Two-Photon Absorption Spectra Predicted by Semiempirical Methods

The molecular markers with large two-photon absorption cross-sections are highly sought for many applications, from photodynamic therapy to optical information processing. One may envision computational design of the chromophores combining desirable one-and two-photon absorption (1PA and 2PA) optical features with other properties important for the specific applications (solubility, toxicity etc.). Semiempirical methods are fast and seem to be well suited for the high throughput computational screening, which may be a part of such rational design strategy. Here we apply several semiempirical methods for prediction of 2PA spectra for substituted olygophenylvinylenes of donor-acceptor type. The predictions are compared to the experiment and previously reported results of the density functional theory (DFT) methods. We find ZINDO/S spectroscopic parameterization combined with single configuration interaction (CIS) method to perform far superior to the general purpose PM6 parameterization combined with various CI schemes. The poor performance in the prediction of 2PA cross-sections at PM6 level is traced to inaccurate excitation energies and incorrect transition dipoles between the 1PA and 2PA states. The state-to-state transition dipoles predicted by the ZINDO/S method are much better, as 2PA cross-sections reach the accuracy comparable to TD-DFT predictions after empirical corrections to the excitation energies. We conclude that semiempirical parameterizations can be used only for qualitative analysis of the 2PA properties. The excitation energies are in need of empirical correction, such as QSPR a linear regression for specific class of molecules. The ZINDO/S parameterization combined with multi-reference CI schemes (such as MR-CIS) could yield a good description of the intensities of 1PA and 2PA excitations, provided the right selection in reference configurations, order of excitation and active space. It can not be considered a black-box method, which makes it unsuitable choice for the high throughput screening technique.