머신러닝 기반 소재의 원자구조와 물성간의 양방향 맵핑
Bilateral Mapping Between Atomic Structures and Properties of Materials Via Machine-Learning
|
한상수 KIST For acceleration of material design, understanding a correlation between material structures and their properties is critical. Of various material properties, an electronic density of states (DOS) is one of key factors in condensed matter physics and material science that determines the properties of materials. First-principles density-functional theory (DFT) calculations have typically been used to obtain the DOS despite the considerable computation cost. Recently, we have developed a extremely fast machine-learning method for predicting the DOS patterns of not only bulk or surface structures but also nanoparticles in multi-component alloy systems by a combination of principle component analysis and crystal graph convolutional neural network. In the first part of this talk, I will cover the machine-learning method for the DOS prediction. And then, I will move to an inverse design concept via a machine-learning, in particular, for extracting material information (e.g., crystal structure, element, and composition) from the targeted electronic structure properties. In the second part of this talk, I will briefly discuss our idea for the inverse design and present relevant results that includes a crystal structure classification from diffraction pattern images and an element representation through color mapping. |