Research data supporting "Signatures of paracrystallinity in amorphous silicon"

This dataset supports the following study: https://www.nature.com/articles/s41467-025-57406-4

Dataset ranging from disorder to order

The dataset contains 3,069 a-Si structures, for a total of approximately 1.3 million atomic environments. These structures range from highly disordered to more crystalline-like. It is separated into .xyz files for each cell structure size.

A pickled dataframe is also provided with additional information on the structures, which can be loaded with pandas as follows:

import pandas as pd
df=pd.read_pickle('./data/df_rev1.pckl.gzip',compression="gzip")
df.keys()

Index(['ase_atoms', 'nb_atoms', 'size', 'vol_per_atom', 'label', 'nnb',
   'Category_2', 'Category_color_2', 'gap_energy', 'dE_gap', 'gap_at_E_NN',
   'mtp_energy', 'dE_mtp', 'mtp_at_E_NN', 'forces', 'F_max',
   'soap_sim_cSi', 'atomistic_soap_sim_cSi', 'PTM', 'CNA', 'stein_sim'],
  dtype='object')

Selected structures

We analyze four structural models (denoted I through IV) of 1,000 atoms of increasing paracrystallinity, which corresponds to the indices 2512, 2545, 2561 and 2568 of the dataframe, respectively.

We also compare prototypical structures from each category: thse indices are 2491 for CRN, 2576 for Paracrystalline and 2604 for Polycrystalline.

Large-scale structures

We also provide larger structural models of 100,000 atoms, namely a paracrystalline and a polycrystalline structure generated with quench rates of $10^{11}$ (top) and $10^{10}$ K/s (bottom) respectively.

Structure generation

All structures were generated following the protocol described in the manuscript. These simulations were carried out in LAMMPS using the $M_{16}^{''}$ ML potential. Only the ultimate frame from each trajectory was added to the database, as shown below:

Data analysis

Scripts for the analysis and plotting of all figures in the manuscript are provided in the scripts and src folders.