of large-scale network applications. As a solution to the bottleneck of the von Neumann architecture, neuromorphic computing, which mimics the human brain to perform complex computations through massively parallel computations, has been proposed. The human brain consists of more than 10 11 neurons and more than 10 14 synapses. The neurons and synapses are connected in parallel to perform memory, computation, reasoning, and learning simultaneously, even with a low power of ≈20 W. In particular, the human brain learns by reconstructing the connection strength between synapses, called synaptic plasticity. As a result, developing artificial synaptic devices capable of mimicking synaptic plasticity for the implementation of artificial neural networks (ANNs) is a critical challenge. Recently, it has been reported that memristors, which have multiple resistance states that can be continuously modulated by an external electrical stimulus, can mimic the function of biological synapses. [4] Various devices, such as resistive switching memory, [5][6][7][8] phase change memory, [9][10][11] ferroelectric memory, [12][13][14][15] and others, [16][17][18] have been proposed as candidates for high-efficiency and highperformance memristors. A common feature of these devices is that they mimic synaptic weights by expressing values between 0 and 1 in an analog form, as opposed to the current digitalbased information communication devices that only use 0 and 1. Among them, the HfZrO 2 (HZO)-based ferroelectric tunnel junction (FTJ) device controls the partial polarization reversal of a ferroelectric thin film to mimic the synaptic weight. Consequently, it implements a multi-resistance state between the high-resistance state (HRS) and low-resistance state (LRS). [19,20] In addition, synaptic characteristics can be secured using a two-terminal metal-semiconductor-metal (MFM) device structure, which has the advantage of being able to design for high integration of 4F 2 . In this study, we investigated the synaptic properties of the HZO FTJ device with an MFM structure for neuromorphic computing applications. HZO ferroelectric thin films with a non-perovskite structure are appropriate for ultra-thin three dimensional capacitors, because they exhibit ferroelectric properties even in a thin film close to 1 nm and have large bandgaps. Moreover, Ferroelectric doped-HfO 2 is considered an alternative to ferroelectric perovskites because of its full CMOS (complementary metal-oxide-semiconductor) process compatibility, high scalability and easy stabilization Owing to the limited processing speed and power efficiency of the current computing method based on the von Neumann architecture, research on artificial synaptic devices for implementing neuromorphic computing capable of parallel computation is accelerating. The potential application of artificial synapses composed of ferroelectric tunnel junctions based on metal-hafnium zirconium oxide-metal structure for neuromorphic computing is investigated. Multiple resistance levels are implemented t...

No abstract

Owing to the 4th Industrial Revolution, the amount of unstructured data, such as voice and video data, is rapidly increasing. Brain-inspired neuromorphic computing is a new computing method that can efficiently and parallelly process rapidly increasing data. Among artificial neural networks that mimic the structure of the brain, the spiking neural network (SNN) is a network that imitates the information-processing method of biological neural networks. Recently, memristors have attracted attention as synaptic devices for neuromorphic computing systems. Among them, the ferroelectric doped-HfO2-based ferroelectric tunnel junction (FTJ) is considered as a strong candidate for synaptic devices due to its advantages, such as complementary metal–oxide–semiconductor device/process compatibility, a simple two-terminal structure, and low power consumption. However, research on the spiking operations of FTJ devices for SNN applications is lacking. In this study, the implementation of long-term depression and potentiation as the spike timing-dependent plasticity (STDP) rule in the FTJ device was successful. Based on the measured data, a CrossSim simulator was used to simulate the classification of handwriting images. With a high accuracy of 95.79% for the Mixed National Institute of Standards and Technology (MNIST) dataset, the simulation results demonstrate that our device is capable of differentiating between handwritten images. This suggests that our FTJ device can be used as a synaptic device for implementing an SNN.

No abstract

No abstract