Université de Montréal

Published purpose

This project proposes to develop the capability to simulate the vibrational properties of 100k atom size systems using deep neural networks (DNN) methods. Such simulations will advance the ability to directly compute what is currently measured with novel Raman techniques. In this project, theoretical and experimental groups will collaborate to bring simulation tools to the next level, where generated data can be directly compared to experimental Raman spectra of complex matter. This testing platform will consist of Raman imaging of defective graphene sheets, systems that cannot be addressed with standard numerical methods because of the size of the problem. The goal is to develop AI-assisted tools to understand better the type of defects that cause a certain Raman response. These tools will use density-functional theory (DFT) simulations as input to guarantee the accuracy and predictability of the approach. These tools will help in the analysis of 2D materials. One possible use of 2D materials is in gas detection and having a better understanding of Raman signals on these surfaces before and after being exposed to certain gases will provide a guide on which materials have the better potential for this type of application.