Applications of Isometry Invariants on Material Property Prediction.

14^th May 2024, 14:00
Jonathan Balasingham
MIF

Abstract

Periodic material or crystal property prediction using machine learning has grown popular in recent years as it provides a computationally efficient replacement for classical simulation methods. A crucial first step for any of these algorithms is the representation used for a periodic crystal. While similar objects like molecules and proteins have a finite number of atoms and their representation can be built based upon a finite point cloud interpretation, periodic crystals are unbounded in size, making their representation more challenging. Isometry invariants offer a consistent way to represent and compare crystal structures. In the present work, we adapt two isometry invariants, the Pointwise Distance Distribution (PDD) and Average Minimum Distance (AMD) as a representation for our learning algorithm. The PDD and AMD distinguished all (more than 660 thousand) periodic crystals in the Cambridge Structural Database as purely periodic sets of points without atomic types. For machine learning models, the composition of a crystal contains crucial information needed for making accurate predictions. We introduce a method for combining the geometric information contained in the isometry invariants with compositional information and show its effectiveness on the commonly used Jarvis-DFT and Materials Project datasets.