Roll-to-roll gravure printed large-area flexible carbon nanotube synaptic photogating transistor arrays for image recognitions

Abstract

The development of novel large-area flexible optoelectronic synaptic devices is of great interest for visual neuromorphic computing. In this paper, we have successfully manufactured 33 × 34 flexible carbon nanotube synaptic photogating transistor arrays via the roll-to-roll gravure printing technology using the mixture of photosensitive rhodamine 6G and poly(vinylidene fluoride-co-hexa-fluoropropylene) as the composite dielectric layer. The resulting thin-film transistor devices exhibit excellent electrical properties with high Ion/Ioff (>105), the maximum carrier mobility (10.71 cm2V−1s−1) and low operating voltage (−1.5 V∼ 0.5 V), and excellent optoelectronic synaptic properties, which are capable of performing not only the excitatory postsynaptic current and paired pulse facilitation but also complex biological synaptic functions such as optical writing and electrical erasure, dynamic learning and forgetting processes, and Pavlov's dog experiment. Significantly, the ultralow energy consumption per light spike event as low as 0.03 fJ can be achieved under the pulse UV light illumination. Moreover, according to the trend that the LTP/LTD curve conforms to the concave function, we successfully embed the specific update rules of our synaptic devices into the generative adversarial network (GAN) that can generate high-quality (7680 × 4096 pixels) images, which is impossible for the human eye to distinguish the difference between the original labels and real labels.