It is demonstrated that significant accuracy improvements in MRSF-TDDFT can be achieved by introducing two different exchange−correlation (XC) functionals for the reference Kohn−Sham DFT and the response part of the calculations, respectively. Accordingly, two new XC functionals of doubly tuned Coulomb attenuated method-vertical excitation energy (DTCAM-VEE) and DTCAM-AEE were developed on the basis of the "adaptive exact exchange (AEE)" concept in the framework of the Coulomb-attenuating XC functionals. The values by DTCAM-VEE are in excellent agreement with those of Thiel's set [mean absolute errors (MAEs) and the interquartile range (IQR) values of 0.218 and 0.327 eV, respectively]. On the other hand, DTCAM-AEE faithfully reproduced the qualitative aspects of conical intersections (CIs) of transbutadiene and thymine and the nonadiabatic molecular dynamics (NAMD) simulations on thymine. The latter functional also remarkably exhibited the exact 1/R asymptotic behavior of the charge-transfer state of an ethylene−tetrafluoroethylene dimer and the accurate potential energy surfaces (PESs) along the two torsional angles of retinal protonated Schiff base model with six double bonds (rPSB6). Overall, DTCAM-AEE generally performs well, as its MAE (0.237) and IQR (0.41 eV) are much improved as compared to BH&HLYP. The current idea can also be applied to other XC functionals as well as other variants of linear response theories, opening a new way of developing XC functionals.
■ INTRODUCTIONSA widely used methodology for studying molecular excited states is the spin-conserving linear response time-dependent density functional theory (LR-TDDFT). 1−3 Despite the success of conventional LR-TDDFT in describing the lowlying excitation energies of molecules, there are well-known limitations. These include the energy of long-range charge transfer excitations within pure and hybrid exchange− correlation (XC) functionals, 4−8 excited states with substantial double excitation character, 9−12 states of molecules undergoing bond breaking, 13,14 and the topology of conical intersections (CIs). 15−18 There have been various attempts to address the issues of charge transfer 19,20 and double excitations. 10,21,22 These limitations can also be overcome more explicitly by the spin-flip model, 23 especially the spin-flip (SF)-TDDFT, 24−27 which utilizes spin-flip excitations from high spin triplet reference (M s = +1). Thus, the incorrect description of the CI, single bond breaking, and the poor description of multireference electronic states can be corrected by it. 24−26 However, SF-TDDFT may suffer from considerable spin contamination. 28−30 As it is due to the incomplete configurations of its response space, a fundamental solution to the problem is to include the missing ones. However, unlike wave function theories, a considerable challenge remains with respect to TDDFT when going beyond the adiabatic approximation to account for more than single excitations. 21 The introduction of the tensor equation-of-motion (TEOM)