Physical Reservoir\nComputing Using Tellurium-Based\nGate-Tunable Artificial Photonic Synapses

We report tellurium (Te) thin-film-based artificial photonic\nsynapses\nand their application to physical reservoir computing (PRC). The Te-based\nartificial photonic synapses were fabricated by using sputtered Te\nthin films and spray-coated MXene (Ti<sub>3</sub>C<sub>2</sub>) electrodes.\nA thorough investigation of the field-dependent persistent photoconductivity\n(PPC) of the Te channel revealed that the relaxation speed of the\ntransient photocurrent depended on the gate bias. Utilizing the PPC\nproperty, the Te device served as an excellent photonic synapse under\nlight pulse stimulus, exhibiting multiple synaptic characteristics\nsuch as excitatory postsynaptic current and paired-pulse facilitation,\nas well as highly linear potentiation-depression characteristics;\na simulation-based study further confirmed the effectiveness of the\ndevice. Most importantly, by exploiting the nonlinear and fading memory\ncharacteristics of the Te photonic synapse, we demonstrate two advanced\nexamples of PRC. In classifying handwritten digits, our system carried\nout successful digit recognition without binarization or another simplification\nprocess with reduced computational cost compared to conventional systems.\nTo solve second-order nonlinear equations, we introduce the strategy\nof utilizing historical nodes. The combination of historical nodes\nand the gate-tunable responses of the photonic synapses, which provide\nan enriched reservoir state, yielded excellent prediction accuracy.\nOverall, this work will offer an understanding of Te-based optoelectronic\ndevices and their synergetic integration with neuromorphic devices\nand PRC.