Cation Effects on the Adsorbed Intermediates of CO2 Electroreduction Are Systematic and Predictable

The electrode–electrolyte interface, and in particular the nature of the cation, has considerable effects on the activity and product selectivity of the electrochemical reduction of CO2. Therefore, to improve the electrocatalysis of this challenging reaction, it is paramount to ascertain whether cation effects on adsorbed intermediates are systematic. Here, DFT calculations are used to show that the effects of K+, Na+, and Mg2+, on single carbon CO2 reduction intermediates can either be stabilizing or destabilizing depending on the metal and the adsorbate. Because systematic trends are observed, cation effects can be accurately predicted in simple terms for a wide variety of metals, cations and adsorbed species. These results are then applied to the reduction of CO2 to CO on four different catalytic surfaces (Au, Ag, Cu, Pd) and activation of weak-binding metals is consistently observed by virtue of the stabilization of the key intermediate *COOH.


INTRODUCTION
In the transition toward carbon neutrality, many large CO 2 emitting industries will face difficulties to moving swiftly and smoothly to utilization of alternative sources of energy.Therefore, effective and selective electrochemical CO 2 reduction reaction (CO 2 RR) electrolyzers will be an important technology for transforming and valorising CO 2 emissions. 1,2espite its great potential, the industrial application of CO 2 RR electrolyzers is yet to be realized due in part to the large number of possible products.−10 Density functional theory (DFT) calculations have been used in the past to develop a deeper understanding of experimental results.For example, on Cu(100) at low overpotentials, mainly C 2 products are observed in experiments, and the formation of ethylene is shown to be independent of pH. 11,12Results from DFT calculations showed that at low overpotential, the formation of an adsorbed C 2 O 2 intermediate (which was observed in its hydrogenated form in FTIR experiments), via a decoupled proton−electron transfer step, is the rate-determining step for the reduction of CO to ethylene on this surface. 13,14Further calculations proposed that at high overpotential, the reaction mechanism changes and the C 2 intermediate is formed from the coupling of *CO and *CHO 15 or the dimerization of *CH 2 .−19 DFT screening of various facets of transition metals was able to show that *COH and *CHO are stabilized differently by various surfaces and so different strategies are required to stabilize each intermediate. 20,21In addition, the selectivity between CO and HCOOH is intricate and has been postulated to depend on a number of factors. 22,23All this highlights the difficulty in optimizing the CO 2 RR since there may be no single approach which will stabilize all intermediates appropriately, which is aggravated by scaling relations between adsorbed intermediates. 24caling relations are the basis of the most widely used method of screening many surfaces for catalytic purposes. 25,26he adsorption energy of similar intermediates correlates in a linear fashion, for example, *OH and *OOH which are key intermediates in oxygen catalysis, 27−29 and C, S, N and O with their respective hydrogenated forms and other species. 25,30,31hese scaling relations allow for the elaboration of Sabatiertype volcano plots where the most active catalysts sit at the apex.However, scaling relations are thought to cause intrinsic limitations to the minimization of the overpotential for reactions in which more than two electrons are transferred due to the constant separation of energy between intermediates that scale together with a unity slope. 28−37 There are multiple ways to include solvent and/or electrolyte effects to try to improve the predictions of DFT calculations. 38There are two commonly used methods to mimic the effect of the surrounding solution.−46 Explicit solvent models include water molecules in the calculations and, although these models have a greater accuracy, they are highly demanding in computational terms. 44An affordable approach is "micro-solvation", in which a few explicit solvent and/or electrolyte species are included around the adsorbate.−48 The composition of the electrolyte can have many effects on the selectivity and the activity of the CO 2 RR in aqueous 49,50 and non-aqueous electrolytes. 7,51Previous works studied the electrochemical conversion of CO 2 in concentrated Mg-(ClO 4 ) 2 and NaOH brines at subfreezing temperatures for applications on Mars and observed drastic changes in the selectivity with respect to experiments at room temperature.While the changes were mainly attributed to the temperature effect, the influence of the identity and charge of the cation, as well as the difference in hydration layers cannot be disregarded. 52In fact, it has been observed that the size of the cation can tune the selectivity of the reaction in aqueous electrolytes.Previous reports have shown that on Ag electrodes, larger cations promoted CO production over H 2 production 53 and increased the current density due to formation of CO and HCOO − . 54,55On Cu electrodes, *K and *I modify the adsorption energies of *CO and its first hydrogenation products, 56 and increasing cation size also favors CO 2 reduction over H 2 evolution as well as promoting C 2 products rather than C 1 . 57DFT calculations have provided an explanation by showing that larger cations provide greater stabilization to C 2 intermediates. 58However, the reason for this stabilization is a subject of much debate.Hori and coworkers 57,59 hypothesized that the cations adsorbed to the electrode surface changing the potential profile of the electric double layer.In this regard, smaller cations such as Li + are strongly solvated and not able to adsorb to the surface, whereas larger cations such as Cs + , which are weakly solvated, can adsorb to the surface more easily and have a greater effect on the potential.In more recent studies implementing attenuated total reflection surface-enhanced infrared spectroscopy (ATR-SEIRAS), Ayemoba and Cuesta 60 showed that larger cations are more effective at buffering the pH since the polarization of water in the solvation shell is greater, closer to the electrode surface.Such an increased buffering capability lowers the pH near the electrode surface, favoring an increase of the CO 2 solubility. 61−64 In perspective, the afore cited works showed various separate cases in which cation effects were present but did not ascertain whether and how cation effects are systematic.Herein, DFT calculations incorporating explicit water molecules and cations in the vicinity of the adsorbed species on metal electrodes are used to investigate the trends in stabilization of key CO 2 RR intermediates by metal cations.First, we show by means of scaling relations the qualitative and quantitative variations in the way cations affect the intermediates of CO 2 RR.Second, by virtue of the systematic trends, we illustrate how it is possible to predict cation effects in various ways.Finally, we show the activation of otherwise inactive sites by cations during CO 2 reduction to CO via *COOH stabilization on several experimentally relevant electrodes.

METHODS
All calculations were performed using the Vienna ab initio simulation package 65 using the PBE exchange-correlation functional 66 and the projector augmented-wave (PAW) method 67 to describe the ion-electron interactions.Previous works showed that PBE and RPBE 68 cannot simultaneously predict with reasonable accuracy the reaction energy and onset potentials of CO 2 RR to CO on metals. 69While errors in the reaction energy are similar (∼0.4 eV), errors in the onset potentials are larger for RPBE.Gas-phase corrections help both functionals match the experimental reaction energy and predict onset potentials close to experiments.Hence, the choice of PBE or RPBE is facultative when gas-phase corrections are used.Extensive analyses with more reactions, functionals, and adsorbates can be found in the literature that confirm our assessment. 70Geometry optimization was carried out using the conjugate gradient method with a plane wave cut-off of 450 eV.For calculations of slabs with adsorbates, relaxation was stopped when the maximal residual forces on all atoms was less than 0.05 eV/Å.The Methfessel−Paxton method 71 was used to sample the Brillouin zone with an electronic temperature of 0.2 eV and all energies were then extrapolated to 0 K.The vertical distance between slabs was greater than 12 Å and dipole corrections were added to avoid spurious electrostatic interactions between vertically stacked images.
The gas-phase energies of species (E A(g) , where A(g) is C, CH, CH 2 , CH 3 , CH 4 , CO, COOH, COH, CHO) were used as a reference for the corresponding adsorption energies: , with E * : energy of the clean surface, E *A : energy of the surface with the adsorbate A. In presence of a cation: , where E *cat is the energy of the surface with the cation (cat: K, Na, Mg), and the brackets indicate co-adsorption in close proximity.The values of E A(g) were either taken from the literature or calculated spinunrestricted and using Gaussian smearing with an electronic temperature of 0.001 eV and again, all energies were extrapolated to 0 K. 21 For gas-phase calculations a 15 Å box was used and the Brillouin zone was sampled using a (1 × 1 × 1) Monkhorst−Pack mesh. 72.1.Reaction Energies.For the reduction of CO 2 to CO, the reaction energy, ΔG, in vacuum was calculated using eq 1.
where ΔE DFT is the calculated DFT reaction energy, ΔE ZPE is the zero-point energy change, T is the absolute temperature (set to 298.15 K), and ΔS is the entropic contribution.ΔE ZPE was calculated with DFT using the vibrational frequencies obtained using the harmonic oscillator approximation.For gas and liquid molecules, ΔS was taken from standard entropies reported in the literature. 73A liquid-phase correction was applied to ΔS for water. 13,73,74For adsorbed intermediates, ΔS includes only the vibrational contribution.Heat capacity effects were not included since their effects are negligible on formation energies at 298.15 K. 75 To account for the errors intrinsic to DFT calculation of gas phase energies, for CO 2 and CO specifically, semiempirical corrections have been applied. 69hen water solvation or cationic corrections are applied, it is assumed that any changes in ΔE ZPE and ΔS are negligible. 46his was verified for the co-adsorption of *K and *CO on Ag(111) where ΔE ZPE and ΔS were calculated and applied and the difference was shown to be less than 0.05 eV.

Effect of Water Solvation on CO 2 Reduction to CO.
When investigating the effect of solvation by water on the electroreduction of CO 2 to CO, (3 × 3) slabs were used of the (111) surface.Four layers of atoms were used where the bottom 2 layers were fixed at the optimized bulk distance.Again, a Monkhorst−Pack mesh was used to sample the Brillouin zone but with grid size (4 × 4 × 1).For explicit solvation, stepwise addition of H 2 O molecules was used. 46,48he additional stabilizing energy provided by each additional water molecule solvating the CO 2 reduction intermediate A,  111) slabs were used.These were large enough to avoid lateral interactions between adsorbates, as shown in Table S11 and Figure S7.Due to computational resource limits, 3 layers of atoms were used, where the bottom layer was fixed at the optimized bulk distances and the top two layers and any absorbates were allowed to fully relax.The most stable coadsorption configurations for *K and C 1 species on Cu are provided in Section S5 and Figure S8.To find those configurations we ran several calculations using the configuration in vacuum and displacing the cations around the adsorbate.The Brillouin zone was sampled using a (3 × 3 × 1) Monkhorst−Pack mesh.Ag, Au, Co, Cu, Ni, Ir, Pt, Pd and Rh surfaces were used in their spin-restricted face-centered cubic (fcc) configuration.Previous work shows that this configuration is acceptable for Co and that, although there are differences in the spin-restricted and unrestricted values for Co and Ni surfaces, the trends are unaffected. 21The stabilization of each intermediate by a co-adsorbed cation is modeled using eq 2.

* + * * [ + ] + *
Again, the brackets indicate co-adsorption in proximity.And so, the stabilization energy provided by *cat, Ω *cat , is given by eq 3: This definition of cation effects is analogous to that of solvation effects (see Section 2.2) and enables a direct comparison of Ω A n , Ω Hd 2 O and Ω *cat .

RESULTS AND DISCUSSION
3.1.Scaling Relations.Scaling relations for the adsorption energies of CO 2 RR intermediates are shown in Figure 1.In addition, Figure S1 presents the same data plotted with the same scale on all panels, Figure S2 contains labeled data points, and Figure S3 classifies the data depending on the presence or absence of co-adsorbed cations.The results for each intermediate in those figures are plotted vs ΔE C and the trends are described using linear regressions the general formula of which is provided in eq 4.
The slope or gradient of the line can be predicted using a simple electron counting relation, m = N A /N C , where N A is the number of electrons required for an intermediate A to reach a full outer shell and N C is the number of electrons for C to complete its outer shell, that is 4. 25,30 This relation is applicable in many situations where the interaction between the surface and each species is similar.In the light of this, stepwise hydrogenation progressively lowers the gradient, for example for CH x vs C scaling relations, the predicted slopes are 0.75, 0.50, and 0.25 for x = 1, 2, 3 (see Table S1). 25In addition, atoms with the same number of valence electrons (e.g., C and Si) scale linearly with a gradient of 1. 30 However, other factors can influence the slope such as solvation or covalent character in the interaction between the adsorbed intermediate and the surface. 35Furthermore, scaling relations may or may not exist depending on the chemical nature of the adsorbates and the composition and coordination number of the adsorption sites. 77Table S1 in the Supporting Information contains the slopes, offsets, correlation coefficients, mean and maximum absolute deviations (MADs and MAXs) of all linear fits in Figure 1.The overall MAD and MAX are 0.09 and 0.39 eV.We note that *Mg co-adsorption generally lowers the correlation coefficients and increases the MADs and MAXs.
Figure 2 condenses all the slopes and offsets of the scaling relations in Figure 1.The data in vacuum, with *Na and *Mg are presented as a function of the data with *K.As discussed in the next paragraphs, Figure 2 shows that the slopes are linearly related, and so are the offsets, which is why we posit that cation effects are systematic and predictable among C 1 species on transition metals.
For the intermediates alone on the surface, shown in blue in Figure 1, the slopes follow the literature and abide by the expected valence electron relationship.As shown in the inset of Figure 2, intermediates where N A is 3, (*CH and *COH) have gradients close to 0.75, where N A is 2 (*CH 2 , *CO, and *CHO) the slope is close to 0.5, and where N A is 1 (*CH 3 and *COOH) the slope is close to 0.25 see Table S1 for details.More importantly, the inset of Figure 2 and Figure S4 show that upon co-adsorption of *K, *Na, and *Mg, cation effects modify the scaling relations observed in vacuum: the slopes of *COH and *CO increase, those of *CH 2 , *CHO, *COOH decrease, and those of *CH and *CH 3 remain approximately constant (except for *CH 3 with *Mg).These changes in the slopes suggest that N A is partially modified by the interaction of the adsorbates with the cations.For instance, for CO, as the electronegativity of the cation increases, the slope increases from close to 0.5 to close to 0.75 suggesting a loss of the double bond character in *CO and a change from 2 to 1 electron needed to complete the valence shell.Previous work has shown that for CO the stretching vibration, which is inversely proportional to the bond strength, is affected by approaching cations, which supports the change in slope and, hence, in relative bond orders, observed here. 58,78,79ppreciable changes are also observed in the offset upon addition of a co-adsorbed cation, as shown in Figure 2 and Figure S5.The offsets of the scaling relations are linearly related in presence and absence of cations and some offsets are made more negative than in vacuum by cation co-adsorption and some others are made less negative.Interestingly, the variations have the same direction as those of the slopes: the offsets of *COH and *CO increase, those of *CH 2 , *CHO, *COOH decrease, and those of *CH and *CH 3 remain approximately constant (except for *CH 3 with *Mg).In principle, Figure 2 and Figures S4 and S5 imply that cation effects can either stabilize or destabilize the adsorption of a given species on a series of metals, and the magnitude and direction of the effects can be predicted.Our main conclusion from Figure 2 is that cations can be used to modify adsorption energies without changing the morphology of the active sites.As will be shown later, this can be used in electrocatalysis to activate or deactivate specific active sites.
Figure 3a,b are parity plots to illustrate how the stabilization varies depending on the cation.Figure 3a shows how the difference in binding energy between the sole intermediate and the intermediate in the presence of *K, Ω *K (see eq 3), compares to the presence of *Na, Ω *Na , for each of the intermediates.If the ions affect the intermediates in the same manner, then there will be a linear relationship between Ω *K and Ω *Na with a slope of 1 and a null offset.To show the expected relationship, the parity line y = x is included with an error band around it of ±0.10 eV.For these two cations the mean absolute deviation (MAD) is 0.03 eV.We observe that all the intermediates, apart from *COOH, are close to the parity line, indicating that they are stabilized in a similar way by both *K and *Na for all metals.*COOH is stabilized more by *Na than *K by 0.10 eV, on average, and the effect is essentially constant for all metals.
The same analysis was carried out to determine the correlation between stabilization by *Mg and *K and the results plotted in Figure 3b.Here, the MAD was 0.18 eV, and we observe that the deviations grow as Ω *K becomes increasingly negative.For each intermediate, the parameters for the linear fit for Ω *Mg vs Ω *K in Figure 3b are shown in Table S2.Altogether, Figure 3a,b clearly indicates that potassium and sodium stabilize the adsorbates approximately in the same way, while magnesium affects the intermediates differently.Hence, one can easily predict the adsorbate stabilization provided by *Na based on the one provided by *K and vice versa, but this is not the case for the stabilization granted by *Mg.
This poses an important question: Is it possible to predict the effects due to *Mg co-adsorption based on those of *K and/or *Na?Seeking an answer, we resorted to four approaches: (i) a linear regression of the data in Figure 3b; (ii) a correlation of the Ω *Mg vs Ω *K parameters with those of Ω *Na vs Ω *K (see Table S2), which are found in Figure 4a,b to be linearly related; (iii) a multivariate regression of Ω *Mg as a function of Ω *K and Ω *Na ; and (iv) per-adsorbate multivariate regressions using 3 input variables, namely Ω *K , the number of valence electrons of the transition metals (9 for Co, Rh, Ir; 10 for Ni, Pd, Pt; 11 for Cu, Ag, Au) and their d-series (3d, 4d, 5d).
Figure 4c shows that the linear fit is the least accurate of the three approaches, and the MAD of the predictions is 0.17 eV while the maximum absolute deviation (MAX) is 0.65 eV.Only 65% of the data are within ±0.20 eV of the parity line in Figure 4c.This was expected judging by the large departures from the parity plot in Figure 3b.Furthermore, capitalizing on the linear relationships in Figure 4a,b and the parameters in Table S2, the MAD can be lowered to 0.11 eV and the MAX to 0.46 eV.Now, 83% of the data are within ±0.20 eV of the parity line in Figure 4c.Similar accuracy, namely MAD/MAX of 0.10/0.47eV, is delivered by a simpler multivariate regression using Ω *Na and Ω *K as inputs.In addition, 85% of the data are within ±0.20 eV of the parity line in Figure 4c.Multivariate regressions using Ω *K , the number of valence electrons of the transition metals and the location in the dseries lead to MAD and MAX values as low as 0.06 and 0.21 eV, see Section S4.In fact, 99% of the predictions are within ±0.20 eV of the parity line, and 86% are in the narrow range of ±0.10 eV of the parity line.In the light of Figure 4c, we conclude that it is possible to predict *Mg stabilization effects on C 1 species with reasonable accuracy by combining the respective data for *K and *Na or using the data for *K and two simple descriptors trivially obtained from the periodic table.3.2.Application to the Reduction of CO 2 to CO.The reduction of CO 2 to form CO via *COOH is the first stage common to the production of numerous CO 2 reduction products with one carbon atom with the exception of formic acid, which is typically formed via *OCHO. 50To illustrate the electrocatalytic effects of cation co-adsorption, in the following we compare the reduction of CO 2 to CO on four experimentally relevant surfaces with and without cations.Previous experimental work showed an increase of the CO 2 RR activity and selectivity on Cu, Pd and Au single-crystal electrodes at subzero temperatures in highly concentrated brines. 52While the authors proposed that the change in catalytic activity and selectivity was mainly associated with an increase of the CO 2 solubility at low temperatures, they also echoed on possible effects of the highly concentrated cations and the water availability at the interface. 52In addition to those metals, Ag has also been highlighted in Figure 5 given its academic and industrial relevance in the CO 2 RR to CO.The first proton−electron transfer is * + CO 2 + H + + e − → *COOH (ΔG 1 = ΔG COOH ).In turn, the second proton− electron transfer is *COOH + H + + e − → * + CO + H 2 O (ΔG 2 = ΔG 0 − ΔG COOH , where ΔG 0 is the overall reaction energy).The limiting potential is given as U L = −max (ΔG 1 , ΔG 2 )/e − .We emphasize that, as only electrochemical steps are considered in Figure 5, the volcano stands alone for metals that bind *CO weakly or moderately, such as Au, Ag and Cu.However, it is advisable to collate the volcano with the separate free-energy diagram for strong-binding metals, such as Pd.The free-energy diagrams for the four metals are provided in Figure S6.
According to Figure 5, the first proton−electron transfer is the potential limiting step when *COOH is in contact with *H 2 O for Au(111), Ag(111), and Cu(111) surfaces, which implies that they are all on the weak-binding side of the volcano.Cation stabilization of *COOH generally enhances the activity of Au(111), Ag(111) and Cu(111) and *Mg has a more pronounced effect than *Na and *K.Interestingly, stabilization by *Mg changes the potential limiting step on these three metals to the second electrochemical step.On the other hand, Pd(111) with *H 2 O is on the strong-binding side of the volcano and close to the apex, the presence of the three cations does not change the potential limiting step and in all cases increases the overpotential.In addition, we stress that the desorption of *CO to CO (g) is rather energetically unfavorable on Pd(111) with and without cations, as illustrated in Figure S6, such that Pd(111) is generally expected to be blocked by *CO.
The main piece of information we extract from Figure 5 is that cations activate weak-binding metals and facets for CO 2 electroreduction to CO.−82 Furthermore, only a few works have probed the effect of Mg 2+ during the CO 2 RR.Besides the well-known work by Monteiro et al. 83 in nearly acidic media on Au electrodes, in a previous work 52 the onset potential for the CO 2 RR on Cu electrodes in Mg(ClO 4 ) 2 brines was observed as low as −0.25 V vs RHE.Wang et al. 84 also reported an enhancement of the CO 2 RR on CoPc in the presence of Mg 2+ ions.Along the same line, Lyu et al. 85  To close this section, we emphasize that our calculations were carried out considering only low surface coverage of species on the (111) facet of transition metals.Future, more comprehensive works ought to systematically examine coverage effects of adsorbates, cations, and spectators on scaling relations and electrocatalytic activity predictions with structural sensitivity, as done elsewhere. 56,88,89

CONCLUSIONS
This work provides a trends-based view of cation effects on Cbased species.The presence of *K, *Na and *Mg changes the adsorption energies of C 1 species, and the magnitude and sign of the effect depend on the specific transition metals which they adsorb on.Because we found that the effects are systematic and gradually change the slopes and the offsets of adsorption-energy scaling relations, we conclude that cations change surface−adsorbate interactions both qualitatively and quantitatively.
Since the stabilization granted by *K and *Na is similar, a simple linear regression suffices to predict one provided that the other is known.However, the stabilization granted by *Mg is different from those of *K and *Na and the differences accentuate alongside the magnitude of the stabilization effect.In view of that, multivariate regressions were provided that enable the accurate prediction of stabilization due to *Mg in terms of (i) those of *K and *Na, and (ii) that of *K and two simple descriptors obtained from the periodic table.
Analyzing CO 2 electroreduction to CO by means of a Sabatier-type volcano plot and free-energy diagrams led us to conclude that cation effects activate weak-binding metals, such as Au, Ag, and Cu, and weak-binding facets such as the (111).The enhancement is because *COOH formation is usually energy intensive, but the presence of cations sizably lowers its potential requirements.
In summary, we have shown here that cation effects on transition metals follow systematic trends.This opens new perspectives in electrocatalysis, as cation effects are most often analyzed separately, that is considering only a specific cation, on a given facet of a specific metal.The fact that the effects are systematic gives hope for a straightforward incorporation of cation effects in materials screening routines, circumventing the need for demanding calculations with numerous atoms or advanced yet arduous treatments of the double layer.

Figure 2 .
Figure 2. Offsets of the scaling relations with and without coadsorbed cations shown in Figure 1 as a function of those with *K.Inset: slopes of the scaling relations in vacuum, with *Na, and with *Mg as a function of those with *K.Following the nomenclature of eq 4 and Table S1, the main panel depicts i A vac , i A * Na and i A * Mg vs i A * K , and the inset depicts m A vac , m A * Na and m A * Mg vs m A * K .The dashed lines are parity lines.

Figure 3 .
Figure 3. Energy stabilization provided by (a) *Na (Ω *Na ) and (b) *Mg (Ω *Mg ) plotted against the energy stabilization provided by *K (Ω *K ) for C 1 intermediates in the reduction of CO 2 on nine transition metals.Data points over the parity line y = x indicate that both cations influence the adsorbate identically.The MAEs are 0.03 and 0.18 eV for Ω *Na vs Ω *K and Ω *Mg vs Ω *K , respectively, and the gray bands correspond to ±0.10 eV around the parity line.

Figure 4 .
Figure 4. Correlations between (a) the slopes and (b) the offsets of the linear fits of Ω *Na vs Ω *K and Ω *Mg vs Ω *K .(c) Parity plot comparing the DFT-calculated stabilization granted by *Mg (Ω *K ) and the stabilization predicted by means of a linear fit (red), an equation based on the parameters in panels (a) and (b) (white), a multivariate regression using Ω *K and Ω *Na (blue), and multivariate regressions using Ω *K , the number of valence electrons of the transition metal and the d-series it belongs to (cyan).The equations of the regressions appear in Section S4.The gray band covers an area of ±0.20 eV around the parity line.

Figure 5 .
Figure 5. Volcano plot for the reduction of CO 2 to CO via *COOH on Au(111) (blue), Ag(111) (red), Cu(111) (orange), and Pd(111) (gray).Results in vacuum and with three different cations are provided for each metal.The data for Au(111) and Ag(111) with *H 2 O with water were taken from ref 69.

. Effect of Cations on C 1 Intermediates.
is compared to the stabilizing energy provided when the water molecule is solvated by other water molecules, Ω Hd 2 O , and if Ω A To investigate the effect of cations on CO 2 RR intermediates, (4 × 4) ( reported and enhancement in the CO 2 RR on CoPc anchored to Mg(OH) 2 substrates, which allegedly facilitate the formation of the *CO 2 − intermediate.Similarly, according to the in situ measurements in ATR configuration reported by Zhu et al., 86 during the CO 2 RR in carbonate solutions on polycrystalline Cu electrodes, the CO bands start appearing at potentials as low as −0.2 V vs RHE.Finally, we note that Seong et al. 87 observed the formation of *CO on Au sites during the CO 2 RR on AuCu catalyst in phosphate buffer solution (pH 7−2) at potentials as low as −0.3 V vs RHE.