Discovering mechanisms for materials microstructure optimization via reinforcement learning of a generative model

Abstract The design of materials structure for optimizing functional properties and potentially, the discovery novel behaviors is a keystone problem in science. In many cases microstructural models underpinning functionality are available well understood. However, optimization average via engineering often leads to combinatorically intractable problems. Here, we explore use reinforcement learning (RL) microstructure targeting physical mechanisms behind enhanced functionalities. We illustrate that RL can provide insights into driving interest 2D discrete Landau ferroelectrics simulator. Intriguingly, find non-trivial phenomena emerge if rewards assigned favor physically impossible tasks, which through rewarding agents rotate polarization vectors energetically unfavorable positions. further strategies induce curl be non-intuitive, based on analysis learned agent policies. This study suggests promising machine method material better understanding dynamics simulations.