Font Size: a A A

Emulation Of Synaptic Clustered Plasticity Based On Polymer Memristive Devices

Posted on:2022-03-17Degree:MasterType:Thesis
Country:ChinaCandidate:J L WuFull Text:PDF
GTID:2568306626998739Subject:Materials Science and Engineering
Abstract/Summary:PDF Full Text Request
In this age of information,the explosive increase of data poses challenge to traditional computer systems based on von Neumann architecture.Neuromorphic systems,taking the intelligent information processing system that emulates the human brain as the starting point,are expected to realize parallel computing for complex data with low power consumption and high efficiency.The basic hardware unit of the neuromorphic system is analogous to the memory unit in the human brain—synapses.The plasticity and cooperative behavior of neuronal synapses is crucial for the implementation of the brain’s learning and memory capabilities.Memristive device,as the most typical synaptic biomimicking device,has been studied extensively on its working mechanism,device structure and materials development,along with numerous research achievements on the emulation of synaptic plasticity.Although the development of inorganic memristive devices is relatively mature,organic memristive devices have superiority in the field of neurosynaptic bionics,considering the advantages of high biocompatibility,good malleability,and low production cost,especially the high consistency between the attribute of some polymer that can transfer both ions and electrons and the working mechanism of synaptic information transmission in our brain.However,present emulations on synaptic behavior are mainly focused on the function of a single synapse,and little attention is paid to the cooperative behavior between synapses.The capability of human brain to process information parallelly and efficiently highly depends on the coordination mechanism between synapses.Therefore,the objectives of this paper are to design polymer memristive device arrays,develop a prototype model for the interactions between devices,and implement device emulations on synaptic interaction with the sandwich-structured polymer memristive device arrays.The specific contents are as follows:(1)Research on the electrical behavior of polymer memristive devices and the statistical behavior of the devices in the array:the polymer memristive device array is prepared with a solution-processed method and two types of lithium salts are employed to dope the semiconductor layer.Both inorganic and organic lithium salt-doped polymer memristive devices show good memristive performance and synaptic plasticity simulation potential,verifying the universality of the device structure and working mechanism.The regular hexagonal top electrode was selected for synaptic biomimetic research after the comparison with other electrode patterns.The small device to device fluctuation in the array provides a reliable basis for the subsequent research on the interaction between devices and the simulation of synaptic clustered plasticity.(2)Research on the electrical behavior and the mechanism of the interaction between memristive devices:The mutual influence of the electrical behavior between different devices is embodied in the hysteresis phenomenon and reading current.When the device current increased after applying a positive bias,the hysteresis window of adjacent devices disappeared under positive voltage sweep and the device resistance also decreased.By means of establishing an equivalent circuit of horizontal and vertical resistors in series and parallel,a physical model is proposed that the migration of ions in the horizontal and vertical direction regulates the distribution of hole concentration and brings significant increment in the horizontal conductivity of the doped semiconductor film.(3)Simulation of synaptic clustered plasticity based on the interaction between multiple memristive devices:The emulation of synaptic interaction based on the ion migration mechanism is implemented in sandwich-structure polymer memristive devices for the first time.Systematic simulations of synaptic clustered plasticity are carried out by employing the polymer memristive device array,including the interaction between adjacent synapses and non-adjacent synapses within a certain distance.The increase/decrease of the current in the programmed device induces the increased/decreased reading current value of adjacent devices in the array,and the amplitude of variation exhibits an inverse relationship with the spacing distance,in accord with the excitatory and inhibitory behaviors of synaptic clustered plasticity.In the case of the programmed device located in the center of an annulus,the same amplitude of variation in reading current has been found in the surrounding memristive devices with identical spacing distance and it also applies to the memristive devices in the second ring area,thereby the spatial symmetry of synaptic clustered plasticity is simulated.For synaptic biomimicking,the migration of ions among different devices resembles the diffusion of neurotransmitters in synaptic clusters,which provides a new approach for the realization of fully functional neuromorphic networks.In summary,based on the polymer memristive device array,this thesis studies the interaction between memristive devices and realize the simulation of synaptic interaction in sandwich-structure polymer memristive devices for the first time,which provides a novel route and valuable experimental information for the design of neuromorphic networks.
Keywords/Search Tags:polymer memristive devices, device interaction, synaptic interaction, synaptic clustered plasticity
PDF Full Text Request
Related items