Multi-Output Virtual Metrology for Physical Vapor Deposition Using Projective Selection Algorithm
Technologies such as virtual metrology (VM), which monitors fabrication processes and predict product properties without physical measurements have numerous positive impacts. In this paper, we propose a VM system that predicts multiple physical properties of metal layers after the physical vapor deposition. We employ the Projective Selection (ProjSe) algorithm, which is suitable for variable selection in multioutput problems, to investigate the relationship between process parameters and layer properties. The effectiveness of the feature selection process combined with different regression models is demonstrated on real-world datasets collected from semiconductor manufacturer Infineon Technologies AG.