ZeoSyn: A Comprehensive Zeolite Synthesis Dataset Enabling Machine-Learning Rationalization of Hydrothermal...

Area of Science:

• Materials Science
• Nanotechnology
• Chemical Engineering

Background:

• Zeolites are versatile nanoporous aluminosilicates crucial for catalysis, gas separation, and ion exchange.
• Hydrothermal synthesis offers control over zeolite properties but requires understanding complex synthesis-structure relationships.
• Existing zeolite synthesis databases are limited in scale and parameter scope.

Purpose of the Study:

• To introduce ZeoSyn, a comprehensive dataset of 23,961 zeolite hydrothermal synthesis routes.
• To develop and validate machine learning models for predicting zeolite products from synthesis parameters.
• To identify key synthesis parameters influencing zeolite crystallization using explainable AI.

Main Methods:

• Compilation of a large-scale dataset (ZeoSyn) with detailed synthesis parameters for 233 zeolite topologies and 921 organic structure-directing agents (OSDAs).
• Development of a machine learning classifier to predict zeolite products based on synthesis conditions.
• Application of SHapley Additive exPlanations (SHAP) to interpret model predictions and identify critical synthesis factors.

Main Results:

• The ZeoSyn dataset contains 23,961 unique zeolite hydrothermal synthesis routes.
• A machine learning classifier achieved >70% accuracy in predicting zeolite products from synthesis parameters.
• SHAP analysis revealed key parameters driving the crystallization of over 200 zeolite frameworks.

Conclusions:

• The ZeoSyn dataset and associated machine learning models provide a powerful tool for understanding and optimizing zeolite synthesis.
• Identifying pivotal synthesis parameters can guide the targeted synthesis of desired zeolites.
• This work facilitates advancements in materials design for catalysis, separation, and other applications.