The simulation of fluid flow on the nano-scale in the field of process engineering involves a large number of relatively small molecules. The systems we are simulating can be modelled using rigid molecular models assembled from sites with non-bonded short-range pair potentials Each of the sites is described by a set of parameters which are required for the calculation of interactions between sites of the same type. For the interaction of unequal sites, mixed parameter sets have to be calculated. This has to be done for each possible pair of sites. We describe an approach to precalculate and store those mixed parameter sets in a stream, which allows efficient access and gives the flexibility to add new site types easily.

Another focus of our work has been on software engineering techniques. Using the adapter design pattern, we achieved a complete decoupling of the physical parts of the simulation (e.g. molecule models and interactions) from the data structures and the parallelisation. This eases the further concurrent development of the software and reduces the complexity of the different modules. It also gives us the opportunity to swap modules in a plug-in like fashion.

Finally, we demonstrate the advantages of a “pair ownership” of processes for the parallelisation which allows the joint calculation of macroscopic values and the forces on molecules.