Enhancing Ultrasonic Thin Film Measurement Using PCA and DFT Driven Machine Learning Models

Ultrasonic measurement of ultra-thin films, ranging from a few nanometers to tens of nanometers, has proven to be a reliable and non-destructive technique widely used in semiconductor manufacturing. Precise process control at each step of semiconductor manufacturing is crucial for the performance of semiconductor devices. In this paper, we demonstrate that machine learning (ML) can effectively address the challenges of ultrasonic measurement for multi-layer thin films, where traditional modeling techniques may fall short. We also show that Principal Component Analysis (PCA) and Discrete Fourier Transform (DFT) can be employed for feature extraction in data processing for ML to improve efficiency while perfectly retaining the most significant features.