Atomic Ensemble Effects and Non-Covalent Interactions at the Electrode–Electrolyte Interface

Cyanide-modified Pt(111) electrodes have been recently employed to study atomic ensemble effects in electrocatalysis. This work, which will be briefly reviewed, reveals that the smallest site required for methanol dehydrogenation and formic acid dehydration is composed of three contiguous Pt atoms. By blocking these trigonal sites, the specific adsorption of anions, such as sulfate and phosphate, can be inhibited, thus increasing the rate of oxygen reduction reaction by one order of magnitude or more. Moreover, alkali metal cations affect hydrogen adsorption on cyanide-modified Pt(111). This effect is attributed to the non-covalent interactions at the electrical double layer between specifically adsorbed anions or dipoles and the alkali metal cations. A systematic investigation is conducted on the effect of the concentration of alkali metal cations. Accordingly, a simple model that reproduces the experimental observations accurately and enables the understanding of the trends in the strength of the interaction between M and CNad when moving from Li + to Cs, as well as the deviations from the expected trends, is developed. This simple model can also explain the occurrence of super-Nernstian shifts of the equilibrium potential of interfacial proton-coupled electron transfers. Therefore, the model can be generally applied to explain quantitatively the effect of cations on the properties of the electrical double layer. The recently reported effects of alkali metal cations on several electrocatalytic reactions must be mediated by the interaction between these cations and chemisorbed species. As these interactions seem to be adequately and quantitatively described by our model, we expect the model to also be useful to describe, explain, and potentially exploit these effects.


Introduction
Among all the challenges faced by humankind in the 21st century, generating enough energy to keep our greatly energy-consuming modern societies functioning while causing a minimum impact on the natural environment is probably the central and most important one. Electrochemical energy conversion and storage offers high theoretical efficiency because it can escape the limitations imposed by the second law of thermodynamics to the efficiency of a thermal engine. However, the actual efficiency of an electrochemical device is directly associated with the kinetics of the electron-transfer reactions occurring at the anode and cathode. Thus, a strong link is created between electrochemical science and technology and catalysis. Electrocatalysis is the science and art of finding a material on which the electrochemical reaction of interest can occur at the fastest rate possible. Eventually, an efficient electrochemical device is created.
Electrocatalysis, like heterogeneous catalysis, involves the interaction of reactants, products and intermediates with the surface of the catalytic material, as the breaking and forming of bonds is necessary to transform the reactants into new chemical species. The rate at which an elementary reaction step proceeds on the surface of a given catalyst not only depends on the strength of these interactions with the catalyst surface (generally considered as electronic effects) but also on the availability of specific atomic groupings that can provide the number of surface atoms necessary for chemisorption and act as active sites for a reaction or a reaction step (atomic ensemble effects).
Another effect that has traditionally been neglected is that of electrolyte cations or generally of any species in the electrolyte that does not form strong covalent chemical bonds with the atoms of the catalyst surface. However, cations can become attached to the electrode surface through non-covalent interactions (e.g., hydrogen bonds, ionic bonds, van der Waals forces, and hydrophobic interactions, which do not involve the sharing of pairs of electrons [1,2]) with chemisorbed species. Thus, the structure and the properties of the interface are affected, and the processes occurring at the electrochemical double layer are influenced as has been shown for some important electrocatalytic processes, such as the methanol oxidation reaction (MOR) [3] and the oxygen reduction reaction (ORR) [3,4].
This brief review offers an overview of recent works in my group using model cyanide-modified Pt(111) electrodes to achieve a deeper understanding of the atomic ensemble effects and non-covalent interactions at the electrode-electrolyte interface.
Cyanide-modified Pt(111): a peculiar surface. Saturated, irreversibly adsorbed cyanide adlayers can be formed on Pt(111) upon immersion in a cyanide-containing solution [5][6][7][8][9]. Infrared studies on cyanide adsorbed on Pt(111) electrodes have consistently found a single vibrational band corresponding to the stretching vibration of a CN group adsorbed on top of a platinum atom through its carbon end [5,6,8,9]. Hubbard and co-workers [7] showed using low-energy electron diffraction (LEED) that saturated, irreversibly adsorbed cyanide adlayers on Pt(111) formed an ordered (2 3 x 2 3 )R30º structure that resists electrode emersion and transfer into an ultrahigh vacuum (UHV) chamber. The same structure was observed later using electrochemical scanning tunneling microscopy (EC-STM) in cyanide-free alkaline solutions [9] and in cyanide-containing acidic and alkaline solutions [10]. The (2 3 x2 3 )R30º structure consists of hexagonally packed hexagons, each containing six CN groups bonded linearly to a hexagon of Pt atoms through their carbon ends ( Figure 1). The Pt atoms around the CN hexagons remain uncovered and lie at the bottom of one-atom-wide troughs. The platinum atom at the center of the hexagon also remains uncovered, thereby yielding a coverage by CN ad of 0.5 ML.
As shown in Figure 1, cyanide-modified Pt(111) does not contain any sites composed of three contiguous platinum atoms. All the threefold-hollow adsorption sites have disappeared from the Pt(111) surface upon the formation of the cyanide (2 3 x2 3 )R30º structure, thus fulfilling one of the conditions (removing only one  of the several kinds of sites present on the surface) required for studying atomic ensemble effects using the site-knockout strategy [11]. The other condition, which requires the electronic properties of the free platinum atoms not bonded to the CN groups to remain unaltered or only negligibly affected, is also fulfilled as we will show below.
Both the underpotential-deposited hydrogen (H upd , i.e., hydrogen adsorbed at potentials more positive than the equilibrium potential for the hydrogen evolution reaction) and the adsorbed OH (OH ad ) regions are shifted in the positive direction in a cyanide-modified Pt(111) electrode compared with an unmodified Pt(111) ( Figure 2). The observed potential shift implies that the G 0 of H upd and OH ad at  = 0 are approximately 19 kJmol -1 more negative and approximately 24 kJmol -1 more positive, respectively, than that of an unmodified Pt(111) [12]. This result could be an indication of an electronic disturbance of the platinum atoms that remain uncovered by CN on a cyanide-modified Pt(111) electrode. However, the weakening of the Pt-OH ad can be most simply understood as being due to the electrostatic through-space repulsive interactions between the electronegative cyanide adlayer and the large negative dipole associated with OH ad [4].
Regarding the increase in G 0 of H upd , we have recently shown that it is due to the formation of (CN ad ) x -H clusters [13]. The formation of adsorbed hydrogen isocyanide (CNH ad ) was also argued by Schardt et al. [14] to explain the effect of pH on the amount of Cs + retained on the surface of a cyanide-modified Pt(111) electrode after immersion in a 0.1 mMCsCl solution.
The formation of (CN ad ) x -H clusters (where x must decrease with decreasing potential within the H upd region) can also explain the change in the Stark tuning rate of the CN stretching frequency of a cyanidemodified Pt(111) electrode occurring at potentials more negative than 0.6V vs. RHE (coinciding with the onset  Table 1 Figure 7B); the value given for CNH in column two corresponds to the Eads of a CNH molecule on Pt(111) calculated with Eq. 6 in reference [13], and the value given for H corresponds to Eads of an H atom on the N atom of an adsorbed CN moiety calculated with Eq. 5 in reference [13]. c) CN and H adsorbed on-top and at a site intermediate between bridge and hcp-hollow, respectively ( Figure 7C in reference [13]); the value given for CN in column two corresponds to the Eads of a CN radical on Pt(111) calculated with Eq. 1 in reference [13], and the value given for H corresponds to Eads of an H atom on a CN-free Pt atom (H = 1/9) of a Pt(111) surface covered by 1/9 ML of CN calculated with Eq. 4 in reference [13]. of H upd formation) [5,15] and the incomplete blocking of hydrogen adsorption after adsorption of CO to saturation on a cyanide-modified Pt(111) electrode [15][16][17].

. Adsorption Energy (E ads ), Structural Parameters, and Mulliken Charges for the CN Radical Adsorbed On-top on Pt(111) at Low Coverage, and for the H Atom Adsorbed on the Nitrogen Atom of the CN (Pt(111)-CNH) and at a Pt Site Intermediate between Bridge and Hcp-hollow (NC-Pt(111)-H) in the Presence of a Low Coverage of CN ad
The most solid proof of the lack of electronic effects on cyanide-modified Pt(111) electrodes, though, is provided by density functional theory (DFT) calculations of the density of states (DOS) and hydrogen adsorption energies on clean Pt(111) and cyanide-modified Pt(111) surfaces [13,18]. As shown in Figure 3, the DOS of the Pt atoms directly bonded to CN ad is shifted to lower energies as expected for strong bonding, but the DOS of the atoms that are not bonded to CN ad remains similar to that of the Pt atoms on the clean (111) surface. Similarly, as shown in Table 1, the adsorption energy of hydrogen on a cyanide-free Pt atom of a cyanidemodified Pt(111) is identical to that of hydrogen on an equivalent site of a Pt(111) surface within the error typical of DFT calculations. Furthermore, the transfer of a proton and an electron to the N atom of a CN group is clearly favored over adsorption on the cyanide-free atoms of a cyanide-modified Pt(111) surface.

Atomic ensemble effects on the electrocatalytic oxidation of C1 organic fuels and on the ORR.
Contrary to what is observed on Pt electrodes in general and on Pt(111) electrodes in particular, the cyclic voltammogram (CV) of a cyanide-modified electrode in a methanol-containing perchloric acid solution (Figure 4a) shows no change in the hydrogen adsorption region compared with the CV in the absence of methanol, and nearly no hysteresis occurs between the positive-and the negative-going sweeps [19]. These are two clear indications that no adsorbed CO is formed on the surface of the cyanide-modified Pt(111) electrode, as confirmed using infrared reflection-absorption spectroscopy (IRRAS) (Figure 4b) and later using differential electrochemical mass spectrometry (DEMS) [15]. These facts, along with the clear observation by IRRAS of CO 2 formation during methanol electro-oxidation on cyanide-modified Pt(111) electrodes ( Figure 4b, left panel), are a clear proof that a minimum of three contiguous Pt atoms (a trigonal site) is required for methanol dehydrogenation to adsorb CO, whereas two adjacent Pt atoms suffice for complete electrooxidation to CO 2 through the direct path. A similar conclusion was reached on the dehydration of formic acid [15]. Neurock et al. found the same result using DFT calculations [20].
The effect of removing the trigonal sites from the surface of a Pt(111) electrode on the ORR is even more dramatic. Figure 5 shows the polarization curves obtained using a rotating disc electrode (RDE) in O 2saturated 0.1 M HClO 4 , 0.05 M H 2 SO 4 , and 0.05 M H 3 PO 4 solutions. In 0.1 M HClO 4 , barely any difference can be observed between the activity of Pt(111) and that of a cyanide-modified Pt(111) electrode. This result is expected because perchlorate adsorbs only weakly, at the most, on Pt, and it is consistent with the idea that CN ad acts simply as an inert site blocker without affecting the electronic properties of platinum. On the contrary, a 25-fold and a 10-fold increase in the activity for the ORR (measured at 0.9 V) is observed for cyanide-modified Pt(111) electrodes in 0.05 M H 2 SO 4 and 0.05 M H 3 PO 4 , respectively [4]. This finding can be attributed to the suppression of the trigonal sites necessary for the specific adsorption of the anions present in sulfuric and phosphoric acid solutions.

Quantitative description of interfacial non-covalent interactions and interfacial proton-coupled electron transfer (PCET).
Any cation in solution can be expected to interact with the negative end of the CN ad dipole on the surface of a cyanide-modified Pt(111) electrode, blocking some of the CN ad groups, which will not be available for the formation of (CN ad ) x -H clusters, and provoking a change in the shape and/or position of the hydrogen adsorption region in the CV of cyanidemodified Pt(111) electrodes. This effect is illustrated in , which shows the variations observed in the CV of a cyanide-modified Pt(111) electrode as the concentration of K + increases. In Figure 6b, the potential at which a given coverage by (CN ad ) x -H clusters has been achieved (i.e., at which a given adsorption charge has crossed the interface) is plotted as a function of the logarithm of the cation concentration for Li + , Na + , K + , and Cs + . As can be seen, at low concentrations, the peak potential remains constant in all the cases and equal to that in M + -free 0.1 M H 2 SO 4 . Above a cation-dependent threshold concentration, the H upd peak potential begins to deviate from the value in the absence of the cation. At high enough concentrations, the peak potential decreases linearly with the logarithm of the cation concentration. This behavior suggests that the cations are retained on the surface as (CN ad ) x -M + clusters in equilibrium with M + in the solution.

Atomic Ensemble Effects and Non-Covalent Interactions
A similar effect is observed if the ionic strength and the cation concentration are kept constant and the pH is changed [21] as shown in Figure 7a for the case of Na +containing solutions. Figure 7b shows plots of the potential (vs. the standard hydrogen electrode, SHE) at constant charge density (i.e., at constant  CNH ) against pH. A line with a slope of -0.059 V is included as a guide to the eye. At all coverages, the potential shift is larger than the -0.059 V per pH unit expected for a PCET. The super-Nernstian contribution decreases slightly as  CNH approaches the maximum coverage possible, but at all  CNH , the shift of the equilibrium potential with pH remains far from the expected Nernstian value.
We developed a simple model that accounts for these two effects [18,21]. The model is illustrated in Figure  8a, which shows a ball model of the structure of the cyanide-modified Pt(111) surface with cations adsorbed at sites surrounded by three CNs, as in the experimentally observed honeycomb structure [18]. Figure 8b shows a cross-section along the line in Figure  8a and a schematic representation of the electrodeelectrolyte interface. The PCET occurs at plane a, which is defined by the position of the specifically adsorbed hydrogen acceptor/proton donor, namely, the N atom of CN ad /CNH ad . The presence of the cations interacting non-covalently with CN ad defines a plane of maximum approach b, where the proton donor (H 3 O + in acidic solutions and H 2 O in alkaline solutions) or acceptor (H 2 O in acidic solutions and OHin alkaline solutions) has to be transferred from the limit of the diffuse double layer (which, in concentrated solutions like the ones used here, coincides with the outer Helmholtz plane, OHP) before reaching a for the PCET to occur.
The assumptions made in the model are summarized as follows: a). The proton-coupled electron transfer corresponding to the electroadsorption of hydrogen on a cyanide-modified Pt(111) electrode does not take place at the metal surface, where the electrostatic potential is  m , but at a plane a, which is somewhere between the metal surface and the OHP, where the electrostatic potential is  a ; b). In the presence of a cation in the solution, a cation layer in equilibrium with cations in the bulk electrolyte is anchored at plane b by the cyanidemodified Pt(111) electrode through electrostatic interactions. At high enough electrolyte concentrations, plane b can be assumed to coincide with the OHP. Therefore, the energy required to remove the cation from plane b contributes to G of the hydrogen electroadsorption reaction; c). The hydrogen electroadsorption reaction may involve the transfer of less than one electron per proton transferred because of the polar nature of the Pt-CN surface bond. Together with the previous conditions, the hydrogen electroadsorption reaction to be considered is (1) d). The potential drop across the interface is linear. Consequently, the potential difference between plane a and the solution, ϕ a is a fraction of the total potential drop across the interface, ϕ m , given by where C m/OHP is the integral capacity of the electrical double layer, and C a/OHP is the integral capacity of the dielectric slab limited by plane a and the OHP (i.e., 0 ω a  1). With the assumptions above, the equilibrium condition for Reaction (1) is   (3), which is equivalent to Equation (2) in reference [18], predicts a constant equilibrium potential for 1   M as a K . A fit of the experimental results to Equation (2) allows to obtain [1-(1-ω a )] and K as . Table  2 lists the values obtained for the cases of Li + , Na + , K + , and Cs + , together with their standard Gibbs free energy of hydration and their ionic radii.
K as would be expected to increase with decreasing cation radius. However, by contrast, it increases strongly from Li + to K + and decreases from K + to Cs + . This observation cannot be explained only by the decrease in the hydration energy with the increasing cation radius, and it suggests that additional stability of the (CN ad ) x M + cluster is provided by an optimal fit of the cation into a cavity formed by the CN groups, similar to some [2]-cryptate inclusion complexes formed by macrobicyclic ligands and alkali metals [22]. The size of the cavity formed by the CN groups must be close to the atomic diameter of Pt (2.77 Å), and K + must fit well inside it. Although Cs + has a lower hydration energy than K + , it is too large to fit in the cavity, and therefore its K as is smaller.
The cation size also affects the value of [1- (1-ω a )]. Smaller cations can approach closer to plane a to yield a large C a/OHP , a small ω a , and a small [1-(1-ω a )]. Again, K + deviates from the expected trend because of the excellent match between its ionic radius and the size of the (CN ad ) 3 cavity.
The cation concentration is kept constant and the pH is changed similar to the experiments illustrated in Figure  6. Under these conditions, Equation (2) becomes   Atomic Ensemble Effects and Non-Covalent Interactions making  a ≠ 0 and decreasing the slope of the pH dependence. The increase in the hydrogen coverage as the potential is made more negative must also provoke an increase in the separation between plane a and the OHP, which provokes an increase in ω a and a decrease in the super-Nernstian contribution (see Figure 7). As predicted by Equations (3) and (4), the same value of [1-(1-ω a )] = 0.61 is obtained from both kinds of experiments for Na + at a constant charge of 20 C cm -2 more negative than expected from a merely Nernstian shift [23,24]. This has been suggested to be the reason for the higher electrocatalytic activity of Pt and other metals in alkaline media toward the oxidation of some organic molecules compared with acidic environments [25][26][27][28]. Our work provides an explanation for these phenomena.

Conclusions
We have shown that the site-knockout strategy enables us to study atomic ensemble effects in electrocatalytic reactions without interference of electronic effects. Atomic ensemble effects can be used to improve both the activity and the selectivity of an electrocatalyst.
Our work has also shown that cations can adsorb on the electrode surface as ion pairs form with specifically adsorbing anions. These electrostatic (or generally noncovalent) interactions, which can affect the properties of the electrode-electrolyte interface and therefore the processes occurring in this region including electrocatalytic reactions, are suitably described by the proposed simple model.
PCETs to specifically adsorbed anions exhibit a super-Nernstian shift with pH if the chemisorption bond is polar and if the plane at which the proton-electron transfer occurs does not coincide with the electrode surface. If above a cation-dependent threshold concentration, the presence in the electrolyte of cations other than H + can have a double effect: (i) cations provoke an additional negative shift of the PCET because of the additional energy required to remove the cation layer before the proton can access the plane at which hydrogenation occurs, and (ii) they affect the magnitude of the super-Nernstian shift by separating the plane of hydrogenation and the OHP.
In addition to the most usually considered electronic effect, atomic ensemble effects and the electrolyte composition can be used to tune the catalytic activity of electrode-electrolyte interfaces.