Global Optimization of Thin-Film Properties in PECVD System Harnessed by Complex-System-Response Platform

The silicon nitride films produced by plasma enhanced chemical vapor deposition (PECVD) are crucial for semiconductor devices due to their excellent mechanical and electrical properties. However, conventional methods for controlling film properties rely on extensive process testing, leading to inefficiencies and high costs. This study introduces the Complex-System-Response (CSR) optimization method to predict the optimal parameter combinations for deposition rate, residual stress, dielectric constant, and surface roughness with minimal experimental trials. The results demonstrate that the optimal parameter combinations predict the global optimal values of these four properties with an accuracy exceeding 92.8%. Furthermore, the global optimal property values achieve enhancements from 1.02 to 73.5 times than those reported in previous studies, highlighting the potential of the CSR method in optimizing the deposition process. Additionally, based on CSR method, we propose an approach for achieving multi-objective optimization, balancing all properties to maximize their proximity to their respective global optimal values.