Carbon Nanotube Synaptic Transistor Network for Pattern Recognition

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.

Multi-Layered Triboelectric Nanogenerators with Controllable Multiple Spikes for Low-Power Artificial Synaptic Devices

In the domains of wearable electronics, robotics, and the Internet of Things, there is a demand for devices with low power consumption and the capability of multiplex sensing, memory, and learning. Triboelectric nanogenerators (TENGs) offer remarkable versatility in this regard, particularly when integrated with synaptic transistors that mimic biological synapses. However, conventional TENGs, generating only two spikes per cycle, have limitations when used in synaptic devices requiring repetitive high-frequency gating signals to perform various synaptic plasticity functions. Herein, a multi-layered micropatterned TENG (M-TENG) consisting of a polydimethylsiloxane (PDMS) film and a composite film that includes 1H,1H,2H,2H-perfluorooctyltrichlorosilane/BaTiO3 /PDMS are proposed. The M-TENG generates multiple spikes from a single touch by utilizing separate triboelectric charges at the multiple friction layers, along with a contact/separation delay achieved by distinct spacers between layers. This configuration allows the maximum triboelectric output charge of M-TENG to reach up to 7.52 nC, compared to 3.69 nC for a single-layered TENG. Furthermore, by integrating M-TENGs with an organic electrochemical transistor, the spike number multiplication property of M-TENGs is leveraged to demonstrate an artificial synaptic device with low energy consumption. As a proof-of-concept application, a robotic hand is operated through continuous memory training under repeated stimulations, successfully emulating long-term plasticity.

High-performance asymmetric electrode structured light-stimulated synaptic transistor for artificial neural networks

Photonics neuromorphic computing shows great prospects due to the advantages of low latency, low power consumption and high bandwidth. Transistors with asymmetric electrode structures are receiving increasing attention due to their low power consumption, high optical response, and simple preparation technology. However, intelligent optical synapses constructed by asymmetric electrodes are still lacking systematic research and mechanism analysis. Herein, we present an asymmetric electrode structure of the light-stimulated synaptic transistor (As-LSST) with a bulk heterojunction as the semiconductor layer. The As-LSST exhibits superior electrical properties, photosensitivity and multiple biological synaptic functions, including excitatory postsynaptic currents, paired-pulse facilitation, and long-term memory. Benefitting from the asymmetric electrode configuration, the devices can operate under a very low drain voltage of 1 × 10-7 V, and achieve an ultra-low energy consumption of 2.14 × 10-18 J per light stimulus event. Subsequently, As-LSST implemented the optical logic function and associative learning. Utilizing As-LSST, an artificial neural network (ANN) with ultra-high recognition rate (over 97.5%) of handwritten numbers was constructed. This work presents an easily-accessible concept for future neuromorphic computing and intelligent electronic devices.

Reinforcing Synaptic Plasticity of Defect-Tolerant States in Alloyed 2D Artificial Transistors

While two-dimensional (2D) materials possess the desirable future of neuromorphic computing platforms, unstable charging and de-trapping processes, which are inherited from uncontrollable states, such as the interface trap between nanocrystals and dielectric layers, can deteriorate the synaptic plasticity in field-effect transistors. Here, we report a facile and effective strategy to promote artificial synaptic devices by providing physical doping in 2D transition-metal dichalcogenide nanomaterials. Our experiments demonstrate that the introduction of niobium (Nb) into 2D WSe2 nanomaterials produces charge trap levels in the band gap and retards the decay of the trapped charges, thereby accelerating the artificial synaptic plasticity by encouraging improved short-/long-term plasticity, increased multilevel states, lower power consumption, and better symmetry and asymmetry ratios. Density functional theory calculations also proved that the addition of Nb to 2D WSe2 generates defect tolerance levels, thereby governing the charging and de-trapping mechanisms of the synaptic devices. Physically doped electronic synapses are expected to be a promising strategy for the development of bioinspired artificial electronic devices.

Biomedical application of 2D nanomaterials in neuroscience

Two-dimensional (2D) nanomaterials, such as graphene, black phosphorus and transition metal dichalcogenides, have attracted increasing attention in biology and biomedicine. Their high mechanical stiffness, excellent electrical conductivity, optical transparency, and biocompatibility have led to rapid advances. Neuroscience is a complex field with many challenges, such as nervous system is difficult to repair and regenerate, as well as the early diagnosis and treatment of neurological diseases are also challenged. This review mainly focuses on the application of 2D nanomaterials in neuroscience. Firstly, we introduced various types of 2D nanomaterials. Secondly, due to the repairment and regeneration of nerve is an important problem in the field of neuroscience, we summarized the studies of 2D nanomaterials applied in neural repairment and regeneration based on their unique physicochemical properties and excellent biocompatibility. We also discussed the potential of 2D nanomaterial-based synaptic devices to mimic connections among neurons in the human brain due to their low-power switching capabilities and high mobility of charge carriers. In addition, we also reviewed the potential clinical application of various 2D nanomaterials in diagnosing and treating neurodegenerative diseases, neurological system disorders, as well as glioma. Finally, we discussed the challenge and future directions of 2D nanomaterials in neuroscience.

Recent progress in three-terminal artificial synapses based on 2D materials: from mechanisms to applications

Synapses are essential for the transmission of neural signals. Synaptic plasticity allows for changes in synaptic strength, enabling the brain to learn from experience. With the rapid development of neuromorphic electronics, tremendous efforts have been devoted to designing and fabricating electronic devices that can mimic synapse operating modes. This growing interest in the field will provide unprecedented opportunities for new hardware architectures for artificial intelligence. In this review, we focus on research of three-terminal artificial synapses based on two-dimensional (2D) materials regulated by electrical, optical and mechanical stimulation. In addition, we systematically summarize artificial synapse applications in various sensory systems, including bioplastic bionics, logical transformation, associative learning, image recognition, and multimodal pattern recognition. Finally, the current challenges and future perspectives involving integration, power consumption and functionality are outlined.

Improving the linearity of synaptic plasticity of single-walled carbon nanotube field-effect transistors via CdSe quantum dots decoration

The linearity of synaptic plasticity of single-walled carbon nanotube field-effect transistor (SWCNT FET) was improved by CdSe quantum dots decoration. The linearity of synaptic plasticity in SWCNT FET with decorating QDs was further improved by reducing the P-type doping level from the atmosphere. The synaptic behavior of SWCNT FET is found to be dominated by the charging and discharging processes of interface traps and surface traps, which are predominantly composed of H₂O/O₂redox couples. The improved synaptic behavior is mainly due to the reduction of the interface trap charging process after QDs decoration. The inherent correlation between the device synaptic behavior and the electron capture process of the traps are investigated through charging-based trap characterization. This study provides an effective scheme for improving linearity and designing new-type SWCNT synaptic devices.

Dielectric-Engineered High-Speed, Low-Power, Highly Reliable Charge Trap Flash-Based Synaptic Device for Neuromorphic Computing beyond Inference

The coming of the big-data era brought a need for power-efficient computing that cannot be realized in the Von Neumann architecture. Neuromorphic computing which is motivated by the human brain can greatly reduce power consumption through matrix multiplication, and a device that mimics a human synapse plays an important role. However, many synaptic devices suffer from limited linearity and symmetry without using incremental step pulse programming (ISPP). In this work, we demonstrated a charge-trap flash (CTF)-based synaptic transistor using trap-level engineered Al₂O₃/Ta₂O₅/Al₂O₃ gate stack for successful neuromorphic computing. This novel gate stack provided precise control of the conductance with more than 6 bits. We chose the appropriate bias for highly linear and symmetric modulation of conductance and realized it with very short (25 ns) identical pulses at low voltage, resulting in low power consumption and high reliability. Finally, we achieved high learning accuracy in the training of 60000 MNIST images.

Progress and Challenges for Memtransistors in Neuromorphic Circuits and Systems

Due to the increasing importance of artificial intelligence (AI), significant recent effort has been devoted to the development of neuromorphic circuits that seek to emulate the energy-efficient information processing of the brain. While non-volatile memory (NVM) based on resistive switches, phase-change memory, and magnetic tunnel junctions has shown potential for implementing neural networks, additional multi-terminal device concepts are required for more sophisticated bio-realistic functions. Of particular interest are memtransistors based on low-dimensional nanomaterials, which are capable of electrostatically tuning memory and learning behavior at the device level. Herein, a conceptual overview of the memtransistor is provided in the context of neuromorphic circuits. Recent progress is surveyed for memtransistors and related multi-terminal NVM devices including dual-gated floating-gate memories, dual-gated ferroelectric transistors, and dual-gated van der Waals heterojunctions. The different materials systems and device architectures are classified based on the degree of control and relative tunability of synaptic behavior, with an emphasis on device concepts that harness the reduced dimensionality, weak electrostatic screening, and phase-changes properties of nanomaterials. Finally, strategies for achieving wafer-scale integration of memtransistors and multi-terminal NVM devices are delineated, with specific attention given to the materials challenges for practical neuromorphic circuits.

Programmable ferroelectric bionic vision hardware with selective attention for high-precision image classification

Selective attention is an efficient processing strategy to allocate computational resources for pivotal optical information. However, the hardware implementation of selective visual attention in conventional intelligent system is usually bulky and complex along with high computational cost. Here, programmable ferroelectric bionic vision hardware to emulate the selective attention is proposed. The tunneling effect of photogenerated carriers are controlled by dynamic variation of energy barrier, enabling the modulation of memory strength from 9.1% to 47.1% without peripheral storage unit. The molecular polarization of ferroelectric P(VDF-TrFE) layer enables a single device not only multiple nonvolatile states but also the implementation of selective attention. With these ferroelectric devices are arrayed together, UV light information can be selectively recorded and suppressed the with high current decibel level. Furthermore, the device with positive polarization exhibits high wavelength dependence in the image attention processing, and the fabricated ferroelectric sensory network exhibits high accuracy of 95.7% in the pattern classification for multi-wavelength images. This study can enrich the neuromorphic functions of bioinspired sensing devices and pave the way for profound implications of future bioinspired optoelectronics.

Reconfigurable Artificial Synapses with Excitatory and Inhibitory Response Enabled by an Ambipolar Oxide Thin-Film Transistor

A gate-tunable synaptic device controlling dynamically reconfigurable excitatory and inhibitory synaptic responses, which can emulate the fundamental synaptic responses for developing diverse functionalities of the biological nervous system, was developed using ambipolar oxide semiconductor thin-film transistors (TFTs). Since the balanced ambipolarity is significant, a boron-incorporated SnO (SnO:B) oxide semiconductor channel was newly developed to improve the ambipolar charge transports by reducing the subgap defect density, which was reduced to less than 10¹⁷ cm^-3. The ambipolar SnO:B-TFT could be fabricated with a good reproductivity at the maximum process temperature of 250 °C and exhibited good TFT performances, such as a nearly zero switching voltage, the saturation mobility of ∼1.3 cm² V^-1 s^-1, s-value of ∼1.1 V decade^-1, and an on/off-current ratio of ∼8 × 10³ for the p-channel mode, while ∼0.14 cm² V^-1 s^-1, ∼2.2 V decade^-1and ∼1 × 10³ for n-channel modes, respectively. The ambipolar device imitated potentiation/depression behaviors in both excitatory and inhibitory synaptic responses by using the p- and n-channel transports by tuning a gate bias. The low-power consumptions of <20 and <2 nJ per pulse for the excitatory and inhibitory operations, respectively, were also achieved. The presented device operated under an ambient atmosphere and confirmed a good operation reliability over 5000 pulses and a long-term air environmental stability. The study presents the high potential of an ambipolar oxide-TFT-based synaptic device with a good manufacturability to develop emerging neuromorphic perception and computing hardware for next-generation artificial intelligence systems.

Nonvolatile Memories in Spiking Neural Network Architectures: Current and Emerging Trends

A sustainable computing scenario demands more energy-efficient processors. Neuromorphic systems mimic biological functions by employing spiking neural networks for achieving brain-like efficiency, speed, adaptability, and intelligence. Current trends in neuromorphic technologies address the challenges of investigating novel materials, systems, and architectures for enabling high-integration and extreme low-power brain-inspired computing. This review collects the most recent trends in exploiting the physical properties of nonvolatile memory technologies for implementing efficient in-memory and in-device computing with spike-based neuromorphic architectures.

Impact of Zr top electrode on tantalum oxide-based electrochemical metallization resistive switching memory: towards synaptic functionalities

Electrochemical metallization memory (ECM) devices have been made by sub-stoichiometric deposition of a tantalum oxide switching film (Ta₂O_5-x ) using sputtering. We investigated the influence of zirconium as the active top electrode material in the lithographically fabricated ECM devices. A simple capacitor like (Pt/Zr/Ta₂O_5-x /Pt) structure represented the resistive switching memory. A cyclic voltammetry measurement demonstrated the electrochemical process of the memory device. The I-V characteristics of ECMs show stable bipolar resistive switching properties with reliable endurance and retention. The resistive switching mechanism results from the formation and rupture of a conductive filament characteristic of ECM. Our results suggest that Zr can be considered a potential active electrode in the ECMs for the next generation of nonvolatile nanoelectronics. We successfully showed that the ECM device can work under AC pulses to emulate the essential characteristics of an artificial synapse by further improvements.

Synaptic Emulation via Ferroelectric P(VDF-TrFE) Reinforced Charge Trapping/Detrapping in Zinc-Tin Oxide Transistor

Brain inspired artificial synapses are highly desirable for neuromorphic computing and are an alternative to a conventional computing system. Here, we report a simple and cost-effective ferroelectric capacitively coupled zinc-tin oxide (ZTO) thin-film transistor (TFT) topped with ferroelectric copolymer poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)) for artificial synaptic devices. Ferroelectric dipoles enhance the charge trapping/detrapping effect in ZTO TFT, as confirmed by the transfer curve (I_D-V_G) analysis. This substantiates superior artificial synapse responses in ferroelectric-coupled ZTO TFT because the current potentiation and depression are individually improved. The ferroelectric-coupled ZTO TFT successfully emulates the essential features of the artificial synapse, including pair-pulsed facilitation (PPF) and potentiation/depression (P/D) characteristics. In addition, the device also mimics the memory consolidation behavior through intensified stimulation. This work demonstrates that the ferroelectric-coupled ZTO synaptic transistor possesses great potential as a hardware candidate for neuromorphic computing.

Synaptic transistors based on a tyrosine-rich peptide for neuromorphic computing

In this article, we propose an artificial synaptic device based on a proton-conducting peptide material. By using the redox-active property of tyrosine, the Tyr-Tyr-Ala-Cys-Ala-Tyr-Tyr peptide film was utilized as a gate insulator that shows synaptic plasticity owing to the formation of proton electric double layers. The ion gating effects on the transfer characteristics and temporal current responses are shown. Further, timing-dependent responses, including paired-pulse facilitation, synaptic potentiation, and transition from short-term plasticity to long-term plasticity, have been demonstrated for the electrical emulation of biological synapses in the human brain. Herein, we provide a novel material platform that is bio-inspired and biocompatible for use in brain-mimetic electronic devices.

Artificial Synapse Based on a 2D-SnO2 Memtransistor with Dynamically Tunable Analog Switching for Neuromorphic Computing

A new type of two-dimensional (2D) SnO2 semiconductor-based gate-tunable memristor, that is, a memtransistor, an integrated device of a memristor and a transistor, was demonstrated to advance next-generation neuromorphic computing technology. The polycrystalline 2D-SnO2 memristors derived from a low-temperature and vacuum-free liquid metal process offer several interesting resistive switching properties such as excellent digital/analog resistive switching, multistate storage, and gate-tunability function of resistance switching states. Significantly, the gate tunability function that is not achievable in conventional two-terminal memristors provides the capability to implement heterosynaptic analog switching by regulating gate bias for enabling complex neuromorphic learning. We successfully demonstrated that the gate-tunable synaptic device dynamically modulated the analog switching behavior with good linearity and an improved conductance change ratio for high recognition accuracy learning. The presented gate-tunable 2D-oxide memtransistor will advance neuromorphic device technology and open up new opportunities to design learning schemes with an extra degree of freedom.

Tunable biological nonvolatile multilevel data storage devices

The speed with which electronic products are updated is continuously increasing. Consequently, since waste electronic products can cause serious environmental pollution, the demand for electronic products made of biological materials is becoming increasingly urgent. Although biological memristors have significant advantages, their electrical characteristics still do not meet the requirements to be used in future nonvolatile memories. Therefore, how to control their electrical characteristics has become a popular topic of research. In this study, tunable biomemristors with an Al/tussah blood (TB)-carbon nanotube (CNT)/indium tin oxide (ITO)/glass structure were fabricated. Such a device exhibits stable bipolar resistance switching behavior and good retention characteristics (10⁴ s). Experimental results show that the ON/OFF current ratio can be effectively controlled by modifying the CNT concentration in the TB-CNT composite film. Multilevel (8 levels, 3 bits per cell) storage capabilities can be achieved in the device by controlling its compliance current in order to achieve high-density storage. The resistance switching behavior originates from the formation and rupture of conductive oxygen vacancy filaments. TB is a promising natural biomaterial in the field of green electronics, and this research could blaze a new trail for the development of biological memory devices. Biomemristors with multilevel resistance states can be used as electronic synapses and are one of the choices for simulating biological synapses.

Biocompatible artificial synapses based on a zein active layer obtained from maize for neuromorphic computing

Artificial synaptic devices based on natural organic materials are becoming the most desirable for extending their fields of applications to include wearable and implantable devices due to their biocompatibility, flexibility, lightweight, and scalability. Herein, we proposed a zein material, extracted from natural maize, as an active layer in an artificial synapse. The synaptic device exhibited notable digital-data storage and analog data processing capabilities. Remarkably, the zein-based synaptic device achieved recognition accuracy of up to 87% and exhibited clear digit-classification results on the learning and inference test. Moreover, the recognition accuracy of the zein-based artificial synapse was maintained within a difference of less than 2%, regardless of mechanically stressed conditions. We believe that this work will be an important asset toward the realization of wearable and implantable devices utilizing artificial synapses.

Multimodal Artificial Neurological Sensory-Memory System Based on Flexible Carbon Nanotube Synaptic Transistor

As the initial stage in the formation of human intelligence, the sensory-memory system plays a critical role for human being to perceive, interact, and evolve with the environment. Electronic implementation of such biological sensory-memory system empowers the development of environment-interactive artificial intelligence (AI) that can learn and evolve with diversified external information, which could potentially broaden the application of the AI technology in the field of human-computer interaction. Here, we report a multimodal artificial sensory-memory system consisting of sensors for generating biomimetic visual, auditory, tactile inputs, and flexible carbon nanotube synaptic transistor that possesses synapse-like signal processing and memorizing behaviors. The transduction of physical signals into information-containing, presynaptic action potentials and the synaptic plasticity of the transistor in response to single and long-term action potential excitations have been systematically characterized. The bioreceptor-like sensing and synapse-like memorizing behaviors have also been demonstrated. On the basis of the memory and learning characteristics of the sensory-memory system, the well-known psychological model describing human memory, the "multistore memory" model, and the classical conditioning experiment that demonstrates the associative learning of brain, "Pavlov's dog's experiment", have both been implemented electronically using actual physical input signals as the sources of the stimuli. The biomimetic intelligence demonstrated in this neurological sensory-memory system shows its potential in promoting the advancement in multimodal, user-environment interactive AI.

Core-Shell Dual-Gate Nanowire Charge-Trap Memory for Synaptic Operations for Neuromorphic Applications

This work showcases the physical insights of a core-shell dual-gate (CSDG) nanowire transistor as an artificial synaptic device with short/long-term potentiation and long-term depression (LTD) operation. Short-term potentiation (STP) is a temporary potentiation of a neural network, and it can be transformed into long-term potentiation (LTP) through repetitive stimulus. In this work, floating body effects and charge trapping are utilized to show the transition from STP to LTP while de-trapping the holes from the nitride layer shows the LTD operation. Furthermore, linearity and symmetry in conductance are achieved through optimal device design and biases. In a system-level simulation, with CSDG nanowire transistor a recognition accuracy of up to 92.28% is obtained in the Modified National Institute of Standards and Technology (MNIST) pattern recognition task. Complementary metal-oxide-semiconductor (CMOS) compatibility and high recognition accuracy makes the CSDG nanowire transistor a promising candidate for the implementation of neuromorphic hardware.

Synaptic transistors and neuromorphic systems based on carbon nano-materials

Carbon-based materials possessing a nanometer size and unique electrical properties perfectly address the two critical issues of transistors, the low power consumption and scalability, and are considered as a promising material in next-generation synaptic devices. In this review, carbon-based synaptic transistors were systematically summarized. In the carbon nanotube section, the synthesis of carbon nanotubes, purification of carbon nanotubes, the effect of architecture on the device performance and related carbon nanotube-based devices for neuromorphic computing were discussed. In the graphene section, the synthesis of graphene and its derivative, as well as graphene-based devices for neuromorphic computing, was systematically studied. Finally, the current challenges for carbon-based synaptic transistors were discussed.

High photosensitivity light-controlled planar ZnO artificial synapse for neuromorphic computing

A light-controlled artificial synapse, which mimics the human brain has been considered to be one of the ideal candidates for the fundamental physical architecture of a neuromorphic computing system owing to the possible abilities of high bandwidth and low power calculation. However, the low photosensitivity of synapse devices can affect the accuracy of recognition and classification in neuromorphic computing tasks. In this work, a planar light-controlled artificial synapse having high photosensitivity (Ion/Ioff > 1000) with a high photocurrent and a low dark current is realized based on a ZnO thin film grown by radiofrequency sputtering. The synaptic functions of the human brain such as sensory memory, short-term memory, long-term memory, duration-time-dependent-plasticity, light-intensity-dependent-plasticity, learning-experience behavior, neural facilitation, and spike-timing-dependent plasticity are successfully emulated using persistent photoconductivity characteristic of a ZnO thin film. Furthermore, the high classification accuracy of 90%, 92%, and 86% after 40 epochs for file type datasets, small digits, and large digit is realized with a three-layer neural network based on backpropagation where the numerical weights in the network layer are mapped directly to the conductance states of the experimental synapse devices. Finally, characterization and analysis reveal that oxygen vacancy defects and chemisorbed oxygen on the surface of the ZnO film are the main factors that determine the performance of the device.

Solution-processed electronics for artificial synapses

Artificial synaptic devices and systems have become hot topics due to parallel computing, high plasticity, integration of storage, and processing to meet the challenges of the traditional Von Neumann computers. Currently, two-terminal memristors and three-terminal transistors have been mainly developed for high-density storage with high switching speed and high reliability because of the adjustable resistivity, controllable ion migration, and abundant choices of functional materials and fabrication processes. To achieve the low-cost, large-scale, and easy-process fabrication, solution-processed techniques have been extensively employed to develop synaptic electronics towards flexible and highly integrated three-dimensional (3D) neural networks. Herein, we have summarized and discussed solution-processed techniques in the fabrication of two-terminal memristors and three-terminal transistors for the application of artificial synaptic electronics mainly reported in the recent five years from the view of fabrication processes, functional materials, electronic operating mechanisms, and system applications. Furthermore, the challenges and prospects were discussed in depth to promote solution-processed techniques in the future development of artificial synapse with high performance and high integration.

Emerging Materials for Neuromorphic Devices and Systems

Neuromorphic devices and systems have attracted attention as next-generation computing due to their high efficiency in processing complex data. So far, they have been demonstrated using both machine-learning software and complementary metal-oxide-semiconductor-based hardware. However, these approaches have drawbacks in power consumption and learning speed. An energy-efficient neuromorphic computing system requires hardware that can mimic the functions of a brain. Therefore, various materials have been introduced for the development of neuromorphic devices. Here, recent advances in neuromorphic devices are reviewed. First, the functions of biological synapses and neurons are discussed. Also, deep neural networks and spiking neural networks are described. Then, the operation mechanism and the neuromorphic functions of emerging devices are reviewed. Finally, the challenges and prospects for developing neuromorphic devices that use emerging materials are discussed.

Flexible Carbon Nanotube Synaptic Transistor for Neurological Electronic Skin Applications

There is an increasing interest in the development of memristive or artificial synaptic devices that emulate the neuronal activities for neuromorphic computing applications. While there have already been many reports on artificial synaptic transistors implemented on rigid substrates, the use of flexible devices could potentially enable an even broader range of applications. In this paper, we report artificial synaptic thin-film transistors built on an ultrathin flexible substrate using high carrier mobility semiconducting single-wall carbon nanotubes. The synaptic characteristics of the flexible synaptic transistor including long-term/short-term plasticity, spike-amplitude-dependent plasticity, spike-width-dependent plasticity, paired-pulse facilitation, and spike-time-dependent plasticity have all been systematically characterized. Furthermore, we have demonstrated a flexible neurological electronic skin and its peripheral nerve with a flexible ferroelectret nanogenerator (FENG) serving as the sensory mechanoreceptor that generates action potentials to be processed and transmitted by the artificial synapse. In such neurological electronic skin, the flexible FENG sensor converts the tactile input (magnitude and frequency of force) into presynaptic action potential pulses, which are then passed to the gate of the synaptic transistor to induce change in its postsynaptic current, mimicking the modulation of synaptic weight in a biological synapse. Our neurological electronic skin closely imitates the behavior of actual human skin, and it allows for instantaneous detection of force stimuli and offers biological synapse-like behavior to relay the stimulus signals to the next stage. The flexible sensory skin could potentially be used to interface with skeletal muscle fibers for applications in neuroprosthetic devices.

Neuromorphic nanoelectronic materials

Memristive and nanoionic devices have recently emerged as leading candidates for neuromorphic computing architectures. While top-down fabrication based on conventional bulk materials has enabled many early neuromorphic devices and circuits, bottom-up approaches based on low-dimensional nanomaterials have shown novel device functionality that often better mimics a biological neuron. In addition, the chemical, structural and compositional tunability of low-dimensional nanomaterials coupled with the permutational flexibility enabled by van der Waals heterostructures offers significant opportunities for artificial neural networks. In this Review, we present a critical survey of emerging neuromorphic devices and architectures enabled by quantum dots, metal nanoparticles, polymers, nanotubes, nanowires, two-dimensional layered materials and van der Waals heterojunctions with a particular emphasis on bio-inspired device responses that are uniquely enabled by low-dimensional topology, quantum confinement and interfaces. We also provide a forward-looking perspective on the opportunities and challenges of neuromorphic nanoelectronic materials in comparison with more mature technologies based on traditional bulk electronic materials.

Electret-Based Organic Synaptic Transistor for Neuromorphic Computing

Neuromorphic computing inspired by the neural systems in human brain will overcome the issue of independent information processing and storage. An artificial synaptic device as a basic unit of a neuromorphic computing system can perform signal processing with low power consumption, and exploring different types of synaptic transistors is essential to provide suitable artificial synaptic devices for artificial intelligence. Hence, for the first time, an electret-based synaptic transistor (EST) is presented, which successfully shows synaptic behaviors including excitatory/inhibitory postsynaptic current, paired-pulse facilitation/depression, long-term memory, and high-pass filtering. Moreover, a neuromorphic computing simulation based on our EST is performed using the handwritten artificial neural network, which exhibits an excellent recognition accuracy (85.88%) after 120 learning epochs, higher than most reported organic synaptic transistors and close to the ideal accuracy (92.11%). Such a novel synaptic device enriches the diversity of synaptic transistors, laying the foundation for the diversified development of the next generation of neuromorphic computing systems.

Implementation of PPI with Nano Amorphous Oxide Semiconductor Devices for Medical Applications

BACKGROUND

Electronic devices which mimic the functionality of biological synapses are a large step to replicate the human brain for neuromorphic computing and for numerous medical research investigations. One of the representative synaptic behaviors is paired-pulse facilitation (PPF). It has been widely investigated because it is regarded to be related to biological memory. However, plasticity behavior is only part of the human brain memory behavior.

METHODS

Here, we present a phenomenon which is opposite to PPF, i.e., paired-pulse inhibition (PPI), in nano oxide devices for the first time. The research here suggests that rather than being enhanced, the phenomena of memory loss would also be possessed by such electronic devices. The device physics mechanism behind memory loss behavior was investigated. This mechanism is sustained by historical memory and degradation manufactured by device trauma to regulate characteristically stimulated origins of artificial transmission behaviors.

RESULTS

Under the trauma of a memory device, both the signal amplitude and signal time stimulated by a pulse are lower than the first signal stimulated by a previous pulse in the PPF, representing a new scenario in the struggle for memory. In this way, more typical human brain behaviors could be simulated, including the effect of age on latency and error generation, cerebellar infarct, trauma and memory loss pharmacological actions (such as those caused by hyoscines and nitrazepam).

CONCLUSION

Thus, this study developed a new approach for implementing the manner in which the brain works in semiconductor devices for improving medical research.

Artificial Synapse Based on van der Waals Heterostructures with Tunable Synaptic Functions for Neuromorphic Computing

Two-dimensional (2D) materials and van der Waals heterostructures have attracted tremendous attention because of their appealing electronic, mechanical, and optoelectronic properties, which offer the possibility to extend the range of functionalities for diverse potential applications. Here, we fabricate a novel multiterminal device with dual-gate based on 2D material van der Waals heterostructures. Such a multiterminal device exhibited excellent nonvolatile multilevel resistance switching performance controlled by the source-drain voltage and back-gate voltage. Based on these features, heterosynaptic plasticity, in which the synaptic weight can be tuned by another modulatory interneuron, has been mimicked. A tunable analogue weight update (both on/off ratio and update nonlinearity) of synapse with high speed (50 ns) and low energy (∼7.3 fJ) programming has been achieved. These results demonstrate the great potential of the artificial synapse based on van der Waals heterostructures for neuromorphic computing.

Rational Band Engineering of an Organic Double Heterojunction for Artificial Synaptic Devices with Enhanced State Retention and Linear Update of Synaptic Weight

Herein, we propose an organic double heterojunction to enable a nonvolatile step modulation of the conductance of an artificial synapse; the double heterojunction is composed of N,N'-dioctyl-3,4,9,10-perylene tetracarboxylic diimide (PTCDI-C₈), copper phthalocyanine (CuPc), and para-sexiphenyl (p-6P). The carrier confinement in the CuPc region present in the double-heterojunction structure enabled the nonvolatile modulation of the postsynaptic current. The proposed organic synapse exhibited an excellent conductance change, characteristic with a nonlinearity (NL) value below 0.01 in the long-term potentiation (LTP) region. Furthermore, the NL value for long-term depression (LTD) could be reduced effectively from 45 to 3.5 by a pulse modulation technique. A simple artificial neural network (ANN) was theoretically designed using the LTP/LTD characteristic curves of such organic synapses, and then, learning and recognition tasks were performed using Modified National Institute of Standards and Technology digit images. A four-amplitude weight update method enabled considerable enhancement of the recognition rate from 53 to 70%. Although the designed ANN was based on a single-layer perceptron model, a high maximum accuracy of 75% was achieved. These newly studied techniques for synaptic devices are expected to open up new possibilities for the realization of artificial synapses based on organic double heterojunctions.

Gate-Tunable Synaptic Dynamics of Ferroelectric-Coupled Carbon-Nanotube Transistors

Artificial neural networks (ANNs) based on synaptic devices, which can simultaneously perform processing and storage of data, have superior computing performance compared to conventional von Neumann architectures. Here, we present a ferroelectric coupled artificial synaptic device with reliable weight update and storage properties for ANNs. The artificial synaptic device, which is based on a ferroelectric polymer capacitively coupled with an oxide dielectric via an electric-field-permeable, semiconducting single-walled carbon-nanotube channel, is successfully fabricated by inkjet printing. By controlling the ferroelectric polarization, synaptic dynamics, such as excitatory and inhibitory postsynaptic currents and long-term potentiation/depression characteristics, is successfully implemented in the artificial synaptic device. Furthermore, the constructed ANN, which is designed in consideration of the device-to-device variation within the synaptic array, efficiently executes the tasks of learning and recognition of the Modified National Institute of Standards and Technology numerical patterns.

Parallel weight update protocol for a carbon nanotube synaptic transistor array for accelerating neuromorphic computing

Brain-inspired neuromorphic computing has the potential to overcome the inherent inefficiency of the conventional von Neumann architecture by using the massively parallel processing power of artificial neural networks. Neuromorphic parallel processing can be implemented naturally using the crossbar geometry of synaptic device arrays with Ohm's and Kirchhoff's laws. However, selective and parallel weight updates of the synaptic crossbar array are still very challenging due to the unavoidable crosstalk between adjacent devices and sneak path currents. Here, we experimentally demonstrate a weight update protocol in a carbon nanotube synaptic transistor array, where selective and parallel weight updates can be executed by exploiting the individually controllable three terminals of the synaptic device via a localized carrier trapping mechanism. The trained 9 × 8 synaptic array solves four different convolution operations simultaneously for the feature extraction of an image. The massive parallelism and robustness of the weight update protocol are important features toward effective manipulation of big data through neuromorphic computing systems.

Artificial 2D van der Waals Synapse Devices via Interfacial Engineering for Neuromorphic Systems

Despite extensive investigations of a wide variety of artificial synapse devices aimed at realizing a neuromorphic hardware system, the identification of a physical parameter that modulates synaptic plasticity is still required. In this context, a novel two-dimensional architecture consisting of a NbSe₂/WSe₂/Nb₂O₅ heterostructure placed on an SiO₂/p+ Si substrate was designed to overcome the limitations of the conventional silicon-based complementary metal-oxide semiconductor technology. NbSe₂, WSe₂, and Nb₂O₅ were used as the metal electrode, active channel, and conductance-modulating layer, respectively. Interestingly, it was found that the post-synaptic current was successfully modulated by the thickness of the interlayer Nb₂O₅, with a thicker interlayer inducing a higher synapse spike current and a stronger interaction in the sequential pulse mode. Introduction of the Nb₂O₅ interlayer can facilitate the realization of reliable and controllable synaptic devices for brain-inspired integrated neuromorphic systems.

Nanoscale Patterning of Carbon Nanotubes: Techniques, Applications, and Future

Carbon nanotube (CNT) devices and electronics are achieving maturity and directly competing or surpassing devices that use conventional materials. CNTs have demonstrated ballistic conduction, minimal scaling effects, high current capacity, low power requirements, and excellent optical/photonic properties; making them the ideal candidate for a new material to replace conventional materials in next-generation electronic and photonic systems. CNTs also demonstrate high stability and flexibility, allowing them to be used in flexible, printable, and/or biocompatible electronics. However, a major challenge to fully commercialize these devices is the scalable placement of CNTs into desired micro/nanopatterns and architectures to translate the superior properties of CNTs into macroscale devices. Precise and high throughput patterning becomes increasingly difficult at nanoscale resolution, but it is essential to fully realize the benefits of CNTs. The relatively long, high aspect ratio structures of CNTs must be preserved to maintain their functionalities, consequently making them more difficult to pattern than conventional materials like metals and polymers. This review comprehensively explores the recent development of innovative CNT patterning techniques with nanoscale lateral resolution. Each technique is critically analyzed and applications for the nanoscale-resolution approaches are demonstrated. Promising techniques and the challenges ahead for future devices and applications are discussed.

Precision-extension technique for accurate vector-matrix multiplication with a CNT transistor crossbar array

Most machine learning algorithms involve many multiply-accumulate operations, which dictate the computation time and energy required. Vector-matrix multiplications can be accelerated using resistive networks, which can be naturally implemented in a crossbar geometry by leveraging Kirchhoff's current law in a single readout step. However, practical computing tasks that require high precision are still very challenging to implement in a resistive crossbar array owing to intrinsic device variability and unavoidable crosstalk, such as sneak path currents through adjacent devices, which inherently result in low precision. Here, we experimentally demonstrate a precision-extension technique for a carbon nanotube (CNT) transistor crossbar array. High precision is attained through multiple devices operating together, each of which stores a portion of the required bit width. A 10 × 10 CNT transistor array can perform vector-matrix multiplication with high accuracy, making in-memory computing approaches attractive for high-performance computing environments.

A Neuromorphic Device Implemented on a Salmon-DNA Electrolyte and its Application to Artificial Neural Networks

A bioinspired neuromorphic device operating as synapse and neuron simultaneously, which is fabricated on an electrolyte based on Cu2+-doped salmon deoxyribonucleic acid (S-DNA) is reported. Owing to the slow Cu2+ diffusion through the base pairing sites in the S-DNA electrolyte, the synaptic operation of the S-DNA device features special long-term plasticity with negative and positive nonlinearity values for potentiation and depression (αp and αd), respectively, which consequently improves the learning/recognition efficiency of S-DNA-based neural networks. Furthermore, the representative neuronal operation, "integrate-and-fire," is successfully emulated in this device by adjusting the duration time of the input voltage stimulus. In particular, by applying a Cu2+ doping technique to the S-DNA neuromorphic device, the characteristics for synaptic weight updating are enhanced (|αp|: 31→20, |αd|: 11→18, weight update margin: 33→287 nS) and also the threshold conditions for neuronal firing (amplitude and number of stimulus pulses) are modulated. The improved synaptic characteristics consequently increase the Modified National Institute of Standards and Technology (MNIST) pattern recognition rate from 38% to 44% (single-layer perceptron model) and from 89.42% to 91.61% (multilayer perceptron model). This neuromorphic device technology based on S-DNA is expected to contribute to the successful implementation of a future neuromorphic system that simultaneously satisfies high integration density and remarkable recognition accuracy.

Hybrid Silicon Nanowire Devices and Their Functional Diversity

In the pool of nanostructured materials, silicon nanostructures are known as conventionally used building blocks of commercially available electronic devices. Their application areas span from miniaturized elements of devices and circuits to ultrasensitive biosensors for diagnostics. In this Review, the current trends in the developments of silicon nanowire-based devices are summarized, and their functionalities, novel architectures, and applications are discussed from the point of view of analog electronics, arisen from the ability of (bio)chemical gating of the carrier channel. Hybrid nanowire-based devices are introduced and described as systems decorated by, e.g., organic complexes (biomolecules, polymers, and organic films), aimed to substantially extend their functionality, compared to traditional systems. Their functional diversity is explored considering their architecture as well as areas of their applications, outlining several groups of devices that benefit from the coatings. The first group is the biosensors that are able to represent label-free assays thanks to the attached biological receptors. The second group is represented by devices for optoelectronics that acquire higher optical sensitivity or efficiency due to the specific photosensitive decoration of the nanowires. Finally, the so-called new bioinspired neuromorphic devices are shown, which are aimed to mimic the functions of the biological cells, e.g., neurons and synapses.

Recent Progress in Three-Terminal Artificial Synapses: From Device to System

Synapses are essential to the transmission of nervous signals. Synaptic plasticity allows changes in synaptic strength that make a brain capable of learning from experience. During development of neuromorphic electronics, great efforts have been made to design and fabricate electronic devices that emulate synapses. Three-terminal artificial synapses have the merits of concurrently transmitting signals and learning. Inorganic and organic electronic synapses have mimicked plasticity and learning. Optoelectronic synapses and photonic synapses have the prospective benefits of low electrical energy loss, high bandwidth, and mechanical robustness. These artificial synapses provide new opportunities for the development of neuromorphic systems that can use parallel processing to manipulate datasets in real time. Synaptic devices have also been used to build artificial sensory systems. Here, recent progress in the development and application of three-terminal artificial synapses and artificial sensory systems is reviewed.

Laser-reconfigured MoS2/ZnO van der Waals synapse

Inspired by biological neural systems, neuromorphic devices may lead to new computing paradigms for exploring cognition, learning and limits of parallel computation. Synapses form the basis of neuromorphic computing and have attracted significant research interest in recent years. Herein, a three-terminal transistor based on a transition metal sulfide and zinc oxide heterojunction is proposed for emulating biological synapses. The transistor exhibits an ON/OFF ratio (10⁴) and significant rectifying behavior with forward-to-reverse bias current ratios of 10⁴. The device demonstrates the essential synaptic behaviors, such as excitatory postsynaptic current, modulation of synaptic weight and paired-pulse facilitation. Furthermore, we show that the hysteretic effect of the transfer curves and the post-synapse current triggered by the presynaptic pulses can be modulated by illumination, and the current under illumination conditions is about 10 times greater than that in the dark. These synapses combine photonic with electric neuromorphic functions, thus showing the application prospects of the optoelectronic interfaces for integrated photonic circuits based on mixed-mode electro-optical operation. Hence, this work offers a new landscape for 2D-material electronics and encourages future research on neuro-electronics.

A Ferrite Synaptic Transistor with Topotactic Transformation

Hardware implementation of artificial synaptic devices that emulate the functions of biological synapses is inspired by the biological neuromorphic system and has drawn considerable interest. Here, a three-terminal ferrite synaptic device based on a topotactic phase transition between crystalline phases is presented. The electrolyte-gating-controlled topotactic phase transformation between brownmillerite SrFeO_2.5 and perovskite SrFeO_{3- δ} is confirmed from the examination of the crystal and electronic structure. A synaptic transistor with electrolyte-gated ferrite films by harnessing gate-controllable multilevel conduction states, which originate from many distinct oxygen-deficient perovskite structures of SrFeO_x induced by topotactic phase transformation, is successfully constructed. This three-terminal artificial synapse can mimic important synaptic functions, such as synaptic plasticity and spike-timing-dependent plasticity. Simulations of a neural network consisting of ferrite synaptic transistors indicate that the system offers high classification accuracy. These results provide insight into the potential application of advanced topotactic phase transformation materials for designing artificial synapses with high performance.

An Evolvable Organic Electrochemical Transistor for Neuromorphic Applications

An evolvable organic electrochemical transistor (OECT), operating in the hybrid accumulation-depletion mode is reported, which exhibits short-term and long-term memory functionalities. The transistor channel, formed by an electropolymerized conducting polymer, can be formed, modulated, and obliterated in situ and under operation. Enduring changes in channel conductance, analogous to long-term potentiation and depression, are attained by electropolymerization and electrochemical overoxidation of the channel material, respectively. Transient changes in channel conductance, analogous to short-term potentiation and depression, are accomplished by inducing nonequilibrium doping states within the transistor channel. By manipulating the input signal, the strength of the transistor response to a given stimulus can be modulated within a range that spans several orders of magnitude, producing behavior that is directly comparable to short- and long-term neuroplasticity. The evolvable transistor is further incorporated into a simple circuit that mimics classical conditioning. It is forecasted that OECTs that can be physically and electronically modulated under operation will bring about a new paradigm of machine learning based on evolvable organic electronics.

Artificial synapses based on nanomaterials

Artificial synapses emulate biological synaptic signals in neuromorphic systems to attain brain-like computation and autonomous learning behaviors in non-von-Neumann systems. Several classes of materials have been applied to this field to achieve numerous functionalities of biological synapses. Nanomaterials (NMs), such as one-dimensional (1D) and two-dimensional (2D) NMs have shown great potential due to their nanometer feature size (1D) and molecular-level thickness (2D). In this paper, we review the development of artificial synapses, and discuss state-of-the-art artificial synapses based on NMs.

Truly Concomitant and Independently Expressed Short- and Long-Term Plasticity in a Bi2 O2 Se-Based Three-Terminal Memristor

Concomitance of diverse synaptic plasticity across different timescales produces complex cognitive processes. To achieve comparable cognitive complexity in memristive neuromorphic systems, devices that are capable of emulating short-term (STP) and long-term plasticity (LTP) concomitantly are essential. In existing memristors, however, STP and LTP can only be induced selectively because of the inability to be decoupled using different loci and mechanisms. In this work, the first demonstration of truly concomitant STP and LTP is reported in a three-terminal memristor that uses independent physical phenomena to represent each form of plasticity. The emerging layered material Bi₂ O₂ Se is used for memristors for the first time, opening up the prospects for ultrathin, high-speed, and low-power neuromorphic devices. The concerted action of STP and LTP allows full-range modulation of the transient synaptic efficacy, from depression to facilitation, by stimulus frequency or intensity, providing a versatile device platform for neuromorphic function implementation. A heuristic recurrent neural circuitry model is developed to simulate the intricate "sleep-wake cycle autoregulation" process, in which the concomitance of STP and LTP is posited as a key factor in enabling this neural homeostasis. This work sheds new light on the development of generic memristor platforms for highly dynamic neuromorphic computing.

Artificial optic-neural synapse for colored and color-mixed pattern recognition

The priority of synaptic device researches has been given to prove the device potential for the emulation of synaptic dynamics and not to functionalize further synaptic devices for more complex learning. Here, we demonstrate an optic-neural synaptic device by implementing synaptic and optical-sensing functions together on h-BN/WSe₂ heterostructure. This device mimics the colored and color-mixed pattern recognition capabilities of the human vision system when arranged in an optic-neural network. Our synaptic device demonstrates a close to linear weight update trajectory while providing a large number of stable conduction states with less than 1% variation per state. The device operates with low voltage spikes of 0.3 V and consumes only 66 fJ per spike. This consequently facilitates the demonstration of accurate and energy efficient colored and color-mixed pattern recognition. The work will be an important step toward neural networks that comprise neural sensing and training functions for more complex pattern recognition.

Artificial neural networks can emulate the human vision because of their spike-based operation by employing memristors as synapses. Here, Seo et al. integrate synaptic and optical sensing functions in a single heterostructure, which enables accurate and energy-efficient recognition of colored patterns.

Synaptic Device Network Architecture with Feature Extraction for Unsupervised Image Classification

For the efficient recognition and classification of numerous images, neuroinspired deep learning algorithms have demonstrated their substantial performance. Nevertheless, current deep learning algorithms that are performed on von Neumann machines face significant limitations due to their inherent inefficient energy consumption. Thus, alternative approaches (i.e., neuromorphic systems) are expected to provide more energy-efficient computing units. However, the implementation of the neuromorphic system is still challenging due to the uncertain impacts of synaptic device specifications on system performance. Moreover, only few studies are reported how to implement feature extraction algorithms on the neuromorphic system. Here, a synaptic device network architecture with a feature extraction algorithm inspired by the convolutional neural network is demonstrated. Its pattern recognition efficacy is validated using a device-to-system level simulation. The network can classify handwritten digits at up to a 90% recognition rate despite using fewer synaptic devices than the architecture without feature extraction.

Aligned Carbon Nanotube Synaptic Transistors for Large-Scale Neuromorphic Computing

This paper presents aligned carbon nanotube (CNT) synaptic transistors for large-scale neuromorphic computing systems. The synaptic behavior of these devices is achieved via charge-trapping effects, commonly observed in carbon-based nanoelectronics. In this work, charge trapping in the high- k dielectric layer of top-gated CNT field-effect transistors (FETs) enables the gradual analog programmability of the CNT channel conductance with a large dynamic range ( i. e., large on/off ratio). Aligned CNT synaptic devices present significant improvements over conventional memristor technologies ( e. g., RRAM), which suffer from abrupt transitions in the conductance modulation and/or a small dynamic range. Here, we demonstrate exceptional uniformity of aligned CNT FET synaptic behavior, as well as significant robustness and nonvolatility via pulsed experiments, establishing their suitability for neural network implementations. Additionally, this technology is based on a wafer-level technique for constructing highly aligned arrays of CNTs with high semiconducting purity and is fully CMOS compatible, ensuring the practicality of large-scale CNT+CMOS neuromorphic systems. We also demonstrate fine-tunability of the aligned CNT synaptic behavior and discuss its application to adaptive online learning schemes and to homeostatic regulation of artificial neuron firing rates. We simulate the implementation of unsupervised learning for pattern recognition using a spike-timing-dependent-plasticity scheme, indicate system-level performance (as indicated by the recognition accuracy), and demonstrate improvements in the learning rate resulting from tuning the synaptic characteristics of aligned CNT devices.

Polymer-Nanocarbon Topological and Electronic Interface

The electronic structure of semiconducting carbon nanotubes selected through polymer functionalization is routinely verified by measuring the spectral van Hove singularity signature under ultraclean vacuum conditions. Interpreting the effect of unperturbed polymer adsorption on the nanotube energetic bands in solvent media is experimentally challenging owing to solvent molecular crowding around the hybrid complex. Here, a liquid-based scanning tunneling microscope and spectroscope operating in a noise-free laboratory is used to resolve the polymer-semiconducting carbon-nanotube-underlying graphene heterostructure in the presence of encompassing solvent molecules. The spectroscopic measurements highlight the role of polymer packing and graphene landscape on the electronic shifts induced in the nanotube energy bands. Together with molecular dynamics simulations, our experimental findings emphasize the necessity of recording physicochemical and electronic properties of liquid-phase solubilized hybrid materials in their native state.

Mimicking Biological Synaptic Functionality with an Indium Phosphide Synaptic Device on Silicon for Scalable Neuromorphic Computing

Neuromorphic or "brain-like" computation is a leading candidate for efficient, fault-tolerant processing of large-scale data as well as real-time sensing and transduction of complex multivariate systems and networks such as self-driving vehicles or Internet of Things applications. In biology, the synapse serves as an active memory unit in the neural system and is the component responsible for learning and memory. Electronically emulating this element via a compact, scalable technology which can be integrated in a three-dimensional (3-D) architecture is critical for future implementations of neuromorphic processors. However, present day 3-D transistor implementations of synapses are typically based on low-mobility semiconductor channels or technologies that are not scalable. Here, we demonstrate a crystalline indium phosphide (InP)-based artificial synapse for spiking neural networks that exhibits elasticity, short-term plasticity, long-term plasticity, metaplasticity, and spike timing-dependent plasticity, emulating the critical behaviors exhibited by biological synapses. Critically, we show that this crystalline InP device can be directly integrated via back-end processing on a Si wafer using a SiO₂ buffer without the need for a crystalline seed, enabling neuromorphic devices that can be implemented in a scalable and 3-D architecture. Specifically, the device is a crystalline InP channel field-effect transistor that interacts with neuron spikes by modification of the population of filled traps in the MOS structure itself. Unlike other transistor-based implementations, we show that it is possible to mimic these biological functions without the use of external factors (e.g., surface adsorption of gas molecules) and without the need for the high electric fields necessary for traditional flash-based implementations. Finally, when exposed to neuronal spikes with a waveform similar to that observed in the brain, these devices exhibit the ability to learn without the need for any external potentiating/depressing circuits, mimicking the biological process of Hebbian learning.

Impact of Synaptic Device Variations on Pattern Recognition Accuracy in a Hardware Neural Network

Neuromorphic systems (hardware neural networks) derive inspiration from biological neural systems and are expected to be a computing breakthrough beyond conventional von Neumann architecture. Interestingly, in neuromorphic systems, the processing and storing of information can be performed simultaneously by modulating the connection strength of a synaptic device (i.e., synaptic weight). Previously investigated synaptic devices can emulate the functionality of biological synapses successfully by utilizing various nano-electronic phenomena; however, the impact of intrinsic synaptic device variability on the system performance has not yet been studied. Here, we perform a device-to-system level simulation of different synaptic device variation parameters in a designed neuromorphic system that has the potential for unsupervised learning and pattern recognition. The effects of variations in parameters such as the weight modulation nonlinearity (NL), the minimum-maximum weight (G_min and G_max), and the weight update margin (ΔG) on the pattern recognition accuracy are analyzed quantitatively. These simulation results can provide guidelines for the continued design and optimization of a synaptic device for realizing a functional large-scale neuromorphic computing system.