Informatics-based optimization of crystallographic descriptors for framework structures

Over the years there have been numerous attempts at classifying zeolites and other network structures. Zeolites and other network structures have complicated structures that have to be described using multiple types of descriptors. Researchers have tried different approaches for the development of such descriptors, based on experiments and theory. Each of these tries to take into account (separately) the topology and the symmetry of the structure being investigated. The symmetry aspects are usually dealt with in terms of space groups while the topological aspects are dealt with by using combinations of planar nets, chains, secondary building units, coordination sequences, and ring statistics.1-4 Our approach is informatics based, one that combines insights from both science and statistics to develop descriptors that can be used to study structure relationships. Here we use Wigner-Seitz cells in an attempt to encode simultaneously both the topological and symmetry properties of the network.5 We use these cells as a basis for development of “secondary” descriptors, which are then used to study structure relationships.