Dynamic Classification of Defect Structures in Molecular Dynamics Simulation Data.

摘要: In this application paper we explore techniques to classify anomalous structures (defects) in data generated from abinitio Molecular Dynamics (MD) simulations of Silicon (Si) atom systems. These systems are studied understand the processes behind formation various defects as they have a profound impact on electrical and mechanical properties Silicon. our prior work presented for defect detection [11, 12, 14]. Here, present two-step dynamic classifier defects. The first step uses up third-order shape moments provide smaller set candidate classes. second assigns correct class structure by considering actual spatial positions individual atoms. is robust scalable size Each phase immune noise, which characterized after study simulation data. We also validate proposed solutions using physical model lattices. demonstrate efficacy correctness approach several large datasets. Our able recognize previously seen identify new real time.