AbstractA novel approach is demonstrated to identify glucose concentration in aqueous solutions based on the combined effect of its frequency‐dependent interaction with microwaves propagating in graphene channels and the modification of graphene radio frequency (RF) conductivity caused by physisorbed molecules. This approach combines broadband microwave sensing and chemical field effect transistor sensing in a single device, leading to information‐rich, multidimensional datasets in the form of scattering parameters. A sensitivity of 7.30 dB(mg/L)−1 is achieved, significantly higher than metallic state‐of‐the‐art RF sensors. Different machine learning methods are applied to the raw, multidimensional datasets to infer concentrations of the analyte, without the need for parasitic effect removals via de‐embedding or circuit modeling, and a classification accuracy of 100% is achieved for aqueous glucose solutions with a concentration variation of 0.09 mgL−1.
Read full abstract