Fig. 3: Clustering of diamond-like environments in structures of 1,000 atoms.
From: Signatures of paracrystallinity in amorphous silicon from machine-learning-driven molecular dynamics
a Count of the number of atoms per cluster, presented for structures with different SOAP similarity values from 0.89 to 0.94, a range which contains only paracrystalline structures. The boxes show the interquartile range (IQR), with the median shown as a line in the box, and the whiskers show the range of the distribution within 1.5 × IQR and exclude outliers, or points lying beyond this threshold, shown in gray. b Count of the number of clusters per structure, presented for every other SOAP similarity values ranging from close to CRN (0.89) to diamond (0.99). The violin plots present a box plot of the distribution inside a kernel density estimate of the distributions, with kernel bandwidth of 1. c Distribution of shortest distance between clusters. For all three plots, structures were sorted into categories (binned) according to SOAP similarity to c-Si, and the colormap indicates the total fraction of polycrystalline structures within a given bin. The threshold between paracrystalline and polycrystalline structures is 15% of locally crystal-like atoms, as determined by Polyhedral Template Matching. Source data are provided in the Source Data file.
