Artificial Neural Network Extraction of Complex Conductivity of Thin Graphene Layers Using Terahertz Time-Domain Spectrometry

Terahertz time-domain spectroscopy (THz-TDS) has shown significant potential for characterising the electrical properties of 2D materials, including graphene, in a non-invasive manner. However, extracting material parameters is analytically complicated. Furthermore, it requires fitting a transfer function for a physical model of conductivity such as a Drude model, which in many cases does not accurately represent real world samples. Here we present a neural network trained using simulated datasets that's capable of extracting the complex conductivity of thin graphene layers from experimentally acquired data. Our end goal is to create a neural network, trained on multiple theoretical models and experimental measurements, capable of extracting electronic parameters directly from the time domain and able to classify which conductivity model represents the sample.