Multi-Scale Neural Network Molecular Dynamics Simulations for Polar Topology Control in Next Generation Ferroelectric Materials.

Quantum simulations methods such as quantum molecular dynamics offer highly accurate understandings of material vibrational and structural properties, but are seriously limited in the length and time scales they are able to investigate due to the large computational cost. Over the recent years the incorporation of machine-learning based inter-atomic potentials trained on quantum data has opened the possibility to study much larger spatiotemporal scales at quantum accuracy due to orders of magnitude reduction in computation cost of the machine-learning model. In this talk we will focus on the use of machine learning based molecular dynamics in the rapidly developing field of polar “topotronics”, or the control of polar topologies in ferroelectric materials for development of next generation topological based electronics. We will discuss the use of multi-scale neural network quantum molecular dynamics simulations to explore optical, mechanical, and electrical control of ferroelectric materials. The demonstrated multi-scale quantum simulation and machine learning frameworks are not limited to ferroelectrics and offer an exciting new avenue for exploring quantum material dynamics