HIPNN

通过机器学习预测磷光能和推断波函数定位

sets using HIPNN and HIP-loc, training and testing parity plots of predicted versus true D E on thermal conformers sampled around equilibria of S 0 and T 1 using HIPNN and HIP-loc, RMSD of optimized geometries using the HIP-loc T 1 potential and energy error plots at those geometries, absolute errors in D E as a function of number of atoms, parity plots of predicted versus true D E for the extensibility set categorized by chemical similarity, localization of singlet – triplet transition for select molecules of the extensibility set computed from DFT spin density and HIP-loc weights, conformation-dependent localization of singlet – triplet transitions in molecules with a single torsional angle, and molecular animations of torsional scans including that of the molecule in Fig.5。参见doi：10.1039/d1Sc02136b