Atomistic simulations of photophysical processes require efficient electronic structure methods, that can describe excited states at least qualitatively. While TD-DFT is already quite efficient, for non-adiabatic molecular dynamics simulations on large molecules a still faster method is needed, since the electronic structure has to be solved thousands of times for each simulation run. In tight-binding TD-DFT (TD-DFTB) approximations are made that are reminiscent of semiempirical methods such as AM1 or MNDO, with the difference that the underlying theory that is approximated is not wavefunction- but density-based.
In our group we develop our own code for tight-binding DFT in combination with surface-hopping dynamics , which has the advantage that new ideas can be tested and incorporated quickly. One of them is a long-range correction , which supresses the low-lying charge-transfer states that uncorrected tight-binding DFT produces erroneously in large molecular aggregates.