Crystal structure identification with 3D convolutional neural networks with application to high-pressure phase transitions in SiO$_2$

TL;DR

This paper develops and benchmarks 3D convolutional neural network architectures, PointNet and DG-CNN, for atomic structure identification, demonstrating their effectiveness on crystal structures and complex SiO2 phases, with applications to understanding crystallization under shock compression.

Contribution

It introduces neural network-based methods for atomic structure recognition, extending their application to complex SiO2 phases and integrating with visualization tools.

Findings

Neural networks outperform traditional methods on simple crystal structures.

The approach reveals insights into SiO2 crystallization under shock compression.

Integration with OVITO enhances usability for atomic simulations.

Abstract

Efficient, reliable and easy-to-use structure recognition of atomic environments is essential for the analysis of atomic scale computer simulations. In this work, we train two neuronal network (NN) architectures, namely PointNet and dynamic graph convolutional NN (DG-CNN) using different hyperparameters and training regimes to assess their performance in structure identification tasks of atomistic structure data. We show benchmarks on simple crystal structures, where we can compare against established methods. The approach is subsequently extended to structurally more complex SiO$_2$ phases. By making use of this structure recognition tool, we are able to achieve a deeper understanding of the crystallization process in amorphous SiO$_2$ under shock compression. Lastly, we show how the NN based structure identification workflows can be integrated into OVITO using its python interface.