The papers in this Special Issue address these challenges by balancing physical fidelity with computational efficiency. They deepen our understanding of device physics while providing models that are both insightful and practical, with applications spanning cryogenic electronics, wide-bandgap devices, and radiation-hardened systems.
Su et al. present a charge-based analytical model for bulk MOSFETs, that is, valid down to 10 mK. Their work clarifies the interface-trap-dominated mechanisms that lead to threshold voltage divergence between NMOS and PMOS devices and quantifies significant analog parameter enhancements, including a 73% increase in PMOS cutoff frequency at 4 K. These findings are essential for quantum-control electronics. Complementing this, Mao et al. provide a comprehensive review of four physics-based compact models for GaN HEMTs, namely MVSG, ASM HEMT, EPFL, and QPZD. They analyze how each model addresses challenges such as trapping effects, self-heating, and process variability, and highlight emerging opportunities for combining physical models with machine learning to accelerate parameter extraction and quantify uncertainties. In the area of radiation-tolerant electronics, Xu et al. introduce a machine-learning approach using an ant-colony-optimized neural network. By adaptively sampling critical waveform regions, their method achieves an RMS error of only 0.82% in predicting single-event transient currents, surpassing the fidelity limits of traditional double-exponential pulse models and enabling high-precision radiation effect simulation for aerospace applications. Meanwhile, Deng et al. demonstrate a practical strategy for AI-assisted SPICE integration. They employ geometry-parameterized scaling laws for spiral inductors and machine-augmented Power MOS trans-conductance models to accelerate parameter extraction by an order of magnitude while preserving full SPICE compatibility. This approach significantly streamlines industrial design workflows.
Collectively, these contributions point to a trend toward physics-informed, data-driven co-design methodologies. By combining rigorous physical insight with computationally efficient, machine learning–aware workflows, they enable robust optimization of devices and circuits across a wide range of applications, from quantum interfaces to aerospace systems.
Future research should prioritize the development of standardized interfaces between AI tools and physical models, the extension of models to three-dimensional integrated wide-bandgap architectures, and the establishment of co-design frameworks for emerging ultra wide–bandgap materials capable of operating in environments ranging from near-zero Kelvin to orbital radiation conditions. We sincerely thank all authors for their outstanding contributions, which have advanced the frontier of semiconductor modeling science.
You received this message because you are subscribed to the Google Groups "mos-ak" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mos-ak+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/mos-ak/CALp-Rj9nAVJk0aGWTZwufdoECgKNXaV7Bq4j3FQ0531OdYNjog%40mail.gmail.com.