Deep-Learning-Based Proxy Modeling for Microscopic Process of Plasma Etching

Dynamic modeling of the microscopic process of plasma etching is crucial for reducing production costs and improving production efficiency. To alleviate issues such as poor real-time performance and limited transferability of typical numerical simulation algorithms, researchers have employed neural networks. Nonetheless, the considerable expenses often pose challenges in acquiring sufficient real datasets. Hence, this paper proposes the Deep-Learning-Based proxy model, combined with the Convolutional Autoencoder and Recurrent Neural Network. The model not only accelerates the simulation speed of plasma etching, but also achieves accurate predictions.