Deriving effective electrode–ion interactions from free-energy profiles at electrochemical interfaces

Fabrice Roncoroni, Abrar Faiyad, Yufeng Li, Tao Ye, Ashlie Martini, David PrendergastThe Journal of Chemical Physics 2026 · DOI · arXiv

TL;DR

This paper demonstrates that standard "mixing rules" used in classical molecular simulations often fail to accurately predict how ions interact with metal electrodes. By using a state-of-the-art machine-learned interatomic potential (MLIP) as a high-accuracy benchmark, the authors developed a systematic way to tune classical models, revealing that these microscopic "stickiness" factors are the primary drivers of macroscopic electrochemical properties like capacitance and the potential of zero charge.

What Problem Does This Paper Address?

The electrochemical interface—the thin region where a metal electrode meets a liquid electrolyte—is the "engine room" of batteries, fuel cells, and electrolyzers. To design better devices, researchers need to model how ions like Sodium ($Na^+$) or Chloride ($Cl^-$) adsorb onto these surfaces.

However, there is a major "Goldilocks" problem in simulation:

  1. Ab initio methods (AIMD): These are highly accurate but computationally "expensive," making it impossible to simulate the long timescales (nanoseconds) needed to see ions move and settle.
  2. Classical Molecular Dynamics (MD): These are fast but rely on simplified "force fields." Specifically, they use "mixing rules" to guess how an ion interacts with a metal atom based on their individual properties.

The authors highlight that these standard guesses are often qualitatively wrong, leading to simulations where ions either stick too much or not at all, contradicting experimental reality.

Methods

The researchers employed a multi-scale approach to bridge the gap between atomic interactions and macroscopic physics:

  • Enhanced Sampling: They used Metadynamics and Umbrella Sampling to calculate "free-energy profiles"—essentially maps of the energy barriers an ion faces as it approaches the gold ($Au(111)$) surface.
  • Machine Learning (MLIP): They utilized Meta’s Universal Models for Atoms (UMA). This MLIP acts as a surrogate for expensive quantum mechanical calculations, providing near-DFT accuracy at a fraction of the cost.
  • Classical Tuning: They tested four popular classical force fields (Jorgensen, Cheatham, Merz, and Dang) and developed a methodology to "tune" their Lennard-Jones (LJ) parameters to match the high-fidelity UMA results.
  • Continuum Modeling: Finally, they fed these molecular-level energy profiles into a modified Poisson-Boltzmann continuum model to predict macroscopic observables like differential capacitance.

Datasets

  • Training Data: The UMA model used was pre-trained on Meta’s massive open-source datasets: OMat24 (Open Materials), OMC25 (Open Molecular Crystals), and OMol25 (Open Molecules), comprising millions of quantum mechanical reference calculations.
  • Evaluation System: The study focused on $Na^+$, $Cl^-$, and $F^-$ ions at the $Au(111)$-water interface.

Key Insights and Ideas

  • Mixing Rules are Unreliable: The most striking "aha moment" is that standard mixing rules (like Lorentz-Berthelot) are fundamentally insufficient for metal-liquid interfaces. They fail to capture the complex electronic response of the metal, often leading to "unphysical" results where ions are repelled when they should be attracted.
  • The Specificity of Adsorption: The UMA model correctly predicted a clear hierarchy: $Cl^-$ adsorbs strongly (it wants to touch the metal), $F^-$ adsorbs weakly, and $Na^+$ doesn't specifically adsorb at all (it stays hydrated). Classical models only got this right after significant manual tuning.
  • Micro-to-Macro Link: The paper proves that a tiny change in the Lennard-Jones "well depth" ($\epsilon$) at the atomic scale can shift the macroscopic Potential of Zero Charge (PZC) by hundreds of millivolts. This underscores why force field parameterization isn't just a technical detail—it's the foundation of predictive electrochemistry.
  • MLIPs as the New Benchmark: The study positions MLIPs not just as a simulation tool, but as a "gold standard" to help fix and refine older, faster classical models.

Key Results

  • Chloride Adsorption: UMA predicted a strong specific adsorption for $Cl^-$ with a free energy minimum of -8.40 kcal/mol at 2.3 Å from the surface.
  • Force Field Sensitivity: Without tuning, classical models were all over the map. For example, the Jorgensen parameters suggested $F^-$ would spontaneously adsorb, while Cheatham parameters suggested it would be repelled.
  • Capacitance Curves: By integrating UMA energies into continuum models, the authors reproduced the characteristic "camel-shaped" differential capacitance curves seen in experiments, which classical models often miss.
  • PZC Shifts: Specific ion adsorption was shown to substantially alter the interfacial ion population, shifting the PZC by approximately 100-200 mV in $NaCl$ systems compared to non-adsorbing models.

Limitations and Future Work

The authors acknowledge a few key hurdles:

  • Short-Range Cutoffs: Current MLIPs like UMA use relatively short cutoffs (around 6 Å), which may struggle to capture long-range electrostatic effects or the full dielectric response of the solvent.
  • Polarization and Charge Transfer: While the "tuned" classical models perform better, they still don't explicitly model electronic polarization or charge transfer between the ion and the metal—effects that are physically present but mathematically complex to include in simple force fields.
  • Future Directions: The authors suggest that future MLIPs should incorporate van der Waals corrections (like DFT-D3) to better model water density and ion solvation near surfaces. They also aim to extend this framework to more complex, multi-component electrolytes.

Summary updated

Abstract

Understanding ion adsorption at electrified metal-electrolyte interfaces is essential for accurate modeling of electrochemical systems. Here, we systematically investigate the free energy profiles of Na+, Cl-, and F- ions at the Au(111)-water interface using enhanced sampling molecular dynamics with both classical force fields and machine-learned interatomic potentials (MLIPs). Our classical metadynamics results reveal a strong dependence of predicted ion adsorption on the Lennard-Jones parameters, highlighting that-without due care-standard mixing rules can lead to qualitatively incorrect descriptions of ion-metal interactions. We present a systematic methodology for tuning the cross term LJ parameters to control adsorption energetics in agreement with more accurate models. As a surrogate for an ab initio model, we employed the recently released Universal Models for Atoms MLIP, which validates classical trends and displays strong specific adsorption for chloride, weak adsorption for fluoride, and no specific adsorption for sodium, in agreement with experimental and theoretical expectations. By integrating molecular-level adsorption free energies into continuum models of the electric double layer, we show that specific ion adsorption substantially alters the interfacial ion population, the potential of zero charge, and the differential capacitance of the system. Our results underscore the critical importance of force field parameterization and advanced interatomic potentials for the predictive modeling of ion-specific effects at electrified interfaces and provide a robust framework for bridging molecular simulations and continuum electrochemical models.
BibTeX
@article{roncoroni2026deriving,
  author = {Fabrice Roncoroni and Abrar Faiyad and Yufeng Li and Tao Ye and Ashlie Martini and David Prendergast},
  title = {{Deriving effective electrode–ion interactions from free-energy profiles at electrochemical interfaces}},
  journal = {The Journal of Chemical Physics},
  year = {2026},
  doi = {10.1063/5.0321463},
}
Figure