Dissertation Defense

Next Generation Graphene Photonics Enabled by Ultrafast Light-Matter Interactions and Machine Learning

Dehui Zhang

Featuring an atomically-thin structure, graphene was first experimentally studied in 2004. Since then, many works reported unique properties of graphene and other 2D materials. However, additional efforts are necessary to bridge these findings in physics with successful industrial applications. In this talk, I will present our efforts on exploiting the ultrafast light-matter interactions in graphene to meet the growing demands in THz light source, IR sensing, and 3D detection. Our works suggest possibilities in the synergic design of nanophotonic devices with machine learning algorithms.

I will start with our graphene heterostructure THz emitter with an 80-time larger emission amplitude than typical photoconductive switches. The enhancement originates from ultrafast hot-carrier transport in the 2D-3D heterojunction. Next, we show a phototransistor design that is electrically tunable in a spectrally-resolved manner. A computational method can extract spectral information of incident light from a single pixel. Further modeling proves the system’s potential as an ultra-compact on-chip spectrometer for hyperspectral imaging. Lastly, I will discuss how the transparent graphene detectors enable a 3D imaging system. Synergic design of the device, the optical architecture, and the machine learning algorithm enables 3D ranging and tracking of a point source. We further explore an extension of the system into a more complex, high-performance 3D camera.

I will conclude the talk with a generalized discussion on the question: Why and how machine learning and artificial intelligence will boost the development of nanophotonic applications?

Chair: Professor Zhaohui Zhong