Ability of Local Structure to Predict Particle Rearrangements in Varying Spatial Dimension

In glassy systems, it is difficult to predict which particles will rearrange under thermal fluctuations or applied load using only local structural information. The most successful and least computationally intensive of the many approaches that have been developed to address this problem uses machine learning to construct a scalar function of local structure, “softness.” This particle-based quantity has been shown to be highly predictive of rearrangements in several model disordered systems in two and three dimensions. One might expect that the ability of local structure to predict rearrangements in a given model will decrease with increasing spatial dimension, e.g. due to greater homogeneity in particles' local environments. We use soft-sphere systems under athermal shear to study systematically the ability of softness to predict rearrangements in different spatial dimensions and test this hypothesis.