide of Co-O-C active centers (i.e., Co-O-C@GO, designed as Co1@GO for simplicity) with a high Co mass loading up to 1.8% with atomic dispersion. These single Co atoms form a Co-O bond with a bond length of 1.47 Å, resulting in a sufficiently high oxidation state for the Co sites for oxygen reduction via a two-electron pathway. As a result, the newly developed Co1@GO exhibited an excellent 2e ORR activity with an onset potential of 0.91 V and a high selectivity of 81.4% for H2O2 production. Of particular importance, the H2O2 yield of Co1@GO is 1.0 mg cm−2 h−1, which is a 19-fold enhancement compared with GO and about 2-fold improvement compared with Co nanoparticles/GO. Our combined experimental observations and density functional theory (DFT) calculations indicate that the high reactivity and selectivity of Co1@GO for electrogeneration of H2O2 is arising from the Co-O-C active center with synergistic effects between the O-bonded single Co atoms and adjacent oxygen functional groups (C-O bond) of the GO.

Results and discussion

The synthesis and structural characterization of Co1@GO

Co1@GO was synthesized through acid leaching of Co nanoparticles (Co NPs/GO) (Figs. 1 and 2a and Supplementary Fig. 1), in which Co nanoparticles (~1.9 nm) were acid etched into Co clusters (Co clusters/GO) (Supplementary Fig. 2) and then etched into single Co atoms and stabilized by oxygen functional groups of the GO (Supplementary Fig. 3), leading to a stable Co1@GO electrocatalyst. It can be seen that the intermediate Co clusters/GO possesses Co clusters and single Co atom (Co1) (Supplementary Fig. 2c), though dominated by the Co clusters. The presence of Co1 suggested that part of Co clusters were dissolved by acid leaching to form single Co atoms, which were captured by the oxygen functional group of GO via Co-O bonding (Fig. 1). Further acid etching removed the Co clusters. High-angle annular dark field scanning transmission electron microscope images of Co1@GO are given in Fig. 2b, c and Supplementary Fig. 3, which show no Co NPs or Co clusters, but single Co atoms homogeneously interdispersed onto the GO support. Figure 2a, b reveals that Co clusters dissolved by acid leaching to form single Co atoms while Fig. 2c suggests that the resultant single Co atoms are stabilized on the GO support by Co-O bonding (vide infra). The energy-dispersive X-ray spectroscopy result in Supplementary Fig. S3c shows the presence of Co element in GO, further referring to the stability of single Co atoms anchored on the GO support. The X-ray diffraction (XRD) pattern of Co1@GO does not reveal additional diffraction peaks of Co apart from the typical broad peak centered around 24.9° from GO (Supplementary Fig. 4), indicating the successful transformation of the Co NPs to single Co atoms in Co1@GO (cf. Figs. 1 and 2 and Supplementary Figs. 1 and 2). The mass loading of Co in Co1@GO was determined by inductively coupled plasma–optical emission spectroscopic measurements to be about 1.8%. To investigate the GO structure of Co1@GO, we first performed the Raman spectroscopic measurements. As shown in Supplementary Fig. 5, the D band and G band of Co1@GO are at 1342.33 and 1587.49 cm−1, respectively, similar to those for GO, which indicates that acid leaching did not cause an obvious change of the GO structure in Co1@GO32,33. The Co 2p XPS spectrum of Co NPs/GO (Supplementary Fig. 6) could be separated into Co2+ (780.9 eV) to Co3+ (782.7 eV). The Co2+ (780.9 eV) state indicates that the surface of Co NPs has been oxidized into CoO. The Co2+ (781.1 eV) state of Co clusters/GO shows a positive shift of 0.2 eV compared with that of Co NPs/GO, which suggests that, after the HCl etching, the surface of Co clusters adsorbs hydroxyl (-OH). It could be confirmed that the C-O state of Co clusters/GO (533.5 eV) shows a down-shift of 0.2 eV compared with Co NPs/GO (C-O 1s, 533.7 eV), indicating that it has -OH on the surface of Co clusters and GO. The Co 2p XPS spectra of Co1@GO (Supplementary Fig. 6c) indicate that the Co atom site in Co1@GO is at Co2+ to Co3+, suggesting the presence of Co-O bond. The high-resolution O1s XPS spectrum of Co1@GO (Supplementary Fig. 6f) shows three peaks: C-O-C (533.6 eV), C-O-H (532.5 eV) and C=O (531.1 eV). The C1s XPS spectra of these three samples also demonstrate the presence of C-O-C, C-O-H and C=O (Supplementary Fig. 6g–i). In addition, the ratio of C-O-C:C-O-H:C=O in Co NPs/GO, Co clusters/GO, and Co1@GO are 0.8:0.6:1, 1.7:1.4:1, and 2.3:1.5:1, respectively. The high content of C-O-C in Co1@GO suggests the high selectivity for H2O231. Interestingly, the C-O-C bind energy in O1s peaks of Co1@GO is the lowest among these three samples, which indicated that the high oxide state of single-atom Co may contribute electron to the C-O, leading to the formation of a Co-O-C electron conjugated complex. The as-measured oxygen contents of Co NPs/GO, Co clusters/GO, and Co1@GO are 12.3%, 14.5% and 15.3%, respectively. The presence of the Co-O-C electron conjugated complex and the high oxygen content in Co1@GO suggests the high performance of two-electron pathway for ORR.

Fig. 1: Schematic of synthesis of Co1@GO.
figure 1

Co nanoparticles were firstly prepared via reducing CoCl2 in 5 vol% H2 in N2 (Co NPs/GO). Co NPs/GO was leached to form Co1@GO.

Full size image
Fig. 2: Electron microscopy images.
figure 2

HAADF-STEM images of a Co NPs/GO and b, c Co1@GO, with the red cycle areas of Co1.

Full size image

The presence of single Co atom on GO was further confirmed by X-ray absorption spectroscopic (XAS) measurements. The X-ray absorption near-edge structure spectra in Fig. 3a shows that the absorption edge intensity of Co1@GO is higher than that of Co foil and close to those of Co3O4 and CoO. In addition, the absorption edge peak position of Co1@GO is positively shifted compared with those of Co3O4 and CoO. These results indicate that Co is atomically dispersed in the Co1@GO, and that the Co center is coordinated with O atoms by interacting with oxygen functional groups in the GO support to form the Co-O bond34,35,36. The Fourier-transformed k3-weighted extended X-ray absorption fine structure spectra of Co1@GO, Co3O4, CoO, and Co foil are reproduced in Fig. 3b, which clearly show no notable peak from 2.0 to 3.0 Å for the Co-Co bond in Co1@GO, but a prominent peak at 1.47 Å for the Co-O bond37. Clearly, therefore, the Co NPs have been dissolved and remained as single Co atoms to coordinate with O (Co-O-C electron conjugated complex) from GO in the Co1@GO (Fig. 1 and Supplementary Fig. 1). Furthermore, the peak position at 1.47 Å is in good agreement with that of Co3O4, suggesting that the Co atom sites in Co1@GO are at a high oxidation state38, from +2 to +3, which should show higher activity than that of a low oxidation state39. This result is in agreement with the XPS results, suggesting high electrochemical performance for Co1@GO (vide infra). The fitting results of the R space in Fig. 3c and the k-space in Supplementary Fig. 7 are consistent with the Co-O3-C structure in DFT calculation results (vide infra), and also indicate the existence of the Co-O-C electron conjugated complex in the Co1@GO.

Fig. 3: X-ray absorption results of Co1@GO.
figure 3

a XANES spectra and b Fourier transforms (FTs) of the Co K-edge of Co1@GO, Co foil, CoO and Co3O4. c Corresponding fitting of the EXAFS spectra of Co1@GO in R space.

Full size image

The electrosynthesis performance of H2O2

To investigate the electrochemical performance of single Co atom sites and the possible influence of the oxygen functional group in Co1@GO, we performed the ORR measurements on Co1@GO, Co NPs/GO and GO in 0.1 M KOH (Fig. 4) using a rotating ring-disk electrode (RRDE). Figure 4a reproduces the linear sweep voltammetry (LSV) curves for Co1@GO, Co NPs/GO and GO in the 0.1 M KOH solution, in which the disk electrode detected the oxygen reduction currents while the Pt ring electrode monitored the yield of H2O2. As can be seen, Co1@GO possesses a higher ORR reactivity with an onset potential of 0.91 V versus a reversible hydrogen electrode (RHE) and a diffusion-limited current of ~0.5 mA, which are much higher than the corresponding values of 0.68 V and 0.028 mA for GO, 0.81 V and 0.34 mA for Co NPs/GO, and 0.76 V and 0.19 mA for Co clusters/GO (Supplementary Figs. 8 and 9). The high onset potential of Co1@GO is attributed to the synergistic effect between the O-bonded single Co atoms and adjacent oxygen functional groups (C-O bond) of the GO, which could decrease the free energy (−0.06 eV) of OOH* (Fig. 5d). The low free energy could improve the reactivity of ORR performance, thus the enhanced onset potential. The production of H2O2 was quantified at 1.2 V by the Pt ring electrode, which avoided the ORR current on the ring with H2O2 oxidation only. Besides, Fig. 4a shows a much higher ring current for Co1@GO than those of GO and Co NPs/GO, indicating that the presence of single Co atoms also enhanced the yield of H2O2. Figure 4b, c shows the selectivity and the yield of H2O2, respectively, for Co1@GO, GO and Co NPs/GO. As can be seen, the H2O2 selectivity of Co1@GO, GO and Co NPs/GO are 81.4%, 76.3%, and 68.7% at 0.6 V, respectively. The observed relatively small change in the selectivity for these samples indicates that the C-O may be the important factor for the selectivity. In addition, the Co1@GO shows a H2O2 yield of 1.0 mg cm−2 h−1 at 0.50 V, which is 19-fold enhancements compared with GO (0.052 mg cm−2 h−1) and about 2-fold improvement compared with Co NPs/GO (0.55 mg cm−2 h−1). The observed big change in the H2O2 yield among these samples suggests that the Co-O is the main reason for the activity for the 2e ORR reactivity. These results indicate that single Co atoms on the GO support could enhance the 2e ORR reactivity and improve the H2O2 selectivity, simultaneously.

Fig. 4: Oxygen reduction performance of Co1@GO and GO.
figure 4

a, b Oxygen reduction performance of Co1@GO and GO in 0.1 M KOH at 1600 rpm (solid lines) and H2O2 product currents (ring electrode, dashed lines). c Electrosynthesis of H2O2 on Co1@GO and GO in O2‐saturated 0.10 M KOH electrolyte. d, e Stability tests of Co1@GO and GO at 0.1 M KOH. f Comparison of the onset potential and selectivity for electrosynthesis H2O2 on Co1@GO and other reported catalysts at 0.6 V (RHE).

Full size image
Fig. 5: Mechanistic investigation through DFT.
figure 5

a The model of Co-O3-C structure used for DFT calculation. The brown, red and blue balls represent C, O and Co atoms, respectively. b The reaction cycle scheme of the intermediates in four-electron and two-electron ORR process on the Co-O3 catalytic center. c Free energy diagrams of four-electron under equilibrium state (U = 0.41 V), and d free energy diagrams of two-electron ORR under equilibrium state (U = −0.13 V).

Full size image

Importantly, Co1@GO also maintains a high H2O2 selectivity and activity when applied to a two-electrode system. In 0.1 M KOH, Co1@GO delivers a steady-state H2O2 at 0.76 V with corresponding to an H2O2 production rate of ~5.7 mol g−1 h−1 or 28 mol m−2 h−1 (Supplementary Figs. 10 and 11). To our best knowledge, the onset potential and H2O2 production performance of Co1@GO are superior to most reported catalysts, such as Co1@NG(O) (0.42 mol g−1 h−1)40, Fe-CNT (1.6 mol g−1 h−1)41 and Co-POC-O (2.98 mol m−2 h−1)42 (Supplementary Table 1). Figure 4d, e presents the stability results, showing a disk and ring current of 0.32 and 0.024 mA, respectively, for Co1@GO at 36,000 s, which are 40 and 3.6 times of the corresponding date for GO (disk current of 0.008 mA and ring current of 0.0067 mA). Clearly, therefore, Co1@GO is a superb catalyst for the electrochemical generation of H2O2.

Figure 4f shows the onset potential versus the H2O2 selectivity for Co1@GO, along with the corresponding data from various previously reported catalysts for comparison.31,40,41,42,43,44,45,46,47,48,49,50 As can be seen, Co1@GO has the most positive potential with a relatively high H2O2 selectivity. For Co1@GO, the high oxidation state of single Co atoms and oxygen functional group of GO in the Co-O-C electron conjugated complex contributed to the high reactivity and high selectivity for electrosynthesis of H2O2. As shown in Supplementary Fig. 12, Co1@GO, Co NPs/GO, Co clusters/GO, and GO all show electron transfer numbers within the range of 2.3–2.7 over the potential from 0.4 to 0.7 V, indicating a 2e pathway for the oxygen reduction. The Co1@GO maintains the lowest electron transfer numbers, demonstrating again that synergistic effects between the O-bonded single Co atoms and adjacent oxygen functional groups (C-O bond) of the GO in the Co1@GO have significantly improved the kinetics for 2e ORR. Clearly, therefore, anchoring single Co atoms of a high oxidation state onto a GO support could afford a feasible strategy to develop advanced electrocatalysts with excellent reactivity, high selectivity, and good stability toward electrosynthesis of H2O2.

Mechanistic investigation on H2O2 electrogeneration

To obtain a better understanding of the Co-O-C active centers and more insights into the catalytic mechanism of the Co1@GO catalysts, we further performed DFT calculations (Supplementary Methods) to determine the free energy, overpotential, and reaction pathway for ORR on Co1@GO. According to the experimental results for the coordination environment near the Co metal, the Co-O-C electron conjugated complex is detected as the active center in the Co1@GO catalyst. Therefore, we built DFT models with possible Co-O structures (Supplementary Fig. 13 and Fig. 5a). With the DFT calculations, the catalytic center is identified as Co-O3-C structure (Fig. 5a). The formation energy of this structure is −2.78 eV, indicating the thermodynamically spontaneous formation of the structures. The typical ORR cycle of four-electron pathway producing H2O (A-B-C-D) and two-electron pathway producing H2O2 (A-E) on the Co-O3 active center is illustrated in Fig. 4b. The free energy diagrams of four-electron and two-electron ORR for the Co-O3-C structure are shown in Fig. 5c, d, respectively. For four-electron ORR, the reaction center shows a quite low catalytic activity with overpotential of 1.07 V. The rate-limiting step of four-electron ORR is identified to be the final step (OH* desorption), which is caused by the strong absorption of OH*. For two-electron ORR, the free energy change is extremely close to that of the ideal catalyst, and the overpotential is only 0.06 V, showing a high activity of producing H2O2. Thus, from the results of DFT calculations, the two-electron ORR is favored for the identified catalytic structure, consistent with the experimental results of high reactivity and selectivity of Co1@GO for electroreduction of O2 to H2O2 via a 2e pathway.

The free energy diagrams for Co-O4-C structures (Supplementary Fig. 13) were also calculated and shown in Supplementary Fig. 14, and their formation energy and overpotential values considered are listed in Supplementary Tables 2 and 3, respectively. It is noted that the Co-O4-C-(OH)4 structure shown in Supplementary Fig. 13c also has good activity and selectivity of two-electron ORR toward H2O2, with four-electron overpotentials of 0.88 V and two-electron overpotential of 0.01 V, which is another possible active center. The density of states (DOS) and Bader charge distribution of the Co-O3-C structure and other Co-O4 structures are also analyzed and shown in Supplementary Figs. 15 and 16, respectively. In general, the metal ion Co in Co-O4 structures carries a higher charge than that on Co-O3 one, which may lead to stronger adsorption to the intermediates. The high electronic LDOS of the Co atom in Co-O3-C structure is close to the Fermi level, promoting the electron transfer from the active center to the absorbates during the reaction and leading to unique two-electron ORR activity, whereas there is a large bandgap in the Co-O4-C structures, which inhibits the electron transfer from the graphene sheet and consequently depresses the 2 electron ORR. Therefore, Co-O3-C structure is more possibly the active center for H2O2 production via a 2e pathway for ORR.

Conclusion

In summary, we have demonstrated that high oxidation of single Co atoms anchored on a GO support (i.e., Co1@GO) exhibited excellent performance for electrosynthesis of H2O2 from ORR via a 2e pathway. The newly developed Co1@GO possessed a unique novel Co-O-C electron conjugated complex as the active center with synergistic effects between the O-bonded single Co atoms and adjacent oxygen functional groups (C-O bond) of the GO. Consequently, Co1@GO delivered the highest reactivity and selectivity when compared with GO and reported catalysts (cf. Fig. 3) with good stability. DFT calculations confirmed Co-O3-C as the catalytic center in Co1@GO. This work represents a breakthrough in the electrosynthesis of H2O2 by rationally designing and developing highly effective, selective, and stable electrocatalysts with new active centers. The methodology developed in this study offers a novel and efficient approach for developing new electrocatalysts for various electrochemical reactions.

Methods

Synthesis of Co1@GO samples

Co1@GO was prepared through the dispersion of 6.2 mg CoCl2 and 50 mg GO in Millipore water (18.2 MΩ·cm) under ultrasonication for 30 min (Ultrasonic baths, 480 W, Bandelin Electonic). The resultant mixture was freeze-dried (Alpha 1–2 LDplus Entry Freeze Dryer, Refrigeration system 0.43 kW, Martin Christ) for 48 h to remove water, followed by thermal treatment at 300 °C for 2 h in 5 vol% H2 in N2, forming Co nanoparticles (Co NPs/GO). The resultant powder was leached in 2 M HCl for 24 h, forming Co clusters (Co clusters/GO), and then 2 M HNO3 for 24 h to remove Co clusters. Finally, the sample was collected by centrifugation, washed with water and ethanol several times. The electrocatalyst thus produced was designated as Co1@GO. For comparison, GO was prepared through the same procedure without the addition of CoCl2.

Structural characterization

XRD results were recorded by a powder XRD (GBC MMA diffractometer) with Cu Kα radiation at the scan rate of 2° min−1. XPS spectra were recorded on Thermo ESCALAB250i XPS by using Al Kα radiation, which was fixed in an analyzer transmission mode. The XAS results of Co K-edge were collected at an XAS station of the Beijing Synchrotron Radiation Facility. The morphological study of the newly developed catalysts was performed on a scanning TEM (JEOL ARM-200F, 200 keV).

Electrochemical measurements

The electrochemical performance of the resultant catalysts was carried out by using a three-electrode system at an electrochemical workstation (CHI 760 E, CH Instrument, USA). A Pt wire was used as the counter electrode and a saturated calomel electrode was applied as the reference electrode. The working electrode was using a RRDE. The ring electrode was Pt ring with an outside diameter of 0.75 mm and an inside diameter of 6.5 mm; the disk electrode was a glass carbon disk with a diameter of 5.0 mm. The slurry was prepared by adding 5.0 mg catalysts into a mixture of 50 μL Nafion and 950 μL ethanol, and then was sonicated for 1 h to form a homogenous dispersion. 5 μL of the ink was dropped on the disk electrode. The working electrode was ready for testing after the solvent evaporated. The potential was calibrated to a RHE by applying the following equation:

$${{{{{\rm{E}}}}}}\,({{{{{\rm{RHE}}}}}})={{{{{\rm{E}}}}}}({{{{{\rm{SCE}}}}}})+0{{{{{\rm{.224V}}}}}}+0.0592\,{{{{{\rm{pH}}}}}}$$

All the tests were carried out in an O2 saturated 0.1 M KOH solution at room temperature with a rotated rate of 1600 rpm. LSV was performed at a scan speed of 10 mV/s, and the potential of the ring electrode was set at 1.5 V (vs. RHE) to detect the production of H2O2. The selectivity, productivity, and electron transfer number were calculated by the following two equations:

$${{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\;{{{{{\rm{selectivity}}}}}}{:}{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}( \% )=200\times ({i}_{r}/N)/({i}_{d}+{i}_{r}/N)$$
$${{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}\;{{{{{\rm{productivity}}}}}}\,({{{{{\rm{mg}}}}}}\,{{{{{{\rm{cm}}}}}}}^{-2}{{{{{{\rm{h}}}}}}}^{-1})=1/2\times ({i}_{r}/N)\times {M}_{{{{{{{\rm{H}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{2}}\times 3600/({{{{{\rm{F}}}}}}/A)$$
$${{{{{\rm{Electron}}}}}}\,{{{{{\rm{transfer}}}}}}\,{{{{{\rm{number}}}}}}\,(n)=\frac{4\,{i}_{r}}{{i}_{r}+{i}_{d}/N}$$

ir was the ring current and id was the disk current, N is the current collection efficiency of Pt ring electrode (N = 0.26), MH2O2 is the molecular weight of H2O2 (MH2O2 = 34.01 g mol−1), F is Faraday constant (F = 96485.3 C mol−1), A is the area of disk electrode (A = 0.196 cm−2). The collection efficiency of the electrode defined as the fraction of product from the disk to the ring is 0.26.

The stability test of these catalysts was performed by the chronoamperometric at a constant potential of −0.77 V (vs. RHE).

The bulk H2O2 production in 0.1 M KOH of Co1@GO was conducted in an H-cell electrolyzer, with 0.5 mg cm−2 Co1@GO onto a 2 × 2 cm2 Freudenberg GDL electrode (Fuel Cell Store) as the cathode, and Pt foil as the anode for H2O electro-oxidation. A Fumasep FAA-3-PK-130 anion exchange membrane (Fuel Cell Store) was used to separate the chambers. The H2O2 concentration can be quantified by Ce(SO4)2 titration based colorimetric method (2Ce4+ + H2O2 → 2Ce3+ + 2 H+ + O2)31. The H2O2 concentration-absorbance liner was calibrated by adding certain amount of commercial H2O2 solution with 0.5 mM Ce(SO4)2. The Ce4+/H2O2 concentration was measured on a Cary 5000 UV-Vis-NIR spectrometer (Agilent) and the absorption at 320 nm wavelength was applied to determine its concentration.

Computational method

The DFT simulation on the free energies and overpotentials of two-electron ORR and four-electron ORR was carried out by DFT using the Vienna Ab initio simulation package (VASP)51,52, where the soft projector-augmented wave and Perdew–Burke–Ernzerhof exchange correlation were implemented53. The unit cells of the structures used in this study are composed of a single-layer graphene sheet embedded with Co and O atoms with the unit size set to 12.82 Å × 12.34 Å × 20.06 Å. Spin-polarization was considered in all calculations. The cutoff energy was set to 550 eV and the geometries were fully relaxed until the residual force convergence value on each atom being less than 0.01 eV Å−1. The structures were sampled by a 3 × 3 × 1 k-point Monkhorst-Pack k-point mesh. In the calculation of reaction substrate and absorbates, the implicit solvent effect is considered to simulate the aqueous environment by using the VASPsol code. Since Co is 3d-transition metal, the Hubbard U (DFT+U) corrections were also considered in the calculations54.