Density Functional Theory of Water with the Machine-Learned DM21 Functional

• DOI: 10.26434/chemrxiv-2022-73d0t
• Publication Date: 2022-03-10T10:25:48Z
• Authors: Etienne Palos; Eleftherios Lambros; Saswata Dasgupta; Francesco Paesani
• Keywords: Energetics; Coupled cluster

The delicate interplay between functional-driven and density-driven errors in density functional theory (DFT) has hindered traditional approximations (DFAs) from providing an accurate description of water for over 30 years. Recently, the deep-learned DeepMind 21 (DM21) been shown to overcome limitations DFAs as it is free delocalization error. To determine if DM21 can enable a molecular-level physical properties aqueous systems within Kohn-Sham DFT, we assess accuracy neutral, protonated, deprotonated clusters. We find that ability accurately predict energetics clusters varies significantly with cluster size. Additionally, introduce many-body MB-DM21 potential derived data expansion energy use simulations liquid function temperature at ambient pressure. size-dependent identified analysis small calculated result systematically overestimating hydrogen-bond strength and, consequently, predicting more ice-like local structure room temperature.