Dissecting the ingredients of optimally tuned range-separated hybrid models for reliable description of non-adiabatic couplings.

Perusing the non-radiative processes requires a reliable prediction of non-adiabatic couplings (NACs) describing the interaction of two Born-Oppenheimer surfaces. In this regard, the development of appropriate and affordable theoretical methods that accurately account for the NAC terms between different excited-states is desirable. In this work, we develop and validate several variants of the optimally tuned range-separated hybrid functionals (OT-RSHs) for investigating NACs and related properties, such as excited states energy gaps and NAC forces, within the time-dependent density functional theory framework. Particular attention is paid to the influence of the underlying density functional approximations (DFAs), the short- and long-range Hartree-Fock (HF) exchange contributions, and the range-separation parameter. Considering several radical cations and sodium-doped ammonia clusters with the available reference data for the NACs and related quantities as the working models, we have evaluated the applicability and accountability of the proposed OT-RSHs. The obtained results unveil that any combination of the ingredients in the proposed models is not proper for describing the NACs, but a particular compromise among the involved parameters is needed to achieve reliable accuracy. Scrutinizing the results of our developed methods, the OT-RSHs based on the PBEPW91, BPW91, and PBE exchange and correlation DFAs, including about 30% HF exchange at the short-range regime, appeared to be the best performers. We also find that the newly developed OT-RSHs with correct asymptotic exchange-correlation potential have superior performances as compared to their standard counterparts with the default parameters and many earlier hybrids with both fixed and interelectronic distance-dependent HF exchange. The recommended OT-RSHs in this study can hopefully be applicable as computationally efficient alternatives to the expensive wave function-based methods for the systems prone to non-adiabatic properties as well as to screen out the novel candidates prior to their challenging synthesis.