Photoelectric memristor based on Ga2O3 / ITZO heterojunction, preparation method and application thereof

Abstract

The invention discloses a photoelectric memristor based on a Ga2O3 / ITZO heterojunction, a preparation method and application. The photoelectric memristor comprises a bottom electrode layer, a Ga2O3 layer, an IZTO layer and a top electrode layer which are sequentially arranged from bottom to top. Wherein the Ga2O3 layer and the IZTO layer are used as heterostructure layers. And finally, a photoelectric memristor with a sandwich structure is formed. The preparation process is simple, large-scale preparation can be achieved, the performance is excellent and stable, the memristive effect is obvious under electric signals and multi-wavelength optical signals, the resistance value can change according to different voltage values or different illumination intensities, and the resistance value change is highly similar to nerve synapses of the human brain. Due to the characteristic, the photoelectric memristor can simulate the behavior of biological nerve synapse, and has the capability of a visual bionic system integrating perception and memory.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor electronic devices, and in particular to a Ga2O3 / ITZO heterojunction photoelectric memristor, a preparation method and applications thereof. Background Art

[0002] At present, complementary metal oxide semiconductor (CMOS) technology is still the core support of neural network algorithms, but its computing efficiency lags significantly behind that of the human brain, which is mainly attributed to the limitations of the von Neumann architecture. With the advent of the big data era and the rapid development of Internet of Things technology, traditional computer systems are facing the dual challenges of low efficiency and high energy consumption. In order to cope with these problems, simulating the architecture and operating principles of the human brain to achieve brain-like computing with strong parallel processing capabilities, high efficiency and low energy consumption has become an important strategy. Given that the human brain achieves learning and memory through a large number of neurons and their complex synaptic connections, the development of artificial synaptic devices that can simulate the functional characteristics and operating mechanisms of biological neural synapses has become a key measure to promote brain-like neural network computing from the hardware level.

[0003] In addition, in nature, most information is transmitted through optical signals, and completing the response to optical signals is the basis for realizing optical information decoding. The impact of light on the device is mainly to regulate the device through the impact of light on the resistive layer material. Finding new materials that can keenly sense and respond to optical signals is a hot research direction in materials science today. At present, photoresponsive materials are mainly divided into two categories: inorganic materials and organic materials. Inorganic materials include metal oxides, etc., and organic materials include perovskites, etc. However, there are many problems in the preparation or preservation of existing photoresponsive materials. For example, perovskite materials have strict requirements on storage conditions. Changes in environmental factors such as humidity and temperature may have irreversible effects on their performance, which undoubtedly increases the difficulty of material storage and transportation. These problems limit the promotion of photoelectric memristors and hinder the development of the optoelectronic industry. Therefore, it is of great significance to use simple materials through simple processing technology to prepare stable and easy-to-preserve optoelectronic devices. The realization of this goal is not only expected to reduce production costs and improve production efficiency, but also greatly broaden the application scope of photoresponsive materials, injecting new vitality into the vigorous development of optoelectronic devices.

[0004] Therefore, in order to overcome the above challenges, the existing technology needs to be improved and developed. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art and provide a preparation method and application of a Ga2O3 / ITZO heterojunction photoelectric memristor.

[0006] To achieve the above object, the technical solution of the present invention is:

[0007] In a first aspect, the present invention provides a Ga2O3 / ITZO heterojunction photoelectric memristor, comprising a bottom electrode layer, a Ga2O3 layer, an IZTO layer and a top electrode layer arranged in sequence from bottom to top; wherein the Ga2O3 layer and the IZTO layer serve as heterostructure layers.

[0008] Optionally, the Ga2O3 layer is in the form of a nanoscale thin film layer;

[0009] Alternatively, the IZTO layer is attached to the Ga2O3 layer in the form of a nanoscale thin film.

[0010] Optionally, the bottom electrode layer is made of one of ITO conductive glass, fluorine-doped tin dioxide conductive substrate or aluminum-doped zinc oxide conductive substrate.

[0011] Optionally, the top electrode layer is made of any one or more of gold, platinum, aluminum, copper, silver, and titanium.

[0012] Optionally, the thickness of the Ga2O3 layer and the IZTO heterojunction thin film layer is 170nm to 200nm;

[0013] Optionally, the thickness of the top electrode layer is 70nm-100nm; the thickness of the bottom electrode layer is 150nm.

[0014] In a second aspect, the present invention provides an application of a Ga2O3 / ITZO heterojunction photoelectric memristor, wherein the Ga2O3 / ITZO heterojunction photoelectric memristor realizes visual heterosynaptic bionics under two wavelengths of optical signals;

[0015] or,

[0016] The Ga2O3 / ITZO heterojunction photoelectric memristor is used for handwritten digit recognition based on neural network.

[0017] In a third aspect, the present invention provides a method for preparing a Ga2O3 / ITZO heterojunction photoelectric memristor, comprising the following steps:

[0018] Step 1: The bottom electrode layer is made of ITO conductive glass and pre-treated;

[0019] Step 2, preparing a Ga2O3 thin film by a magnetron sputtering method, attaching the Ga2O3 thin film layer to the bottom electrode layer, and obtaining a Ga2O3 thin film grown in situ on the surface of the bottom electrode layer;

[0020] Step 3, preparing an IZTO film by magnetron sputtering, and attaching the IZTO film to the Ga2O3 film to obtain a heterojunction film layer composed of Ga2O3 and IZTO;

[0021] Step 4: magnetron sputtering a metal top electrode on the heterojunction thin film layer to form a photoelectric memristor with multi-wavelength controllable performance.

[0022] Optionally, the pretreatment includes: firstly ultrasonically cleaning the bottom electrode layer with acetone, deionized water and anhydrous ethanol in sequence, and then drying.

[0023] Optionally, the magnetron sputtering in step 2 is performed under a working gas pressure of 0.5 Pa, a sputtering power of 90 to 110 W, and a deposition time of 10 to 20 min;

[0024] The magnetron sputtering in step 3 is carried out under the condition of a working gas pressure of 0.5 Pa, a sputtering power of 70 to 100 W, and a deposition time of 5 to 7 minutes;

[0025] The magnetron sputtering in step 4 is carried out under the condition of working pressure of 1Pa, sputtering power of 50-60W, and deposition time of 10-15min.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] Ga2O3 is a wide bandgap semiconductor, which allows it to absorb UV light efficiently. Due to its wide bandgap properties, Ga2O3 has a high absorption coefficient in the UV band, so it has been widely studied and used in applications such as UV photodetectors, UV photodiodes, and solar cells. In addition, Ga2O3 is relatively stable chemically and can maintain its semiconductor properties under high temperature environments, which makes it advantageous for applications under extreme environmental conditions. IZTO (indium zinc tin oxide) is a transparent conductive oxide material whose photoelectric properties can be adjusted by changing the ratio of indium, zinc, and tin. This adjustability allows IZTO to customize its conductivity and transmittance according to specific application requirements to optimize device performance. In addition, IZTO has high transparency and good conductivity. Using transparent IZTO electrodes in memristors can make the devices easier to integrate into transparent electronic devices. Moreover, the high mobility of IZTO helps to achieve memristive switching at low currents, which enables memristors to operate under low power consumption conditions, thereby extending the service life of the device. Ga2O3 and IZTO combine multiple advantages such as low cost, high carrier mobility, excellent light absorption ability, and stable thermal performance. Therefore, it has broad application prospects in memory, optoelectronic devices, and neuromorphic computing devices, and is expected to achieve large-scale industrial applications and further promote the development of related technologies.

[0028] Based on Ga2O3 / ITZO heterojunction photoelectric memristor, it can realize adjustable photoelectric synaptic function and visual bionic system with perception and memory. The proposed device successfully obtained the electrical synaptic function by changing the pulse interval of electrical stimulation. In addition, the device has a good light-induced response to multiple wavelengths such as deep ultraviolet, ultraviolet, blue, and green, and realizes the light-induced synaptic function. Since the synergistic effect of multiple optical signals can make the two-terminal device show a function similar to that of a three-terminal synapse (i.e., the facilitation of heterosynapses), based on this, the Pavlov classical conditioning experiment was simulated by applying a mixed signal of ultraviolet light (365nm) and blue light (470nm) at the same time and realized the visual heterosynaptic function. Heterosynaptic facilitation helps to improve the learning and memory ability of the device and has great potential in all aspects of neural networks and neuromorphic computing. In addition, an innovative architecture of a cross-array of photoelectric memristors is proposed, which combines the advantages of electrical modulation and 275nm / 365nm / 470nm / 550nm optical modulation memristors, and uses the slight difference in the forward conductance of the two types of memristors to accurately represent the weights of the neural network. This method expands the weight adjustment range, enhances the anti-interference ability, and improves the fault tolerance in the weight mapping process. Even in the presence of hardware non-idealities and random noise, it still maintains a high recognition rate, demonstrating its good robustness and potential in neuromorphic computing applications.

[0029] Finally, the materials used in the photoelectric memristor are environmentally friendly and simple, the preparation process used is simple, and it is easy to preserve, suitable for large-scale promotion, and can be widely used in fields such as information technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a schematic diagram of the structure of the photoelectric memristor of the present invention;

[0031] Figure 2 is a cross-sectional SEM characterization image of the photoelectric memristor of Example 2;

[0032] Figure 3 is a turn-on and erase curve diagram of the photoelectric memristor in Example 2;

[0033] Figure 4 This is a diagram showing the effect of double pulse facilitation or inhibition performance when the electrical signal of the photoelectric memristor in Example 2 is used for synapse simulation;

[0034] Figure 5 The photoresponses of the photoelectric memristor in Example 2 to deep ultraviolet (275 nm), ultraviolet light (365 nm), blue light (470 nm) and green light (550 nm) at different light intensities, respectively;

[0035] Figure 6is a diagram showing the double pulse facilitation effect when four optical signals of the photoelectric memristor in Example 2 are used for synapse simulation;

[0036] Figure 7 This is a diagram showing the application effect of the photoelectric memristor in Example 2 for visual heterosynaptic bionics.

[0037] Figure 8 The photoelectric memristor in Example 2 is used for handwritten digit recognition based on a neural network. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0039] The present disclosure is further described below in conjunction with the accompanying drawings and examples. It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those of ordinary skill in the art to which the present disclosure belongs.

[0040] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments in the present disclosure and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0041] In the technical solutions disclosed in one or more embodiments, Figure 1 As shown, the photoelectric memristor is composed of, from bottom to top: a bottom electrode layer ITO conductive glass, a Ga2O3 layer, an IZTO layer and a top electrode layer Ag; wherein the Ga2O3 layer and the IZTO layer between the top electrode and the bottom electrode serve as heterostructure layers.

[0042] Ga2O3 is a wide bandgap semiconductor, which allows it to absorb UV light efficiently. Due to its wide bandgap properties, Ga2O3 has a high absorption coefficient in the UV band, so it has been widely studied and used in applications such as UV photodetectors, UV photodiodes, and solar cells. In addition, Ga2O3 is relatively stable chemically and can maintain its semiconductor properties in high temperature environments, which makes it advantageous for applications under extreme environmental conditions. IZTO (indium zinc tin oxide) is a transparent conductive oxide material that can adjust its photoelectric properties by changing the ratio of indium, zinc, and tin. This adjustability allows IZTO to customize its conductivity and transmittance according to specific application requirements to optimize device performance. In addition, IZTO has high transparency and good conductivity. Using transparent IZTO electrodes in memristors can make devices easier to integrate into transparent electronic devices. In addition, the high mobility of IZTO helps to achieve memristive switching at low currents, which enables memristors to operate under low power consumption conditions, thereby extending the service life of the device. Ga2O3 and IZTO combine multiple advantages such as low cost, high carrier mobility, excellent light absorption ability and stable thermal performance. Therefore, they have broad application prospects in memory, optoelectronic devices and neuromorphic computing devices, and are expected to achieve large-scale industrial applications and further promote the development of related technologies.

[0043] Thus, by adopting Figure 1The Ga2O3 / ITZO heterojunction photoelectric memristor with the structure shown can realize adjustable photoelectric synaptic function and visual bionic system with perception and memory. The proposed device successfully obtains the electrical synaptic function by changing the pulse interval of electrical stimulation. In addition, the device has a good light-induced response to multiple wavelengths such as deep ultraviolet, ultraviolet, blue, and green, and realizes the light-induced synaptic function. Since the synergistic effect of multiple optical signals can make the two-terminal device show a function similar to that of a three-terminal synapse (i.e., the facilitation of heterosynapses), based on this, the Pavlov classical conditioning experiment was simulated by applying a mixed signal of ultraviolet light (365nm) and blue light (470nm) at the same time and realizing the visual heterosynaptic function. Heterosynaptic facilitation helps to improve the learning and memory ability of the device and has great potential in all aspects of neural networks and neuromorphic computing. In addition, an innovative architecture of a cross-array of photoelectric memristors is proposed, which combines the advantages of electrical modulation and 275nm / 365nm / 470nm / 550nm optical modulation memristors, and uses the slight difference in the forward conductance of the two types of memristors to accurately represent the weights of the neural network. This method expands the weight adjustment range, enhances the anti-interference ability, and improves the fault tolerance in the weight mapping process. Even in the presence of hardware non-idealities and random noise, it still maintains a high recognition rate, indicating its good robustness and potential in neuromorphic computing applications. Finally, based on the environmentally friendly and simple materials used in the photoelectric memristor, the preparation process used is simple and easy to preserve, suitable for large-scale promotion, and can be widely used in information technology and other fields.

[0044] In some embodiments, the Ga2O3 layer 3 is in the form of a nanoscale thin film layer;

[0045] In some embodiments, the IZTO layer 2 is in the form of a nano-scale thin film; the IZTO layer is attached to the Ga2O3 thin film layer in the form of a nano-scale thin film. The thickness of the Ga2O3 layer and the IZTO heterojunction thin film layer is 170nm to 200nm.

[0046] Optionally, the bottom electrode layer may be made of ITO conductive glass, fluorine-doped tin dioxide (FTO) conductive substrate or aluminum-doped zinc oxide (AZO) conductive substrate.

[0047] In this embodiment, it is preferred to use ITO as the conductive glass substrate because Ga2O3 can be grown more fully on the ITO film.

[0048] In some embodiments, the top electrode can be made of any one or more of gold, platinum, aluminum, copper, silver, and titanium. Preferably, it can be made of precious metals such as gold, silver, and platinum, which are not easily oxidized during use. The thickness of the Ga2O3 layer and the IZTO heterojunction thin film layer is 170nm to 200nm.

[0049] In some embodiments, the Ga2O3 / ITZO heterojunction photo-memristor can be applied to the following aspects:

[0050] Application 1: Visual heterosynaptic biomimetic:

[0051] Heterosynapses are widely present in various parts of the visual system, such as the retina and visual cortex, and play a key role in regulating human visual function. The process of heterosynaptic facilitation in synaptic devices was verified by the classic Pavlovian conditioned reflex experiment. In the experiment, "food vision" was used as an unconditioned stimulus (US), and "ring tone" was used as a neutral stimulus (NS). Light pulses with a wavelength of 365nm (as a pre-signal) and 275nm light pulses (as a modulation signal) were used to simulate "ring tone" and "food vision", respectively. Through the training process, NS was successfully converted into a conditioned stimulus (CS), indicating that heterosynaptic facilitation has achieved visual heterosynaptic bionics. Visual heterosynaptic bionics shows great potential in improving the learning and memory capabilities of neural networks, and is expected to promote the development of high-performance artificial visual memory systems.

[0052] Application 2: Digital handwriting recognition based on neural network

[0053] In traditional computer architecture, high-precision handwritten digit recognition usually requires a lot of computing resources and storage capacity. However, the unique photoelectric synaptic plasticity of memristors opens up new possibilities for implementing neural networks at the hardware level. To illustrate this capability, a handwritten digit recognition system based on convolutional neural networks was developed to implement neuromorphic computing functions in a memristor array, using the conductance values ​​of the memristors to replace the weight parameters in the CNN. During the training process, the conductance of the memristors is dynamically adjusted by external voltage or photoelectric signals to achieve learning and updating of network weights. By normalizing the conductance values, they can be accurately mapped to the weights in the neural network, which is conducive to efficient storage and updating of weights in hardware. This method not only simulates the synaptic plasticity of biological neural networks, but also significantly reduces energy consumption and computing resource requirements. In addition, an innovative architecture of a cross-array of photoelectric memristors is proposed, which combines the advantages of electrically modulated and 275nm / 365nm / 470nm / 550nm optically modulated memristors, and uses the subtle differences in the forward conductance of the two types of memristors to accurately represent the neural network weights. This method expands the weight adjustment range, enhances anti-interference ability, and improves fault tolerance in the weight mapping process. Even in the presence of hardware non-idealities and random noise, the use of optoelectronic modulation still maintains a high recognition rate, demonstrating its good robustness and potential in neuromorphic computing applications.

[0054] Example 1

[0055] The preparation method of the photoelectric memristor provided in this embodiment firstly grows a Ga2O3 thin film layer on an ITO conductive glass substrate by magnetron sputtering, then magnetron sputters an IZTO nano-thin film layer on the Ga2O3 thin film layer, and finally magnetron sputters an Ag electrode on the IZTO nano-thin film layer, thereby forming a photoelectric memristor with an "ITO / Ga2O3 / IZTO / Ag structure". The specific steps can be as follows:

[0056] Step 1: pre-treat the ITO conductive glass. The specific operation of the pre-treatment is: firstly ultrasonically clean the bottom electrode layer with acetone, deionized water and anhydrous ethanol in sequence, and then dry it.

[0057] Specifically, the specification of the ITO conductive glass is 2×2 cm, the ultrasonic cleaning time can be 5 min to 10 min, and the drying is performed at a temperature of 55° C. to 65° C. for 1 h to 2 h.

[0058] Step 2: Prepare Ga2O3 thin film by magnetron sputtering, and grow the Ga2O3 thin film layer on the bottom electrode layer:

[0059] Specifically, the preset target material is: a Ga2O3 ceramic target with a purity of 99.99%;

[0060] The Ga2O3 thin film was prepared by radio frequency magnetron sputtering process. The parameters of magnetron sputtering are as follows: working pressure is 0.5Pa; gas flow rate is 40sccm; sputtering power is 100W; sputtering time is 15min.

[0061] Step 3: Prepare an IZTO film by magnetron sputtering, and attach the IZTO film to the Ga2O3 film to obtain a Ga2O3 / IZTO heterojunction film layer:

[0062] Specifically, the preset target materials are: ZnO ceramic target and ITO ceramic target with a purity of 99.99%;

[0063] The IZTO film with a thickness of 100 nm was deposited by magnetron co-sputtering of two targets: ZnO ceramic target and ITO ceramic target. The parameters of magnetron sputtering are as follows: working pressure is 0.5 Pa; gas flow rate is 34 sccm; ZnO sputtering power is 100 W, ITO sputtering power is 70 W; sputtering time is 10 min.

[0064] Step 4: Prepare a Ag top electrode on the IZTO film using a DC magnetron sputtering method.

[0065] By direct current sputtering, a 40 nm thick Ag top electrode was deposited through a metal mask. The parameters of magnetron sputtering were as follows: working pressure was 0.5 Pa; gas flow rate was 30 sccm; sputtering energy was 60 W; and sputtering time was 10 min.

[0066] Example 2

[0067] The preparation method of the photoelectric memristor provided in this embodiment is completely the same as that of the embodiment 1 except that the sputtering time of the IZTO thin film prepared by the magnetron sputtering method in step 2 is adjusted from "10 min" to "6 min". The specific steps are as follows:

[0068] Step 1: pre-treat the ITO conductive glass. The specific operation of the pre-treatment is: firstly ultrasonically clean the bottom electrode layer with acetone, deionized water and anhydrous ethanol in sequence, and then dry it.

[0069] Step 2: Prepare Ga2O3 thin film by magnetron sputtering, and grow the Ga2O3 thin film layer on the bottom electrode layer:

[0070] Specifically, the preset target material is: a Ga2O3 ceramic target with a purity of 99.99%;

[0071] The Ga2O3 thin film was prepared by radio frequency magnetron sputtering process. The parameters of magnetron sputtering are as follows: working pressure is 0.5Pa; gas flow rate is 40sccm; sputtering power is 100W; sputtering time is 15min.

[0072] Step 3: Prepare an IZTO film by magnetron sputtering, and attach the IZTO film to the Ga2O3 film to obtain a Ga2O3 / IZTO heterojunction film layer:

[0073] Specifically, the preset target materials are: ZnO ceramic target and ITO ceramic target with a purity of 99.99%;

[0074] The IZTO film with a thickness of 80nm was deposited by magnetron co-sputtering of two targets: ZnO ceramic target and ITO ceramic target. The parameters of magnetron sputtering are as follows: working pressure is 0.5Pa; gas flow rate is 34sccm; ZnO sputtering power is 100W, ITO sputtering power is 70W; sputtering time is 6min.

[0075] Step 4: Prepare a Ag top electrode on the IZTO film using a DC magnetron sputtering method.

[0076] By direct current sputtering, a 40 nm thick Ag top electrode was deposited through a metal mask. The parameters of magnetron sputtering were as follows: working pressure was 0.5 Pa; gas flow rate was 30 sccm; sputtering energy was 60 W; and sputtering time was 10 min.

[0077] Figure 2 This is a cross-sectional SEM characterization image of the photoelectric memristor in Example 2. The thickness of IZTO is about 80 nm, and the thickness of Ga2O3 is about 100 nm.

[0078] Figure 3 The IV curve diagram of the opening and erasing of the photoelectric memristor in Example 2. When testing the device, the ITO conductive glass layer is grounded and a voltage is applied to the Ag electrode. In the figure, "1" refers to the curve of the scanning process from 0-1.5V, "2" refers to the curve of the scanning process from 1.5-0V; "3" refers to the curve that appears from the 0--1.5V scanning, and "4" refers to the curve that appears from the -1.5-0V scanning. In the "1" to "4" curves, "1" and "2" are the writing process of the memristor, which is equivalent to the state of information storage; "3" and "4" are the erasing process of the memristor, that is, the state of information erasure. Figure 4 It shows that the photoelectric memristor with the "ITO / Ga2O3 / IZTO / Ag" structure prepared in Example 2 exhibits typical non-volatile resistance conversion characteristics and self-rectification characteristics.

[0079] Figure 4 A pair of continuous electrical signal pulse stimulations were applied to both ends of the ITO / Ga2O3 / IZTO / Ag device prepared in Example 2, and the current value corresponding to each stimulation was recorded, successfully simulating the PPF / PPD function of the biological synapse.

[0080] Figure 5 The light response dependence of the ITO / Ga2O3 / IZTO / Ag photoelectric memristor prepared in Example 2 on deep ultraviolet (275nm), ultraviolet (365nm), blue light (470nm) and green light (550nm). It can be concluded that as the light intensity continues to increase, the response current of the device will gradually increase.

[0081] Figure 6 A pair of continuous light signal pulse stimulations are applied to both ends of the ITO / Ga2O3 / IZTO / Ag device in Example 2. The light signal pulses are respectively excited by deep ultraviolet (275nm), ultraviolet light (365nm), blue light (470nm) and green light (550nm), and the current value corresponding to each stimulation is recorded, successfully simulating the PPF function of biological synapses.

[0082] Figure 7This is an application effect diagram of the photoelectric memristor in Example 2 for visual heterosynaptic bionics. In neural networks, heterosynapses have a profound impact on high-level cognitive functions such as learning and memory, sensory information processing, and emotional regulation by regulating the interactions between multiple synapses. Focusing on the complete behavior of individual optical synapses, the synergistic effect of multiple light signals can make a two-terminal device exhibit functions similar to those of a three-terminal synapse (i.e., the facilitation of heterosynapses). In biology, heterosynapses refer to connections between neurons. Although they are not directly involved in the transmission of visual information, they can significantly affect the activity of visual neurons. These synapses are present in various components of the visual system, such as the retina and visual cortex, and have an important influence on the regulation of human visual function (such as Figure 7 In order to verify the process of heterosynaptic facilitation in the synaptic device, this embodiment conducted a classic Pavlovian conditioned reflex experiment. In the experiment, "food vision" was used as an unconditioned stimulus (US), and "ringing" was used as a neutral stimulus (NS). The ability of the memristor to achieve heterosynaptic facilitation under different light stimuli was studied in depth. Figure 7 As shown in b, a light pulse with a wavelength of 460 nm (as a pre-signal, a pulse width of 1 ms, and a pulse intensity of 120 μW / cm 2 ) and 365nm light pulses (as modulation signals, with a pulse width of 1ms and a pulse intensity of 120μW / cm 2 ) simulate “ring sound” (NS) and “food vision” (US), respectively. Figure 7 c shows that the photocurrent caused by neutral stimulation alone is 16nA, which is set as the threshold current; while the photocurrent caused by unconditioned stimulation is 26.7nA, which is significantly higher than the current of neutral stimulation. During the training phase, the simultaneous application of NS and US caused significant changes in the photocurrent. After the training, NS stimulation alone can induce a photocurrent of 18.3nA, which exceeds the threshold current. Therefore, NS was successfully transformed into a conditioned stimulus (CS), and visual heterosynaptic bionics was achieved through heterosynaptic facilitation. The above experimental results show that the ITO / Ga2O3 / IZTO / Ag structure successfully simulated Pavlov's classical conditioning experiment and achieved visual heterosynaptic bionics. Visual heterosynaptic bionics has great potential in improving neural networks and computing power, and is expected to significantly improve the learning and memory capabilities of devices, paving the way for the application of powerful artificial visual memory systems in the future.

[0083] Figure 8The photoelectric memristor in Example 2 is used for handwritten digit recognition based on neural network. In traditional computer architecture, high-precision handwritten digit recognition usually requires a lot of computing resources and storage capacity. However, the unique photoelectric synaptic plasticity of the memristor opens up new possibilities for implementing neural networks at the hardware level, demonstrating the huge computing potential of handwritten digit recognition. To illustrate this capability, a handwritten digit recognition system based on convolutional neural network (CNN) was developed, as shown in Figure 8 As shown in a. The CNN model consists of two convolutional layers, two pooling layers and one fully connected layer. The first layer has 32 channels and a size of 28×28; the second layer has 64 channels and a size of 14×14. The fully connected layer of the neural network contains 3136 input nodes, 1024 hidden nodes and 10 output nodes. The input nodes come from the handwritten digital images of the National Institute of Standards and Technology (MNIST) data set. In order to realize the neuromorphic computing function in the memristor array, the conductance value of the memristor is used to replace the weight parameters in the CNN. During the training process, the conductance of the memristor is dynamically adjusted by external voltage or photoelectric signal to realize the learning and updating of network weights. By normalizing the conductance value, it can be accurately mapped to the weight in the neural network, which is conducive to the efficient storage and updating of weights by hardware. This method not only simulates the synaptic plasticity of biological neural networks, but also significantly reduces energy consumption and computing resource requirements. As Figure 8 As shown in Figure 2, an innovative architecture of a cross-array of photoelectric memristors is proposed, which combines the advantages of electrical modulation and 275nm / 365nm / 470nm / 550nm optical modulation memristors. This design uses the slight difference in the forward conductance of the two types of memristors to accurately represent the weights of the neural network. By combining electrical modulation and optical modulation, this method expands the weight adjustment range, enhances anti-interference ability, and improves fault tolerance during weight mapping. Figure 8 c shows the relationship between the number of training iterations and the recognition rate under conditions such as ideal modulation, electrical modulation, and 275nm / 365nm / 470nm / 550nm photoelectric co-modulation. It can be seen that the recognition rate increases with the increase in the number of iterations under the six conditions. The results show that the memristor neural network still maintains a high recognition rate using photoelectric modulation in the presence of hardware non-idealities and random noise, indicating its good robustness and potential in neuromorphic computing applications. Figure 8 d is the highest recognition rate in 100 iterations, which is 98.40% in the ideal case, 95.69% / 97.13% / 95.08% / 91.72% under 275nm / 365nm / 470nm / 550nm photoelectric modulation, and 85.08% under pure electricity. Figure 8EI further demonstrated a strong correspondence between the predicted and actual output values ​​of the confusion matrix after 100 training cycles under electrical modulation and 275nm / 365nm / 470nm / 550nm optoelectronic co-modulation. The training process of the confusion matrix is ​​visually represented by the color gradient along its diagonal. Under pure electrical modulation, the matrix appears chaotic. However, after adding 275nm / 365nm / 470nm / 550nm optoelectronic co-modulation, the matrix transforms to a consistent color, indicating successful recognition of all patterns (0-9). This transformation reflects the reduction of recognition errors and the gradual learning of pattern recognition, emphasizing the enhancement of synaptic learning, achieving higher image recognition capabilities through optoelectronic co-modulation, and highlighting its potential for image recognition applications in neuromorphic computing systems.

[0084] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A Ga2O3 / ITZO heterojunction photoelectric memristor, characterized in that: It includes a bottom electrode layer, a Ga2O3 layer, an ITZO layer and a top electrode layer arranged in sequence from bottom to top; wherein the Ga2O3 layer and the ITZO layer serve as heterostructure layers.

2. The Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 1, characterized in that: The Ga2O3 layer is in the form of a nanoscale thin film layer; Alternatively, the ITZO layer is attached to the Ga2O3 layer in the form of a nanoscale thin film.

3. The Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 1, characterized in that: The bottom electrode layer is made of ITO conductive glass, fluorine-doped tin dioxide conductive substrate or aluminum-doped zinc oxide conductive substrate.

4. The Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 1, characterized in that: The top electrode is made of any one or more of gold, platinum, aluminum, copper, silver, and titanium.

5. The Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 1, characterized in that: The thickness of the Ga2O3 layer and the IZTO heterojunction thin film layer is 170nm-200nm.

6. The Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 1, characterized in that: The thickness of the top electrode layer is 70nm-100nm; the thickness of the bottom electrode layer is 150nm.

7. The application of the Ga2O3 / ITZO heterojunction photoelectric memristor according to any one of claims 1 to 6, characterized in that: The Ga2O3 / ITZO heterojunction photoelectric memristor realizes visual heterosynaptic bionics under two wavelengths of light signals; or, The Ga2O3 / ITZO heterojunction photoelectric memristor is used for handwritten digit recognition based on neural network.

8. A method for preparing a Ga2O3 / ITZO heterojunction photoelectric memristor, characterized in that: The steps include: Step 1: The bottom electrode layer is made of ITO conductive glass and pre-treated; Step 2, preparing a Ga2O3 thin film by a magnetron sputtering method, attaching the Ga2O3 thin film layer to the bottom electrode layer, and obtaining a Ga2O3 thin film grown in situ on the surface of the bottom electrode layer; Step 3, preparing an IZTO film by magnetron sputtering, and attaching the IZTO film to the Ga2O3 film to obtain a heterojunction film layer composed of Ga2O3 and IZTO; Step 4: magnetron sputtering a metal top electrode on the heterojunction thin film layer to form a photoelectric memristor with multi-wavelength controllable performance.

9. The method for preparing a Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 8, characterized in that: The pretreatment comprises: firstly ultrasonically cleaning the bottom electrode layer with acetone, deionized water and anhydrous ethanol in sequence, and then drying.

10. The method for preparing a Ga2O3 / ITZO heterojunction photoelectric memristor according to claim 8, characterized in that: The magnetron sputtering in step 2 is carried out under the condition of a working gas pressure of 0.5 Pa, a sputtering power of 90 to 110 W, and a deposition time of 10 to 20 minutes; The magnetron sputtering in step 3 is carried out under the condition of a working gas pressure of 0.5 Pa, a sputtering power of 70 to 100 W, and a deposition time of 5 to 7 minutes; The magnetron sputtering in step 4 is carried out under the condition of a working gas pressure of 1 Pa, a sputtering power of 50 to 60 W, and a deposition time of 10 to 15 minutes.

Images