A Cheminformatics Approach for Zeolite Framework Determination

Knowledge of the framework topology of zeolites is essential for multiple applications. Framework type determination relying on the combined information of coordination sequences and vertex symbols is appropriate for crystals with no defects. In this work we present an alternative machine learning model to classify zeolite crystals according to their framework types. The model is based on an eighteen-dimensional feature vector generated from the crystallographic data of zeolite crystals that contains topological, physical-chemical and statistical descriptors. Trained with sufficient known data, this model predicts the framework types of unknown zeolite crystals within 1-2 % error and shows to be better suited when dealing with real zeolite crystals, all of which always have geometrical defects even when the structure is resolved by crystallography.