Forecasting low framework density zeolites from synthesis descriptors using machine learning

Abstract

The framework density (FD) of zeolites can be defined as the number of tetrahedrally coordinated atoms (T-atoms) per 1000 Å3 and it is used to distinguish zeolites from tectosilicates and is an important index for evaluating the structure-void ratio of a zeolite. While the synthesis of extra-large pore zeolites is well known in the literature, it is extremely challenging to prepare extra-large pore zeolites with a low framework density. Only a few extra-large pore zeolites with low framework density have been synthesized to date, which raises the question of why such zeolites are so difficult to synthesize? To obtain such novel zeolites, experimentalists use qualitative analysis to identify structural patterns in experimental data and fine-tune synthesis descriptors. The aim of this paper is to demonstrate that machine learning can be utilized to predict the framework density of zeolites from synthesis descriptors, thus providing a platform for developing new zeolite structures. A machine learning approach is applied in this study to correlate the molar composition of a gel containing 11 metal oxides, fluoride, water, hydroxide ions, and organic structure agents, as well as operating parameters, using a large collection of synthetic records of 151 zeolite frameworks compiled from the literature, along with the framework density and the type of ring that results. Besides providing excellent accuracy, the decision tree provides a heuristic insight into the chemistry of zeolite synthesis by showing the conditions that result in low framework density. The data as presented does not provide details. However, machine learning algorithms can be used to find hidden patterns in the data, providing a deeper understanding of the chemistry involved in the formation of low framework density zeolites. This approach can, in principle, be applied to the synthesis of any material to rationalize empirical knowledge and fine-tune the synthesis conditions to achieve the desired properties.