Article ID Journal Published Year Pages File Type
6929549 Journal of Computational Physics 2016 16 Pages PDF
Abstract
Ginzburg-Landau theory is one of the most powerful phenomenological theories in physics, with particular predictive value in superconductivity. The formalism solves coupled nonlinear differential equations for both the electronic and magnetic responsiveness of a given superconductor to external electromagnetic excitations. With order parameter varying on the short scale of the coherence length, and the magnetic field being long-range, the numerical handling of 3D simulations becomes extremely challenging and time-consuming for realistic samples. Here we show precisely how one can employ graphics-processing units (GPUs) for this type of calculations, and obtain physics answers of interest in a reasonable time-frame - with speedup of over 100× compared to best available CPU implementations of the theory on a 2563 grid.
Related Topics
Physical Sciences and Engineering Computer Science Computer Science Applications
Authors
, , , , ,