Introduction

Alkali metals have been extensively employed as promoters in various important chemical reactions such as CO oxidation1,2, CO2 hydrogenation3,4,5, ammonia synthesis6,7 and water-gas shift reaction8,9. Upon adsorption on transition metal surfaces, alkali metals are highly susceptible to donate electrons to the substrates due to the low ionization energies, resulting in the formation of alkali metal cations. Such an interfacial charge transfer induces the reduction of work function on transition metal surfaces, which enables to tune the adsorption and activation of molecules, including CO, O2, N2, H2O and CO210,11,12,13,14,15,16. Despite the recognized importance of these cationic promoters, an atomic-scale understanding of how alkali coordinate with adsorbates and reaction intermediates on metal surfaces remains incomplete. Low-temperature scanning tunneling microscopy (STM) and non-contact atomic force microscopy (nc-AFM) have emerged as powerful techniques for directly probing alkali ion–molecule interactions at submolecular resolution11,17,18,19. Interestingly, interfacial alkali metal cations could aggregate into one-dimensional chains through water-mediated attractive interactions18. The specific alkali ion-water coordination structures at metal surfaces influence the reactivities of water molecules20. Moreover, alkali metals promote the chemisorption of O2 on gold surface, forming K2O2/K2O complex and their aggregates, which could act as active sites in heterogeneous catalysis11,12,21. Thus, unraveling the precise coordination structure of alkali metals and reactant molecules in real-space is critical for elucidating the catalytic mechanism.

The activation and transformation of CO2 have long been a significant topic of extensive interest due to its crucial role in carbon neutral cycle and the production of valuable chemicals22,23,24. CO2 is chemically inert and adsorbs weakly on transition metal surfaces25,26,27,28. Alkali metals have been reported to facilitate the chemisorption of CO2 on metal crystals through electron donation29,30,31,32,33. This significantly facilitates the activation and reaction of CO2 under ultra-high vacuum (UHV) conditions, even below room temperature34,35,36,37,38. Previous studies, primarily relying on vibrational spectroscopy and X-ray photoelectron spectroscopy (XPS), have identified the formation of various negatively charged species, such as activated anionic CO2δ-, oxalate (C2O42-) and carbonate (CO32-) on alkali metal modified metal surfaces31,37,38,39. However, the precise mechanism of how alkali metals activate the CO2 molecules and stabilize the negatively charged intermediates remains controversial, particularly concerning whether alkali metals directly coordinate with reaction intermediates and whether single or multiple alkali are required for the activation of CO2. The lack of direct atomic-scale characterization on the coordination structures of alkali metal and the negatively charged intermediates has impeded a comprehensive understanding of the surface chemistry of CO2, which is critical for improving CO2 fixation efficiency and optimizing catalyst design.

In this work, we utilize STM and nc-AFM in combination with density function theory (DFT) calculations to investigate CO2 activation and reaction on alkali metal (K, Cs) modified Cu(111) and Cu(100) surfaces (Methods, Supplementary Note 1 and Supplementary Table 1). The alkali could assemble into trimer clusters with CO2 molecules on both Cu(100) and Cu(111) surfaces, in which CO2 is activated as CO2δ-, residing at the center of the trimer. Then we find the formation of oxalate on Cu(100) surface that is coordinated with four alkali ions. The generation of oxalate is further validated by tip manipulation experiment and theoretical calculations. Further increasing CO2 pressure leads to the formation of two-dimensional (2D) alkali carbonate film on copper surfaces, which is further confirmed by XPS. These results elucidate the pivotal role of aggregating a specific number of alkali metals in promoting the activation and reaction of CO2 on copper.

Results and discussion

Aggregation of alkali ions to stabilize anionic CO2 δ-

Figure 1a shows the STM image of Cu(111) surface covered with 0.08 ML Cs. The Cs adatoms ionize to Cs+ cations by transferring electrons to the Cu substrate. The ionization results in the dipole-dipole repulsion, which drives the Cs+ cations to arrange in an ordered hexagonal lattice40 with the distance of 0.88 nm (Fig. 1a). After depositing CO2 at the pressure of 7 × 10−6 mbar for 2 min at 300 K, Cs+ cations aggregated into trimers (Fig. 1b), in which the Cs+ cations are close-packed with a distance of ~5.6 Å. Similar Cs+ cation trimers were also observed on Cu(100) surface after dosing CO2 molecules (2 × 10−6 mbar for 5 min) on the Cs/Cu(100) surface at 230 K (Fig. 1c, e, Supplementary Note 1 and Supplementary Table 1). The Cs+ cations in the trimer could be directly resolved in the high-resolution AFM images, which are shown as round depressions at a large tip height (left panel of Fig. 1f, attractive force dominated region) and bright round protrusions at a small tip height (right panel of Fig. 1f, Pauli repulsion force dominated region)17,18.

Fig. 1: Formation of the alkali trimer coordinated with a CO2δ- radical anion on copper.
Fig. 1: Formation of the alkali trimer coordinated with a CO2δ- radical anion on copper.
Full size image

a Constant-current STM image of Cs/Cu(111) surface. b, c STM images of 3Cs+∙CO2δ- clusters on Cu(111) (b) and Cu(100) (c) surfaces. d Top and side views of the geometric model of 3Cs+∙CO2δ- cluster on Cu(100). Cs, Cu, C, O atoms are denoted by green, yellow, grey and red spheres, respectively. e–g Magnified STM image (e), constant-height AFM images (\(\Delta f\)) (f) and simulated AFM images (g) of the 3Cs+∙CO2δ- cluster on Cu(100). The black arrows in d, f, g denote the lower Cs+ at the bridge site of Cu(100) surface. Set point of STM images: −300 mV, 20 pA (a); 100 mV, 10 pA (b); 300 mV, 5 pA (c, e). The tip heights of experimental and simulated AFM images are marked on the images. The tip heights in (f) are referenced to the STM set point on the Cu substrate (100 mV, 10 pA). The tip heights in (g) are defined as the vertical distance between the apex atom of the metal tip and the outermost atom of the Cu substrate. The oscillation amplitude of the experimental and simulated AFM images is 100 pm. All the STM/AFM images were obtained at 5 K.

The aggregation of Cs+ and the formation of Cs+ trimers on copper suggests that a CO2 molecule is trapped in the center of the trimer cluster. DFT calculations reveal that the CO2 molecule is activated to a CO2δ- radical anion (Fig. 1d, Supplementary Note 2, Supplementary Figs. 1 and 2). The CO2δ- within the Cs+ trimer on Cu(100) (Fig. 1d) adopts a bent structure with an O-C-O bond angle of 117°, elongated C-O bond lengths (1.36 Å and 1.27 Å) compared to the 1.16 Å in linear geometry, and a delocalized extra electron in the antibonding orbital with a net charge of −1.3 |e | . The charge analysis results reveal the alkali coupled electron transfer from copper surface to the CO2 molecule (Supplementary Note 3, Supplementary Fig. 3 and Table 2), resulting in the formation of an activated CO2δ- anion, which is stabilized by non-covalent interactions with three alkali ions. The anionic CO2δ- chemisorbs on Cu(100) with one oxygen sitting between the cations and directly contacting with the copper surface, and the other oxygen coordinated with the Cs+ cation that adsorbed at the bridge site of Cu(100) surface (Fig. 1d). The higher lying O in the CO2δ- is visualized as a bright spot in the center of the trimer (Fig. 1f). The Cs+ cation at the bridge site shows a less bright contrast in the AFM image (black arrow in Fig. 1f) comparing with the other two Cs+ due to the slight height difference, which is confirmed by the force curve and DFT calculations (Supplementary Fig. 4). All those features in the experimental AFM images are nicely reproduced by the simulated AFM images (Fig. 1g) based on the DFT model shown in Fig. 1d. Moreover, other metastable trimer structures were also observed on Cu(100), which can readily undergo interconversion (Supplementary Note 2, Supplementary Figs. 2 and 5, Supplementary Table 3). The trimer structure on Cu(111) shows the identical adsorption configuration as that on Cu(100) (Supplementary Note 2 and Supplementary Fig. 1).

Additionally, we conducted the same experiment with K+ and observed the similar aggregation behavior of K+ into trimers on both Cu(111) and Cu(100) surfaces (Supplementary Fig. 6). However, the coverage of 3 K+ ∙ CO2δ- clusters is much lower than that of 3Cs+∙CO2δ- on both surfaces, suggesting the lower CO2 activation activity of K compared to Cs. In order to understand the prevalent aggregation of three alkali ions in stabilizing the anionic CO2δ- on copper surfaces, we have calculated the energies of 1 K+ + CO2, 2 K+ + CO2, 3 K+ + CO2, 4 K+ + CO2 on Cu(100) before and after aggregation into clusters. DFT calculations show that the total energy reduction caused by the aggregation of three K+ is maximal, making it the thermodynamically most stable configuration in stabilizing CO2δ- (Supplementary Fig. 7). This section highlights the aggregation of specific number of alkali ions on copper to activate and stabilize CO2δ- radicals, which could further serve as reactants, facilitating the conversion of CO2 into oxalate and carbonate that will be discussed in the following sections.

Formation of alkali-oxalate coordinated clusters on Cu(100)

To elucidate the role of alkali in CO2 conversion, we examined the reaction intermediates on alkali metal (K, Cs) modified Cu(100) surface at an elevated temperature of approximately 300 K (Supplementary Note 1 and Supplementary Table 1). In contrast with the trimer features that appeared at 230 K, room temperature induced the formation of homogeneous tetrameric structures for both K+ and Cs+, which present two different orientations with respect to the Cu(100) lattice (yellow and green arrows in Fig. 2a, Supplementary Figs. 8 and 9). High-resolution AFM images revealed a central rectangular feature surrounded by four lobes which are attributed to an oxalate anion (C2O42-), coordinated by four K+ cations (Fig. 2b). The central oxalate anion (C2O42-) is resolved as bright rectangle due to the repulsive interaction between the CO-terminated tip and the four negatively charged oxygen in oxalate at a small tip height (middle panel in Fig. 2b). Further decreasing the tip height would induce the pronounced lateral relaxation of CO molecule at the tip apex and the appearance of a bright sharp line along the central of the repulsive saddle between the oxygen atoms (bottom panel in Fig. 2b).

Fig. 2: Formation of oxalate anion on Cu(100) surface.
Fig. 2: Formation of oxalate anion on Cu(100) surface.
Full size image

a Large-scale STM image of the 4 K+ ∙ C2O42- clusters on Cu(100) surface (−200 mV, 50 pA). The yellow and green arrows denote two types of tetramers. b–f Constant-height AFM images (\(\Delta f\)) (b), simulated AFM images (c, d) and the structural models (e, f) of type-Ⅰ and type-Ⅱ tetramers. K, Cu, C, O atoms are denoted by purple, yellow, grey and red spheres, respectively. The white arrows in (b) denote the CO molecules adsorbed on Cu(100). The CO molecules adsorbed around the tetramer could stabilize the tetramer, protecting it from disturbances induced by the tip during scanning. The tip heights marked in (b) are referenced to the STM set point on the Cu substrate (100 mV, 50 pA). The tip heights marked in (c, d) are defined as the vertical distance between the apex atom of the metal tip and the outermost atom of the Cu substrate. g Frequency shift (\(\Delta f\)) measured as a function of bias above the K, C and O atoms in the 4 K+ ∙ C2O42- tetramer (amplitude: 100 pm, tip height: 100 pm).

Furthermore, DFT calculations confirmed the thermodynamical stability of 4 K+ ∙ C2O42- clusters on Cu(100). The oxalate anion adopts a planar D2h configuration that is stabilized through electrostatic interactions with four K+ cations occupying bridge (Type-Ⅰ) or top (Type-Ⅱ) site on Cu(100) (Fig. 2e, f and Supplementary Table 4). The distance between the diagonal K+ cations along the C-C bond is ~95 pm larger than that perpendicular to the C-C bond in the oxalate ion (Supplementary Fig. 9 and Supplementary Table 4). Each type of tetramer has two different orientations that are perpendicular with each other, which are clearly resolved in the high-resolution AFM images (Supplementary Fig. 9). These configurations were further validated by the good match of experimental and simulated AFM images obtained at different tip heights (Fig. 2b–d). Charge analysis revealed a net positive charge of +0.85 |e| for each K⁺ and a net negative charge of −1.8 |e| for the oxalate, which are discriminated by measuring the local contact potential difference (LCPD)41. The VCPD of the oxalate anion shifted by +150 meV with respect to potassium cations because of the reversed direction of the dipole induced by the charged ions (Fig. 2g). The VCPD of C and O atoms in the oxalate anion is almost indistinguishable due to the delocalization of the electron over the entire molecule.

In order to further confirm the formation of oxalate, we performed a controlled tip manipulation experiment on the 4 K+ ∙ C2O42- cluster. Figure 3a shows the STM images of the initial tetramer and the constructed trimer, dimer and monomer by removing the K+ from the tetramer sequentially with a voltage pulse at 2 ~ 4 eV. The zoom-in experimental and simulated AFM images and the optimized geometry models are shown in Fig. 3b–e. Removing one K+ leads to the side falling down of the oxalate ion (second panels in Fig. 3d, e, black arrows), so as to enhance the binding with the copper substrate. The slide-like configuration of the planar oxalate anion in the 3 K+ ∙ C2O42- cluster is visualized in the high-resolution AFM images (second panels in Fig. 3b, c and Supplementary Fig. 10), in which two lower oxygen atoms became invisible. Interestingly, when removing two K+, the remaining two K+ would lift the oxalate ion into a nearly upright configuration, in which the upward two oxygen atoms are vertically higher than the potassium cations (third panels in Fig. 3d, e, green arrows), and are visualized as a bright dimer feature (third panels in Fig. 3b, green arrow). Finally, the planar oxalate ion would fall down again when there is only one K+ (bottom panels in Fig. 3d, e) and the K+ was directly visualized as a bright sphere (bottom panels in Fig. 3b, c). Simulated AFM images demonstrated excellent agreement with experimental observations, thereby validating the identity of oxalate and its dynamic configurations under different cation coordination conditions. Moreover, we have considered the possibility of (CO2)2⁻, which is known to be stable in the gas phase42. However, under our experimental conditions, the (CO2)2⁻ anion appears to be unstable and undergoes transformation into oxalate (Supplementary Fig. 10). This behavior is attributed to the specific coordination environment and non-covalent interactions provided by the alkali metal trimer.

Fig. 3: Tip manipulation induced the adsorption configuration change of the oxalate anion on Cu(100) surface.
Fig. 3: Tip manipulation induced the adsorption configuration change of the oxalate anion on Cu(100) surface.
Full size image

a STM images of the tetramer and constructed trimer, dimer and monomer (from the top to the bottom panels, highlighted by dashed yellow circles, STM set point: 100 mV, 50 pA. b–e Constant-height AFM images (\(\Delta f\)) (b), simulated AFM images (c) and the structural models (d, e) of the 4 K+ ∙ C2O42-, 3 K+ ∙ C2O42-, 2 K+ ∙ C2O42-, K+ ∙ C2O42- clusters (from the top to the bottom panels). K, Cu, C, O atoms are denoted by purple, yellow, grey and red spheres, respectively. The tip heights of (b) are 30 pm, 40 pm, 110 pm, 40 pm (from top to bottom), which are referenced to the STM set point on the Cu substrate (100 mV, 50 pA). The tip heights of c are 10.73 Å, 11.21 Å, 12.35 Å, 10.88 Å, which are defined as the vertical distance between the apex atom of the metal tip and the outermost atom of the Cu substrate. The black arrows denote the lower lying oxygen atoms in the 3 K+ ∙ C2O42- cluster. The green arrows denote the upward oxygen atoms in the 2 K+ ∙ C2O42- cluster.

Reaction mechanism for the formation of oxalate

To figure out the CO2 reaction pathway and mechanism for the formation of oxalate on alkali modified Cu(100) surface, we performed climbing image nudged elastic band (CI-NEB) calculations on the conversion from activated CO2δ- to C2O42- (Fig. 4a, b). Experimental results have shown that the CO2δ- is initially coordinated with three alkali ions on copper (Fig. 1). So, 3 K+ ∙ CO2δ- can be regarded as the initial state, which inclines to adsorb an additional CO2 molecule with the oxygen atoms attached to the K+ cations (Fig. 4a, top panel and Supplementary Fig. 11). Subsequently, the linear CO2 within the alkali trimer would be activated into a CO2α- anion with a net charge of −0.5 |e| and C-C coupling happens, leading to the formation of C2O42- with a reaction barrier of 0.49 eV (Fig. 4a, b and Supplementary Fig. 12). For contrast, we have calculated the formation pathway of oxalate on Cu(100) without alkali and with only one K+, which present the reaction barrier of 1.18 eV and 0.55 eV, respectively (Fig. 4a, b). These results indicate that the reaction of CO2 to produce oxalate on Cu(100) surface is significantly boosted by K+ trimer, which shows the lowest reaction barrier. This is primarily attributable to the ability of K+ trimer to stabilize the key transition CO2δ- ∙ CO2α- (T3) structure (Fig. 4a, top panel) through cooperative K+-CO2δ-/CO2α- attractive interactions, effectively mitigating the Coulomb repulsion between the CO2δ- ∙ CO2α- anions.

Fig. 4: Calculated reaction pathway and barrier for the formation of oxalate on Cu(100).
Fig. 4: Calculated reaction pathway and barrier for the formation of oxalate on Cu(100).
Full size image

a Snapshots along the reaction paths with 3 K+, 1 K+ and without K+. K, Cu, C, O atoms are denoted by purple, yellow, grey and red spheres, respectively. The number n in the Tn denotes the (n + 1)th data point in (b). b Energy profiles for the formation of oxalate with 3 K+, 1 K+ and without K+. c STM images of the key intermediate trimers (3 K+ ∙ C2O42-). The yellow dashed circles show the conversion from trimer to tetramer by attaching one K+. Set point: 100 mV, 10 pA (left); 300 mV, 10 pA (right). d Energy profile for the diffusion of K+ to form a 4 K+ ∙ C2O42-. Snapshots along the reaction path are presented above the energy profile. The number n in the Tn denotes the (n + 1)th data point along the reaction path.

Moreover, the energy profiles of three different reaction pathways reveal that the final state of 3 K+ ∙ C2O42- represents the most energetically favorable configuration (Fig. 4b), which is also captured in the experiment (Supplementary Fig. 13). Figure 4c (left panel) shows the STM image of 3 K+ ∙ C2O42- clusters. We noticed that the STM tip would trigger the diffusion of isolated K+, which attached to the trimer (3 K+ ∙ C2O42-), and subsequently led to the formation of tetramer (4 K+ ∙ C2O42-) (dash circles in Fig. 4c). Such a transition process is energetically barrierless as supported by theoretical calculations (Fig. 4d and Supplementary Fig. 12) and is driven by the final state due to the favorable geometric match between the Cu(100) lattice and the square-like arrangement of the 4 K+ ∙ C2O42- clusters. The tetramer structure represents the most stable configuration for oxalate on Cu(100), consistent with experimental observations with the tetramer being dominant. However, we did not observe the oxalate intermediates on Cu(111) in our current experiment, despite the previous report of potassium oxalate formation through vibrational characterization38. One possible reason is that oxalate only appears at specific CO2:K ratios, making it difficult to capture experimentally with STM/AFM. Another possible reason could be the instability of oxalate intermediates on Cu(111), which tend to dissociate and further transform into carbonate (see the next section).

Formation of 2D alkali carbonate film

We further increase the pressure of CO2 to explore the reaction products on alkali metal (K, Cs) modified Cu(100) and Cu(111) surfaces (Supplementary Note. 1 and Supplementary Table. 1). Interestingly, STM images revealed the formation of two dimensional (2D) periodic films on both Cu(100) and Cu(111) surfaces (Fig. 5a, b and Supplementary Figs. 14–17). According to the XPS measurement (Supplementary Fig. 18), the resolved peaks of 531.4 eV in the O 1 s spectrum and 289 eV in the C 1 s spectrum39,43 enable us to assign the 2D lattice to the formation of alkali carbonate on copper surfaces. The potassium carbonate films exhibit the square-like lattice with monolayer height in the STM images (Fig. 5a, b and Supplementary Figs. 15 and 17). In contrast, high-resolution AFM images could directly identify the positions of K+ cations and CO32- anions. The CO32- anions appeared as a bright square lattice due to repulsive interactions between the CO32- and the negatively charged CO-terminated tip (Fig. 5c, d, white dash circles). Additionally, the features of K+ cations change from depressions to round protrusions with decreasing tip height (Fig. 5c, d). Such a contrast inversion behavior is quite prevalent in height-dependent AFM images of various alkali compounds18 (Fig. 1f and Fig. 2b).

Fig. 5: Formation of alkali carbonate islands on copper surfaces.
Fig. 5: Formation of alkali carbonate islands on copper surfaces.
Full size image

a, b STM images of the KCO3 islands on Cu(100) (a) and Cu(111) (b) surfaces. Inset of (b) shows the zoom-in STM image of KCO3 on Cu(111). c Constant-height AFM images (\(\Delta f\)) of KCO3 on Cu(100) at the tip heights of 100 pm (upper panel) and 20 pm (lower panel). d Constant-height AFM images (\(\Delta f\)) of KCO3 on Cu(111) at the tip heights of 40 pm (upper panel) and 10 pm (lower panel). The tip heights of (c, d) are referenced to the STM set point on the Cu substrates (100 mV, 20 pA). The white dashed circles in (c, d) denote the features of CO32- in the AFM images. e, f Geometric models of KCO3 islands on Cu(100) (e) and Cu(111) (f) surfaces. K, Cu, C, O atoms are denoted by purple, yellow, grey and red spheres, respectively. The K+ column that is highlighted by green arrows is ~6 pm vertically higher than that denoted by yellow arrows.

Based on these experimental results, we constructed the initial structures of KCO3 on Cu(100) and Cu(111) surfaces and the optimized geometric models are displayed in Fig. 5e, f (Supplementary Note 4 and 5, Supplementary Fig. 19). The K+ and CO32- ions arrange themselves in a checkerboard-like configuration, in which each K+ is surrounded by four CO32- and vice versa. DFT calculations revealed that the carbonate anions in the 2D KCO3 lattice sit on the Cu(100) surface adopting an upright bidentate configuration with one oxygen pointing upward away from the surface (Fig. 5e). As a result, the higher protruding oxygen atoms of the carbonate anions are visualized as small bright spots in the AFM images (denoted as dash circles in Fig. 5c, d). The KCO3 film on Cu(100) shows the perfect square lattice with the constant of 4.3 Å (Fig. 5e), in which the molecular planes of adjacent carbonate anions are perpendicular to each other, and all the K+ cations are identical, possessing the identical brightness in the AFM images (Fig. 5c, lower panel). In contrast, the K+ cations in the periodic KCO3 lattice on Cu(111) present different brightness in the AFM image (Fig. 5d, lower panel). This contrast difference is primarily attributed to the vertical height variation of K+ cations caused by the rotation of the surrounding carbonate anions (Supplementary Note 5 and Supplementary Fig. 20). The experimental height-dependent AFM images are perfectly reproduced by the AFM simulations based on the geometric models (Supplementary Figs. 15 and 17). Moreover, we have realized the atomic-resolution imaging of cesium carbonate on copper surfaces with nc-AFM and the atomic structures were determined unambiguously in combination with DFT calculations and AFM simulations (Supplementary Note 4 and Supplementary Figs. 14, 16, 21).

Interestingly, the 2D alkali carbonate films on both Cu(100) and Cu(111) surfaces are identified as non-stoichiometric KCO3 and CsCO3 structures with periodic square-like lattice, in which the K+/Cs+ and CO32- ions are arranged in a checkerboard configuration. Such a non-stoichiometric compound is stabilized by the copper substrate due to the interlayer charge transfer, in which the K+/Cs+ and CO32- ions have a net charge of +0.88 |e| and −1.47 |e | , respectively, and the copper surface is positively charged (Supplementary Table 5). The square alkali carbonate ionic lattice deviates from the hexagonal symmetry of Cu(111) substrate owing to the strong electrostatic attraction between the alkali cations and carbonate anions. It’s worth mentioning that a number of structures of alkali carbonate films on Cu(111) and Cu(100) surfaces have been excluded because of the inconsistency with the experimental results (Supplementary Note 4, Supplementary Figs. 19 and 21). In addition, isolated carbonate anions were observed on Cu(111) surface (Supplementary Fig. 22), further corroborating the production of carbonate at higher CO2 pressures. The reaction mechanism of CO2 to form alkali carbonate complexes on Cu(111) and Cu(100) surfaces is discussed in Supplementary Note 6, Supplementary Figs. 23 and 24.

Alkali metals on Cu(100) exhibit higher reactivity than on Cu(111). On Cu(100), CO2 activation occurs at low temperature, and the CO2 pressure required for the formation of alkali carbonate films is also lower compared to Cu(111). Furthermore, alkali metals on Cu(100) surface can promote the CO2 dissociation and the produced chemisorbed oxygen are observed by XPS and STM/AFM (Supplementary Figs. 18 and 23), which is absent on Cu(111). We also compared the reactivity of different alkali metals, revealing the trend Cs>K>Na (Supplementary Note 6). STM/AFM experiments with Na detected only the most stable carbonate species (Supplementary Fig. 25), while other intermediates (such as CO2-, C2O42-) were not observed, possibly due to their instability and short lifetimes on the surface.

This work reveals the aggregation of specific number of alkali metals in promoting the activation and reaction of CO2 on copper surfaces. In combination with STM/AFM and DFT calculations, we have demonstrated that alkali ions aggregate into trimers to facilitate the activation of CO2 into CO2δ- anion radicals. Alkali trimers can stabilize the chemisorbed CO2δ- and further assist the adsorption and activation of an additional CO2 molecule, leading to the formation of oxalate on Cu(100), which is stabilized by four alkali ions. The cooperative interactions between alkali trimer and CO2δ- anions effectively overcome Coulomb repulsion, enabling the formation of oxalate at a significantly reduced reaction barrier. At higher CO2 pressures, alkali metals promote the formation of alkali carbonate, in which the alkali cations and carbonate ions are assembled into ordered 2D films. Alkali metals on Cu(100) exhibit higher reactivity than on Cu(111) and can facilitate the dissociation of CO2 into CO and oxygen. This suggests that the activation and transformation of CO2 might be more favorable on transition metals or oxide surfaces that offer stronger interaction with CO2 and facilitate charge transfer. The incorporation of alkali metals could not only enhance the reaction activity, but also modulate the reaction pathways6. Elucidating the adsorption configurations and coordination structures of CO2-derived intermediates with both alkali metals and the underlying substrate is essential for understanding the reaction mechanism.

Our findings provide atomic-scale insights into the mechanisms underlying CO2 activation and transformation on alkali metal modified copper surfaces. These results not only provide valuable guidance for designing efficient catalysts in carbon capture, utilization, and storage technologies but also lay a solid foundation for advancing CO2 conversion in heterogeneous catalysis. An increasing number of studies have demonstrated the feasibility of integrating electrochemical cell with UHV characterization facilities44,45,46,47,48. This offers a valuable opportunity to investigate electrochemical processes using advanced surface science techniques and to elucidate the atomic-scale structure of electrode-electrolyte interfaces.

Methods

Sample preparation

The Cu(111) and Cu(100) single crystal were cleaned by several cycles of 1.0 keV Ar+ sputtering followed by annealing at 830 K. Alkali metals (K/Cs) were deposited on copper via a commercially available alkali metal dispenser (SAES Getters). The high-purity CO2 gas (99.999%) was dosed to the alkali metal modified copper surfaces through a precision leak valve with a dosing tube pointing towards the sample with a distance of about 7 cm in the preparation chamber and about 3 cm in the load-lock chamber of STM/AFM system. The alkali carboxylate and alkali oxalate clusters were prepared by dosing CO2 in the ultra-high vacuum (UHV) preparation chamber of the STM/AFM system. The alkali carbonate films were prepared by dosing CO2 in the UHV load-lock chamber of the STM/AFM system. The experimental details are described in Supplementary Note 1 and Supplementary Table 1.

STM and AFM experiments

All the experiments were carried out with an UHV Scienta Omicron POLAR-STM/AFM combined system operated at 4.8 K using a qPlus sensor equipped with a W tip (resonance frequency f0 = 24.7 kHz, spring constant k0 ≈ 1.8 kN m−1, quality factor Q ≈ 60000). All of the STM images were obtained in constant-current mode and the AFM images were obtained in constant-height mode with the oscillation amplitude of 100 pm at 4.8 K. All of the STM and AFM images were acquired with a CO-terminated tip. The CO-tip was obtained by positioning the tip over a CO molecule adsorbed on the Cu(100) and Cu(111) surfaces, followed by ramping the bias around 3.2 V and 2.5 V, respectively. The STM/AFM images were processed by Nanotec WSxM49. All of the STM and AFM images are row data without filtering.

XPS experiments

All the XPS experiments were performed in a UHV chamber Scienta Omicron XPS Lab System. This system is equipped with Argus CU analyzer and dual Al Kα anode X-ray source (resolution ∼0.5 eV) with a photon energy of 1486.7 eV. The samples were prepared in POLAR-STM/AFM system and transferred to the UHV-XPS system through a UHV suitcase with the base vacuum of 1 × 10−10 mbar. The XPS energies were calibrated by setting the Cu 2p3/2 binding energy position to 933.0 eV. The spectra were analyzed with a commercial CasaXPS software and were fitted with the combined Gaussian-Lorentzian shape after background (using a Shirley-type baseline) removal.

DFT calculations

DFT calculations were performed using the MedeA-VASP (version 5.4.4) software package50,51. Projector augmented wave pseudopotentials52 were used with an energy cutoff of 550 eV to expand the electronic wave functions. Van der Waals corrections for dispersion forces were taken into consideration using the optB86b-vdW functional53,54. The Cu(111) substrate was constructed with a four-layer slab model, while the Cu(100) substrate was constructed with a three-layer slab model, and the lattice constant of Cu was set to 3.615 Å. Cu atoms in the substrate apart from that in the top layer were fixed during geometry optimizations. The thickness of the vacuum slab was no less than 15 Å and the force criterion used for geometry optimizations was 0.01 eV/Å. Energy barriers and reaction paths were calculated using the CI-NEB method55 with a force criterion of 0.1 eV/Å. Geometric models shown in the figures (both main text and Supplementary Information) were produced by VESTA56.

AFM simulations

The probe particle model method as described in refs. 57,58 was used to simulate the AFM images. For the probe-particle tip model, we utilized the following parameters: the effective lateral stiffness k = 0.75 N/m and atomic radius Rc = 1.661 Å. The CO-tip was simulated by a quadrupole-like charge distribution at the tip apex59 with Q = −0.20 e (e is the elementary charge, and Q is the magnitude of quadrupole charge at the tip apex). The input electrostatic potentials of different reaction intermediates of CO2 on Cu(100) and Cu(111), employed in AFM simulations, were obtained from DFT calculations. The tip height in the AFM simulation is defined as the vertical distance between the metal tip apex and the topmost Cu layer. The oscillation amplitudes of the tip were 1 Å. Parameters of Lennard-Jones pairwise potentials for all elements are listed in the Supplementary Table 6.