Research On The Optoelectronic Properties Of Oxide-Based Memristors And The Application Of Synaptic Bionics

In the age of artificial intelligence,where information is booming,the amount of data that must be processed and stored has exploded.However,the large-scale integrated circuits that serve massive intelligent tasks are facing problems of low integration,low transfer speeds,and high-power consumption with the demise of Moore’s Law and the limitations of traditional von Neumann computing system architectures.Inspired by the human brain,neuromorphic computing with superior processing capabilities has emerged as an advanced strategy to address the above dilemma.Moreover,the introduction of a non-volatile electronic device capable of integrating memory and computation,i.e.,memristor,offers an effective solution to this neuromorphic strategy.Memristors are used in highly extensible and low-power artificial synapses due to their similar structure and conductivity modulation mode to biological synapses,which are arranged in an integrated form to enable efficient neuromorphic chips with powerful computational capabilities.It requires memristors to be able to constantly expose excellent sensing and memory functions in response to external signal stimuli.Currently,signals such as electricity,light,gas,and sound have been stretched over memristors to simulate advanced biological functions such as vision,smell,and hearing.However,the relevant bionic capabilities of memristors are not comprehensive enough and there is a lack of multi-signal editable features for efficient applications,which will certainly limit the potential of artificial synapses in real and complex scenarios.To address the above issues,two advanced optoelectronic artificial synapses based on oxide materials have been prepared and their potential for multi-modulation bionic capabilities and multi-functional applications has been verified in this thesis.The main works of this study are as follows:（1）Multi-modulated optoelectronic memristor（MMOM）based on Ga₂O₃/Mo S₂heterojunction for bionic synapses and artificial visual system:A memristor with Ag/Ga₂O₃/Mo S₂/ITO structure was prepared by hydrothermal and magnetron sputtering methods.The performance of the devices as artificial synapses was tested using electrical and optical signals,respectively.Following the construction of the basic properties of the device,the arrayed artificial synapses were assembled to simulate a human visual system with both perception and memory capabilities and check the yield and performance consistency.Further,an artificial heterologous synapse is constructed by introducing electrically modulated synaptic ports into the completed two-terminal visual synapse which fills the missing heterologous signal input in the artificial visual synapse.（2）Amorphous gallium oxide homojunction-based multi-functional optoelectronic synapse（MFOS）for signal processing:Al/Ga₂O₃（OR）/Ga₂O₃（OD）/ITO structured memristors were prepared by magnetron sputtering with different experimental environments.The devices were also tested for their synaptic performance with electrical and optical signals,respectively.The high-pass filtering property of the device for a single wavelength of light was tested and applied to an artificial visual system consisting of a small-scale array.Further,their different response mode under another wavelength of light was measured,and a classic example of associative learning of Pavlovian conditioning reflex experience was realized under co-modulation with dual-band wavelength illumination.Similarly,the logic gate of IMP was finally implemented with the read voltage switches to the positive.