Bilayered Oxide‐Based Cognitive Memristor with Brain‐Inspired Learning Activities

Abstract

our brain would solve this dilemma, that is, the von Neumann bottleneck. Thus, neuromorphic engineering comes into being. Basically, data-driven neuromorphic engineering needs to complete a large number of data processing tasks. Presently, one of the main tasks in the state-ofthe-art neuromorphic computing system is to optimize neuromorphic algorithm. Due to the use of von Neumann architecture, such neuromorphic systems always consume extremely high energy. In fact, our brain nervous system is consisted of ≈10 11 neurons connected by ≈10 15 synapses. [3] Neurons and synapses are basic units in brain information processing. Brain cognitive behaviors occur through synaptic responses and neural functions. And the synaptic plasticity is the most important characteristics of synapse. Such architecture makes our brain operate in extremely high energy efficiency with certain fault tolerance to respond to our surroundings. Thus, brain-inspired neuromorphic devices have been proposed for mimicking biological synaptic responses and neural functions. [4][5][6] It is getting a new branch for neuromorphic engineering.Recently, with the developments of microelectronics, optoelectronics, and material technologies, solid-state neuromorphic devices have been proposed to mimic biological synaptic functions. Such devices include two-terminal resistance change memory devices [7][8][9][10] and field-effect transistor based neuromorphic transistors. [11][12][13][14][15] Due to the simple sandwich structure, two-terminal memristors have the inherent priority in 3D integrity. They have been widely investigated for high density storage applications. [16,17] With nonlinear electrical characteristics and nonvolatile resistance modulation effects, two-terminal memristors are very suitable for neuromorphic device applications. Such devices include ferroelectric random access memory (FeRAM), phase change random access memory (PCRAM), resistive random access memory (RRAM), etc. [18][19][20][21][22] In fact, most of the recent reported neuromorphic devices are based on two-terminal memristors. Especially with the deep understanding of the operation mechanisms of memristors, neural operation and typical Modified National Institute of Standards and Technology (MNIST) pattern recognition have also been demonstrated by using memristor arrays. [23][24][25] In addition, energy consumption is another important index in neuromorphic device applications. Recently, several works have been reported on neuromorphic device with low energy consumption. [26][27][28][29][30] Xu et al. [26] obtained organic nanowire synaptic transistors with energy consumption of ≈1.23 fJ per Recently, neuromorphic devices have attracted great attention due to their potential to overcome the von Neumann bottleneck. Due to their nonlinear electrical characteristics and nonvolatile resistance, memristors have been proposed for use in neuromorphic device applications. Bilayered HfO 2 /TiO x -based cognitive memristors are proposed. They demonstrate conductance-modulation capabilit...

Engineering Method for Tailoring Electrical Characteristics in TiN/TiOx/HfOx/Au Bi-Layer Oxide Memristive Devices

“…Those memristive devices showed a significant improvement in the resistance modulation linearity (Li et al, 2018a) and the number of resistance states (Stathopoulos et al, 2017) along with the reduced variability in the resistive switching characteristics (Wang et al, 2010). The bi-layer metal oxide devices typically consist of an oxide layer that serves as a reservoir of oxygen vacancies and a solid state electrolyte layer which builds a Schottky-like interface contact with the adjacent metallic electrode (Huang et al, 2012;Bousoulas et al, 2016;Kim et al, 2018;Xiong et al, 2019). The resistive switching mechanism can be described as follows (Cüppers et al, 2019): under an external bias voltage oxygen vacancies are injected from the reservoir layer into the solid state electrolyte layer in which the oxygen vacancies are forming a filamentary conduction path toward the metallic electrode.…”

Emulating Artificial Synaptic Plasticity Characteristics from SiO2-Based Conductive Bridge Memories with Pt Nanoparticles

“…In addition, the higher the triggering amplitude the smaller pulsing sequence is required in order to impose the targeted conductance state. The above effects can be viewed as direct evidence during the implementation of bio-synaptic properties [ 37 ].…”

“…This behavior is closely associated with the steep transitions during that the hysteresis pattern presents and it is detrimental in terms of efficiency of the backpropagation algorithm at the training stage of deep neural networks (DNNs) [ 47 ]. Several optimization procedures have been proposed toward obtaining a more linear response, such as the incorporation of bilayer structures [ 37 ], but at the expense of the capability of emulating the STP to LTP transition [ 48 , 49 ].…”

Optically Modulated Threshold Switching in Core–Shell Quantum Dot Based Memristive Device

“…[ 14–17 ] The ions motion in memristors naturally emulates the biological functions of synapses while the programmable conductance can represent the synaptic weight. [ 18–28 ] Crossbar array constructed by memristor units can realize the hardware neuromorphic network with in‐memory computing capability and high parallelism. [ 29,30 ] In future, there are three primary application fields of memristive technology, including on‐chip memory and storage, in‐memory computing, and biologically inspired computing, as suggested by Lu and co‐workers.…”