Master thesis defense in Quantum Information Science by Måns Johan Palm

Accelerating Monte Carlo Trajectory Simulations of Cold Atoms -- A GPU-Parallel pylcp Rewrite

Simulation software targeting computationally dempanding Monte Carlo simulations rely mostly on serial integration of independent atomic trajectories. This thesis presents a JAX-based, GPU-accelerated rewrite of pylcp that retains the original API while modernizing the codebase. Benchmarks demonstrate that batched GPU execution delivers 11.1x to 35.0x faster per-atom performance than the optimal parallel CPU configurations. The rewrite is validated against a four-stage 88Sr cooling cascade, successfully reproducing experimental steady-state temperatures and qualitative spatial behaviors. Overall, the updated package provides order-of-magnitude speedups for statistically significant ensembles without compromising the physics of the original model.