Neuromorphic computing is a promising alternative to conventional systems as it could enable parallel computation and adaptive learning process. However, the development of energy efficient neuromorphic hardware has been hindered by limited performance analog synaptic devices. Here, we demonstrate conductance modulation behavior in ferroelectric thin-film transistors (FeTFT) that have nanoscale material oxide semiconductors. Accurate control polarization changes layer induces linear...

The development of electronic devices possessing the functionality biological synapses is a crucial step toward replicating capabilities human brain. Of various materials that have been used to realize artificial synapses, renewable natural advantages being abundant, inexpensive, biodegradable, and ecologically benign. In this study, we report biocompatible synapse based on matrix biopolymer ι-carrageenan (ι-car), which exploits Ag dynamics. This emulates short-term plasticity (STP),...

A number of synapse devices have been intensively studied for the neuromorphic system which is next-generation energy-efficient computing method. Among these various types devices, photonic recently attracted significant attention. In particular, using persistent photoconductivity (PPC) phenomena in oxide semiconductors are receiving much attention due to similarity between relaxation characteristics PPC and Ca2+ dynamics biological synapses. However, limitations its controllability...

A unique three-dimensional integration strategy is provided for high-performance, ultrahigh-density ferroelectric memory.

Abstract Hardware-based neural networks (NNs) can provide a significant breakthrough in artificial intelligence applications due to their ability extract features from unstructured data and learn them. However, realizing complex NN models remains challenging because different tasks, such as feature extraction classification, should be performed at memory elements arrays. This further increases the required number of arrays chip size. Here, we propose three-dimensional ferroelectric NAND (3D...

Convolutional neural networks (CNNs) have gained much attention because they can provide superior complex image recognition through convolution operations. Convolution processes require repeated multiplication and accumulation operations, which are difficult tasks for conventional computing systems. Compute-in-memory (CIM) that uses parallel data processing is an ideal device structure CIM based on two-terminal synaptic devices with a crossbar has been developed, but unwanted leakage current...

This paper reviews recent developments in artificial synapses that exploit various emerging memory devices. The emulation of synaptic plasticity and operation mechanism using materials structures are presented.

Neuromorphic computing that mimics the biological brain has been demonstrated as a next-generation method due to its low power consumption and parallel data processing characteristics. To realize neuromorphic computing, diverse neural networks such deep (DNNs) spiking (SNNs) have introduced. DNNs require artificial synapses analog conductance modulation characteristics, whereas SNNs characteristics controlled by temporal relationships between signals, so development of multifunctional...

High-performance optoelectronic synaptic transistors are reported with a long-term memory by using organic–inorganic halide perovskites and oxide semiconductors.

Artificial synapse is the basic unit of a neuromorphic computing system. However, there need to explore suitable synaptic devices for emulation dynamics. This study demonstrates photonic multimodal device by implementing perovskite quantum dot charge-trapping layer in organic poly(3-hexylthiophene-2,5-diyl) (P3HT) channel transistor. The proposed presents favorable band alignment that facilitates spatial separation photogenerated charge carriers. serves as basis optically induced trapping,...

A low-voltage, low-power CMOS delay element is proposed. With a unit inverter load, from 2.6 ns to 76.3 ms achieved in 0.8 /spl mu/m technology. Based on thyristor concept, the value of proposed can be varied over wide range by control current. The inherent advantage low voltage domains enables this work down supply 1 V while threshold nMOS and pMOS transistors are 840 mV -770 mV, respectively. designed less sensitive temperature variation than RC-based or inverter-based elements....

The resistive switching behaviors of AlO_X/HfO_X bilayer structures were investigated.

Low power consumption is the important requirement in memory devices for saving energy. In particular, improved energy efficiency essential implantable electronic operation under a limited supply. Here, we demonstrate use of κ-carrageenan (κ-car) as resistive switching layer to achieve that has low consumption. A carboxymethyl (CM) group introduced κ-car increase its ionic conductivity. Ag was doped CM:κ-car improve properties devices. Memory based on Ag-doped showed electroforming-free...

Hafnia-based ferroelectric thin-film transistors (FeTFTs) are regarded as promising candidates for future nonvolatile memory devices owing to their low power consumption, high operational speed, and complementary metal–oxide–semiconductor compatibility. However, the scalability of hafnia-based materials feasibility three-dimensional (3D) device fabrication should be confirmed ultrahigh-density applications. In this work, we demonstrate that FeTFTs can scaled down a 10-nm dimension using...

This paper presents synaptic transistors that show long-term weight modulation <italic>via</italic> injection of ions. Linear and symmetric update is achieved, which enables high recognition accuracy in artificial neural networks.

Abstract Flexible threshold switch devices are essential for low‐power and high‐speed semiconductor devices. Especially, bidirectional has been regarded as the ideal switching device ultrahigh‐density crosspoint memory Here, a flexible Pt/Ag‐doped ZnO/Pt on plastic substrate synthesized by electrochemical bottom‐up deposition is introduced. The behavior with ultralow off‐current, high selectivity (≈10 7 ), super‐steep slope. related to migration of silver ions form Ag filament. This shows...

Programmable memory characteristics of electrodeposited CuOx-based resistive random access (ReRAM) can be significantly improved by adopting a bilayer structure with built-in current limiter. To control the on-current and enhance device uniformity, Pt/CuOx (switching layer)/CuOx (current limiter)/Pt is proposed. This synthesized controlling solution pH during electrochemical deposition (ECD). The (synthesized at 9)/CuOx 11.5)/Pt exhibits reliable uniform self-compliant switching behavior....

Zirconium-doped hafnium oxide (HfZrOx) is one of the promising ferroelectric materials for next-generation memory applications. To realize high-performance HfZrOx applications, formation defects in HfZrOx, including oxygen vacancies and interstitials, needs to be optimized, as it can affect polarization endurance characteristics HfZrOx. In this study, we investigated effects ozone exposure time during atomic layer deposition (ALD) process on 16-nm-thick films showed different depending time....

Oxide semiconductors are promising channel materials for hafnia-based ferroelectric transistor memories because they can constrain the formation of an unwanted interfacial layer that deteriorate stability device. A major obstacle is limited memory window, originating from insufficient polarization switching <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${n}$ </tex-math></inline-formula> -type oxide cannot...

Films of CeO 2 were deposited by atomic layer deposition ( ALD ) using a Ce (mmp) 4 [mmp = 1‐methoxy‐2‐methyl‐2‐propanolate] precursor and H O reactant. The growth characteristics film properties investigated. process produced highly pure, stoichiometric films with polycrystalline cubic phases. Using the process, effects doping into an HfO gate dielectric systematically Regardless /( + Hf composition, all x 1− exhibited constant rates approximately 1.3 Å/cycle, which is essentially identical...

The aim of the neuromorphic computing is to emulate energy-efficient and smart data-processing ability biological brain, which achieved by massively interconnected neurons synapses. strength a connection between two modified homosynaptic heterosynaptic plasticity. As current research in device mainly focused on emulating plasticity, complex functions are not easy mimic because they require both We demonstrate use liquid crystal–carbon nanotube (LC–CNT) composite as resistive switching...

State-of-the-art deep convolutional neural networks (CNNs) are widely used in current AI systems, and achieve remarkable success image/speech recognition classification. A number of recent efforts have attempted to design custom inference engine based on various approaches, including the systolic architecture, near memory processing, processing-in-memory (PIM) approach with emerging technologies such as resistive random access (RRAM). However, a comprehensive comparison these approaches...