locating the time between events or collisions and advancing the system to the next event.
Molecular dynamics simulation techniques for systems interacting with discontinuous potentials are discussed. Optimization Due to the simplicity of its time integration, DMD offers and efficiency techniques are summarized for performing discontinthe opportunity to explore phenomena that occur on long uous molecular dynamics on serial computers with direct applicatime scales using relatively inexpensive hardware. Even tion to polymer-like fluids. Comparisons are presented for two though long-time phenomena can be studied using continualgorithms: (1) single-event scheduling, and (2) multiple-event ous potentials ( which are efficient on vector supercomputscheduling. The single-event scheduling algorithm is approximately 75% faster than the multiple-event scheduling algorithm for molecuers) this is not always practical due to the small time steps lar fluids but yields equivalent performance for atomic fluids. For involved and the unfavorable economics associated with the single-event scheduling method, a combination of link lists and large allocations of machine time. In contrast, the DMD neighbor lists are used when searching for possible particle interacmethod can cover large time domains on less expensive tions. The combination of efficiency techniques permits multibillion time step simulations for relatively large systems on desktop work-desktop workstations and workstation clusters. Recent adstations. Both discontinuous molecular dynamics codes for single vances in the speed and performance of serial computers and multiple-event scheduling algorithms are available on the Inmakes simulation of large systems interacting with disconternet. The utility of the method is demonstrated for entangled tinuous potentials attractive for investigating phenomena chains, tethered chains, and heteronuclear chain mixtures. แฎ 1997 operating on time scales that cover many orders in magni-Academic Press tude. Furthermore, the development of low-cost, highspeed memory makes treatment of large systems possible * Corresponding author. 16