Analysis of the Relationships between Synthesis Conditions and Structures of Zeolites Using Machine Learning

Abstract

The relationships between synthesis conditions and obtained structures of SDA-free aluminosilicate zeolites were analyzed using machine learning. Topology prediction indicated that misclassified topologies often shared a common composite building unit (CBU) with the actual topology. Binary classification characterized whether a specific CBU was observed in the products and visualized the distribution trends using t-SNE. The analysis results imply that zeolites containing mor or sod as CBUs in the framework are more likely to crystallize through these CBUs than zeolites containing can or d6r.

Graphical Abstract