Flexible Photo-synaptic Devices Based On Negative Photoconductance And Relaxation In Graphene

Driven by the development of science and technology,a new era of the internet of everything comes.A large amount of information is interacted between humans and machines.The contemporary computing system relies on the von Neumann structure,and its memory unit and calculating unit are separated from each other.This restricts the intelligent development centered on the huge amount of information.The solution method of only optimizing the algorithm is not enough to break the von Neumann bottleneck.Therefore,researchers draw inspiration from the fast and parallel information processing mode of the brain nervous system,and propose a new computing system with an artificial neural network structure that integrates storage and computing.By preparing a single electronic device with synaptic function,the construction of artificial neural network is realized.With the in-depth study of semiconductor devices by researchers,the preparation process of the device has become increasingly mature,and its functions have become more diversified.It has been applied in various fields and has become a potential object for building artificial synapses.Semiconductor optoelectronic devices based on the photoelectric effect can realize the mutual conversion of optical signals and electrical signals.It has advantages in constructing artificial synapses that directly obtain external light information.Under the photoexcitation of the device relying on the photoelectric effect,the separation of electron-hole pairs in the semiconductor increases the carrier concentration and the conduction current,thereby realizing the photoelectric signal conversion of the device.Therefore,most of the devices constructed are optoelectronic synaptic devices that regulate excitability,but lack the inhibitory synaptic plasticity regulated by light.This paper proposes a flexible optoelectronic device with negative photoconductivity constructed by a heterointerface of highly conductive graphene and flexible polyimide.The photogenerated electrons can be transferred at the heterointerface to realize the negative photoconductivity effect.In addition,the device adopts chemical vapor growth（CVD）single-layer graphene,which can obtain the phenomenon of electron-hole recombination relaxation because it has a small amount of defects.From this,the inhibitory plasticity of synapses can be realized,and various basic functions of synapses can be simulated under the control of light,which is applied to the accurate recognition of handwritten digits.The research content of this paper mainly includes two parts:1.A simple preparation process of fixed-point transfer and prefabricated electrodes was designed.Then,a flexible optoelectronic device with negative photoconductive effect was prepared with polyimide as the substrate and CVD monolayer graphene as the channel material,and the performance and mechanism were explored.Using the strong UV absorption ability of polyimide,combined with CVD single-layer graphene to form a heterogeneous interface,it was proved by Raman and PL spectroscopy,as well as the UV light experiment of mechanically exfoliated graphene.Under the stimulation of ultraviolet light,the photogenerated electrons inside the polyimide enter the graphene through the interface and recombine with the holes therein.As a result,the channel conductance is reduced,resulting in a negative photoconductivity phenomenon.In addition,CVD graphene itself has defects,so relaxation occurs when electrons are recovered after irradiation.The 365 nm band light source was used to control and test under different source-drain voltage,optical power,and pulsed light conditions.When the source-drain voltage increases from 0.05 V to1 V,the corresponding I_on/I_off increases from 9.59%to 15.64%,and the current recovery time decreases accordingly.As the light intensity increases,the change value of the photocurrent increases,and the current recovery time increases.Pulse width and number of pulses are positively related to current change and current recovery time,while pulse interval is inversely related.2.Based on flexible PI-graphene optoelectronic devices,artificial synapses and their arrays were constructed to realize the basic function of synaptic plasticity.It mainly realizes short/long-term plasticity,double-pulse facilitation（PPF）,and spike rate-dependent plasticity（SRDP）.And the ability of synaptic devices to transition from short-term plasticity to long-term plasticity is obtained through the modulation of pulsed light parameters.In addition,it is proved that the flexible PI-graphene synapse device has good flexibility,and the device can work stably under different degrees of deformation measured under the bending strain of 0 to 0.024.Then,the LTP/LTD curve of the synaptic device was obtained by triboelectric nanogenerator and UV light modulation,and its linearity and symmetry were analyzed.The corresponding linearity was 1.9,2.3,and the symmetry was 1.28,indicating that LTP/LTD.The LTD curve has good linearity and symmetry.Therefore,when it is applied to image recognition,the data set from MNIST is used for learning and training without introducing a hidden layer.When the training period is less than 2100,the recognition accuracy of the handwritten digit"8"image reaches 84%.It provides more possibilities for building flexible artificial neural network systems.