Application of artificial synapse based on all-inorganic perovskite memristor in neuromorphic computing

Artificial synapse inspired by the biological brain has great potential in the field of neuromorphic computing and artificial intelligence.The memristor is an ideal artificial synaptic device with fast operation and good tolerance.Here,we have prepared a memristor device with Au/CsPbBr_(3)/ITO structure.The memristor device exhibits resistance switching behavior,the high and low resistance states no obvious decline after 400 switching times.The memristor device is stimulated by voltage pulses to simulate biological synaptic plasticity,such as long-term potentiation,long-term depression,pair-pulse facilitation,short-term depression,and short-term potentiation.The transformation from short-term memory to long-term memory is achieved by changing the stimulation frequency.In addition,a convolutional neural network was constructed to train/recognize MNIST handwritten data sets;a distinguished recognition accuracy of~96.7%on the digital image was obtained in 100 epochs,which is more accurate than other memristor-based neural networks.These results show that the memristor device based on CsPbBr3 has immense potential in the neuromorphic computing system.

Two-Terminal Lithium-Mediated Artificial Synapses with Enhanced Weight Modulation for Feasible Hardware Neural Networks

Recently,artificial synapses involving an electrochemical reaction of Li-ion have been attributed to have remarkable synaptic properties.Three-terminal synaptic transistors utilizing Li-ion intercalation exhibits reliable synaptic characteristics by exploiting the advantage of nondistributed weight updates owing to stable ion migrations.However,the three-terminal configurations with large and complex structures impede the crossbar array implementation required for hardware neuromorphic systems.Meanwhile,achieving adequate synaptic performances through effective Li-ion intercalation in vertical two-terminal synaptic devices for array integration remains challenging.Here,two-terminal Au/LixCoO_(2)/Pt artificial synapses are proposed with the potential for practical implementation of hardware neural networks.The Au/LixCoO_(2)/Pt devices demonstrated extraordinary neuromorphic behaviors based on a progressive dearth of Li in LixCoO_(2)films.The intercalation and deintercalation of Li-ion inside the films are precisely controlled over the weight control spike,resulting in improved weight control functionality.Various types of synaptic plasticity were imitated and assessed in terms of key factors such as nonlinearity,symmetricity,and dynamic range.Notably,the LixCoO_(2)-based neuromorphic system outperformed three-terminal synaptic transistors in simulations of convolutional neural networks and multilayer perceptrons due to the high linearity and low programming error.These impressive performances suggest the vertical two-terminal Au/LixCoO_(2)/Pt artificial synapses as promising candidates for hardware neural networks.

A Flexible Tribotronic Artificial Synapse with Bioinspired Neurosensory Behavior

As key components of artificial afferent nervous systems,synaptic devices can mimic the physiological synaptic behaviors,which have attracted extensive attentions.Here,a flexible tribotronic artificial synapse(TAS)with bioinspired neurosensory behavior is developed.The triboelectric potential generated by the external contact electrification is used as the ion-gel-gate voltage of the organic thin film transistor,which can tune the carriers transport through the migration/accumulation of ions.The TAS successfully demonstrates a series of synaptic behaviors by external stimuli,such as excitatory postsynaptic current,paired-pulse facilitation,and the hierarchical memory process from sensory memory to short-term memory and long-term memory.Moreover,the synaptic behaviors remained stable under the strain condition with a bending radius of 20 mm,and the TAS still exhibits excellent durability after 1000 bending cycles.Finally,Pavlovian conditioning has been successfully mimicked by applying force and vibration as food and bell,respectively.This work demonstrates a bioinspired flexible artificial synapse that will help to facilitate the development of artificial afferent nervous systems,which is great significance to the practical application of artificial limbs,robotics,and bionics in future.

An artificial synapse by superlattice-like phase-change material for low-power brain-inspired computing

Phase-change material(PCM)is generating widespread interest as a new candidate for artificial synapses in bioinspired computer systems.However,the amorphization process of PCM devices tends to be abrupt,unlike continuous synaptic depression.The relatively large power consumption and poor analog behavior of PCM devices greatly limit their applications.Here,we fabricate a GeTe/Sb2Te3 superlattice-like PCM device which allows a progressive RESET process.Our devices feature low-power consumption operation and potential high-density integration,which can effectively simulate biological synaptic characteristics.The programming energy can be further reduced by properly selecting the resistance range and operating method.The fabricated devices are implemented in both artificial neural networks(ANN)and convolutional neural network(CNN)simulations,demonstrating high accuracy in brain-like pattern recognition.

Artificial synapses based on organic electrochemical transistors with self-healing dielectric layers

Organic electrochemical transistors(OECTs)have emerged as one type of promising building block for neuromorphic systems owing to their capability of mimicking the morphology and functions of biological neurons and synapses.Currently,numerous kinds of OECTs have been developed,while self-healing performance has been neglected in most reported OECTs.In this work,the OECTs using self-healing polymer electrolytes as dielectric layers are proposed.Several important synaptic behaviors are simulated in the OECTs by doping the channel layers with ions from the electrolytes.Benefitting from the dynamic hydrogen bonds in the self-healing polymer electrolytes,the OECTs can successfully maintain their electrical performance and the ability of emulating synaptic behaviors after self-healing compared with the initial state.More significantly,the sublinear spatial summation function is demonstrated in the OECTs and their potential in flexible electronics is also validated.These results suggest that our devices are expected to be a vital component in the development of future wearable and bioimplantable neuromorphic systems.

Ultrasound:A new strategy for artificial synapses modulation

Due to its non-invasive nature,ultrasound has been widely used for neuromodulation in biological systems,where its application influences the synaptic weights and the process of neurotransmitter delivery.However,such modulation has not been emulated in physical devices.Memristors are ideal electrical components for artificial synapses,but up till now they are hardly reported to respond to ultrasound signals.Here we design and fabricate a HfOx-based memristor on 64Y-X LiNbO_(3) single crystal substrate,and successfully realize artificial synapses modulation by shear-horizontal surface acoustic wave(SH-SAW).It is a prominent short-term resistance modulation,where ultrasound has been shown to cause resistance drop for various resistance states,which could fully recover after the ultrasound is shut off.The physical mechanism illustrates that ultrasound induced polarization potential in the HfOx dielectric layer acts on the Schottky barrier,leading to the resistance drop.The emulation of neuron firing frequency modulation through ultrasound signals is demonstrated.Moreover,the joint application of ultrasound and electric voltage yields fruitful functionalities,such as the enhancement of resistance window and synaptic plasticity through ultrasound application.All these promising results provide a new strategy for artificial synapses modulation,and also further advance neuromorphic devices toward system applications.

Band-tailored van der Waals heterostructure for multilevel memory and artificial synapse

Two-dimensional(2D)van der Waals heterostructure(vdWH)-based floating gate devices show great potential for next-generation nonvolatile and multilevel data storage memory.However,high program voltage induced substantial energy consumption,which is one of the primary concerns,hinders their applications in lowenergy-consumption artificial synapses for neuromorphic computing.In this study,we demonstrate a three-terminal floating gate device based on the vdWH of tin disulfide(SnS2),hexagonal boron nitride(h-BN),and few-layer graphene.The large electron affinity of SnS2 facilitates a significant reduction in the program voltage of the device by lowering the hole-injection barrier across h-BN.Our floating gate device,as a nonvolatile multilevel electronic memory,exhibits large on/off current ratio(105),good retention(over 104 s),and robust endurance(over 1000 cycles).Moreover,it can function as an artificial synapse to emulate basic synaptic functions.Further,low energy consumption down to7 picojoule(pJ)can be achieved owing to the small program voltage.High linearity(<1)and conductance ratio(80)in long-term potentiation and depression(LTP/LTD)further contribute to the high pattern recognition accuracy(90%)in artificial neural network simulation.The proposed device with attentive band engineering can promote the future development of energy-efficient memory and neuromorphic devices.

Tactile tribotronic reconfigurable p-n junctions for artificial synapses

The emulation of biological synapses with learning and memory functions and versatile plasticity is significantly promising for neuromorphic computing systems.Here,a robust and continuously adjustable mechanoplastic semifloating-gate transistor is demonstrated based on an integrated graphene/hexagonal boron nitride/tungsten diselenide van der Waals heterostructure and a triboelectric nanogenerator(TENG).The working states(p-n junction or n;-n junction)can be manipulated and switched under the sophisticated modulation of triboelectric potential derived from mechanical actions,which is attributed to carriers trapping and detrapping in the graphene layer.Furthermore,a reconfigurable artificial synapse is constructed based on such mechanoplastic transistor that can simulate typical synaptic plasticity and implement dynamic control correlations in each response mode by further designing the amplitude and duration.The artificial synapse can work with ultra-low energy consumption at 74.2 f J per synaptic event and the extended synaptic weights.Under the synergetic effect of the semifloating gate,the synaptic device can enable successive mechanical facilitation/depression,short-/long-term plasticity and learning-experience behavior,exhibiting the mechanical behavior derived synaptic plasticity.Such reconfigurable and mechanoplastic features provide an insight into the applications of energyefficient and real-time interactive neuromodulation in the future artificial intelligent system beyond von Neumann architecture.

Flexible artificial synapse based on single-crystalline BiFeO_(3) thin film

Realization of functional flexible artificial synapse is a significant step toward neuromorphic computing.Herein,a flexible artificial synapse based on ferroelectric tunnel junctions(FTJs)is demonstrated,using BiFeO_(3)(BFO)thin film as the functional layer.The inorganic single crystalline FTJs grown on rigid perovskite substrates at high temperatures are integrated with the flexible plastic substrates,by using the water-soluble Sr_(3)Al_(2)O_(6)(SAO)as the sacrificial layer and the following transfer.The transferred freestanding BFO thin film exhibits excellent ferroelectric properties.Moreover,the memristive properties and the brain-like synaptic learning performance of the flexible FTJs are investigated.The results show that multilevel resistance states were maintained well of the flexible artificial synapse,together with their stable synaptic learning properties.Our work indicates the promising opportunity of ferroelectric thin film based flexible synapse used in the future neuromorphic computing system.

Homologous gradient heterostructure-based artificial synapses for neuromorphic computation

Gradient heterostructure is one of fundamental interfaces and provides an effective platform to achieve gradually changed properties in mechanics,optics,and electronics.Among different types of heterostructures,the gradient one may provide multiple resistive states and immobilized conductive fila-ments,offering great prospect for fabricating memristors with both high neuromorphic computation capability and repeatability.Here,we invent a memristor based on a homologous gradient heterostructure(HGHS),compris-ing a conductive transition metal dichalcogenide and an insulating homolo-gous metal oxide.Memristor made of Ta–TaS_(x)O_(y)–TaS 2 HGHS exhibits continuous potentiation/depression behavior and repeatable forward/backward scanning in the read-voltage range,which are dominated by multi-ple resistive states and immobilized conductive filaments in HGHS,respec-tively.Moreover,the continuous potentiation/depression behavior makes the memristor serve as a synapse,featuring broad-frequency response(10^(-1)–10^(5) Hz,covering 106 frequency range)and multiple-mode learning(enhanced,depressed,and random-level modes)based on its natural and moti-vated forgetting behaviors.Such HGHS-based memristor also shows good unifor-mity for 5?7 device arrays.Our work paves a way to achieve high-performance integrated memristors for future artificial neuromorphic computation.

Forming-free artificial synapses with Ag point contacts at interface

Ag/Ta_(2)O_(5)/CuO/Pt memristive devices with Ag point contacts at the interface exhibit forming-free and partial volatile analog resistive switching properties.Versatile synaptic functions,like the short-term plasticity,the long-term potentiation and the paired-pulse facilitation,are emulated with these devices.The Ag point contacts in the Ta_(2)O_(5)layer are verified through transmission electron microscope(TEM)and X-ray photoelectron spectroscope(XPS).The Ag point contacts at the interface endow the device the transition from the electrochemical metallization mode to the valence change mode,and the analog resistive switching behavior and neuromorphic functions.This interface engineering of introducing point contacts at the interface provides a way for the development of neuromorphic devices with low power consumption.

A gate-tunable artificial synapse based on vertically assembled van der Waals ferroelectric heterojunction

Memtransistor,a multi-terminal device that combines both the characteristics of a memristor and a transistor,has been intensively studied in two-dimensional layered materials(2 DLM),which show potential for applications in such as neuromorphic computation.However,while often based on the migration of ions or atomic defects in the conduction channels,performances of memtransistors suffer from the poor reliability and tunability.Furthermore,those known 2 DLM-based memtransistors are mostly constructed in a lateral manner,which hinders the further increasing of the transistor densities per area.Until now,fabricating non-atomic-diffusion based memtransistors with vertical structure remains challenging.Here,we demonstrate a vertically-integrated ferroelectric memristor by hetero-integrating the 2 D ferroelectric materials CuInP_(2)S_(6)(CIPS)into a graphite/CuInP_(2)S_(6)/MoS_(2)vertical heterostructure.Memristive behaviour and multi-level resistance states were realized.Essential synaptic behaviours including excitatory postsynaptic current,paired-pulse-facilitation,and spike-amplitude-dependent plasticity are successfully mimicked.Moreover,by applying a gate potential,the memristive behaviour and synaptic features can be effectively gate tuned.Our findings pave the way for the realization of novel gate-tunable ferroelectric synaptic devices with the capability to perform complex neural functions.

Intrinsic polarization coupling in 2Dα‐In_(2)Se_(3)toward artificial synapse with multimode operations

Emulation of advanced synaptic functions of the human brain with electronic devices contributes an important step toward constructing high‐efficiency neuromorphic systems.Ferroelectric materials are promising candidates as synaptic weight elements in neural network hardware due to their controllable polarization states.However,the increased depolarization field at the na-noscale and the complex fabrication process of the traditional ferroelectric materials hamper the development of high‐density,low‐power,and highly sensitive synaptic devices.Here,we report the implementation of two‐dimensional(2D)ferroelectricα‐In_(2)Se_(3)as an active channel material to emulate typical synaptic functions.Theα‐In_(2)Se_(3)‐based synaptic device fea-tures multimode operations,enabled by the coupled ferroelectric polarization under various voltage pulses applied at both drain and gate terminals.Moreover,the energy consumption can be reduced to~1 pJ by using high‐κdielectric(Al2O3).The successful control of ferroelectric polarizations inα‐In_(2)Se_(3)and its application in artificial synapses are expected to inspire the implementation of 2D ferroelectric materials for future neuromorphic systems.

Humidity-induced synaptic plasticity of ZnO artificial synapses using peptide insulator for neuromorphic computing

Neuromorphic devices inspired by the human brain have attracted significant attention because of their excellent ability for cognitive and parallel computing.This study presents ZnO-based artificial synapses with peptide insulators for the electrical emulation of biological synapses.We demonstrated the dynamic responses of the device under various environmental conditions.The proton-conducting property of the tyrosine-rich peptide enables time-dependent responses under ambient conditions such that various aspects of synaptic behaviors are emulated by the devices.The transition from short-term memory to longterm memory is achieved via electrochemical doping of ZnO by protons.Furthermore,we demonstrate an image classification simulation using a multi-layer perceptron model to evaluate the potential of the device for use in neuromorphic computing.The neural network based on our device achieved a recognition accuracy of 87.47% for the MNIST handwritten digit images.This work proposes a novel device platform inspired by biosystems for brain-mimetic hardware systems.

Low operating voltage memtransistors based on ion bombarded p-type GeSe nanosheets for artificial synapse applications

Two-dimensional(2D)layered materials have many potential applications in memristors owing to their unique atomic structures and electronic properties.Memristors can overcome the in-memory bottleneck for use in brain-like neuromorphic computing.However,exploiting additional lateral memtransistors based on 2D layered materials remains challenging.There are few studies on p-type semiconductors that have not been theoretically analyzed.In this study,a lateral memtransistor based on p-type GeSe nanosheets is investigated.A threeterminal GeSe memtransistor that modulated the interfacial barrier height was fabricated using low-energy ion irradiation;the memtransistor exhibited a low operating voltage.The memtransistor successfully mimics biological synapse,including neuroplasticity functions,such as short-term plasticity,long-term plasticity,paired-pulse facilitation,and spike-timing-dependent plasticity.The mechanism of interfacial modulation was verified by experimental results and theoretical calculations.The results show that it is feasible to modulate the interface of 2D GeSe nanosheets using low-energy ion irradiation to realize a lateral memtransistor.This may provide promising opportunities for artificial neuromorphic system applications based on 2D layered materials.

A light-emitting electrochemical artificial synapse with dual output of photoelectric signals

Despite recent remarkable progress in multiple synaptic devices,searching for artificial synapses with new functions is still an important task in the construction of artificial neural networks.The parallel output functionality of photoelectric signals in artificial synaptic devices is interesting and desirable as on-chip optoelectronic interconnection technology allows the connections between neurons weighted by current and light.In turn,it provides degrees of freedom and reduces circuit lead density in the design of large-scale neural networks.Hence,for the first time,a light-emitting electrochemical artificial synapse(LEEAS)based on poly[2-methoxy-5-(2'-ethylhexyloxy)-1,4-phenylenevinylene]/poly(ethylene oxide)/lithium salt blends with dual output of photoelectric signals was developed in this study.The electrochemical redox reaction enables the device to achieve synaptic plasticity in biology and emulate the memory enhancement process,high-pass filtering characteristic,and classical Pavlov’s conditioned reflex experiment.In addition,the transient luminescence intensity of the LEEAS induced by identical electric spikes exhibits a synaptic-like potentiation behavior.Owing to the combination of electroluminescence(EL)and synaptic memory behavior,an LEEAS array exhibits a unique image display and storage functions that can memorize displayed images.The LEEAS proposed in this work enriches the diversity of artificial synapses,promoting the diversified design and development of next-generation optoelectronic hybrid artificial neural networks.

Van der Waals ferroelectric transistors:the all-round artificial synapses for high-precision neuromorphic computing

State number,operation power,dynamic range and conductance weight update linearity are key synaptic device performance metrics for high-accuracy and low-power-consumption neuromorphic com-puting in hardware.However,high linearity and low power consump-tion couldn’t be simultaneously achieved by most of the reported synaptic devices,which limits the performance of the hardware.This work demonstrates van der Waals(vdW)stacked ferroelectric field-effect transistors(FeFET)with single-crystalline ferroelectric nanoflakes.Ferroelectrics are of fine vdW interface and partial polar-ization switching of multi-domains under electric field pulses,which makes the FeFETs exhibit multi-state memory characteristics and ex-cellent synaptic plasticity.They also exhibit a desired linear conduc-tance weight update with 128 conductance states,a sufficiently high dynamic range of G_(max)/G_(min)>120,and a low power consumption of 10 fJ/spike using identical pulses.Based on such an all-round device,a two-layer artificial neural network was built to conduct Modified Na-tional Institute of Standards and Technology(MNIST)digital num-bers and electrocardiogram(ECG)pattern-recognition simulations,with the high accuracies reaching 97.6%and 92.4%,respectively.The remarkable performance demonstrates that vdW-FeFET is of obvious advantages in high-precision neuromorphic computing applications.

Flexible organic artificial synapse with ultrashort-term plasticity for tunable time-frequency signal processing

A flexible organic artificial synapse(OAS)for tunable time-frequency signal processing was fabricated using a tri-blend film that had been fabricated using a one-step solution method.When combined with a chitosan film,this OAS can achieve an ultrashort-term retention time of only 49 ms for instant electricalcomputing applications;this is the shortest retention time yet achieved by a two-terminal artificial synapse.An array of these flexible OASs can withstand a high bending strain of 5%for 10^(4) cycles;this deformation endurance is a new record.The OAS was also sensitive to the number and frequency of electrical inputs;a tunable cut-off frequency enables dynamic filtering for use in image detail enhancement.This work provides a new resource for development of future neuromorphic computing devices。

Ferroelectric artificial synapse for neuromorphic computing and flexible applications

Research of artificial synapses is increasing in popularity with the development of bioelectronics and the appearance of wearable devices.Because the high-temperature treatment process of inorganic materials is not compatible with flexible substrates,organic ferroelectric materials that are easier to process have emerged as alternatives.An organic synaptic device based on P(VDF-TrFE)was prepared in this study.The device showed reliable P/E endurance over 104 cycles and a data storage retention capability at 80℃ over 104 s.Simultaneously,it possessed excellent synaptic functions,including short-term/long-term synaptic plasticity and spike-timing-dependent plasticity.In addition,the ferroelectric performance of the device remained stable even under bending(7 mm bending radius)or after 500 bending cycles.This work shows that low-temperature processed organic ferroelectric materials can provide new ideas for the future development of wearable electronics and flexible artificial synapses.

A Low-Power Artificial Synapse Could One Day Interface with the Brain

Battery technology inspires a flexible,organic,nonvolatile device for neuromorphic circuits that needs only millivolts to change state.The researchers have created a new form of'artificial synapse'that may one day be used to create flexible circuitry that could directly interface with the brain.