We have elaborated a parallel program for molecular dynamics simulations
of 2-D Lennard-Jones (LJ) systems with large number of molecules on
networked workstations and on multiprocessors. The program is
appropriate for mixtures of several kinds of particles interacting via
LJ potentials characterised by different values of 
and
parameters.
The program is written in C language and optimised for the execution time and memory requirements. The computational complexity is reduced by use of distributed neighbor (Verlet) and link (Hockney) lists. On heterogeneous or loaded networks load balance is obtained by an adaptive repartitioning of the computational box during a simulation run. The load balancing procedure is based on the assumption that time of computation scales linearly with the number of particles.