A multi-state neural synapse device based on probability flip and a modulation method thereof

Abstract

The application provides a multi-state neural synapse device based on probability flip and a modulation method thereof, and relates to the technical field of semiconductor devices.The multi-state neural synapse device comprises a plurality of probability devices connected in series; wherein the resistance state of each probability device presents a probabilistic high resistance state or a low resistance state according to different input pulse current densities; the total output voltage of a plurality of probability devices connected in series or the total output voltage of multiple pulse currents of the same probability device is used to simulate the weight of a neuron synapse, so that an artificial synapse device for neuromorphic computing is realized.The multi-state neural synapse device can change the paradigm of neuromorphic computing from deterministic simulation to random bionic simulation, and has unique advantages in adaptive learning and uncertain reasoning scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor devices, in particular to a multi-state neural synapse device based on probability flipping and a modulation method thereof. BACKGROUND

[0002] In recent years, neuro-morphic computing technology imitating the human brain has attracted widespread attention. In order to enable a semiconductor device to implement the brain information transmission method, a high-performance artificial synapse device capable of simulating various synaptic connections is needed.

[0003] The probability device can have a deterministic magnetization flipping when a varying pulse current is injected under an out-of-plane magnetic field, and the thermal stability can cause the magnetic layer to have a random magnetization flipping when the current is injected. This phenomenon, in which the deterministic flipping and the probabilistic magnetization flipping based on the spin-orbit torque effect (SOT) or the spin transfer torque effect (STT) exist simultaneously, is called the rebound phenomenon.

[0004] By utilizing the characteristics of the above-mentioned probability device, an artificial synapse device for neuro-morphic computing can be realized. SUMMARY

[0005] The present application aims to provide a multi-state neural synapse device based on probability flipping and a modulation method thereof, which utilizes the multi-resistance state of the total voltage of a plurality of probability devices in series or the total output voltage of a plurality of pulse currents of the same probability device to simulate the weight of a neuron synapse, and realizes an artificial synapse device for neuro-morphic computing.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In one aspect, a multi-state neural synapse device based on probability flipping is provided, which comprises a plurality of probability devices in series; wherein the resistance state of each probability device presents a probabilistic high-resistance state or a low-resistance state according to different input pulse current densities; and the total voltage of a plurality of probability devices in series or the total output voltage of a plurality of pulse currents of the same probability device is used to simulate the weight of a neuron synapse.

[0008] Optionally, the probability device is a magnetic tunnel junction device or a Hall bar device.

[0009] Optionally, the magnetic tunnel junction device comprises, from bottom to top, a substrate, a spin current generation layer, a ferromagnetic layer, an intermediate non-magnetic insulating barrier layer, a magnetic pinning layer, and a readout electrode layer.

[0010] Optionally, the Hall bar device comprises, from bottom to top, a substrate, a first spin current generation layer, a second spin current generation layer, a ferromagnetic layer, and a cover layer.

[0011] Optionally, the probability device has different magnetization reversal for different pulse current density:

[0012] When the input pulse current density is greater than the critical reversal current density and less than the first threshold current density J0, a deterministic reversal occurs; when the input pulse current density is greater than the first threshold current density J0, a probability reversal occurs; when the input pulse current density continues to increase, greater than the second threshold current density J1, a random reversal occurs, that is, the upward and downward reversal probabilities of the magnetization state are both 50%.

[0013] Optionally, when the deterministic reversal occurs, the input pulse current density is 10×10 10 A·m -2 ~20×10 10 A·m -2 ; when the probability reversal occurs, the input pulse current density is 20×10 10 A·m -2 ~30×10 10 A·m -2 ; when the random reversal occurs, the input pulse current density is greater than 30×10 10 A·m -2 , and less than the maximum current density that the probability device can withstand.

[0014] In another aspect, a modulation method based on the multi-state neural synapse device of any of the above is provided, the method comprising:

[0015] For N probability devices, the high resistance state is 1 and the low resistance state is 0;

[0016] A pulse current with a pulse current density of J m is input to each of the probability devices, and the pulse current density corresponds to a reversal probability of p m , and the output total voltage expectation is N×p m , realizing multi-resistance state regulation;

[0017] Alternatively, a pulse current with different pulse current densities J n is input to each of the probability devices, respectively corresponding to different reversal probabilities p n , realizing multi-resistance state regulation.

[0018] Optionally, one probability device is reused to replace multiple probability devices; the input pulse current density is the same or different each time, and the number of resistance states depends on the number of input pulse currents, to realize multi-resistance state regulation.

[0019] The technical solution provided by the application has at least the following beneficial effects:

[0020] The application provides a multi-state neural synapse device based on probability flip and a modulation method thereof. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0022] Figure 1 Fig. 1 is a structural schematic diagram of a multi-state neural synapse device provided by the embodiments of the present application;

[0023] Figures 2a-2b Fig. 2 is a schematic diagram of a magnetic tunnel junction device based on SOT and STT provided by the embodiments of the present application;

[0024] Figure 3 Fig. 3 is a schematic diagram of a Hall bar device provided by the embodiments of the present application;

[0025] Figure 4 Fig. 4 is a schematic diagram of a rebound phenomenon provided by the embodiments of the present application;

[0026] Figure 5 Fig. 5 is a schematic diagram of a probability curve varying with current density provided by the embodiments of the present application. DETAILED DESCRIPTION

[0027] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be described clearly and completely in combination with the drawings of the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the described embodiments of the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the protection scope of the present application.

[0028] The embodiments of the present application provide a multi-state neural synapse device based on probability flip, as shown in Fig. 1. Figure 1As shown, the polymorphic neural synapse device comprises multiple probabilistic devices connected in series, with "+" indicating series connection. The resistance state of each probabilistic device exhibits a probabilistic high-resistance or low-resistance state depending on the input pulse current density. The total voltage of the series connection of multiple probabilistic devices or the total output voltage of multiple pulse currents from the same probabilistic device simulates the weight of the neuronal synapse, thus realizing an artificial synapse device for neuromorphic computation.

[0029] In this embodiment of the invention, the probability device is a magnetic tunnel junction device (MTJ) or a Hall bar device.

[0030] like Figures 2a-2b The figures shown are schematic diagrams of magnetic tunnel junction devices based on the spin-orbit moment (SOT) effect and the spin-transfer moment (STT) effect, respectively. The magnetic tunnel junction device, from bottom to top, includes: a substrate 1, a spin current generation layer 2, a ferromagnetic layer 3, an intermediate non-magnetic insulating barrier layer 4, a magnetic pinning layer 5, and a readout electrode layer 6. The arrows in the figures represent the first current (write current I). in ) and the second current (readout current V) out ) direction. Through the direction. Figures 2a-2b The spin flow generation layer 2 and the ferromagnetic layer 3 are sputtered separately to reduce thermal stability. When a pulsed current is injected, a phenomenon of deterministic flip and probabilistic magnetization flip will occur simultaneously based on the spin orbital moment effect (SOT) or spin transfer moment effect (STT), i.e., the rebound phenomenon.

[0031] Figure 3 This is a schematic diagram of a Hall bar device, which, from bottom to top, includes: a substrate 1, a first spin current generation layer 7, a second spin current generation layer 8, a ferromagnetic layer 3, and a capping layer 9. Similarly, the arrows represent the first current (write current I). in ) and the second current (readout current V) out ) direction. Through the direction. Figure 3 The first spin flow generation layer 7 and the second spin flow generation layer 8 are sputtered separately to reduce thermal stability. When a pulse current is injected, a phenomenon occurs where deterministic flipping and probabilistic magnetization flipping coexist, i.e., the rebound phenomenon.

[0032] This invention designs a polymorphic neural synapse device that can be used for neuromorphic computing based on the rebound phenomenon of probabilistic devices. Figure 4 This is a schematic diagram of the rebound phenomenon, where the probabilistic device exhibits different magnetization reversals in response to different pulse current densities:

[0033] When the input pulse current density is greater than the critical flipping current density and less than the first threshold current density J0, deterministic flipping occurs; when the input pulse current density is greater than the first threshold current density J0, probabilistic flipping occurs; when the input pulse current density continues to increase and is greater than the second threshold current density J1, random flipping occurs, that is, the flipping probabilities of the magnetization state upward and downward are both 50%. Figure 4 In the formula, ±J0 is deterministic flipping, ±J0 ~ ±J1 is probabilistic flipping, and greater than ±J1 is random flipping. The probability curve with the change of the current density is shown in FIG. 2. Figure 5

[0034] As an optional implementation, when the deterministic flipping occurs, the input pulse current density is 10×10 10 A·m -2 ~20×10 10 A·m -2 (the range of the deterministic flipping changes according to the size of the device); when the probabilistic flipping occurs, the input pulse current density is 20×10 10 A·m -2 ~30×10 10 A·m -2 (the range of the probabilistic flipping changes according to the size of the device, and does not coincide with the deterministic flipping current density); when the random flipping occurs, the input pulse current density is greater than 30×10 10 A·m -2 , and less than the maximum current density that the probabilistic device can withstand, that is, the upper limit of the current density depends on the current density that the probabilistic device can withstand.

[0035] Based on the above characteristics, in the embodiment of the application, the multi-resistance state of the total voltage of a plurality of probabilistic devices in series or the total output voltage of a plurality of pulse currents of the same probabilistic device is used to simulate the weight of a neuron synapse, and a multi-state neural synapse device based on probabilistic flipping is constructed.

[0036] Further, the application further provides a modulation method of the multi-state neural synapse device, comprising:

[0037] For N probabilistic devices, the high resistance state is 1, and the low resistance state is 0;

[0038] A pulse current with a pulse current density of J m is input to each of the probabilistic devices, the flipping probability corresponding to the pulse current density is p m , the output total voltage expectation is N×p m , and multi-resistance state adjustment is realized.

[0039] Alternatively, different pulse current densities J n ​pulse current, respectively, corresponding to different flipping probabilities p n , to realize multi-resistance state regulation.

[0040] As another embodiment of the application, one probability device can be reused to replace multiple probability devices; the input pulse current density is the same or different each time, and the number of resistance states depends on the number of input pulse currents, to realize multi-resistance state regulation.

[0041] Specifically, the embodiment of the application provides the following modulation schemes:

[0042] Scheme 1: For N probability devices, each probability device J m is given a pulse current density, and the pulse current density corresponds to a probability p m , and the output total voltage expectation is N x p m . In theory, in the case that the number of probability devices is sufficient and the voltage values can be distinguished, there can be infinite resistance states (in the traditional device, N devices have at most N resistance states).

[0043] Scheme 2: For N probability devices, each probability device is given a different current density J n , corresponding to different probabilities p n . Similarly, there can be infinite resistance states, but the adjustability of the synapse is greatly increased.

[0044] Scheme 3: On the basis of Scheme 1 and Scheme 2, multiple probability devices can be replaced by one probability device for reuse, and the current density injected each time can be the same or different, increasing the operability of neuromorphic computing, and the number of resistance states depends on the number of input currents, and in theory, the number of resistance states has no upper limit.

[0045] Compared with the prior art, the application provides a new multi-state neural synapse device based on probability flipping, which is composed of multiple probability devices in series, and the resistance state of each probability device presents a probabilistic high resistance state or low resistance state according to different input pulse current densities. The multi-resistance state of the total output voltage of the series connection of multiple probability devices or the total output voltage of multiple pulse currents of the same probability device can simulate the weight of a neuron synapse, thereby realizing an artificial synapse device for neuromorphic computing. The multi-state neural synapse device can change the paradigm of neuromorphic computing from deterministic simulation to random bionic simulation, has unique advantages in adaptive learning and uncertain reasoning scenarios, and can become a key component of the next generation of brain-like chips.

[0046] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a", "comprising", or "comprises" does not, without more constraints, preclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0047] Reference throughout this specification to "one embodiment", "an embodiment", "exemplary embodiment", "some embodiments", and so on means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment.

[0048] It should be understood that the term "and / or" as used herein merely indicates that there are three possible relationships between associated objects, for example, A and / or B can mean that A exists alone, A and B exist together, or B exists alone, where A and B can be singular or plural. In addition, the character " / " in the present document generally represents an "or" relationship between the associated objects, but can also represent an "and / or" relationship. The specific meaning can be understood according to the context before and after.

[0049] In the present document, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or similar expressions means any combination of the items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be singular or plural.

[0050] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0051] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0052] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0053] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0054] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0055] The present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be completely understood without the description of these details by those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, flows, elements and circuits, etc. are not described in detail.

[0056] The above merely describes preferred embodiments of the present application and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A polymorphic neural synapse device based on probability flipping, characterized in that, The polymorphic neural synapse device includes multiple probabilistic devices connected in series; wherein, the resistance state of each probabilistic device presents a probabilistic high-resistance state or a low-resistance state depending on the different input pulse current densities; the weight of the neuronal synapse is simulated by the total voltage of the series connection of the multiple probabilistic devices or the total output voltage of multiple pulse currents of the same probabilistic device.

2. The polymorphic neural synapse device according to claim 1, characterized in that, The probability device is a magnetic tunnel junction device or a Hall bar device.

3. The polymorphic neural synapse device according to claim 2, characterized in that, The magnetic tunnel junction device comprises, from bottom to top: a substrate, a spin flow generation layer, a ferromagnetic layer, an intermediate non-magnetic insulating barrier layer, a magnetic pinning layer, and a readout electrode layer.

4. The polymorphic neural synapse device according to claim 2, characterized in that, The Hall bar device comprises, from bottom to top: a substrate, a first spin current generation layer, a second spin current generation layer, a ferromagnetic layer, and a capping layer.

5. The polymorphic neural synapse device according to claim 1, characterized in that, The probability device exhibits different magnetization reversals for different pulse current densities: When the input pulse current density is greater than the critical reversal current density but less than the first threshold current density J0, a deterministic reversal occurs; when the input pulse current density is greater than the first threshold current density J0, a probabilistic reversal occurs; when the input pulse current density continues to increase and exceeds the second threshold current density J1, a random reversal occurs, meaning that the probability of reversing the magnetization state upwards and downwards is 50%.

6. The polymorphic neural synapse device according to claim 5, characterized in that, When a deterministic flip occurs, the input pulse current density is 10 × 10⁻⁶. 10 A·m -2 ~20×10 10 A·m -2 When a probability flip occurs, the input pulse current density is 20 × 10⁻⁶. 10 A·m -2 ~30×10 10 A·m -2 When random flips occur, the input pulse current density is greater than 30 × 10⁻⁶. 10 A·m -2 And it is less than the maximum current density that the probability device can withstand.

7. A modulation method based on the polymorphic neural synapse device according to any one of claims 1-6, characterized in that, The method includes: For N probability devices, let the high-resistance state be 1 and the low-resistance state be 0; The input pulse current density for each of the probabilistic devices is J m The pulse current, the switching probability corresponding to the pulse current density is p. m The expected total output voltage is then N×p. m To achieve multi-resistance state regulation; Alternatively, a different pulse current density J can be input to each of the probabilistic devices. n The pulse currents correspond to different flip probabilities p. n This enables multi-resistance state regulation.

8. The modulation method for the polymorphic neural synapse device according to claim 7, characterized in that, One probability device can be reused to replace multiple probability devices; the input pulse current density may be the same or different each time, and the number of resistance states depends on the number of input pulse currents, so as to achieve multi-resistance state regulation.

Images