Data as Waves: A Relational Approach to Semiconductor Process Analysis Using Relative Phase

This study investigates a relational representation of semiconductor process data based on relative phase and its progression over time. Rather than treating process data only as collections of individual sensor measurements, the study examines whether relationships formed among sensors can provide additional information about process state and process outcomes. Using data from a BOSCH plasma-etching process, a recurrent temporal structure was first identified directly from the sensor data and subsequently compared with the process recipe. Relative phase was then used to represent relationships between sensor signals within this recurrent structure. The relationship between Platen RF Load Power and Pressure was examined together with its cycle-wise relational progression and compared with wafer-level Si etch results. The results show that the overall relational state and its progression contain complementary information. This supports extending process-data representation from individual sensor values toward relational states and the paths through which those states evolve over time. This Zenodo record includes the manuscript and the public reproducibility code used to reproduce the principal analyses reported in the study, including recurrent-structure discovery, relative-phase analysis, relational progression, and out-of-fold predictive evaluation.