Designing Energy Materials Based on Machine Learning Potentials
발표자
()
초록
내용
Machine learning (ML) is increasingly being used to predict the physical and chemical properties of materials accurately by utilizing large databases. With the growing importance of precise nanoscale control in material science, predicting atomic-level phenomena with ML has become crucial. Traditional methods like density functional theory (DFT) are often used for this purpose but are limited by their high computational cost and applicability to simple conditions. Thus, developing force fields (FF) that can quickly and accurately predict nanoscale structural changes is essential. This presentation discusses the development of ML-based force fields (ML-FFs) for nanomaterials, using high-dimensional neural network potentials and Gaussian processes, trained on data from DFT and ab-initio molecular dynamics simulations. This approach offers a computationally efficient and accurate way to understand the physical properties and design principles of high-performance energy materials.