Oxide Optoelectronic Memristor And Synaptic Function Emulation

Brain-inspired computing is considered as the next generation of artificial intelligence,which emulates the neural structure and operation of the human brain at the physical level and therefore can perform advanced computing tasks such as learning,recognition,and cognition,in a fast and energy-efficient way.Memristors are promising candidates for brain-inspired computing due to their simple two-terminal structure,high operation speed and low energy consumption.For traditional electrical memristors,a high voltage or current is generally required to tune the conductance,which always causes ion or atom migration and a large amount of Joule heat thus resulting in microstructure change.It may induce significant cycle-to-cycle and device-to-device performance variations,which is considered a major obstacle to their practical application.In this doctoral dissertation,we mainly focus on improving the stability of memristors.First,a purely electronic memristor based on an oxide homojunction was fabricated with a conductance tuning mechanism of electron trapping and detrapping.Second,an optoelectronic memristor with broadband spectral response was obtained by introducing optical signals into the purely electronic memristor.Then,an all-optical control(AOC)memristor was realized by designing and optimizing the light signal schemes.Finally,human visual functions were emulated with our AOC memristor.The main research contents are described as follows.(1)Purely electronic memristor based on an oxide homojunctionIn this device,a relatively mature amorphous oxide material,In Ga Zn O(a-IGZO),was used as the active layer.Through the analysis of the energy band structure and electrical performance of the device,it was found that the memristive switching behavior originates from the electrons trapping and detrapping at the oxygen vacancy defects.Purely electronic memristor does not involve ion migrantion during its operation.Therefore,it is expected to exhibit excellent endurance performance and time retention.At the same time,the device also shows self-rectifying characteristics with the highest rectification ratio of 10~5,which can help to solve the current crosstalk problem in memristive crossbar arrays.(2)Purely electronic optoelectronic memristorBased on the above purely electronic memristor,an optoelectronic memristor with a broad spectrum(280-1000 nm)response was obtained.This device exhibits significant persisitent photoconductivity under ultraviolet,visible and near-infrared light irradiation,with a retention time>10~4s.Therefore,the optical signal can effectively increase the device conductance.A hybrid control scheme composed of optical and electrical signals was used to realize the reversible regulation of conductance and emulation of synaptic functions.(3)AOC memristor and visual functions emulationBased on the above optoelectronic memristor,an AOC memristor was realized for the first time.The conductance of the device is reversibly tunable over a continuous range by varying only the wavelength of the controlling light.It was found that more than 100 nonvolatile conductance states can be obtained in this AOC memristor.The optical power densities required for operating the device is very low(～20μW/cm~2),which is beneficial to further improving stability and reducing power consumption.Furthermore,spike-timing-dependent synaptic plasticity learning can be mimicked in the AOC memristor.Finally,simple human visual functions were emulated with this device.