Quantification of Crystal Packing Similarity from Spherical Harmonic Transform

In this work, we present a new computational approach to characterize and classify molecular packing in the solid states. The key idea is to project each neighboring molecule (or short contact) from the centered molecule into a unit sphere according to the interaction energy. Consequently, the similarity between two spherical images can be evaluated from the spherical harmonics expansion based on the maximum cross-correlation. We apply this approach to successfully reproduce the previous packing assignment on a small amount of data with an improved categorization. Furthermore, we conduct a packing similarity analysis over 2000 hydrocarbon crystal data sets and uncover a set of abundant packing motifs. Unlike the previous approaches based on the subjective visual comparison at the real space, our approach provides a more robust way to measure the packing similarity, thus paving the way for a rapid classification of large scale crystal data.