MSDE Symposium 2023: Frontiers in Molecular Engineering

Advanced featurization and characterization of MOF pores for adsorption applications Arun Gopalan 1 , Kaihang Shi 2 , Randall Q. Snurr 2 , Lev Sarkisov 1 1 University of Manchester, United Kingdom, 2 Northwestern University, U.S.A State-of-the-art approaches to the computational discovery of MOFs can be thought of in terms of three key components. 1) Representation : Network of atoms and bonds, a connected string of building blocks[1], etc. 2) Prediction : Molecular simulations, ab initio calculations, machine learning, etc. 3) S tructure sampler/generator : Databases like CoRE-MOFs [2] , ToBaCCo [3] , etc. or, optimizers based on Genetic Algorithm [4] , Bayes [5] , etc. How we represent materials is vital to all three steps. While the popular string and graph representations of MOFs primarily gear towards the chemical structure, adsorption properties correlate better with the characteristics of the 3D confining environment, which is unique to each MOF [6]. Hence, we present ‘pore graphs’ which are flexible enough to retain any cavity information relevant to the specific application, while being easy to compute, store and retrieve when used in characterization or machine learning problems. The void space in a material is converted into a network of interconnected pockets, using 3D image segmentation [7] . For example, let’s look at the pore graph workflow applied to Cu-BTC (HKUST-1) shown below.

Breaking down the void space into constituent pockets allows us to get the most out of existing pore descriptors. As examples, solutions to two challenging characterization problems are illustrated here:

• Detect pore windows (internal and periodic to the unit cell) • Detect the types of pockets (of different sizes/shapes)

Apart from characterization, pore graphs can be annotated with any node (pocket) or edge (connection) attribute (LCD, window size, energy histogram, etc.) and fed into graph ML algorithms to predict complex adsorption properties that are often inaccessible to simpler architectures, without having to work with cumbersome 3D images [8] . To summarize, pore graphs help us understand the cavities in a material better by representing it as a network of interconnected pockets. This reveals the underlying similarities in pockets, helps us solve many challenging characterization problems, and opens the possibility of adsorption property prediction via graph ML algorithms [9] . References 1. Bucior et al. , Cryst Growth Des, 2019, 19 , 6682–6697. Chung et al. , J Chem Eng Data, 2019, 64 , 5985–5998. Anderson et al. ,Crystengcomm, 2019, 21 , 1653–1665. 2. Lee et al. , ACS Appl Mater Inter, 2023, 13 , 23647–23654. 3. Taw et al., Adv Theory Simulations, 2022, 5 , 2100515. Bucio et al. , Mol Syst Des Eng, 2018, 4 , 162–174. Soille et al. , Signal Process, 1990, 20 , 171–182. Hung et al. , J Phys Chem C, 2022, 126 , 2813–2822. 4. Zhou et al. , AI Open, 2020, 1 , 57–81.

IP04

© The Author(s), 2023

Made with FlippingBook Learn more on our blog