(556f) Machine Learning Approach to Predict Adsorption Properties in Cofs

Adsorbent materials synthesized in modular fashion such as MOFs and COFs offer a choice of framework structures; scientist may tune the framework properties or choose from a large number of candidate materials to achieve the desired capacity or selectivity for gas separations and storage applications. In general, for a given adsorbent material only limited adsorption property data will be available in the literature.

Attempts have been made to impute the missing data and allow one to choose the best adsorbent material from the list for a given task [1]. Sturluson et al. proposed a linear low-rank matrix factorization approach to solve such a missing data problem for COF systems [1]. It is worth mentioning that this approach is suitable only if there exists a linear relationship between variables in the data set. Our preliminary analysis shows that some of the properties possess a nonlinear relationship.

In this work, we use a Kernel-based Machine Learning approach to capture the non-linearity in solving the above COF missing data problem. We hypothesize that this method will be able to extract the inherent non-linearity present in the data. As in the earlier work [1], the data for our work was also taken from the database reported by Ongaro et al. [2] and consisted of structure and adsorption properties of more than 800 COFs. In addition, we also included select structure properties of a given COF (such as pore size, pore volume and surface area) as an input to our model. We demonstrate the effect of inclusion of structure properties of the COFs and non-linear models in solving the missing data problem for prediction of their adsorption properties.

[1] Sturluson A, Raza A, McConachie GD, Siderius DW, Fern XZ, Simon CM. Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials. Chemistry of Materials. 2021; 33(18): 7203–7216.

[2] Ongari D, Yakutovich AV, Talirz L, Smit B. Building a Consistent and Reproducible Database for Adsorption Evaluation in Covalent–Organic Frameworks. ACS Central Science. Oct. 2019; 5(10): 1663–1675.