Wafer‐Scale 2D Hafnium Diselenide Based Memristor Crossbar Array for Energy‐Efficient Neural Network Hardware

Advanced Materials, Journal Year: 2021, Volume and Issue: 34(25)

Published: Sept. 12, 2021

Memristor crossbar with programmable conductance could overcome the energy consumption and speed limitations of neural networks when executing core computing tasks in image processing. However, implementation array (CBA) based on ultrathin 2D materials is hindered by challenges associated large-scale material synthesis device integration. Here, a memristor CBA demonstrated using wafer-scale (2-inch) polycrystalline hafnium diselenide (HfSe2 ) grown molecular beam epitaxy, metal-assisted van der Waals transfer technique. The exhibits small switching voltage (0.6 V), low (0.82 pJ), simultaneously achieves emulation synaptic weight plasticity. Furthermore, enables artificial network high recognition accuracy 93.34%. Hardware multiply-and-accumulate (MAC) operation narrow error distribution 0.29% also demonstrated, power efficiency greater than 8-trillion operations per second Watt achieved. Based MAC results, hardware convolution processing can be performed kernels (i.e., soft, horizontal, vertical edge enhancement), which constitutes vital function for hardware.

Language: Английский

---

Reconfigurable neuromorphic memristor network for ultralow-power smart textile electronics

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Dec. 2, 2022

Neuromorphic computing memristors are attractive to construct low-power- consumption electronic textiles due the intrinsic interwoven architecture and promising applications in wearable electronics. Developing reconfigurable fiber-based is an efficient method realize that capable of neuromorphic function. However, previously reported artificial synapse neuron need different materials configurations, making it difficult multiple functions a single device. Herein, textile memristor network Ag/MoS

Language: Английский

---

High‐Performance Memristor Based on 2D Layered BiOI Nanosheet for Low‐Power Artificial Optoelectronic Synapses

Advanced Functional Materials, Journal Year: 2022, Volume and Issue: 32(25)

Published: April 3, 2022

Abstract Artificial optoelectronic synapses with both electrical and light‐induced synaptic behaviors have recently been studied for applications in neuromorphic computing artificial vision systems. However, the combination of visual perception high‐performance information processing capabilities still faces challenges. In this work, authors demonstrate a memristor based on 2D bismuth oxyiodide (BiOI) nanosheets that can exhibit bipolar resistive switching (RS) performance as well plasticity eminently suitable low‐power synapses. The fabricated exhibits RS high ON/OFF ratio up to 10 5 , an ultralow SET voltage ≈0.05 V which is one order magnitude lower than most reported memristors materials, low power consumption. Furthermore, demonstrates not only voltage‐driven long‐term potentiation, depression plasticity, paired‐pulse facilitation, but also short‐ plasticity. Moreover, photonic synapse be used simulate “learning experience” human brain. Consequently, BiOI shows ultra‐low consumption, provides new material strategy construct retina‐like sensors functions perceiving information.

Language: Английский

---

Porous crystalline materials for memories and neuromorphic computing systems

Chemical Society Reviews, Journal Year: 2023, Volume and Issue: 52(20), P. 7071 - 7136

Published: Jan. 1, 2023

This review highlights the film preparation methods and application advances in memory neuromorphic electronics of porous crystalline materials, involving MOFs, COFs, HOFs, zeolites.

Language: Английский

---

Simultaneous emulation of synaptic and intrinsic plasticity using a memristive synapse

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: May 19, 2022

Neuromorphic computing targets the hardware embodiment of neural network, and device implementation individual neuron synapse has attracted considerable attention. The emulation synaptic plasticity shown promising results after advent memristors. However, neuronal intrinsic plasticity, which involves in learning process through interactions with been rarely demonstrated. Synaptic occur concomitantly process, suggesting need simultaneous implementation. Here, we report a neurosynaptic that mimics single cell. Threshold switch phase change memory are merged threshold switch-phase device. Neuronal is demonstrated based on bottom layer, resembles modulation firing frequency biological neuron. also introduced nonvolatile switching top layer. Intrinsic simultaneously emulated cell to establish positive feedback between them. A loop retraining system implemented array for accelerated training.

Language: Английский

---

Low-Power Memristor Based on Two-Dimensional Materials

The Journal of Physical Chemistry Letters, Journal Year: 2022, Volume and Issue: 13(31), P. 7130 - 7138

Published: July 28, 2022

The memristor is an excellent candidate for nonvolatile memory and neuromorphic computing. Recently, two-dimensional (2D) materials have been developed use in memristors with high-performance resistive switching characteristics, such as high on/off ratios, low SET/RESET voltages, good retention endurance, fast speed, power energy consumption. Low-power are highly desired recent fast-speed energy-efficient artificial networks. This Perspective focuses on the progress of low-power based 2D materials, providing a condensed overview relevant developments memristive performance, physical mechanism, material modification, device assembly well potential applications. detailed research status has reviewed different from insulating hexagonal boron nitride, semiconducting transition metal dichalcogenides, to some newly materials. Furthermore, brief summary introducing perspectives challenges included, aim insightful guide this field.

Language: Английский

---

In‐Memory Computing using Memristor Arrays with Ultrathin 2D PdSeO_x/PdSe₂ Heterostructure

Advanced Materials, Journal Year: 2022, Volume and Issue: 34(26)

Published: April 8, 2022

In-memory computing based on memristor arrays holds promise to address the speed and energy issues of classical von Neumann system. However, stochasticity ions' transport in conventional oxide-based memristors imposes severe intrinsic variability, which compromises learning accuracy hinders implementation neural network hardware accelerators. Here, these challenges are addressed using a low-voltage array an ultrathin PdSeOx /PdSe2 heterostructure switching medium realized by controllable ultraviolet (UV)-ozone treatment. A distinctively different mechanism is revealed that can confine formation conductive filaments, leading remarkable uniform with low set reset voltage variability values 4.8% -3.6%, respectively. Moreover, convolutional image processing further implemented various crossbar kernels achieve high recognition ≈93.4% due highly linear symmetric analog weight update as well multiple conductance states, manifesting its potential beyond computing.

Language: Английский

---

High‐Performance Memristors Based on Ultrathin 2D Copper Chalcogenides

Advanced Materials, Journal Year: 2022, Volume and Issue: 34(9)

Published: Jan. 6, 2022

Copper chalcogenides represent a class of materials with unique crystal structures, high electrical conductivity, and earth abundance, are recognized as promising candidates for next-generation green electronics. However, their 2D structures the corresponding electronic properties have rarely been touched. Herein, series ultrathin copper chalcogenide nanosheets thicknesses down to two unit cells successfully synthesized, including layered Cu2 Te, well nonlayered CuSe Cu9 S5 , via van der Waals epitaxy, nonvolatile memristive behavior is investigated first time. Benefiting from highly active Cu ions low migration barriers, memristors based on crystals exhibit relatively small switching voltage (≈0.4 V), fast speed, uniformity, wide operating temperature range (from 80 420 K), stable retention good cyclic endurance. These results demonstrate tangible applications in future low-power, cryogenic, harsh

Language: Английский

---

Compact artificial neuron based on anti-ferroelectric transistor

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Nov. 17, 2022

Abstract Neuromorphic machines are intriguing for building energy-efficient intelligent systems, where spiking neurons pivotal components. Recently, memristive with promising bio-plausibility have been developed, but limited reliability, bulky capacitors or additional reset circuits. Here, we propose an anti-ferroelectric field-effect transistor neuron based on the inherent polarization and depolarization of Hf 0.2 Zr 0.8 O 2 film to meet these challenges. The intrinsic accumulated polarization/spontaneous films implements integration/leaky behavior neurons, avoiding external Moreover, exhibits low energy consumption (37 fJ/spike), high endurance (>10 12 ), uniformity stability. We further construct a two-layer fully ferroelectric neural networks that combines synapses, achieving 96.8% recognition accuracy Modified National Institute Standards Technology dataset. This work opens way emulate materials provides approach high-efficient neuromorphic hardware.

Language: Английский

---

Memristor-based neural networks: a bridge from device to artificial intelligence

Nanoscale Horizons, Journal Year: 2023, Volume and Issue: 8(6), P. 716 - 745

Published: Jan. 1, 2023

This paper reviews the research progress in memristor-based neural networks and puts forward future development trends.

Language: Английский

---