A self-assembled avalanche response device and preparation and application thereof
By forming a conductive percolation network in a self-assembled avalanche response device, the problems of insufficient electrical stability and response complexity are solved, the mapping quality and calculation accuracy of the reservoir computing system are improved, and efficient information processing is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- FUDAN UNIVERSITY
- Filing Date
- 2022-06-25
- Publication Date
- 2026-07-14
AI Technical Summary
Existing self-assembled network response devices suffer from poor electrical stability and insufficient response complexity in reservoir computing systems, resulting in inadequate mapping quality and computational accuracy.
By fabricating metal electrodes and conductive polymer response layers on an insulating substrate, and utilizing the assembly of nano-conductive fillers and interface modifiers to form a conductive percolation network, avalanche-like exponential response and electrical uniformity of the device can be achieved.
This improved the response amplitude and electrical response uniformity of the device, enhanced the mapping quality and computational accuracy of the reservoir network, and enabled efficient information classification and prediction.
Smart Images
Figure CN117332826B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electronic device technology, and relates to a self-assembled avalanche response device and its preparation and application. Background Technology
[0002] Reservoir computing is a neuromorphic computing paradigm adept at handling temporal tasks. It boasts advantages such as easy and fast training and high energy efficiency, making it crucial for applications in information processing, including speech recognition, image recognition, and chaotic system prediction. A reservoir computing system consists of three parts: an input layer, a reservoir network layer, and an output layer. The reservoir network layer, composed of an arbitrarily connected network of neurons, can nonlinearly map input information to a high-dimensional state space for the output layer to read, train, and recognize. Therefore, to realize the functionality of a reservoir computing system, the reservoir network layer needs to possess two key characteristics: dynamic nonlinear response and short-term memory.
[0003] Hardware implementation of reservoir computing systems offers unique advantages in terms of computational speed and energy consumption. Currently, hardware implementation of reservoir computing systems mainly focuses on developing physical devices with dynamic nonlinear response and short-time memory characteristics to simulate the mapping function of reservoir network layers. These different types of physical devices mainly include dynamic volatile resistive switching memories, spin-torque nano-oscillators, silicon-based photonic devices, and ferroelectric memories. However, as reservoir network unit devices, these devices still have some key shortcomings in information processing: the electrical stability of the devices is poor, thus reducing the mapping quality of the reservoir network to input information; the insufficient response complexity of the unit devices limits the mapping capability of network nodes, resulting in insufficient computational accuracy of the system.
[0004] Response devices constructed using self-assembled networks of conductive nanoparticles exhibit dynamic nonlinear responses and spontaneous recovery characteristics to external stimuli, thus enabling their use in reservoir computing. These self-assembled networks can also directly and effectively simulate the complex topological network structure of the brain and the avalanche response characteristics of neurons, and are simple to fabricate and have low development costs. However, as... Figure 1 and Figure 2 As shown, due to the random and disordered distribution of self-assembled networks, the electrical response uniformity of such devices is poor, which severely limits the spatial mapping accuracy of information in the reservoir network. Summary of the Invention
[0005] The purpose of this invention is to provide a self-assembled avalanche response device, its preparation, and its application.
[0006] The objective of this invention can be achieved through the following technical solutions:
[0007] One of the technical solutions of the present invention provides a self-assembled avalanche response device, which comprises an insulating substrate, metal electrodes deposited at both ends of the insulating substrate, and a response layer deposited on the insulating substrate and located between the two metal electrodes. Preferably, the insulating substrate is a silicon substrate.
[0008] The second technical solution of the present invention provides a method for fabricating a self-assembled avalanche response device, comprising the following steps:
[0009] (1) Take an insulating substrate and deposit metal electrodes at both ends of its upper surface;
[0010] (2) Take nano-conductive filler and interface modifier, add them to the elastic matrix solution, stir and mix, and then ultrasonically disperse to obtain a conductive polymer solution;
[0011] (3) The obtained conductive polymer solution is spin-coated onto an insulating substrate by a spin coating process, and the ends of the two metal electrodes facing each other are covered. After spin coating is completed, the conductive polymer film is dried to obtain a conductive polymer film as a response layer, thus completing the preparation of the self-assembled avalanche response device.
[0012] Furthermore, in step (1), the insulating substrate is a single-crystal silicon wafer or a silicon oxide wafer.
[0013] Furthermore, in step (1), the material of the metal electrode is one or more of Au, Ag, or Pt, Cu, and alloys such as TiN, CuAg, and AuAg.
[0014] Furthermore, in step (2), the elastic matrix in the elastic matrix solution is selected from one or more of the following: polyacrylate elastomers, thermoplastic polyurethanes, waterborne polyurethanes, natural rubber, epoxidized natural rubber, dimethylsiloxane, and styrene-ethylene-butadiene-styrene block copolymers.
[0015] The solvent used in the elastic matrix solution is water or an organic solvent, wherein the organic solvent is selected from one or more of methanol, ethanol, toluene, chloroform, acetone, dimethylformamide, and dimethyl sulfoxide;
[0016] The mass fraction of the elastic matrix solution is 5%–80%.
[0017] Furthermore, in step (2), the interface modifier is an amphiphilic molecule or amphiphilic polymer with hydrophilic and lipophilic chemical groups, specifically selected from one or more of sodium linear alkylbenzene sulfonate, lignin sulfonate, sodium fatty alcohol polyoxyethylene ether sulfate, sodium lauryl sulfate, glyceryl stearate, alkyl polyether, fatty alcohol polyoxyethylene ether, nanocellulose, etc.
[0018] Furthermore, in step (2), the nano-conductive filler is a metal nanoparticle, a metal nanowire, a carbon nanotube, graphene, or other two-dimensional conductors or semiconductor materials. Metal nanoparticles include gold nanoparticles and silver nanoparticles, and metal nanowires include gold nanowires and silver nanowires, etc.
[0019] Furthermore, in step (2), the mass ratio of the nano-conductive filler to the elastic matrix is 0.5 to 100:100;
[0020] The mass ratio of nano-conductive filler to interface modifier is 0.1 to 10:1.
[0021] Furthermore, in step (2), the stirring and mixing process conditions are as follows: at room temperature, magnetic stirring is performed for 20 to 40 minutes, preferably 30 minutes, and the stirring speed is 300 to 500 rpm, preferably 400 rpm; the ultrasonic dispersion process is as follows: ultrasonic dispersion is performed using a cell disruptor, the ultrasonic power is 200 to 400 W, preferably 300 W, and the total ultrasonic duration is 20 to 40 minutes, preferably 30 minutes.
[0022] Furthermore, in step (3), the spin coating speed is 1000-7000 rpm, preferably 5000 rpm, and the spin coating time is 0.5-1.5 min, preferably 1 min;
[0023] The drying temperature is 30–80℃, preferably 50℃, and the drying time is 2–6 hours, preferably 4 hours.
[0024] Furthermore, the thickness of the metal electrode is 40nm-100nm.
[0025] This invention utilizes a controllable assembly strategy to fabricate a responsive layer with a conductive percolation network structure. This responsive layer comprises an elastic matrix with thermal expansion properties, interface-modifying molecules, and nano-conductive fillers. When a pulsed voltage is applied, the conductive network generates Joule heating, causing the elastic matrix to expand. This volume expansion of the matrix disrupts the formed conductive pathways, resulting in a sharp decrease in the device's conductivity. When the voltage is removed, the elastic matrix returns to its original volume, allowing the conductive pathways to reconnect, and the conductivity returns to its initial state. Therefore, this device exhibits an avalanche-like exponential response amplitude to stimuli.
[0026] Furthermore, interface-modulating molecules with hydrophilic and oleophilic properties can connect the nano-conductive filler and the elastic matrix through chemical bonding, enabling the nano-conductive filler to be stably assembled at the micro-interface of the elastic matrix, thereby giving the device good electrical uniformity.
[0027] The third technical solution of the present invention provides an application of a self-assembled avalanche response device, which serves as a reservoir network layer and is used in a reservoir computing system.
[0028] The process of applying the self-assembled avalanche response device of the present invention to a reservoir computing system is as follows:
[0029] (1) The input layer is used to preprocess the raw data, thereby converting the raw data into a time-series input pulse sequence. The reservoir network layer is implemented through hardware substitution using the self-assembled network response device prepared above. This hardware reservoir network layer can respond to the time-series input pulses to achieve high-dimensional spatial feature mapping of the input information. The output layer is used to read and train the feature mapping data.
[0030] (2) Connect the response device to the electrical test platform. Connect one electrode port of the device to an arbitrary waveform generator and the other port to ground. This is to enable the response to and reading of the input signal.
[0031] (3) Reading and recording of the response signal. The programmed timing input pulse sequence is applied to the response device. The pulse width and amplitude of the pulse voltage are adjustable. Pulse width programming range: 10µs-100ms; pulse amplitude programming range: 1V-10V. After the device passes through the pulse voltage sequence, its response state is measured and read. The read pulse voltage is 0.1V, the read pulse width is consistent with the input pulse width, and the read interval is determined by the response curve recovery time. After each response, the device can spontaneously recover to its initial state. Therefore, by repeating the above process, the feature mapping of the input information and its state acquisition can be completed.
[0032] (4) Training and testing of the read signal. The read function is trained using machine learning algorithms. Taking image recognition as an example, the corresponding dataset is obtained by reading the response state data of the reservoir network for 20,000 training images. The dataset is divided into training and test sets according to a certain ratio. An appropriate machine learning algorithm (such as logistic regression, Adam algorithm, stochastic gradient descent algorithm, etc.) is selected to iteratively train the connection weights of the training set samples until a certain error accuracy is achieved, thus completing the training. Finally, the recognition accuracy of the reservoir computing system is evaluated using the test set samples.
[0033] Compared with the prior art, the present invention has the following advantages:
[0034] (1) A self-assembled avalanche response two-end device was developed using a controllable assembly strategy. This controllable assembly strategy can, on the one hand, construct a critical percolation network, enabling the device to exhibit an exponential avalanche response characteristic when subjected to electrical stimulation, thereby improving the response amplitude of the device; on the other hand, it can enable the conductive filler to assemble in an orderly and stable manner at the interface of the elastic matrix, constructing a conductive network with stable dynamic structural response. Due to the chemical bonding effect, this network can exhibit good electrical response uniformity to stimulation.
[0035] (2) A hardware-based reservoir computing system based on a self-assembled avalanche response device for information classification and prediction. Due to the dynamic nonlinear response and short-time memory characteristics of the aforementioned response device, it is used to implement the reservoir network layer of the reservoir computing system in hardware, thereby replacing the traditional software-generated reservoir network layer. Attached Figure Description
[0036] Figure 1 The diagram shows the structure and characterization of an existing hardware reservoir computing system based on self-assembled network response devices.
[0037] Figure 2 Performance diagram of existing hardware reservoir computing systems based on self-assembled network response devices;
[0038] Figure 3 This is a schematic diagram illustrating the structure and principle of the self-assembled avalanche response device of the present invention;
[0039] Figure 4 This is a process flow diagram for the fabrication and use of the self-assembled avalanche response device of the present invention;
[0040] Figure 5 The diagram shows the performance of the hardware storage pool computing system of this invention under a continuous pulse voltage of 1V.
[0041] Figure 6 The resistance response variation performance of the hardware storage pool computing system of this invention under pulse voltage stimulation of different magnitudes is shown in the figure.
[0042] Figure 7 This is a graph showing the performance of the hardware reserve pool computing system of the present invention in terms of recognition accuracy for the MNIST handwritten digit dataset. Detailed Implementation
[0043] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.
[0044] In the following embodiments, unless otherwise specified, the raw materials or processing techniques are all commercially available materials in the art.
[0045] Example 1:
[0046] This embodiment first provides a method for fabricating a self-assembled avalanche response device, specifically including the following steps:
[0047] Step 1:
[0048] First, metal electrodes at both ends are deposited on a silicon substrate. The electrode material is Au, and the thickness of the metal electrodes is about 80 nm.
[0049] Step 2:
[0050] 1) Preparation of the conductive polymer solution for the response layer and 2) Deposition of the response layer film between the two metal electrodes.
[0051] 1) Silver nanowires and nanocellulose were added to an aqueous polyurethane solution at a mass ratio of 1:1. The aqueous polyurethane emulsion was purchased from BASF, brand name Joncryl FLX 5002. The solution was diluted to 20% by adding water, resulting in a silver nanowire to aqueous polyurethane solution mass ratio of 1:50. The solution was magnetically stirred at 400 rpm for 30 minutes at room temperature. Subsequently, it was ultrasonically dispersed using a cell disruptor at a power of 300 W for a total sonication time of 30 minutes. To prevent the solution temperature from becoming too high due to prolonged sonication, an intermittent sonication strategy was adopted: the solution was sonicated for 5 minutes, then paused for 10 minutes to allow the solution to cool to room temperature. Through these methods, a well-dispersed conductive polymer solution for the response layer was obtained.
[0052] (2) The conductive polymer solution was spin-coated onto both ends of a photolithographically patterned metal electrode using a spin coating process. The polymer solution was positioned between the two metal electrodes and partially covered the ends of the metal electrodes, ensuring good contact between the polymer solution and the metal electrodes. The thickness of the deposited film was controlled by the concentration of the conductive polymer solution. The spin coating speed was 5000 rpm, and the spin coating time was 1 minute. Subsequently, the film was dried on a hot stage at 40°C for 4 hours to obtain a dried conductive polymer response layer. Thus, the fabrication of the self-assembled avalanche response device was completed.
[0053] The self-assembled avalanche response device prepared above is used as the hardware replacement for the reservoir mesh layer and is applied to the hardware reservoir computing system. The specific operation process of the resulting hardware reservoir computing system is as follows:
[0054] 1. The input layer preprocesses the raw data, transforming it into a time-series input pulse sequence. The reservoir network layer is implemented using the self-assembled network response device described above, which provides a hardware replacement. This hardware reservoir network layer responds to the time-series input pulses, enabling high-dimensional feature mapping of the input information. The output layer is used for reading and training the feature-mapped data.
[0055] 2. Connect the response device to the electrical test platform. Connect one electrode port of the device to an arbitrary waveform generator, and ground the other port. This enables the response to and reading of the input signal.
[0056] 3. Reading and Recording the Response Signal. The programmed timing input pulse sequence is applied to the response device. The pulse width and amplitude are adjustable. Pulse width programming range: 10µs-100ms; pulse amplitude programming range: 1V-10V. After the device passes through the pulse voltage sequence, its response state is measured and read. The read pulse voltage is 0.1V, the read pulse width is consistent with the input pulse width, and the read interval is determined by the response curve recovery time. After each response, the device can spontaneously recover to its initial state. Therefore, by repeating the above process, feature mapping of the input information and state acquisition can be completed.
[0057] 4. Training and Testing of the Reading Signal. The reading function is trained using machine learning algorithms. Taking image recognition as an example, a dataset is obtained by reading the response state data of a reservoir network from 20,000 training images. This dataset is divided into training and testing sets according to a certain ratio. An appropriate machine learning algorithm (such as logistic regression, Adam algorithm, stochastic gradient descent, etc.) is selected to iteratively train the connection weights of the training set samples until a certain error accuracy is achieved, completing the training. Finally, the recognition accuracy of the reservoir computing system is evaluated using the test set samples.
[0058] Figure 3 This diagram illustrates the microstructural changes of the response layer in a self-assembled avalanche response device, using a silicon substrate as an example. In the diagram, the spherical microspheres represent latex particles of an aqueous polyurethane emulsion, with an ordered conductive network aggregated at the interfaces of adjacent latex particles. The ordered assembly of the conductive filler at the latex particle interfaces is due to the "connecting" effect of the interface modifier, allowing the conductive filler to be uniformly dispersed and stably aggregated at the latex particle interfaces. When a pulsed voltage is applied, the Joule heating effect causes the elastic matrix to expand, breaking the formed conductive pathway and drastically reducing the device's conductivity. When the voltage is removed, the elastomer returns to its original volume, reconnecting the conductive pathway and restoring the conductivity to its initial state. Therefore, the device exhibits a high response amplitude and good electrical uniformity.
[0059] Figure 4 This is a process flow diagram illustrating the fabrication of the self-assembled avalanche responder device of the present invention and its application in reservoir computing (using a silicon substrate as an example for illustration). By using the self-assembled avalanche responder device of the present invention in the reservoir network layer, high-dimensional spatial feature mapping of time-series input pulse sequences can be effectively achieved. Figure 5It can be seen that under continuous pulse voltage stimulation of 1V, the hardware reservoir computing system exhibits good electrical response characteristics and short-time plasticity. During the pulse voltage application-removal process, the resistance increases and the R / R0 ratio increases when the pulse voltage is applied; when the voltage is removed, the resistance value recovers nonlinearly. When the interval between two consecutive pulse voltages is short, the resistance value has not yet recovered to its initial state before responding to a new stimulus. Therefore, a sequence of consecutive pulse voltages with short intervals can increase the R / R0 ratio of the device, exhibiting a synaptic-like paired-pulse facilitation (PPF) characteristic, which can improve the abundance of reservoir network nodes.
[0060] from Figure 6 It can be seen that the response amplitude of the hardware-based memory pool computing system increases with the increase of the pulse voltage. When a 10V stimulus is applied, the resistance ratio changes from 10... 0 Increase to approximately 10 6 It exhibits an "avalanche-like" exponential change response. Furthermore, the reservoir network layer response device can recover to its initial state, and the device shows good uniformity in its electrical response to the same stimulus, which is beneficial for improving the mapping quality of the reservoir network to input information.
[0061] from Figure 7 It can be seen that the reservoir computing system based on this self-assembled avalanche response device can recognize the MNIST handwritten digit dataset with an accuracy of over 95%, demonstrating high recognition accuracy.
[0062] Comparative Example 1:
[0063] It is largely the same as Example 1, except that the addition of the interface modifier is omitted.
[0064] In Comparative Example 1, due to the lack of an interface modifier to disperse the conductive filler and to regulate the stable assembly of the conductive filler at the elastic matrix interface, the conductive network could not assemble orderly at the elastic matrix interface. Therefore, the resulting device exhibited poor uniformity and stability, making subsequent electrical testing difficult.
[0065] Example 2:
[0066] The majority of the components are the same as in Example 1, except that in this example, the mass ratio of the nano-conductive filler to the elastic matrix is limited to 0.5:100.
[0067] The mass ratio of the nano-conductive filler to the interface modifier is 0.1:1.
[0068] Example 3:
[0069] The majority of the contents are the same as in Example 1, except that in this example, the mass ratio of the nano-conductive filler and the elastic matrix is limited to 100:100.
[0070] The mass ratio of nano-conductive filler to interface modifier is 10:1.
[0071] Example 4:
[0072] The majority of the contents are the same as in Example 1, except that in this example, the mass ratio of the nano-conductive filler and the elastic matrix is limited to 50:100.
[0073] The mass ratio of nano-conductive filler to interface modifier is 5:1.
[0074] Examples 5-10:
[0075] Compared to Example 1, most of the components are the same, except that in this example, the elastic matrix is replaced with equal masses of thermoplastic polyurethane, waterborne polyurethane, natural rubber, epoxidized natural rubber, dimethylsiloxane, and styrene-ethylene-butadiene-styrene block copolymer.
[0076] Examples 11-14:
[0077] Compared with Example 1, most of them are the same, except that in this example, the nano-conductive filler is replaced with metal nanoparticles, metal nanowires, carbon nanotubes, and graphene of equal mass.
[0078] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.
Claims
1. A self-assembled avalanche response device, characterized in that, It consists of an insulating substrate, metal electrodes deposited at both ends of the insulating substrate, and a response layer deposited on the insulating substrate and located between the two metal electrodes; The self-assembled avalanche-responsive device is prepared through the following steps: (1) Take an insulating substrate and deposit metal electrodes at both ends of its upper surface; (2) Take the nano-conductive filler and interface modifier, add them to the elastic matrix solution, stir and mix, and then ultrasonically disperse them to obtain a conductive polymer solution; (3) The obtained conductive polymer solution is spin-coated onto an insulating substrate by a spin coating process, and the ends of the two metal electrodes facing each other are covered. After spin coating is completed, the substrate is dried to obtain a conductive polymer film as a response layer. In step (2), the elastic matrix in the elastic matrix solution is selected from one or more of the following: polyacrylate elastomers, thermoplastic polyurethanes, waterborne polyurethanes, natural rubber, epoxidized natural rubber, dimethylsiloxane, and styrene-ethylene-butadiene-styrene block copolymers. The solvent used in the elastic matrix solution is water or an organic solvent, wherein the organic solvent is selected from one or more of methanol, ethanol, toluene, chloroform, acetone, dimethylformamide, and dimethyl sulfoxide; The mass fraction of the elastic matrix solution is 5-80%; In step (2), the interface modifier is an amphiphilic molecule or amphiphilic polymer with hydrophilic and lipophilic chemical groups, specifically selected from one or more of linear alkylbenzene sulfonate, lignin sulfonate, sodium fatty alcohol polyoxyethylene ether sulfate, sodium lauryl sulfate, glyceryl stearate, alkyl polyether, fatty alcohol polyoxyethylene ether, and nanocellulose. In step (2), the nano-conductive filler is metal nanoparticles, metal nanowires, carbon nanotubes, graphene or other two-dimensional conductors or semiconductor materials; In step (2), the mass ratio of the nano-conductive filler and the elastic matrix is 0.5~100:100; The mass ratio of nano-conductive filler to interface modifier is 0.1~10:
1.
2. The method for fabricating a self-assembled avalanche response device according to claim 1, characterized in that, Includes the following steps: (1) Take an insulating substrate and deposit metal electrodes at both ends of its upper surface; (2) Take the nano-conductive filler and interface modifier, add them to the elastic matrix solution, stir and mix, and then ultrasonically disperse them to obtain a conductive polymer solution; (3) The obtained conductive polymer solution is spin-coated onto the insulating substrate by spin coating process, and the ends of the two metal electrodes facing each other are covered. After spin coating is completed, the conductive polymer film is dried to obtain the response layer, thus completing the preparation of the self-assembled avalanche response device. In step (2), the elastic matrix in the elastic matrix solution is selected from one or more of the following: polyacrylate elastomers, thermoplastic polyurethanes, waterborne polyurethanes, natural rubber, epoxidized natural rubber, dimethylsiloxane, and styrene-ethylene-butadiene-styrene block copolymers. The solvent used in the elastic matrix solution is water or an organic solvent, wherein the organic solvent is selected from one or more of methanol, ethanol, toluene, chloroform, acetone, dimethylformamide, and dimethyl sulfoxide; The mass fraction of the elastic matrix solution is 5-80%; In step (2), the interface modifier is an amphiphilic molecule or amphiphilic polymer with hydrophilic and lipophilic chemical groups, specifically selected from one or more of linear alkylbenzene sulfonate, lignin sulfonate, sodium fatty alcohol polyoxyethylene ether sulfate, sodium lauryl sulfate, glyceryl stearate, alkyl polyether, fatty alcohol polyoxyethylene ether, and nanocellulose. In step (2), the nano-conductive filler is metal nanoparticles, metal nanowires, carbon nanotubes, graphene or other two-dimensional conductors or semiconductor materials; In step (2), the mass ratio of the nano-conductive filler and the elastic matrix is 0.5~100:100; The mass ratio of nano-conductive filler to interface modifier is 0.1~10:
1.
3. The method for fabricating a self-assembled avalanche response device according to claim 2, characterized in that, In step (1), the insulating substrate is a single-crystal silicon wafer or a silicon oxide wafer; The material of the metal electrode is selected from one or more of Au, Ag, Pt, Cu, and TiN, CuAg, and AuAg alloys.
4. The method for fabricating a self-assembled avalanche-response device according to claim 2, characterized in that, In step (2), the stirring and mixing process conditions are as follows: at room temperature, magnetic stirring is performed for 20 to 40 minutes at a stirring speed of 300 to 500 rpm; the ultrasonic dispersion process is as follows: ultrasonic dispersion is performed using a cell disruptor with an ultrasonic power of 200 to 400 W and a total ultrasonic duration of 20 to 40 minutes. In step (3), the spin coating speed is 1000~7000 rpm and the spin coating time is 0.5~1.5 min; The drying temperature is 30~80℃, and the time is 2~6 hours.
5. The method for fabricating a self-assembled avalanche-response device according to claim 2, characterized in that, The thickness of the metal electrode is 40nm-200nm.
6. The application of the self-assembled avalanche response device as described in claim 1, characterized in that, This self-assembled avalanche response device serves as a reservoir network layer and is used in a reservoir computing system.