Introduction

The electrochemical reduction of carbon dioxide (CO2) into valuable chemical feedstocks and fuels, such as ethylene, acetate, and ethanol, offers a sustainable pathway to store renewable electricity while mitigating carbon emissions1,2,3,4. Currently, alkaline CO2 electrolysis enables efficient C-C coupling for C2+ production, but bicarbonate/carbonate formation under high pH conditions (e.g., pH>14) leads to poor CO2 single-pass conversion efficiency (SPCE, <25% for C2 products)5,6. Switching CO2 reduction reaction (CO2RR) from an alkaline to an acidic environment can circumvent the carbonate issue and improve the CO2 utilization; nevertheless, the proton-rich condition makes the competing HER more kinetically favorable7,8. Using a bipolar membrane can improve CO2 utilization and minimize liquid product crossover, but it leads to low energy efficiency due to the excess overpotential to drive water dissociation9,10. To surmount the above issues, a two-step cascade electrolysis process has been proposed, in which CO2 is initially reduced to CO and CO is further converted to C2+ products11,12. Such a process shows a promising application prospect given that the CO2-to-CO conversion can reach a FE greater than 95% and CO reduction reaction (CORR) can be operated in alkaline electrolytes, benefitting C2+ formation13,14,15. However, CORR commonly undergoes low selectivity among different C2+ chemicals. At present, steering CORR to a single principal C2+ product with high selectivity remains a daunting challenge.

Among various C2+ products, acetic acid is an important chemical reagent in the manufacture of textiles, medicines, polymers, and foods, with a large annual production capacity of more than 18 million tons and a market size of $13 billion USD/year16,17. Today, approximately 75% of acetic acid is produced in the chemical industry through the thermal carbonylation of methanol and CO, which, however, results in substantial CO2 emission18,19. Electrochemical CORR using renewable electricity provides an appealing carbon-free approach to produce acetic acid (acetate in alkaline electrolytes)20. Unfortunately, CO-to-acetate selectivity is limited by the competition of ethylene and alcohol formation and the parasitic HER21,22. Previous studies have shown that *CO surface coverage plays a vital role in determining the CORR selectivity, i.e., high *CO coverage favors acetate formation, whereas low *CO coverage leads to ethylene23,24. To achieve high *CO surface coverage, strategies like constructing a copper-palladium (CuPd) intermetallic catalyst25, employing a high-pressure flow reactor24, and creating a Cu catalyst with stacking-fault structure17 have been adopted. These strategies have shown great effectiveness to improve surface *CO coverage, which, however, led to cost issues and process complexity at the same time. Therefore, a simple, low-cost and effective strategy to increase acetate selectivity is highly desirable.

Research on Cu-catalyzed CORR has revealed that the common C2+ intermediate *(OH)C=COH, under the H*-deficient environment, will proceed with *CH2-C=O via *C=C=O, followed by OH- attacking to form acetate25,26. By contrast, under H*-rich environment, it will be transformed to *C=COH, followed by multiple proton-coupled electron transfer (PCET) steps to generate ethylene26. Herein, we report a Ce-SAs modified crystalline-amorphous dual-phase Cu catalyst, which can steer CORR toward acetate under a pulsed electrolysis condition. Comprehensive operando spectroscopic characterizations reveal that Ce-SAs can reduce the electron density of dual-phase Cu, which weakens K+ adsorption on the outer Helmholtz plane (OHP), leading to a low proportion of K·H2O and, therefore a low H* coverage. Moreover, Ce-SAs also attract electrons from crystalline Cu particles, which forms Cu+ species that enhance COatop adsoption for favorable COatop-COatop dimerization. As a result, we achieved CORR toward acetate with a high FE of 71.3 ± 2.1% and a time-averaged acetate current density of 110.6 ± 2.0 mA cm−2. The catalyst also showed no evidence of performance drop after 138 h of operation under a pulsed electrolysis mode.

Results

Synthesis and structure characterizations of catalysts

Ce has been well demonstrated to promote CO2RR performance due to its unique electronic structure arising from the spin-orbit coupling of the 4 f and valence orbitals27,28. The 4f orbital leads to the formation of Ce3+/Ce4+ redox pair, which can stabilize Cu+ active sites during CO2RR29. Furthermore, coupling Ce 4f orbital with the C 2p orbital enhances CO2 and *CO adsorption, which facilitates C2+ formation30. Recently, a Ce-SA alloyed Ru catalyst has been reported to boost alkaline HER in which the oxyphilic Ce SAs could enrich OH- and repel Had to moderately adsorb on the Ru atoms, leading to enhanced HER energetics31. Since low H* coverage and high *CO coverage favors acetate formation, we reasoned that CORR can be steered towards acetate by using Ce-SAs to regulate surface H* and *CO coverage of the Cu-based catalysts.

In our previous work, we reported a crystalline-amorphous dual-phase Cu-based catalyst that exhibited marked CO2-to-C2+ selectivity due to the highly enhanced chloride adsorption at the interface32. Here, based on this unique structure, we developed a Ce-SAs modified dual-phase Cu catalyst, which comprises Ce-SAs dispersed uniformly on crystalline Cu and amorphous Cu oxide matrix (termed Ce-d-Cu hereafter) (Fig. 1a and b, Supplementary Fig. 1). We synthesized this catalyst by adding Ce (NO3)3·6H2O and CuCl2·2H2O in ethylene glycol with tannic acid at room temperature (25 °C) (see Methods for details), in which ethylene glycol was the reaction solvent, while tannic acid could chelate with Cu2+ ions to form amorphous networks, prohibiting complete transformation of Cu2+ to Cu33. We controlled the Ce amount on dual-phase Cu by regulating the molar ratio of Ce (NO3)3·6H2O to CuCl2·2H2O (Supplementary Table 1). An excessive amount of tannic acid, however, leads to Ce-SAs modified amorphous Cu oxide phase without crystalline Cu, which we abbreviated as Ce-a-Cu (Fig. 1c and d, Supplementary Fig. 2 and Table 1). For comparison, dual-phase Cu and amorphous Cu oxide without Ce-SAs incorporation were also prepared, which we abbreviated as d-Cu and a-Cu, respectively (Supplementary Figs. 3 and 4). Moreover, crystalline Cu nanoparticles (termed c-Cu) were prepared by a similar method as a reference (Supplementary Fig. 5).

Fig. 1: Structural characterization of Ce-d-Cu and Ce-a-Cu catalysts.
Fig. 1: Structural characterization of Ce-d-Cu and Ce-a-Cu catalysts.
Full size image

Schematic illustration of Ce-d-Cu (a) and Ce-a-Cu (c). HAADF-STEM images and EDS elementary mapping images of Ce-d-Cu (b) and Ce-a-Cu (d). The orange circles mark the Ce SAs. Ce L3-edge XANES spectra (e), Ce L3-edge Fourier transform-EXAFS spectra (f) and Wavelet transform EXAFS signals (g) of Ce-d-Cu, Ce-a-Cu and reference CeO2. Source data for Fig. 1 are provided as a Source Data file.

Transmission electron microscopy (TEM) reveals that the average particle size of Cu nanoparticles in Ce-d-Cu is 7.4 nm (Supplementary Fig. 1), analogous to that of d-Cu and c-Cu counterparts (Supplementary Figs. 3 and 5). High-resolution TEM (HRTEM) demonstrates the high crystallinity of these Cu nanoparticles with average lattice spacings of 0.209 nm, corresponding to Cu (111) planes (Supplementary Figs. 6–8). No lattice fringes are observed on Ce-a-Cu, a-Cu, and the substrate of Ce-d-Cu and d-Cu, showing their amorphous nature (Supplementary Figs. 2 and 9). Energy-dispersive X-ray spectroscopy (EDS) elemental mapping exhibits that Cu and O are mainly distributed in the crystalline phase and amorphous phase of Ce-d-Cu and d-Cu, respectively, while Ce is uniformly dispersed in both the crystalline phase and amorphous phase of Ce-d-Cu (Supplementary Figs. 10 and 11). For Ce-a-Cu and a-Cu, Cu and O are uniformly distributed in their amorphous phase, and Ce is evenly dispersed on the amorphous phase of Ce-a-Cu (Supplementary Figs. 12 and 13).

Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was performed to observe the dispersion status of Ce in Ce-d-Cu and Ce-a-Cu. Based on the difference in Z contrast, individual Ce atoms can be distinguished from Cu and O atoms since the atomic number of Ce is larger than those of Cu and O elements and is thus brighter in the dark field34. It can be seen that Ce species are atomically dispersed in both Ce-d-Cu and Ce-a-Cu catalysts (Fig. 1b and d, yellow circles, Supplementary Fig. 14). By contrast, no individual Ce species are observed in d-Cu and a-Cu (Supplementary Figs. 9 and 15). The measurement of Ce-Ce distance further reveals that there are no Ce-Ce bonds in both Ce-d-Cu and Ce-a-Cu catalysts (Supplementary Fig. 16). X-ray absorption spectroscopy (XAS) was further employed to probe the local atomic structure around Ce centers. The Cu extended X-ray absorption fine structure (EXAFS) fitting results show that the coordination number (CN) of Cu-Ce is 7.46 and the Cu-Ce bond length is 3.15 Å in Ce-d-Cu, versus those of 7.86 and 3.25 Å in Ce-a-Cu (Supplementary Fig. 17). As Fig. 1e shows, the X-ray absorption energy near-edge structure (XANES) spectra at Ce L3-edge of Ce-d-Cu and Ce-a-Cu exhibit strong single peaks (around 5727 eV), which are different from the CeO2 reference with two well-separated maximum peaks at 5731 and 5737 eV. Fourier-transform EXAFS of CeO2 presents coordination peaks at 3.6 Å and 1.9 Å (Fig. 1f), which are assigned to Ce-Ce and Ce-O bonds, respectively35. By contrast, Ce-d-Cu and Ce-a-Cu only exhibit Ce-O coordination peaks at 2.0 Å (Fig. 1f), suggesting no Ce-Ce bonds in the two catalysts. Additionally, EXAFS wavelet transform (WT) was performed to provide powerful resolution in k space and R space. As displayed in Fig. 1g, the reference CeO2 exhibits intensity maximums at 3.6 Å−1 and 9.2 Å−1, which originated from Ce-O and Ce-Ce contribution, respectively31. As contrast, Ce-d-Cu and Ce-a-Cu only show intensity maximums at 3.6 Å−1. These results ulteriorly validate the presence of single-atom Ce sites in Ce-d-Cu and Ce-a-Cu catalysts.

We employed electron energy-loss spectroscopy (EELS) at the atom level to probe the valence of Cu and Ce sites. The EELS spectra acquired from a Cu particle region of Ce-d-Cu displays the intensity of the Cu L2,3 edge and Ce M4,5 edge (Supplementary Fig. 18), whose white-line ratios of the Ce M4,5 edges and Cu L2,3 can be used to distinguish the oxidation state of Ce and Cu. The EELS spectra yield Ce M5/M4 peak intensity of 1.1, indicative of +3 valence of Ce SAs36. Meanwhile, the intensity ratio of Cu L3/L2 is 2.1, which falls in between Cu0 and Cu+37,38. The increase in the valence of the non-interface region of metallic Cu particles suggests the electron transfer from Cu particles to Ce SAs (Supplementary Fig. 18). This phenomenon was further explored by X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) techniques. The peak area ratio of Cu+/Cu0 2p3/2 and Cu2+ 2p3/2 reveals the (Cu++Cu0) concentration. As shown in Supplementary Fig. 19, the Ce-d-Cu exhibits a (Cu++Cu0)/Cu2+ ratio of 0.46, higher than that of 0.33 of d-Cu. Moreover, Gaussian fitting of Cu LMM Auger spectra of Ce-d-Cu yields a 29.9% Cu+, higher than the 20.8% of d-Cu, whereas the Cu0 content in Ce-d-Cu is 19.0%, lower than the 22.6% of d-Cu (Supplementary Fig. 20). Additionally, the Ce 3d XPS uncovers that Ce3+ constitutes more proportion than Ce4+ in Ce-d-Cu (Supplementary Fig. 21). We further performed density functional theory (DFT) calculations to elucidate the origin of Cu+ in Ce-d-Cu. Charge-density difference calculations reveal that crystalline-phase Cu loses electrons while the Ce SAs gain electrons (Supplementary Figs. 22 and 23). These results together confirm the electron transfer from the crystalline Cu particles to the Ce SAs, leading to the formation of Cu+ species.

CORR performance under pulsed electrolysis

Recently, pulsed electrolysis has emerged as an intriguing technique to improve C2+ selectivity and catalyst lifetime during CO2RR39. It has been reported to regenerate Cu+ and enrich electrolyte anions (e.g. OH-, Cl-) due to the periodic switch between the anodic potential and cathodic potential, thereby enhancing *CO adsorption and promoting C-C coupling for C2+ formation40,41. Moreover, the dynamic potential change during pulsed electrolysis can reduce poisonous adsorbates such as carbon deposition and carbonate, increasing the catalyst lifetime42. We took this view and operated CORR under a pulsed electrolysis condition. We surmised that pulsed electrolysis could enrich OH- at the catalyst surface under alkaline electrolytes, which would boost acetate formation due to its nucleophilic attack toward *CH2-C=O intermediates.

We evaluated CORR performance of various catalysts in a flow-cell reactor with 3 M KOH (Supplementary Fig. 24). CO gas with a flow rate of 30 sccm was fed at the cathode and the outlet gas flow rate was measured for accurate product analysis. Solution-phase and gas-phase product quantifications were performed by the means of nuclear magnetic resonance (NMR) spectroscopy and in-line gas chromatography (GC), respectively. The impact of pulsed parameter on CORR activity of Ce-d-Cu was first evaluated by applying a square-wave pulsed with varied anodic current densities (ia = -5, -10, and -20 mAcm-2) and pulsed periods (ta=tc = 0.02, 0.1, 0.2, 0.5, 0.75, 1.0, 5.0 and 10 s) (Supplementary Figs. 25 and 26). Our measurements show that the Ce-d-Cu catalyst attains peak acetate FE of 71.3 ± 2.1% under the pulsed condition of ia = -10 mA cm−2, ic = −300 mA cm−2, ta=tc = 0.5 s (Fig. 2a).

Fig. 2: CORR performance.
Fig. 2: CORR performance.
Full size image

a FE values (top) and time-averaged current densities (bottom) on Ce-d-Cu under different cathodic current densities while keep ia = -10 mA cm−2, ta=tc = 0.5. Acetate FE (b), time-averaged acetate current densities (c), FE ratio of acetate and ethylene (d) and H2 FE (e) on various catalysts at cathodic current densities ranging from 100 to 500 mA cm-2. f Continuous 138 h stability test on Ce-d-Cu at the optimal pulsed parameter. g Comparison of this work with current CORR electrocatalysts, in terms of acetate FE, C2+ FE, FEacetate/FEC2+, H2 FE, and operating lifetime. Error bars in a–e were obtained by at least three independent measurements. Source data for Fig. 2 are provided as a Source Data file.

We found that the loading amount of Ce SAs had an important effect on the acetate selectivity, which reaches the maximum when the loading amount of Ce is 4.15 wt.% (Supplementary Fig. 27). Further increasing the Ce loading to 5.98 wt.%, however, leads to the decrease of the acetate and C2+ selectivity, probably the result of overloading of Ce-SAs that cover the Cu active sites. Compared with static CORR, the pulsed CORR enables significant enhancements of acetate selectivity and HER suppression (Supplementary Fig. 28). The increased acetate selectivity was attributed to the enriched OH- near the catalytic surface, as evidenced by in-situ Raman measurements (Supplementary Fig. 29), which could attack *CH2-C=O intermediates to promote acetate formation.

Figure 2a shows that acetate dominates the product distribution on Ce-d-Cu over the cathodic current density ranging from 100 to 500 mA cm-2 at the optimal pulsed parameter (ia = −10 mA cm−2, ta=tc = 0.5 s), achieving a maximum acetate FE of 71.3 ± 2.1% with a time-averaged acetate current density of 110.6 ± 1.9 mA cm−2. By contrast, the peak acetate FE on d-Cu and c-Cu are 32.9% and 19.3%, which are 2.2-fold and 3.7-fold lower than that of Ce-d-Cu (Fig. 2b and c, Supplementary Figs. 30 and 31, Supplementary Tables 2–6). For Ce-a-Cu and a-Cu, H2 becomes the major products, while the acetate FE is below 0.5%.

Since ethylene is the major competitor to acetate, we used the FE ratio of acetate-to-ethylene to describe the priority for acetate formation on different catalysts. Among different catalysts, Ce-d-Cu yields the highest acetate/ethylene ratio of 9.6 ± 1.9 at 300 mA cm-2, much greater than that of 2.0 for the counterpart catalysts (Fig. 2d). The stark contrast of acetate/ethylene ratio between Ce-d-Cu (9.6 ± 1.9) and d-Cu (1.0 ± 0.1) reveals that Ce-SAs can shift CORR pathway from ethylene towards acetate. Besides, d-Cu exhibits a higher acetate/ethylene ratio (1.0 ± 0.1) than that of c-Cu (0.5 ± 0.03), probably owing to the increased *CO coverage created by the favorable Cu+ at the interface32. Figure 2e shows that the a-Cu and Ce-a-Cu catalysts were both dominated by HER with H2 FEs higher than 70% at all current densities examined. Compared to d-Cu and a-Cu, the lower H2 FEs on Ce-d-Cu and Ce-a-Cu imply that Ce-SAs might suppress HER, owing to its weak affinity for H* intermediates31.

To examine the long-term stability of Ce-d-Cu catalyst, we performed continuous CO electrolysis at the most profitable pulsed parameter (ia = −10 mA cm−2, ic = −300 mA cm−2, ta=tc = 0.5 s) (Fig. 2f). This catalyst exhibits a stable acetate FE of >60% over 138 hours with no sign of improvement in overpotentials (Fig. 2f, Supplementary Fig. 32), which allowed us to produce 130.2 mmol of acetate throughout the test (Fig. 2f). Post-mortem HAADF-STEM characterizations of the spent Ce-d-Cu reveals that the Ce-SAs and dual-phase structure are both well retained (Supplementary Fig. 33), showing the structural and chemical robustness of the Ce-d-Cu catalyst under the pulsed electrolysis condition. In stark contrast, the acetate FE on Ce-d-Cu catalyst undergoes a significant drop from 31.2% to 22.5% within 28 hours under the non-pulsed condition (Supplementary Fig. 34). This indicates that the pulsed electrolysis can greatly mitigate the flooding issues, thereby increasing the catalyst lifetime.

We compared the CO-to-acetate performance of our Ce-d-Cu catalyst with those previously reported in terms of acetate FE, C2+ FE, FEacetate/FEC2+, H2 FE and operating lifetime17,43,44,45,46,47 (Fig. 2g). The performance metrics obtained on Ce-d-Cu outperform the most prior Cu-based catalysts achieved by other approaches (Fig. 2g), demonstrating the effectiveness of Ce-SAs modified dual-phase Cu for high-performance selective CO electrolysis toward acetate.

In situ ATR-SEIRAS studies

To investigate the origins of the marked CORR performance of Ce-d-Cu catalyst, we performed in-situ attenuated total reflection surface-enhanced infrared absorption (ATR-SEIRAS) spectroscopy to monitor CORR intermediates on different catalysts under the pulsed conditions (Ec = −0.1~−1.2 V, Ea = 0 V, ta=tc = 0.5 s) (Supplementary Fig. 35). Figure 3a shows clear spectral modes on various catalysts in the frequency ranging from 1740 to 2100 cm−1, belonging to the *CO intermediates48. As the overpotential increases, two vibration modes arise for Ce-d-Cu, d-Cu, c-Cu and Ce-a-Cu: the high-frequency band at ∼2050 cm−1 is attributed to *COatop configuration, while the low-frequency band at ∼1820 cm−1 originates from *CObridge configuration49,50 (Fig. 3a). By contrast, no obvious ATR-SEIRAS signals were found on a-Cu, explaining its poor CORR performance. A redshift of the ν(COatop) band was observed over these catalysts at more negative potentials, among which Ce-d-Cu displays the largest Stark tuning slope of 16.1 cm−1 V−1, suggesting the strongest interaction between *COatop and its surface32 (Supplementary Fig. 36).

Fig. 3: In-situ ATR-SEIRAS investigation.
Fig. 3: In-situ ATR-SEIRAS investigation.
Full size image

In-situ ATR-SEIRAS spectra ranging from 1750 to 2100 cm−1 (a) and 1280 to 1440 cm-1 (b) for various catalysts under the pulsed conditions (Ec = −0.1~−1.2 V vs. RHE, Ea = 0 V vs. RHE, without iR correction, ta = tc = 0.5 s). Peak area ratio of *COatop / *CObridge (c) and normalized peak area of *OCO (d) (normalized against the largest ν(O-H) band of *H2O) as a function of cathodic potential. Source data for Fig. 3 are provided as a Source Data file.

On Cu (111), prior studies have demonstrated that the reaction energy barrier of *COatop-*COatop dimerization is lower than that of *COatop-*CObridge and *CObridge-*CObridge dimerization51,52. Moreover, a higher *COatop coverage can boost the kinetics of *COatop-*COatop dimerization, facilitating C2+ formation53,54. Mathematical integration and statistical analysis disclose that the peak ratio of *COatop / *CObridge on Ce-d-Cu is much higher than that of other studied catalysts over the potential ranging from −0.1~−1.2 V, indicating its highest *COatop coverage (Fig. 3c). We found that the *COatop / *CObridge ratio follows the order a-Cu <Ce-a-Cu <c-Cu <d-Cu <Ce-d-Cu (Fig. 3c), which agrees with trend of Cu+ content observed over these catalysts (Supplementary Fig. 20). We thus speculate that the superior *COatop surface coverage on Ce-d-Cu is attributed to its abundant Cu+ sites induced by Ce-SAs. Once CO adsorbs on Cu, the electrons in Cu d orbitals will transfer to CO 2π* antibonding orbitals to form d → 2π* back donation55,56. Compared with metallic Cu0, Cu+ sites have a low-coordination state, which makes dz2 → 2π* more favorable than dxz/dyz → 2π* interaction, thus inhibiting *CObridge adsorption but promoting *COatop formation57.

The formation of acetate is signaled by the bands at 1402 and 1330 cm-1 (Fig. 3b), which are attributable to the νs(OCO) and δ(CH3) modes of the bridge-bonded acetate58. To reveal the formation of acetate, the integrated *OCO peak areas were normalized with respect to the respective largest ν(O-H) band of *H2O. Figure 3d shows that Ce-d-Cu exhibits the maximal normalized *OCO peak area across the potential range of −0.1~−1.2 V, followed by d-Cu, c-Cu, Ce-a-Cu, and a-Cu, which is consistent with the order of acetate FE and current density observed in Fig. 2b and c. Additionally, the normalized *OCO peak area versus cathodic potentials shows a similar trend with that of *COatop coverage (Supplementary Fig. 37), corroborating that the surface *COatop coverage contributes to acetate formation because *COatop can prompt the conversion of *(OH)C = COH to monodentate *C=C=O adsorbate—a key acetate intermediate24,25.

Interfacial water structure analysis

Interfacial water structure has been reported to play an important role in CO2RR and CORR, which directly affects the kinetics of hydrogenation steps occurring in the CO2RR and CORR as well as the competitive HER59,60. To explore the source of high acetate selectivity on Ce-d-Cu, we focus our attention on the O-H stretching mode (3000–3800 cm−1) of interfacial water via the ATR-SEIRAS spectra, which is associated with the distinct hydrogen bonding environments of water61. We chose the pulsed condition of Ec = -0.8 V, Ea = 0 V, ta=tc = 0.5 s to study the interfacial water structure of various catalysts (Fig. 4a, Supplementary Fig. 38). Through Gaussian fitting, the broad peak in 3000-3800 cm−1 can be divided into three peaks at about 3600, 3450 and 3250 cm−1 through Gaussian fitting, which can be assigned to K+ ion-hydrated water (K·H2O), water with weak hydrogen bonds (2-coordinated hydrogen-bonded water, 2-HB-H2O) and strong hydrogen bonds (4-coordinated hydrogen-bonded water, 4-HB-H2O), respectively61.

Fig. 4: Interfacial water structure of catalysts.
Fig. 4: Interfacial water structure of catalysts.
Full size image

a In-situ ATR-SEIRAS spectra of interfacial water on the various catalysts under the pulsed condition of Ea = 0 V, Ec = -0.8 V, ta=tc = 0.5 s (without iR correction). Gaussian fitting of three O-H stretching modes of 4-HB-H2O, 2-HB-H2O and K·H2O are shown in green, red and blue, respectively. Potential-dependent population of K·H2O (b) and frequency of O-H stretching modes of K·H2O varied with potential (c) on different catalysts (without iR correction). H2 FE (d), acetate FE, ethylene FE and FEacetate/FEethylene ratio (e) as a function of K·H2O populations. Source data for Fig. 4 are provided as a Source Data file.

K·H2O is formed due to the hydration process of K+ cations, where the K+ attracts the O atoms of the H2O molecules, allowing the H2O molecules to coordinate around the K+ ions to form K·H2O (ref62). Supplementary Fig. 39 show that the proportion of K·H2O on Ce-d-Cu decreases as the pulsed frequency increases. This is because that the K+ ions will undergo adsorption and desorption to rearrange in the double electrode layer when switching the operation potentials, resulting in the loss of partial K+ ions during the pulsed electrolysis63. Peak fittings of ATR-SEIRAS spectra at 2 Hz present the distribution of three water components depending on the applied potentials over the five catalysts (Supplementary Fig. 40). Among different catalysts, a-Cu yields the highest K·H2O population of 33.8%, followed by 14.6% for Ce-a-Cu, 5.2% for c-Cu and 4.1% for d-Cu, while Ce-d-Cu gives the lowest population of mere 2.5% (Fig. 4a). As the applied potential becomes more negative, the K·H2O population on the studied catalysts all increase due to the improved electrostatic enrichment of K+ cations (Fig. 4b). Moreover, in comparison to d-Cu and a-Cu, the lower K·H2O population on Ce-d-Cu and Ce-a-Cu indicates that Ce-SAs may repel K+ ions from migrating to the catalyst surface because of the electrostatic repulsion between Ce3+ and K+ cations. Figure 4c shows that the Stark slopes of K·H2O on a-Cu, Ce-a-Cu, c-Cu, d-Cu, and Ce-d-Cu are 50.5, 37.5, 31.8, 25.2, and 14.9 cm−1V-1, respectively. The steeper Stark slope indicates that the vibrational dipole moment of K·H2O (the direction of the OH bond) is more parallel to the direction of the electric field at the cathodic potentials61. Nevertheless, given the high oxophilicity of Ce element, Ce-SAs prefer the O-down structure of K·H2O while hindering the formation of H-down orientation31, leading to smaller stark slopes of Ce-a-Cu and Ce-d-Cu compared to their counterparts without Ce-SAs.

Because of the lack of constraint of hydrogen bonds, K·H2O is more inclined to transform into the two-H-down structure at negative potentials than 2-HB-H2O and 4-HB-H2O61. Such water orientation facilitates the water dissociation process (H2O+e-→H* + OH-, Volmer step) to generate H* species59,61, which could participate in HER and the hydrogenation steps during CORR. Thus, we used the K·H2O population to probe the impact of K·H2O on CORR. Figure 4d shows the H2 FEs increase monotonically with the K·H2O populations, confirming that K·H2O can generate H* for HER promotion. We also established the plots of acetate FE, ethylene FE, FEacetate/FEethylene ratio with respect to the K·H2O populations. By contrast, the FE toward ethylene shows a volcano-shaped dependence on the K·H2O population, where the c-Cu with 5.2% K·H2O yields the highest FE of 39.6% (Fig. 4e), suggesting that the moderate H* coverage favors ethylene formation. However, we observed that the FE toward acetate decreases as the K·H2O population increases. Additionally, the FEacetate/FEethylene ratio also exhibits a similar decreasing trend with the K·H2O populations, suggesting that low H* coverage could promote the formation of acetate while preventing ethylene generation. Taken together, we conclude that Ce-SAs in Ce-d-Cu could lower the surface H* coverage by reducing the K·H2O content, which enables a H*-deficient environment for the transformation of *(HO)C=COH to *C=C=O via a dehydration step, thereby leading to selective acetate production with a FE of 71.3 ± 2.1%.

Mechanism of selective acetate production

Last, we conducted in-situ differential electrochemical mass spectroscopy (DEMS) tests to gain mechanistic insights into the improved acetate selectivity enabled by Ce-SAs incorporation. We first performed real-time DEMS monitoring of the H2 and C2H4 formed over different catalysts (Supplementary Fig. 41). As Fig. 5a shows, for all the catalysts, when switching from the open-circuit voltage (OCV) to the constant current mode, the mass to charge signal (m/z) of H2 (H2+, m/z = 2) shows instant rising steps, corresponding to the H2 formation. Intriguingly, the intensity of H2 signals drops largely under the pulsed mode, which was caused by the enrichment of OH- ions that suppress HER41. The integration of the DEMS curve against time reveals the sum of H2 produced during this time. Our results show that when switching from the static mode to the pulsed mode, the H2 yield declines 73.6% on Ce-d-Cu, 52.7% on d-Cu, 39.6% on c-Cu, 20.8% on Ce-a-Cu, and 12.5% on a-Cu, respectively. The decrease of H2 production on Ce-d-Cu and Ce-a-Cu are greater than that on d-Cu and a-Cu. We observe that the DEMS signals of ethylene (C2H2+, m/z = 26) exhibit a similar trend on the studied catalysts (Fig. 5b), among which the ethylene production decreases 80.5% and 25.7% on Ce-d-Cu and Ce-a-Cu under the pulsed mode, much greater than that of 58.7% and 13.7% on d-Cu and a-Cu, respectively. These results further confirm that Ce SAs can inhibit ethylene and H2 generation, thus promoting acetate formation.

Fig. 5: Mechanism of enhanced acetate selectivity.
Fig. 5: Mechanism of enhanced acetate selectivity.
Full size image

DEMS signals of C2H2+ (a) and H2+(b) over time for various catalysts. The DEMS tests were performed in 3 M KOH (pH=14.43 ± 0.05) with constant current density (i = −10 mA cm−2) for 10 min, followed by a pulsed mode (ia = −1 mA cm−2, ic = −10 mA cm-2, ta = tc = 0.5 s) for 10 min. PZC values (c) and plots of Cφ vs. η (d) on various catalysts (without iR correction). Detailed schematic illustration showing how the Ce-SAs tailor interfacial water structure of Ce-d-Cu (e) and Ce-a-Cu (f) to promote CORR to acetate and suppress HER, respectively. Source data for Fig. 5 are provided as a Source Data file.

We further compared the potential of zero charge (PZC)—a intrinsic characteristic can reveal the amount of charge on the catalyst surface—of various catalysts using the Gouy-Chapman capacitance minimum method64,65 (Fig. 5c). In comparison to d-Cu with PZC of 0.213 V, the lower PZC value on Ce-d-Cu (0.157 V) suggests that Ce-SAs can reduce the surface electron density of Ce-d-Cu, which hampers the K+ enrichment for K·H2O formation, in consistent with interfacial water structure analysis above. To further explore the hydrogen adsorption behavior, we performed the operando electrochemical impedance spectra (EIS) measurements on the studied catalysts (Supplementary Fig. 42 and Supplementary Table 7). The double-parallel equivalent circuit model was employed to simulate the recorded Nyquist plots66. The second parallel components (Cφ and R2) reflect the hydrogen adsorption behavior on catalyst surface, where Cφ and R2 represent the hydrogen adsorption pseudo-capacitance and resistance, respectively67 (Supplementary Figs. 42 and 43, Supplementary Table 7). As shown in Fig. 5d, the integration of Cφ versus overpotential (η) provides information on the hydrogen adsorption charge (QH) on the catalyst surface, which thus can reflect the surface H* coverage. Among different catalysts, the Ce-d-Cu yields the lowest QH value of 426 μC, substantially smaller than that of 664 μC for d-Cu counterpart. The decreased hydrogen adsorption on Ce-d-Cu and Ce-a-Cu further validates that Ce-SAs can reduce H* coverage, which favors the formation of key *C=C=O intermediate for acetate.

Last, we performed DFT calculations to attain additional mechanistic insights. Our computational results show that COatop and H could adsorb on d-Cu with adsorption energy of -1.42 and -2.97 eV, respectively. In contrast, the incorporation of Ce SAs leads to strengthened COatop and weakened H adsorption with adsorption energies of -1.72 and -2.70 eV, respectively (Supplementary Data 1 and Supplementary Figs. 44–46). The reaction barrier of *(OH)C=COH→*C=C=O—a key step associated with the pathway toward acetate—was further calculated. Our simulations reveal that a high free energy barrier of −0.35 eV was required for *(OH)C=COH→*C=C=O conversion in the absence of Ce SAs. Nevertheless, the barrier was reduced to -0.68 eV after the incorporation of Ce SAs (Supplementary Data 1 and Supplementary Figs. 47–49). Overall, our theoretical results predict that the incorporation of Ce SAs can result in improved *COatop coverage and decreased H* adsorption, offering catalytic sites energetically more favorable for the *(OH)C=COH→*C=C=O transformation.

On the basis of above results, we become clear about the origin of the notable acetate selectivity on Ce-d-Cu catalyst under the pulsed electrolysis condition. First, Ce-SAs with positive charge on the dual-phase Cu diminish the surface electron density, which reduce K+ adsorption on the OHP, hampering the formation of K·H2O. Meanwhile, the oxyphilic Ce-SAs prevent the H-down orientation of K·H2O, which makes the dissociation of H* difficult, thus limiting the hydrogenation steps of CORR. Second, the electron transfer from crystalline Cu to Ce-SAs yields Cu+ species, which favor the active *COatop formation. As a result, *COatop undergoes the *COatop dimerization and hydrogenation steps to form *(OH)C=COH, which then transform into *CH2-C=O via *C=C=O under low H* coverage. Furthermore, OH- enriched at the catalyst surface under pulsed condition can attack *CH2-C=O, leading to the high selective production of acetate.

Discussion

In summary, we have demonstrated a Ce-SAs modified crystalline-amorphous dual-phase Cu catalyst, where Ce-SAs act as a structure modifier to enable highly selective CO-to-acetate conversion under pulsed electrolysis conditions. Such Ce-SAs modifier inhibits K·H2O formation by reducing the surface electron density, resulting in a low surface H* coverage. Moreover, Ce-SAs also facilitate the formation of Cu+ species through electron transfer from crystalline Cu to Ce SAs, which improves local *COatop coverage that benefits *(OH)C=COH formation. The low H* coverage results in a preference for the *(OH)C=COH→*CH2-C=O pathway during CORR. Consequently, *CH2-C=O will react with the enriched OH- under pulsed condition, yielding an acetate FE of 71.3 ± 2.1% and a time-averaged acetate current density of 110.6 ± 2.0 mA cm-2. This catalyst is also shown to operate continuously over 138 h in a flow-cell reactor. Our work demonstrates a distinct catalyst design based on the atomic dispersion of rare-earth elements that tailor interfacial water structure for highly selective CO2RR/CORR.

Methods

Chemicals

All chemicals were used as received without further purification. Cerium (III) nitrate (Ce(NO3)3 · 6H2O, ≥99.0%), copper (II) chloride (CuCl2·2H2O, ≥99.0%), sodium hydroxide (NaOH, ≥96.0%), ethylene glycol (C2H4(OH)2, 99.5%), acetone (CH3COCH3, 99.5%), DL-1-Amino-2-propanol (CH3CH(OH)CH2NH2, 93.0%), copper(II) acetate (Cu(CH3COO)2·H2O, ≥99.0%), hydrazine hydrate (N2H4·H2O, 85.0%), tannic acid (C76H52O46, ≥99.0%) and potassium hydroxide (KOH, ≥99.0%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Nitrogen (N2, 99.999%) and carbon monoxide (CO, 99.999%) gas were purchased from Nanjing Special Gas Factory Co., Ltd, China.

Synthesis of Ce-d-Cu

For the synthesis of 4.15 wt.% Ce-d-Cu, 1.02 g Ce (NO3)3 · 6H2O was dissolved in 10 mL ultrapure water (18.25 MΩ). Next, 0.2 g CuCl2·2H2O was dissolved in 40 mL ethylene glycol in a 100 mL beaker, followed by adding 1 mL aforesaid aqueous Ce (NO3)3 · 6H2O, and the obtained system was magnetically stirred for 30 min to form a homogeneous and stable solution. After that, 40 mg tannic acid was added, and the solution was magnetically stirred for 30 min until the tannic acid was dissolved. Subsequently, 2 mL NaOH (1 M) was added dropwise and stirred for another 10 min. The final product was collected by centrifuging at 9000 rpm, and washed with water and acetone three times each to remove the residual ethylene glycol in the sample, and finally dried in vacuum overnight. All the experiments described above were performed at room temperature (25 °C). For the synthesis of 2.12 wt. % and 5.98 wt.% Ce-d-Cu, 0.510 and 1.53 g Ce (NO3)3 · 6H2O were used respectively.

Synthesis of d-Cu

All the procedures were same with the synthesis of Ce-d-Cu except for not adding Ce (NO3)3 · 6H2O.

Synthesis of Ce-a-Cu

All the procedures were same with the synthesis of 4.15 wt.% Ce-d-Cu except that 160 mg tannic acid was added.

Synthesis of a-Cu

All the procedures were same with the synthesis of Ce-a-Cu except for not adding Ce (NO3)3 · 6H2O.

Synthesis of c-Cu

10 mL DL-1-Amino-2-propanol was added in 30 mL ethylene glycol in a 150 mL three-necked flask to form a homogeneous and stable solution. Subsequently, 2.73 g Cu (CH3COO)2·H2O was dissolved in the above solution and magnetically stirred for 30 min under N2 atmosphere in an ice bath. After that, 7.3 mL N2H4·H2O was added in the solution at one time under magnetic stirring, and stirred for another 12 h under an ice bath and N2 atmosphere. The resulting solution was settled overnight after adding acetone 1 : 1 by volume. The final product was collected by centrifuging at 10,000 rpm, washed with acetone and n-hexane for three times, respectively, and then dried in vacuum overnight.

Material characterizations

The synthesized catalysts were characterized by various analytic techniques. TEM images were obtained from H-7650. JEM-F200 and FEI Talos F200X, equipped with Super X-EDS system (four systematically arranged windowless silicon drift detectors) at 200 kV were carried out to acquire HRTEM and EDS mapping images. HAADF images and EELS spectra were measured on Titan Themis Z Atomic Resolution Analytical Microscope with an acceleration voltage of 300 kV. XPS data were gathered using an ESCALAB-MKII x-ray photoelectron spectrometer with Mg Kα radiation as an exciting source (Mg Kα = 1253.6 eV). XAFS measurements were performed on the RapidXAFS in Anhui Absorption Spectroscopy Analysis Instrument Co, Ltd.

Preparation of GDEs

Briefly, 30 mg catalyst, 100 μL nafion, and 1 mL isopropanol were mixed by ultrasonication for at least 30 min to form an ink. The catalyst loading amount can be quantified by measuring the mass difference before and after catalyst deposition, yielding a value of 2.4 mg cm−2. For different weights of catalysts, the nafion and isopropanol used should be scaled down or up, accordingly. The obtained ink was sprayed on a 2 × 2 cm2 gas-diffusion carbon papers (YLS-30T, purchased from Suzhou Sinero China) using an airbrush. The resultant GDEs were dried in a vacuum for CORR tests.

Electrochemical measurements

The electrochemical measurements were performed with a CHI760E workstation (CH Instruments, Shanghai) under the atmospheric environment with the temperature of 25 °C. The applied potentials were measured against an Ag/AgCl reference electrode (saturated in 3.0 M KCl) and converted to the RHE reference scale by the following equation:

$$E({{{\rm{vs}}}}.{{{\rm{RHE}}}})=E({{{\rm{vs}}}}.{{{\rm{{Ag}}}/{AgCl}}})+0.197+0.0591\times {{{\rm{pH}}}}$$
(1)

The calibration of the reference electrode involves the following steps: The reference electrode to be calibrated is designated as the working electrode, while a standard Ag/AgCl electrode serves as the reference electrode and a Pt foil as the counter electrode. During the calibration process, 3 M KOH solution is used as the electrolyte. The calibration test employs the open circuit potential monitoring method of the electrochemical workstation. The stable numerical values obtained from this test represent the potential difference E between the two reference electrodes. The actual potential Ex of the reference electrode can be calculated using the formula: E = Ex−0.197 V, where 0.197 V is the theoretical potential of the Ag/AgCl electrode.

CORR tests were performed in a flow cell. The pulsed electrolysis measurements were set with certain programs: ia = −10 mA cm−2, ic = −100, −200, −300, −400 and −500 mA cm-2, ta = tc = 0.5 s, the repeated times na = nc = 3600. The static electrolysis measurements were set with constant current density: i = −100, −200, −300, −400 and −500 mA cm−2, the electrolysis time t = 1800s. The prepared GDEs with the effective geometric area of 1 cm2, Ag/AgCl electrode, and a piece of IrOx coated Ti mesh were used as the cathode, reference electrode and anode, respectively. 168.3 g KOH are dissolved in 1 L H2O to form 3 M KOH solution (pH = 14.43 ± 0.05), which is then sealed in bottles at 25 °C. 20 mL of catholytes (3 M KOH) and 20 mL of anolytes (3 M KOH) were circulated by a peristaltic pump (BT100-2J, Longer Pump) in the cathode and anode chambers separated by a 2 cm × 2 cm cation exchange membrane (Nafion 117, 183 µm). The nafion 117 was treated before use as follows: it was first soaked with 5 wt.% H2O2 at 80 °C for 1 hour, then soaked in deionized water for half an hour. Next, it was boiled in 5 wt.% dilute H2SO4 at 80 °C for 1 hour and then was soaked in deionized water for another half an hour. The internal resistance of the flow cell was determined to be 4.71 ± 0.13 KΩ using a multimeter. The flow rate of the electrolytes was kept at 20 mL min−1 during the CORR tests. Gaseous CO was fed to the back side of GDEs through the gas chamber, whose flow rate was maintained at 30 sccm using a gas mass flow controller (C100L, Sierra). Before entering the chromatograph, the outlet gas underwent homogenization in a hermetically sealed gas chamber, with real-time flow rate monitoring conducted via a gas flowmeter to ensure measurement accuracy of gaseous emissions.

CORR product analysis

A gas chromatograph (Trace-1300, ThermoFisher) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID) were used to quantify the gas-phase product. The formula below was used to calculate the FE of the gas products:

$${{FE}}_{i}=\frac{{Q}_{i}}{{Q}_{{total}}}\times 100\%=\frac{{n}_{i}\times {C}_{i}\times u\times t\times F}{({I}_{a}\times {T}_{a}+{I}_{c}\times {T}_{c})\times {V}_{M}}$$
(2)

Where FEi: Faradaic efficiency for the gas product i; Qi: charge transferred for product i formation; Qtotal: charge passed through the working electrode; ni: electrons transferred for reduction to a molecule product i; Ci: volume fraction of the product i detected by GC; u: outlet gas flowrate; t: CO electrolysis time; F: Faradaic constant, 96485 C mol−1; Ia: anodic current during CO electrolysis; Ta: total anodic pulsed time; Ic: cathodic current during CO electrolysis; Tc: total cathodic pulsed time; VM: gas molar volume at room temperature (25 °C), 24.5 L mol−1.

Liquid-phase products were quantified by nuclear magnetic resonance (NMR) spectroscopy (Bruker 400 MHz spectrometer). For each sample, 1H NMR spectra was collected with water suppression mode. Typically, 500 µl of the catholyte was mixed with 100 µl of D2O containing 50 ppm (m/m) dimethyl sulphoxide (DMSO) as the internal standard. Same spectral acquisition parameters were used for all measurements to ensure complete relaxation and quantification. The FEs of liquid products are then calculated by the following formula:

$${{FE}}_{i}=\frac{{Q}_{i}}{{Q}_{{total}}}\times 100\%=\frac{{n}_{i}\times {C}_{i}\times V\times F}{{I}_{a}\times {T}_{a}+{I}_{c}\times {T}_{c}}$$
(3)

Where FEi: Faradaic efficiency for the liquid product i; Qi: charge transferred for product i formation; Qtotal: charge passed through the working electrode; ni: electrons transferred for reduction to a molecule product i; Ci: concentration of the product i in catholyte gained by NMR; V: total volume of catholyte; t: CO2 electrolysis time; F: Faradaic constant, 96,485 C mol−1; Ia: anodic current during CO electrolysis; Ta: total anodic pulsed time; Ic: cathodic current during CO electrolysis; Tc: total cathodic pulsed time.

Preparation of working electrodes for surface-enhanced infrared absorption spectroscopy (SEIRAS)

The silicon prism was first immersed in an aqua regia solution to remove the surface impurities. Then, the prism was polished with a 0.05 μm Al2O3 slurry to make the surface hydrophobic, followed by ultrasonic removal of surface residues in DIW, acetone, and DIW each for 5 min. The silicon prism was then dried and soaked 120 s in the NH4F solution to make the surface superhydrophobic. Then, the silicon prism was immersed in a mixture of gold-plated solution and a 2 wt % HF aqueous solution under a 55 °C water bath for 7 min to perform chemical gold plating. After that, the silicon prism deposited with Au film was gently cleaned with DIW and naturally dried. The electrodes were prepared by dropping ink onto the obtained Au film. First, 20 mg catalyst powder was dispersed in 1 mL isopropanol, followed by adding 30 μL Nafion solution, and the mixture was sonicated for 30 min. After that, 200 μL of the suspension was uniformly dropped onto the Au film, and finally dried at 60 °C to obtain the working electrode.

In-situ ATR-SEIRAS tests

In-situ ATR-SEIRAS tests were conducted in a spectroelectrochemical cell with a three-electrode configuration, where the as-prepared electrodes, an Ag/AgCl electrode, and a graphite rod were used as working electrodes, reference electrode, and counter electrode, respectively. The cell was assembled into a NICOLET iS50 FTIR spectrometer equipped with a liquid nitrogen-cooled MCT detector. Before the in-situ SEIRAS tests, the working electrode was first reduced at −0.1 V vs. RHE for 10 min in Ar-saturated 3 M KOH. During the in-situ SEIRAS tests under the pulse electrolysis, the spectra was collected for each 0.1 V decrease in Ec while Ea was fixed at 0 V, ta and tc were fixed at 0.5 s. Gaseous CO was kept bubbling into the electrolyte (3 M KOH) during the tests. Each spectrum consisted of 32 single beams with a resolution of 4 cm−1. The spectrum collected at 0.1 V vs. RHE in CO-saturated 3 M KOH was used as a background.

In-situ DEMS measurements

The in-situ DEMS experiments were conducted in a custom cell. The experiments were carried out in 3 M KOH, and the electrolyte is saturated by CO bubbling before the electrochemical measurements. A cold trap cooled by dry ice was installed between the vacuum chamber and the electrochemical cell to capture the water vapor generated during the experiments, thus protecting the mass spectrometer. A graphite rod and an Ag/AgCl electrode and were used as the counter electrode and reference electrode, respectively. The working electrode was prepared by airbrushing the catalyst ink on an Au film that was sputtered on the hydrophobic polytetrafluoroethylene membrane. Due to the pressure difference, the gas products generated in-situ will be sucked into the vacuum chamber through the cold trap for further detection by the mass spectrometer. The electrochemical measurements were carried out in 3 M KOH saturated with CO bubbling. The catalysts were subjected to the galvanostatic or pulsed electrolysis tests, while the mass signals of the gaseous products were recorded.

DFT calculations

The DFT calculations were carried out using the plane-wave code Vienna ab-initio simulation package (VASP)68 program with projector augmented wave (PAW)69 method. The convergence criterion of the electronic self-consistent iteration was set to be 10−6 eV, and the kinetic energy cut off is 500 eV. In addition, the atomic positions were relaxed until the force on each atom is below 0.02 eV Å−1. The Perdew-Burke-Ernzerhof (PBE)70 generalized gradient approximation (GGA) exchange-correlation functional was used throughout.

By analyzing the EXAFS fitting results in Supplementary Fig. 17, we uncovered that the average coordination number of Cu-O in a-Cu was 3.53, and the Cu-O bond was 1.91 Å. Based on this, we constructed the amorphous Cu oxide matrix model and then combine it with Cu (111) model to obtain the preliminary d-Cu slab model. According to the loading amount of Ce SAs (4.15 wt.%) in Supplementary Table 1, we further placed two Ce atoms on the amorphous-phase Cu oxides and crystalline-phase Cu of the d-Cu model respectively to obtain the preliminary Ce-d-Cu slab model. These slab models were separated by a 15 Å vacuum layer in a direction between the slab and its periodic images. During structural optimizations of the surface models, a 1 × 2 × 2 gamma-point-centered k-point grid for the Brillouin zone was used. All of the atomic layers were allowed to fully relax.

The adsorption energy (Eads) of an adsorbate A was defined as:

$${E}_{{ads}}={E}_{A/{surf}}-{E}_{{surf}}-{E}_{A}$$
(4)

where EA/surf, Esurf, and EA are the energy of A adsorbed on the surface slab, the energy of surface slab, and the energy of adsorbate A, respectively.

The Gibbs free energies are calculated by:

$$G=H-T\varDelta S={E}_{{DFT}}+{E}_{{ZPE}-}{TS}$$
(5)

Where the EDFT is the total energy from the DFT calculation. EZPE is the zero-point energy, S is the entropy, and T is the temperature (298 K). The vibrational analysis was carried out by relaxing the surface adsorbates and keeping the rest of the system fixed.