Machine learning based band gap screening for ultrathin materials

Abstract

The basics of Application of Machine Learning in Density functional Theory (DFT) to help us in electronic structure calculations for ultrathin 2D materials, and the quest of new ultrathin 2D materials have become a new field of research. To explain how we are approaching and progressing in this project and what will be the motivation, we will start by drawing an analogy. Let us suppose that, this DFT is my car, and we need to learn how to drive a car. So, to start with this, I will first get myself acquainted with the working principle of the car. For instance, the role of engine, gear box, shaft, handbrakes inside my car. Until we do not know these stuffs, we would not be able to understand why we drive the car in lower gear on steep hills and on higher gear on a flat road. So, before we start doing electronic calculations for different materials, we must know the functioning and the physics phenomena working behind DFT. So, the first phase of our project was to understand the working principle of DFT. The second phase is to drive the car, under the supervision of an experienced and learned person who knows driving. And to indentify what modifications are needed to be done in my car to optimize it. In this case, right now we are studying the things which can modify my car aka. DFT and getting used to the tools required to add those things in my car, such as programming/coding in different languages (for example: c++ and Python). That thing is called Machine learning. The third phase is to make use of my car for different purposes. For instance, going to the office, or grocery store, or hospital. In our case, we will use DFT for calculating electronic structures of ultrathin 2D materials. And then we will think of a process to apply Machine Learning (ML) in my car (DFT) to speed up our car so that we can reach our desired location (where we use to go very often) quickly. ML methodologies have capability to speed up our electronic structure calculation. we will also think of a way to make use of ML methodologies to predict new ultrathin 2D materials as well. We have made a Machine Learning model, which takes the different energy band gaps of corresponding 2D materials, and have successfully screened them on the basis of band gap between 1.2 eV to 1.3 eV.