Implementations of Semiclassical Approximations for Spectroscopic and Vibrational Eigenfunction Calculations

Semiclassical Initial Value Representaiton (SCIVR) molecular dynamics has been known since a long time to be able to calculate accurately vibrational power spectra of small, isolated molecules with inclusion of quantum effects, such as zero-point energies, overtones, tunneling splittings, and quantum resonances.[1] Some recent methodological advancements introduced by my group and based on a divide-and-conquer approach (DC SCIVR)[2] have allowed to apply SCIVR to large molecular systems, up to several dozens of atoms, by reproducing not only nuclear power spectra but also vibrational eigenfunctions. In this talk I will mainly focus on applications. First, I will show how we implemented DC SCIVR with a machine learning algorithm and applied it to NMA spectra calculations and how SCIVR can deal with the solvation issue by studying water clusters[3] and discussing the issue of the OH band in the formic acid dimer. Simulations of some relevant spectral features of nucleobases, nucleosides, solvated thymidine will be used to compare results based on precise ab initio on-the-fly semiclassical dynamics with those relying on force fields.[4] In the second part, I will show how one can calculate the vibrational energy eigenfunctions of molecules starting from classical trajectory information only and how multidimensional anharmonic vibrational eigenfunctions can provide a picture of the molecule, which is quite different from the standard normal mode one [5] Eventually, I will show how we can deal with the molecular adsorption on TiO2(101) Anatase surface using direct ab initio molecular dynamics and reproducing spectra of adsorbed molecules.