CDS&E: Machine-Learning-Driven Methods for Multiobjective and Inverse Design of van-der-Waals-Material-Based Devices

University Of Florida, Gainesville FL

Investigators

Abstract

Advanced semiconductor device technologies play a critical role in computing, data storage, artificial intelligence (AI), and quantum technologies. Two-dimensional materials and their heterojunctions form a promising material and structure platform to develop future semiconductor devices. Despite promising technological potentials, their technological adoption has been hindered by several major challenges including predicting device properties accurately and assessing device performance systematically. In this project, computer-aided simulation and design capabilities will be developed to address these challenges by combining AI methods with advanced device simulation. The development of AI-guided computer simulation methods will facilitate fast and accurate prediction of device properties, and enable automatic and efficient design of these nanoscale devices. This project will result in an integrated testbed for research and education on applications of AI methods in nanoelectronics. The project will engage and train high school, undergraduate, and graduate students in the fields of semiconductor and AI technologies. The modeling tools developed in this project will be disseminated as an open-source online resource. The goals of the project are to develop physics-informed machine-learning (ML) models and device design methods for efficient and automatic simulation and design of van der Waals (wdW) semiconductor devices. The proposed research activities include: (1) develop physics-informed ML models in an embedded or hybrid manner for quantum transport device simulations, with consideration of improving efficiency, respecting device physics, and reducing the amount of training data required; (2) develop a multi-objective optimization method to systematically assess and comprehensively optimize vdW-material and vdW-heterojunction devices, by simultaneously considering multiple technologically important device performance metrics; (3) develop an efficient gradient-based inverse design method that is integrated with quantum transport device simulations, by applying auto-differentiation methods over the quantum transport equation and its solution algorithms; (4) test and apply the methods proposed to simulate and design a set of vdW-material and vdW-heterojunction devices, which have shown promising logic and memory device performance. The project develops an essential knowledge base for harnessing rapidly developing machine learning methods and theory to advance the modeling, simulation, and design capabilities of nanoelectronic devices. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.