Research and Application of Genetic Algorithms in Semiconductor Devices

Abstract

Semiconductor technology underpins a vast array of modern electronic systems, yet continued device scaling and rising design complexity now confront fundamental physical and manufacturing limits. This thesis explores the use of genetic algorithms (GAs)—population-based, gradient-free optimizers inspired by natural selection—to address key challenges in semiconductor development. We first review GA methodologies, detailing their evolutionary operators and workflow steps. Next, we survey GA applications across three domains: (1) calibration of material-property models, where GAs tune machine-learning hyperparameters to accurately predict band gaps in doped metal-oxides; (2) inverse design of novel semiconductor compounds, combining GA search with first-principles-trained surrogates to identify compositions exhibiting target optoelectronic properties; and (3) device and process optimization, demonstrating GA–simulation hybrids that substantially reduce expensive Simulation Program with Integrated Circuit Emphasis (SPICE) and Technology Computer-Aided Design(TCAD) evaluations while discovering superior multilayer device configurations and manufacturing recipes. Representative case studies include multi-objective optimization of concentrating photovoltaic–thermoelectric modules, GA-driven production scheduling, and yield enhancement via feature-selection frameworks. Comparative analyses show that GAs outperform gradient-based and exhaustive search methods in both solution quality and computational efficiency. Finally, we present new GA experiments on benchmark MOSFET and photonic device models, quantifying performance gains over classical techniques. Collectively, our results highlight genetic algorithms as a versatile, scalable toolset for navigating the high-dimensional, multi-objective landscapes of contemporary semiconductor design and manufacturing.