MlCOFSyn: A Machine Learning Framework To Facilitate the Synthesis of 2D Covalent Organic Frameworks

Two-dimensional covalent organic frameworks (2D COFs) have been historically synthesized empirically, often resulting in uncontrolled crystallization and inferior crystal sizes, which limit their performance in various applications. Recently, crystallization models tailored for 2D COFs have been developed that demonstrate great potential in facilitating their rational synthesis. Nevertheless, effective strategies to leverage these models for 2D COF synthesis remain underdeveloped, and the specialized expertise required, combined with the high computational costs of exploring the vast chemical space, poses additional barriers to their practical application. In this work, we present a machine learning framework, named MlCOFSyn, that is designed to assist in the synthesis of 2D COFs. This framework explores the application of 2D COF crystallization models by implementing three pivotal functionalities: predicting crystal sizes based on the input monomer addition sequence, reverse-engineering monomer addition sequences to achieve desired crystal sizes, and optimizing monomer addition sequences to produce larger crystals. These functionalities are critical for the controlled synthesis of 2D COFs but have been largely underexplored due to the lack of accessible theoretical tools. The MlCOFSyn framework leverages efficient machine-learning algorithms and features an intuitive graphical interface, enabling its use on consumer-grade computers by nonexperts. By addressing these gaps, the MlCOFSyn framework represents a substantial advancement in facilitating 2D COF research and synthesis.