New techniques for parallel simulation of high-temperature superconductors

We discuss several new techniques used for the simulation of high-temperature superconductors on parallel computers. We introduce an innovative methodology to study the effects of temperature fluctuations on the vortex lattice configuration of these materials. We have found that the use of uniform orthogonal meshes results in several limitations. To address these limitations, we consider nonorthogonal meshes and describe a new discrete formulation that solves the difficult problem of maintaining gauge invariance on nonorthogonal meshes. With this discretization, adaptive refinement strategies are used to concentrate grid points where error contributions are large (in this case, near vortex cores). We describe the algorithm used for the parallel implementation of this refinement strategy, and we present computational results obtained on the Intel DELTA.<<ETX>>