Use of a polarized light source to enhance machine learning based identification of graphene

The application of machine learning in physics research has been on the rise recently. Combining physics with machine learning has led to creating more efficient systems. We propose to use machine learning combined with polarization dependent optical properties of graphene to enhance the identification of monolayer graphene device. In graphene, the reflection of light depends on the polarization of light incident on it and the thickness of graphene. In this talk, we present our work on using retrofitted optical microscope in conjunction with a polarized light source with rotational degree of freedom for autonomous identification of graphene. The integrated system consists of motor control board, stepper motors, custom designed mounts for motor control and rotational control of the polarizing filters. The system collects reflected light from samples at various polarization angle and then automatically store it to a database for training the machine learning algorithm. The system will dictate the movement and lighting of the microscope based on the sample, find and log the locations of graphene and photograph it. The algorithms currently being implemented are object detection, image processing, and then data classification and storage.