TDDFT-as(p): a series of high accuracy semi-empirical TDDFT model using minimal auxiliary basis.

We report a general strategy to efficiently approximate Time-Dependent Density Functional Theory (TDDFT) calculations of molecular response properties such as UV absorption spectra. This strategy leads to a series of highly accurate semiempirical TDDFT models: TDDFT-as(p). In our benchmarking on 39 small to medium sized molecules, TDDFT-as has an energy error of 0.06 eV, 25% of currently best method sTDDFT with an energy error of 0.24 eV. The advantage of our approach is that we can improve the TDDFT-as model by adding one more p function per atom, which makes the TDDFT-asp model with an energy error of 0.04 eV. Moreover, TDDFT-as(p) needs a nearly trivial implementation and is naturally ready for existing quantum chemistry softwares such as Turbomole.

This work was supported by a startup fund from Case Western Reserve University. This work made use of the High Performance Computing Resource in the Core Facility for Advanced Research Computing at Case Western Reserve University.