Introduction

The electrochemical carbon dioxide reduction reaction (CO2RR) has emerged as an attractive method for simultaneously closing the carbon cycle and producing important industrial feedstocks due to its compatibility with renewable, carbon-free energy sources1,2. Cu-based materials are particularly unique among the various types of CO2RR catalysts developed today, as they can produce multi-electron (>2 e−) reduced products like methane3, as well as multi-carbon (C2+) products consisting mainly of ethylene4,5 and ethanol6, regardless of their different structures, such as Cu foil7, nanoparticles8, and even molecular catalysts9. However, the activity and selectivity of these catalysts have not yet reached the level required for industrial application. Therefore, many efforts have been made to improve their electrochemical performance further, especially for more valuable C2+ products. These efforts include facet design10, defect engineering11, pretreatments such as oxide-derived Cu (OD-Cu)12, and electrolyte engineering13,14. However, the determining factors and atomic-level insights into the underlying mechanisms are still unclear.

CO2RR, as a representative aqueous electrochemical reaction, is significantly affected not only by the catalyst but also by the environment at the solid–electrolyte interface. Since the 1990s, Hori et al. have discovered that alkali metal cations (M+) can enhance CO2RR activity by more than an order of magnitude and concurrently improve C2+/C1 selectivity on the Cu electrode. This promoting effect becomes more significant with larger cations from Li+ to Cs+13. The promoting effect of alkali metal cations has been confirmed15,16,17 and extended to other CO2RR catalysts, such as CO-selective Ag18, Au19, and HCOOH-selective SnO220. Several explanations have been developed so far. The surface pH model attributed the differences between alkali metal cations to the interfacial proton and CO2 concentrations resulting from the equilibration between carbonate-group anions13,18, but the relationship between this model and the reaction mechanism is not well understood. Subsequently, a dipole-field interaction model was proposed, which emphasizes the stabilizing effect of the electric field between solvated alkali metal cations and the negatively charged metal surface on polar intermediates15,21,22. Therefore, intermediates with large dipoles involving *CO2 and *OCCO (the asterisk * denotes the adsorption site) would be strongly stabilized, enhancing the selectivity of corresponding products such as CO on Ag/Au and C2+ on Cu. However, this model is based on the uniform interfacial electric field model and attributes the discrepancy between different alkali metal cations to the tendency of larger cations to accumulate more at the interface while ignoring the intrinsic difference of individual alkali metal cations and possible local interactions between cations and intermediates.

Recently, it has been further demonstrated that the CO2RR cannot be initiated on Cu, Ag, Au, and SnO2 without alkali metal cations, even to CO and HCOOH20,23,24, which cannot be fully explained by the above models. In addition, spectroscopic results have shown that the introduction of Li+, but not Cs+ cations, results in the highest electric field in the Stern layer, while the strength of the solvation-induced reaction field, the Onsager field, varies with the cation size25. These findings suggest that different alkali metal cations have inherent differences rather than a simple concentration disparity and that cations at the interface could greatly influence the efficiency of the CO2-to-CO conversion process, potentially playing a more active role rather than merely acting as spectators. To provide a more detailed explanation of the influence of solvated alkali metal cations on reaction mechanisms, ab initio molecular dynamics (AIMD) simulations were performed and revealed the existence of local interactions, such as coordination between solvated cations and surface intermediates26. Notable examples include the interaction between K+ cation and adsorbed *CO2 molecule on Au surface23,27, and the interaction between Li+/K+/Cs+ cations and *CO + *CO dimer on Cu surface28. However, a comprehensive understanding of the effect of different alkali metal cations on CO2 activation and their critical role in CO2RR initialization on the Cu surface is still lacking.

In this work, we have systematically studied the effect of alkali metal cations on both CO2RR and the main competing hydrogen evolution reaction (HER), using AIMD simulations with enhanced sampling methods and explicitly taking into account the effects of solvation and potential. By calculating the free energy barriers of CO2 activation, a monotonically decreasing trend from Li+ to Cs+ is found, which is due to the coordination ability between the alkali metal cation and the *CO2 intermediate. The bridge coordination between larger cations and *CO2 helps to facilitate CO2RR more compared to the side configuration preferred by smaller cations. Furthermore, we show that the necessary role of alkali metal cations in CO2RR on Cu surfaces cannot be fully understood without simultaneously considering the competing HER. Our results shed light on the intrinsic short-range stabilizing effect of individual alkali metal cations on CO2RR and highlight the necessity of explicitly considering the environment, such as cation and solvation, as well as competing reactions, in electrochemical simulations in order to obtain more reliable insights.

Results and discussion

Electrical double-layer response of applied potential

To provide a benchmark for the subsequent analysis, we first performed simulations of the Cu–water interface system with and without CO2 adsorption, denoted as Cu–*CO2–H2O and Cu–H2O, respectively, to investigate the response of the electrical double layer (EDL) to different applied potentials. Figure 1a–c shows the density distribution of water perpendicular to the interface under different conditions. Three primary peaks could be observed above the Cu–water interface at approximately 2.1, 2.8, and 5.2 Å, respectively. The first peak (marked with the gray dashed line in Fig. 1a–c) is attributed to water molecules chemisorbed on the Cu surface, while the second and third peaks indicate two partially ordered water layers. No evident peaks were observed further away from the interface, representing the diffusion layer and the bulk water with a density of about 1 g cm−3, and indicating the sufficient thickness of the water layer as well. Furthermore, as the applied potential decreases, the intensity of the first peak decreases, as shown in Fig. 1a, which could be attributed to the increasing repulsive interaction between the more negatively charged substrate and the partially negatively charged O atom of the H2O molecule (as schematically shown in Fig. 1d). Meanwhile, H atoms would be attracted by the surface, causing the O–H bond to rotate toward the surface, as reflected by the change in interfacial water dipoles within 5 Å of the Cu surface (Supplementary Fig. 1). The dipole distribution changes from random to a single peak, indicating a more aligned distribution of water molecules in the first water layer.

Fig. 1: Response of interfacial structure to the applied potential.
figure 1

a–c Distribution of water density along the direction perpendicular to the Cu(100)–water interface under different conditions, with the topmost Cu layer at the distance of 0. The gray dashed line indicates the first peak of chemisorbed water. d Schematic of the interfacial structure change under corresponding conditions to (a–c). CO2 molecules are depicted as gray and red spheres, while H2O molecules are represented as stick models for clarity. The applied potentials are obtained from the work function of the whole system. (Supplementary Note 1).

Full size image

A similar response of the interfacial water arrangement to the applied potential was also observed in the Cu–*CO2–H2O system (Supplementary Fig. 2), where a *CO2 molecule was introduced onto the Cu surface of the Cu–H2O system. Energetically, the adsorption enthalpies of CO2 decrease with decreasing applied potential, indicating the more stable adsorption of CO2 (Supplementary Table 1). During the additional 15 ps of AIMD simulation, CO2 was observed to be stably adsorbed on the Cu surface in a bidentate configuration through the C–Cu and O–Cu bonds, with the other O atom facing upward and forming hydrogen bonds with the surrounding water molecules (adsorption configuration in Fig. 1d). Bader charge analysis29 shows that the chemisorbed CO2 has a charge of about −1.05 e, indicating a fully activated adsorption state. In addition, the introduction of CO2 enhances the adsorption of chemisorbed water (Fig. 1b), but this is due to the lack of potential control (Const-Q case in Fig. 1d). After adjusting the applied potential to be consistent with the Cu–H2O system by electron exchange with the electron reservoir, the intensity of the first peak is almost identical to the reference (Fig. 1c and Const-U case in Fig. 1d). The above results confirm the validity of our model and provide the basis for further investigation of the electrochemical interface during the CO2RR process30,31,32,33.

Furthermore, it is worth noting that the bidentate configuration of *CO2 on the Cu(100) surface differs from that observed in a fully implicit solvent environment. In the latter case, CO2 tends to be physisorbed on the Cu(100) surface over a wide potential range (Supplementary Fig. 3), while the bidentate configuration is less stable by ~0.3 eV at −0.6 V vs. SHE (standard hydrogen electrode), which is the concerned potential range in this work. This difference indicates the significant contribution of surrounding hydrogen bonds to the stabilization of *CO2 from explicit solvation and thus raises a cautionary note regarding the use of fully implicit solvents for investigating molecular explanations of reaction mechanisms, especially when the interactions between intermediates and the surrounding electrochemical environment, such as solvated water molecules, play a substantial role.

Promoting effects of alkali metal cations on CO2 adsorption

Experimental results have shown that CO2RR products, including CO, cannot be detected on coinage metals without M+ in the electrolyte20,23,24. This suggests that CO2RR is inhibited in the early stages of the CO2-to-CO conversion process. To investigate the underlying mechanism, here we first focused on the effect of alkali metal cations on the rate-limiting step of CO2RR to CO, the CO2 activation process34,35,36, which is expected to have electrostatic interactions with alkali metal cations due to its transformation from a linear non-polar configuration to a bent polar one. We studied the Cu–*CO2–H2O system in more detail and subsequently added M+–F− (M+ = Li+, Na+, K+, or Cs+) ion pairs into it to keep the applied potential close to the PZC of Cu(100) as well as the onset potential of CO2RR at Cu (−0.6 V vs. SHE)14, which are referred to as Cu–*CO2–M+(aq) systems. Bader charge analysis confirms full ionization of the added ions, with +0.9 e for alkali ions and −0.85 e for F ions, respectively. Using the angle of CO2 as a collective variable (CV) during the adsorption process, we found that the CVs of the initial state (IS, a free linear CO2) and the final state (FS, a chemisorbed bent CO2) do not vary much with different cations, peaking at ~174.5° and ~118°, respectively (Supplementary Fig. 4). The angle of CO2 at the IS deviates from 180° due to thermal fluctuations and interactions with water, and the chemisorbed configuration of *CO2 at the FS is identical to that in the Cu–*CO2–H2O system (Fig. 2 and Supplementary Fig. 4). These results further demonstrate the significant role of hydrogen bonds formed between *CO2 and surrounding H2O molecules.

Fig. 2: Schematic of the representative *CO2 adsorption structures.
figure 2

Local interactions between *CO2 at Cu(100) surface and the surrounding environment: a Schematic of the bidentate adsorption of *CO2 at Cu surface in an explicit solvent with C and one of the O atoms coordinating with Cu, and the other O atom forming hydrogen bonds (light blue dashed lines) with surrounding water. b Schematic of the side configuration of Li+ or Na+ coordinated with only one O atom of *CO2 (magenta dashed line). c Schematic of the bridge configuration of K+ or Cs+ coordinated with both two O atoms of *CO2.

Full size image

More importantly, we found that all alkali metal cations can stably coordinate with *CO2 but with different coordination features. Li+ and Na+ ions coordinate with only one O atom of *CO2 (side configuration, as shown in Fig. 2b), while K+ and Cs+ would coordinate with both O atoms of *CO2 (bridge configuration, as shown in Fig. 2c). One could naturally assume that the bridge coordination results from the larger cation radius, which is supported by the increasing first solvation shell radius (\({r}_{{{{{{\rm{M}}}}}}({{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}})}\)) from Li+ to Cs+ (Supplementary Table 2). If the configuration differences are maintained during the activation process, this phenomenon could be an intrinsic cause of the different promoting effects on CO2RR of M+ with different sizes. The bridge configuration is more favorable for CO2 bending as a consequence of the electrostatic repulsion between M+ and the partially positively charged C atom of CO2. From an energetic point of view, we firstly considered the CO2 adsorption free energy with different alkali metal cations as an approximation of the activity descriptor. The system with Li+ shows the strongest adsorption of *CO2, while the adsorption free energies with the other three cations do not differ much from each other (Fig. 3 and Supplementary Table 3). Consequently, no clear trend with cation size could be observed, which is also consistent with previous results on Au(111)23. As repeatedly demonstrated35,37, the reaction energy alone cannot fully describe the activity of a system, while the activation energy from the IS to the transition state (TS) might be more important to serve as the real energetic criteria. Therefore, in order to obtain more accurate and detailed information about the cation effects on the CO2 adsorption process, we next concentrated on the free energy barrier and transition states of the CO2 adsorption process at Cu(100).

Fig. 3: Free energy profiles and representative structures for CO2 activation process on Cu(100) surface.
figure 3

a Free energy profile and representative initial state (IS), transition state (TS), and final state (FS) structures for the CO2 adsorption process in the pure water (Cu–*CO2–H2O) system without M+. The gray dashed lines indicate the zero value. b Corresponding free energy gradient with error bar at each calculated collective variable. Crossover points with dashed baseline indicate the positions of IS, TS, and FS. The error bars represent standard errors of free energy gradients calculated using block average at the corresponding collective variable. c–f Free energy profiles and representative structures at IS, TS, and FS for CO2 adsorption with different cations (Cu–*CO2–M+) of c Li+, d Na+, e K+, and f Cs+. The shadows in all five free energy profiles represent the cumulative errors ( ~ 0.01 eV at TS) estimated with the block average method (Supplementary Notes 2 and 3). In the local structures, CO2 molecules are depicted as gray and red spheres, Cu substrates are shown as reddish-brown spheres, H2O molecules are illustrated as red and white sticks, and M+, including Li+, Na+, K+, and Cs+ are represented by green, yellow, purple and blue spheres, respectively. The magenta dashed lines around M+ illustrate their coordination with O atoms in the first solvation shells.

Full size image

Figure 3a shows the free energy profile of CO2 activation in the Cu–*CO2–H2O system in the absence of M+, where a kinetic barrier of 0.22 eV is indicated. For comparison, Fig. 3c–f shows the free energy profiles of CO2 activation with ions in the solvent. The kinetic barriers exhibit a clear, monotonically decreasing trend from the Cu–*CO2–H2O system to the Cu–*CO2–M+(aq) systems, with the inclusion of Li+, Na+, K+, and Cs+. This trend is consistent with the observed activity trend of CO2RR with alkali metal cations in experiments15,16,17,18,19. Moreover, cations are generally divided into two groups during the activation process based on the activation barriers and their coordination characteristics with *CO2: smaller cations of Li+/Na+ and larger cations of K+/Cs+ (validation of the stable configurations is discussed in Supplementary Note 4). The free energy barrier difference between the two groups is about 0.05 eV, which can increase the corresponding reaction rate by a factor of 5, which is comparable to the experimental results15,16,17,18,19,25.

The activation barrier in the CO2 bending process and the concerted charge transfer result from the degeneracy breaking of the 2πu (lowest unoccupied molecular orbital, LUMO) and 1πg (highest occupied molecular orbital, HOMO) states of CO2. Concurrently, the lower LUMO level of the bent CO2 facilitates the back donation of electrons from the substrate to CO2, resulting in the negatively charged *CO2−36. This process can be further enhanced by alkali metal cations (Fig. 3). In the Cu–*CO2–H2O system, the CV at the TS is approximately 153°, but it moves closer to the IS with alkali metal cations. Meanwhile, the negative charges on the CO2 at the TS show a similar monotonically decreasing trend, with about –0.3 e transferred to CO2 in the Cu–*CO2–H2O system, while only about –0.17 e is transferred in the Cu–*CO2–Cs+(aq) system (Supplementary Table 4). This also implies that the TS is closest to the IS with the Cs+ cation, which benefits its lowest activation barrier.

The coordination structures of M+ in the first solvation shell with and without *CO2 at the TS were further analyzed to understand the underlying molecular mechanism of the CO2 activation (Fig. 4a), with the systems denoted as Cu–*CO2–M+(aq) and Cu–M+(aq), respectively. For Li+ and Na+ cations, the coordination with CO2 involves the exchange of an oxygen atom from H2O in the first solvation shell, resulting in an apparently unchanged total coordination number. This exchange is mainly due to the hard solvation shells of small cations, which can hardly accept more solvation molecules with the environment, as shown by the platform in the integrated radial distribution function (RDF) at the radius of the first solvation shell in Supplementary Fig. 5. However, for K+ and Cs+ cations, the coordination numbers increased by 0.9 and 2.2, respectively. This is because K+ and Cs+ could coordinate with more than one oxygen from *CO2 and maintain the coordination with H2O. The higher coordination number of K+ and Cs+ cations with their unique bidentate coordination structure with *CO2 could facilitate the electron transfer process and help to activate CO2 more effectively than smaller cations, providing important insights into the molecular origin of their higher promoting effect on CO2RR activity. In addition, there is a difference in the CO2 activation barrier between Cs+ and K+ cations, but experimentally, they were found to activate CO2 similarly25, which could also be explained by their coordination structure. Comparing the total coordination numbers of alkali metal cations in the Cu–M+(aq) system with those in the Cu–*CO2–M+(aq) systems (Supplementary Fig. 6), most remain unchanged except for a decrease in the Cs+ cation at the interface without *CO2. In the AIMD simulation, we observed that the Cs+ cation loses some solvated H2O molecules at the Cu–M+(aq) interface (Supplementary Fig. 7), but when it coordinates with *CO2, its solvation ability with H2O is slightly recovered at the Cu–*CO2–M+(aq) interface (Fig. 4a). The partially desolvated Cs(H2O)x complex behaves like quasi-specific adsorption that may block the active sites and be unfavorable for surface reactions, ultimately resulting in its similar activating effect on CO2 with K+.

Fig. 4: Coordination analysis of the first solvation shell of alkali metal cations around TS.
figure 4

a The number of coordinated O for Li+, Na+, K+, and Cs+ cations at the Cu(100)–electrolyte interface with and without the *CO2. b Contour map of the distance distribution between alkali metal cations and two oxygens of CO2, statistically derived from snapshots around the transition state. The results of Li+, Na+, K+, and Cs+ are represented by green, orange, purple, and blue, respectively.

Full size image

Figure 4b further shows the statistical distribution of the distance between alkali metal cations and two oxygens of CO2, O1, and O2. Li+ and Na+ cations were observed to coordinate stably with only one oxygen, resulting in a larger M+–O2 distance compared to M+–O1. In contrast, for K+ and Cs+ cations, the preferred bridge configuration leads to an equal distance of M+–O1 and M+–O2. Another difference between these two cation groups is the density of the distributions, with the larger cations showing a broader distribution, consistent with the absence of platforms in the integrated RDF (Supplementary Fig. 5c, d). This suggests a more dynamic coordination around CO2 compared to Li+ and Na+ cations. In addition, we observed an accumulation of electrons between the Li+ cation and the coordinated O of CO2, which is the characteristic feature of a bond. However, no significant charge accumulation was observed for the Cs+ cation (Supplementary Fig. 8), supporting the above analysis. In this way, more dynamically coordinated K+/Cs+ with *CO2 could also accelerate the reaction cycle by flexibly moving away to help activate CO2 at other active sites and avoid inhibiting further reaction steps compared to more strongly coordinated Li+/Na+.

Based on the kinetic barrier simulations of CO2 activation and the coordination analysis of M+, we could conclude that the distinct coordination ability of alkali metal cations is responsible for their different promotional roles in CO2 activation on the Cu surface. The bridge configuration between K+/Cs+ and *CO2 could both facilitate the CO2 bending and electron transfer process, and the flexible coordination between them could accelerate the reaction cycle. These effects together result in the obvious activity and selectivity enhancement of K+ and Cs+ cations compared with Li+ and Na+. These results are consistent with the trend of Onsager reaction field strengths from the spectroscopic exploration25.

Furthermore, similar behavior could also be expected on other coinage metal surfaces or facets due to their similar response to alkali metal cations18,19. Basically, the effect of different metal surfaces or facets on an electrochemical system can be divided into two main aspects. First, it can alter the active sites, thereby affecting the binding strengths between the substrate and intermediates due to changes in substrate properties or the coordination number of the active sites. Second, it can alter the potential of the zero charge, thus influencing the surface charge density under specific applied potentials. Since the activation of CO2 involves electron transfer processes, changes in the surface charge density can significantly affect the driving forces provided by the applied potentials. For example, in our simulations, we observed stable adsorption of bent *CO2 on Cu(100) surfaces without any constraints or external factors such as cations or potentials. In contrast, previous reports on the Au(111) surface indicated that bent *CO2 could not be stably adsorbed even in the presence of K+ ions23,37. However, it is further demonstrated that bent *CO2 can become a stable configuration at higher surface charge densities on the Au(110) surface27. Despite the different stable adsorption configurations observed on different metal surfaces, the promoting role of alkali metal cations remains a consistent factor. Therefore, we expect that some of the findings from our work can be applied to other metal surfaces and facets and that additional effects will also be explored in our future work.

Necessary effect of M+ on CO2RR

The above analysis successfully explains the increasingly promoting role of alkali metal cations from Li+ to Cs+ on CO2 activation. However, the relatively low kinetic barrier of 0.22 eV in the Cu–*CO2–H2O system cannot fully account for the necessary role of alkali metal cations on CO2RR observed in the experiment. To determine this, we need to consider not only the main reaction CO2 activation but also the side reaction HER. In the absence of alkali metal cations, the highly acidic electrolytes used in the experiment could lead to a high concentration of surface protons, causing the competing HER to dominate20,23,24. Although HER is known to be a major competing side reaction during CO2RR, the molecular understanding of the effects of alkali metal cations on HER has rarely been studied. Therefore, we further investigated the possible influence of alkali metal cations on proton accessibility and HER activity on the Cu surface.

Figure 5 shows the distributions of an excess proton with and without the alkali metal cations at the interface. Cumulative averages of the results are shown in Supplementary Fig. 9. The excess proton is solvated to form a hydronium ion (H3O+), which can move along the hydrogen bond network by hopping between different water molecules. After equilibrating for more than 20 ps, the solvated proton reaches a dynamically stable configuration at a distance of about 5.2 Å away from the Cu surface. Based on the water distribution shown in Fig. 1, the excess proton diffuses into the second water layer (third peak in Fig. 1a), which satisfies the three hydrogen bonds required by the H3O+ ion. However, when alkali metal cations are introduced at the interface, the proton-interface distance increases, except for Li+ (Fig. 5a). The change in proton-interface distance results from the disruption of the interfacial hydrogen bond network and the electrostatic repulsion of solvated alkali metal cations24,38,39. For larger cations, including Na+, K+, and Cs+, the excess proton is repelled away from the second water layer and confined in the first solvation shells of these cations, as indicated by the comparable proton-interface distances and the diameters of the first solvation shells of alkali metal cations. The average proton-interface distance with Cs+ is slightly less than twice the radius of the first solvation shell due to its partial desolvation behavior, similar to the case in Supplementary Fig. 7. As for the case with Li+, the proton-interface distance seems to be even smaller than in the case with a single H3O+, which is due to two aspects. On the one hand, the solvation shell of Li+ is smaller and harder than that of larger alkali metal cations, as indicated in the previous sections, which means that the interfacial hydrogen bond network is relatively more complete, leaving space for H3O+ to form three required hydrogen bonds at the interface. In this way, the H3O+ and solvated Li+ could coexist between the first and second water layers despite of the electrostatic repulsion between them. Furthermore, after the introduction of alkali metal cations, the negative charge density at the Cu(100) surface increases compared to that of single H3O+, which further attracts the proton to approach the interface. Therefore, the attraction from the negatively charged catalyst surface and the inhibition from alkali metal cations combine to result in the distinct proton-interface distances with various alkali metal cations near the interface. Furthermore, the hydrogen bond lifetime decreases with the inclusion of the Li+ cation and is further suppressed with increasing cation size (Supplementary Note 5), which provides additional support for the influence of solvated alkali metal cations on solvent dynamics.

Fig. 5: Interaction between solvated proton (H3O+) and alkali metal cations.
figure 5

a Distribution of H3O+ in the solvent with and without alkali metal cations. The red and purple dashed lines represent the average distance of the excess proton from the Cu surface in the production timescale of 10 ps with and without K+ cation, respectively. The corresponding representative interface structures are shown in (b). H3O+ and K+ are highlighted and shown as spheres where O is represented in red, H in white, and K in purple, while other atoms are hidden for visual clarity. Magenta dashed lines around K+ illustrate the coordination with O atoms in the first solvation shell.

Full size image

In addition, since H2O molecules could also be potential proton sources, we further investigated the kinetic barriers of the Volmer step from H2O molecules, such as free H2O, surface-adsorbed H2O, and K+-coordinated H2O, as shown in Supplementary Fig. 10. The free and adsorbed H2O molecules exhibit high free energy barriers of 0.75 eV and 0.95 eV, respectively, while the K+-coordinated H2O shows the highest barrier of 1.05 eV, which are all far more difficult to happen than CO2RR. Therefore, by combining the analysis of cation effects on both CO2RR and competing HER reactions, we were able to fully explain the necessary role of alkali metal cations in CO2RR, which resulted from the inhibition of protons derived from H3O+ approaching the surface, as well as the much higher kinetic barrier of protons derived from H2O molecules compared to CO2 activation.

Note that only the qualitative trend is provided here. Typically, determining the reaction network requires a quantitative criterion based on differences in activation barriers. However, in this case, the critical factor is the inhibition of protons approaching the surface by cations. This implies that the concentration of reactants, influenced by potential, ionic activities, and mass transport, may play the dominant role. Furthermore, in reality, the interfacial concentration of alkali metal cations may deviate from one another under a specific potential due to different capacitance responses40,41,42,43. Since only a few ions can be introduced at the nanoscale simulations due to scale limitations, it is challenging to reflect the concentration change accurately. Hence, this necessitates further exploration of both interfacial evolution and simulation methods.

Furthermore, our results derived from the fully explicitly solvated interfacial model offer significant insights into future electrochemical simulation work. Notably, the model indicates different adsorption configurations of the CO2 molecule in implicit and explicit solvents, distinct coordination structures of solvated alkali metal cations and intermediates, and the interaction between protons and alkali metal cations. The results suggest the importance of explicitly considering the electrolyte in simulation models, particularly when dealing with species that could potentially interact with the interfacial hydrogen bond network or other components within the electrical double layer, including solvated ions. Meanwhile, our investigations into the copper surface may aid in gaining a broader understanding of how alkali metal cations impact CO2RR performance. This research can be combined with studies conducted on gold surfaces27,37,44, where the alkali metal cations could differentiate the CO2 activation pathways between the inner- and outer-sphere mechanisms. After clarifying the various roles that alkali metal cations play on different coinage metals and even other CO2RR catalysts, it is expected that a more systematic relationship between EDL design and CO2RR performance can be established, which demands further exploration.

In conclusion, the results of a series of explicitly solvated AIMD simulations on CO2RR and competing HER have provided systematic insights into the promotional roles and the necessary effects of alkali metal cations for CO2RR. Energetically, the CO2 activation barrier decreases with increasing cation size. The different promotional roles can be attributed to the coordination abilities, or Onsager field strengths, of various alkali metal cations. In particular, smaller Li+ and Na+ cations can only coordinate with one oxygen atom of CO2 and simultaneously release a water molecule, whereas K+ and Cs+ cations can flexibly coordinate with both oxygen atoms of CO2, which is more favorable for CO2 activation and reaction cycling. However, the partial desolvation of Cs+ cation slightly inhibits the increasing tendency, resulting in its similarity with K+ cation. On the other hand, the necessary role is attributed to the competition between HER and CO2RR. The interfacial proton accessibility can be largely hindered by alkali metal cations, and the much higher kinetic barrier of protons derived from H2O molecules helps CO2RR to outcompete HER. Given the similarity in catalytic behavior with alkali metal cations, these phenomena are expected for other CO-selective catalysts, such as Ag and Au. Our results help to understand the molecular origin of the possible role of the electrochemical environment emphasize the importance of explicit solvation, ions, and systematic exploration of competing reactions in electrochemical simulations, and show the effectiveness of theoretical tools in exploring the solid–electrolyte interface.

Methods

Solid–electrolyte–Ne double electrode model

Based on experimental results, it was determined that the Cu(100) surface would dominate after a prolonged electroreduction process45. To represent the electrode, we chose the Cu(100) surface modeled as a three-layer slab with the p(3 × 3) supercell to simulate the CO2 activation processes, and larger p(4 × 4) supercell were utilized when considering the interactions between alkali metal cations and protons. The space above the Cu surface was filled with water molecules with a density of 1 g cm−3. In order to maintain the water density, one water molecule was replaced with a CO2 molecule. The volume of the solvent space was expanded with the addition of ions. Alkali metal cations M+ (Li+, Na+, K+, or Cs+) were initially positioned ~5 Å away from the Cu(100) surface, while an anion F- was placed far away on the other side of the solvent to serve as an ionization pair with the cation, maintaining the applied potential close to the CO2 onset potential. An additional Ne atomic layer and a 12 Å vacuum layer were added over the interface model to act as a counter electrode and monitor the electrochemical potential of the entire system46,47.

AIMD simulations

All AIMD simulations were performed using the Vienna Ab initio Simulation Package (VASP)48. The projector augmented wave (PAW) potential49 and the generalized gradient approximation (GGA) with the revised Perdew–Burke–Ernzerhof (RPBE) functional50 were employed to describe the electron-ion interactions and the exchange-correlation energy, respectively. The van der Waals interactions were described using the DFT-D3 correction51. The energy cutoff for the plane wave basis expansion was set to 500 eV. For the Brillouin zone sampling, only the gamma point was used for AIMD simulations, while a denser 3 × 3 × 1 grid was used for electronic structure calculations52. The Nose–Hoover thermostat was used for canonical sampling at 300 K53,54. The kinetic barriers and the free energy profile of the CO2 activation process were studied using a combination of slow-growth55 and blue-moon sampling56 methods with the SHAKE algorithm57, with the angle of the CO2 molecule chosen as the collective variable (CV). The choice of CV is discussed in Supplementary Note 2.

Considering the applied potential, the Ne counter electrode46,47 was used to monitor the potential change of the reaction system on the fly. A dipole correction58 along the direction perpendicular to the interface was applied in the vacuum layer. The vacuum level of the metal side could be considered as constant due to the electrostatic screening of the metal, and thus the absolute potential of the system, i.e., the applied potential, could be calculated accordingly: \(U \, {{{{{\rm{vs}}}}}}\,{{{{{\rm{SHE}}}}}}={\Phi }_{{{{{{\rm{Interface}}}}}}}-4.44\,{{{{{\rm{V}}}}}},\, {\Phi }_{{{{{{\rm{Interface}}}}}}}={{\Phi }_{{{{{{\rm{metal}}}}}}}}+\frac{4{{{{{\rm{\pi }}}}}} {\mu}}{A}*{k}\), where μ is the dipole moment of the system, A is the surface area, and k is Coulomb constant. Then, constant potential corrections were applied to account for the slight change in the electrochemical potential for different reaction states along the reaction paths (Supplementary Note 6)59.