Molecular Cocrystal Strategy for Retinamorphic Vision with UV-Vis-NIR Perception and Fast Recognition.

Abstract

Neuromorphic vision sensors capable of multispectral perception and efficient recognition are highly desirable for bioretina emulation, but their realization is challenging. Here, we present a cocrystal strategy for preparing an organic nanowire retinamorphic vision sensor with UV-vis-NIR perception and fast recognition. By leveraging molecular-scale donor-acceptor interpenetration and charge-transfer interfaces, the cocrystal nanowire device exhibits ultrawide photoperception ranging from 350 to 1050 nm, fast photoresponse of 150 ms, high specific detectivity of 8.2 × 10¹² Jones, and responsivity of 15 A W^-1, as well as retina-like photosynaptic plasticity behaviors. Utilizing the sensor nerve and convolutional neural network, the architecture achieves 90% accuracy in recognizing colorful images. The cocrystal design offers an effective method for constructing nanowire photosynases with high performance in artificial visual systems.