Artificial neuron with excitation and inhibition dual modes and operation method thereof

By selecting the appropriate electrode material and dielectric layer in artificial neurons, and using polar electrical signals to achieve excitation and suppression mode switching, the problem that artificial neurons in the prior art is difficult to integrate excitation and suppression signals, achieving low power consumption and a multifunctional neuromorphic computing system integrated with high density.

CN119031826APending Publication Date: 2024-11-26NINGBO INST OF MATERIALS TECH & ENG CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410992302.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing artificial neurons have difficulty integrating excitatory and inhibitory signals from synapses spatially and temporally, and can only work in a single excitatory mode, lacking dual-mode function.

Method used

An artificial neuron with dual modes of excitation and inhibition was designed to promote the use of intermediate dielectric layers by selecting active and inert materials on the electrode layer and dielectric layer, and achieving excitation and inhibition patterns under the action of electrical signals with opposite polarities. Ion migration realizes coordinated processing of electrical signals.

Benefits of technology

The excitation and suppression mode switching under the action of polar electrical signals is realized, which reduces power consumption, simplifies the device structure, facilitates high-density integration, and simulates the biotrustworthiness and complex signal processing capabilities of neurons.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119031826A_ABST
    Figure CN119031826A_ABST
Patent Text Reader

Abstract

The invention discloses an artificial neuron with excitation and inhibition dual modes and an operation method thereof, and belongs to the technical field of artificial neurons. The artificial neuron has a two-end structure and comprises a substrate, a bottom electrode layer, an intermediate dielectric layer and a top electrode layer which are stacked in sequence, any one of the top electrode layer or the bottom electrode layer is an active electrode, and the artificial neuron can be used for forming an active electrode under the action of electric signals with opposite polarities by setting and selecting the material of the intermediate dielectric layer. And dual modes of excitability and inhibition are realized. Cooperative processing of excitability and inhibitory signals can be realized based on a single two-end device, a complex peripheral circuit is not needed, power consumption is low, the device structure is simple, and development of a high-density integrated multifunctional neuromorphic computing system is facilitated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of artificial neuron devices, and in particular relates to an artificial neuron with excitation and inhibition dual modes and an operation method thereof. Background Art

[0002] With the rapid development of modern information technology, artificial intelligence (AI) has been widely used in fields such as driverless cars, face recognition, voice interaction, and smart medical care, providing many conveniences for people's lives. At present, AI technology is mainly implemented based on traditional von Neumann computing devices, whose memory and processor are physically separated, facing the challenges of power consumption wall and storage wall. In contrast, the human brain has the advantages of low energy consumption, high parallelism, and strong fault tolerance. It can efficiently perform complex tasks such as cognition, learning, and reasoning. Therefore, brain-inspired neuromorphic computing has attracted widespread attention from researchers.

[0003] The core of building a neuromorphic computing architecture is to develop new artificial synapses and neuron devices. At present, a series of important progress has been made in the research of artificial synaptic devices. Researchers have simulated the plasticity of biological synapses based on emerging memory devices such as memristors. Compared with synaptic devices, there are relatively few studies on artificial neurons. At present, artificial neurons are mainly constructed using CMOS circuits, which have a relatively complex structure and are not conducive to high-density integration. In recent years, some emerging electronic devices, such as threshold resistive memristors, phase change devices, spin devices and transistors, can simulate neuron functions, but generally can only work in a single excitation mode. For example, the invention patent with publication number CN115843215A and the document "An Artificial Neuron Based on a Threshold Switching Memristor" [IEEE ELECTRON DEVICE LETTERS, 2018, 39 (2), 308-311] respectively disclose an artificial neuron based on a threshold resistive memristor, but the disclosed neurons can only work in the excitation mode because their threshold voltage cannot be adjusted. In the human brain, neurons can integrate excitatory and inhibitory signals from synapses in space and time, that is, they have both excitatory and inhibitory modes. However, in existing artificial neuron research, it is difficult to achieve both excitatory and inhibitory modes at the same time. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an artificial neuron with dual modes of excitation and inhibition and an operation method thereof, so as to solve the problem in the prior art that artificial neurons are difficult to integrate excitatory and inhibitory signals from synapses in space and time.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An artificial neuron with dual modes of excitation and inhibition, comprising a substrate, a bottom electrode layer, an intermediate medium layer and a top electrode layer arranged in sequence from bottom to top;

[0007] Any one of the top electrode layer and the bottom electrode layer is an active electrode, and the other is an inert electrode;

[0008] The artificial neuron realizes a dual mode of excitation and inhibition under the action of two electrical signals with adjacent time and opposite polarities, wherein the electrical signals are current signals or voltage signals;

[0009] The electrical signal is applied to the active electrode or the inert electrode.

[0010] A further improvement of the present invention is:

[0011] Preferably, the inert electrode is one or more of an inert metal, a conductive oxide, a conductive nitride and a conductive carbon material;

[0012] The intermediate dielectric layer is one or more of sulfide, oxide and perovskite materials;

[0013] The active electrode is any one or more of copper, silver, aluminum, titanium, nickel, zinc, manganese and iron.

[0014] Preferably, the inert metal is one or more of platinum, palladium, iridium and gold, the conductive oxide is one or more of indium tin oxide, fluorine-doped tin oxide, aluminum-doped zinc oxide, the conductive nitride is TiN, and the conductive carbon material is graphite or graphene.

[0015] Preferably, the substrate is a thermally oxidized silicon wafer, the bottom electrode layer is platinum, the middle dielectric layer is zinc sulfide, and the top electrode layer is copper.

[0016] Preferably, when the intermediate dielectric layer is zinc sulfide, it is prepared by thermal evaporation.

[0017] A method for operating any of the above-mentioned artificial neurons with dual modes of excitation and inhibition, comprising the following four operating methods:

[0018] Type 1: First, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0019] Type 2: First, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0020] Type three: first, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0021] Type 4: firstly, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then, a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0022] When the electrical signal is applied to the active electrode, the inert electrode is grounded; when the electrical signal is applied to the inert electrode, the active electrode is grounded.

[0023] Preferably, the positive electrical signal is a positive DC voltage, and the negative electrical signal is a negative DC voltage.

[0024] Preferably, the amplitudes of the positive DC voltage and the negative DC voltage are 1 mV to 100 V, and the scanning step length is 0.001 to 1 V / s.

[0025] Preferably, the positive electrical signal is a positive pulse voltage, and the negative electrical signal is a negative pulse voltage.

[0026] Preferably, the amplitude of the positive pulse voltage and the negative pulse voltage is 1 mV to 100 V, and the frequency is 0.1 Hz to 1 MHz.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention discloses an artificial neuron with dual modes of excitation and inhibition. The artificial neuron has special settings on each layer of material. One end is an active electrode that is easy to ionize, and the other end is an inert electrode that is difficult to ionize. An intermediate medium layer material that is conducive to ion migration is set between the top electrode layer and the bottom electrode layer. The selection of this material, combined with the application of electrical signals of opposite polarity to the artificial neuron, enables the artificial neuron to exhibit different modes under electrical signals of different polarities. The present invention is based on a single two-terminal device, and can achieve coordinated processing of excitatory and inhibitory signals without complex peripheral circuits. It has low power consumption and a simple device structure, which is conducive to the development of high-density integrated multifunctional neuromorphic computing systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The schematic diagram of the artificial neuron structure disclosed in the present invention, in which: 1-top electrode layer, 2-middle medium layer, 3-bottom electrode layer, 4-substrate;

[0030] Figure 2 A schematic diagram of a current-voltage characteristic curve of a bipolar threshold resistive switching (BTS) device disclosed in the present invention;

[0031] Figure 3 It is a schematic diagram of the operation of the BTS dual-mode neuron disclosed in the present invention;

[0032] Figure 4 The current-voltage characteristic curve of the prepared BTS neuron under positive voltage sweep;

[0033] Figure 5 The threshold voltage and holding voltage distribution diagram of the prepared BTS neuron in 100 cycles;

[0034] Figure 6 Distribution diagram of threshold voltage and holding voltage of 10 different BTS neuron devices prepared. Each device was measured 5 times and the average value was taken;

[0035] Figure 7 The current-voltage characteristic curve of the prepared BTS neuron under positive and negative voltage scanning;

[0036] Figure 8 The relationship between the threshold voltage of the prepared BTS neuron and the negative scanning voltage amplitude;

[0037] Fig. 9 Leak-accumulation-fire (LIF) behavior of the prepared BTS neurons under positive voltage pulses, (pulse amplitude (A) = 1.0 V, width (W) = 50 ms, interval (I) = 100 ms);

[0038] Fig.10 The relationship between the firing probability of the prepared BTS neuron and the amplitude of the forward pulse voltage;

[0039] Fig.11 The complete discharge behavior of the prepared BTS neuron under a positive voltage pulse (A = 1.0 V, W = 50 ms, I = 50 ms);

[0040] Fig.12 The prepared BTS neuron exhibited complete inhibition behavior under alternating positive and negative voltage pulses (wherein the positive voltage pulse A=1.0V, W=50ms, and the negative voltage pulse A=1.0V, W=50ms).

[0041] Fig.13 This is the relationship between the firing probability of the prepared BTS neuron and the amplitude of the negative pulse voltage. DETAILED DESCRIPTION

[0042] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0043] In the description of the present invention, it should be noted that the orientations or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. are based on the orientations or positional relationships shown in the drawings, and are 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 cannot be understood as limiting the present invention; the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance; in addition, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0044] In the human brain, neurons have dual modes of excitation and inhibition, which can efficiently integrate excitatory and inhibitory signals to complete complex learning and cognitive functions. Emerging neuron devices generally can only work in a single mode and have low biological credibility. Although traditional CMOS neuron circuits can achieve dual modes of excitation and inhibition, they have complex structures and high power consumption, which is not conducive to large-scale integration. In order to solve the above problems, the present invention discloses an artificial neuron device with dual modes of excitation and inhibition.

[0045] The first aspect of the present invention discloses an artificial neuron device with dual modes of excitation and inhibition. The dual-mode neuron structure of the device is as follows: Figure 1As shown, from bottom to top, there are substrate 4, bottom electrode layer 3, intermediate dielectric layer 2 and top electrode layer 1. The artificial neuron device has dual modes of excitation and inhibition. In the following embodiments, electrical signals are input through the top electrode layer 1, and the bottom electrode layer 3 is grounded. The top electrode layer 1 and the bottom electrode layer 3 have one end of active metal and the other end of one or more of inert metal, conductive oxide, conductive nitride and conductive carbon material. The intermediate dielectric layer 2 is one or more of sulfide, oxide, perovskite material and carbon material. There are four options for device electrodes: 1) Active metal as top electrode and inert metal as bottom electrode. When a positive voltage is applied to the device, active metal atoms are easily ionized into charged ions and migrate under the action of the electric field, forming a metal conductive channel in the intermediate dielectric layer. At this time, the neuron firing threshold is small and is in the excitation mode; when a negative voltage is applied to the device, the inert metal is not easy to ionize, and the metal conductive channel will be broken under the action of the electric field. At this time, the neuron firing threshold increases, showing an inhibition mode. 2) Inert metal as top electrode and active metal as bottom electrode. When a negative voltage is applied to the device, the active metal atoms are easily ionized into charged ions and migrate under the action of the electric field, forming a metal conductive channel in the intermediate dielectric layer. At this time, the neuron firing threshold is small and is in an excitation mode; when a positive voltage is applied to the device, the inert metal is not easy to ionize, and the metal conductive channel will break under the action of the electric field. At this time, the neuron firing threshold increases, showing an inhibition mode. 3) Both the top electrode and the bottom electrode are active metals. Regardless of whether the device applies a positive or negative voltage, the active electrode is ionized, promoting the formation of a metal conductive channel, so the neuron can only be in a single excitation mode. 4) Both the top electrode and the bottom electrode are inert metals. Regardless of whether the device applies a positive or negative voltage, the inert electrode is not easy to ionize, and a metal conductive channel cannot be formed, making it difficult for the neuron to achieve dual mode. Therefore, the combination of active and inert electrodes, and matching the corresponding electrical signal application method, gives artificial neurons dual-mode characteristics of excitation and inhibition.

[0046] The artificial neuron device can realize opposite excitation mode and inhibition mode under the action of electrical signals with opposite polarities; the electrical signals are current signals or voltage signals. In the excitation mode, the neuron has a higher firing probability; in the inhibition mode, the neuron has a lower firing probability. The present invention realizes the coordinated processing of excitatory and inhibitory signals by neurons based on a single device, can simulate neuron behaviors such as all-or-nothing and dynamically adjustable neuron firing probability, and endows the neuron device with higher biological credibility.

[0047] In some embodiments of the present invention, the active electrode is one or more of copper, silver, aluminum, titanium, nickel, zinc, manganese and iron, etc. The other end is an inert electrode, which is less active and is not easily ionized under the action of an electrical signal, wherein the inert metal is one or more of platinum, palladium, iridium and gold, etc., the conductive oxide is one or more of indium tin oxide (ITO), fluorine-doped tin oxide (FTO), aluminum-doped zinc oxide (AZO), the conductive nitride is TiN, and the conductive carbon material is graphite or graphene.

[0048] The bottom electrode layer, the top electrode layer and the intermediate dielectric layer are all prepared by a coating process, including magnetron sputtering, atomic layer deposition, thermal evaporation, electron beam evaporation, and chemical vapor deposition. A suitable coating method is selected according to the materials of the electrode layer and the intermediate dielectric layer.

[0049] Preferably, the thickness of the top electrode layer 1 is 1-500 nm.

[0050] Preferably, the top electrode layer 1 is made of copper and is prepared by electron beam evaporation at a growth rate of The deposition temperature is room temperature. Copper metal is easy to ionize, and the generated copper ions migrate quickly under the action of the electric field, forming a metal conductive channel at a lower voltage, which is beneficial to reducing the energy consumption of the device.

[0051] Preferably, the thickness of the bottom electrode layer 3 is 1-500 nm.

[0052] Preferably, the bottom electrode layer 3 is platinum and is prepared by magnetron sputtering with a sputtering power of 20 to 200 W and a deposition temperature of room temperature to 300° C.

[0053] The intermediate dielectric layer 2 is one or more of sulfide, oxide, perovskite material, carbon material, etc. The sulfide is zinc sulfide, copper sulfide, silver sulfide, etc., the oxide is zinc oxide, titanium oxide, tantalum oxide, hafnium oxide, etc., the perovskite material is an inorganic perovskite material, and the carbon material is graphite or graphene.

[0054] Preferably, the intermediate dielectric layer 2 is zinc sulfide, which is prepared by thermal evaporation at a growth rate of The deposition temperature is room temperature to 300° C., and the thickness is 1 to 500 nm. The intermediate dielectric layer 2 prepared by thermal evaporation is loose and porous. Metal ions are easy to migrate in zinc sulfide, and a metal conductive channel can be formed at a relatively low voltage, which is beneficial to reducing device energy consumption.

[0055] In a specific embodiment of the present invention, the substrate 4 is a thermally oxidized silicon wafer, the bottom electrode layer 3 is made of platinum with a thickness of 200 nm, the intermediate dielectric layer 2 is made of zinc sulfide with a thickness of 200 nm, and the top electrode layer 1 is made of copper with a thickness of 165 nm.

[0056] In one embodiment of the present invention, a method for preparing an excitatory and inhibitory dual-mode neuron is disclosed as follows:

[0057] (1) Preparation of substrate and bottom electrode: A thermally oxidized silicon wafer was selected as the substrate 4, and a Pt layer was selected as the bottom electrode layer 3. The structure from bottom to top was Si / SiO2 / Ti / Pt, wherein the thickness of the SiO2 layer was 300 nm. A Ti layer with a thickness of 50 nm was grown on the SiO2 layer to increase the mechanical bonding force between the Pt layer and the thermally oxidized silicon layer. The thickness of the Pt layer was 200 nm and was grown using magnetron sputtering.

[0058] (2) A zinc sulfide film is prepared on the bottom electrode layer prepared in (1) by a thermal evaporation method.

[0059] The preparation parameters are as follows:

[0060] Using 99.9% pure zinc sulfide as the evaporating material, the substrate temperature is 200°C, and the deposition rate is maintained at A zinc sulfide thin film was prepared with a thickness of 200 nm.

[0061] (3) A copper top electrode layer was prepared on the zinc sulfide film prepared in (2) by electron beam evaporation combined with a mask plate method, and the growth rate was The deposition temperature was room temperature and the thickness of the copper electrode layer was 165 nm.

[0062] The second aspect of the present invention discloses an operating method of an artificial neuron device with dual modes of excitation and inhibition, which realizes reversible switching of neurons between dual modes of excitation and inhibition by introducing excitatory signals and inhibitory signals. Among them, the excitatory signal can reduce the neuron firing threshold and increase the neuron firing probability; the inhibitory signal can increase the neuron firing threshold and reduce the neuron firing probability. The excitatory signal and the inhibitory signal are voltage (current) signals with opposite polarity (direction). There are the following four situations:

[0063] Type 1: First, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0064] Type 2: First, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0065] Type three: first, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0066] Type 4: firstly, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then, a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode;

[0067] When the electrical signal is applied to the active electrode, the inert electrode is grounded; when the electrical signal is applied to the inert electrode, the active electrode is grounded.

[0068] In the above application method, no matter which type, when the active electrode matches the positive electrical signal and the inert electrode matches the negative electrical signal, the artificial neuron is in the excitation mode; when the active electrode matches the negative electrical signal and the inert electrode matches the positive electrical signal, the artificial neuron is in the inhibition mode.

[0069] As a preferred solution, when applying a signal, the electrical signal and the corresponding electrode material applied first are required to put the artificial neuron into an excitation mode, and then the subsequent modes are alternated in sequence.

[0070] Figure 2 The figure is a schematic diagram of the current-voltage characteristic curve of the BTS device disclosed in the present invention. It can be seen from the figure that under the forward voltage scan, the device resistance state changes suddenly when the threshold voltage is reached, switching from a high resistance state to a low resistance state; when the voltage is swept back to the holding voltage, the device resistance will change from a low resistance state back to a high resistance state. Under the negative voltage scan, the device resistance state changes from a high resistance state to a higher resistance state, which is called a reset operation. At this time, a forward voltage scan is performed again, and the threshold voltage becomes smaller. This type of device can be called a BTS memristor. Dual-mode neurons can be realized based on this device. Figure 3 As shown, BTS neurons can operate in both excitatory and inhibitory modes, achieving efficient integration of excitatory and inhibitory signals.

[0071] The present invention realizes an excitation and inhibition dual-mode artificial neuron based on a unique bipolar threshold resistance switching (BTS) memristor. The device exhibits threshold resistance switching (TS) characteristics during positive voltage scanning and has a relatively small threshold voltage (V th ); After the negative voltage scan, in the subsequent forward voltage scan, V th increases, and under continuous forward voltage sweep, V th Then it gradually decreases to the initial value, and then shows the BTS characteristics. BTS devices are mainly based on the on-off operation of copper conductive filaments. Under the forward voltage, it is conducive to the formation of copper conductive channels inside the device, V th The smaller the voltage, the greater the probability of neuron firing, corresponding to its excitation mode; under negative voltage, the electric field and Joule heat cause the filament to break, V thAs the voltage increases, the probability of neuron firing decreases, corresponding to its inhibition mode. The conductive filaments are switched on and off by adjusting the positive and negative voltages, and the location where the conductive filaments break occurs near the inert electrode. The reconfigurable morphology of the conductive filaments inside the device gives neurons the ability to reversibly switch between the excitation and inhibition modes.

[0072] In some embodiments of the present invention, the voltage signal is a direct current voltage or a pulse voltage signal.

[0073] Preferably, the excitatory signal is a positive DC voltage, and the inhibitory signal is a negative DC voltage; wherein the DC voltage amplitude is 1mV to 100V, and the scanning step length is 0.001 to 1V / s;

[0074] More preferably, the excitatory signal is a 2V positive DC voltage, and the scanning step is 0.01V / s; the inhibitory signal is a -3V negative DC voltage, and the scanning step is 0.01V / s.

[0075] Preferably, the excitatory signal is a positive pulse voltage, and the inhibitory signal is a negative pulse voltage; the pulse voltage amplitude is 1mV to 100V, and the frequency is 0.001Hz to 1GHz.

[0076] More preferably, the excitatory signal is a 1V positive pulse voltage with a frequency of 10Hz; the inhibitory signal is a -1V negative pulse voltage with a frequency of 10Hz.

[0077] The following is further described in conjunction with specific embodiments.

[0078] Example 1

[0079] In this embodiment, an artificial neuron device with dual modes of excitation and inhibition is disclosed, and the preparation process of the device includes the following steps:

[0080] (1) Preparation of substrate and bottom electrode: A thermally oxidized silicon wafer was selected as the substrate 4, and a Pt layer was selected as the bottom electrode layer 3. The structure from bottom to top was Si / SiO2 / Ti / Pt, wherein the thickness of the SiO2 layer was 300 nm. A Ti layer with a thickness of 50 nm was grown on the SiO2 layer to increase the mechanical bonding force between the Pt layer and the thermally oxidized silicon layer. The thickness of the Pt layer was 200 nm and was grown using magnetron sputtering.

[0081] (2) A zinc sulfide film is prepared on the bottom electrode layer prepared in (1) by a thermal evaporation method.

[0082] The preparation parameters are as follows:

[0083] Using 99.9% pure zinc sulfide as the evaporating material, the substrate temperature is 200°C, and the deposition rate is maintained at A zinc sulfide thin film was prepared with a thickness of 200 nm.

[0084] (3) A copper top electrode layer was prepared on the zinc sulfide film prepared in (2) by electron beam evaporation combined with a mask plate method, and the growth rate was The deposition temperature was room temperature, the thickness of the copper electrode layer was 165 nm, and a device having a structure from top to bottom of substrate (copper), bottom electrode layer (platinum), middle dielectric layer (zinc sulfide) and top electrode layer (copper) was obtained.

[0085] In this embodiment, a DC voltage scanning method is used to implement the BTS neuron dual-mode operation. Figure 4 The current-voltage characteristic curve of the prepared BTS neuron under forward voltage scanning. The bottom electrode of the device is grounded and the voltage is applied to the top electrode. As can be seen from the figure, the forward voltage scanning process is 0-1-0V, the threshold voltage is about 0.8V, and the holding voltage is about 0.2V. Figure 5 The distribution diagram of threshold voltage and holding voltage of the prepared BTS neuron in 100 consecutive cycles. As can be seen from the figure, the threshold voltage and holding voltage show small fluctuations around 0.8 and 0.2V respectively, indicating that the device has excellent stability. Figure 6 The distribution diagram of threshold voltage and holding voltage of 10 prepared BTS neurons, where each device was measured 5 times and the average value was taken. It can be seen from the figure that the threshold voltage and holding voltage distribution of the 10 randomly selected devices are relatively consistent, indicating that the device has good device repeatability.

[0086] It is worth noting that the threshold voltage of the prepared BTS neurons can be modulated by applying a negative sweep voltage. Figure 7 As shown, the initial threshold voltage of BTS neurons is about 0.8V ( Figure 7 When a negative sweep voltage is applied, a reset process is observed ( Figure 7 The device resistance state changes from a high resistance state to a higher resistance state, indicating that the device has semi-volatile characteristics. The semi-volatility of resistance exhibited by BTS devices is the key to simulating excitatory and inhibitory dual-mode neurons based on such devices. After reset, the device threshold voltage increases significantly to +1.5V ( Figure 7 When the device is continuously subjected to a positive scan voltage, the threshold voltage gradually decreases to its initial value. In addition, the effect of the negative voltage scan amplitude on the device threshold voltage is also studied. Figure 8 The results show that the device threshold voltage increases with the increase of negative voltage sweep amplitude.

[0087] Example 2

[0088] Dual-mode operation of BTS neurons was implemented using pulse voltage stimulation. Fig. 9The figure shows the change of device current over time under the stimulation of a voltage pulse with an amplitude of 1V. As can be seen from the figure, the device current did not change significantly at the beginning, and after the first 8 seconds, a significant current pulse signal appeared. Setting the current emission threshold to 10μA, it can be seen that the device reaches the emission threshold at 11 seconds, and then the emission probability increases steadily, reaching nearly 100% emission after 28 seconds. Therefore, the device exhibits typical neuronal LIF behavior. In addition, Fig.10 As shown, the device emission probability has a strong dependence on the forward pulse amplitude. When the pulse amplitude is lower than 0.6V, the emission probability is close to 0, and when the pulse amplitude is higher than 1V, the emission probability is close to 100%.

[0089] Fig.11 The results show that the prepared BTS neurons exhibited a fully emitting state under 1V positive pulse stimulation. When 1V positive and negative alternating voltage pulses were applied, the device exhibited a completely non-emitting state, indicating that negative voltage pulses can effectively inhibit the emission of BTS neurons ( Fig.12 ). This inhibition mechanism stems from the fact that negative voltage pulses increase the threshold voltage of BTS neurons. In addition, the device firing probability has a strong dependence on the amplitude of negative pulses. Fig.13 As shown, as the amplitude of the negative voltage pulse increases, the device emission probability gradually decreases.

[0090] Example 3

[0091] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, palladium metal as a bottom electrode with a thickness of 100nm, zinc oxide as an intermediate dielectric layer with a thickness of 100nm, and silver as a top electrode with a thickness of 100nm from bottom to top.

[0092] Among them, the bottom electrode is prepared by magnetron sputtering, the sputtering power is 100W, and the deposition temperature is 100℃.

[0093] The intermediate dielectric layer is prepared by thermal evaporation method, and the deposition temperature is 100°C.

[0094] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0095] Example 4

[0096] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, iridium metal as a bottom electrode with a thickness of 1nm, an inorganic perovskite material (CsPbI3) as an intermediate dielectric layer with a thickness of 1nm, and aluminum as a top electrode with a thickness of 1nm from bottom to top.

[0097] Among them, the bottom electrode is prepared by magnetron sputtering with a sputtering power of 20W and a deposition temperature of 300℃.

[0098] The intermediate dielectric layer is prepared by thermal evaporation at a deposition temperature of 200°C.

[0099] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0100] Example 5

[0101] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, gold metal as a bottom electrode with a thickness of 500nm, titanium oxide as an intermediate dielectric layer with a thickness of 500nm, and titanium as a top electrode with a thickness of 500nm from bottom to top.

[0102] Among them, the bottom electrode is prepared by magnetron sputtering, the sputtering power is 200W, and the deposition temperature is room temperature.

[0103] The intermediate dielectric layer is prepared by thermal evaporation method, and the deposition temperature is 25°C.

[0104] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0105] Example 6

[0106] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, ITO as a bottom electrode with a thickness of 300nm, zinc sulfide as an intermediate dielectric layer with a thickness of 300nm, and nickel as a top electrode with a thickness of 300nm from bottom to top.

[0107] Among them, the bottom electrode is prepared by magnetron sputtering with a sputtering power of 50W and a deposition temperature of 200°C.

[0108] The intermediate dielectric layer is prepared by thermal evaporation method, and the deposition temperature is 50°C.

[0109] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0110] Example 7

[0111] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, AZO as a bottom electrode with a thickness of 400nm, zinc sulfide as an intermediate dielectric layer with a thickness of 400nm, and zinc as a top electrode with a thickness of 400nm from bottom to top.

[0112] Among them, the bottom electrode is prepared by magnetron sputtering, the sputtering power is 150W, and the deposition temperature is 250℃.

[0113] The intermediate dielectric layer is prepared by thermal evaporation method, and the deposition temperature is 150°C.

[0114] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0115] Example 8

[0116] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, TiN as a bottom electrode with a thickness of 300nm, zinc oxide as an intermediate dielectric layer with a thickness of 300nm, and manganese as a top electrode with a thickness of 300nm from bottom to top.

[0117] Among them, the bottom electrode is prepared by magnetron sputtering with a sputtering power of 120W and a deposition temperature of 80°C.

[0118] The intermediate dielectric layer is prepared by thermal evaporation method, and the deposition temperature is 250°C.

[0119] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0120] Example 9

[0121] The dual-mode neuron uses a thermally oxidized silicon wafer as a substrate, graphite as a bottom electrode with a thickness of 400nm, zinc oxide as an intermediate dielectric layer with a thickness of 400nm, and iron as a top electrode with a thickness of 400nm from bottom to top.

[0122] Among them, the bottom electrode is prepared by magnetron sputtering with a sputtering power of 180W and a deposition temperature of 260℃.

[0123] The intermediate dielectric layer is prepared by thermal evaporation method with a deposition temperature of 300°C.

[0124] The top electrode is prepared by electron beam evaporation with a growth rate of The deposition temperature was room temperature.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An artificial neuron with dual modes of excitation and inhibition, It is characterized in that It comprises a substrate (4), a bottom electrode layer (3), an intermediate dielectric layer (2) and a top electrode layer (1) which are arranged in sequence from bottom to top; Any one of the top electrode layer (1) and the bottom electrode layer (3) is an active electrode, and the other is an inert electrode; The artificial neuron realizes a dual mode of excitation and inhibition under the action of two electrical signals with adjacent time and opposite polarities, wherein the electrical signals are current signals or voltage signals; The electrical signal is applied to the active electrode or the inert electrode.

2. An artificial neuron having dual modes of excitation and inhibition according to claim 1, It is characterized in that The inert electrode is one or more of an inert metal, a conductive oxide, a conductive nitride and a conductive carbon material; The intermediate dielectric layer (2) is one or more of sulfide, oxide and perovskite materials; The active electrode is any one or more of copper, silver, aluminum, titanium, nickel, zinc, manganese and iron.

3. An artificial neuron with dual modes of excitation and inhibition according to claim 2, It is characterized in that The inert metal is one or more of platinum, palladium, iridium and gold, the conductive oxide is one or more of indium tin oxide, fluorine-doped tin oxide, and aluminum-doped zinc oxide, the conductive nitride is TiN, and the conductive carbon material is graphite or graphene.

4. An artificial neuron having dual modes of excitation and inhibition according to claim 1, It is characterized in that The substrate (4) is a thermally oxidized silicon wafer, the bottom electrode layer (3) is platinum, the middle dielectric layer (2) is zinc sulfide, and the top electrode layer (1) is copper.

5. An artificial neuron having dual modes of excitation and inhibition according to claim 4, It is characterized in that When the intermediate dielectric layer (2) is zinc sulfide, it is prepared by thermal evaporation.

6. A method for operating an artificial neuron having dual modes of excitation and inhibition according to any one of claims 1 to 5, It is characterized in that There are four operation methods: Type 1: First, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode; Type 2: First, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode; Type three: first, a positive electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then a negative electrical signal is applied to the active electrode of the artificial neuron, and the artificial neuron is in the inhibition mode; Type 4: firstly, a negative electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the excitation mode; then, a positive electrical signal is applied to the inert electrode of the artificial neuron, and the artificial neuron is in the inhibition mode; When the electrical signal is applied to the active electrode, the inert electrode is grounded; when the electrical signal is applied to the inert electrode, the active electrode is grounded.

7. The method for operating an artificial neuron with dual excitation and inhibition modes according to claim 6, It is characterized in that The positive electrical signal is a positive direct current voltage, and the negative electrical signal is a negative direct current voltage.

8. The method for operating an artificial neuron with dual excitation and inhibition modes according to claim 7, It is characterized in that The amplitudes of the positive DC voltage and the negative DC voltage are 1 mV to 100 V, and the scanning step length is 0.001 to 1 V / s.

9. The method for operating an artificial neuron with dual excitation and inhibition modes according to claim 6, It is characterized in that The positive electrical signal is a positive pulse voltage, and the negative electrical signal is a negative pulse voltage.

10. The method for operating an artificial neuron with dual excitation and inhibition modes according to claim 9, It is characterized in that The amplitude of the positive pulse voltage and the negative pulse voltage is 1mV to 100V, and the frequency is 0.1Hz to 1MHz.