ANN Based Cryogenic DC Modeling of a 16 nm nMOS FinFET

An artificial neural network (ANN) based cryogenic DC I/V model for a 16-nm nMOS FinFET is presented and validated over the temperature range from 2.95 K to 300 K. The model is trained on measured drain-current data and compared against an analytical empirical model to evaluate accuracy, generalization capability, and computational efficiency. Several multilayer perceptron architectures are investigated, focusing on interpolation and extrapolation performance beyond the training bias space. The results demonstrate that the proposed ANN models accurately reproduce the drain current behavior and achieve excellent performance. When implemented in a circuit-level simulator, compact ANN architectures exhibit simulation times comparable to analytical models. Overall, the proposed approach represents a viable and efficient alternative for cryogenic FinFET modeling.