Platinum/Cation Interactions in Alkaline Media Observed by Constant Potential Molecular Dynamics

Abstract

The influence of cations on the properties of a chemisorbed species-free Pt(100)/1 M (X)­OH (X = Na, K, Cs) interface is explored using the MetalWalls software. Specifically, tools are developed to gather observations on (i) the distribution of charges within the electrode, (ii) the distribution of cations and anions within the electrochemical double layer, and (iii) the water mobility and orientation. They are discussed as a function of the nature of the cation and the magnitude of the potential. Differences are observed in the individual charges taken by interfacial platinum atoms that translate into drastically modified charge profiles as a function of the nature of the cation. These are intertwined with the distribution of the cations and anions (i.e., OH^–) at a distance <1 nm from the electrode. Not only said distribution is influenced, but also the solvation of cations, and the mobility and orientation of water molecules. Ultimately, this work (i) presents a new application of the MetalWalls software for technologies that are of great relevance nowadays and (ii) discusses from a computational standpoint the layered influence of the nature of the cation onto properties of the Pt(100)/electrolyte interface in alkaline media.

Introduction

Paraphrasing from electrochemical methods: fundamentals and applications, electrochemistry is the science that studies the controlled addition or removal of electrons, often one by one, by molecules or ions arriving very near a metallic or semiconducting electrode. This “very near” notion is pivotal in the field, and this region, often referred to as the “local reaction environment,” the “reaction zone,” or the “electrode/electrolyte interface,” is the heart of most electrochemical systems. The introduction of an electrode into an electrolyte will most likely induce a reorganization of the charge within the “electrode/electrolyte interface” and alter the local distribution of species that are not necessarily participating in the reactions (e.g., ions or water from the supporting electrolyte). If a discrepancy in the species orientation/distribution is observed vs. the electrolyte bulk, this implies that an “electrochemical double layer” (EDL) was formed. The structure of the electrochemical double layer is most often discussed as the Gouy–Chapman–Stern representation, albeit more recent representations have also been proposed and debated upon, , which consists in a compact layer near the electrode and a diffuse layer from the electrode toward the bulk. However, it is nowadays agreed upon that these mathematical models are not fully representative of the complexity of the electrode/electrolyte interface. , Namely, the Gouy–Chapman–Stern model only takes into consideration the potential distribution and the ion charges but completely occults the finite ionic size and exclusion volume effects, polarizability, presence of structured hydration shells and the associated solvation energies, specific chemical or surface binding interactions, and ion–ion correlation effects. Furthermore, it also considers the electrode as a homogeneously charged system. These simplifications of the interface complex nature have recently shown their limits, as recent works highlighted the influence of the nature of the cationic spectator charge carrier (i.e., not, in theory, involved in the reaction) on electrocatalysts’ activity for various electrochemical reactions. An example of this influence can be found when considering platinum as an electrocatalyst for the hydrogen evolution reaction (HER) in alkaline media. Specifically, on Pt(111), the HER/HOR (hydrogen oxidation reaction) is independent of the nature and concentration of the cations, while on stepped surfaces (e.g., Pt(110)), cations can either promote or inhibit the HER depending on their concentration and on the pH of the solution. This influence outreaches the sole HER/HOR. For example, the onset potential of CO electrooxidation on Pt single crystals is lowered in the presence of Li⁺ comparatively to Na⁺ containing electrolytes, and the presence of partially desolvated alkali cations stabilizes the CO₂ ^– intermediate during the CO₂ reduction reaction, with the nature and, specifically, the radius of the cations, influencing the product generation rate and type. In short, several effects are ascribed to cations’ influence on electrocatalysis, namely, (i) modification of the water interfacial arrangement owing to cation-induced hydration effect and water depletion at the interface for larger cations; (ii) modification of the chemisorption and physisorption behavior of reactants owing to steric and electrostatic effects; (iii) modification of the potential distribution in the EDL; (iv) modification of the electrocatalyst electronic structure, i.e., the local surface charge density of the electrode. These are only a few effects among many, and the interested reader can find further insights into the cation effects in electrocatalysis in the review by Khani et al. or Wu and Xu. Such effects induced by nonreactive species hint toward a key role played by the electrolyte properties, at the reactive interface (e.g., by modifying the water interfacial structure and properties, blocking the adsorption sites, interacting with the reactive species and intermediates, etc.), which is yet to be unveiled.

As gathering insights on the cation distribution, solvation, and their influence on their interfacial properties from an experimental standpoint proved to be extremely challenging, the solution may lie in the development of computational chemistry tools in parallel to the experimental methods, combined to the now well-established knowledge about force field parametrization, as it offers tools to better understand the influence of the cations. This has not always been possible, as molecular dynamics simulations did not initially allow to consider more than simple systems (e.g.,) with uncharged surfaces due to computational cost limitations. However, while the Ewald summation method dates from 1921, its application in molecular dynamics simulations during the 1980s marked a significant improvement in handling long-range electrostatic interactions, particularly for charged species. Unlike classical cutoff or shifted cutoff approaches, which may introduce abrupt truncations of the potential, the Ewald method ensured a smoother treatment beyond the cutoff distance. Although short-range interactions are still computed directly, the Ewald summation better accounts for long-range contributions, thereby improving the global electrostatic consistency of the simulation rather than strictly refining the local environment. Furthermore, the later development of Constant Potential Molecular Dynamics (CP-MD) simulations, with variable charges on electrode atoms, enabled a more realistic behavior compared with the previous classical models by allowing the charge to evolve during the simulation at a given potential. This came at the cost of the size of the simulation system and the resulting boundary effects, as well as the time scale of the simulation. Finally, the recent reconsideration of the electrode charge nature and distribution proposed by Scalfi et al. further increased the realism of the computational observations. This specifically consisted in defining the charge within the electrode no longer as a global, constant value that is distributed equally throughout the entire electrode but rather as a fluctuating distribution of charges that can locally evolve with time, as long as the global condition (i.e., often the potential difference between two electrodes) is respected, with an illustration of this distribution of charges provided in Figure a. The charge evolution is here stated by an electrical potential fixed and defined as

$$\frac{\partial V_c(Q)}{\partial Q_i}={\mathrm{\Psi }}_i$$ 1

where Ψ _i is the applied potential difference on the electrode sites, Q_i is the integrated charges, and V _c is the Coulombic energy of the system. This improvement cannot be understated as the charge distribution within the electrode is now affected by the ion distribution at the interface, thus yielding a better description of the cation effect on the electrode/electrolyte interface properties. Furthermore, Scalfi et al. introduced the consideration of the “metallicity” of the electrode to refine this charge distribution. This was formalized in the MetalWalls software package and showed promising results on bulk carbon, nanoporous carbons, and gold electrodes. ,

1.

(a) Visual representation of the electrolyte-exposed (100) layer of the Pt electrode (with the periodicity on (x, y) illustrated) when a potential difference of 0 V is applied between the two electrodes, on 4 frames spaced by 25 ps with the color of platinum atoms varying between blue and red (gray corresponding to a neutral charge and the maximum/minimum charge reached being +0.06/–0.07 elementary chargee) as a function of their local charge, as illustrated in Figure . (b) Side view of the simulation system presenting 5 Pt cubic-face-layers in contact with the electrolyte (in the electrolyte, oxygen in red, hydrogen in white, Na⁺ in blue). Note that the first layer of a given electrode in (b) corresponds to the layer discussed in (a).

The computational study of the cation effect using classical MD is few and far between, , with recent examples being the work by Johnson and Haussener, which focuses on a comparative evaluation of the cation effect by the use of a continuum model or of constant potential molecular dynamics on silver in the frame of the CO₂ reduction and exhibiting different interfacial cation distribution as a function of the cation nature, as well as the need for molecular dynamics to map the properties of the electrode/electrolyte interface or the work of Tran et al., which investigates the Ag(111) interface in 1 M XOH, with X = Li, Na, K and Cs. By opposition, it is important to stress that a growing number of works have been performed using ab initio molecular dynamics (AIMD) to investigate the cation effect. AIMD is based on the Hamiltonian resolution. This allows consideration of the electronic structure and the quantum nature of the interactions involved. However, the resolution of the Hamiltonian often requires DFT calculation methods, consequently increasing the computation time assigned for the system. Moreover, those calculations are possible only for limited size systems, whereas classical MD calculations (such as the ones performed by MetalWalls) are using an empirical force field. The empirical character of these simulations enables the modeling of larger systems and simulation times, but the use of fixed interaction terms between particles limits their ability to really capture real physical behavior. Each method has its strengths and weaknesses, making them complementary. Hence, they have to be considered based on their suitability for a given system: while AIMD is more suitable for energetic evaluation and small system computation, classical MD is more appropriate for studying large systems and dynamics. Nevertheless, AIMD allowed to better rationalize the metal/electrolyte interactions, as extensively discussed in several reviews and opinion letters by Groß or Groß and Sakong, , who emphasized AIMD’s ability to rationalize the water conformation at the interface as a function of the electrode chemistry, geometry, and adsorbate nature. They also highlighted the sensitivity of the results to the choice of DFT parameters (e.g., constrained/unconstrained). It also provided insights on the cation effect in electrocatalysis, supporting some of the earlier described observations (e.g., in the works of Monterio et al. or Bender et al.).

Here, we decided to extend the application of the MetalWalls software to this problem, i.e., the influence of the nature of the cation on the reactivity of the electrochemical interface to the HER. MetalWalls cannot simulate an electron exchange at the interface and, thus, the reaction itself, but the depiction of the interface it provides as a function of the cation nature and of the applied potential can help in building hypotheses to explain the phenomena occurring during electrocatalysis. Therefore, we here are assessing the properties of the electrode/electrolyte interface of a Pt (100)/1 M (X)­OH (X⁺ = Na⁺, K⁺, Cs⁺) interface with MetalWalls to obtain information (i) on the charge distribution within the electrode as a function of X; (ii) the cation and anion distribution and solvation within the EDL, this as a function of the potential and of the nature of the cations; (iii) the water molecules mobility and orientation near the interface.

Methodology

The investigated systems are composed of 3000 water molecules represented using the TIP4P/2005 (Transferable Intermolecular Potential with 4 Points), in which the charges are distributed as follows: (i) −1.1128 e on a virtual site M _w close to the O _w and (ii) 0.5564e on the hydrogen sites, and 54 X⁺ OH^– ions pairs (with X⁺ = Na⁺, K⁺ and Cs⁺), corresponding to a concentration of 1 M (X)­OH. The parameters of the added ion pairs, i.e., Na⁺, K⁺, and Cs⁺ coupled with OH^–, are used from the Madrid-2019 force field that induces a charge scaling by 0.85 , to consider implicitly the polarization. The Madrid-2019 force field diffusion, viscosity, solvation profiles, and dielectric charges correspond to experimental values up to a concentration of 4 M, hence implying that higher localized concentrations should be discussed with caution. No charge transfer or screening effect is explicitly considered, meaning the charge of the electrolyte is maintained constant during the simulation. The TIP4P model, developed in , was chosen to be consistent with the development of the force field (FF) parameters and to better represent the electrostatic charge distribution within the water molecules, as the water molecule is considered as a reactant in a lot of electrochemical reactions of interest. Concerning the water–platinum interactions, the GAL17 force field was used. It was specifically designed to characterize the Pt(111)–water interactions. Hence, one might question the rationale for not using its later version, GAL19, which is adapted to both Pt(111) and Pt(100), but the GAL19 force field was not fitted with a 4-point water model, compared with the GAL17 force field which was tested for the TIP4P-Ew, not exactly but closely related to the type of water model chosen in our simulations for our particular representation of the water dipole. The initial configuration was obtained by the fftool software using the packmol environment to create a first box with the following dimensions (i.e., ∼3 × ∼3 × ∼14 nm). The system was initially relaxed in the NPT ensemble (constant number of particles, pressure, and temperature). Then, 5 layers of Pt in the cubic-face-centered configuration with the (100) facet exposed to the electrolyte were added. The box is 3D periodic with a vacuum space introduced both after and before the first electrode to avoid any parasite interactions between the positive and negative electrodes. This electrode was generated with the Pt cubic-face-centered elementary mesh repeated in each direction of space and then cut to the size of the system. This electrode is parametrized with a metallicity parameter which is introduced to set the electron displacement into it and take account of the metallic character of the electrode. This parameter is set using eq 2:

$$l_{\mathrm{TF}}={\left(\frac{\pi }{192}\frac{a_0^3}{{\rho }_{\mathrm{e}}}\right)}^{1/6}\approx 0.5{\left(\frac{a_0^3}{{\rho }_{\mathrm{e}}}\right)}^{1/6}$$ 2

with a ₀ the Bohr length and p _e the electronic density of valence electrons in the metal.

Eq leads to the obtention of a Thomas–Fermi length. This parameter increase leads to a larger charge distribution in the electrode thickness. For a low value, the charge will be mainly distributed on the first plan. A first equilibration in the NVT ensemble with a Maxwell–Boltzmann velocity redistribution every 1000 steps is first performed to relax the system (constant number of particles, constant volume, and constant temperature). Here, no thermostat is used due to the high potential energy, as the velocities are unlikely to converge to a reasonable value. This is followed by a simulation in the NVT ensemble where a potential difference between the two electrodes of 0 or 1 V is applied during 10–20 ns, with the stopping time being determined by the stability of the density profile after simulation time. Temperature and pressure are controlled with a Nose–Hoover thermostat and an isotropic barostat method with a relaxation time of 10 fs. The cutoff is set to 9 Å and the intermolecular potential is computed with the Coulombic and a 6–12 Lennard–Jones potential expressed in eqs and , respectively.

$$U_{\mathsf{Co}\mathsf{u}\mathsf{l}\mathsf{o}\mathsf{m}\mathsf{b}\mathsf{i}\mathsf{c},\mathit{Coulombiij})}=\frac{q_i{q}_j}{r_{\mathit{ij}}}$$ 3

$$U_{\mathit{LJ},(\mathit{ij})}=4{\epsilon }_{\mathit{ij}}({\left(\frac{{\sigma }_{\mathit{ij}}}{r_{\mathit{ij}}}\right)}^{12}-{\left(\frac{{\sigma }_{\mathit{ij}}}{r_{\mathit{ij}}}\right)}^6)$$ 4

where q_i is the charge of the atom i and r_ij is the distance between atoms i and j. ε _ij and σ _ij are, respectively, the epsilon and sigma parameters of a combination developed in eqs and for i–j Lennard–Jones pairs. Or, in the case of the electrode electrolyte Coulombic interaction, by

$$U_{\mathrm{Coulombic},(\mathit{ij})}=\frac{Q_i{q}_j}{r_{\mathit{ij}}}\mathrm{erf}[r_{\mathit{ij}}]$$ 5

where Q_i is the charge of an electrode atom and q_j the charge of an electrolyte atom. erf represents the error function. The Lorentz–Berthelot rules have been used to compute cross parameters for LJ interactions:

$${\epsilon }_{\mathit{ij}}=\sqrt{{\epsilon }_{\mathit{ii}}{\epsilon }_{\mathit{jj}}}$$ 6

$${\sigma }_{\mathit{ij}}=\frac{{\sigma }_{\mathit{ii}}+{\sigma }_{\mathit{jj}}}{2}$$ 7

with ε _ij the depth of the potential well in kJ/mol, σ _ij the distance parameters in Å, and r_ij the distance between two particles at a given time t. Long-range electrostatics are calculated using the Ewald summation method, i.e., water molecules are kept rigid during the various simulations, while positions of the atoms of the electrodes have been constrained. The RATTLE algorithm is used to maintain constraints. The simulation was done using the MetalWalls software, which, in addition to the electrode polarizability elaborated on during the introduction, allows 3D-slabperiodic conditions. The Velocity Verlet algorithm has been used to integrate the equations of motion, with an integration step of 1 fs. Position of the atoms in the systems has been saved every 1000 steps. Energy and temperature were saved every 100 steps. The simulation outputs have been visualized using homemade VMD and Tcl scripts that were used to determine orientation and position of each molecule/atom, as well as the state of charge of the system. The density profiles and charge profiles were performed with the VMD profile tool developed in . The angle of the water molecule is represented through a simplified value defined as the orientation order parameter (S):

$$S=\frac{(3⟨{\cos }^2(x)⟩-1)}{2}$$ 8

with “x” as the angle of the molecule dipole to the “z” vector, perpendicular to the electrode. “S” can take three values of importance, i.e., (i) S = 0 for a system with a random water molecule orientation; (ii) S = −0.5 for a system in which the water molecules are perpendicular to the “z” vector; (iii) S = 1 for a system in which the water molecules are aligned with the “z” vector.

Results and Discussion

As elaborated upon in the introduction, Figure presents the MetalWalls software’s ability to perform constant potential molecular dynamics simulation with a variable distribution of charges through time and space on the electrode, while ensuring that the potential between the two electrodes remains constant through the simulation. This is translated in the simulation by an evolution of each Pt atom’s individual charge as a function of time (see Figure a). This charge is mainly impacted by the near-interfacial species in the electrolyte that modify largely the potential nonbounded energy of the system (e.g., a close-to-interface cation at a given time will result in a local accumulation of electrons to compensate for this charge), but the influence of every electrolyte species is ultimately considered, resulting in the complex and dynamic charge distribution illustrated in Figure a. A side view of the box is provided in Figure b. This phenomenon, i.e., the diversity of charges on the electrode, is further illustrated in Figure , which presents (i) the distribution of all of the charge values taken by each individual platinum atom directly exposed to the electrolyte during the entire simulation (Figure a–c) and (ii) the distribution of the charge density taken by the whole electrode (including atoms of layers not directly exposed to the electrolyte) during the simulation (Figure d–f) for various applied voltages and cations. It is important to underline that, for the 0 V situation (i.e., no potential difference between the two electrodes), the two electrodes can exchange electrons freely to maintain the “0 V” condition imposed by MetalWalls, i.e., they are short-circuited. Hence, in Figure a,d, the charge distributions on the two electrodes are jointly represented. This decision is further elaborated on in the Supporting Information (Figure S1 and associated discussion).

2.

In a simulation box (dimensions = 26.5 × 26.5 × 142.25 Å) of a 1 M XOH (X = Na, K, Cs) between 2 × 5 Pt layers in cubic-face-centered configuration with the Pt(100) facet facing the electrolyte, containing 3000 TIP4P water molecules and 54 OH^– and X⁺ ions, (a–c) the charge probability distribution of a single Pt atom in the atomic layer directly exposed to the electrolyte over the entire simulation duration and (d–f) the charge density probability distribution of a given electrode over the entire simulation duration. For (a) and (d), i.e., when the potential difference between the two electrodes is 0 V, the two electrodes are able to freely exchange electrons (i.e., short-circuited) and are therefore represented as one electrode, which includes the charge taken for all atoms from both electrodes exposed to the electrolyte (a–c) or for the total charge for both electrodes (d–f). This peculiarity is discussed extensively in the Supporting Information.

When discussing Figure a–c, it is important to underline that this spread in charges is mainly observed on the first atomic layer of the electrodes (see Figure S3 for the distribution of the charge in the second atomic layer of the electrodes), illustrating that the electrolyte is mainly affecting the charge in the first atomic layer of the electrode, directly in contact with the electrolytic solution. This characteristic is a consequence of the material-dependent Thomas–Fermi approximation used in the model, which links charge distribution through the planes to the electronic density of the observed material assuming the validity of the local density approximation. When 0 V is applied, the individual charge and the charge density repartitions are centered around zero, as illustrated in Figure a,d. The effect of an applied voltage is further illustrated in the panels b,c,e,f. As a result of the 1 V applied voltage (i.e., the positive electrode = +1 V vs. the negative one, and both electrodes would ultimately be at ±0.5 V vs. a hypothetical reference electrode situated in the middle of the simulation box), the atoms and charge distributions adjust in response to the applied voltage. The atomic charge distribution is centered around ±3 × 10^–3 to ∼15 × 10^–3 elementary charges (1.6 × 10^–19 C) for the positive and negative electrodes, respectively.

Compared with the distribution profiles presented in Figure a–c, the distributions presented in Figure d–f represent the probability of each electrode to take a total charge density. At 0 V, the charge density distribution is symmetric and centered around zero. Whereas in pure water it presents a Gaussian shape with a low dispersion, and the presence of the XOH electrolyte leads to a wider dispersion (Figure d). This results from the interactions between the ions and the electrodes, leading to (i) local variations of the electric charge that in turn (ii) influence the global electrode charge that fluctuates over time, due to the ion mobility near the electrode/electrolyte interface. Interestingly, the presence of K⁺ leads to a multimodal charge density distribution at 0 V, hinting toward the fact that in the presence of K⁺, the total interfacial charge distribution within the electrode fluctuates between two groups of values. This is indicative of the fact that, when K⁺ is within the electrochemical double layer, it has a high probability of physisorbing in a “4-circled” configuration, i.e., in between four platinum atoms. This brings it extremely close to the Pt surface (i.e., below <1.75 Å) and substantially impacts its global charge density, creating a large difference with a case where low/no K⁺ are being physisorbed, hence explaining the bimodal distribution. This is confirmed by the observations presented in Figures S4 and S5, which discuss the different configurations in which the cations are observed in the electrode vicinity for the KOH and NaOH electrolytes, confirming that both 1 M NaOH at 1 V and 1 M KOH at 0 V exhibit the highest relative probability of having a cation in a “4-circled” configuration.

Both cation size and strength can be put in relation with the energy interaction profile and the resulting position as proposed in . Figure e,f present the charge density distribution when 1 V is applied between the two electrodes. Additionally, Figure S6 presents examples of charge density distribution for several potentials, i.e., 0, 0.5, 1, and 2 V applied between the two electrodes for the Pt(100)/1 M NaOH interface. The applied voltage leads to a clear difference between the positive (Figure e) and negative (Figure f) electrodes, as well as to a sharpening of the distribution in the case of K⁺ and Cs⁺. Interestingly, a multimodal distribution of the charge density on the electrode is observed in the presence of Na⁺ under a potential, as shown in Figure e,f. These distributions are still affected by the nature of the cations, with the average charge density deviating vs. H₂O as follows: Cs⁺ < Na⁺ < K⁺. Using the results from Figure e,f, the differential double-layer capacitances of the Pt/solution interface in the absence of adsorbate species were also extracted and are summarized in Table S2. These values fall in the same ranges of experimental observations in the literature , and also evidence the dependence of the capacitance to the cation nature. Interestingly, the trend reported in the present work differs from that reported by Xue et al. (i.e., Cs⁺ < Na⁺ < K⁺ in 1 M XOH in our work vs. Na⁺ < K⁺ < Cs⁺ in 0.05 M XClO₄ in Xue et al.’s work, likely due to variation in electrolyte composition and concentration as well as surface orientation).

Figure focuses on the distribution of species and charges at the interface, on the side of the electrolytic solution, i.e., specifically on the cations and anions, as well as the resulting charge distribution. At 0 V, fluctuations in the ion density are observed in the electrode vicinity, with the main peaks being observed at ∼2.5 and 5 Å from the electrode surface. This layered structure arises from the interface presence, and the distribution is symmetrical for both electrodes. By summing the anion and cation charges, the overall ion charge distribution is obtained (Figure , right panel). This charge distribution exhibits substantial variations in the 0–9 Å region from the electrode, which are summing to a null elementary charge (as they are compensating for the excess of charge within the Pt electrode, which is null). When considering the cation-induced discrepancies within the observations of Figure , a few key features need to be highlighted:

• (i)The intensity of the local charge and local distribution of ions in the ∼2.5 Å region is higher for KOH, and its maxima is closer to the Pt(100) compared with the other electrolytes, NaOH and CsOH. This is consistent with the results in Figure , confirming the stronger influence of K⁺ on the interfacial charge distribution.
• (ii)The OH^– is mainly located in the vicinity of the cations. This layered organization is further confirmed by the distribution of water molecules presented in Figure S6, as the water itself exists close to the ions.

3.

Distribution of the ions and the charges, excluding the water contribution, within the electrolyte of the different simulation boxes used within this work in 10–20 ns simulations. Specifically, (a–c) correspond to the (a) Na⁺, (b) OH^–, and (c) ion-induced charge distribution as a function of the distance from the left/positive electrode in the Pt(100)/1 M NaOH simulation box, this with an applied potential difference of 0 and 1 V; (d–f) correspond to the (d) Cs⁺, (e) OH^–, and (f) ion-induced charge distribution as a function of the distance from the left/positive electrode in the Pt(100)/1 M CsOH simulation box, this with an applied potential difference of 0 and 1 V; (g–i) correspond to the (g) K⁺, (h) OH^–, and (i) ion-induced charge distribution as a function of the distance from the left/positive electrode in the Pt(100)/1 M KOH simulation box, this with an applied potential difference of 0 and 1 V.

When a potential difference is applied between the two electrodes, e.g., 1 V, the ions reorganize at the interface to compensate for the charges on the Pt electrodes shown in Figure . When focusing on the positive electrode, almost all of the Cs⁺ are expelled from the 2.5 Å layer, while a significant amount of K⁺ can still be found. At ∼5 Å from the electrode, a significant decrease of the local Na⁺ and Cs⁺ concentrations is also observed at the positive electrode, whereas the concentration of K⁺ still remains similar. Surprisingly, a partial depletion of OH^– in the 2.5 Å layer is also observed in 1 M CsOH and 1 M NaOH, with an increase of abundance in the 5 Å region, indicating that the charge compensation of the electrode occurs through cation removal and water reorganization in the positive electrode vicinity (2.5 Å) rather than an increase of the OH^– concentration. Interestingly, for 1 M KOH, the distribution of OH^– remains qualitatively the same, with the main difference being the decrease in concentration >5 Å, incidentally indicating that other species (e.g., K⁺ removal or H₂O) are compensating for the electrode charge. At the negative electrode, an increase of the cation density is observed, in both the 2.5 and the 5 Å layers for the Na⁺ and the K⁺. Surprisingly, a decrease of the Cs⁺ concentration is observed at ∼2.5 Å from the Pt surface, accompanied by a substantial increase of Cs⁺ and OH^– concentration in the 5 Å region (notably, the Cs⁺ is locally higher than 4 M, i.e., beyond the concentration at which the Madrid-2019 force field was experimentally validated, thus stressing that the observed trends should be qualitatively, rather than quantitatively, discussed). This hints toward the fact that the Cs⁺ reorganization at 1 V completely hinders the OH^– mobility at the negative electrode. The overall peculiar behavior of Cs⁺ was also observed by experimental and theoretical means on other systems, namely, Au-poly in XHCO₃ where it was shown that a substantially different interfacial profile was observed with Cs⁺ as the cation while, on Ag(111), a decreased concentration in the closest vicinity of the electrode was also observed with Cs⁺ as the cation in 1 M XOH, vs. Li⁺, Na⁺, and K⁺.

Hence, the nature of the cation drastically influences the interfacial properties of a free platinum (100) surface (i.e., in the absence of chemisorbed species). A first explanation behind the cation effect can be found within the properties of each cation, as described in Table S1. For example, the proximity and intensity (i.e., relative to the second peak) of the 2.5 Å increase with the ε _ii value (i.e., Cs⁺ < Na⁺ < K⁺), further stressing the key role played by the noncovalent interactions onto the EDL structure. Similarly, when a voltage is applied, the cations with the lower ε _ii are expelled from the positive electrode, the closest layer of ions, and, in the case of Cs⁺, even from the negative electrode, whereas K⁺ remains present in the electrode vicinity. This proximity to the interface, both at 0 and 1 V, is consistent with the changes observed in the electrode charge in Figure , i.e., closer and denser cation presence in the 2.5 Å strongly influences the distribution of electrode charge vs. the one observed in pure H₂O. Another parameter by which the nature of the cation could influence the interfacial properties would be the size of the latter, i.e., the van der Waals radius of Cs⁺ > the van der Waals radius of K⁺ and Na⁺. However, the solvation number as a function of the distance electrode should be considered in the frame of this discussion and is illustrated at 0 V and at the positive electrode at 1 V in Figure for the different cations discussed in this work.

4.

First shell solvation number in the Pt(100)/1 M XOH interface according to the distance to the electrode for (a) Na⁺ at 0 and 1 V (negative electrode); (b) K⁺ at 0 and 1 V (negative electrode); (c) Cs⁺ at 0 and 1 V (negative electrode).

Here, a couple of points ought to be highlighted: (i) independently of the voltage, the solvation numbers are reaching a constant value when the distance to the electrode is >5.5–6 Å, namely, ∼5 for K⁺ and Na⁺, and 4 for Cs⁺. This corresponds, in Figure , to the region beyond the second peak of the EDL in which the charge and ion distributions are close to the bulk values. The fact that the solvation number reaches a constant value further confirms that this region does not belong to the EDL. Then, (ii) both K⁺ and Na⁺ experience a partial desolvation when nearing the interface, down to ca. 3.5–4 molecules of H₂O, and this desolvation seems voltage independent. Finally, (iii) the Cs⁺ solvation does not substantially decrease at distances from the electrode ranging from 1 to 4 Å at 0 V. When a potential is applied, the solvation in this region decreases. However, it is important to stress that, at 1 V, at the positive electrode, there is almost no Cs⁺ < 4 Å (see Figure d). However, it appears that within the 5 Å region, Cs⁺ is more solvated than in the bulk (average number of solvation is ∼6) and, according to Figure d, in high concentration.

Another key result to discuss is that the redistribution of charges considering only the ions (Figure ) near the interface only partially corresponds to what we would expect from the classic EDL theories (e.g., Gouy–Chapman–Stern), even qualitatively. Namely, in 1 M KOH, a gradual decrease of the charge opposite to the electrode charge is observed and is qualitatively coherent with the EDL theory. However, when considering the 1 M NaOH and, especially, the 1 M CsOH, the distribution of charges reaches a maximum not at the interface, but rather in the 5 Å region. This indicates that, in the <4 Å, the compensation of the electrode charge may arise, not from the ions, but most likely from the solvent. This is also hinted by Figure S7, as water molecules experience a significant reorganization when a voltage is applied between the two electrodes in 1 M CsOH, with the oxygen (from H₂O) moving “away” from the negative electrode and “toward” the positive electrode, indicating a reorientation and charge compensation by the water. Further interesting observations are obtained using the orientation order parameter (S; see eq ). This parameter aims to illustrate the different orientations taken by the water molecules near an electrode, as illustrated in Figure a–c. Namely, S = 0 indicates water with the dipolar moment of water aligned at an angle of ∼53 or 128° perpendicular to the platinum surface. S = 1 is indicative of water with its hydrogen pointing either toward or opposite to the electrode surface. Finally, S = −0.5 corresponds to the bisector of the hydrogen–oxygen–hydrogen angle being parallel to the surface. Detailed values of the water dipolar moment can be found in Figure S8. It is important to underline that there is much more possible orientation for S ∼ 0 than for the other significant numbers, thus implying that free water molecules will most likely present an S parameter centered around 0. The different values of S taken for distances of 0–10 Å from the (negative) electrode surface, as a function of the nature of the cation and of the potential, are represented in Figure d. One should underline the presence of the S “standard deviation” on Figure d, which corresponds to the dispersion in S values taken at a given distance of the electrode (i.e., a low standard deviation would indicate H₂O constrained in each configuration, whereas a high standard deviation is indicative of a more mobile H₂O with a more random orientation).

5.

Evolution of the water orientation at the electrode/electrolyte interface at 0 V and at 1 V (negative electrode): (a–c) illustration of the key values taken by the orientation order parameter (S) as a function of the water angle and (d) profile of the orientation order parameter (S) as a function of the distance to the electrode and the type of electrolyte/potential. The colors on (d) correspond to the relative standard deviation of the S parameter and thus witness the freedom of rotation of the water molecules.

In the >6 Å region, independently of the cation nature and the potential, the water is outside the EDL and does not compensate for the electrode charge. Therefore, it is mostly free, except for the water in the solvation layer of the ions, and a large variety of dipolar moment orientations (S = 0 ± 0.4) are observed. This extends up to the ∼5 Å region (i.e., the second peak in the ion density profiles, see Figure ), in which S ∼ 0. In the <4 Å region, the water presents a strong antialignment (S ∼ −0.2) with a low standard deviation, indicative of little flexibility in the possible taken angles. This is possibly due to the abundance of cations in the 2.5 Å layer and their partial solvation in the electrode vicinity (see Figure ), which leads to a constrained orientation of the water molecules and indicates that these solvating molecules are to some extent presenting a dipolar moment parallel to the surface. Then, S changes from −0.2 to −0.1 to 0 when being nearest to the electrode/electrolyte interface, and this peculiar trend is believed to result from the compensation of the electrode charges in the nearest region to the electrode (i.e., < 2 Å), which leads to some of the water molecule hydrogen pivoting toward the electrode while part of it still contributes to the partial solvation of the cations, explaining why an S = 1 is not observed. Although, as the antialignment <4 Å is also observed in a pure H₂O electrolyte (see Figure d), it is also possible that this reorganization is a direct consequence of the water orientation nearest to the electrode surface and at <4 Å, as well as of the water density in said regions (see Figure S6) that both constraint the angles taken by this intermediate layer. The main “cation-induced” difference is the minimal value taken by S, i.e., the value is modified for 1 M NaOH at 0 vs. 1 V (−0.25 and −0.15, respectively), indicating that the water molecules that are not solvated are compensating for the electric field. This confirms that the compensation for the electrode charge is not carried by the sole ions but also by the water molecules and their local orientation.

A third aspect that has yet to be tackled is the H₂O mobility near the interface. This is illustrated in Figure . To perform this plot, we established a threshold at 5 Å (i.e., considering as “out” the water molecules beyond the maximum of the second peak in the cation distributions, see Figure ) and followed the water mobility as illustrated in Figure a–c. The residence time of the molecules closer to the electrode than 5 Å was then monitored, and the probability of residence times was plotted in Figure d for all of the systems discussed within this work.

6.

Illustration of (a–c) how the residence time of a given water molecule was established (i.e., each picosecond, the 0–5 Å domain was assessed and the H₂O molecules inside were counted while the H₂O molecules outside were not counted. If a H₂O molecule moves outside, the counting stops) and (d) average residence time of a water molecule at the Pt(100)/1 M (X)­OH (X⁺ = Na⁺, Cs⁺, K⁺) interface (either at 0 V or at the negatively charged electrode when a 1 V voltage is applied).

The interfacial H₂O molecules in 1 M CsOH, 1 M NaOH, and 1 M KOH have a similar residence time at 0 V, with the latter being substantially lower in 1 M KOH. This indicates that in 1 M KOH, the water molecules <5 Å exhibit higher mobility and freedom, which is consistent with the standard deviation values of S observed in Figure d. Further differences are observed when applying a 1 V potential difference between the two electrodes, which can be tied to the observation of Figures and : as the Cs⁺ are accumulating in the ∼5 Å layer and are highly solvated in this region, they greatly limit the ability of water to move outside the EDL. Furthermore, the low density of Cs⁺ in the electrode vicinity (∼2.5 Å) implies that the water needs to compensate for a substantial part of the electrode charge. In contrast, the water mobility (i.e., lower residence time) substantially increases with the voltage for NaOH, as the cations are now more abundant in the ∼2.5 Å layer and compensate for more of the electric charge (see Figure ), hence confirming that, as a function of the cation nature, the water contribution to the electric field compensation and, thus, its mobility are evolving.

Hence, it becomes apparent that the nature of the cation influences the interfacial concentration of water, its orientation, and its mobility for Pt(100)/1 M (X)­OH interfaces. All three effects are expected, in turn, to influence water availability, ease of adsorption onto the platinum surface, and thus their reactivity. An additional case could be made here on the charge availability at the electrode/electrolyte interface, i.e., that when a high density of cations is at the electrode/electrolyte interface, the electrons within the electrode will mirror its charge and thus be less accessible to the water molecules adsorbed onto the surface prior to their reduction. In light of these comments, one would expect that an ideal interface for HER electrocatalysis should exhibit a less constrained but retained in the EDL (i.e., high residence time) water molecule, with high water density near the electrode and low cation concentration at the electrode surface to avoid a charge-mirroring effect from the electrode material. This would ultimately indicate the Cs⁺ as an ideal cation, whereas K⁺ (high mobility, high density of K⁺ in the ∼2.5 Å) would result in depreciated electrocatalytic properties for the HER. The latter observation would be consistent with the results obtained by Monterio et al. and Goyal et al., on Pt-polycrystalline and Ru-modified Pt(553), which showed deteriorated performances for the HER in 0.1 M KOH vs. 0.1 M NaOH. However, one ought to be cautious in directly comparing the experimentally observed HER trends and the results presented in this manuscript. Indeed, this model considers an ideally polarizable electrode, without any chemisorbed species on the Pt(100) surface, whereas it is established that, in the HER potential range, the Pt surface is covered by H*. The presence of “partial charge transferred chemisorbed species” will affect the solvent and cation interactions with the surface.

Hence, said cautious nature should be kept when considering the conclusions of this work, despite the fact that the simulated water structure (Figure S7) exhibits a profile consistent with earlier observations, , as it is generally accepted that, even in so-called double-layer region of Pt (i.e., 0 V in this system), OH* absorbates are present on Pt(100) and Pt(110) surfaces. , Furthermore, when 1 V is applied between the two electrodes, (i) the positive electrode reaches a potential range where OH* species form on the platinum surface. Since OH* may induce charge localization and a steric effect, our simulations considering an ideally polarizable electrode should be considered carefully in this potential range, and thus the positive electrode was little discussed in this manuscript; (ii) the negative electrode reaches a potential range in which H* is formed onto the Pt surface. H* does not lead to steric hindrance owing to its limited space occupancy. We studied the charge on Pt and H* as a function of the number of platinum layers for one H* in bridge configuration on the Pt(100) surface, as the bridge is identified as a possible adsorption conformation for H*. These results are presented in Figure S9 and show that, for 10 atoms of Pt, i.e., an equivalent thickness to our MetalWalls slab, the charge delocalization would be of 0.012e per platinum atom, i.e., substantially lower than the cation effect observed on the “chemisorbed species-free” electrode. This value, however, is only valid if the influence of the hydrogen charge is distributed homogeneously throughout the electrode. However, the hydrogen might mainly impact the topmost platinum atoms and the presence of an H* also implies the presence of a negative (i.e., −0.118e per hydrogen) charge onto the Pt surface, which itself is expected to influence the cation and water behavior. This naturally sparks the direction for future investigations, i.e., how will the cation effect evolve in the presence of H* onto the platinum surface, using the set of tools and methods developed within this work? To this, one may add that even if an experimentally coherent surface coverage in adsorbates was achieved in MetalWalls on Pt(100), experimental works to compare it with remain sparse on low index crystals, e.g., none exist on Pt(100), and such works should always be discussed carefully, as alkaline electrolytes are known to contain impurities (e.g., 3d-transition metal) in various concentrations as a function of the provider, type of cation, etc. that ought to also influence the HER activity.

Conclusions

This study presents a comparative molecular dynamics (MD) investigation of 1 M (XOH, where X = Na, K, Cs) aqueous systems in contact with Pt(100) electrodes as a function of the applied voltage between two electrodes. The variable charges, constant potential method, applied for the first time on Pt electrode/XOH interfaces, reveals the strong impact of the nature of the cation on the interfacial structure and charge distribution. The observed charge variations follow a trend linked to the Lennard–Jones parameters, where a denser electronic cloud leads to higher polarization of the electrode at the interface. The cation concentration profiles exhibit a layered structure in the electrode’s vicinity, with two primary peaks observed at ∼2.5 and 5 Å. Upon applying a voltage, the local concentration of Na⁺ and K⁺ increases at both peaks on the negative electrode. In contrast, the Cs⁺ concentration increases solely in the second cation layer. This unique density profile of Cs⁺ is attributed to its solvation behavior, which undergoes more significant changes at 1 V compared with Na⁺ and K⁺, particularly around ∼4 Å from the electrode. The simulations reveal that the cation cannot fully compensate for the applied voltage. Instead, water molecules orient themselves near the electrode to balance the electrode charges. The TIP4P/2005 water model effectively captures this dipole orientation, driven by van der Waals interactions and the electric field. Water residence time at the interface varies with the cation, following the trend K⁺ < Na⁺ < Cs⁺. This suggests that cation solvation governs water behavior near the electrode. Among the three cations, K⁺ exhibits the strongest impact on the electrode/electrolyte properties possibly due to its dense electronic structure, while Cs⁺ is the least constraining owing to its weaker electronic structure. Despite limitations in system size and simulation time, this study provides valuable insights and paves the way for further research on Pt molecular dynamics at a constant potential. Future work may involve comparisons with continuum formulations as well as extending this method to investigate different Pt geometries and surface coverages, aiming to bridge the gap between simulations and experimental conditions.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c07205.

• The Supporting Information contains Figures S1–S9, Tables S1 and S2, and the associated discussions. The latter provides/illustrates the force field parameters (Table S1), an illustration and explication of the system behavior when the 0 V is applied between the two electrodes (Figures S1 and S2), of the charge distribution as a function of atomic layer within the electrode (Figure S3), of the cation relative abundancy as a function of the type of physisorption site (Figures S4 and S5), of the total charge distribution onto the right electrode as a function of the voltage (Figure S6), of the specific capacitance as a function of the cation nature (Table S2), of the water density profiles (Figure S7), of a water angle to orientation order parameter correlation (Figure S8), and of the charge delocalization induced by the presence of a hydrogen atom calculated from density functional theory (Figure S9) (PDF)

The authors declare no competing financial interest.