Data science shows that entropy correlates with accelerated zeolite crystallization in Monte Carlo simulations

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Zeolite nanoporous crystals formation is a slow process, often taking weeks.
• Organic structure-directing agents (OSDAs) were previously shown to accelerate zeolite crystallization, but the underlying physical factors remain unclear.

Purpose of the Study:

• To apply data science techniques to Monte Carlo (MC) simulation trajectories to identify factors accelerating zeolite crystallization.
• To understand the differences between one-OSDA (1OSDA) and two-OSDA (2OSDA) synthesis conditions in accelerating zeolite formation.

Main Methods:

• Utilized the smooth overlap of atomic positions (SOAP) approach to encode simulation snapshots, capturing two- and three-body atomic correlations.
• Employed Support Vector Machine (SVM) for classification of simulation datasets, achieving high accuracy (AUC=0.99 with 3-body, 0.94 with 2-body correlations).
• Performed pair entropy calculations to investigate information heterogeneity and its correlation with crystallization speed.

Main Results:

• Principal Component Analysis (PCA) failed to distinguish between 1OSDA and 2OSDA simulations.
• SVM successfully classified simulation structures, indicating significant differences in information heterogeneity between 1OSDA and 2OSDA.
• Entropy differences calculated during crystallization semi-quantitatively explained the observed acceleration in MC simulations.

Conclusions:

• Altering synthesis conditions to significantly change the entropy of labile silica networks can accelerate zeolite crystallization.
• The study suggests that controlling network entropy is a key factor for optimizing zeolite synthesis speed.
• Findings provide insights for developing new strategies to accelerate the formation of nanoporous materials.