Synapse Array Research Articles

Tin Doping Induced High‐Performance Solution‐Processed Ga2O3 Photosensor toward Neuromorphic Visual System

AbstractGa2O3 is an emerging wide‐bandgap semiconductor with high deep ultraviolet absorption, tunable persistent photoconductivity, and excellent stability toward electric fields, making it a promising component for neuromorphic visual systems (NVSs). However, Ga2O3‐based photosensors with high responsivity and long response decay times are required for efficient NVSs. A solution‐processed doping strategy for fabrication of Ga2O3 is proposed with tin foil as a dopant source. Tin‐doped Ga2O3 (Ga2O3:Sn) photosensors are obtained with ultrahigh responsivity and extremely long response decay times. These behaviors are attributed to substitutional tin and oxygen vacancies that modulate defect‐related hole trapping. High‐performance Ga2O3:Sn photosensors can mimic photonic synaptic behaviors and image pre‐processing functions. NVSs based on a Ga2O3:Sn photonic synapse array perform pattern recognition with an accuracy of 97.3% under an unprecedented low‐light pulse stimuli of 0.5 µW cm−2. This work provides a low‐cost solution‐processed approach to ultrasensitive Ga2O3:Sn NVSs and will facilitate developments in artificial intelligence technology.

Flexible, Transparent, and Wafer-Scale Artificial Synapse Array Based on TiOx /Ti3 C2 Tx Film for Neuromorphic Computing.

A high-density neuromorphic computing memristor array based on 2Dmaterials paves the way for next-generation information-processing components and in-memory computing systems. However, the traditional 2D-materials-based memristor devices suffer from poor flexibility and opacity, which hinders the application of memristors in flexible electronics. Here, a flexible artificial synapse array based on TiOx /Ti3 C2 Tx film is fabricated by a convenient and energy-efficient solution-processing technique, which realizes high transmittance (≈90%) and oxidation resistance (>30 days). The TiOx /Ti3 C2 Tx memristor shows low device-to-device variability, long memory retention and endurance, a high ON/OFF ratio, and fundamental synaptic behavior. Furthermore, satisfactory flexibility (R = 1.0mm) and mechanical endurance (104 bending cycles) of the TiOx /Ti3 C2 Tx memristor are achieved, which is superior to other film memristors prepared by chemical vapor deposition. In addition, high-precision (>96.44%) MNIST handwritten digits recognition classification simulation indicates that the TiOx /Ti3 C2 Tx artificial synapse array holds promise for future neuromorphic computing applications, and provides excellent high-density neuron circuits for new flexible intelligent electronic equipment.

Organic Memristor-Based Flexible Neural Networks with Bio-Realistic Synaptic Plasticity for Complex Combinatorial Optimization.

Hardware neural networks with mechanical flexibility are promising next-generation computing systems for smart wearable electronics. Several studies have been conducted on flexible neural networks for practical applications; however, developing systems with complete synaptic plasticity for combinatorial optimization remains challenging. In this study, the metal-ion injection density is explored as a diffusive parameter of the conductive filament in organic memristors. Additionally, a flexible artificial synapse with bio-realistic synaptic plasticity is developed using organic memristors that have systematically engineered metal-ion injections, for the first time. In the proposed artificial synapse, short-term plasticity (STP), long-term plasticity, and homeostatic plasticity are independently achieved and are analogous to their biological counterparts. The time windows of the STP and homeostatic plasticity are controlled by the ion-injection density and electric-signal conditions, respectively. Moreover, stable capabilities for complex combinatorial optimization in the developed synapse arrays are demonstrated under spike-dependent operations. This effective concept for realizing flexible neuromorphic systems for complex combinatorial optimization is an essential building block for achieving a new paradigm of wearable smart electronics associated with artificial intelligent systems.

Organic All-Photonic Artificial Synapses Enabled by Anti-Stokes Photoluminescence.

All-photonic synaptic devices with the merits of visible signals and high spatiotemporal resolution are promising to break the Von Neumann bottleneck. Although organic synapses outperform their inorganic counterpart for easy molecular modulation and lower energy consumption, the organic all-photonic artificial synapse has never been reported. Here, all-photonic synaptic characteristics were unprecedentedly observed in an organic semiconductor, (3,6-dimethyl-9H-carbazol-9-yl)(thiophen-2-yl) methanone (S2OC), with anti-Stokes photoluminescence. Impressively, the intensity of fluorescence from the higher excited state (S3) exhibited synaptic performance, which constantly increased with irradiation time through a channel composed of intersystem crossing, triplet-triplet annihilation, and energy transfer. More importantly, the relationship between the molecular structure and synaptic performance was established. Based on the synaptic photoplasticity property, noncontacted multilevel anticounterfeiting and imaging recognition were realized in all-photonic synapse arrays. This work provides a universal strategy for tuning the performances of organic synapses upon regulating the molecular structures, which paves the way for the application of organic semiconductors in artificial intelligence.

Optoelectronic artificial synapse based on Si1-xSnx alloyed film

The emergence of big data has increased the requirements for computing and processing systems. Memristors–electrical components that simulate the synapses and neurons of the human brain—have been proposed to help meet these requirements. Optoelectronic synapses are excited by optical signals and have the advantages of low crosstalk, high bandwidth, and low power consumption. Herein, a MOS structure optoelectronic synapse device based on an amorphous silicon (a-Si1-xSnx; x = 0.0457) alloy thin film and an AlOy thin film was fabricated via oxygen vacancy energy band and heterostructure design. Significantly, the device exhibited ultralong decay time of 3576.07 s and a broadband response from the visible to the near-infrared-I region. A series of important synaptic functions, including excitatory postsynaptic current, paired-pulse facilitation, short-term memory to long-term memory transition, and learning experience behaviour, was successfully simulated on the device. We simulated a 3 × 3 optoelectronic synapse array, which can simulate the visual memory of images observed by the human eye. The effect of ‘pupil’ scaling on image memory was simulated by adjusting the read voltage. Based on these advantages, the optoelectronic synaptic device developed in this study has significant potential for application in the new generation of artificial synaptic systems.

Flexibly Photo-Regulated Brain-Inspired Functions in Flexible Neuromorphic Transistors.

As an attractive prototype for neuromorphic computing, the difficultly attained three-terminal platforms have specific advantages in implementing the brain-inspired functions. Also, in these devices, the most utilized mechanisms are confined to the electrical gate-controlled ionic migrations, which are sensitive to the device defects and stoichiometric ratio. The resultant memristive responses have fluctuant characteristics, which have adverse influences on the neural emulations. Herein, we designed a specific transistor platform with light-regulated ambipolar memory characteristics. Also, based on its gentle processes of charge trapping, we obtain the impressive memristive performances featured by smooth responses and long-term endurable characteristics. The optoelectronic samples were also fabricated on flexible substrates successfully. Interestingly, based on the optoelectronic signals of the flexible devices, we endow the desirable optical processes with the brain-inspired emulations. We can flexibly emulate the light-inspired learning-memory functions in a synapse and further devise the advanced synapse array. More importantly, through this versatile platform, we investigate the mutual regulation of excitation and inhibition and implement their sensitive-mode transformations and the homeostasis property, which is conducive to ensuring the stability of overall neural activity. Furthermore, our flexible optoelectronic platform achieves high classification accuracy when implemented in artificial neural network simulations. This work demonstrates the advantages of the optoelectronic platform in implementing the significant brain-inspired functions and provides an insight into the future integration of visible sensing in flexible optoelectronic transistor platforms.

Design of Synaptic Driving Circuit for TFT eFlash-Based Processing-In-Memory Hardware Using Hybrid Bonding

This paper presents a synaptic driving circuit design for processing in-memory (PIM) hardware with a thin-film transistor (TFT) embedded flash (eFlash) for a binary/ternary-weight neural network (NN). An eFlash-based synaptic cell capable of programming negative weight values to store binary/ternary weight values (i.e., ±1, 0) and synaptic driving circuits for erase, program, and read operations of synaptic arrays have been proposed. The proposed synaptic driving circuits improve the calculation accuracy of PIM operation by precisely programming the sensing current of the eFlash synaptic cell to the target current (50 nA ± 0.5 nA) using a pulse train. In addition, during PIM operation, the pulse-width modulation (PWM) conversion circuit converts 8-bit input data into one continuous PWM pulse to minimize non-linearity in the synaptic sensing current integration step of the neuron circuit. The prototype chip, including the proposed synaptic driving circuit, PWM conversion circuit, neuron circuit, and digital blocks, is designed and laid out as the accelerator for binary/ternary weighted NN with a size of 324 × 80 × 10 using a 0.35 μm CMOS process. Hybrid bonding technology using bump bonding and wire bonding is used to package the designed CMOS accelerator die and TFT eFlash-based synapse array dies into a single chip package.

Highly uniform photonic synapse arrays based on TIPS-pentacene nanowires/CsPbBr3 heterojunctions

Organic nanowire thin films was fabricated by template imprinting method. The photonic synaptic array simulated the characteristics of artificial light synapses. Based on a composite synaptic array, an artificial vision system was achieved.

Fiber-shaped artificial optoelectronic synapses for wearable visual-memory systems

<h2>Summary</h2> Biomimetic visual systems that are enabled with receptor-like exquisite visual sensing and synapse-like learning/memory functions offer exciting opportunities for futuristically wearable electronics. However, attempts to develop wearable visual systems remain limited, and present-day challenges in artificial visual systems are unsatisfactory geometric compatibility with irregular surfaces and poor breathability. Here, a fiber-shaped artificial optoelectronic synapse (FAOS) with double-twisted architecture consisting of high-density TiO_2-x and MoS₂ arrays is successfully constructed. Significantly, the proposed FAOS can emulate both electrical and light-induced synaptic functions such as excitatory postsynaptic current, short/long-term plasticity, and "learning-forgetting-relearning" behavior. The FAOS also exhibits excellent flexibility without obvious degradation of optoelectronic synaptic performance during bending. As a proof-of-concept demonstration, multiple FAOSs are woven into commercial textiles to form optoelectronic synapse arrays for perception and memory of image information. This work provides guidance on rational design for advanced FAOSs and accelerates practical application in wearable electronic textiles.

All Two‐dimensional Integration‐Type Optoelectronic Synapse Mimicking Visual Attention Mechanism for Multi‐Target Recognition

AbstractThe human visual attention mechanism enables them to rapidly perceive important information and objects in a complex external scene; this effectively solves the problems of data redundancy, low‐resolution images, and substantial computing resources. The process by which the attention system reconstructs the visual information can be considered as integrating internal attention signals with external visual details in the postsynaptic neuron. However, electronic devices that simulate visual attention modulation by incorporating device characteristics into neuromorphic vision systems (NVSs) to achieve visual attention behavior are rarely reported. Herein, a synapse device that integrates optical and electrical stimulation is designed using ReS2/hBN/monolayer graphene heterojunction to mimic attention regulation and integrate multiple neuron signals successfully. The synapse array can imitate perceptual learning of the human visual system (HVS) to realize visual preprocessing, such as image contrast improvement and weak signal enhancement at the sensory terminal, and overcome data redundancy. Moreover, by applying gate voltage pulses, electric‐tunable synaptic plasticity is successfully observed, attributed to the carrier trapping and de‐trapping mechanism in the floating layer. Attention stabilization, fluctuation, distraction, and reinforcement are exhibited, simulating the attention behaviors of the HVS. Thus, an NVS with attention mechanism is established depending on the optoelectronic hybrid synaptic plasticity of the device, which successfully mimics the HVS to perform a multi‐target recognition task. Furthermore, the effect of device defects on the NVS is rarely evaluated, in which a method is provided to analyze the application results of the NVS when considering uniformity and fault rate. This study may provide new inspiration for developing neuromorphic vision systems for autonomous driving and brainwave control in the future.

Robust Organic–Inorganic Heterosynapses with High PPF and Broad Photoperception

AbstractHybrid heterostructures are promising neurotransmitter platforms for photonic synapses in bioinspired optoelectronics. Organic–inorganic heterostructures have the features of a wide photoresponse and wet synthesis but still face poor stability and limited comprehensive performance. Here, it has been reported ultrastable and broadband photonic synaptic behaviors from solution‐processed cross‐linked conjugated polymer/zinc oxide nanorod heterostructures, which can be formed by in situ polymerization and hydrothermal methods. Light‐tunable synaptic function enables neuromorphic behaviors, such as high paired‐pulse facilitation, short‐ and long‐term plasticity, over the UV‐to‐visible spectral range. Impressively, the heterosynapse can maintain clear synaptic plasticity after being annealed in ambient air at 300 °C, showing robust thermal stability in harsh conditions. Moreover, visual perception and recognition can be mimicked with our synapse arrays, providing a reliable pathway to construct stable neurotransmitters toward practical neuromorphic optoelectronics.

An Efficient Training Methodology of Hardware Neural Network Based on Wafer‐Scale MoS2 Synaptic Array

AbstractNeuromorphic computing has attracted broad attention in recent years, owing to its break in the bottleneck of traditional von Neumann architecture with its advantages of high efficiency and low power consumption. However, many reported works only focused on the improvement of the performance of a single device, or assumed that each neuron in the artificial neural network (ANN) is ideal and equivalent, which cannot realize the application of device in hardware neural network (HNN). Herein, a top‐gate structured floating‐gate synapse (FG‐Synapse) array is fabricated. Wafer‐level molybdenum sulfide (MoS2 ) channel and hafnium oxide (HfO2 ) dielectric layer are grown by atomic layer deposition. Besides, long‐term potential/depression (LTP/LTD) performance of each device are explored, which showed high consistency. Finally, a three‐layer ANN and proposed as an efficient approximation method to realize the recognition of digital pictures. It is the first time that a 2D device array based on a floating gate structure applied to ANNs. The recognition rate of a single device reaches 92.2%, and the recognition rate of the device array reaches 91.3%. This work is of great significance to provide the possibility for the application of larger‐scale arrays in neuromorphic computing in the future.

Organizing Reliable Polymer Electrode Lines in Flexible Neural Networks via Coffee Ring-Free Micromolding in Capillaries.

With an increase in the demand for smart wearable systems, artificial synapse arrays for flexible neural networks have received considerable attention. A synaptic device with a two-terminal configuration is promising for complex neural networks because of its ability to scale to a crossbar array architecture. To realize practical crossbar arrays with a high density, it is essential to achieve reliable electrode lines that act as signal terminals. However, an effective method to develop intrinsically flexible signal lines in artificial neural networks has not been developed. In this study, we achieved reliable polymer signal lines for flexible neural networks using coffee ring-free micromolding in capillaries (MIMIC). In a typical MIMIC, the outward convective flow of the polymer solution inherently deteriorates the pattern fidelity. To achieve reliable conducting polymer (CP) lines, we precisely controlled the flow of the polymer solution in the MIMIC by inducing the Marangoni force. When the convective and Marangoni flows for the solution were balanced in the MIMIC, the CP line patterns were reliably produced with high fidelity. The developed CP lines exhibited superior conductivity and high mechanical flexibility. Moreover, flexible memristor arrays consisting of CP signal lines demonstrated a high potential for realizing practical neuromorphic systems linked to artificial intelligence.

Controlled Growth of Wafer-Scale Transition Metal Dichalcogenides with a Vertical Composition Gradient for Artificial Synapses with High Linearity.

Artificial synapses are promising for dealing with large amounts of data computing. Great progress has been made recently in terms of improving the on/off current ratio, the number of states, and the energy efficiency of synapse devices. However, the nonlinear weight update behavior of a synapse caused by the uncertain direction of the conductive filament leads to complex weight modulation, which degrades the delivery accuracy of information. Here we propose a strategy to improve the weight update behavior of synapses using chemical-vapor-deposition-grown transition metal dichalcogenides (TMDCs) with a vertical composition gradient, where the sulfur concentration decreases gradually along the thickness direction of TMDCs and thus forms a certain direction of the conduction filament for synapse devices. It is worth noting that the devices show an excellent linear conductance of potentiation and depression with a high linearity of 0.994 (surpassing most state-of-the-art synapses), have a large number of states, and are able to fabricate synapse arrays with wafer-scale. Furthermore, the devices based on the TMDCs with the vertical composition gradient exhibit an asymmetric feature of potentiation and depression behaviors with high linearity and follow the simulated linear Leaky ReLU function, resulting in a high recognition accuracy of 94.73%, which overcomes the unreliability issue in the Sigmoid function due to the vanishing gradient phenomenon. This study not only provides a universal method to grow TMDCs with a vertical composition gradient but also contributes to exploring highly linear synapses toward neuromorphic computing.

Impact of the leakage current of an AND-type synapse array on spiking neural networks

In this paper, we analyze the impact of the leakage current of an AND-type synapse array on a spiking neural network (SNN). In order to evaluate the firing rate modulation affected by the leakage current of a synapse array, the electrical characteristics of a fabricated dual-gate synapse are measured and modeled. A SPICE circuit simulation shows a severe change in the firing rate in the SNN, and a system-level simulation with MNIST datasets exhibits frequency-dependent performance degradation due to the leakage current. Compared to a case with negligible leakage current, a 6.98% accuracy loss at 50 kHz of input frequency, which is the firing rate of conventional CMOS neurons.

Systematic Engineering of Metal Ion Injection in Memristors for Complex Neuromorphic Computing with High Energy Efficiency

Neuromorphic electronics attract significant attention as a new computing architecture. Despite much effort for achieving practical neuromorphic systems, it is still challenging to construct a synapse array ideal for complex neural networks. Herein, a novel strategy for developing a highly integrated crossbar array of a one‐selector–one‐memory (1S–1R) synapse by systematically engineering ion injection is demonstrated. In the proposed synapse, an electrochemical metallization (ECM) memristor consisting of unstable filaments and a typical ECM device with stable filaments act as a selector with a low leakage current and a stable memory device, respectively. To overcome the voltage‐matching issues in constructing the 1S–1R synapse with high integration density, ion injection related with the electrical properties is optimized in the ECM devices via the distribution of active metal nanoparticles at the interface. The developed synapse possesses a high on/off ratio, superior selectivity, low operating current, and stable multilevel conductance, compared to the previously reported devices. High feasibility for complex neuromorphic systems is demonstrated, and the neural network based on the developed synapse array exhibits reliable parallel computation with high energy efficiency. This promising concept of realizing complex neuromorphic electronics is a fundamental building block for the practical artificial intelligence.

Back‐End‐of‐Line SiC‐Based Memristor for Resistive Memory and Artificial Synapse

AbstractTwo‐terminal memristor has emerged as one of the most promising neuromorphic artificial electronic devices for their structural resemblance to biological synapses and ability to emulate many synaptic functions. In this work, a memristor based on the back‐end‐of‐line (BEOL) material silicon carbide (SiC) is developed. The thin film memristors demonstrate excellent binary resistive switching with compliance‐free and self‐rectifying characteristics which are advantageous for the implementation of high‐density 3D crossbar memory architectures. The conductance of this SiC‐based memristor can be modulated gradually through the application of both DC and AC signals. This behavior is demonstrated to further emulate several vital synaptic functions including paired‐pulse facilitation (PPF), post‐tetanic potentiation (PTP), short‐term potentiation (STP), and spike‐rate‐dependent plasticity (SRDP). The synaptic function of learning‐forgetting‐relearning processes is successfully emulated and demonstrated using a 3 × 3 artificial synapse array. This work presents an important advance in SiC‐based memristor and its application in both memory and neuromorphic computing.

Analog synaptic devices applied to spiking neural networks for reinforcement learning applications

In this work, we implement hardware-based spiking neural network (SNN) using the thin-film transistor (TFT)-type flash synaptic devices. A hardware-based SNN architecture with synapse arrays and integrate-and-fire (I&F) neuron circuits is presented for executing reinforcement learning (RL). Two problems were used to evaluate the applicability of the proposed hardware-based SNNs to off-chip RL: the Cart Pole balancing problem and the Rush Hour problem. The neural network was trained using a deep Q-learning algorithm. The proposed hardware-based SNNs using the synapse model with measured characteristics successfully solve the two problems and show high performance, implying that the networks are suitable for executing RL. Furthermore, the effect of variations in non-ideal synaptic devices and neurons on the performance was investigated.

Ultralow‐Power and Multisensory Artificial Synapse Based on Electrolyte‐Gated Vertical Organic Transistors

AbstractBioinspired electronics have shown great potential in the field of artificial intelligence and brain‐like science. Low energy consumption and multifunction are key factors for its application. Here, multisensory artificial synapse and neural networks based on electrolyte‐gated vertical organic field‐effect transistors (VOFETs) are first developed. The channel length of the organic transistor is scaled down to 30 nm through cross‐linking strategy. Owing to the short channel length and extremely large capacitance of the electric double layer formed at the electrolyte–channel interface, the minimum power consumption of one synaptic event is 0.06 fJ, which is significantly lower than that required by biological synapses (1–10 fJ). Moreover, the artificial synapse can be trained to learn and memory images in a 5 × 5 synapse array and emulate the human brain's spatiotemporal information processing and sound azimuth detection. Finally, the artificial tongue is designed using the synaptic transistor that can discriminate acidity. Overall, this study provides new insights into realizing energy‐efficient artificial synapses and mimicking biological sensory systems.

Topochemical Synthesis of Copper Phosphide Nanoribbons for Flexible Optoelectronic Memristors (Adv. Funct. Mater. 14/2022)

Flexible Optoelectronic Memristors In article number 2110900, Cong Ye, Jiahong Wang, and co-workers present a topochemical method to transform crystalline phosphorus nanoribbons into copper phosphide nanoribbons (Cu3P NRs) with well-maintained morphology. The flexible optoelectronic memristor constructed with Cu3P NRs exhibit excellent performance, and the memristor-based synapse arrays mimic the memory backtracking and pattern recognition.