ABSTRACT
Over the last two decades, classical trajectory simulations (CTS) [1] have emerged as the method of choice, for computing the dynamics of atomic nuclei in molecular and condensed matter systems. Such methods are relatively easy to implement, scale linearly or quadratically in the number of atoms, and can be trivially parallelized across many cores of a supercomputing cluster or grid. On the other hand, quantum dynamical effects (tunneling, dispersion, interference, zero-point energy, etc.) are known or suspected to be important for many systems of current interest, and there is a demand for reliable numerical methods that treat quantum effects well. This has motivated a host of theoretical and computational improvements, e.g., mixed quantum-classical methods [2], semiclassical methods [3,4], centroid dynamics [5], and trajectory surface hopping (TSH) [6,7], all of which are designed, in principle, to handle large and complex systems of the sort routinely treated using CTS [1]. However, as these methods all treat quantum effects only approximately, and generally do not enable systematic convergence to the exact quantum result, it remains a vital but largely unanswered question to what extent, and for which applications, these methods actually capture the relevant quantum dynamical effects for the system in question [8].
