Compact artificial neuron based on TaOx multifunctional device

Abstract

The invention belongs to the related field of artificial nerve morphology, and discloses a compact artificial neuron based on TaOx multifunctional devices, the artificial neuron is composed of two TaOx multifunctional devices with the same structure, and the two TaOx multifunctional devices are connected in series back to back; wherein when voltage in a first direction is applied to the TaOx multifunctional device, the TaOx multifunctional device shows a charge trapping and releasing mechanism and is used for providing simulated volatile characteristics; and when the voltage in the second direction is applied, a conductive wire mechanism is shown and is used for providing digital volatile characteristics. The invention further discloses a structural design of the TaOx multifunctional device. According to the invention, the required neuron function can be realized by only relying on the back-to-back connection of two same TaOx multifunctional devices without additionally introducing capacitors, resistors and other devices, the circuit structure is compact and convenient to process and manufacture, and the problems of low integration level, high process difficulty and the like of the existing artificial neuron are effectively improved. Therefore, large-scale integration of the artificial neural network is facilitated.

Description

Technical Field

[0001] The present invention belongs to the field of artificial neural morphology, and more specifically, relates to a TaO-based x Compact artificial neurons for multifunctional devices. Background Art

[0002] The spiking neural network (SNN) belongs to the third generation of neural network models. It is different from the second generation neural network based on multilayer perceptrons, and it achieves a higher level of biological neural simulation. Among them, the most important difference lies in the neuron part. The neurons of the second generation network are fixed activation functions and only have zero-order dynamic characteristics, while the neurons in the spiking neural network have richer dynamic characteristics, which can better process spatiotemporal information. These artificial neurons have a variety of different model types, ranging from the most accurate biophysical neuron model to its simplest release model. Among them, the neuron model based on leakage-integration-ignition is one of the most commonly used artificial neuron models because of its simple setting and efficient coding ability, which is in the best position between biological rationality and practicality.

[0003] In the prior art, when the related functions of leak-integrate-fire neurons are realized in hardware, two different units are generally used to respectively realize the membrane potential function of leak-integrate and the threshold switching function of ignition. Among them, the membrane potential function can be realized by a unit with time dynamic characteristics such as a resistor-capacitor circuit (RC circuit) and a dynamic memristor, and the threshold function can be realized by a threshold circuit, a threshold device such as a metal conductive wire memristor and a metal-insulator transition device.

[0004] However, further research shows that the above existing solutions still have the following defects or shortcomings: they all require different types of hardware combinations, the circuit of a single leakage-integration-ignition artificial neuron is complex, and the capacitor is difficult to miniaturize; in large-scale artificial neural network circuits, complex neuron circuits will multiply the difficulty of the process and make it difficult to improve the integration of the system. Accordingly, this field urgently needs to conduct research and improvements to improve the technical problems of low integration and high process difficulty of existing artificial neurons. Summary of the invention

[0005] In view of one or more of the above defects or improvement requirements of the prior art, the present invention provides a TaO-based xA compact artificial neuron of a multifunctional device, wherein by combining the working characteristics and specific needs of artificial neurons, the specific structural composition and working mechanism of its key components are redesigned in a targeted manner, so that a more compact artificial neuron can be obtained, which not only does not require the participation of additional devices such as capacitors and resistors, but also simplifies the manufacturing process and does not need to consider issues such as resistance matching during the series connection. Therefore, it can effectively solve the problems of low integration and complex process of existing artificial neurons, and is particularly beneficial to large-scale integrated applications of artificial neural networks.

[0006] To achieve the above object, according to the present invention, a TaO-based x A compact artificial neuron with multifunctional devices, characterized by:

[0007] The artificial neuron consists of two TaO x The multifunctional devices are connected in series in a back-to-back manner; wherein,

[0008] The TaO x The multifunctional device exhibits a charge trapping and releasing mechanism for providing analog volatile characteristics when a voltage is applied in a first direction, and exhibits a conductive filament mechanism for providing digital volatile characteristics when a voltage is applied in a second direction;

[0009] When the above compact neuron is working, one of the TaO x Multifunctional device, the first TaO x The resistive switching mechanism of the multifunctional device is the charge trapping and release mechanism. Another TaO x Multifunctional devices, also known as the second TaO x The resistive switching mechanism of the multifunctional device is a conductive wire mechanism, which is used to realize the leakage-integration-ignition function of artificial neurons.

[0010] As a further preferred embodiment of the present invention, for the TaO x For a multifunctional device, it is preferred that a first electrode layer, a functional layer, a barrier layer and a second electrode layer are stacked in sequence from bottom to top, wherein the first electrode layer and the second electrode layer are made of inert metal materials and active metal materials respectively to form different electrodes; the functional layer is made of TaO x The material is made and x is less than 2.5, so that oxygen vacancies exist in the functional layer to form trap energy levels; the barrier layer is made of an inert metal material and is used to block the metal atoms of the second electrode layer from entering the functional layer.

[0011] As a further preferred embodiment of the present invention, the first electrode layer and the barrier layer are respectively made of one of the following inert metal materials: Pt, Pd, Au, Ni, ITO, Ti, W, Al, Ta, Hf, TaN, TiN and TiW; the second electrode layer is made of the following active metal materials: Ag, Cu or other similar materials.

[0012] As a further preferred embodiment of the present invention, the thickness of the first electrode layer is designed to be 5nm-2000nm, the thickness of the functional layer is designed to be 2nm-100nm, the thickness of the barrier layer is designed to be 2nm-10nm, and the thickness of the second electrode layer is designed to be 10nm-2000nm.

[0013] As a further preferred embodiment of the present invention, for the first TaO x For the multifunctional device, when the first electrode layer is grounded and a negative voltage is applied to the second electrode layer, an analog volatile characteristic is exhibited, which can be used to realize the leakage and integration functions of neurons.

[0014] As a further preferred embodiment of the present invention, for the second TaO x For a multifunctional device, when the first electrode layer is grounded and the second electrode layer is initialized by applying a positive bias, a forward voltage is applied, and a digital volatile characteristic is exhibited, which can be used to realize the ignition and leakage functions of neurons. As a further preferred embodiment of the present invention, in order to realize the leakage-integration-ignition function of the above-mentioned artificial neuron, the first TaO x Multifunctional devices, second TaO x The multifunctional devices can be connected in series via the respective first electrode layers, and wherein the first TaO x A pulse in a negative bias direction is applied to the second electrode layer of the multifunctional device, and the second TaO x The second electrode layer of the multifunctional device is grounded.

[0015] As a further preferred embodiment of the present invention, when the first TaO x When the amplitude of the negative bias pulse applied to the second electrode layer of the multifunctional device is greater than the preset amplitude, the neuron realizes the ignition function, and the number of pulses required decreases as the amplitude or frequency of the applied pulse increases.

[0016] In general, the above technical solution conceived by the present invention has the following technical advantages compared with the prior art:

[0017] (1) The artificial neuron of the present application consists of only two identical TaO xThe multifunctional devices can be connected in series back to back, and voltages in different directions are applied to make them have different electrical characteristics to achieve different functions. When a negative voltage is applied, the device resistance switching mechanism is a charge capture and release mechanism, which exhibits analog volatile characteristics and can be used to achieve the functions of integration and leakage in neurons. When a positive voltage is applied, the device resistance switching mechanism is a conductive wire mechanism, which exhibits digital volatile characteristics and can be used to achieve the functions of ignition and leakage in neurons. In this way, not only a more compact product can be provided, but also the functions of integration leakage ignition of artificial neurons can be achieved without the need for additional devices such as capacitors and resistors.

[0018] (2) This application further discloses TaO x The specific structural composition of the multifunctional device and its key processing parameters have been optimized, and accordingly, there is no need to consider issues such as resistance state matching during the series connection process, and the preparation of two devices can be achieved in a single lithography process. Therefore, the compact artificial neuron of the present application effectively improves the problems of low integration and high process difficulty of existing artificial neurons, and can improve the integration of large-scale neuromorphic computing circuits while reducing their process complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 According to a preferred embodiment of the present invention, a TaO x Schematic diagram of a compact artificial neuron for a multifunctional device;

[0020] Figure 2 yes Figure 1 The TaO x Schematic diagram of the structure of the multifunctional device;

[0021] Figure 3 This is a specific embodiment of the present invention (Pt / TaO x / Ti / Ag multifunctional device) was subjected to 100 consecutive negative bias scans and obtained a voltage-current curve;

[0022] Figure 4 This is a specific embodiment of the present invention (Pt / TaO x / Ti / Ag multifunctional device) applied a pulse test response diagram in the negative voltage direction, where the amplitude of the applied pulse is -6V and the frequency is 5Hz;

[0023] Figure 5 This is a specific embodiment of the present invention (Pt / TaO x / Ti / Ag multifunctional device) after applying 1uA current limiting and 0-0.5V initialization operation, and applying 100 consecutive forward bias scans to obtain the voltage-current diagram;

[0024] Figure 6The relevant test circuit diagram for the compact artificial neuron of the present invention is shown exemplarily;

[0025] Figure 7 is a response diagram obtained after applying continuous pulses to the compact neuron of the present invention, wherein the applied pulse amplitude is -6V and the frequency is 6.25Hz;

[0026] Figure 8 1 is a response diagram of the compact artificial neuron of the present invention under pulse tests at different amplitudes, wherein the applied pulse amplitudes are respectively -5.8V, -5.9V, -6V, -6.1V, and -6.2V, and the frequencies are all 6.66Hz;

[0027] Fig. 9 It is a response diagram of the compact artificial neuron of the present invention under pulse test at different frequencies, wherein the amplitude of the applied pulse is -6.1V, and the frequencies are 5Hz, 5.56Hz, and 7.14Hz respectively. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] It should be understood that expressions such as "include" and "may include" used in this application indicate the existence of the disclosed functions, operations or constituent elements, and do not limit one or more additional functions, operations and constituent elements. In this application, terms such as "include" and / or "have" may be interpreted as indicating specific characteristics, numbers, operations, constituent elements, components or combinations thereof, but may not be interpreted as excluding the existence or possibility of adding one or more other characteristics, numbers, operations, constituent elements, components or combinations thereof.

[0030] It should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0031] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0032] In this application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0033] Figure 1 According to a preferred embodiment of the present invention, a TaO x Schematic diagram of a compact artificial neuron with multifunctional devices. Figure 1 To explain the present invention in more detail.

[0034] like Figure 1 As shown in FIG. , the present invention is based on TaO x The compact artificial neuron of the multifunctional device consists of two structurally identical TaO x The multifunctional devices are connected in series back to back; when TaO x When a voltage is applied to the second electrode layer of the multifunctional device and the first electrode layer is grounded, a charge capture and release mechanism is exhibited when a negative voltage is applied, which is used to provide analog volatile characteristics; after the initialization operation is performed, a conductive filament mechanism is exhibited when a positive voltage is applied, which is used to provide digital volatile characteristics; when the above compact neuron is working, one of the TaO x The multifunctional device (shown as device A in the figure) is the first TaO x Multifunctional device, its resistance switching mechanism is charge capture and release mechanism, another TaO x The multifunctional device (shown as device B in the figure) is also the second TaO x Multifunctional device, its resistance switching mechanism is the conductive wire mechanism

[0035] See also Figure 2 According to a preferred embodiment of the present invention, for the TaO xFor a multifunctional device, it is preferred that a first electrode layer, a functional layer, a barrier layer and a second electrode layer are stacked in sequence from bottom to top, wherein the first electrode layer and the second electrode layer are made of inert metal materials and active metal materials respectively to form different electrodes; the functional layer is made of TaO x The material is made and x is less than 2.5, so that oxygen vacancies exist in the functional layer to form trap energy levels; the barrier layer is made of an inert metal material and is used to block the metal atoms of the second electrode layer from entering the functional layer.

[0036] According to another preferred embodiment of the present invention, the first electrode layer and the barrier layer are respectively made of one of the following inert metal materials: Pt, Pd, Au, Ni, ITO, Ti, W, Al, Ta, Hf, TaN, TiN and TiW; the second electrode layer is made of the following active metal materials: Ag, Cu or other similar materials.

[0037] According to another preferred embodiment of the present invention, the thickness of the first electrode layer is designed to be 5nm-2000nm, the thickness of the functional layer is designed to be 2nm-100nm, the thickness of the barrier layer is designed to be 2nm-10nm, and the thickness of the second electrode layer is designed to be 10nm-2000nm.

[0038] In a specific application scenario, for the first TaO x For a multifunctional device, if its first electrode layer is grounded and a negative voltage is applied to the second electrode layer, the device's resistive switching mechanism is a charge capture and release mechanism with analog volatility, which can be used to realize the leakage and integration functions of neurons.

[0039] In another specific application scenario, for the second TaO x For a multifunctional device, after it performs the initialization operation of current limiting forward voltage, if its first electrode layer is grounded and a forward voltage is applied to the second electrode layer, the device's resistive switching mechanism is a conductive wire mechanism with digital volatility, which can be used to realize the ignition and leakage functions of neurons.

[0040] In another specific application scenario, when the above compact artificial neuron is working, the first TaO x The first electrode layer and the second TaO x The first electrode layer of the multifunctional device is connected, and the first TaO x A negative bias pulse is applied to the second electrode layer of the multifunctional device, and the second TaO x The second electrode layer of the multifunctional device is grounded. When the amplitude of the applied negative bias pulse is greater than the preset amplitude, the neuron can realize the ignition function, and the number of required pulses decreases as the amplitude or frequency of the applied pulse increases.

[0041] exist Figure 2 In the specific application scenario shown, the arrows indicate the direction of the current between the two devices A and B when the compact artificial neuron is working; x The multifunctional devices, i.e., devices A and B, are connected in series through the first electrode layer; in addition, a negative voltage pulse is input to the second electrode layer of device A, and the second electrode layer of device B is grounded. Accordingly, device A will use its analog volatile characteristics under negative bias to realize the functions of integration and leakage in neurons; after device B performs an initialization operation with a current limit of 1uA and 0-0.5V, it uses its digital volatile characteristics under positive bias to realize the functions of ignition and leakage in neurons.

[0042] Some specific embodiments are given below to explain the present invention more clearly.

[0043] Example 1

[0044] In this embodiment, the TaO x The multifunctional device comprises a first electrode layer, a functional layer, a barrier layer and a second electrode layer stacked in sequence from bottom to top; wherein the functional layer is TaO x Thin film, wherein x=2.042; the material of the first electrode layer is Pt; the material of the barrier layer is Ti; the material of the second electrode layer is Ag; in addition, the thickness of the first electrode layer is 100nm, the thickness of the functional layer is 10nm, the thickness of the barrier layer is 5nm, and the thickness of the second electrode layer is 100nm.

[0045] Example 2

[0046] In this embodiment, the TaO x The multifunctional device comprises a first electrode layer, a functional layer, a barrier layer and a second electrode layer stacked in sequence from bottom to top; wherein the functional layer is TaO x Thin film, wherein x=2.042; the material of the first electrode layer is Au; the material of the barrier layer is Pt; the material of the second electrode layer is Ag; in addition, the thickness of the first electrode layer is 5nm, the thickness of the functional layer is 2nm, the thickness of the barrier layer is 2nm, and the thickness of the second electrode layer is 10nm.

[0047] Example 3

[0048] In this embodiment, the TaO x The multifunctional device comprises a first electrode layer, a functional layer, a barrier layer and a second electrode layer stacked in sequence from bottom to top; wherein the functional layer is TaO xThin film, wherein x=2.042; the material of the first electrode layer is ITO; the material of the barrier layer is TaN; the material of the second electrode layer is Cu; in addition, the thickness of the first electrode layer is 30nm, the thickness of the functional layer is 10nm, the thickness of the barrier layer is 3nm, and the thickness of the second electrode layer is 300nm.

[0049] Example 4

[0050] In this embodiment, the TaO x The multifunctional device comprises a first electrode layer, a functional layer, a barrier layer and a second electrode layer stacked in sequence from bottom to top; wherein the functional layer is TaO x Thin film, wherein x=2.042; the material of the first electrode layer is TiN; the material of the barrier layer is W; the material of the second electrode layer is Cu; in addition, the thickness of the first electrode layer is 2000nm, the thickness of the functional layer is 100nm, the thickness of the barrier layer is 10nm, and the thickness of the second electrode layer is 2000nm.

[0051] Example 5

[0052] In this embodiment, the TaO x The multifunctional device comprises a first electrode layer, a functional layer, a barrier layer and a second electrode layer stacked in sequence from bottom to top; wherein the functional layer is TaO x Thin film, wherein x=2.042; the material of the first electrode layer is Ni; the material of the barrier layer is TaN; the material of the second electrode layer is Ag; in addition, the thickness of the first electrode layer is 1000nm, the thickness of the functional layer is 70nm, the thickness of the barrier layer is 7nm, and the thickness of the second electrode layer is 1000nm.

[0053] The following is the Pt / TaO x / Ti / Ag multifunctional devices were subjected to a series of tests.

[0054] like Figure 3 As shown in Figure 2, the Pt / TaO x The voltage-current curve of the / Ti / Ag multifunctional device obtained by applying 100 consecutive negative bias scans. The Pt electrode is grounded and the Ag electrode is subjected to a negative scanning bias of 0 to -3 to 0 V. The test shows that the device has simulated volatility and can realize the function of neuronal leakage.

[0055] like Figure 4 As shown in Figure 2, the Pt / TaO xThe pulse test response diagram of the / Ti / Ag multifunctional device applied in the negative voltage direction; the Pt electrode is grounded, and the Ag electrode is applied with ten consecutive pulses in the negative bias direction, with a pulse amplitude of -6V and a frequency of 5Hz. The test shows that the device response increases with the increase of the number of pulses, and the function of neuron integration can be realized.

[0056] like Figure 5 As shown, the Pt / TaO x The current-voltage curve of the Ti / Ag multifunctional device obtained by applying a current limit of 1uA and an initialization operation of 0-0.5V and then applying a forward bias of 0-0.3-0V to continuously scan. The Pt electrode is grounded and the Ag electrode is applied with voltage. The test shows that the device has digital volatility and can realize the neuron ignition and leakage functions.

[0057] Based on the above electrical tests, it can be found that: (1) Multifunctional device Pt / TaO x When a negative bias is applied to the Ag electrode in / Ti / Ag, it exhibits a simulated volatile characteristic, which can be used to realize the integration and leakage processes; (2) Multifunctional device Pt / TaO x After the initialization operation of applying 1uA current-limiting forward voltage direction to the Ag electrode in / Ti / Ag, a forward bias is applied, which exhibits digital volatile characteristics and can be used to realize the functions of ignition and leakage.

[0058] Figure 6 The relevant test circuit diagram for the compact artificial neuron of the present invention is shown as an example. Figure 5 The first electrode layer connecting device A and device B, where the potential change is used to represent the change of membrane potential in biological neurons.

[0059] Figure 7 This is a response diagram obtained after applying continuous pulses to the compact neuron of the present invention, where the applied pulse amplitude is -6V and the frequency is 6.25Hz. As shown in the figure, as the number of applied pulses increases, device A integrates the input pulses, and the potential at point M increases continuously until the device resistance change threshold voltage is reached, and the resistance state of device B changes from a high resistance state to a low resistance state, realizing the ignition function.

[0060] Figure 8 It is a response diagram of the compact artificial neuron of the present invention under pulse tests at different amplitudes, wherein the applied pulse amplitudes are respectively -5.8V, -5.9V, -6V, -6.1V, and -6.2V, and the frequencies are all 6.66Hz; as shown in the figure, the compact artificial neuron can only achieve ignition when the applied pulse amplitude is greater than -5.8V, and the number of applied pulses required decreases as the applied pulse amplitude increases.

[0061] Fig. 9The following is a response diagram of the compact artificial neuron of the present invention under pulse tests at different frequencies, wherein the amplitude of the applied pulses is -6.1V, and the frequencies are 5Hz, 5.56Hz, and 7.14Hz, respectively. As shown in the figure, the number of pulses required for the compact neuron to achieve ignition decreases as the applied pulse frequency increases.

[0062] Based on the above electrical measurements, it can be found that (1) the compact artificial neuron can realize the functions of leakage, integration and ignition in neurons; (2) the response of the compact artificial neuron can be controlled by adjusting the input pulse frequency and amplitude; among them, the compact artificial neuron needs to apply a pulse amplitude greater than the threshold to achieve ignition, and the number of pulses required to achieve ignition decreases with the increase of the applied pulse amplitude; the number of pulses required for the compact artificial neuron to achieve ignition decreases with the increase of the applied pulse frequency.

[0063] In summary, the compact artificial neuron system according to the present invention can be realized by relying on only two identical TaO x By connecting the multifunctional devices back to back, the integral ignition function can be realized, and there is no need to introduce additional devices such as resistors and capacitors. The circuit structure is simple, which is conducive to the large-scale integration of artificial neural networks. Therefore, it is particularly suitable for large-scale integration applications of artificial neural networks and has good practical value and application prospects.

[0064] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A TaO-based x A compact artificial neuron with multifunctional devices, characterized by: The artificial neuron consists of two TaO x The multifunctional devices are connected in series in a back-to-back manner; wherein, The TaO x The multifunctional device exhibits a charge trapping and releasing mechanism for providing analog volatile characteristics when a voltage is applied in a first direction, and exhibits a conductive filament mechanism for providing digital volatile characteristics when a voltage is applied in a second direction; When the above compact neuron is working, one of the TaO x Multifunctional device, the first TaO x The resistive switching mechanism of the multifunctional device is the charge trapping and release mechanism. Another TaO x Multifunctional devices, also known as the second TaO x The resistive switching mechanism of the multifunctional device is a conductive wire mechanism, which is used to realize the leakage-integration-ignition function of artificial neurons.

2. The compact artificial neuron according to claim 1, characterized in that For the TaO x For a multifunctional device, it is preferred that a first electrode layer, a functional layer, a barrier layer and a second electrode layer are stacked in sequence from bottom to top, wherein the first electrode layer and the second electrode layer are made of inert metal materials and active metal materials respectively to form different electrodes; the functional layer is made of TaO x The material is made and x is less than 2.5, so that oxygen vacancies exist in the functional layer to form trap energy levels; the barrier layer is made of an inert metal material and is used to block the metal atoms of the second electrode layer from entering the functional layer.

3. The compact artificial neuron according to claim 2, characterized in that The first electrode layer and the barrier layer are respectively made of one of the following inert metal materials: Pt, Pd, Au, Ni, ITO, Ti, W, Al, Ta, Hf, TaN, TiN and TiW; the second electrode layer is made of the following active metal materials: Ag, Cu or other similar materials.

4. The compact artificial neuron according to claim 2 or 3, characterized in that The thickness of the first electrode layer is designed to be 5nm-2000nm, the thickness of the functional layer is designed to be 2nm-100nm, the thickness of the barrier layer is designed to be 2nm-10nm, and the thickness of the second electrode layer is designed to be 10nm-2000nm.

5. The compact artificial neuron according to any one of claims 2 to 4, characterized in that: For the first TaO x For the multifunctional device, when the first electrode layer is grounded and a negative voltage is applied to the second electrode layer, an analog volatile characteristic is exhibited, which can be used to realize the leakage and integration functions of neurons.

6. The compact artificial neuron according to any one of claims 2 to 5, characterized in that: For the second TaO x For the multifunctional device, when the first electrode layer is grounded and the second electrode layer is initialized by applying a positive bias direction, a forward voltage is applied, exhibiting a digital volatility characteristic, which can be used to realize the ignition and leakage functions of neurons.

7. The compact artificial neuron according to any one of claims 2 to 6, characterized in that: In order to realize the leakage-integration-firing function of the artificial neuron, the first TaO x Multifunctional devices, second TaO x The multifunctional devices can be connected in series via the respective first electrode layers, and wherein the first TaO x A pulse in a negative bias direction is applied to the second electrode layer of the multifunctional device, and the second TaO x The second electrode layer of the multifunctional device is grounded.

8. The compact artificial neuron according to claim 7, characterized in that When the first TaO x When the amplitude of the negative bias pulse applied to the second electrode layer of the multifunctional device is greater than the preset amplitude, the neuron realizes the ignition function, and the number of pulses required decreases as the amplitude or frequency of the applied pulse increases.

Images