|
Authors: S.D. Ivanov, A. Witt, M. Shiga and D. Marx
Vibrational spectroscopy in the infrared (IR) regime is an ubiquitous tool widely used to probe complex dynamics of molecular systems, and thus the accurate prediction and interpretation of vibrational spectra is one of the important goals in theoretical chemistry. Although classical ab initio molecular dynamics (MD) methods have shown real success simulating complex many-body systems, still the most complicated and, therefore, interesting classes of systems are waiting for tools that would allow computing their dynamical real-time properties. Indeed, shallow potential energy landscapes, light reduced masses, effects of isotope substitution all require quantum effects of nuclei to be taken into account explicitly. Techniques based on Feynman's path integral (PI) formalism provide an elegant and, in principle, exact way to compute static properties. Nevertheless, methods that allow for computation of dynamic properties are called for in the framework of PIs. Currently, there exist two competing approaches, namely, Centroid and Ring Polymer PI molecular dynamics. The former, introduced by Cao and Voth [1] relies on the concept of the path centroids whose dynamics is shown to be a quasi-classical approximation to quantum dynamics. The latter, developed by Craig and Manolopolous [2] uses PIMD under certain conditions directly to extract the quasi-classical real-time evolution. Marx, Tuckerman and Martyna [3] developed an ab initio generalization of adiabatic centroid path integral molecular dynamics. The same idea can be naturally extended from ab initio CMD to ab initio RPMD as well. Importantly, these techniques are not restricted to finite systems such as molecules or clusters, but have considerable prospects to be applied to extended condensed matter systems in the future. Up to now, the two methods were compared using short time limit analysis and few model systems, like para-hydrogen. Also several ab initio computations of small molecules were reported. However, the advantages, drawbacks and limitations of the methods, especially in connection to IR spectroscopy of chemically interesting systems, have not been investigated systematically, which is the goal of this contribution. Since PI techniques do not depend on the way of computing forces, our investigations were performed using simple classical force fields. Even for tiny systems, like diatomic, water and methane molecules in the chemically interesting range of temperatures, we have found that both methods have intrinsic problems when it comes to computing IR spectra. The reasons of these failures are analysed and a possible remedy for adiabatic ab initio CMD is discussed. References: [1] J. Cao and G.A. Voth, J. Chem. Phys., v. 99, p. 10070 (1993) [2] I.R. Craig and D.E. Manolopolous, J. Chem. Phys, v. 121, p. 3368 (2004) [3] D. Marx, M.E. Tuckerman, and G.J. Martyna, Comput. Phys. Commun., v. 118, p. 166 (1999) |
|