CO2 on Graphene: Benchmarking Computational Approaches to Non-Covalent Interactions

Designing and optimising graphene-based gas sensors in silico entails constructing appropriate atomistic representations for the physisorption complex of an analyte on an infinite graphene sheet, then selecting accurate yet affordable methods for geometry optimisations and energy computations. In this work, diverse density functionals (DFs), coupled cluster theory, and symmetry-adapted perturbation theory (SAPT) in conjunction with a range of finite and periodic surface models of bare and supported graphene were tested for their ability to reproduce experimental adsorption energies of CO2 on graphene in a low-coverage regime. Periodic results are accurately reproduced by the interaction energies extrapolated from finite clusters to infinity. This simple yet powerful scheme effectively removes size dependence from the data obtained using finite models, and the latter can be treated at more sophisticated levels of theory relative to periodic systems. While for small models inexpensive DFs such as PBE afford surprisingly good agreement with the gold standard of quantum chemistry, CCSD(T), interaction energies closest to experiment are obtained by extrapolating the SAPT results, and with non-local van der Waals functionals in the periodic setting. Finally, none of the methods and models reproduce the experimentally observed CO2 tilted adsorption geometry on Pt(111) support, calling for either even more elaborate theoretical approaches, or a revision of the experiment.