[NanoIC topic] Physics-inspired ML framework for Compact Model Calibration

Device compact models (CMs) are critical for linking semiconductor device physics with circuit-level design, providing simplified yet accurate representations of complex device behaviour for efficient simulation. As technology scales into nanometer dimensions and introduces new materials and architectures, these models must capture non-ideal effects such as short-channel phenomena, variability, and temperature dependence. This capability enables designers to predict performance, power, and reliability early in the design cycle, reducing development time and cost. However, calibrating CMs at advanced nodes is challenging due to intricate device physics like quantum confinement, strong parasitic interactions, and parameter interdependencies. Variability from process imperfections and the complexity of emerging architectures such as FinFETs and Gate-All-Around FETs further complicate calibration, requiring sophisticated techniques and large datasets while balancing accuracy and computational efficiency.