粒子群算法在薄膜参数反演中的参数选择

Selection of the PSO Parameters for Inverting of Ellipsometry

为确保用于薄膜椭偏参数反演计算的粒子群算法收敛于全局最优值,采用标准粒子群算法,学习参数c1和c2采用线性关系kc1+c2=d,选取不同惯性系数ω,通过数值实验分析算法反演Si基底上SiO2薄膜光学常数的优化性能,结果表明粒子群算法的惯性系数ω取值范围在0.5～0.8之间,学习参数c1,c2配对之和不超过3,较小的c1与较大的c2相配对,可以保证用于椭偏参数反演的粒子群算法具有较好的优化性能.

The standard particle swarm optimization is used to guarantee the convergent global optima of the PSO which is applied to the inverting of ellipsometry. The linear relationship of kc1 ＋ c2 = d is applied to the parameters of c1 and c2. With different ωs selected, the optimal performance of the PSO for inverting the optical constant of film SiO2/Si was analyzed by numerical experiment . The result shows that,with the inertia weight co ranging from 0. 5 to 0. 8, the sum of learning parameters c1and c2 preferably is not more than 3,and that matching the smaller Cl with the larger c2 guarantees the better optimization performance of the PSO.