PiNNwall: Heterogeneous Electrode Models from Integrating Machine Learning and Atomistic Simulation.

Area of Science:

• Computational materials science
• Electrochemistry
• Machine learning

Background:

• Electrochemical energy storage relies on capacitive processes, often simulated using models like Siepmann-Sprik for metal electrodes.
• Existing models struggle with heterogeneous electrodes requiring chemical specificity, lacking analytical solutions.
• Recent extensions addressed electrode metallicity but not complex material compositions.

Purpose of the Study:

• To develop a novel computational approach for simulating heterogeneous electrode materials in electrochemical energy storage.
• To overcome limitations of current models in capturing chemical specificity for complex electrode systems.
• To enable accurate molecular simulations of advanced electrode-electrolyte interfaces.

Main Methods:

• Integration of atomistic machine learning (PiNN) for charge and response kernel generation.
• Coupling PiNN with classical molecular dynamics (MetalWalls) for electrochemical system modeling.
• Development of the PiNNwall interface for heterogeneous electrode simulations.

Main Results:

• The PiNNwall interface successfully models chemically doped graphene and graphene oxide electrodes.
• Demonstrated capability to simulate polarized oxide surfaces with coexisting proton and electronic charges.
• Established a new framework for incorporating chemical specificity into electrode simulations.

Conclusions:

• The PiNNwall interface provides a powerful tool for modeling complex and heterogeneous electrode materials.
• This advancement is crucial for designing and optimizing next-generation energy storage systems.
• Opens new avenues for molecular-level understanding of electrochemical interfaces.