Structure prediction of crystals, surfaces and nanoparticles

Abstract

We review the current techniques used in the prediction of crystal structures and their surfaces and of the structures of nanoparticles. The main classes of search algorithm and energy function are summarized, and we discuss the growing role of methods based on machine learning. We illustrate the current status of the field with examples taken from metallic, inorganic and organic systems. This article is part of a discussion meeting issue ‘Dynamic in situ microscopy relating structure and function’.

Energy landscapes of perfect and defective solids: from structure prediction to ion conduction

“…More generally, there are a number of diverse approaches to exploration of the energy landscape, as outlined, for example, by Woodley and Catlow [5], Oganov et al [6], and Woodley et al [14]. Methods include:…”

“…Methods based on machine learning are rapidly becoming more important [14]. All methods rely on the accurate and rapid calculation of energies either by quantum-mechanical methods, chiefly DFT, or by interatomic potentials (force fields), which are increasingly developed by machine learning trained on quantum-mechanical results for smaller systems.…”

Big data and machine learning for materials science

“…The prediction of crystal structures is among the most important applications of high-throughput experiments [ 71 ], which rely on ab initio calculations. DFT has been combined with ML to exploit interatomic potentials for searching and predicting carbon allotropes [ 72 ].…”

Discovery of a Low Thermal Conductivity Oxide Guided by Probe Structure Prediction and Machine Learning

“… [1] The initial identification of such lead materials is challenging because both the properties and the stability of a new material are determined by its structure and its composition in concert, neither of which can be known at the outset. [2] The vast chemical space impedes selection of composition, [3] while the absence of bounds on unit cell metrics and dimensionality [4] obstructs identification of structure. Here we tackle this challenge by fusing the prediction of unexplored structural motifs that will provide experimentally accessible new compositions with assessment of their properties by machine learning.…”