Ionic Sieving Through One-Atom-Thick 2D Material Enables Analog Nonvolatile Memory for Neuromorphic Computing.

The first report on ion transport through atomic sieves of atomically thin 2D material is provided to solve critical limitations of electrochemical random-access memory (ECRAM) devices. Conventional ECRAMs have random and localized ion migration paths; as a result, the analog switching efficiency is inadequate to perform in-memory logic operations. Herein ion transport path scaled down to the one-atom-thick (≈0.33 nm) hexagonal boron nitride (hBN), and the ionic transport area is confined to a small pore (≈0.3 nm2 ) at the single-hexagonal ring. One-atom-thick hBN has ion-permeable pores at the center of each hexagonal ring due to weakened electron cloud and highly polarized B-N bond. The experimental evidence indicates that the activation energy barrier for H+ ion transport through single-layer hBN is ≈0.51 eV. Benefiting from the controlled ionic sieving through single-layer hBN, the ECRAMs exhibit superior nonvolatile analog switching with good memory retention and high endurance. The proposed approach enables atomically thin 2D material as an ion transport layer to regulate the switching of various ECRAM devices for artificial synaptic electronics.

Excellent synaptic behavior of lithium-based nano-ionic transistor based on optimal WO2.7 stoichiometry with high ion diffusivity.

In this study, we introduce a lithium (Li) ion-based three-terminal (3-T) synapse device using WO x as a channel. Our study reveals a key stoichiometry of WO2.7 for excellent synaptic characteristics that is related to Li-ion diffusivity. The open-lattice structure formed by oxygen deficiency promoted Li-ion injection and diffusion. The optimized stoichiometry and improved Li-ion diffusivity were confirmed by x-ray photoelectron spectroscopy analysis and cyclic voltammetry, respectively. Furthermore, the transient conductance change that inevitably occurs in ion-based synaptic transistors was resolved by applying a two-step voltage pulse scheme. As a result, we achieved a symmetric and linear weight-update characteristic with reduced program/erase operation time.

Understanding of proton induced synaptic behaviors in three-terminal synapse device for neuromorphic systems.

In this study, we investigate a proton-based three-terminal (3-T) synapse device to realize linear weight-update and I-V linearity characteristics for neuromorphic systems. The conductance states of the 3-T synapse device can be controlled by modulating the proton concentration in the WOx channel. Therefore, we estimate the dynamic change of proton concentration in the channel region, which directly affects synaptic behaviors. Our findings indicate that the supply of an excess number of protons from the SiO2-H electrolyte and low proton diffusivity in the WOx channel result in asymmetric and non-linear weight-update characteristics. In addition, though the linear I-V characteristics can be obtained using non-stoichiometric WOx, we observe that significant oxygen deficiency in the channel region increases the operating current levels. Thus, based on this information, we introduce optimized conditions of each component in the 3-T synapse device and shape of the gate voltage pulses. As a result, an excellent classification accuracy is achieved using linear weight-update and I-V linearity characteristics under optimized device and pulse conditions.

Strategies to Improve the Synaptic Characteristics of Oxygen-Based Electrochemical Random-Access Memory Based on Material Parameters Optimization.

Oxygen-based electrochemical random-access memories (O-ECRAMs) are promising synaptic devices for neuromorphic applications because of their near-ideal synaptic characteristics and compatibility with complementary metal-oxide-semiconductor processes. However, the correlation between material parameters and synaptic properties of O-ECRAM devices has not yet been elucidated. Here, we propose the critical design parameters to fabricate an ideal ECRAM device. Based on the experimental data and simulation results, it is revealed that consistent ion supply from the electrolyte and rapid ion diffusion in the channel are critical factors for ideal synaptic characteristics. To optimize these parameters, crystalline WO2.7 exhibiting fast ion diffusivity and ZrO1.7 exhibiting an appropriate ion conduction energy barrier (1.1 eV) are used as a channel and an electrolyte, respectively. As a result, synaptic characteristics with near-ideal weight-update linearity in the nanosiemens conductance range are achieved. Finally, a selector-less O-ECRAM device is integrated into a 2 × 2 array, and high recognition accuracy (94.83%) of the Modified National Institute of Standards and Technology pattern is evaluated.

Linear Frequency Modulation of NbO2-Based Nanoscale Oscillator With Li-Based Electrochemical Random Access Memory for Compact Coupled Oscillatory Neural Network.

Oscillatory neural network (ONN)-based classification of clustered data relies on frequency synchronization to injected signals representing input data, showing a more efficient structure than a conventional deep neural network. A frequency tunable oscillator is a core component of the network, requiring energy-efficient, and area-scalable characteristics for large-scale hardware implementation. From a hardware viewpoint, insulator-metal transition (IMT) device-based oscillators are attractive owing to their simple structure and low power consumption. Furthermore, by introducing non-volatile analog memory, non-volatile frequency programmability can be obtained. However, the required device characteristics of the oscillator for high performance of coupled oscillator have not been identified. In this article, we investigated the effect of device parameters of IMT oscillator with non-volatile analog memory on coupled oscillators network for classification of clustered data. We confirmed that linear conductance response with identical pulses is crucial to accurate training. In addition, considering dispersed clustered inputs, a wide synchronization window achieved by controlling the hold voltage of the IMT shows resilient classification. As an oscillator that satisfies the requirements, we evaluated the NbO2-based IMT oscillator with non-volatile Li-based electrochemical random access memory (Li-ECRAM). Finally, we demonstrated a coupled oscillator network for classifying spoken vowels, achieving an accuracy of 85%, higher than that of a ring oscillator-based system. Our results show that an NbO2-based oscillator with Li-ECRAM has the potential for an area-scalable and energy-efficient network with high performance.

Neural Network Training Acceleration With RRAM-Based Hybrid Synapses.

Hardware neural network (HNN) based on analog synapse array excels in accelerating parallel computations. To implement an energy-efficient HNN with high accuracy, high-precision synaptic devices and fully-parallel array operations are essential. However, existing resistive memory (RRAM) devices can represent only a finite number of conductance states. Recently, there have been attempts to compensate device nonidealities using multiple devices per weight. While there is a benefit, it is difficult to apply the existing parallel updating scheme to the synaptic units, which significantly increases updating process's cost in terms of computation speed, energy, and complexity. Here, we propose an RRAM-based hybrid synaptic unit consisting of a "big" synapse and a "small" synapse, and a related training method. Unlike previous attempts, array-wise fully-parallel learning is possible with our proposed architecture with a simple array selection logic. To experimentally verify the hybrid synapse, we exploit Mo/TiOx RRAM, which shows promising synaptic properties and areal dependency of conductance precision. By realizing the intrinsic gain via proportionally scaled device area, we show that the big and small synapse can be implemented at the device-level without modifications to the operational scheme. Through neural network simulations, we confirm that RRAM-based hybrid synapse with the proposed learning method achieves maximum accuracy of 97 %, comparable to floating-point implementation (97.92%) of the software even with only 50 conductance states in each device. Our results promise training efficiency and inference accuracy by using existing RRAM devices.

Near ideal synaptic functionalities in Li ion synaptic transistor using Li3POxSex electrolyte with high ionic conductivity.

All solid-state lithium-ion transistors are considered as promising synaptic devices for building artificial neural networks for neuromorphic computing. However, the slow ionic conduction in existing electrolytes hinders the performance of lithium-ion-based synaptic transistors. In this study, we systematically explore the influence of ionic conductivity of electrolytes on the synaptic performance of ionic transistors. Isovalent chalcogenide substitution such as Se in Li3PO4 significantly reduces the activation energy for Li ion migration from 0.35 to 0.253 eV, leading to a fast ionic conduction. This high ionic conductivity allows linear conductance switching in the LiCoO2 channel with several discrete nonvolatile states and good retention for both potentiation and depression steps. Consequently, optimized devices demonstrate the smallest nonlinearity ratio of 0.12 and high on/off ratio of 19. However, Li3PO4 electrolyte (with lower ionic conductivity) shows asymmetric and nonlinear weight-update characteristics. Our findings show that the facilitation of Li ionic conduction in solid-state electrolyte suggests potential application in artificial synapse device development.