Predicting mechanical properties of crystalline materials through topological analysis

Abstract

With the aim to develop simple, programmatically generated, topology-based descriptors of crystal structures for application to mechanical properties prediction methods, we have developed a new geometric analysis protocol using the CSD Python API. By scanning a crystal structure for Miller planes with the least physical impediment to translational slipping, we are able to predict plausible slip planes for a crystal structure – a feature shown to correlate strongly with tabletability. A simple, automatic hydrogen-bond network dimensionality analysis method has also been developed which, when used in conjunction with the slip plane analysis, can detect whether the proposed slip plane is bridged by hydrogen bonding interactions. These methods are combined with other calculated topological features into a set of crystal structure descriptors, which provide a fast, qualitative way to describe the aspect of crystal plasticity that results from packing geometry. While intended for use in conjunction with other methods, these descriptors alone are shown to correctly predict the relative tabletability measured for multiple drug systems.