AI-Based Nano-Scale Material Property Prediction for Li-Ion Batteries

Abstract

In this work, we propose a machine learning (ML)-based technique that can learn interatomic potential parameters for various particle–particle interactions employing quantum mechanics (QM) calculations. This ML model can be used as an alternative for QM calculations for predicting non-bonded interactions in a computationally efficient manner. Using these parameters as input to molecular dynamics simulations, we can predict a diverse range of properties, enabling researchers to design new and novel materials suitable for various applications in the absence of experimental data. We employ our ML-based technique to learn the Buckingham potential, a non-bonded interatomic potential. Subsequently, we utilize these predicted values to compute the densities of four distinct molecules, achieving an accuracy exceeding 93%. This serves as a strong demonstration of the efficacy of our proposed approach.

Keywords

• quantum mechanics
• molecular dynamics
• artificial intelligence
• machine learning
• nanotechnology