Accelerated calculation of configurational free energy using a combination of reverse Monte Carlo and neural network models: Adsorption isotherm for 2D square and triangular lattices

TL;DR

This paper introduces a hybrid method combining reverse Monte Carlo and neural networks to efficiently calculate configurational free energy and adsorption isotherms for 2D lattices, significantly reducing computational time.

Contribution

The study presents a novel ANN-RMC approach that accurately estimates free energy and adsorption isotherms, outperforming traditional Monte Carlo simulations in speed.

Findings

Accurate adsorption isotherms obtained within seconds.

Method validated against Monte Carlo calculations.

Applicable to various adsorbate interactions and conditions.

Abstract

We demonstrate the application of artificial neural network (ANN) models to reverse Monte Carlo based thermodynamic calculations. Adsorption isotherms are generated for 2D square and triangular lattices. These lattices are considered because of their importance to catalytic applications. In general, configurational free energy terms that arise from adsorbate arrangements are challenging to handle and are typically evaluated using computationally expensive Monte Carlo simulations. We show that a combination of reverse Monte Carlo (RMC) and ANN model can provide an accurate estimate of the configurational free energy. The ANN model is trained/constructed using data generated with the help of RMC simulations. Adsorption isotherms are accurately obtained for a range of adsorbate-adsorbate interactions, coverages and temperatures within few seconds on a desktop computer using this method. The results are validated by comparing to MC calculations. Additionally, H adsorption on Ni(100) surface is studied using the ANN/RMC approach.