20

2025

-

03

Journal of Luminescence - Invited Review | Progress in the Research of Oxide All-Optical Synapses

Author:


Introduction
Brain-inspired Neuromorphic ComputingIt is expected to overcome the bottlenecks of traditional von Neumann computing architecture, achieve low power consumption and efficient information processing, and thus promote the development of artificial intelligence technology. Artificial synapses are the key hardware for building neuromorphic systems. Among them, optoelectronic synapses combine the advantages of electronics and photonics, and have multiple functions such as optical sensing, information calculation, and storage. In the emerging all-optical optoelectronic synapse, both the non-volatile increase and decrease of conductance are achieved through optical signals, which can effectively avoid the damage of electrical signals to the device microstructure, improve operating stability, and endow the synaptic device with new functions, which has important research value. Recently, the research group led by Researcher Zhuge Feiat the Ningbo Institute of Materials Technology and Engineering, Chinese Academy of Sciences, published a Journal of Luminescenceinvited review article titled Research Progress on Oxide All-Optical Synapses. The paper, with master's student Shan Hai as the first author, comprehensively and systematically reviews the research progress of oxide all-optical synaptic devices with long-term plasticity, discusses in detail the device structure, material selection, and photoelectric response mechanism of all-optical synapses based on optical wavelength and optical power control, and finally analyzes the challenges facing the development of current all-optical synapses.

Introduction

The conductance of synaptic devices is generally reversibly controlled using voltage or current, which is prone to ion migration and Joule heat in the device, thus facing problems such as high power consumption and poor stability. If light is used instead of electricity to control the synaptic conductance, changes in the device microstructure can be avoided, and Joule heat can be eliminated, thereby achieving low-power, high-stability all-optical synapses. In addition, all-optical synaptic devices are directly photosensitive, which helps to realize the integrated function of sensing, storage, and computation, and can be used to develop new-generation artificial vision chips. For all-optical synapses, the choice of photosensitive functional layer materials is crucial. Currently, many functional materials have been reported to have excellent photoresponse capabilities, including oxides (such as IGZO, ZnO, CeO x、TaO x、TiO 2), low-dimensional nanomaterials (such as graphene, selenides), organic materials, and perovskite materials. Among them, oxide materials have many application advantages. First, oxides can be prepared by various methods, such as magnetron sputtering, chemical vapor deposition, atomic layer deposition, etc., which provides convenience for the preparation of all-optical synapses; second, oxides are compatible with existing CMOS processes, which helps to achieve high-integration density synaptic arrays and chips, which is crucial for building large-scale neuromorphic computing systems; finally, oxides have good chemical stability and thermal stability, which helps to improve the long-term stability of devices. Based on this, researchers have tried to use oxide materials to construct all-optical synaptic devices to achieve all-optical synapses with excellent photoresponse ability and high stability, and explore their applications in neuromorphic computing, high-density storage, and optical sensors.

 

 

Wavelength-Controlled Oxide All-Optical Synapses

(1) Oxide All-Optical Memristors

In oxide all-optical memristors, it is common to control the device conductance by changing the wavelength of the irradiation light. The research team at the Ningbo Institute of Materials, Chinese Academy of Sciences, based on the oxygen-deficient InGaZnO (OD-IGZO)/oxygen-rich InGaZnO (OR-IGZO) homojunction, realized the all-optical memristor for the first time internationally, and built an in-sensor encryption architecture integrating image sensing, encryption calculation, and post-processing functions based on the InGaZnO all-optical memristor. Northeast Normal University developed a plasma all-optical memristor based on Ag-TiO2nanocomposites. Fudan University reported a NiO/TiO2all-optical self-rectifying memristor array with multilevel storage function under light stimulation.

Figure 1: Wavelength-controlled oxide all-optical memristors (a) Schematic diagram of all-optical mode of double-layer IGZO memristor; (b) Reversible conductance control of IGZO memristor under 420 nm blue light and 800 nm near-infrared light pulses; (c) In-situ encrypted neuromorphic vision sensor in IGZO memristor; (d) Ag-TiO2(e) Silver-Titanium2(f) NiO/TiO2(g) Nickel/Titanium2All-optical conductance control of memristor

ZnO is an important optoelectronic semiconductor material with advantages such as low deposition temperature and high thermal conductivity, and is widely used in optoelectronic devices, sensors, solar cells, etc. The Ningbo Institute of Materials, Chinese Academy of Sciences, prepared a single-layer ZnO all-optical memristor grown by pure argon magnetron sputtering. National Yang Ming Chiao Tung University, Taiwan, prepared a ZnO/Zn2Cerium4heterojunction all-optical memristor, and also reported an ITO/Ga2O3/ZnO/ITO structure all-optical memristor. Central South University developed a p-Si/ZnO heterojunction all-optical memristor. Northeast Normal University reported a ZnO/MoO+all-optical memristor that can achieve SET and RESET processes under 350 nm and 570 nm illumination, respectively. Jilin University designed a chlorophyll/ZnO heterojunction memristor that achieves conductance increase and decrease under 430 nm and 730 nm light stimulation, respectively.xFigure 2: Wavelength-controlled ZnO-based all-optical memristors (a) Single-layer ZnO

(b) ZnO/Zn (c) Schematic diagram of memristor structure; (c) Schematic diagram of Ga2Cerium4/ZnO memristor structure; (d) Schematic diagram of p-Si/ZnO memristor structure; (e) ZnO/MoO2O3(f) Schematic diagram of chlorophyll/ZnO heterojunction memristor structure+-Si/ZnO忆阻器结构示意图;(e) ZnO/MoOx(2) Oxide All-Optical Transistors

(2)氧化物全光控晶体管

In addition to oxide memristors, there are also all-optical synapses based on oxide transistors. National Cheng Kung University in Taiwan reported an IGZO/ZrOxall-optical transistor. The device achieved light enhancement and light suppression behavior under visible light and near-infrared light irradiation, respectively. Tianjin University used graphene/TiO2heterostructure to achieve an all-optical transistor. Under 635 nm red light stimulation, the conductance of the device increased; under 365 nm ultraviolet light irradiation, the conductance decreased. Central South University used WSe2/WOxheterostructure to fabricate a photoelectric transistor synapse, exhibiting all-optical behavior under red and ultraviolet light stimulation.

Figure 3: Wavelength-controlled ZnO-based all-optical memristor (a) All-optical control performance of single-layer ZnO memristor; (b) Schematic diagram of ZnO/Zn2SnO4 memristor; (c) Schematic diagram of Ga2O3/ZnO memristor structure; (d) Schematic diagram of p+-Si/ZnO memristor structure; (e) Schematic diagram of ZnO/MoOx memristor structure; (f) Schematic diagram of chlorophyll/ZnO heterojunction memristor structure; (g) Schematic diagram of WSe2/WOx transistor structure; (h) Current increase under 648 nm red light; (i) Current decrease under 375 nm ultraviolet light

(3) Others Two-terminal or three-terminal all-optical devices

In addition to memristors and transistors, some two-terminal or three-terminal devices similar in structure to memristors or transistors have also exhibited conductance all-optical control characteristics, mainly two-terminal structures. Shenzhen University fabricated an artificial synapse based on MAPbBr3/ZnO heterojunction, which can achieve long-term increase and decrease of device conductance under 365 nm ultraviolet light and 520 nm green light irradiation, respectively. Zhejiang University developed a ZnAlSnO/SnO heterojunction all-optical synapse for accurately simulating excitatory and inhibitory synaptic behavior. The device can achieve conductance enhancement and suppression under 635 nm and 532 nm light irradiation, respectively. In addition, Zhejiang University achieved a reconfigurable all-optical synapse based on the IGZO/SnO/SnS structure. This device integrates sensing, storage, and processing functions, and can exhibit synaptic excitation and inhibition under ultraviolet and visible light stimulation.

Figure 4: Wavelength-controlled oxide heterojunction all-optical synapse; (a) Schematic diagram of MAPbBr3/ZnO device structure; (b) Current increase under 365 nm ultraviolet light; (c) Current decrease under 520 nm green light; (d) Schematic diagram of ZnAlSnO/SnO device structure; (e) Current increase under 635 nm light; (f) Current decrease under 532 nm light; (g) Schematic diagram of IGZO/SnO/SnS device structure; (h) All-optical conductance modulation behavior

 

Power-controlled oxide all-optical synapse

The University of Strasbourg in France developed a ferroelectric material/graphene heterojunction transistor, where the ferroelectric layer is Pb[(Mg1/3Nb2/3)0.70Ti0.30]O3crystal. The device can achieve reversible control of device conductance under 365 nm ultraviolet light pulses. The all-optical control performance of the device comes from the polarization and depolarization of the oxide ferroelectric layer. At lower light intensity, free charges are generated in the ferroelectric layer, leading to depolarization; at higher light intensity, excess photogenerated carriers recombine with previously existing shielding charges, leading to polarization recovery.

Figure 5: Power-controlled oxide ferroelectric material/graphene all-optical transistor (a) Schematic diagram of the device structure; (b) Photoresponse performance under different light powers; (c) Reversible conductance control under different light powers

Summary and Outlook

Artificial synapses are key hardware for building neuromorphic systems. Among them, optoelectronic synapses combine the advantages of electronics and photonics, and have multiple functions such as optical sensing, information calculation, and storage. In emerging all-optical optoelectronic synapses, both non-volatile increase and decrease of conductance are achieved through optical signals, which can effectively avoid the damage of electrical signals to the device microstructure, improve operational stability, and endow the synaptic device with new functions. This is of great significance for building high-performance neuromorphic systems and promoting the development of artificial intelligence technology. Although some progress has been made in the research of oxide all-optical synapses, the development of this field is still in its initial stage, and faces many challenges in photosensitive materials, control mechanisms, device performance, light signal application, and functional applications. With the collaborative in-depth research of material systems, device fabrication, circuit architecture, and artificial intelligence algorithms, it is expected that in the near future, all-optical synapses will be used in next-generation neuromorphic computing systems such as brain-like vision chips, thereby promoting the informatization and intellectualization development of human society.

 

 

Paper Information

Dan Hai, Ying Hongwei, Cheng Peihong, Hu Lingxiang, Wang Jingrui, Ye Zhizhen, Zhuge Fei. Research Progress of Oxide All-Optical Synapses [J]. Journal of Luminescence, 2025, 46(02): 245-259. DOI: 10.37188/CJL.20240251

https://cjl.lightpublishing.cn/zh/article/doi/10.37188/CJL.20240251/