Distributed memory parallelization of the multi-configuration time-dependent hartree method

The multiconfiguration time-dependent Hartree (MCTDH) method is an algorithm to efficiently propagate wavepackets. Since its development it has been established as one of the most important methods to study the quantum dynamics, e.g. excitations of vibrations and rotations or scattering events, of molecular systems. The method has been developed in Heidelberg over the last 18 years and is able to treat large systems with ∼9 degrees of freedom in its single processor application. As described in a previous work [1] the code has been parallelized for shared memory hardware and was successfully applied to the investigation of the Zundel cation [2,3] (15 internal degrees of freedom) using up to 8 processors with speedup factors of >7. However, the Heidelberg MCTDH-code was still lacking the ability to use the more powerful distributed memory hardware of computer clusters. This drawback has been removed recently by applying the Message Passing Interface (MPI) to MCTDH, as described in this report. © Springer-Verlag Berlin Heidelberg 2010.