MTPP-based Memristors For Neuromorphic Application

Human brain is a highly parallel,energy efficient,and event-driven information processing system,which is based upon high levels of computation even challenging the largest supercomputer.Inspired by the efficiency of the brain,neuromorphic devices—memristors have recently generated significant interest as new paradigms toward the realization of novel data storage and neuromorphic computing applications,which is one of most promising approaches to overcome the “von Neumann bottleneck” in the post-Moore's law era.Indeed,organic materials distinguish themselves with lowcost and easy-fabrication process,mechanical flexibility,biocompatibility,and more importantly,tunable electronic properties via molecular design strategy.Many attempts so far have been devoted to yield organic memristor devices with excellent performance.Unfortunately,the operation of these organic memristors usually relies either on the slow kinetics of ion diffusion to retain their states or on charge storage in metal nanoparticles characterized by low charge carrier mobility and large charge injection barriers,making them unattractive for high performance memristor applications since the high operational voltage and fluctuant operation parameters.These limitations reduce the accuracy,stability and scalability of organic memristors and pose challenges for these devices to approach the interconnectivity,information density,and energy efficiency of the brain.Therefore,it is highly deserved to fabricate satisfactory organic memristors with adequate electrical characteristics to fully realize the efficiency and performance gains of a neural architecture.This thesis detailedly reviews the recent progress of nanoionics-enabled memristors.On this basis,the preliminary devices ITO/MTPP/Al2O3-x/Al are fabricated to demonstrate the novel coordination-assisted mixed ions and electrons transport,as well as their applications of versatile emulations;specifically,the novel SVDP with the applications of noise filtering are detailed explained.(1)The numerous publications in the literature are critically reviewed,and provided fundamental insights into the strategies for regulating analog switching based on the nanoionics-enabled devices and the correlation among materials.To begin with a critical overview of the proposed nanoionic mechanisms for memristive switching was given,focusing particularly on providing fundamental insights into the strategies for regulating the adaptive memristive characteristics based on the modulation of the effective Schottky barrier and carrier distribution.Subsequently,various active materials including oxides,nanomaterials,polymers,small molecules,bio-macromolecules,dopants and so forth that investigated in recent research efforts were examined for developing high performance devices.(2)Inspired by the deliberate oxygen transport carried by the ferriporphyrin-containing hemoglobin(Hb)in our blood circulation,a novel coordination-assisted mixed electronic/ionic conductor(MEIC)platform ITO/MTPP/Al2O3-x/Al is investigated to delicately regulate the ionic migration for robust molecular multimedia memristors.The microscopic nature of both the oxygen ions migration and the coordination binding is revealed for the first time by in situ STEM-EDX and XPS analysis.It is found that oxygen ions can bind and dissociate with ZnTPP molecules reversibly,which can be regarded as continuous hopping processes.In comparison with the uncontrollable random migration pathway(e.g.,grain boundary and dislocation)in oxide devices,this novel coordination-modulated hopping process can confine the oxygen ions successive translocation along ZnTPP molecules under electrical fields.The resultant redistribution of the oxygen ions will subtly affect the essential interfacial parameters,triggering smooth and tunable memristive swithching.Moreover,except the resultant excitatory trends of intense stimuli,the overall inhibitory trends gradually increase under stronger weak stimuli(i.e.,SVDP),which can be ascribed to the larger multiple polarization processes.(3)The coordination bonding effect on the ionic transport is further investigated by utilizing different MTPP mediums including ZnTPP,Ni TPP,CoTPP,FeTPPCl and TPP.A series of memristors are fabricated with identical configurations as well as uniform thickness of deposited MTPP layers to examine the device performance.The metal-oxygen bond length and EMetal-O are calculated through the method of density functional calculations.It is found that the metal-oxygen bond length gradually prolongs and the EMetal-O gradually attenuates for ZnTPP,NiTPP and CoTPP molecules,corresponding to the increasing overall conductivity and declining anti-saturation.The bonded chlorine in the FeTPPCl molecule leads to dismal oxygen permeability liable for the poor properties.Moreover,owing to the absence of coordination regulation,the oxygen ions are transported crudely through the interspace of TPP molecules at a high speed as indicated by the resultant higher conductive level and effortless saturation.Notably,the memristor behaviors are highly sensitive to different MTPP materials(not limited to the discussed MTPP materials),which can be regarded as metal-dependent plasticity,implying that it is feasible to fabricate versatile neuromorphic circuits using MTPP memristors as programmable building blocks according to demand.(4)The above experimental results unambiguously demonstrate that the ZnTPP memristor features impressive environmental stability,smooth memristive behaviors and low energy consumption potential,as well as a wide range and high density of non-volatile intermediate states available for computation are highly adapted to perform multiple synaptic functions emulation by applying specific programming stimuli.In this context,various scanning modes or stimulation pulse sequences have been explored to emulate multiple synaptic functions,such as STP/LTP(STM/LTM),SRDP,‘learning-experience' behaviors,habituation and sensitization.In particular,the capability of SVDP directly at the device level enable the implement of noise filtering just rely on memristor devices,which seems to be a very promising idea that can lead to the design of efficient visual recognition system,and extends the fresh functions successfully for the neuromorphic hardware.