Brain-inspired Neuromorphic Systems Research Articles

A Self-Driven Ga2O3 Memristor Synapse for Humanoid Robot Learning.

In recent years, the rapid development of brain-inspired neuromorphic systems has created an imperative demand for artificial photonic synapses that operate with low power consumption. In this study, a self-driven memristor synapse based on gallium oxide (Ga2O3) nanowires is proposed and demonstrated successfully. This memristor synapse is capable of emulating a range of functionalities of biological synapses when exposed to 255nm light stimulation. These functionalities encompass peak time-dependent plasticity, pulse facilitation, and memory learning capabilities. It exhibits an ultrahigh paired-pulse facilitation index of 158, indicating exceptional learning performance. The transition from short-term memory to long-term memory can be attributed to the remarkable relearning capabilities. Furthermore, the potential applications of the memristor synapse is showcased through the successful manipulation of a humanoid intelligent robot. Upon establishing artificial intelligence (AI) systems, the control commands originating from the synaptic device can drive the humanoid robot to perform various actions. Based on the memristor synapses, the autonomous feedback system of the humanoid robot facilitates a good collaboration between robotic actions and bio-inspired light perception. Therefore, this research opens up an effective way to advance the development of neuromorphic computing technologies, AI systems, and intelligent robots that demand ultra-low energy consumption.

New-Style Logic Operation and Neuromorphic Computing Enabled by Optoelectronic Artificial Synapses in an MXene/Y:HfO2 Ferroelectric Memristor.

Today's computing systems, to meet the enormous demands of information processing, have driven the development of brain-inspired neuromorphic systems. However, there are relatively few optoelectronic devices in most brain-inspired neuromorphic systems that can simultaneously regulate the conductivity through both optical and electrical signals. In this work, the Au/MXene/Y:HfO2/FTO ferroelectric memristor as an optoelectronic artificial synaptic device exhibited both digital and analog resistance switching (RS) behaviors under different voltages with a good switching ratio (>103). Under optoelectronic conditions, optimal weight update parameters and an enhanced algorithm achieved 97.1% recognition accuracy in convolutional neural networks. A new logic gate circuit specifically designed for optoelectronic inputs was established. Furthermore, the device integrates the impact of relative humidity to develop an innovative three-person voting mechanism with a veto power. These results provide a feasible approach for integrating optoelectronic artificial synapses with logic-based computing devices.

Multilevel resistive switching in hydrothermally synthesized FeWO4 thin film-based memristive device for non-volatile memory application

The memristors offer significant advantages as a key element in non-volatile and brain-inspired neuromorphic systems because of their salient features such as remarkable endurance, ability to store multiple bits, fast operation speed, and extremely low energy usage. This work reports the resistive switching (RS) characteristics of the hydrothermally synthesized iron tungstate (FeWO4) based thin film memristive device. The detailed physicochemical analysis was investigated using Rietveld’s refinement, X-ray photoelectron spectroscopy (XPS), field emission scanning electron microscopy (FE-SEM), and transmission electron microscopy (TEM) techniques. The fabricated Ag/FWO/FTO memristive device exhibits bipolar resistive switching (BRS) behavior. In addition, the devices exhibit negative differential resistance (NDR) at both positive and negative bias. The charge-flux relation portrayed the non-ideal or memristive nature of the devices. The reliability in the RS process was analyzed in detail using Weibull distribution and time series analysis techniques. The device exhibits stable and multilevel endurance and retention characteristics which demonstrates the suitability of the device for the high-density non-volatile memory application. The current conduction of the device was dominated by Ohmic and trap controlled-space charge limited current (TC-SCLC) mechanisms and filamentary RS process responsible for the BRS in the device. In a nutshell, the present investigations reveal the potential use of the iron tungstate for the fabrication of memristive devices for the non-volatile memory application.

Transparent artificial synapses based on Ag/Al-doped ZnO/ITO memristors for bioinspired neuromorphic computing

Today's computing systems to meet the enormous demands of information processing have driven the development of brain-inspired neuromorphic systems that can perform fast and energy-efficient calculations. In this study, Al-doped ZnO (AZO: ZnO to Al2O3 = 98:2 wt%) films with over 95 % light transmission were prepared by RF magnetron sputtering, and a structurally simple Ag/AZO/ITO memristor was fabricated. Moreover, the sustained potentiation/depression behavior of the memristor allows it to mimic artificial synapses that exhibit biological synaptic and neuronal functions such as short-term/long-term plasticity and paired-pulse facilitation.These results will contribute to the research and development of high-performance artificial synapses.

Artificial Spiking Neuron with Bursting Dynamics for Noise-Resistant Neuromorphic Coding

Emulating complex neuronal functionalities in human brains using emerging biomimetic electronics is critical for next-generation brain-like chips and artificial general intelligence. Burst firing, featured with the repeated firing of discrete groups of spikes, is high-order neuronal dynamics for reliable signal transmission, coding, and processing. Here, we report a memristor (Pt/Co3O4–x/ITO) that can natively produce periodic voltage oscillation groups under current stimulation, capturing the burst-firing features of actual biological neurons. This behavior stems from the dynamic formation/rupture of conductive filaments regulated by the VO concentration at the grain boundaries. The burst frequency can be tuned by adjusting the driving current intensity and used to implement burst frequency coding. The bursting neuron-based artificial neural system achieves a recognition accuracy of 86% for the Fashion-MNIST tasks, under the disturbance of random noise spikes. This work provides a promising platform to realize complex neuronal dynamics and to construct robust brain-inspired neuromorphic systems.

Preventing Vanishing Gradient Problem of Hardware Neuromorphic System by Implementing Imidazole-Based Memristive ReLU Activation Neuron.

With advances in artificial intelligent services, brain-inspired neuromorphic systems with synaptic devices have recently attracted significant interest to circumvent the von Neumann bottleneck. However, the increasing trend of deep neural network parameters causes huge power consumption and large area overhead of a nonlinear neuron electronic circuit, and it incurs a vanishing gradient problem. Here, we present a memristor-based compact and energy-efficient neuron device to implement a rectifying linear unit (ReLU) activation function. To emulate the volatile and gradual switching of the ReLU function, we propose a copolymer memristor with a hybrid structure using a copolymer/inorganic bilayer. The functional copolymer film developed by introducing imidazole functional groups enables the formation of nanocluster-type pseudo-conductive filaments by boosting the nucleation of Cu nanoclusters, causing gradual switching. The ReLU neuron device is successfully demonstrated by integrating the memristor with an amorphous InGaZnO thin-film transistors, and achieves 0.5 pJ of energy consumption based on sub-10μA operation current and high-speed switching of 650ns. Furthermore, device-to-system-level simulation using neuron devices on the MNIST dataset demonstrates that the vanishing gradient problem is effectively resolved by five-layer deep neural networks. The proposed neuron device will enable the implementation of high-density and energy-efficient hardware neuromorphic systems. This article is protected by copyright. All rights reserved.

Demonstration of electronic synapses using a sericin-based bio-memristor

The bio-memristor based on biomaterial has ushered in enthusiasm and optimism in brain-inspired computing systems. Here, the bio-memristor based on sericin has been fabricated with the structure of Ag/sericin/W. The sericin-based bio-memristors demonstrated threshold-switching behavior with low set voltage (∼0.25 V), good cycle-to-cycle uniformity (∼400 cycles), and a large switching window (>100). Interestingly, the device conductance was tuned gradually by the modulation of voltage pulses (amplitude, number, and frequency). The synaptic behaviors can be mimicked, i.e., short-term plasticity, spike-rate-dependent plasticity, and spike-timing-dependent plasticity. This work may open new avenues of bio-memristors in brain-inspired neuromorphic systems.

Versatile GeS-based CBRAM with compliance-current-controlled threshold and bipolar resistive switching for electronic synapses

Chalcogenide materials have promising physical and electrical characteristics for use in advanced memory and electronic synaptic devices. However, limited research has been conducted on chalcogenide materials for highly advanced and scalable memory devices. In this study, we investigate compliance current (CC)-controlled resistive switching (RS) in a monochalcogenide GeS-based conductive bridge random access memory (CBRAM) device for use in advanced memory and electronic synapses. A CBRAM device with a structure stacking sequence of Ag/GeS/Pt/Ti/SiO2 is fabricated by depositing a 13 nm GeS layer using a sputtering process. The GeS-based CBRAM device exhibits the reversible CC-controlled transition between threshold resistive switching (T-RS) and bipolar resistive switching (B-RS). Under a high CC (1 mA and above), the device exhibits B-RS behavior with excellent retention and highly reproducible endurance characteristics, illustrated by a very high ION/IOFF ratio of almost ∼2 × 108. Under a low CC (100 µA and below), the device exhibits T-RS, with the automatic transition from a low resistance state (LRS) to a high resistance state (HRS) when sweeping the voltage in the reverse direction. Under all CC conditions, the device has a very low HRS current of around ∼1 × 10−12 A. A model based on the formation and rupture of conductive filaments (CFs) is proposed to explain the coexistence of T-RS and B-RS in GeS-based CBRAM devices. In particular, the change in the size and geometry of the CFs in accordance with the applied CC is speculated to be the main reason for the coexistence of T-RS and B-RS. The retention characteristics of the device in both T-RS and B-RS mode are investigated for possible application to case-sensitive hardware-based data security and the disposal of trusted and untrusted information. The ability of the CBRAM device in T-RS mode to perform important biological synaptic functions for potential use in brain-inspired neuromorphic systems is also investigated.

Interface state-dependent synaptic characteristics of Pt/CeO2/Pt memristors controlled by post-deposition annealing

Artificial synapses with analog resistance changes have been actively investigated since they are essential elements in brain-inspired neuromorphic systems. In this study, interface state-dependent artificial synaptic characteristics of analog, linear, and symmetric synaptic weight update are demonstrated in Pt/CeO2/Pt memristors controlled by post-deposition annealing of CeO2 layer. The devices with optimized annealing temperature, i.e., 400 and 500 °C-annealed devices, show reliable and analog-type resistance decrease upon applying positive voltage, and vice versa upon applying negative voltage. The resistance change emulates wide-range, linear, and symmetric potentiation and depression behaviors, as being tunable by the number of pulses and its amplitude. It also exhibits short-term and long-term plasticity with respect to pulse conditions. The conduction modes are fitted well with Schottky conduction controlled by interface energy barrier. In addition, negligible resistance changes are observed in reference device with SiO2 interfacial layers inserted at top and bottom Pt/CeO2 interfaces, i.e., Pt/SiO2/CeO2/SiO2/Pt device, supporting that the resistance change is associated with tunable Pt/CeO2 interface states. It is concluded that the resistance change is induced by voltage-driven tuning of Pt/CeO2 interface states, and the control of these interfaces is crucial for reliable synaptic operations.

Study of synaptic properties of honey thin film for neuromorphic systems

Besides conventional electrical properties, materials for emerging artificial synaptic devices in brain-inspired neuromorphic systems need to have properties analogous to those of biological synapses, such as synaptic plasticity. In this work, we investigated, for the first time, the synaptic properties of a natural bio-organic material – honey. The honey film was sandwiched between Ag and ITO to form a structure resembling biological synapse. Neural facilitation, an essential synaptic plasticity, was demonstrated by honey thin film with a larger facilitation index, wider interval time range and higher dynamic filtering gain than other natural bio-organic materials. The results indicate that honey is promising for renewable artificial synaptic devices.

Solution-Processed, Electrolyte-Gated In2O3 Flexible Synaptic Transistors for Brain-Inspired Neuromorphic Applications.

Emulating the essential synaptic behaviors using single synaptic transistor has attracted extensive attention for building the brain-inspired neuromorphic systems. However, few reports on synaptic transistors fabricated by solution processes have been reported. In this article, the indium oxide synaptic transistors based on polyimide substrates were fabricated by a nontoxic water-inducement method at a low temperature, and lithium perchlorate (LiClO4) was dissolved in polyethylene oxide as the gate electrolyte. For water-inducement process, comparable electrical properties of the synaptic transistors can be achieved by prolonging the annealing time rather than high-temperature annealing with a relatively short time. The effect of the annealing time on the electrical performance of the electrolyte-gated transistors annealed at various temperatures was investigated. It is found that the electrolyte-gated-synaptic transistor on polyimide substrate annealed at 200 °C exhibits high electrical performance and good mechanical stability. Due to the ion migration relaxation dynamics in the polymer electrolyte, various important synaptic behaviors such as the excitatory postsynaptic current, paired-pulse facilitation, high-pass filtering characteristics, and long-term memory performance were successfully mimicked. The electrolyte-gated synaptic transistors based on solution-processed In2O3 exhibit great potential in neuromorphological applications.

Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network

Brain-inspired neuromorphic systems (hardware neural networks) are expected to be an energy-efficient computing architecture for solving cognitive tasks, which critically depend on the development of reliable synaptic weight storage (i.e., synaptic device). Although various nanoelectronic devices have successfully reproduced the learning rules of biological synapses through their internal analog conductance states, the sustainability of such devices is still in doubt due to the variability common to all nanoelectronic devices. Alternatively, a neuromorphic system based on a relatively more reliable digital-type switching device has been recently demonstrated, i.e., a binarized neural network (BNN). The synaptic device is a more mature digital-type switching device, and the training/recognition algorithm developed for the BNN enables the task of facial image classification with a supervised training scheme. Here, we quantitatively investigate the effects of device parameter variations on the classification accuracy; the parameters include the number of weight states (Nstate), the weight update margin (ΔG), and the weight update variation (Gvar). This analysis demonstrates the feasibility of the BNN and introduces a practical neuromorphic system based on mature, conventional digital device technologies.

Realization of Artificial Neuron Using MXene Bi-Directional Threshold Switching Memristors

Artificial neurons and synapses are critical units for processing intricate information in brain-inspired neuromorphic systems. Memristors are frequently engineered as artificial synapses due to their simple structures, nonlinear dynamics, and high-density integration. However, the development of artificial neurons on memristors has less progress. In this letter, we propose a rich dynamics-driven artificial neuron based on two-dimensional materials MXene. Partial essential neural features of neural processing, including leaky integration, automatic threshold-driven fire, and self-recovery, were successfully emulated in a unified manner. The space-charge-limited current (SCLC) model accompanied by electrochemical metallization effect was used to explain electrical characteristics. This work will provide a useful guideline for designing and manipulating memristor as artificial neurons for brain-inspired systems.

Optoelectronic Properties of Printed Photogating Carbon Nanotube Thin Film Transistors and Their Application for Light-Stimulated Neuromorphic Devices.

Artificial synapses/neurons based on electronic/ionic hybrid devices have attracted wide attention for brain-inspired neuromorphic systems since it is possible to overcome the von Neumann bottleneck of the neuromorphic computing paradigm. Here, we report a novel photoneuromorphic device based on printed photogating single-walled carbon nanotube (SWCNT) thin film transistors (TFTs) using lightly n-doped Si as the gate electrode. The drain currents of the printed SWCNT TFTs can gradually increase to over 3000 times of their starting value after being pulsed with light stimulation, and the electrical signals can maintain for over 10 min. These characteristics are similar to the learning and memory functions of brain-inspired neuromorphic systems. The working mechanism of the light-stimulated neuromorphic devices is investigated and described here in detail. Important synaptic characteristics, such as low-pass filtering characteristics and nonvolatile memory ability, are successfully emulated in the printed light-stimulated artificial synapses. It demonstrates that the printed SWCNT TFT photoneuromorphic devices can act as the nonvolatile memory units and perform photoneuromorphic computing, which exhibits potential for future neuromorphic system applications.

Ultralow-voltage albumen-gated electric-double-layer thin film transistors

In recent years, environment-friendly and biocompatible electronics have received extensive attention. As a kind of natural biological material with rich sources, proteins have been widely used in electronic devices. In this work, electric-double-layer (EDL) thin-film transistors (TFTs) gated by natural chicken albumen are fabricated at room temperature. The indium-tin-oxide (ITO) conductive glass is employed as a substrate. The spin coated chicken albumen film is used as the gate dielectric. The indium-zinc-oxide (IZO) is sputtered on an albumen-coated ITO glass as the channel and the source/drain electrodes with only one shadow mask. The capacitance-frequency measurements demonstrate an ultra-large specific capacitance of the albumen film at low frequencies. For the physical understanding of the capacitive coupling within the albumen film, the phase angle is characterized as a function of frequency. The results indicate that such an ultra-large capacitive coupling can be attributed to the proton migration under the electric field, which results in the EDL effect at the interface of the albumen film. By DC sweep measurements, a low leakage current is observed (<3.0 nA at <i>V</i><sub>gs</sub>=1.5 V), which indicates a good isolation of the albumen-based dielectric. By transfer and output measurements, an ultralow operation voltage of 1.5 V, a high field-effect mobility of 38.01 cm<sup>2</sup>/(V·s), a low subthreshold swing of 164 mV/decade, and a large on-off ratio of 2.4×10<sup>6</sup> are obtained for such albumen-gated TFTs. The ultra-large EDL capacitive coupling is responsible for such good electrical characteristics. The dynamic bias stress stability of the albumen-gated TFTs is also investigated. The device exhibits a good reproducibility in response to the repeatedly pulsed gate voltage. A maintainable on-to-off ratio (>10<sup>6</sup>) and no obvious current loss are observed, which suggests that neither chemical doping nor chemical reaction occurs at the albumen-based dielectric/IZO channel interface when the gate potential is biased. After being aged one day in air ambient without surface passivation, the albumen-gated TFTs show a good stability of the electrical properties. Such ultralow-voltage EDL-TFTs gated by albumen electrolyte will be useful for the bioelectronic and low-energy portable electronic products. And our results will also have potential applications in biocompatible artificial neuron networks and brain-inspired neuromorphic systems.

Silicon synaptic transistor for hardware-based spiking neural network and neuromorphic system

Brain-inspired neuromorphic systems have attracted much attention as new computing paradigms for power-efficient computation. Here, we report a silicon synaptic transistor with two electrically independent gates to realize a hardware-based neural network system without any switching components. The spike-timing dependent plasticity characteristics of the synaptic devices are measured and analyzed. With the help of the device model based on the measured data, the pattern recognition capability of the hardware-based spiking neural network systems is demonstrated using the modified national institute of standards and technology handwritten dataset. By comparing systems with and without inhibitory synapse part, it is confirmed that the inhibitory synapse part is an essential element in obtaining effective and high pattern classification capability.

Mixed protonic and electronic conductors hybrid oxide synaptic transistors

Mixed ionic and electronic conductor hybrid devices have attracted widespread attention in the field of brain-inspired neuromorphic systems. Here, mixed protonic and electronic conductor (MPEC) hybrid indium-tungsten-oxide (IWO) synaptic transistors gated by nanogranular phosphorosilicate glass (PSG) based electrolytes were obtained. Unique field-configurable proton self-modulation behaviors were observed on the MPEC hybrid transistor with extremely strong interfacial electric-double-layer effects. Temporally coupled synaptic plasticities were demonstrated on the MPEC hybrid IWO synaptic transistor, including depolarization/hyperpolarization, synaptic facilitation and depression, facilitation-stead/depression-stead behaviors, spiking rate dependent plasticity, and high-pass/low-pass synaptic filtering behaviors. MPEC hybrid synaptic transistors may find potential applications in neuron-inspired platforms.

Oxide-based synaptic transistors gated by solution-processed gelatin electrolytes

In human brain, a large number of neurons are connected via synapses. Simulation of the synaptic behaviors using electronic devices is the most important step for neuromorphic systems. In this paper, proton conducting gelatin electrolyte-gated oxide field-effect transistors (FETs) were used for emulating synaptic functions, in which the gate electrode is regarded as pre-synaptic neuron and the channel layer as the post-synaptic neuron. In analogy to the biological synapse, a potential spike can be applied at the gate electrode and trigger ionic motion in the gelatin electrolyte, which in turn generates excitatory post-synaptic current (EPSC) in the channel layer. Basic synaptic behaviors including spike time-dependent EPSC, paired-pulse facilitation (PPF), self-adaptation, and frequency-dependent synaptic transmission were successfully mimicked. Such ionic/electronic hybrid devices are beneficial for synaptic electronics and brain-inspired neuromorphic systems.

Recent Progress in Phase-ChangeMemory Technology

We survey progress in the PCM field over the past five years, ranging from large-scale PCM demonstrations to materials improvements for high–temperature retention and faster switching. Both materials and new cell designs that support lower-power switching are discussed, as well as higher reliability for long cycling endurance. Two paths towards higher density are discussed: through 3D integration by the combination of PCM and 3D-capable access devices, and through multiple bits per cell, by understanding and managing resistance drift caused by structural relaxation of the amorphous phase. We also briefly survey work in the nascent field of brain-inspired neuromorphic systems that use PCM to implement non-Von Neumann computing.

Short-Term Plasticity and Synaptic Filtering Emulated in Electrolyte-Gated IGZO Transistors

Artificial synaptic devices have attracted a broad interest for hardware implementation of brain-inspired neuromorphic systems. In this letter, a short-term plasticity simulation in an indium–gallium–zinc oxide (IGZO) electric-double-layer (EDL) transistor is investigated. For synaptic facilitation and depression function emulation, three-terminal EDL transistor is reduced to a two-terminal synaptic device with two modified connection schemes. Furthermore, high-pass and low-pass filtering characteristics are also successfully emulated not only for fixed-rate spike train but also for Poisson-like spike train. Our results suggest that IGZO-based EDL transistors operated in two terminal mode can be used as the building blocks for brain-like chips and neuromorphic systems.