PIC/MCC Simulation Of Low Temperature Plasma In MEMS Etching Process

During MEMS fabrications,low temperature plasma etching plays an important role in etching process.According to different plasma sources,plasma etching can be divided into capacitively coupled plasma etching and inductively coupled plasma etching,among which inductively coupled plasma etching has been widely concerned in recent years due to its high density,high selectivity and simple equipment.It will need a lot of time and cost to study the plasma discharge process only by experimental method.On the contrary,by means of numerical simulations,not only the parameters of the cavity can be changed flexibly,but also the research and development cycle can be shortened,which plays a guiding role in the distribution of plasma in the cavity and the optimal design of the cavity.In this paper,the Particle-In-Cell Monte Carlo method(PIC/MCC)is used to simulate the inductively coupled plasma discharge process.The simulation model is based on a two-dimensional cylindrical coordinate cavity.The program is mainly divided into six modules as follows: particle initialization,electromagnetic field solution,current source solution,particle push,collision treatment and boundary treatment.The finite difference time domain(FDTD)method of the central difference scheme is used for the electromagnetic field solution,the zigzag method is used for the current source solution,the frog leap method is used for the particle propulsion module,the Monte Carlo method is used for the collision treatment,the perfectly match layer method is used for the electromagnetic field boundary,and the reflection and absorption is used for the particle motion boundary.The disadvantage of the PIC/MCC approach is that the program takes a long time to run without parallel computing.Through CUDA programming,our program is accelerated on GPU with parallel computing.The modules with parallel acceleration include electromagnetic field solving module,current source solving module,particle pushing module and collision module,and the memory usage of parallel program is also optimized.Finally,the program acceleration ratio is obtained by comparing the running time of serial program and parallel program.