A GPU implementation of classical density functional theory for rapid prediction of gas adsorption in nanoporous materials

Abstract

Nanoporous materials are promising as the next generation of absorbents for gas storage and separation with ultrahigh capacity and selectivity. The recent advent of data-driven approaches in materials modeling provides alternative routes to tailor nanoporous materials for customized applications. Typically, a data-driven model requires a large amount of training data that cannot be generated solely by experimental methods or molecular simulations. In this work, we propose an efficient implementation of classical density functional theory with a graphic processing unit (GPU) for the fast yet accurate prediction of gas adsorption isotherms in nanoporous materials. In comparison to serial computing with the central processing unit, the massively parallelized GPU implementation reduces the computational cost by more than two orders of magnitude. The proposed algorithm renders new opportunities not only for the efficient screening of a large materials database for gas adsorption but it may also serve as an important stepping stone toward the inverse design of nanoporous materials tailored to desired applications.