Multipass cell design with the random walk and gradient descent optimization algorithms

Abstract

An automated approach is presented for optimizing the multipass cell (MPC) design with dense patterns in this paper. First, a strategy based on the random walk (RW) algorithm is implemented for global exploration to determine the parameters of the target MPC configuration and accelerate the design process. Second, the gradient descent (GD) algorithm is performed for local exploitation to optimize the re-entrant condition in a fast and automatic way. In addition, we apply the clustering method to identify the desired spot patterns with specific properties automatically. Finally, the proposed algorithms are tested in the optimization of two types of densely patterned MPC under the re-entrant condition. The results presented in this paper clearly show that the proposed approach is effective and efficient in optimizing the MPC design automatically and can be further utilized in more complex optical configurations.