AI-Assisted Design of Drain-Extended FinFET With Stepped Field Plate for Multi-Purpose Applications

Abstract

Fin Field-Effect-Transistor (FinFET) has become fundamental components in advanced integrated circuit, while the drain-extended FinFET (DE-FinFET) features a lightly doped drain extension region to improve the device’s breakdown voltage. However, both structural refinement and the optimal integration of various parameters remain limited in achieving comprehensive optimization of device performance. This study introduces a novel DE-FinFET featuring a stepped field plate to improve overall performance of device. Moreover, within an AI-assisted design framework, predictive modeling and multi-objective optimization of the device are accomplished using Kolmogorov–Arnold Networks (KAN) and the Nondominated Sorting Genetic Algorithm (NSGA-III). More importantly, the proposed framework enables efficient device design and performance evaluation, achieving an average prediction accuracy of 98.19% for electrical performance metrics while being over two million times faster than traditional Technology Computer-Aided-Design (TCAD) simulations. In addition, it effectively generates Pareto-optimal solutions, delivering an average improvement of 9.03% across key electrical performance metrics. The proposed novel device of DE-FinFET offers a new route toward tailoring electrical properties. Meanwhile, the methodology of AI-assisted design not only accelerates device design but also enables customizable solutions for multi-purpose applications.