The Office of Dean-Research and Department of Physics, SRM University-AP are jointly organising the 10th edition of the Eminent Guest Lecture Series on June 17, 2026. The university welcomes Prof. Adisak Boonchun, Associate Professor, Department of Physics, Faculty of Science, Kasetsart University, Bangkok, Thailand, to deliver a lecture on “Machine Learning Methods in Materials Simulation”.
Abstract
In computational materials science, we always face a big problem: choosing between accuracy and speed. Density Functional Theory (DFT) is very accurate for studying atoms, but it requires too much computer power. Because of this, standard DFT can only simulate small systems with a few hundred atoms for a very short time. This makes it very difficult to study real-world properties like heat transport, defects, or phase changes.
This lecture introduces how machine learning (ML) is breaking this limit. Today, we demonstrate the use of Machine Learning Force Fields (MLFF). Instead of solving complex systems, an on-the-fly MLFF model can learn the relationship between atomic structures and energy from a small set of DFT data. This allows us to run simulations with DFT-level accuracy but at the speed of simple classical models.
As a real example, this talk will show how we use ML methods to calculate the lattice thermal conductivity of semiconductors. We will discuss how to train the model efficiently to simulate heat transport in large systems, which is impossible with standard DFT alone. Finally, we will talk about the current challenges of AI in physics and look at the future of automated materials simulation.
