Towards the Accurate Simulation of Electrochemical Interfaces by Combining Electron Density and Long-Range Machine-Learning Methods

TH Department Seminar
Date: Feb 22, 2024
Time: 02:00 PM (Local Time Germany)
Speaker: Dr. Andrea Grisafi
Institute of Data and Computational Sciences, Sorbonne Université, Paris, France
Location: https://zoom.us/j/94537687996?pwd=ZGpvVFpQT1owM20xajJnUjF0SU5qUT09
Room: Meeting ID: 945 3768 7996 | Passcode: 693535
Host: TH Department

The accurate study of electrochemical interfaces calls for simulation techniques that can treat the electronic response of metal electrodes under electrostatic perturbations. Despite recent advancements in atomistic machine-learning (ML) methods applied to electronic-structure properties, predicting the non-local behavior of the charge density in electronic conductors remains a majoropen challenge.

I will show how to tackle this problem by incorporating long-range structural information into an equivariant ML model capable of predicting the Kohn-Sham electron density of the system.¹ I will continue by deriving a finite-field extension of the method, which allows us to reproduce the electronic charge transfer through the metal electrode, as generated by the application of an external electric field. Crucially, I will show how these developments can be integrated within the simulation of a gold/electrolyte ionic capacitor under an applied voltage, enabling the accurate prediction of the differential electric capacitance. I will conclude outlining how future applications of the method will make it possible to drive the large-scale simulation of electrochemical cells with the accuracy of finite-field density functional theory.

¹Andrea Grisafi, Augustin Bussy, Mathieu Salanne and Rodolphe Vuilleumier, Predicting the charge density response in metal electrodes, Physical Review Materials 7, 125403 (2023).