Simulation On Semiconductor Devices Based On Evolutionary Computation

The semiconductor devices simulation takes an important part in devices design for it influences the quality and the cost of the product greatly. Recently, the simulation methods and simulation softwares consume time and computer space, and can not get the exact numerical solution, especially when solving high dimension and extremely nonliner equation groups. Wherefore, the evolutionary computation is introduced to solve the problem. We choose the PSO (Partical Swarm Optimization) algorithm based on reappearance group intelligence to simulate semiconductor devices, in order to supply a better method for semiconductor devices simulation technique.The main works are as follows:(1) The analysis of conventional computation method in semiconductor simulation. Based on analysing the semiconductor devices in the physical and mathematical foundation, we summarize the conventional computation method solving high-dimensional nonlinear equations in semiconductor devices simulation, expound the difficulties of solving discrete linear equations, which has a large scaled, sparse and ill conditioned coefficient matrix, analyse the advantages and disadvantages of Gaussian elimination method, Newton iterative method; Based on considering the the boundary conditions, we disperse and linearize the semiconductor devices equations, at last the matrix equation is received for further solving.(2) Introduce the semiconductor devices simulation method based on Cultrual Partical Swarm Optimization Algorithm(CPSO). The Cultural Algorithm(CA) is melt in PSO algorithm, namely, using the knowledge space of the CA to restrain the patical searching optimum point in optimized space, in this way reducing the search time, speeding up convergence rate, and also avoiding a local optimum. The analysis of typical example shows that CPSO is far better than PSO.(3) The semiconductor devices simulation based on CPSO algorithm. Firstly, disperse and linearize the partial differential equations by Finite Difference Method (FDM) to be a large sparse matrix equations, then solve it iteratively by CPSO algorithm. The simulation results of graded PN junction show that the CPSO algorithm is better than Newton iterative method and basic PSO algorithm. Through simulation of steady state diode, we get the electron concentration, hole concentration and the potential distribution, also they all meet the accuracy required, and greatly improve the speed of simulation. The application results show that the CPSO algorithm in the device simulation area is feasible and effective, provide a method for improving the performance of simulation software.