Identifying Interstitials in MD Simulations - Max Space Clustering Method

Abstract

Molecular Dynamics (MD) simulations are suitable to study the details of atomistic processes. Identification of interstitials is one of the main difficulties in the analysis of defect dynamics. To identify the interstitials, different methods use different criteria. All the methods, except the Wigner-Seitz method, assume some input parameters which may lead to an erroneous number of interstitials. We describe an unsupervised clustering algorithm, called Max Space clustering method, to identify interstitials without assuming any input parameters. We show that the algorithm is suited for identifying interstitials at high temperatures when amplitude of atomic oscillations become larger, a fact ignored by other models.