Fig. 11: Photonic nose with in-sensor computing.

a Schematics of the Photonic nose with an in-sensor computing unit. MIR light is coupled into the multiple waveguide channels, and each is encoded with the sensory data signal. The optical signal is then picked up by the graphene photodetector, whose weight is set by adjusting the bias voltage. The multiplied signals are then summed and sent to electrical postprocessing for classification tasks. b Optical images of raphene photodetector. c Measured Spectra of gas mixtures. d The confusion map of the gas mixture group classification. Reproduced with permission from the American Chemical Society (2024)²³⁸