Evolution of two-dimensional grain boundary networks implemented in GPU

We consider a variational model for two-dimensional grain growth that allows for finite mobilities of boundaries
and boundary junctions. We use a collocation method to describe the grain boundaries and posit evolution
equations that ensure energy dissipation. The resulting algorithm is amenable to parallelization and a GPU
implementation described in this work leads to significant reduction in computational time. We benchmark our
code against known numerical and analytical solutions in some special cases. We also conduct an extensive
analysis of statistics that develop for various parameter regimes.