Spin‐flip approach within time‐dependent density functional tight‐binding method: Theory and applications

Abstract

A spin-flip time-dependent density functional tight-binding (SF-TDDFTB) method is developed that describes target states as spin-flipping excitation from a high-spin reference state obtained by the spin-restricted open shell treatment. Furthermore, the SF-TDDFTB formulation is extended to long-range correction (LC), denoted as SF-TDLCDFTB. The LC technique corrects the overdelocalization of electron density in systems such as charge-transfer systems, which is typically found in conventional DFTB calculations as well as density functional theory calculations using pure functionals. The numerical assessment of the SF-TDDFTB method shows smooth potential curves for the bond dissociation of hydrogen fluoride and the double-bond rotation of ethylene and the double-cone shape of H 3 as the simplest degenerate systems. In addition, numerical assessments of SF-TDDFTB and SF-TDLCDFTB for 39 S 0 /S 1 minimum energy conical intersection (MECI) structures are performed. The SF-TDDFTB and SF-TDLCDFTB methods drastically reduce the computational cost with accuracy for MECI structures compared with SF-TDDFT. K E Y W O R D S conical intersection, degenerate phenomena, long-range correction, spin-flip, time-dependent density functional tight-binding method

Understanding intermolecular interactions of large systems in ground state and excited state by using density functional based tight binding methods

“…It is hopeful that DFTB-EDA will be capable of describing excited states of single, double, and mixed excitation character when spin-flip TD-DFTB is employed. 75 With further development of DFTB, improvements in the accuracy of DFTB-EDA can also be expected.…”

Computational and data driven molecular material design assisted by low scaling quantum mechanics calculations and machine learning

“…154 A spin-ip TDDFTB (SF-TDDFTB) method is developed to fast determine S 0 /S 1 minimum energy conical intersection (MECI) structures. 155 In 2019, Nakai and co-workers implemented a GPU-accelerated DC-TDDFTB method, which could reproduce the experimental absorption and uorescence spectra of 2-acetylindan-1,3-dione in explicit acetonitrile solution. 156 It is not straightforward to precisely compare the accurate and costs of various low scaling QM methods because there usually exist some adjustable parameters for achieving the balance between the accuracy and computational costs.…”

“… 154 A spin-flip TDDFTB (SF-TDDFTB) method is developed to fast determine S 0 /S 1 minimum energy conical intersection (MECI) structures. 155 In 2019, Nakai and co-workers implemented a GPU-accelerated DC-TDDFTB method, which could reproduce the experimental absorption and fluorescence spectra of 2-acetylindan-1,3-dione in explicit acetonitrile solution. 156 …”

Excited state, non-adiabatic dynamics of large photoswitchable molecules using a chemically transferable machine learning potential

“…The gap error of 1.89 kcal/mol should be contrasted with the error of 15 kcal/mol in Ref. [9], which applied semi-empirical methods to azobenzene. The errors in the excited state forces are slightly larger, but still quite low.…”

“…Many methods have been developed to alleviate this steep computational cost. Semi-empirical methods [8][9][10] provide qualitatively correct results across many systems, but are ultimately bounded by their approximations, with average energy errors of 15 kcal/mol [9]. Other approaches focus on hardware and software development [11,12], but cannot match the speed of semi-empirical methods.…”