Efficient computational modeling of Raman spectra of liquids and solutions

Abstract

An efficient approach for an accurate quantum mechanical (QM) modeling of Raman spectra of condensed-phase systems is described. Energetically low-lying cluster structures of a molecule surrounded by an explicit shell of solvating molecules are efficiently generated at the semi-empirical tight-binding QM level and then re-optimized at the Density Functional Theory (DFT) level of theory. Such cluster models of a solvated molecule are shown to be sufficient to reproduce experimental vibrational frequencies and relative Raman intensities of several hydrogen-bonded liquids and aqueous solutions with the use of B3LYP-D3/def2-TZVP, ωB97X-3c, or B97-3c DFT methods in harmonic simulations, provided that the first solvation shell is included in the model. Analogous simulations at the computationally less demanding PBE-D3/def2-TZVP level provided less good, but still reasonably accurate, results. With the examples of acetone, acetonitrile, benzene, and their deuterated analogs, and the ionic liquid 1,3-dimethylimidazolium tetrafluoroborate, it is demonstrated that Raman spectra of liquids, where pronounced hydrogen bonds are absent, can be obtained in the gas-phase approximation. A comparison of absolute Raman intensities measured for gaseous and liquid acetone, acetonitrile, and benzene with our cluster simulations suggests that the inclusion of more than the first solvation shell is needed to reproduce the observed increase in Raman scattering cross sections in liquids relative to gases.