Authors:

Spasko Aleksov

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

This paper presents a method for developing a machine learning model, which will reduce the computation costs for modeling semiconductor nanodevices. Modern semiconductor advancements have resulted in transistors being as small as a few nanometers. As a result, there is an increased need for software capable of simulating new designs. We use Purdue's Nanoelectronics Modeling Tools (NEMO5) software on the RCAC computing clusters to model these devices. The mode space method is used to create approximate band structures for several devices. This method has enabled previously impossible simulations to be run, such as accounting for electron scattering due to phonons and crystal impurities. A large data set of these band structures is created by varying parameters such as device size, device geometry, and material type. The approximate band structures are compared to the true band structures and assessed by creating a metric of fit for the model. The data set will then be used to develop and train the model so the simulations can be done more efficiently. Keywords: Machine Learning; NEGF; Nanoelectronics; Mode Space

Source:

Purdue University / 2024

Topics:

No topics listed

Co-authors:

Spasko Aleksov