Emerging Optical In‐Memory Computing Sensor Synapses Based on Low‐Dimensional Nanomaterials for Neuromorphic Networks

Abstract

Emerging optical synapses with in‐memory computing sensor (IMCS) performance are considered to be one of the most effective candidates to circumvent the bottleneck of the current Von Neumann structure while developing neuromorphic systems with higher effectiveness and lower energy consumption. Biomimetic properties of optical IMCS synapses in function and form indicate the higher requirements for utilized functional materials, such as stronger optical sensitivity and lower energy dissipation. Because of properties with high optical‐sensitivity efficiency and excellent electrical conductivity, low‐dimensional nanomaterials have received tremendous interest in modulating optical‐induced synaptic plasticity and emulating optical‐triggered neuromorphic activity of optical IMCS synapses. Herein, a comprehensive summary of optical IMCS synapses based on low‐dimensional nanomaterials is introduced systematically for the first time, including 0D, 1D, and 2D materials. In addition, the content of biomimetic synaptic characteristics, materials classification, operation mechanism, and neuromorphic applications of optical IMCS synapses based on low‐dimensional nanomaterials are also summarized in this work. At last, the challenges and outlook related to artificial optical IMCS synapses with low‐dimensional nanomaterials are provided.

Flexible Memristor Devices Using Hybrid Polymer/Electrodeposited GeSbTe Nanoscale Thin Films

“…The multistate switching behavior is uniform and reproducible for 50 cycles, as demonstrated in Figure b. Such behavior suggests that our hybrid memristor devices have promising potential as candidates of artificial synapses for neuromorphic computing and multilevel data storage applications. − Such CC level-dependent control has been widely demonstrated in resistive switching memories and can be ascribed by the formation of filaments with different dimensions across the device. , Although similar multistate switching behavior was previously observed in our crossbar GeSbTe-based memory devices, the ON/OFF resistance ratio was smaller (by about 1 order of magnitude), and the CC was needed for both the SET and RESET processes to protect the devices. ,, …”

An all-two-dimensional Fe-FET retinomorphic sensor based on the novel gate dielectric In2Se3−xOx

“…The recognition accuracy exhibits enough fault-tolerance capability to apply to an artificial neural network for the visual system. 12…”

Neural-inspired artificial synapses based on low-voltage operated organic electrochemical transistors

“…Different types of LTP behaviors, such as spike timing-dependent plasticity (STDP) (Fig. 5e), spike ratedependent plasticity (SRDP), and spike number-dependent plasticity (SNDP), 145,146 have been reported in artificial synaptic devices.…”