Optimizing impedance matching parameters for single-frequency capacitively coupled plasma via machine learning

Impedance matching plays a critical role in achieving stable and controllable plasma conditions in capacitive coupled plasma (CCP) systems. However, due to the complex circuit system, the nonlinear relationships between components, and the extensive parameter space of the matching network, finding optimal component values pose significant challenges. To address this, we employ an artificial neural network as a surrogate model for the matching system, leveraging its powerful pattern learning capability for a reliable and efficient search for matching parameters. In this paper, we designed four different parameters as optimization objectives and took the modulus of the reflection coefficient as an example to demonstrate the impedance matching optimization process of a CCP in detail using a particle-in-cell/Monte Carlo collision model. Our approach not only provides an effective optimization direction but also furnishes an entire parameter space that aligns with expectations, rather than just a single point. Moreover, the method presented in this paper is applicable to both numerical simulations and experimental matching parameter optimization.