Bruner Adam, LaMaster Daniel, Lopata Kenneth
Department of Chemistry, Louisiana State University , Baton Rouge, Louisiana 70803, United States.
Center for Computation & Technology, Louisiana State University , Baton Rouge, Louisiana 70803, United States.
J Chem Theory Comput. 2016 Aug 9;12(8):3741-50. doi: 10.1021/acs.jctc.6b00511. Epub 2016 Jul 14.
We present a method for accelerating the computation of UV-visible and X-ray absorption spectra in large molecular systems using real-time time-dependent density functional theory (TDDFT). This approach is based on deconvolution of the dipole into molecular orbital dipole pairs developed by Repisky, et al. [Repisky et al., J. Chem. Theory Comput. 2015, 11, 980-911] followed by Padé approximants to their Fourier transforms. By combining these two techniques, the required simulation time is reduced by a factor of 5 or more, and moreover, the transition dipoles yield the molecular orbital contributions to each transition, akin to the coefficients in linear-response TDDFT. We validate this method on valence and core-level spectra of gas-phase water and nickel porphyrin, where the results are essentially equivalent to conventional linear response. This approach makes real-time TDDFT competitive against linear response for large molecular and material systems with a high density of states.