Nanocrystal energetics via quantum similarity measures

Abstract

We first develop a descriptor-based representation of atomic environments by devising two local similarity indices defined from an atom-partitioned quantum-chemical descriptor. Then, we employ this representation to explore the size-, shape- and composition-dependent nanocrystal energetics. For this purpose, we utilize an energy difference mu that is related to the atomic chemical potential, which enables one to characterize energetic heterogeneities. Employing first-principles calculations based on the density functional theory for a set of database systems, namely unary atomic clusters in the shape of regular polyhedra and the bulk solids of C, Si, Pd and Pt, we explore the correlations between the energy difference mu and similarity indices. We find that there exists an interconnection between nanocrystal energetics and quantum similarity measures. Accordingly, we develop a means for computing total energy differences from the similarity indices via interpolation, and utilize a test set comprising a variety of unary nanocrystals and binary nanoalloys/nanocompounds for validation. Our findings indicate that the similarity-based energies could be utilized in computer-aided design of nanoparticles.