A bionic optoelectronic reservoir neuromorphic device and its preparation method

Abstract

The present invention discloses a bionic optoelectronic reservoir neuromorphic device and a preparation method thereof. The optoelectronic reservoir neuromorphic device includes: a substrate; a seed layer formed on the substrate; an isolation layer formed on the seed layer; an oxide nanowire network formed on the surface of the isolation layer, serving as an electronic reservoir, constructing neuron nodes in the reservoir by utilizing the randomly arranged characteristics of the grown oxide nanowires, and realizing randomly connected reservoir neurons similar to the human brain by means of the memristive behavior of the oxide nanowires; semiconductor quantum dots with optoelectronic response, grown on the oxide nanowire network, modifying the device to expand the response of the device from pure electricity to optoelectronic response, and realizing the perception function of optical signals; and a working electrode grown on the nanowire network.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and particularly relates to a bionic optoelectronic reservoir neuromorphic device and a preparation method thereof. Background Art

[0002] Traditional computers are all based on the von Neumann architecture. Limited by the separated design of the storage chip and the computing chip, the access and computing processing of data both require frequent data transmission between the two, making it difficult to achieve breakthroughs in speed and power consumption during the computing process. In order to further improve the computing processing speed of data, it is necessary to develop new computing hardware to help break through the traditional von Neumann computing bottleneck.

[0003] Compared with the working mode of traditional computers, neuromorphic computing based on artificial neural networks takes weight iterative computing as the core and has obvious advantages in processing a large amount of data. However, the cumbersome weight update and the connection of multiple nodes greatly increase the computing difficulty. The weights of each connection layer need to be updated during each training cycle to meet the accuracy and precision requirements of the computing results, increasing the computing energy consumption and cost. Different from traditional artificial neural networks, reservoir computing is an efficient bionic neural network computing mode that only needs to train the output layer. It is more convenient and efficient than traditional neural network computing and has great cost advantages and efficiency advantages in time series prediction, pattern recognition, etc.

[0004] In order to realize reservoir computing, it is necessary to develop corresponding integrated circuit chips to provide hardware guarantee. In order to be closer to the state of randomly connected neuron nodes in a real reservoir, it is necessary to construct a non-linear dynamic physical system and a supporting physical device and material system. However, traditional reservoir computing devices mostly use memristive devices with electrical excitation memory effects, which can only meet the computing requirements and still cause huge energy consumption problems during the process of receiving data from sensor device units. Therefore, developing an optoelectronic reservoir neuromorphic device with optoelectronic sensing function is of great significance for the development of new reservoir computing. Summary of the Invention

[0005] The present invention discloses a bionic optoelectronic reservoir neuromorphic device and a preparation method thereof. By using the random growth process of oxide nanowires to construct a naturally randomly distributed reservoir neuron, realizing the storage and computing of information by means of the memristive function of oxides, and using semiconductor quantum dot modification to expand the optoelectronic information sensing function of the device, it is used to construct an optoelectronic sensing, storage, and computing integrated reservoir computing system.

[0006] A method for preparing a bionic optoelectronic reservoir neuromorphic device includes the following steps: forming a seed layer on a substrate; performing in-situ oxidation treatment on the seed layer to form an isolation layer; growing an oxide nanowire network on the surface of the isolation layer as an electronic reservoir, constructing neuron nodes in the reservoir by using the randomly arranged characteristics of the grown oxide nanowires, and realizing randomly connected reservoir neurons similar to the human brain by means of the memristive behavior of the oxide nanowires; growing semiconductor quantum dots with optoelectronic responses on the oxide nanowire network to modify it, expanding the response of the device from pure electricity to optoelectronic response, and realizing the perception function of optical signals; growing a working electrode on the nanowire network to obtain a bionic optoelectronic reservoir neuromorphic device.

[0007] In the method for preparing a bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the seed layer is Ni, Ti, Zn, Cu or Mo.

[0008] In the method for preparing a bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the oxide nanowire network is nickel oxide, titanium oxide, zinc oxide, copper oxide or nickel oxide.

[0009] In the method for preparing a bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the specific steps for growing the oxide nanowire network include: applying a voltage of 10 - 100 V to the seed layer, then annealing the sample at an annealing temperature of 400 - 600 °C for an annealing duration of 3 - 6 hours to form a randomly distributed and grown oxide nanowire network.

[0010] In the method for preparing a bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the semiconductor quantum dots are MoS2, PtS2, ZnS, CdS, CdSe or PtSe2 quantum dots.

[0011] A bionic optoelectronic reservoir neuromorphic device includes: a substrate; a seed layer formed on the substrate; an isolation layer formed on the seed layer; an oxide nanowire network formed on the surface of the isolation layer as an electronic reservoir, constructing neuron nodes in the reservoir by using the randomly arranged characteristics of the grown oxide nanowires, and realizing randomly connected reservoir neurons similar to the human brain by means of the memristive behavior of the oxide nanowires; semiconductor quantum dots with optoelectronic responses grown on the oxide nanowire network to modify it, expanding the response of the device from pure electricity to optoelectronic response, and realizing the perception function of optical signals; a working electrode grown on the nanowire network.

[0012] In the bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the seed layer is Ni, Ti, Zn, Cu or Mo.

[0013] In the bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the oxide nanowire network is nickel oxide, titanium oxide, zinc oxide, copper oxide or nickel oxide.

[0014] In the bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the semiconductor quantum dots are MoS2, PtS2, ZnS, CdS, CdSe or PtSe2 quantum dots.

[0015] In the bionic optoelectronic reservoir neuromorphic device of the present invention, preferably, the substrate is a highly doped silicon wafer with silicon dioxide grown on it.

[0016] Beneficial effects:

[0017] (1) It breaks the traditional von Neumann computing architecture, introduces reservoir neuromorphic computing to achieve in-memory computing and intelligent computing efficacy. Data does not need to be frequently read between different units, saving a large amount of time and energy consumption. When performing big data calculations, it has lower power consumption and higher efficiency.

[0018] (2) Oxide nanowires are used to fabricate the bionic reservoir neuromorphic device. By means of the natural random stacking of oxide nanowires during the growth process, the construction of random nodes of neurons in the bionic reservoir is realized, and the weight update of reservoir neurons is achieved by using the memristive function of the oxide nanowires themselves.

[0019] (3) By introducing semiconductor quantum dots to modify the oxide nanowires, the response of the reservoir neuromorphic device that only responds to electrical excitation is extended to the response range of optical excitation, realizing the integration of the reservoir from in-memory computing function to sense-in-memory computing integrated function, greatly improving the signal perception ability of the neuromorphic device and reducing the energy consumption of the system. Description of the drawings

[0020] Figure 1 It is a flowchart of the preparation method of the bionic optoelectronic reservoir neuromorphic device.

[0021] Figures 2 to 7 It is a schematic structural diagram of each stage of the preparation method of the bionic optoelectronic reservoir neuromorphic device. Detailed implementation manners

[0022] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. The described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "vertical", "horizontal", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0024] In addition, many specific details of the present invention are described below, such as the structure, materials, dimensions, processing techniques and technologies of the device, in order to understand the present invention more clearly. However, as those skilled in the art can understand, the present invention can be implemented without these specific details. Unless specifically pointed out below, each part of the device can be made of materials well-known to those skilled in the art, or materials with similar functions developed in the future can be used.

[0025] Figure 1 is a flowchart of a method for preparing a bionic optoelectronic reservoir neuromorphic device. As Figure 1 shown, the method for preparing a bionic optoelectronic reservoir neuromorphic device includes the following steps:

[0026] Step S1, as Figure 2 shown, use a highly doped silicon wafer 100 with 100 nm of silicon dioxide 101 grown on it as the preparation substrate for the bionic optoelectronic reservoir neuromorphic device, and ultrasonically clean it with acetone, ethanol and deionized water respectively. The ultrasonic cleaning time is 10 - 30 minutes.

[0027] Step S2, use physical vapor deposition, electron beam evaporation, thermal evaporation or other methods to grow a metal nickel seed layer 102 with a thickness of 50 - 100 nm on the surface of the silicon dioxide 101. The obtained structure is as Figure 3 shown. The seed layer can be selected from Ni, Ti, Zn, Cu, Mo, etc.

[0028] Step S3, in-situ oxidation treatment is performed on the Ni seed layer 102 by means of rapid thermal annealing, vacuum annealing furnace, oxygen Plasma treatment, etc., to obtain a nickel oxide thin film as the isolation layer 103, which is used to prevent leakage current from the top device downward, as Figure 4 shown. The processing gas atmosphere for annealing or oxygen Plasma is an oxygen atmosphere, and the processing time for annealing or oxygen Plasma is controlled within 2 - 10 minutes. The thickness of the isolation layer is preferably 10 - 50 nm.

[0029] Step S4, a nickel oxide nanowire network 104 is grown on the surface of the isolation layer 103 by electroplating, which is used to realize the core functional layer of the bionic reservoir neuromorphic device, simulating the randomly connected nodes in the reservoir, as Figure 5 shown. Specifically, a DC power supply is selected, and a voltage is applied to the Ni seed layer. The range of the voltage magnitude is 10 - 100 V. Subsequently, the sample is annealed by means of rapid thermal annealing, vacuum annealing furnace, etc., to form a randomly distributed and grown nickel oxide nanowire network. The annealing temperature is 400 - 600 °C, and the annealing duration is 3 - 6 hours. As an oxygen vacancy type material system, the oxide nanowire can generate a current response under the excitation of voltage, and has a memory effect on the flowing current, recording and feedback in the form of resistance change, serving as the core functional layer of the reservoir device. During the process of nanowire growth and preparation, through the design of disordered growth, it forms a natural reservoir network in different stacking shapes to complete the calculation task.

[0030] Step S5, MoS2 semiconductor quantum dots 105 are grown on the nickel oxide nanowire network 104 by solution spin coating method to modify it, expanding the response of the device from pure electricity to optoelectronic response, and realizing the function of sensing optical signals, as Figure 6 shown. The concentration of MoS2 quantum dots is 1 mg / ml - 50 mg / ml. The spin coating is divided into two steps. The spin coating speed in the first step is 100 - 500 revolutions per minute, and the spin coating duration is 10 - 30 seconds. The spin coating speed in the second step is 3000 - 5000 revolutions per minute, and the spin coating duration is 40 - 60 seconds. The semiconductor quantum dots can be quantum dots with optoelectronic responses such as MoS2, PtS2, ZnS, CdS, CdSe, PtSe2, etc. By introducing a semiconductor quantum dot with typical optoelectronic response as the modification layer, the response range of the reservoir can be expanded, and a reservoir neuromorphic device with optoelectronic response can be obtained to complete signal sensing, calculation and storage, and improve the information processing efficiency of the reservoir.

[0031] Step S6, an Ag electrode is grown on the nanowire network 104 as the working electrode 106 by means of photolithography and physical vapor deposition, electron beam evaporation, thermal evaporation, etc., and the obtained structure is as Figure 7As shown. The electrode thickness is 50 - 100 nm, and the electrode size ranges from 50 μm × 50 μm to 500 μm × 500 μm. The electrode material can be selected from metals such as Ag, Ti, Ta, Cu, Al, Pt, Pd, etc.

[0032] This application uses the integrated circuit electroplating process to prepare oxide nanowires as an electronic reservoir, uses the characteristics of the randomly arranged grown oxide nanowires to construct neuron nodes in the reservoir, and realizes a randomly connected natural reservoir neuron similar to the human brain by means of the memristive behavior of the oxide nanowires, further conforming to the distribution characteristics of randomly initialized neurons and their random weights in the real reservoir, and avoiding the existence of artificial interference factors. On the other hand, by introducing quantum dots with excellent optoelectronic responses to modify the oxide nanowires, the response mode of the oxide nanowires is expanded, and a neuromorphic device with optoelectronic responses is constructed, which can complete information sensing, processing, and calculation in the same reservoir, greatly improving the information processing efficiency.

[0033] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A preparation method of a bionic optoelectronic reservoir neuromorphic device, characterized in that: The method comprises the following steps: Form a seed layer on a substrate; Perform in-situ oxidation treatment on the seed layer to form an isolation layer; Grow an oxide nanowire network on the surface of the isolation layer as an electronic reservoir, utilize the randomly arranged characteristics of the grown oxide nanowires to construct neuron nodes in the reservoir, and realize randomly connected reservoir neurons similar to the human brain by means of the memristive behavior of the oxide nanowires; Grow semiconductor quantum dots with optoelectronic response on the oxide nanowire network to modify it, expand the response of the device from pure electricity to optoelectronic response, and realize the perception function of optical signals; Grow a working electrode on the nanowire network to obtain a bionic optoelectronic reservoir neuromorphic device.

2. The preparation method of the bionic optoelectronic reservoir neuromorphic device according to claim 1, characterized in that: The seed layer is Ni, Ti, Zn, Cu or Mo.

3. The preparation method of the bionic optoelectronic reservoir neuromorphic device according to claim 1, characterized in that: The oxide nanowire network is nickel oxide, titanium oxide, zinc oxide, copper oxide or nickel oxide.

4. The preparation method of the bionic optoelectronic reservoir neuromorphic device according to claim 3, characterized in that: The specific steps of growing the oxide nanowire network include: applying a voltage to the seed layer, the magnitude of the voltage being 10 - 100 V, then annealing the sample, the annealing temperature being 400 - 600 °C, and the annealing duration being 3 - 6 hours, to form a randomly distributed and grown oxide nanowire network.

5. The preparation method of the bionic optoelectronic reservoir neuromorphic device according to claim 1, characterized in that: The semiconductor quantum dots are MoS2, PtS2, ZnS, CdS, CdSe or PtSe2 quantum dots.

6. A bionic optoelectronic reservoir neuromorphic device, characterized in that: It includes: A substrate; A seed layer formed on the substrate; An isolation layer formed on the seed layer; An oxide nanowire network formed on the surface of the isolation layer as an electronic reservoir, utilize the randomly arranged characteristics of the grown oxide nanowires to construct neuron nodes in the reservoir, and realize randomly connected reservoir neurons similar to the human brain by means of the memristive behavior of the oxide nanowires; Semiconductor quantum dots with optoelectronic response grown on the oxide nanowire network to modify it, expand the response of the device from pure electricity to optoelectronic response, and realize the perception function of optical signals; A working electrode grown on the nanowire network.

7. The bionic optoelectronic reservoir neuromorphic device according to claim 6, characterized in that: The seed layer is Ni, Ti, Zn, Cu or Mo.

8. The bionic optoelectronic reservoir neuromorphic device according to claim 6, characterized in that: The oxide nanowire network is nickel oxide, titanium oxide, zinc oxide, copper oxide or nickel oxide.

9. The bionic optoelectronic reservoir neuromorphic device according to claim 6, characterized in that: The semiconductor quantum dots are MoS2, PtS2, ZnS, CdS, CdSe or PtSe2 quantum dots.

10. The bionic optoelectronic reservoir neuromorphic device according to claim 6, wherein the substrate is a highly doped silicon wafer with silicon dioxide grown thereon.

Images