A Wafer Yield Fluctuation Analysis Model Utilizing Least-Squares Spectral Analysis

Modern semiconductor production lines are complex high technology systems. In this area the complexity of the products is also reflected in the manufacturing itself. The semiconductor production process is generating layers on a silicon wafer using machines that are utilized in sequences to finish consecutive layers. This process amplifies the impact of even small deviations throughout the sequence of machines, which in turn are subject for seasonal fluctuations, based on their control cycles of machines. These fluctuations are reflected in any measurements taken to observe the production process, including the final yield. The yield is regarded as a base quality measure for manufacturing to evaluate the utilization of production means and reflects the quality of the manufacturing. Many approaches exist to conduct a root-cause analysis on manufacturing data, yet applied solutions to identify and analyze fluctuations are still scarce. This work proposes a fluctuation analysis pipeline and highlights how to assign the fluctuations to potential root-causes. The approach is successfully applied in practice and first results are shown.