Machine Learning Assisted Multi-Functional Graphene-Based Harmonic Sensors

Real-time monitoring of multiple (bio)chemical agents, molecules and gases is in a high demand, particularly for the future internet-of-things (IoTs) and point-of-care tests (POCT), that are connected via the 5G ecosystem. Here, we propose a lightweight, multi-agent (bio)-chemical wireless sensor based on graphene field-effect transistor (GFET) circuits, taking advantage of GFET's dual functionalities, i.e., frequency modulation and (bio-)chemical sensing. The GFET-based radio-frequency (RF) modulators circuits can convert the continuous wave (CW) monotonic signal to multiple harmonics, with conversion efficiencies sensitively depending on densities of (bio-)chemical agents. Specifically, we exploit a machine learning (ML)-based readout method to extract the concentration levels of the (bio-)chemical dopants from the harmonic spectrum. Further, we show that by increasing the order of GFET circuits and thus the number of detectable harmonics, the neural network performance and the overall readout accuracy can be enhanced. The proposed GFET-based wireless sensor could be ultracompact, ultralow-profile, portable and flexible, thus potentially benefiting a wide range of applications in IoTs, POCTs, and Industry 4.0.