Superconductivity---Elastic String Model

#raggedright433#

We developed a code for the numerical simulation of the planar motion of a one-dimensional elastic filament (single vortex) under tension, to investigate the properties of the vortex-glass state in high-temperature, Type-II superconductors. The computational problem required the time integration of a stochastic evolution equation; ensemble averages were obtained by considering the long-time behavior of the solution for a large number of realizations. The objective of the numerical simulations is to measure the resulting ``average'' velocity of the filament as a function of the applied force. In the study of the elastic filament model, we observed avalanche-type behavior when the applied forces are in the neighborhood of a critical transition value. Such behavior is characteristic of self-organized criticality phenomena. We have developed other elastic filament models to explore this phenomenon. These studies also require the accumulation of statistics from a large number of events. Each event involves the solution of a stochastic differential equation subject to a random initial perturbation. In the appropriate parametric state space, the system will enter a steady state for a sufficiently large number of events. The calculations are characterized by a large number of independent calculations that can occur simultaneously, a situation ideally suited for coarse-grained parallelism. The most difficult calculations for the elastic filament model occurred for very small applied forces when the system is in a ``glassy'' or ``creep'' state characterized by very slow dynamics which require extremely large amounts of computer time to establish the asymptotic behavior. To further study this, we developed an alternative model based on a static tilted potential, characteristic of creep motion. The calculation is characterized by large numbers of ensembles (#tex2html_wrap_inline464# 12,000) each corresponding to a random realization of a potential tilted by an applied force. For small forces, a very large number of spatial points (#tex2html_wrap_inline466# 75,000) are needed to resolve the potentials. In all three of these cases, the parallel approach used is coarse grained. Since each realization is both time consuming and independent of the other realizations, we were able to run a large number of these jobs in parallel. This work was initially started using a BBN TC2000 and a Sun SPARC workstation network. We ported the programs to the SP1 system, where we have found a significant improvement in execution time and the number of processors available. For some of the most difficult calculations, with very small applied forces, we reduced solution time from approximately five days on Sun SPARC workstations to approximately 17 hours on the SP1 system. %