Stacked Ensemble Machine Learning for Range-Separation Parameters

Abstract

High-throughput virtual materials and drug discovery based on density functional theory has achieved tremendous success in recent decades, but its power on organic semiconducting molecules suffers catastrophically from self-interaction error until the optimally tuned range-separated hybrid (OT-RSH) exchange--correlation functionals were developed. The accurate but expensive fi�rst-principles OT-RSH transitions from a short-range (semi-)local functional to a long-range Hartree--Fock exchange at a distance characterized by the inverse of a molecule-specifi�c, non-empirically-determined range-separation parameter (ω). In the present study, we proposed a promising stacked ensemble machine learning model that provides an accelerated alternative of OT-RSH based on system-dependent structural and electronic con�figurations. We trained ML-ωPBE, the �first functional in our series, using a database of 1,970 organic semiconducting molecules with sufficient structural diversity, and assessed its accuracy and efficiency using another 1,956 molecules. Compared with the �first-principles OT-ωPBE, our ML-ωPBE reached a mean absolute error of 0:00504a0^{-1} for the optimal value of ω, reduced the computational cost for the test set by 2.66 orders of magnitude, and achieved a comparable predictive power in various optical properties.

Accurate Electronic and Optical Properties of Organic Doublet Radicals Using Machine Learned Range-Separated Functionals

“…84,104−109 The source code and database associated with the present study have been uploaded to the GitHub repository of the Lin Group. 182 Computational Details of Quantum Chemistry. All semiempirical tight-binding calculations as part of CMDs were performed using GFN2-xTB developed by Grimme and co-workers and the minimal STO-3G basis set in the xTB package, 101−103 with radical structures generated using RDKit 183 based on their simplified molecular-input line-entry systems (SMILES).…”