Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

Nanoporous materials (NPMs) selectively adsorb and concentrate gases into their pores thus could be used to store, capture, sense many different gases. Modularly synthesized classes of NPMs, such as covalent organic frameworks (COFs), offer a large number candidate structures for each adsorption task. A complete NPM-property table, containing measurements relevant properties in would enable the matching NPMs with tasks. However, practice, matrix is only partially observed (incomplete); have not been measured. The idea this work leverage (NPM, property) values impute missing ones. Similarly, commercial recommendation systems entries an incomplete product–customer ratings recommend products customers. We demonstrate COF system match COFs tasks by training low-rank model COF–adsorption-property constructed from simulated uptakes CH4, H2O, H2S, Xe, Kr, CO2, N2, O2, H2 at various conditions. matrix, fit (COF, values, provides (i) predictions (ii) “map” COFs, wherein represented points, similar (dissimilar) congregate (separate). able rank reasonably well most properties, but imputation performance diminishes precipitously when fraction exceeds 60%. concepts our can applied broadly data pertaining properties.