LIF neuron circuit based on self-powered threshold switch memristor and preparation method thereof

Abstract

The invention discloses an LIF neuron circuit based on a self-powered threshold switch memristor and a preparation method of the LIF neuron circuit. The LIF neuron circuit comprises a friction nanometer generator, a rectifier bridge, a threshold TS memristor, a fixed value resistor and a capacitor. Dependence of a traditional artificial neuron on an external power supply and overall energy consumption of a system are remarkably reduced, meanwhile, key dynamic characteristics such as leakage, integration, distribution and delay of a biological neuron are efficiently simulated through cooperative charging and discharging behaviors of the memristor and the capacitor, a stimulation intensity-dependent frequency coding function is shown, and the simulation performance of the neuron is improved. And an effective hardware implementation scheme is provided for constructing a low-power-consumption and high-integration-level sensing and computing integrated neuromorphic system.

Description

Technical Field

[0001] This invention relates to the field of microelectronic materials and devices, specifically to a LIF neuron circuit based on a self-powered threshold-switched memristor device and its fabrication method. Background Technology

[0002] Artificial neurons mainly consist of an integration component and an activation process. Threshold-type TS memristor devices can be used to perform the activation action, while the integration function (or accumulation process) is usually implemented by peripheral circuits with resistors and capacitors. In traditional neuronal circuit systems, when the sensor module collects analog information, it needs to be converted into digital information by an analog-to-digital converter, stored in memory, and then transmitted to the processing unit. The limited bandwidth between the sensor node and the external processing unit hinders the data processing speed and is limited by the external power supply, thus generating a large amount of energy consumption and safety issues. Since triboelectric nanogenerators can sense external mechanical force signals and directly convert them into electrical signals for output, developing a self-powered threshold-switched memristor device-based LIF neuronal circuit can effectively solve the above problems, laying the foundation for building more complex neuromorphic network hardware, and further developing high-density neuromorphic computing systems. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a self-powered threshold-switched memristor (TS) memristor circuit for LIF neurons and its fabrication method. By utilizing a triboelectric nanogenerator for self-powered operation and a threshold-switched TS memristor device along with other circuit components, the simulation of an artificial neuron is successfully achieved. The artificial neuron circuit of this invention has the advantages of simple and clear design and low power consumption.

[0004] To achieve the above objectives, the present invention provides a LIF neuron circuit based on a self-powered threshold-switched memristor device, comprising: a triboelectric nanogenerator, a rectifier bridge, a threshold-type TS memristor device, a fixed resistor, and a capacitor;

[0005] The triboelectric nanogenerator senses external mechanical signals and converts them into voltage pulses. After passing through a rectifier bridge, the output is a positive voltage pulse. After being divided by a fixed resistor, the voltage pulse is charged and discharged by a capacitor connected in parallel across the threshold-type TS memristor device, resulting in a continuous voltage spike pulse.

[0006] Preferably, the triboelectric nanogenerator comprises: an electrode PET / ITO, a positive electrode material PDMS, a negative electrode material PU, an electrode Al, and a substrate ITO / PET; wherein PET / ITO and Al serve as electrodes for PDMS and PU to collect charges.

[0007] Preferably, utilizing the unidirectional conductivity of diodes, four rectifier diodes are connected in a full-wave bridge rectifier circuit to form the rectifier bridge; the rectifier bridge converts positive and negative input AC into a single positive DC.

[0008] Preferably, in the threshold-type TS memristor device, the bottom electrode is an inert metal electrode Pt, and the top electrode is an active metal electrode Ag; the buffer layer is a phase change material AIST, used to control the Ag content. + Injection and diffusion rates; the medium layer is HfO2.

[0009] Preferably, the fixed resistor adjusts the voltage supplied to the threshold type TS memristor device by voltage division, and the resistance of the fixed resistor is 100MΩ.

[0010] Preferably, the switching behavior of the threshold-type TS device is verified by the charging and discharging behavior of the capacitor, wherein the capacitance value of the capacitor is 10μF.

[0011] Preferably, the working process of the threshold-type TS memristor includes: applying a small voltage and a small limiting current across the top and bottom electrodes of the threshold-type switched memristor to generate Ag. + Forming random conductive filaments; adding a buffer layer AIST to alleviate Ag + Injection speed to prevent Ag + The conductive filament is subjected to a process where the interface of the filament breaks and quickly returns to a high resistance.

[0012] This invention also provides a fabrication method for fabricating the LIF neuron circuit based on the self-powered threshold-switched memristor device described in claim 1. The fabrication process of the threshold-type TS memristor device includes: growing a bottom electrode Pt on an insulating substrate using magnetron sputtering at a power of 100W for 10 minutes; growing a dielectric layer HfO2 on the Pt bottom electrode using atomic layer deposition at a temperature of 150°C for 90 minutes; growing a buffer layer AIST on the dielectric layer HfO2 using magnetron sputtering at a power of 60W for 5 minutes; and finally depositing a top electrode on the buffer layer AIST using thermal evaporation.

[0013] Preferably, the preparation method of the triboelectric nanogenerator includes: first, mixing PDMS prepolymer and curing agent in a ratio of 10:1, then filtration under vacuum and curing in an oven before demolding, and then assembling them in the order of PET / ITO, PDMS, PU, ​​Al, and ITO / PET.

[0014] Preferably, the prepared dielectric layer HfO2 has a thickness of 20 nm; the buffer layer AIST has a thickness of 15 nm; the bottom electrode Pt has a thickness of 50 nm; and the top electrode Ag has a thickness of approximately 50 nm.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0016] This invention significantly reduces the dependence of traditional artificial neurons on external power sources and the overall energy consumption of the system. At the same time, through the coordinated charging and discharging behavior of memristors and capacitors, it efficiently simulates the key dynamic characteristics of biological neurons, such as leakage, integration, firing, and refractory period, and exhibits a frequency encoding function that depends on the stimulus intensity. This provides an effective hardware implementation scheme for building a low-power, highly integrated sensory-computing neuromorphic system. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of the threshold-type TS memristor device prepared according to the present invention;

[0019] Figure 2 This is a structural diagram of the threshold-type TS memristor device prepared according to the present invention;

[0020] Figure 3 Figure IV shows the basic electrical characteristics of the threshold-type TS memristor device prepared in this invention;

[0021] Figure 4 A flowchart illustrating the triboelectric nanogenerator (TENG) prepared according to this invention;

[0022] Figure 5 This is a structural diagram of the triboelectric nanogenerator (TENG) prepared in this invention;

[0023] Figure 6 The diagram shows the electrical properties of the triboelectric nanogenerator (TENG) prepared in this invention.

[0024] Figure 7 The circuit diagram of the LIF neuron based on a self-powered threshold-switched memristor device constructed in this invention is shown below.

[0025] Figure 8 The two-channel response VT diagram of an oscilloscope connected in parallel across the rectifier bridge, the threshold device, and the capacitor in the LIF neuron circuit based on a self-powered threshold switch memristor device constructed for this invention.

[0026] Figure 9In the LIF neuron circuit based on a self-powered threshold-switched memristor device constructed for this invention, the frequency response of the oscilloscope channel connected in parallel across the threshold device and the capacitor is magnified to show five different amplitudes of the VT response of the oscilloscope channel connected in parallel across the rectifier bridge.

[0027] Figure 10 The LIF neuron circuit based on a self-powered threshold-switched memristor device constructed for this invention presents a statistical graph of the frequency response of the oscilloscope channel connected in parallel across the threshold device and the capacitor, along with the voltage frequency response of the oscilloscope channel connected in parallel across the rectifier bridge at five different amplitudes. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Before proceeding with the explanation, the technical terms used in this invention will be explained.

[0031] PDMS: full name polydimethylsiloxane, a widely used organosilicon polymer known for its flexibility, biocompatibility and insulation.

[0032] PU: full name polyurethane, a multifunctional polymer that comes in various forms, from rigid foam to elastic fiber.

[0033] ITO: Indium tin oxide, a transparent conductive oxide thin film.

[0034] PET: full name polyethylene terephthalate, a very common transparent, lightweight, strong plastic with good chemical stability.

[0035] Al: Aluminum, a common metal with good electrical conductivity and cost-effectiveness.

[0036] Ag: Silver, a precious metal with excellent electrical conductivity.

[0037] AIST: short for silver-indium-antimony-tellurium (usually referring to Ag-In-Sb-Te), is a common phase change material, also used in some memristor devices.

[0038] HfO2: Hafnium dioxide, a metal oxide material with a high dielectric constant (high-k), is commonly used as a gate dielectric layer in microelectronics.

[0039] Pt: Platinum, a precious metal with very stable (inert) chemical properties.

[0040] PET / ITO: A transparent ITO (indium tin oxide) conductive film is deposited on a transparent PET film. This provides both a flexible substrate (PET) and a transparent electrode (ITO).

[0041] ITO / PET: It has the same structure as PET / ITO, but emphasizes that the ITO layer is on top.

[0042] Example 1

[0043] This embodiment provides a LIF neuron circuit based on a self-powered threshold-switched memristor device, including: a triboelectric nanogenerator (TENG), a rectifier bridge, a threshold-type TS memristor device, a fixed resistor, and a capacitor.

[0044] The TENG comprises electrodes PET / ITO, a positive electrode material PDMS, a negative electrode material PU, an electrode Al, and a substrate ITO / PET; wherein PET / ITO and Al serve as electrodes for PDMS and PU, respectively, collecting charge. In this embodiment of the TENG, when an external force is applied, the two triboelectric materials, PDMS and PU, come into contact with each other. Due to the triboelectric effect, an equal amount of surface charge with opposite polarity is transferred between them, and electrons are transferred from the amplitude PU to the PDMS, resulting in negatively charged PDMS and positively charged PU. During the release process, a charge flow is generated between the electrodes to maintain charge balance. Due to the electrostatic induction effect, after the electrodes are released, they retain the same amount of opposite charge with the corresponding triboelectric materials. When an external force is applied again, the electrostatic balance between the electrodes and the corresponding triboelectric materials is broken, thereby generating a charge flow. Under the cyclic action of compression and release, a mixed positive and negative transient output voltage is generated.

[0045] In threshold-type TS memristor devices, the bottom electrode is an inert metal electrode Pt, and the top electrode is an active metal electrode Ag; the buffer layer is a phase change material AIST, used to control the Ag content. + Injection and diffusion rates; the dielectric layer is HfO2. In threshold-type switched memristor devices, Ag can be generated by applying a small voltage and a small limiting current to the top and bottom electrodes of the device. + The formation of random conductive filaments and the addition of the buffer layer AIST can effectively alleviate Ag. + Injection speed to prevent the formation of coarse and difficult-to-break Ag. + Conductive filaments are designed to minimize the fragility of the filament's interface and allow it to quickly recover to high resistance.

[0046] Utilizing the unidirectional conductivity of diodes, four rectifier diodes are connected in a full-wave bridge rectifier circuit to form a rectifier bridge; the rectifier bridge converts positive and negative input AC to a single positive DC. A fixed resistor with a resistance of 100MΩ is used to adjust the voltage across the threshold-type TS memristor device via voltage division. The switching behavior of the threshold-type TS device is verified by the charging and discharging behavior of a capacitor with a capacitance of 10μF.

[0047] The principle behind the LIF behavior of biological neurons in this embodiment is as follows:

[0048] First, the triboelectric nanogenerator senses mechanical signals of varying external forces and directly converts them into electrical signals. After rectification by a rectifier bridge and voltage division by a fixed resistor, a stable positive DC signal is output. At this point, the capacitor connected in parallel across the threshold device in the high-resistivity (HRS) state begins to charge. When the voltage across the capacitor exceeds the threshold turn-on voltage (Vth) of the threshold device, the threshold device instantly transitions from the high-resistivity (HRS) state to the low-resistivity (LRS) state, and the capacitor begins to discharge. When the voltage across the capacitor drops to the holding voltage (Vhold) of the threshold device, the threshold device spontaneously returns to the high-resistivity (HRS) state, and the capacitor re-enters the charging state. This circuit continuously cycles, enabling the continuous integration and activation of neurons.

[0049] Integration and activation are important characteristics of neurons. Utilizing this neuronal characteristic, this embodiment achieves the key characteristic that the frequency of the output pulse increases with the increase of the pulse amplitude through the LIF neuron circuit based on the self-powered threshold-switched memristor device described above.

[0050] Example 2

[0051] This embodiment provides a fabrication method for fabricating a LIF neuron circuit based on a self-powered threshold-switched memristor device.

[0052] like Figure 1 The diagram shows a flowchart of the threshold-type TS memristor device fabricated in this embodiment. The specific fabrication process steps are as follows:

[0053] Step 1 (100): Clean the insulating substrate sequentially with trichloroethylene, acetone, ethanol, and deionized water in an ultrasonic cleaner at 60W power for 10 minutes to remove dust and impurities from the substrate surface and then blow it dry with an air gun.

[0054] Step 2 (101): The bottom electrode Pt is grown on the insulating substrate by magnetron sputtering at a growth pressure of 1 Pa, a growth power of 100 W, and a growth time of 10 minutes.

[0055] Step 3 (102): A dielectric layer HfO2 is grown on the Pt bottom electrode using atomic layer deposition. The growth pressure is about 2.67 Pa, the growth temperature is 150 °C, and the growth time is 90 minutes.

[0056] Step 4 (103): A buffer layer AIST is grown on an HfO2 thin film substrate by magnetron sputtering at a growth pressure of 1 Pa, a growth power of 60 W, and a growth time of 5 minutes.

[0057] Step 5 (104): The active metal electrode Ag is deposited on the buffer layer AIST using a thermal evaporation coating method. The growth pressure is 3E-4Pa, the mask aperture is about 300μm, and the thickness is about 50nm.

[0058] Figure 2 The structural diagram of the threshold-type TS memristor device prepared by this invention is shown. Electrical tests were performed on the threshold-type TS memristor device, and its electrical IV characteristic diagram is shown below. Figure 3 As shown, it exhibits obvious bipolar threshold resistive switching behavior. When a 1V voltage is applied across the device, the threshold turn-on voltage (V) of the device is approximately 0.5V. th Under these conditions, the resistor transitions from a high-resistance state to a low-resistance state, and the current limit is set to 100nA. When the voltage retraces back to approximately 0.15V, the voltage is maintained at (V). hold When a negative voltage of the same magnitude is applied, the resistance will spontaneously return from a low resistance state to a high resistance state. When a negative voltage of the same magnitude is applied, the IV diagram exhibits the same symmetry as the positive voltage, also known as the bidirectional threshold characteristic.

[0059] Figure 4 The flowchart for the triboelectric nanogenerator (TENG) prepared according to this invention is as follows:

[0060] Step 1 (400): Mix the PDMS prepolymer and curing agent thoroughly at a ratio of 10:1 and pour the mixture into the mold;

[0061] Step 2 (401): Place the mold in a vacuum chamber under a pressure of 0.06 MPa for 30 minutes to defoam;

[0062] Step 3 (402): After the defoaming process is completed, place the mold in a 65℃ oven to cure for 4 hours;

[0063] Step 4 (403): Cover the peeled PDMS film onto the Al film that encapsulates the PU film;

[0064] Step 5 (404): Assemble in the order of PET / ITO, PDMS, PU, ​​Al, ITO / PET;

[0065] Figure 5The schematic diagram shows the structure of the triboelectric nanogenerator (TENG) prepared in this invention. Electrical tests were performed on the TENG, and its electrical VT characteristic diagram is shown below. Figure 6 As shown, under the conditions of 3Hz frequency and 10N external force, TENG can stably output a positive and negative cycle voltage of up to 15V within 10s. This shows that by controlling different external forces, voltage pulses far exceeding the threshold voltage of the TS memristor device can be easily output. The different amplitude voltages output under different external forces are used to verify the key characteristic that the frequency of the output pulse increases with the increase of the pulse amplitude of the neuron.

[0066] Next, build as follows Figure 7 The self-powered LIF neuron circuit shown generates a mixed positive and negative transient output voltage via TENG and outputs it to the rectifier bridge, where it is converted into a uniform and stable positive voltage. After voltage division by the fixed resistor Rs, the resistance state change of the threshold-type TS memristor device can correspond to the charging and discharging behavior of the capacitor due to the parallel connection of the capacitor and the capacitor. As the capacitor is continuously charged and discharged, the circuit cycles to realize the integration and activation behavior of the neuron.

[0067] The integration and firing behavior of neurons can be visually observed using an oscilloscope connected in parallel across the threshold device and capacitor, such as... Figure 8 As shown: Under the condition of selecting a 100MΩ fixed resistor and a 10Mf capacitor, a voltage with an amplitude of about 3.2V is generated by a triboelectric nanogenerator (obtained by channel 1). Since the capacitor and the threshold device are connected in parallel, the charging and discharging behavior of the capacitor and the resistance switching behavior of the threshold device can be clearly observed after amplification (the magnified view pointed to by the dashed line).

[0068] Next, the voltage amplitudes (1.8V-3.2V) output by the TENG were tested within a local time range of 0-400μs. Figure 9 As shown, the different frequency neuronal firing spikes that can be monitored after amplification are caused by the resistance switching of the memristor, ensuring the generation of continuous voltage spikes. Correspondingly, the peak and trough values ​​of the voltage oscillation curve are V0 and V1, respectively. th and V hold Finally, as Figure 10 As shown, we statistically analyzed five different firing peak frequencies of neurons under different voltage amplitudes (1.8V-3.2V). It can be clearly observed that as the applied voltage amplitude increases, the firing peak frequency of neurons also increases significantly, reaching a firing frequency of up to 18kHz. The two show a clear positive correlation, confirming the intensity-dependent frequency response characteristics of biological neurons.

[0069] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A LIF neuron circuit based on self-powered threshold switching memristive devices, characterized in that, It comprises: a friction nanogenerator, a rectifier bridge, a threshold type TS memristor device, a fixed resistance and a capacitor; The external mechanical signal is perceived by the friction nanogenerator and converted into a voltage pulse output, and the output is a uniform positive voltage pulse after flowing through the rectifier bridge. After voltage division by the fixed resistance, the capacitor connected in parallel to the threshold type TS memristor device performs charging and discharging behavior, and outputs a continuous voltage spike pulse.

2. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 1, wherein, The friction nanogenerator comprises: electrode PET / ITO, positive electrode material PDMS, negative electrode material PU, electrode Al, and substrate ITO / PET; wherein PET / ITO and Al are used as the electrodes of PDMS and PU to collect charges.

3. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 1, wherein, Four rectifier diodes are connected according to a full-wave bridge rectifier circuit to form the rectifier bridge by utilizing the unidirectional conduction characteristic of the diodes; the rectifier bridge converts positive and negative input alternating current into single positive direct current.

4. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 1, wherein, In the threshold type TS memory device, the bottom electrode is an inert metal electrode Pt, the top electrode is an active metal electrode Ag; the buffer layer is a phase change material AIST, used for controlling Ag + Injection and diffusion speed; dielectric layer is HfO2.

5. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 4, wherein, The fixed resistance adjusts the voltage input to the two ends of the threshold type TS memristor device by voltage division, and the resistance value of the fixed resistance is 100MΩ.

6. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 4, wherein, The switching behavior of the threshold type TS device is verified by the charging and discharging behavior of the capacitor, and the capacitance value of the capacitor is 10μF.

7. The LIF neuron circuit based on self-powered threshold switching memristive devices of claim 4, wherein, The working process of the threshold type TS resistive device includes: applying a small voltage and a small limiting current between the top and bottom electrodes of the threshold type switching resistive device to generate Ag + Forming random conductive filaments; adding a buffer layer AIST to alleviate Ag + Injection speed, prevent the emergence of Ag + Conductive filaments, make the interface of the filament break and quickly recover to high resistance.

8. A method of manufacturing, the method for manufacturing a LIF neuron circuit based on a self-powered threshold switching memristive device according to any one of claims 1-7, characterized in that, The preparation process of the threshold type TS memristor device comprises: growing a bottom electrode Pt on an insulating substrate by a magnetron sputtering method, the growth power is 100W, and the growth time is 10 minutes; growing a medium layer HfO2 on the Pt bottom electrode by an atomic layer deposition method, the growth temperature is 150℃, and the growth time is 90 minutes; then growing a buffer layer AIST on the medium layer HfO2 by a magnetron sputtering method, the growth power is 60W, and the growth time is 5 minutes; and finally evaporating a top electrode on the buffer layer AIST by a thermal evaporation method.

9. The preparation method according to claim 8, characterized in that, The preparation method of the friction nanogenerator comprises: firstly, mixing PDMS prepolymer and curing agent in a ratio of 10:1, then filtering under vacuum and curing and demolding in an oven, and then pasting and combining in the order of PET / ITO, PDMS, PU, Al, and ITO / PET.

10. The preparation method according to claim 8, characterized in that, The thickness of the prepared medium layer HfO2 is 20nm; the thickness of the buffer layer AIST is 15nm; the thickness of the bottom electrode Pt is 50nm; and the thickness of the top electrode Ag is about 50nm.

Images