High-fidelity simulation of collective effects in electron beams using an innovative parallel method

Among the most challenging and heretofore unsolved problems in accelerator physics is accurate simulation of the collective effects in electron beams. Electron beam dynamics is crucial in understanding and the design of: (i) high-brightness synchrotron light sources — powerful tools for cutting-edge research in physics, biology, medicine and other fields, and (ii) electron-ion particle colliders, which probe the nature of matter at unprecedented depths. Serial, or even naively parallel, implementation of the electron beam’s self-interaction is prohibitively costly in terms of efficiency and memory requirements, necessitating simulation times on the order of months or years. In this paper, we present an innovative, high-performance, high-fidelity, scalable model for simulation of collective effects in electron beams using state-ofthe-art multicore systems (GPUs, multicore CPUs, and hybrid CPU-GPU platform). Our parallel simulation algorithm implemented on different multicore systems outperforms the sequential simulation, achieving a performance gain of up to 7.7X and over 50X on the Intel Xeon E5630 CPU and GTX 480 GPU, respectively. It scales nearly linearly with the cluster size. Our simulation code is the first scalable parallel implementation on GPUs, multicore CPUs, and on hybrid CPUGPU platform for simulating the collective dynamical effects of electron beams in accelerator physics.