加利福尼亚大学洛杉矶分校

Time: 10:00 am, December 30, 2011

Place: 3楼会议室

Abstract: Particle-in-cell (PIC) simulations employ the most fundamental description of plasmas, that is particles interacting with electromagnetic fields, and because they require no a priori assumptions to model the physics they can be thought of as computer experiments. This property makes the PIC method very expensive yet unparallel in its ability to validate theories and model experiments. Although the method was originally developed to study microscopic physics in collisionless plasmas, such as Landau damping, plasma echo, and Debye shielding, recent advances in computer hardware and algorithms have allowed PIC simulations to produce quantitative agreement with experiments in a wide range of conditions, and the UCLA simulation group has been in the forefront of these developments in the past 40 years. On the other hand, as of November 2011, the top 10 supercomputers in the world have all achieved more than 1 Peta FLOP using LINPAC, with two out of these ten computers located in China. With the emergence of these machines, detailed kinetic physics in HED plasmas, such as parametric instabilities, laser wakefield accelerators and laser driven ion acceleration can be explored with ever greater detail. In this talk I will discuss the current trend in supercomputing, how PIC codes can be adapted for these emerging architectures, and give estimates on the size of the problems that can be studied using peta scale supercomputers such as the Tianhe 1A. Finally, I will present the relevant kinetic physics involved in each of the problems described above, and explain in detail how PIC codes can be properly used to study these problems.