Parallelized two-dimensional particle-in-cell simulation for capacitively coupled plasmas using graphic processing units

A large size chamber with 450 mm and plasma wall-interaction for the nano-scale feature are needed in semiconductor manufacturing industry recently. However, there are difficulties to experiment and simulate a large scale reactor chamber and nano-scale features. In the simulation case, plasma simulation using particle-in-cell (PIC) method shows very high accuracy compared with fluid simulations. However, PIC method has a disadvantage of slow speed caused by individual calculation of lots of computational particles. Recently, new computing method using graphic processing units (GPUs) enables to make a low-cost and low-power personal super computer and there is a study to utilize the GPUs for the parallelization of PIC plasma simulators [1,2]. To overcome the heavy computation problem of a PIC method, we have developed a parallelized PIC code using GPUs. In this work, two-dimensional axisymmetric simulator for a capacitively coupled plasma (CCP) sources is presented with the performance improvement using GPUs. Also, investigated were plasma phenomena with frequency variation as well as dual frequency with phase variation for CCP.