Semiconductor device design using the BiMADS algorithm

Designing high-performance semiconductor devices is a complex optimization problem, which is characterized by multiple and, often, conflicting objectives. In this research work, we introduce a multi-objective optimization design approach based on the Bi-Objective Mesh Adaptive Direct Search (BiMADS) algorithm. First, we assess the performance of the algorithm on the design of a n+-n-n+n+-n-n+ silicon diode using a standard drift–diffusion model, showing that BiMADS is able to find the best solutions and to outperform the state-of-the-art algorithms. Successively, we tackle the design of MESFET and MOSFET devices, using a Maximum Entropy Principle (MEP) model; BiMADS is able to locate new designs that minimize the size of the device and provide an increased output current. Moreover, it is proved that BiMADS is able to locate promising solutions with a tight budget of objective function evaluations, which makes it suitable for large-scale industrial applications.