Phonon spectra from short non-thermally equilibrated molecular dynamics simulations

A novel method for extracting vibrational frequencies and eigenvectors from short non-thermally equilibrated molecular dynamics trajectories is presented. A first global estimation of the frequencies is followed by the determination of the eigenvectors through a least squares fit of the trajectory including orthogonality constraints. Then, the trajectory is projected onto each of the normal modes. At that point each projected trajectory contains mainly one frequency component, which is re-estimated. The frequency estimation is performed using the MUltiple SIgnal Classification (MUSIC) algorithm. The scheme is iterated until self-consistency in the frequencies is achieved. This constitutes a great improvement over the standard MUSIC procedure. The fact that no thermal equilibration is needed makes this method well suited for ab initio molecular dynamics. A calculation of the vibrational spectrum of the C60 molecule is provided as an example of the power of this new method.