Machine learning determination of atomic dynamics at grain boundaries

Abstract

In polycrystalline materials, grain boundaries are sites of enhanced atomic motion, but the complexity of the atomic structures within a grain boundary network makes it difficult to link the structure and atomic dynamics. Here, we use a machine learning technique to establish a connection between local structure and dynamics of these materials. Following previous work on bulk glassy materials, we define a purely structural quantity (softness) that captures the propensity of an atom to rearrange. This approach correctly identifies crystalline regions, stacking faults, and twin boundaries as having low likelihood of atomic rearrangements while finding a large variability within high-energy grain boundaries. As has been found in glasses, the probability that atoms of a given softness will rearrange is nearly Arrhenius. This indicates a well-defined energy barrier as well as a well-defined prefactor for the Arrhenius form for atoms of a given softness. The decrease in the prefactor for low-softness atoms indicates that variations in entropy exhibit a dominant influence on the atomic dynamics in grain boundaries.

Reinforcing materials modelling by encoding the structures of defects in crystalline solids into distortion scores

“…Likewise, the potential energy of atom is positively correlated with its softness 2 , although there is a large spread for a given energy value. In this study, we observe the higher variance of d SVM compared to the statistical distances d RB , consistent with that previously reported by Sharp et al 2…”

“…Present-day materials science enables simulations of defect nucleation, recombination, migration and transition at the atomic scale by means of ultra large scale experiments [1][2][3] . Facilitated by the continuous increase in computational power and parallel computing, these objectives are achieved using traditional molecular dynamics (MD), quantum-classical QM/MM simulations [4][5][6] and by a rapidly growing number of fast exploring, biased in energy 7 or mean force 8,9 methods and other simulation schemes, such as accelerated MD 10 or statistical learning approaches 11 .…”

Machine learning and predicting the time-dependent dynamics of local yielding in dry foams

“…The foam structure exhibits weaker and stronger spots, be it due to the network structure or film properties [23]. In amorphous solids soft spots have been studied using measures for nonaffine deformation [24], Voronoi cell anisotropy [25], and machine learning tools [26,27] and the distribution of local yield stresses is an inherent property of plasticity models [28], as is the local relaxation dynamics after a yield event [29].…”

A model for grain boundary thermodynamics

“…An alternative view, therefore, can be to picture GBs with reference to their corresponding bulk phases. Thus, parallel to the current characterization techniques 57 and automated simulation methods 58,59 that are mainly focused on studying individual GBs, a general approach could be assessing the thermodynamic and kinetic properties of GBs with reference to the known bulk. For this purpose, we need to establish a physical framework that allows an approximation of the GB environment with respect to its reference bulk material.…”