Van der Waals heterostructure-based flexible two-terminal ferroelectric field-effect transistors for in-sensor optoelectronic computing.

The rapid development of artificial intelligence has resulted in the exploitation of bio-inspired visual systems for in-sensor optoelectronic computing. To achieve this, we propose a bioinspired flexible two-terminal optoelectronic synapse enabled by a van der Waals (vdW) ferroelectric field-effect transistor (FeFET) with the metal-ferroelectric-metal-insulator-semiconductor (MFMIS) architecture. The MoS2/h-BN/graphene/CuInP2S6 (CIPS) vdW FeFET exhibits drain voltage (Vds)/light pulses-modulated ferroelectric polarization state and synaptic plasticity. Consequently, the device has the capacity to emulate a range of neural synaptic functions. Based on the flexible two-terminal optoelectronic synapse, we further simulate an optoelectronic neural network based on the two-terminal vdW FeFET synapses array with an image recognition accuracy of >90%. Thus, the demonstrated device characteristics hold the potential for the development of future flexible in-sensor optoelectronic computing applications.