Spin-transfer torque magnetic memory as a stochastic memristive synapse for neuromorphic systems

Quantized Magnetic Domain Wall Synapse for Efficient Deep Neural Networks

The quantization of synaptic weights using emerging nonvolatile memory (NVM) devices has emerged as a promising solution to implement computationally efficient neural networks on resource constrained hardware. However, the practical implementation of such synaptic weights is hampered by the imperfect memory characteristics, specifically the availability of limited number of quantized states and the presence of large intrinsic device variation and stochasticity involved in writing the synaptic states. This article presents ON-chip training and inference of a neural network using quantized magnetic domain wall (DW)-based synaptic array and CMOS peripheral circuits. A rigorous model of the magnetic DW device considering stochasticity and process variations has been utilized for the synapse. To achieve stable quantized weights, DW pinning has been achieved by means of physical constrictions. Finally, VGG8 architecture for CIFAR-10 image classification has been simulated by using the extracted synaptic device characteristics. The performance in terms of accuracy, energy, latency, and area consumption has been evaluated while considering the process variations and nonidealities in the DW device as well as the peripheral circuits. The proposed quantized neural network (QNN) architecture achieves efficient ON-chip learning with 92.4% and 90.4% training and inference accuracy, respectively. In comparison to pure CMOS-based design, it demonstrates an overall improvement in area, energy, and latency by , , and , respectively.

Ferromagnetic resonance measurement with frequency modulation down to 2 K

Ferromagnetic resonance (FMR) spectroscopy is a powerful technique to study the precessional dynamics of magnetization in thin film heterostructures. It provides valuable information about the mechanisms of exchange bias, spin angular momentum transfer across interfaces, and excitation of magnons. A key desirable feature of FMR spectrometers is the capability to study magnetization dynamics over a wide phase space of temperature (T), frequency (f), and magnetic field (B). The design, fabrication, and testing of such a spectrometer, which uses frequency modulation techniques for improved detection of microwave absorption, reduces heat load in the cryostat and allows simultaneous measurements of inverse spin Hall effect (ISHE) induced dc voltages, is described in this paper. The apparatus is based on a 2-port transmitted microwave signal measurement using a grounded co-planar waveguide. The input radio frequency (RF) signal, frequency modulated at a tunable f-band, excites spin precession in the sample, and the attenuated RF signal is measured phase sensitively. The sample stage, inserted in the bore of a superconducting solenoid, allows magnetic field and temperature variability of 0 to ±5 T and 2-310 K, respectively. We demonstrate the working of this Cryo-FMR and ISHE spectrometer on thin films of Ni80Fe20 and Fe60Co20B20 over a wide T, B, and f phase space.

Multistate Compound Magnetic Tunnel Junction Synapses for Digital Recognition

The quest to mimic the multistate synapses for bioinspired computing has triggered nascent research that leverages the well-established magnetic tunnel junction (MTJ) technology. Early works on the spin transfer torque MTJ-based artificial neural network (ANN) are susceptible to poor thermal reliability, high latency, and high critical current densities. Meanwhile, work on spin-orbit torque (SOT) MTJ-based ANN mainly utilized domain wall motion, which yields negligibly small readout signals differentiating consecutive states and has designs that are incompatible with technological scale-up. Here, we propose a multistate device concept built upon a compound MTJ consisting of multiple SOT-MTJs (number of MTJs, n = 1-4) on a shared write channel, mimicking the spin-based ANN. The n + 1 resistance states representing varying synaptic weights can be tuned by varying the voltage pulses (±1.5-1.8 V), pulse duration (100-300 ns), and applied in-plane fields (5.5-10.5 mT). A large TMR difference of more than 13.6% is observed between two consecutive states for the 4-cell compound MTJ, a 4-fold improvement from reported state-of-the-art spin-based synaptic devices. The ANN built upon the compound MTJ shows high learning accuracy for digital recognition tasks with incremental states and retraining, achieving test accuracy as high as 95.75% in the 4-cell compound MTJ. These results provide an industry-compatible platform to integrate these multistate SOT-MTJ synapses directly into neuromorphic architecture for in-memory and unconventional computing applications.

Proposition of Adaptive Read Bias: A Solution to Overcome Power and Scaling Limitations in Ferroelectric-Based Neuromorphic System

Hardware neuromorphic systems are crucial for the energy-efficient processing of massive amounts of data. Among various candidates, hafnium oxide ferroelectric tunnel junctions (FTJs) are highly promising for artificial synaptic devices. However, FTJs exhibit non-ideal characteristics that introduce variations in synaptic weights, presenting a considerable challenge in achieving high-performance neuromorphic systems. The primary objective of this study is to analyze the origin and impact of these variations in neuromorphic systems. The analysis reveals that the major bottleneck in achieving a high-performance neuromorphic system is the dynamic variation, primarily caused by the intrinsic 1/f noise of the device. As the device area is reduced and the read bias (VRead ) is lowered, the intrinsic noise of the FTJs increases, presenting an inherent limitation for implementing area- and power-efficient neuromorphic systems. To overcome this limitation, an adaptive read-biasing (ARB) scheme is proposed that applies a different VRead to each layer of the neuromorphic system. By exploiting the different noise sensitivities of each layer, the ARB method demonstrates significant power savings of 61.3% and a scaling effect of 91.9% compared with conventional biasing methods. These findings contribute significantly to the development of more accurate, efficient, and scalable neuromorphic systems.

Artificial Neuron Devices

Efforts to design devices emulating complex cognitive abilities and response processes of biological systems have long been a coveted goal. Recent advancements in flexible electronics, mirroring human tissue's mechanical properties, hold significant promise. Artificial neuron devices, hinging on flexible artificial synapses, bioinspired sensors, and actuators, are meticulously engineered to mimic the biological systems. However, this field is in its infancy, requiring substantial groundwork to achieve autonomous systems with intelligent feedback, adaptability, and tangible problem-solving capabilities. This review provides a comprehensive overview of recent advancements in artificial neuron devices. It starts with fundamental principles of artificial synaptic devices and explores artificial sensory systems, integrating artificial synapses and bioinspired sensors to replicate all five human senses. A systematic presentation of artificial nervous systems follows, designed to emulate fundamental human nervous system functions. The review also discusses potential applications and outlines existing challenges, offering insights into future prospects. We aim for this review to illuminate the burgeoning field of artificial neuron devices, inspiring further innovation in this captivating area of research.

Memcapacitor Crossbar Array with Charge Trap NAND Flash Structure for Neuromorphic Computing

The progress of artificial intelligence and the development of large-scale neural networks have significantly increased computational costs and energy consumption. To address these challenges, researchers are exploring low-power neural network implementation approaches and neuromorphic computing systems are being highlighted as potential candidates. Specifically, the development of high-density and reliable synaptic devices, which are the key elements of neuromorphic systems, is of particular interest. In this study, an 8 × 16 memcapacitor crossbar array that combines the technological maturity of flash cells with the advantages of NAND flash array structure is presented. The analog properties of the array with high reliability are experimentally demonstrated, and vector-matrix multiplication with extremely low error is successfully performed. Additionally, with the capability of weight fine-tuning characteristics, a spiking neural network for CIFAR-10 classification via off-chip learning at the wafer level is implemented. These experimental results demonstrate a high level of accuracy of 92.11%, with less than a 1.13% difference compared to software-based neural networks (93.24%).

Emerging memristive artificial neuron and synapse devices for the neuromorphic electronics era

Growth of data eases the way to access the world but requires increasing amounts of energy to store and process. Neuromorphic electronics has emerged in the last decade, inspired by biological neurons and synapses, with in-memory computing ability, extenuating the 'von Neumann bottleneck' between the memory and processor and offering a promising solution to reduce the efforts both in data storage and processing, thanks to their multi-bit non-volatility, biology-emulated characteristics, and silicon compatibility. This work reviews the recent advances in emerging memristive devices for artificial neuron and synapse applications, including memory and data-processing ability: the physics and characteristics are discussed first, i.e., valence changing, electrochemical metallization, phase changing, interfaced-controlling, charge-trapping, ferroelectric tunnelling, and spin-transfer torquing. Next, we propose a universal benchmark for the artificial synapse and neuron devices on spiking energy consumption, standby power consumption, and spike timing. Based on the benchmark, we address the challenges, suggest the guidelines for intra-device and inter-device design, and provide an outlook for the neuromorphic applications of resistive switching-based artificial neuron and synapse devices.

Potential of AlP and GaN as barriers in magnetic tunnel junctions

AlP and GaN are wide band gap semiconductors used industrially in light emitting diodes. We investigate their potential as tunnel barriers in magnetic tunnel junctions, employing density functional theory and the non-equilibrium Green's function method for ground state and quantum transport calculations, respectively. We show that the valence band edges are dominated by pz orbitals and the conduction band edges are dominated by s orbitals. Both materials filter Bloch states of Δ1 symmetry at the Γ-point of the Brillouin zone. In the zero bias limit, we find for the Co/AlP/Co junction a high tunnel magnetoresistance of ∼200% at the Fermi energy and for the Co/GaN/Co junction a tunnel magnetoresistance of even ∼300% about 1.4 eV below the Fermi energy.

Realization of Artificial Neurons and Synapses Based on STDP Designed by an MTJ Device

As the third-generation neural network, the spiking neural network (SNN) has become one of the most promising neuromorphic computing paradigms to mimic brain neural networks over the past decade. The SNN shows many advantages in performing classification and recognition tasks in the artificial intelligence field. In the SNN, the communication between the pre-synapse neuron (PRE) and the post-synapse neuron (POST) is conducted by the synapse. The corresponding synaptic weights are dependent on both the spiking patterns of the PRE and the POST, which are updated by spike-timing-dependent plasticity (STDP) rules. The emergence and growing maturity of spintronic devices present a new approach for constructing the SNN. In the paper, a novel SNN is proposed, in which both the synapse and the neuron are mimicked with the spin transfer torque magnetic tunnel junction (STT-MTJ) device. The synaptic weight is presented by the conductance of the MTJ device. The mapping of the probabilistic spiking nature of the neuron to the stochastic switching behavior of the MTJ with thermal noise is presented based on the stochastic Landau-Lifshitz-Gilbert (LLG) equation. In this way, a simplified SNN is mimicked with the MTJ device. The function of the mimicked SNN is verified by a handwritten digit recognition task based on the MINIST database.

Double-Forming Mechanism of TaOx-Based Resistive Memory Device and Its Synaptic Applications

The bipolar resistive switching properties of Pt/TaOx/InOx/ITO-resistive random-access memory devices under DC and pulse measurement conditions are explored in this work. Transmission electron microscopy and X-ray photoelectron spectroscopy were used to confirm the structure and chemical compositions of the devices. A unique two-step forming process referred to as the double-forming phenomenon and self-compliance characteristics are demonstrated under a DC sweep. A model based on oxygen vacancy migration is proposed to explain its conduction mechanism. Varying reset voltages and compliance currents were applied to evaluate multilevel cell characteristics. Furthermore, pulses were applied to the devices to demonstrate the neuromorphic system's application via testing potentiation, depression, spike-timing-dependent plasticity, and spike-rate-dependent plasticity.

Spiking Dynamics in Dual Free Layer Perpendicular Magnetic Tunnel Junctions

Spintronic devices have recently attracted a lot of attention in the field of unconventional computing due to their non-volatility for short- and long-term memory, nonlinear fast response, and relatively small footprint. Here we demonstrate experimentally how voltage driven magnetization dynamics of dual free layer perpendicular magnetic tunnel junctions can emulate spiking neurons in hardware. The output spiking rate was controlled by varying the dc bias voltage across the device. The field-free operation of this two-terminal device and its robustness against an externally applied magnetic field make it a suitable candidate to mimic the neuron response in a dense neural network. The small energy consumption of the device (4-16 pJ/spike) and its scalability are important benefits for embedded applications. This compact perpendicular magnetic tunnel junction structure could finally bring spiking neural networks to sub-100 nm size elements.

Artificial Neuronal Devices Based on Emerging Materials: Neuronal Dynamics and Applications

Artificial neuronal devices are critical building blocks of neuromorphic computing systems and currently the subject of intense research motivated by application needs from new computing technology and more realistic brain emulation. Researchers have proposed a range of device concepts that can mimic neuronal dynamics and functions. Although the switching physics and device structures of these artificial neurons are largely different, their behaviors can be described by several neuron models in a more unified manner. In this paper, the reports of artificial neuronal devices based on emerging volatile switching materials are reviewed from the perspective of the demonstrated neuron models, with a focus on the neuronal functions implemented in these devices and the exploitation of these functions for computational and sensing applications. Furthermore, the neuroscience inspirations and engineering methods to enrich the neuronal dynamics that remain to be implemented in artificial neuronal devices and networks toward realizing the full functionalities of biological neurons are discussed.

Proton-Gated Synaptic Transistors, Based on an Electron-Beam Patterned Nafion Electrolyte

Neuromorphic processors using artificial neural networks are the center of attention for energy-efficient analog computing. Artificial synapses act as building blocks in such neural networks for parallel information processing and data storage. Herein we describe the fabrication of a proton-gated synaptic transistor using a Nafion electrolyte thin film, which is patterned by electron-beam lithography (EBL). The device has an active channel of indium-zinc-oxide (IZO) between the source and drain electrodes, which shows Ohmic behavior with a conductance level on the order of 100 μS. Under voltage applications to the gate electrode, the channel conductance is changed due to the injection and extraction of protons between the IZO channel and the Nafion electrolyte, emulating various synaptic functions with short-term and long-term plasticity. When positive (negative) gate voltage pulses are consecutively applied, the device exhibits long-term potentiation (depression) at the same number of steps as the number of input pulses. Based on these characteristics, an artificial neural network using this transistor shows ∼84% image recognition accuracy for handwritten digits. The subject transistor also successfully mimics paired-pulse facilitation and depression, Hebbian spike-timing-dependent plasticity, and Pavlovian associative learning followed by extinction activities. Finally, dynamical pattern image memorization is demonstrated in a 5 × 5 array of these synaptic transistors. The results indicate that EBL patternable Nafion electrolytes have great potential for use in the fabrication and circuit-level integration of synaptic devices for neuromorphic computing applications.

All oxide based flexible multi-folded invisible synapse as vision photo-receptor

All oxide-based transparent flexible memristor is prioritized for the potential application in artificially simulated biological optoelectronic synaptic devices. SnOx memristor with HfOx layer is found to enable a significant effect on synaptic properties. The memristor exhibits good reliability with long retention, 104 s, and high endurance, 104 cycles. The optimized 6 nm thick HfOx layer in SnOx-based memristor possesses the excellent synaptic properties of stable 350 epochs training, multi-level conductance (MLC) behaviour, and the nonlinearity of 1.53 and 1.46 for long-term potentiation and depression, respectively, and faster image recognition accuracy of 100% after 23 iterations. The maximum weight changes of -73.12 and 79.91% for the potentiation and depression of the synaptic device, respectively, are observed from the spike-timing-dependent plasticity (STDP) characteristics making it suitable for biological applications. The flexibility of the device on the PEN substrate is confirmed by the acceptable change of nonlinearities up to 4 mm bending. Such a synaptic device is expected to be used as a vision photo-receptor.

E-Spin: A Stochastic Ising Spin Based on Electrically-Controlled MTJ for Constructing Large-Scale Ising Annealing Systems

With its unique computer paradigm, the Ising annealing machine has become an emerging research direction. The Ising annealing system is highly effective at addressing combinatorial optimization (CO) problems that are difficult for conventional computers to tackle. However, Ising spins, which comprise the Ising system, are difficult to implement in high-performance physical circuits. We propose a novel type of Ising spin based on an electrically-controlled magnetic tunnel junction (MTJ). Electrical operation imparts true randomness, great stability, precise control, compact size, and easy integration to the MTJ-based spin. In addition, simulations demonstrate that the frequency of electrically-controlled stochastic Ising spin (E-spin) is 50 times that of the thermal disturbance MTJ-based spin (p-bit). To develop a large-scale Ising annealing system, up to 64 E-spins are implemented. Our Ising annealing system demonstrates factorization of integers up to 264 with a temporal complexity of around O(n). The proposed E-spin shows superiority in constructing large-scale Ising annealing systems and solving CO problems.

Lateral double magnetic tunnel junction device with orthogonal polarizer for high-performance magnetoresistive memory

Magnetic tunnel junction (MTJ)-based memory devices have larger switching delay and energy consumption, compared to cache or dynamic random access memory. In order to broaden the applications of the magnetoresistive random access memory, reducing the switching time and energy consumption of the MTJ is required. Here, a novel lateral double MTJ with an orthogonal polarizer is proposed. The proposed device consists of three ferromagnetic regions: the first pinned region (PR1) with perpendicular magnetic anisotropy (PMA), a free region (FR) with PMA, and the second pinned region (PR2) with in-plane magnetic anisotropy (IMA). PR1 and PR2 are placed on top of the oxide barrier, which separates them from the FR, comprising a lateral double MTJ structure. The current pulse through PR2 helps to perturb the magnetization of the FR. Since the angle between PR2 and FR is 90°, the initial torque increases significantly, decreasing switching delay by 4.02 times and energy-delay product by 7.23 times. It is also shown, that the area of the access transistor can be reduced by approximately 10%, while maintaining the same energy-delay product and reducing gate RC delay.

Compact artificial neuron based on anti-ferroelectric transistor

Neuromorphic machines are intriguing for building energy-efficient intelligent systems, where spiking neurons are pivotal components. Recently, memristive neurons with promising bio-plausibility have been developed, but with limited reliability, bulky capacitors or additional reset circuits. Here, we propose an anti-ferroelectric field-effect transistor neuron based on the inherent polarization and depolarization of Hf_0.2Zr_0.8O₂ anti-ferroelectric film to meet these challenges. The intrinsic accumulated polarization/spontaneous depolarization of Hf_0.2Zr_0.8O₂ films implements the integration/leaky behavior of neurons, avoiding external capacitors and reset circuits. Moreover, the anti-ferroelectric neuron exhibits low energy consumption (37 fJ/spike), high endurance (>10¹²), high uniformity and high stability. We further construct a two-layer fully ferroelectric spiking neural networks that combines anti-ferroelectric neurons and ferroelectric synapses, achieving 96.8% recognition accuracy on the Modified National Institute of Standards and Technology dataset. This work opens the way to emulate neurons with anti-ferroelectric materials and provides a promising approach to building high-efficient neuromorphic hardware.

Electric-Field Control of Spin Diffusion Length and Electric-Assisted D'yakonov-Perel' Mechanism in Ultrathin Heavy Metal and Ferromagnetic Insulator Heterostructure

Electric-field control of spin dynamics is significant for spintronic device applications. Thus far, effectively electric-field control of magnetic order, magnetic damping factor and spin-orbit torque (SOT) has been studied in magnetic materials, but the electric field control of spin relaxation still remains unexplored. Here, we use ionic liquid gating to control spin-related property in the ultra-thin (4 nm) heavy metal (HM) platinum (Pt) and ferromagnetic insulator (FMI) yttrium iron garnet (Y₃Fe₅O₁₂, YIG) heterostructure. It is found that the anomalous Hall effect (AHE), spin relaxation time and spin diffusion length can be effectively controlled by the electric field. The anomalous Hall resistance is almost twice as large as at 0 voltage after applying a small voltage of 5.5 V. The spin relaxation time can vary by more than 50 percent with the electric field, from 41.6 to 64.5 fs. In addition, spin relaxation time at different gate voltage follows the reciprocal law of the electron momentum scattering time, which indicates that the D'yakonov-Perel' mechanism is dominant in the Pt/YIG system. Furthermore, the spin diffusion length can be effectively controlled by an ionic gate, which can be well explained by voltage-modulated interfacial spin scattering. These results help us to improve the interface spin transport properties in magnetic materials, with great contributions to the exploration of new physical mechanisms and spintronics device.

NeuroPack: An Algorithm-Level Python-Based Simulator for Memristor-Empowered Neuro-Inspired Computing

Emerging two-terminal nanoscale memory devices, known as memristors, have demonstrated great potential for implementing energy-efficient neuro-inspired computing architectures over the past decade. As a result, a wide range of technologies have been developed that, in turn, are described via distinct empirical models. This diversity of technologies requires the establishment of versatile tools that can enable designers to translate memristors’ attributes in novel neuro-inspired topologies. In this study, we present NeuroPack, a modular, algorithm-level Python-based simulation platform that can support studies of memristor neuro-inspired architectures for performing online learning or offline classification. The NeuroPack environment is designed with versatility being central, allowing the user to choose from a variety of neuron models, learning rules, and memristor models. Its hierarchical structure empowers NeuroPack to predict any memristor state changes and the corresponding neural network behavior across a variety of design decisions and user parameter options. The use of NeuroPack is demonstrated herein via an application example of performing handwritten digit classification with the MNIST dataset and an existing empirical model for metal-oxide memristors.

Spin Reflection-Induced Field-Free Magnetization Switching in Perpendicularly Magnetized MgO/Pt/Co Heterostructures

Field-free magnetization switching is critical towards practical, integrated spin-orbit torque (SOT)-driven magnetic random-access memory with perpendicular magnetic anisotropy. Our work proposes a technique to modulate the spin reflection and spin density of states within a heavy-metal Pt through interfacing with a dielectric MgO layer. We demonstrate tunability of the effective out-of-plane spin torque acting on the ferromagnetic Co layer, enabling current-induced SOT magnetization switching without the assistance of an external magnetic field. The influence of the MgO layer thickness on effective SOT efficiency shows saturation at 4 nm, while up to 80% of field-free magnetization switching ratio is achieved with the MgO between 5 and 8 nm. We analyze and attribute the complex interaction to spin reflection at the dielectric/heavy metal interface and spin scattering within the dielectric medium due to interfacial electric fields. Further, through substituting the dielectric with Ti or Pt, we confirm that the MgO layer is indeed responsible for the observed field-free magnetization switching mechanism.

Complementary Photo-Synapses Based on Light-Stimulated Porphyrin-Coated Silicon Nanowires Field-Effect Transistors (LPSNFET)

Neuromorphic computing has emerged as the high-energy-efficiency and intelligent solution for processing sensory data. As a potential alternative to neuromorphic computing, photo-excited synaptic systems can integrate the functions of optoelectronic sensing and synaptic computing to realize the low-power and high-performance visual perception. However, one major challenge in high-efficient photo-excited synaptic system is to realize the complementarily enhanced and inhibited synaptic behaviors with small hardware cost as possible. Another challenge is to fabricate the photo-synapse devices with complementary metal oxide semiconductor (CMOS)-compatible process to achieve high enough integration density for practical application. Here, a CMOS-compatible Light-stimulated Porphyrin-coated Silicon Nanowire Field Effect Transistor (LPSNFET) technology is proposed and developed to form the complementary photo-synapses with only two CMOS-like transistors. LPSNFET exhibits fivefold improvement in photo-sensitivity compared to the bare silicon nanowire (SiNW) devices, and can still show obvious responses when incident illumination power is as low as 0.1 mW cm^-2 . Moreover, it enables tunable dynamic synaptic plasticity and versatile synaptic functions. Especially, the complementarily enhanced and inhibited behaviors can be realized by modulating SiNW/porphyrin interface via simply changing the MOS type of LPSNFET, which acts like the photonic counterpart of CMOS technology to provide the basic brick for building complex neuromorphic circuits efficiently and economically. Finally, the CMOS process compatibility of LPSNFET provides potential application in future large scale in-sensor computing.

IGZO/CsPbBr3-Nanoparticles/IGZO Neuromorphic Phototransistors and Their Optoelectronic Coupling Applications

Optoelectronic synaptic devices are of great scientific and practical importance because of various potential applications such as ocular simulating and optical-electrical managers based on a new optoelectronic coupling mechanism. In this work, we design a novel channel layer with p-type CsPbBr₃ nanoparticles (NPs) buried in an InGaZnO (IGZO) film to construct the corresponding thin-film transistors (TFTs), which exhibits intense improvement in visible-light photosensitivity and synaptic plasticity as compared to the pure IGZO counterpart. Specifically, the composite device is able to exhibit versatile synaptic behavior under light stimuli with density as low as 0.12 μW/cm² and with the gain 5-20 times higher than that of the IGZO TFT in the visible-light region. Based on the band alignment between the IGZO and NPs, the excitation and decay processes of intrinsic and photoinduced carriers are discussed. Moreover, owing to the gate bias control in a three-terminal configuration, our TFT synapses can imitate complex biological behaviors including the famous "Pavlov's dog" experiment and the "reward and punishment mechanism" of the brain via editing the gate voltage/light pulse stimuli.

Neuromorphic Engineering: From Biological to Spike-Based Hardware Nervous Systems

The human brain is a sophisticated, high-performance biocomputer that processes multiple complex tasks in parallel with high efficiency and remarkably low power consumption. Scientists have long been pursuing an artificial intelligence (AI) that can rival the human brain. Spiking neural networks based on neuromorphic computing platforms simulate the architecture and information processing of the intelligent brain, providing new insights for building AIs. The rapid development of materials engineering, device physics, chip integration, and neuroscience has led to exciting progress in neuromorphic computing with the goal of overcoming the von Neumann bottleneck. Herein, fundamental knowledge related to the structures and working principles of neurons and synapses of the biological nervous system is reviewed. An overview is then provided on the development of neuromorphic hardware systems, from artificial synapses and neurons to spike-based neuromorphic computing platforms. It is hoped that this review will shed new light on the evolution of brain-like computing.

Antiferromagnet-Based Neuromorphics Using Dynamics of Topological Charges

We propose a spintronics-based hardware implementation of neuromorphic computing, specifically, the spiking neural network, using topological winding textures in one-dimensional antiferromagnets. The consistency of such a network is emphasized in light of the conservation of topological charges, and the natural spatiotemporal interconversions of magnetic winding. We discuss the realization of the leaky integrate-and-fire behavior of neurons and the spike-timing-dependent plasticity of synapses. Our proposal opens the possibility for an all-spin neuromorphic platform based on antiferromagnetic insulators.

Prospect of Spin-Orbitronic Devices and Their Applications

Science, engineering, and medicine ultimately demand fast information processing with ultra-low power consumption. The recently developed spin-orbit torque (SOT)-induced magnetization switching paradigm has been fueling opportunities for spin-orbitronic devices, i.e., enabling SOT memory and logic devices at sub-nano second and sub-picojoule regimes. Importantly, spin-orbitronic devices are intrinsic of nonvolatility, anti-radiation, unlimited endurance, excellent stability, and CMOS compatibility, toward emerging applications, e.g., processing in-memory, neuromorphic computing, probabilistic computing, and 3D magnetic random access memory. Nevertheless, the cutting-edge SOT-based devices and application remain at a premature stage owing to the lack of scalable methodology on the field-free SOT switching. Moreover, spin-orbitronics poises as an interdisciplinary field to be driven by goals of both fundamental discoveries and application innovations, to open fascinating new paths for basic research and new line of technologies. In this perspective, the specific challenges and opportunities are summarized to exert momentum on both research and eventual applications of spin-orbitronic devices.

Multi-terminal ionic-gated low-power silicon nanowire synaptic transistors with dendritic functions for neuromorphic systems

Neuromorphic computing systems have shown powerful capability in tasks, such as recognition, learning, classification and decision-making, which are both challenging and inefficient in using the traditional computation architecture. The key elements including synapses and neurons, and their feasible hardware implementation are essential for practical neuromorphic computing. However, most existing synaptic devices used to emulate functions of a single synapse and the synapse-based networks are more energy intensive and less sustainable than their biological counterparts. The dendritic functions such as integration of spatiotemporal signals and spike-frequency coding characteristics have not been well implemented in a single synaptic device and thus play an imperative role in future practical hardware-based spiking neural networks. Moreover, most emerging synaptic transistors are fabricated by nanofabrication processes without CMOS compatibility for further wafer-scale integration. Herein, we demonstrate a novel ionic-gated silicon nanowire synaptic field-effect transistor (IGNWFET) with low power consumption (<400 fJ per switching event) based on the standard CMOS process platform. For the first time, the dendritic integration and dual-synaptic dendritic computations (such as "Add" and "Subtraction") could be realized by processing frequency coded spikes using a single device. Meanwhile, multi-functional characteristics of artificial synapses including the short-term and long-term synaptic plasticity, paired pulse facilitation and high-pass filtering were also successfully demonstrated based on 40 nm wide IGNWFETs. The migration of ions in polymer electrolyte and trapping in high-k dielectric were also experimentally studied in-depth to understand the short-term plasticity and long-term plasticity. Combined with statistical uniformity across a 4-inch wafer, the comprehensive performance of IGNWFET demonstrates its potential application in future biologically emulated neuromorphic systems.

Synaptic Characteristics of Amorphous Boron Nitride-Based Memristors on a Highly Doped Silicon Substrate for Neuromorphic Engineering

In this study, the resistive switching and synaptic properties of a complementary metal-oxide semiconductor-compatible Ti/a-BN/Si device are investigated for neuromorphic systems. A gradual change in resistance is observed in a positive SET operation in which Ti diffusion is involved in the conducting path. This operation is extremely suitable for synaptic devices in hardware-based neuromorphic systems. The isosurface charge density plots and experimental results confirm that boron vacancies can help generate a conducting path, whereas the conducting path generated by a Ti cation from interdiffusion forms is limited. A negative SET operation causes a considerable decrease in the formation energy of only boron vacancies, thereby increasing the conductivity in the low-resistance state, which may be related to RESET failure and poor endurance. The pulse transient characteristics, potentiation and depression characteristics, and good retention property of eight multilevel cells also indicate that the positive SET operation is more suitable for a synaptic device owing to the gradual modulation of conductance. Moreover, pattern recognition accuracy is examined by considering the conductance values of the measured data in the Ti/a-BN/Si device as the synaptic part of a neural network. The linear and symmetric synaptic weight update in a positive SET operation with an incremental voltage pulse scheme ensures higher pattern recognition accuracy.

Enhanced Photoinduced Carrier Generation Efficiency through Surface Band Bending in Topological Insulator Bi2Se3 Thin Films by the Oxidized Layer

Topological insulators (TIs) have become popular in the field of optoelectronic devices because of their broadband and high-sensitivity properties, which are attributed to the narrow band gap of the bulk state and high mobility of the Dirac surface state. Although perfectly grown TIs are known to exhibit strong stability against oxidation, in most cases, the existence of vacancy defects in TIs reacts to air and the characteristics of TIs is affected by oxidation. Therefore, changes in the band structure and electrical characteristics by oxidation should be considered. A significant change occurs because of the oxidation; however, the dependence of the photoresponse of TIs on oxidation has not been studied in detail. In this study, the photoresponsivity of oxidized Bi₂Se₃ films is enhanced, rather than degraded, after oxidation in air for 24 h, resulting in a maximum responsivity of 140 mA W^-1. This responsivity is substantially higher than previously reported values for Bi₂Se₃. Furthermore, a change in the photoresponse time of Bi₂Se₃ due to air exposure is systematically observed. Based on variations in the Fermi level and work function, using photoelectron spectroscopy, it is confirmed that the responsivity is improved from the junction effect of the Bi-based surface oxidized layer.

Improving the Recognition Accuracy of Memristive Neural Networks via Homogenized Analog Type Conductance Quantization

Conductance quantization (QC) phenomena occurring in metal oxide based memristors demonstrate great potential for high-density data storage through multilevel switching, and analog synaptic weight update for effective training of the artificial neural networks. Continuous, linear and symmetrical modulation of the device conductance is a critical issue in QC behavior of memristors. In this contribution, we employ the scanning probe microscope (SPM) assisted electrode engineering strategy to control the ion migration process to construct single conductive filaments in Pt/HfO_x/Pt devices. Upon deliberate tuning and evolution of the filament, 32 half integer quantized conductance states in the 16 G₀ to 0.5 G₀ range with enhanced distribution uniformity was achieved. Simulation results revealed that the numbers of the available QC states and fluctuation of the conductance at each state play an important role in determining the overall performance of the neural networks. The 32-state QC behavior of the hafnium oxide device shows improved recognition accuracy approaching 90% for handwritten digits, based on analog type operation of the multilayer perception (MLP) neural network.

Exploring the Impact of Variability in Resistance Distributions of RRAM on the Prediction Accuracy of Deep Learning Neural Networks

In this work, we explore the use of the resistive random access memory (RRAM) device as a synapse for mimicking the trained weights linking neurons in a deep learning neural network (DNN) (AlexNet). The RRAM devices were fabricated in-house and subjected to 1000 bipolar read-write cycles to measure the resistances recorded for Logic-0 and Logic-1 (we demonstrate the feasibility of achieving eight discrete resistance states in the same device depending on the RESET stop voltage). DNN simulations have been performed to compare the relative error between the output of AlexNet Layer 1 (Convolution) implemented with the standard backpropagation (BP) algorithm trained weights versus the weights that are encoded using the measured resistance distributions from RRAM. The IMAGENET dataset is used for classification purpose here. We focus only on the Layer 1 weights in the AlexNet framework with 11 × 11 × 96 filters values coded into a binary floating point and substituted with the RRAM resistance values corresponding to Logic-0 and Logic-1. The impact of variability in the resistance states of RRAM for the low and high resistance states on the accuracy of image classification is studied by formulating a look-up table (LUT) for the RRAM (from measured I-V data) and comparing the convolution computation output of AlexNet Layer 1 with the standard outputs from the BP-based pre-trained weights. This is one of the first studies dedicated to exploring the impact of RRAM device resistance variability on the prediction accuracy of a convolutional neural network (CNN) on an AlexNet platform through a framework that requires limited actual device switching test data.

Memristive and CMOS Devices for Neuromorphic Computing

Neuromorphic computing has emerged as one of the most promising paradigms to overcome the limitations of von Neumann architecture of conventional digital processors. The aim of neuromorphic computing is to faithfully reproduce the computing processes in the human brain, thus paralleling its outstanding energy efficiency and compactness. Toward this goal, however, some major challenges have to be faced. Since the brain processes information by high-density neural networks with ultra-low power consumption, novel device concepts combining high scalability, low-power operation, and advanced computing functionality must be developed. This work provides an overview of the most promising device concepts in neuromorphic computing including complementary metal-oxide semiconductor (CMOS) and memristive technologies. First, the physics and operation of CMOS-based floating-gate memory devices in artificial neural networks will be addressed. Then, several memristive concepts will be reviewed and discussed for applications in deep neural network and spiking neural network architectures. Finally, the main technology challenges and perspectives of neuromorphic computing will be discussed.

Neuromorphic Spintronics

Neuromorphic computing uses basic principles inspired by the brain to design circuits that perform artificial intelligence tasks with superior energy efficiency. Traditional approaches have been limited by the energy area of artificial neurons and synapses realized with conventional electronic devices. In recent years, multiple groups have demonstrated that spintronic nanodevices, which exploit the magnetic as well as electrical properties of electrons, can increase the energy efficiency and decrease the area of these circuits. Among the variety of spintronic devices that have been used, magnetic tunnel junctions play a prominent role because of their established compatibility with standard integrated circuits and their multifunctionality. Magnetic tunnel junctions can serve as synapses, storing connection weights, functioning as local, nonvolatile digital memory or as continuously varying resistances. As nano-oscillators, they can serve as neurons, emulating the oscillatory behavior of sets of biological neurons. As superparamagnets, they can do so by emulating the random spiking of biological neurons. Magnetic textures like domain walls or skyrmions can be configured to function as neurons through their non-linear dynamics. Several implementations of neuromorphic computing with spintronic devices demonstrate their promise in this context. Used as variable resistance synapses, magnetic tunnel junctions perform pattern recognition in an associative memory. As oscillators, they perform spoken digit recognition in reservoir computing and when coupled together, classification of signals. As superparamagnets, they perform population coding and probabilistic computing. Simulations demonstrate that arrays of nanomagnets and films of skyrmions can operate as components of neuromorphic computers. While these examples show the unique promise of spintronics in this field, there are several challenges to scaling up, including the efficiency of coupling between devices and the relatively low ratio of maximum to minimum resistances in the individual devices.

Magnetic Tunnel Junction Applications

Spin-based devices can reduce energy leakage and thus increase energy efficiency. They have been seen as an approach to overcoming the constraints of CMOS downscaling, specifically, the Magnetic Tunnel Junction (MTJ) which has been the focus of much research in recent years. Its nonvolatility, scalability and low power consumption are highly attractive when applied in several components. This paper aims at providing a survey of a selection of MTJ applications such as memory and analog to digital converter, among others.

Bridging Biological and Artificial Neural Networks with Emerging Neuromorphic Devices: Fundamentals, Progress, and Challenges

As the research on artificial intelligence booms, there is broad interest in brain-inspired computing using novel neuromorphic devices. The potential of various emerging materials and devices for neuromorphic computing has attracted extensive research efforts, leading to a large number of publications. Going forward, in order to better emulate the brain's functions, its relevant fundamentals, working mechanisms, and resultant behaviors need to be re-visited, better understood, and connected to electronics. A systematic overview of biological and artificial neural systems is given, along with their related critical mechanisms. Recent progress in neuromorphic devices is reviewed and, more importantly, the existing challenges are highlighted to hopefully shed light on future research directions.

Towards spike-based machine intelligence with neuromorphic computing

Guided by brain-like 'spiking' computational frameworks, neuromorphic computing-brain-inspired computing for machine intelligence-promises to realize artificial intelligence while reducing the energy requirements of computing platforms. This interdisciplinary field began with the implementation of silicon circuits for biological neural routines, but has evolved to encompass the hardware implementation of algorithms with spike-based encoding and event-driven representations. Here we provide an overview of the developments in neuromorphic computing for both algorithms and hardware and highlight the fundamentals of learning and hardware frameworks. We discuss the main challenges and the future prospects of neuromorphic computing, with emphasis on algorithm-hardware codesign.

Tuning magnetic anisotropy in Co-BaZrO3 vertically aligned nanocomposites for memory device integration

Ferromagnetic nanostructures with strong anisotropic properties are highly desired for their potential integration into spintronic devices. Several anisotropic candidates, such as CoFeB and Fe-Pt, have been previously proposed, but many of them have limitations such as patterning issues or thickness restrictions. In this work, Co-BaZrO3 (Co-BZO) vertically aligned nanocomposite (VAN) films with tunable magnetic anisotropy and coercive field strength have been demonstrated to address this need. Such tunable magnetic properties are achieved through tuning the thickness of the Co-BZO VAN structures and the aspect ratio of the Co nanostructures, which can be easily integrated into spintronic devices. As a demonstration, we have integrated the Co-BZO VAN nanostructure into tunnel junction devices, which demonstrated resistive switching alluding to Co-BZO's immense potential for future spintronic devices.

Two-dimensional materials for synaptic electronics and neuromorphic systems

Synapses in biology provide a variety of functions for the neural system. Artificial synaptic electronics that mimic the biological neuron functions are basic building blocks and developing novel artificial synapses is essential for neuromorphic computation. Inspired by the unique features of biological synapses that the basic connection components of the nervous system and the parallelism, low power consumption, fault tolerance, self-learning and robustness of biological neural systems, artificial synaptic electronics and neuromorphic systems have the potential to overcome the traditional von Neumann bottleneck and create a new paradigm for dealing with complex problems such as pattern recognition, image classification, decision making and associative learning. Nowadays, two-dimensional (2D) materials have drawn great attention in simulating synaptic dynamic plasticity and neuromorphic computing with their unique properties. Here we describe the basic concepts of bio-synaptic plasticity and learning, the 2D materials library and its preparation. We review recent advances in synaptic electronics and artificial neuromorphic systems based on 2D materials and provide our perspective in utilizing 2D materials to implement synaptic electronics and neuromorphic systems in hardware.

Double Pinned Perpendicular-Magnetic-Tunnel-Junction Spin-Valve Providing Multi-level Resistance States

A new design for high density integration greater than gigabits of perpendicular-magnetic-tunnel-junction (p-MTJ) spin-valve, called the double pinned (i.e., bottom and top pinned structures) p-MTJ spin-valve achieved a multi-level memory-cell operation exhibiting four-level resistances. Three key magnetic properties, the anisotropy exchange field (H_ex) of the bottom pinned structure, the coercivity (H_c) of the double free-layer, and the H_c of the top pinned structure mainly determined four-level resistances producing tunneling-magnetoresistance (TMR) ratios of 152.6%, 33.6%, and 166.5%. The three key-design concepts are: i) the bottom pinned structure with a sufficiently large H_ex to avoid a write-error, ii) the H_c of the double free-layer (i.e., ~0.1 kOe) much less than the H_c of the top pinned structure (i.e., ~1.0 kOe), and iii) the top pinned structure providing different electron spin directions.

Reversible uptake and release of sodium ions in layered SnS2-reduced graphene oxide composites for neuromorphic devices

With the advent of brain-inspired computing for complex data processing, emerging nonvolatile memories have been widely studied to develop neuromorphic devices for pattern recognition and deep learning. However, the devices still suffer from limitations such as nonlinearity and large write noise because they adopt a stochastic switching approach. Here, we suggest a biomimetic three-terminal electrochemical artificial synapse that is operated by a conductance change in response to intercalation of sodium (Na+) ions into a layered SnS2-reduced graphene oxide (RGO) composite channel. SnS2-RGO can reversibly uptake and release Na+ ions, so the conductance of the channel in artificial synapse can be controlled effectively and thereby it can emulate essential synaptic functions including short-term plasticity, spatiotemporal signal processing, and transition from short-term to long-term plasticity. The artificial synapse also shows linear and symmetric potentiation/depression with low cycle-to-cycle variation; these responses could improve the write linearity and reduce the write noise of devices. This study demonstrates the feasibility of next-generation neuromorphic memory using ion-based electrochemical devices that can mimic biological synapses with the migration of Na+ ions.

A Swarm Optimization Solver Based on Ferroelectric Spiking Neural Networks

As computational models inspired by the biological neural system, spiking neural networks (SNN) continue to demonstrate great potential in the landscape of artificial intelligence, particularly in tasks such as recognition, inference, and learning. While SNN focuses on achieving high-level intelligence of individual creatures, Swarm Intelligence (SI) is another type of bio-inspired models that mimic the collective intelligence of biological swarms, i.e., bird flocks, fish school and ant colonies. SI algorithms provide efficient and practical solutions to many difficult optimization problems through multi-agent metaheuristic search. Bridging these two distinct subfields of artificial intelligence has the potential to harness collective behavior and learning ability of biological systems. In this work, we explore the feasibility of connecting these two models by implementing a generalized SI model on SNN. In the proposed computing paradigm, we use SNNs to represent agents in the swarm and encode problem solutions with the spike firing rate and with spike timing. The coupled neurons communicate and modulate each other's action potentials through event-driven spikes and synchronize their dynamics around the states of optimal solutions. We demonstrate that such an SI-SNN model is capable of efficiently solving optimization problems, such as parameter optimization of continuous functions and a ubiquitous combinatorial optimization problem, namely, the traveling salesman problem with near-optimal solutions. Furthermore, we demonstrate an efficient implementation of such neural dynamics on an emerging hardware platform, namely ferroelectric field-effect transistor (FeFET) based spiking neurons. Such an emerging in-silico neuron is composed of a compact 1T-1FeFET structure with both excitatory and inhibitory inputs. We show that the designed neuromorphic system can serve as an optimization solver with high-performance and high energy-efficiency.

Flexible organic synaptic device based on poly (methyl methacrylate):CdSe/CdZnS quantum-dot nanocomposites

A synaptic device that functionally mimics a biological synapse is a promising candidate for use as an electronic element in a neuromorphic system. In this study, flexible electronic synaptic devices based on poly (methyl methacrylate) (PMMA):CdSe/CdZnS core-shell quantum-dot (QD) nanocomposites are demonstrated. The current-voltage characteristics for the synaptic devices under consecutive voltage sweeps show clockwise hysteresis, which is a critical feature of an artificial synaptic device. The effect of the CdSe/CdZnS QD concentration on the device performance is studied. The flexible electronic synaptic devices under bending show the similar and stable electrical performances. The memory retention measurements show that the e-synapse exhibits long-term potentiation and depression. The carrier transport mechanisms are analyzed, and thermionic emission and space-charge-limited-current conduction are found to be dominant.

Tunable Resistive Switching Enabled by Malleable Redox Reaction in the Nano-Vacuum Gap

Neuromorphic computing has emerged as a highly promising alternative to conventional computing. The key to constructing a large-scale neural network in hardware for neuromorphic computing is to develop artificial neurons with leaky integrate-and-fire behavior and artificial synapses with synaptic plasticity using nanodevices. So far, these two basic computing elements have been built in separate devices using different materials and technologies, which poses a significant challenge to system design and manufacturing. In this work, we designed a resistive device embedded with an innovative nano-vacuum gap between a bottom electrode and a mixed-ionic-electronic-conductor (MIEC) layer. Through redox reaction on the MIEC surface, metallic filaments dynamically grew within the nano-vacuum gap. The nano-vacuum gap provided an additional control factor for controlling the evolution dynamics of metallic filaments by tuning the electron tunneling efficiency, in analogy to a pseudo-three-terminal device, resulting in tunable switching behavior in various forms from volatile to nonvolatile switching in a single device. Our device demonstrated cross-functions, in particular, tunable neuronal firing and synaptic plasticity on demand, providing seamless integration for building large-scale artificial neural networks for neuromorphic computing.

A Soft-Pruning Method Applied During Training of Spiking Neural Networks for In-memory Computing Applications

Inspired from the computational efficiency of the biological brain, spiking neural networks (SNNs) emulate biological neural networks, neural codes, dynamics, and circuitry. SNNs show great potential for the implementation of unsupervised learning using in-memory computing. Here, we report an algorithmic optimization that improves energy efficiency of online learning with SNNs on emerging non-volatile memory (eNVM) devices. We develop a pruning method for SNNs by exploiting the output firing characteristics of neurons. Our pruning method can be applied during network training, which is different from previous approaches in the literature that employ pruning on already-trained networks. This approach prevents unnecessary updates of network parameters during training. This algorithmic optimization can complement the energy efficiency of eNVM technology, which offers a unique in-memory computing platform for the parallelization of neural network operations. Our SNN maintains ~90% classification accuracy on the MNIST dataset with up to ~75% pruning, significantly reducing the number of weight updates. The SNN and pruning scheme developed in this work can pave the way toward applications of eNVM based neuro-inspired systems for energy efficient online learning in low power applications.

ReStoCNet: Residual Stochastic Binary Convolutional Spiking Neural Network for Memory-Efficient Neuromorphic Computing

In this work, we propose ReStoCNet, a residual stochastic multilayer convolutional Spiking Neural Network (SNN) composed of binary kernels, to reduce the synaptic memory footprint and enhance the computational efficiency of SNNs for complex pattern recognition tasks. ReStoCNet consists of an input layer followed by stacked convolutional layers for hierarchical input feature extraction, pooling layers for dimensionality reduction, and fully-connected layer for inference. In addition, we introduce residual connections between the stacked convolutional layers to improve the hierarchical feature learning capability of deep SNNs. We propose Spike Timing Dependent Plasticity (STDP) based probabilistic learning algorithm, referred to as Hybrid-STDP (HB-STDP), incorporating Hebbian and anti-Hebbian learning mechanisms, to train the binary kernels forming ReStoCNet in a layer-wise unsupervised manner. We demonstrate the efficacy of ReStoCNet and the presented HB-STDP based unsupervised training methodology on the MNIST and CIFAR-10 datasets. We show that residual connections enable the deeper convolutional layers to self-learn useful high-level input features and mitigate the accuracy loss observed in deep SNNs devoid of residual connections. The proposed ReStoCNet offers >20 × kernel memory compression compared to full-precision (32-bit) SNN while yielding high enough classification accuracy on the chosen pattern recognition tasks.

Engineering synaptic characteristics of TaOx/HfO2 bi-layered resistive switching device

We performed various pulse measurements on an atomic layer deposited (ALD) HfO₂-based resistive switching random access memory (RRAM) device and investigated its electronic synaptic characteristics. Unlike requirements for RRAM device application, to achieve the multi-state conductance changes required for the synaptic device, we employed additional sputtered TaO_x thin film formation on the ALD HfO₂ switching medium, which leads to engineering the concentration of oxygen vacancies and modulating the conductive filaments. With this TaO_x/HfO₂ bi-layered device, we attained gradual resistive switching, linear and symmetric conductance change, improved endurance and reproducibility characteristics compared to a single HfO₂ device. Finally, we emulated spike-timing-dependent plasticity based learning rule with pulses inspired by neural action potential, indicating its potential as an electronic synaptic device in a hardware neuromorphic system.

An electronic synaptic device based on HfO2TiOx bilayer structure memristor with self-compliance and deep-RESET characteristics

We reported on a Ti/HfO₂/TiO_x/Pt memristor with self-compliance, deep-RESET characteristics and excellent switching performance, including ultrafast program/erase speed (10 ns), a large memory window (10³) and good pulse endurance (10⁷ cycles). The self-compliance and deep-RESET characteristics are beneficial for protecting the device from permanent breakdown in both SET and RESET processes especially under the pulse operation mode. In addition to bistable state switching, we also achieved multiple or even continuous conductance state switching under a DC sweep and a pulse-train operation mode in the Ti/HfO₂/TiO_x/Pt memristor, which can be seen as a substitution of a biological synapse. The capability of continuous modulation conductance (synaptic weight) in the Ti/HfO₂/TiO_x/Pt memristor was investigated and the potentiation and depression characteristics of the synaptic weight could be precisely tuned by the number or amplitude of the input pulse-train. Moreover, clear experimental evidence of short-term plasticity (STP) and long-term plasticity (LTP) in a single memristor was also demonstrated. Increasing the pulse amplitude or width, or decreasing the interval of two adjacent pulses of the input pulse-train resulted in the memristor behavior transitioning from STP to LTP. The realization of those important synaptic functions in the Ti/HfO₂/TiO_x/Pt memristor may be suitable for applications in artificial neural systems.

Neuromorphic computing with multi-memristive synapses

Neuromorphic computing has emerged as a promising avenue towards building the next generation of intelligent computing systems. It has been proposed that memristive devices, which exhibit history-dependent conductivity modulation, could efficiently represent the synaptic weights in artificial neural networks. However, precise modulation of the device conductance over a wide dynamic range, necessary to maintain high network accuracy, is proving to be challenging. To address this, we present a multi-memristive synaptic architecture with an efficient global counter-based arbitration scheme. We focus on phase change memory devices, develop a comprehensive model and demonstrate via simulations the effectiveness of the concept for both spiking and non-spiking neural networks. Moreover, we present experimental results involving over a million phase change memory devices for unsupervised learning of temporal correlations using a spiking neural network. The work presents a significant step towards the realization of large-scale and energy-efficient neuromorphic computing systems.

Memristive technology is a promising avenue towards realizing efficient non-von Neumann neuromorphic hardware. Boybat et al. proposes a multi-memristive synaptic architecture with a counter-based global arbitration scheme to address challenges associated with the non-ideal memristive device behavior.

Effect of conductance linearity and multi-level cell characteristics of TaOx-based synapse device on pattern recognition accuracy of neuromorphic system

To improve the classification accuracy of an image data set (CIFAR-10) by using analog input voltage, synapse devices with excellent conductance linearity (CL) and multi-level cell (MLC) characteristics are required. We analyze the CL and MLC characteristics of TaO_x-based filamentary resistive random access memory (RRAM) to implement the synapse device in neural network hardware. Our findings show that the number of oxygen vacancies in the filament constriction region of the RRAM directly controls the CL and MLC characteristics. By adopting a Ta electrode (instead of Ti) and the hot-forming step, we could form a dense conductive filament. As a result, a wide range of conductance levels with CL is achieved and significantly improved image classification accuracy is confirmed.

Double MgO-based Perpendicular Magnetic-Tunnel-Junction Spin-valve Structure with a Top Co2Fe6B2 Free Layer using a Single SyAF [Co/Pt]n Layer

A new perpendicular spin-transfer-torque magnetic-tunnel-junction (p-MTJ) spin-valve was developed to achieve a high tunneling magnetoresistance (TMR) ratio. It had a double MgO-based spin-valve structure with a top Co₂Fe₆B₂ free layer and incorporated a single SyAF [Co(0.4 nm)/Pt(0.3 nm)]₃ layer and a new buffer layer of Co(0.6)/Pt(0.3)/Co(0.4). It had a TMR ratio of 180% and anisotropy exchange field (H _ex ) of 3.44 kOe after ex-situ annealing of 350 °C for 30 min under a vacuum below 10^-6 torr and a perpendicular magnetic field of 3 tesla, thereby ensuring a memory margin and avoiding read disturbance failures. Its high level of performance was due to the face-center-cubic crystallinity of the MgO tunneling barrier being significantly improved by decreasing its surface roughness (i.e., peak-to-valley length of 1.4 nm).