CN111415001A - Electronic neuron and artificial neural network based on siganmin - Google Patents

Electronic neuron and artificial neural network based on siganmin Download PDF

Info

Publication number
CN111415001A
CN111415001A CN202010165391.8A CN202010165391A CN111415001A CN 111415001 A CN111415001 A CN 111415001A CN 202010165391 A CN202010165391 A CN 202010165391A CN 111415001 A CN111415001 A CN 111415001A
Authority
CN
China
Prior art keywords
magnetic
current
injection layer
current injection
nano
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010165391.8A
Other languages
Chinese (zh)
Other versions
CN111415001B (en
Inventor
梁雪
张溪超
周艳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chinese University of Hong Kong CUHK
Original Assignee
Chinese University of Hong Kong CUHK
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chinese University of Hong Kong CUHK filed Critical Chinese University of Hong Kong CUHK
Priority to CN202010165391.8A priority Critical patent/CN111415001B/en
Publication of CN111415001A publication Critical patent/CN111415001A/en
Application granted granted Critical
Publication of CN111415001B publication Critical patent/CN111415001B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Theoretical Computer Science (AREA)
  • Molecular Biology (AREA)
  • Neurology (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Hall/Mr Elements (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention relates to an electronic neuron based on a siganus oramin and an artificial neural network. The electronic neuron comprises a sigramin generating and driving device and an annular detecting device; wherein, the Sjgren seed generating and driving device comprises a magnetic nano-film and a current injection layer; the current injection layer and the annular detection device are both arranged on the magnetic nano film, and the current injection layer is arranged in the center of the annular detection device; the current injection layer is arranged to generate a magnetic skullet under the action of a generated current and drive the magnetic skullet to move in a direction far away from the current injection layer under the action of a driving current; wherein the current density of the generated current is greater than the current density of the driving current; the annular detection device is arranged to detect magnetic skamming that moves into its detection range. The size and the cost of the existing artificial neural network can be effectively reduced.

Description

Electronic neuron and artificial neural network based on siganmin
Technical Field
The invention relates to the technical field of artificial neural networks, in particular to a siganmin-based electronic neuron and an artificial neural network.
Background
Similar to biological neural networks, existing artificial neural networks also contain a large number of neurons and synapses. These neurons and synapses are currently composed primarily of Complementary Metal Oxide Semiconductors (CMOS). Thus, existing artificial neural networks will face two major challenges. First, the structure and size of the network will be limited by the size of the neurons and synapses; second, such large hardware facilities must also consume large amounts of energy.
Disclosure of Invention
In view of the above, it is necessary to provide a siganus based electronic neuron and an artificial neural network.
A Schlegunger-based electronic neuron comprises a Schlegunger generation and driving device and an annular detection device; wherein, the Sjgren seed generating and driving device comprises a magnetic nano-film and a current injection layer;
the current injection layer and the annular detection device are both arranged on the magnetic nano film, and the current injection layer is arranged in the center of the annular detection device;
the current injection layer is arranged to generate a magnetic skyrmion through the magnetic nano-film under the action of a generated current, and the magnetic skyrmion is driven to move in a direction far away from the current injection layer under the action of a driving current; wherein the generation current and the driving current are both injected into the current injection layer in a direction perpendicular to the magnetic nano thin film;
the annular detection device is arranged to detect magnetic skamming that moves into its detection range.
In one embodiment, the magnetic nano-film is made of a magnetic nano-material having DM interaction and capable of stabilizing the magnetic skynerger.
In one embodiment, the ferromagnetic nano-thin film is in the shape of a disk, and the diameter of the disk-shaped ferromagnetic nano-thin film ranges from 80nm to 120 nm.
In one embodiment, the current injection layer includes a pinning layer and a spacer layer;
the spacing layer is arranged on the magnetic nano film;
the pinning layer is disposed on the spacer layer.
In one embodiment, the current density of the generated current is greater than the current density of the driving current.
In one embodiment, the surface of the current injection layer contacted with the magnetic nano thin film forms a nano point contact, and the radius r of the nano point contact1And the radius r of the magnetic skynolve2Satisfies the following formula:
r1＝r2±C
wherein the value range of C is between 2nm and 3 nm.
In one embodiment, the ring-shaped detection device is a circular ring-shaped tunnel junction, and the inner diameter of the circular ring-shaped tunnel junction is smaller than or equal to the maximum movement radius of the magnetic sigramins.
In one embodiment, the method further comprises the following steps:
a current injection device connected to the current injection layer for injecting the generation current or the driving current into the current injection layer.
In one embodiment, the method further comprises the following steps:
and the peripheral circuit is electrically connected with the annular detection device and is used for outputting an electric signal when the annular detection device detects the magnetic sigramins.
Based on the same inventive concept, the application also provides an artificial neural network, which comprises a plurality of mutually connected neurons, wherein the neurons are any one of the electronic neurons based on the magnetic siganmin.
The electronic neuron and the artificial neural network based on the magnetic siganus are provided with a siganus generating and driving device and an annular detection device, wherein the siganus generating and driving device is composed of a magnetic nano film and a current injection layer, the current injection layer and the annular detection device are arranged on the magnetic nano film, and the current injection layer is arranged at the center of the annular detection device. And then applying currents (generating current and driving current) with different magnitudes to the current injection layer respectively to promote the magnetic nano film to generate and drive the magnetic siganmin, and finally detecting the magnetic siganmin moving to the detection range of the magnetic nano film by using an annular detection device, so that the process similar to biological neurons can be realized by adjusting the current densities of different currents (generating current and driving current). Because the size of the device (including the skyscraper) for generating and driving the magnetic skyscraper is very small and is in a nanometer level, the size of the electronic neuron can be reduced, and the whole size of the artificial neural network can be further reduced; in addition, the spin polarization current required by the generation and driving of the magnetic skammomum is very small, so that the power consumption of the electronic neuron is reduced, and the overall power consumption of the artificial neural network is further reduced.
Drawings
FIG. 1 is a schematic diagram of a magnetic skyrmion-based electronic neuron in one embodiment;
FIG. 2 is a schematic diagram of a structure of a biological neuron in an exemplary technique;
FIG. 3 is a schematic diagram of the working principle of a biological neuron in an exemplary technique;
FIG. 4 is a schematic diagram of the motion trajectory of a magnetic skamming molecule under the action of a spin-polarized current;
FIG. 5 is a schematic diagram of the movement behavior of magnetic skamming under the action of a periodic, uniform spin-polarized current;
FIG. 6 is a schematic diagram of the behavior of magnetic skamming under periodic non-uniform spin-polarized current;
FIG. 7 is a schematic diagram of the motion behavior of a magnetic skammomum under sinusoidal spin-polarized current.
Wherein:
and Dd: a dendrite Ax: axons
AxT: axon terminals J1: generating an electric current
J2: drive current SK: magnetic Scurrio-Glabra (Sw-Gemini)
Detailed Description
To facilitate an understanding of the present application, the present application will now be described more fully with reference to the accompanying drawings. Preferred embodiments of the present application are given in the accompanying drawings. This application may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements may also be present. The terms "vertical," "horizontal," "left," "right," and the like as used herein are for illustrative purposes only and do not represent the only embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the description of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.
Skyrmions (Skyrmions) is a particle-like model proposed by the nuclear physicist, Tony Skyrme, in 1961. In 2009, german scientists found for the first time in experiments a stable magnetic skullcup, which is a chiral spin structure with a vortex structure (chiral spin configuration). The magnetic skynergenis can stably exist in bulk magnets with extremely strong spin-orbit coupling effect or in nano-films coupled with heavy metals. The central magnetic moment of the magnetic skynerger is opposite to the direction of the surrounding magnetic moment, and the central magnetic moment can be turned to be consistent with the direction of the surrounding magnetic moment by overcoming a certain potential barrier. Thus, the magnetic segmentons are more stable than conventional domain wall structures.
Furthermore, the application of the magnetic sigmin in the neuron has great advantages compared with the traditional CMOS structure, which mainly shows the following three aspects: 1) compared with the traditional magnetic domain wall, the size of the magnetic skullet can be very small, and currently, the size of a single magnetic skullet can be 5 nm; 2) the magnetic segmentum is a special magnetic domain structure with topological protection, is more stable compared with the traditional magnetic domain wall, and can be kept stable when external conditions (magnetic field, temperature, device defects and the like) change, so that the stability of the neuron based on the magnetic segmentum is higher; 3) the firing current density of magnetic segmentons is much less than that of the domain walls, being parts per million of the domain walls, so that magnetic segmenton-based neurons have the property of low power consumption.
Please refer to fig. 2, which is a schematic structural diagram of a biological neuron according to an exemplary technique. The neuron is mainly composed of two parts, a cell body (not shown) and a protrusion (not shown). Wherein the processes comprise dendrites Dd and axons Ax. The dendrite Dd is dendritic and is short and numerous, and is mainly used for receiving signals from other neurons; the axon Ax is a thin, small branch, long process that serves to relay information to the next neuron.
In general, two neurons are connected by synapses, thereby achieving processing and transmission of information. Specifically, a stimulation signal from a pre-neuron is typically converted to a neurotransmitter that acts on the dendrites or cell bodies of a post-neuron, opening an ion channel, thereby causing a potential difference, i.e., a membrane potential, to form outside the membrane of the neuron. Without continuous stimulation, the membrane potential gradually decreases, trying to return to the resting potential, a process also referred to as leakage. As shown in fig. 3, as multiple signals are stimulated, the membrane potential accumulates, and when it reaches a certain potential threshold, the nerve cell generates a new electrical signal that propagates along axon Ax to axon terminal AxT, where it is converted to a neurotransmitter that acts on the next neuron. From the above analysis, the working process of biological neurons mainly includes three main processes: leakage, accumulation, and emission.
Based on the same working principle as that of biological neurons, the application designs an electronic neuron and an artificial neural network based on magnetic siganmin. In order to make the aforementioned and other features and advantages of the invention more comprehensible, embodiments accompanied with figures are described in detail below.
Fig. 1 is a schematic structural diagram of an electronic neuron based on a magnetic sigramin according to the present application. Where J1 denotes a generation current, J2 denotes a drive current, and SK denotes a magnetic skullet. The electronic neuron of the present application may include a squareness sub-generation and driving device (not shown) and a loop detection device 20. The sgemin generating and driving device includes a magnetic nano-film 110 and a current injection layer 120.
The magnetic nano-film 110 may be made of a magnetic nano-material having DM interaction and capable of stabilizing magnetic sigecures. Specifically, the magnetic nanomaterial has perpendicular anisotropy, including but not limited to ferromagnetic materials, antiferromagnetic materials, synthetic antiferromagnetic materials, and ferrimagnetic materials. Its function is to generate and stabilize the magnetic skullet SK. In this embodiment, the magnetic nano-film 110 may be formed by sequentially depositing platinum and cobalt materials on a substrate by a deposition process, and then etching the platinum and cobalt materials to obtain a cobalt-platinum film with a predetermined shape. The shape of the magnetic nano-film 110 formed after etching includes, but is not limited to, rectangle, square and disc. The cobalt platinum material (Co/Pt) has a large spin-orbit coupling moment, so that DM (dzyaloshinski-Moriya) interaction required for stabilizing the magnetic skullons SK can be generated, and due to the DM interaction, after a current J1 is generated by local injection of the magnetic nano-film 110, a stable magnetic skullons SK can be generated at a corresponding position of the magnetic nano-film 110. For convenience of description and illustration, the following examples of the present application are described by taking the magnetic nano thin film 110 as a disk-shaped thin film made of cobalt platinum (Co/Pt), which is referred to as a ferromagnetic nano disk for short.
In this embodiment, when the magnetic nano-film 110 is a ferromagnetic nano-disc, the ferromagnetic parameters used by the ferromagnetic nano-disc may be: saturation magnetization MS580kA/m, 15pJ/m for exchange constant A, 0.8MJ/m for magnetocrystalline anisotropy constant K3DMI constant D is 3mJ/m2. The diameter of the ferromagnetic nanodisk 110 can be selected in the range of 80nm to 120nmAnd nm. It will be appreciated that the ferromagnetic parameters involved may be adjusted accordingly for the magnetic material selected.
Further, with continued reference to fig. 1, the current injection layer 120 and the ring probe 20 are disposed on the magnetic nano-film 110, and more advantageously, the current injection layer 120 is disposed at the center of the ring probe 20. When the magnetic nano-film 110 is a ferromagnetic nano-disc, the current injection layer 120 and the annular detecting device 20 are concentric with a geometric center.
In this embodiment, the current injection layer 120 and the ferromagnetic nanodisk 110 can be integrally constructed as a tunnel junction. Since the ferromagnetic nanodisk 110 is made of cobalt-platinum (Co/Pt), when the current injection layer 120 is selected, only two layers of materials (spacer layer 122/pinning layer 124) are deposited on the cobalt-platinum (Co/Pt) material by a deposition process, and then patterned by an etching process to form the shape of a disk. It is understood that the materials, compositions, etc. of the spacer layer 122 and the pinning layer 124 can be referred to the description in the prior art, which is not the focus of this patent, and therefore, will not be further described herein.
In this embodiment, the surface of the current injection layer 120 in contact with the ferromagnetic nanodisk 110 can be electrochemically formed into a conductive channel, i.e., a nanodot contact. The cross-section of the nano-point contact may be selected to be circular, and preferably the center of the nano-point contact coincides with the center of the ferromagnetic nanodisk 110. Alternatively, in order to ensure that the magnetic skutter SK can be stably generated and driven, the radius of the nano-point contact should be set as close as possible to the radius of the magnetic skutter SK. In this embodiment, the radius r of the nano-point contact1Radius r of magnetic skamming SK2Satisfies the following formula:
r1＝r2±C
wherein the value range of C is between 2nm and 3 nm. Illustratively, when C is taken to be 2nm, r2Taking the radius r of the corresponding nano-point contact at 8nm1It is 10 nm. However, the present application is not limited thereto.
After the magnetic skutter SK is stabilized, a driving current J2 is injected through the nanodot contact, and the magnetic skutter SK gradually moves away from the center of the ferromagnetic nanodisk 110 and starts to move in a circular motion. If the distance d between the magnetic skutter SK and the center of the ferromagnetic nanodisk 110 is regarded as the membrane potential of the neuron, the design can well simulate the basic functions of the biological neuron, namely three main processes: leakage, accumulation and emission.
Specifically, if the driving current J2 is suddenly removed during the process that the magnetic skullet SK is far away from the center of the ferromagnetic nanodisk 110, at this time, due to the boundary action, the magnetic skullet SK will make a spiral motion to gradually return to the center of the ferromagnetic nanodisk 110 (because the total energy of the whole electronic neuron is positively correlated with the distance between the magnetic skullet SK and the center of the ferromagnetic nanodisk 110, the total energy of the whole electronic neuron is the lowest when the magnetic skullet SK is at the center of the ferromagnetic nanodisk 110.
If the driving current J2 is continuously injected, the magnetic sigecum SK gradually moves away from the center of the ferromagnetic nanodisk 110 (see fig. 4), and the magnetic sigecum SK does not start to move circularly until the maximum moving radius (i.e., the stable moving radius) is reached. Which corresponds to the process of gradual accumulation of membrane potential in biological neurons. It is worth mentioning that although the specific position of the magnetic skutter SK is uncertain during the circular motion or the spiral motion, the distance from the center of the magnetic skutter SK (i.e. the center of the ferromagnetic nanodisk 110) is certain. In other words, for the same distance d, the magnetic skutter SK may be at any position on the circumference with radius d. It should be noted that the radius of the circular motion of the magnetic skynerger SK of the present application is mainly determined by the size of the ferromagnetic nanodisk 110, the current density of the driving current J2 and the radius of the nanodot contact. In practical applications, the radius of stable motion of the magnetic skutter SK can be enlarged by suitably increasing the radius of the ferromagnetic nanodisk 110 or the current density of the drive current J2 or the radius of nanodot contact.
At this time, if a distance smaller than the maximum radius is set as a critical distance dc (corresponding to a membrane potential threshold in a biological neuron) and the ring probe 20 is placed at this position, the magnetic strobilurin SK is detected when it moves to the critical distance dc (i.e., enters the detection range of the ring probe 20).
Further, the ring probe 20 may be a circular ring tunnel junction, the inner diameter of which may be set to be less than or equal to the maximum radius of motion of the magnetic sigramins, according to the foregoing description. I.e. the inner diameter of the circular ring tunnel junction is smaller than or equal to the critical distance dc. The reason why the annular detection device 20 can detect the magnetic sigmin SK is that: the magnetic moments of the selected ferromagnetic nanodisks 110 are originally arranged along the same direction, and the magnetic skminster SK generated on the ferromagnetic nanodisks 110 is of a vortex structure, so that the magnetic moments of the positions where the magnetic moments are located can be deflected, and the magnetic resistance of the positions can be changed due to the deflection of the magnetic moments.
When the ring detector 20 detects the magnetic skyrmion SK, an electrical signal can be emitted through a peripheral circuit (not shown) provided in the present application and electrically connected to the ring detector 20. Which corresponds to the process of firing after the accumulation of membrane potential in a biological neuron reaches a threshold.
In this particular embodiment, the generation current J1 and the driving current J2 are both injected into the center of the ferromagnetic nanodisk 110 in a direction perpendicular to the nanoscopic point contact, and the current density of the generation current J1 is greater than the current density of the driving current J2. illustratively, the current density of the generation current J1 may range from 1 × 1012A/m2To 1 × 1015A/m2Meanwhile, the current density of the driving current J2 may be 1 × 1011A/m2。
In one embodiment, the generating current J1 and the driving current J2 can be ordinary currents or spin-polarized currents; in this embodiment, the generating current J1 and the driving current J2 can be normal currents. When the generation current J1 and the driving current J2 are both selected as normal currents, they pass through the tunnel junction formed by the current injection layer 120 and the ferromagnetic nanodisk 110, and become spin-polarized currents, and after the spin-polarized currents act on the ferromagnetic nanodisk 110, a stable magnetic skuller SK is generated at the corresponding position of the ferromagnetic nanodisk 110. Further, when a normal current is applied to the tunnel junction formed by the current injection layer 120 and the ferromagnetic nanodisk 110 and then acts as a spin-polarized current for driving the magnetic skullet SK to move, the spin-polarized current for driving the magnetic skullet SK to move may be any one of a periodic uniform spin-polarized current, a periodic non-uniform spin-polarized current, and a sinusoidal spin-polarized current, or may be any combination of the three. That is, when the spin-polarized current for driving the magnetic skutter SK is any one of or any combination of three of a periodic uniform spin-polarized current, a periodic non-uniform spin-polarized current, and a sinusoidal spin-polarized current, the magnetic skutter SK can be driven.
As an aid, refer to fig. 5, which is a schematic diagram of the motion behavior of the magnetic sigramins under the action of the periodic uniform spin-polarized current. In the figure, d represents the distance of the magnetic skminger SK from the center of the ferromagnetic nanodisk 110. J2 denotes drive current in units of A/m2. Wherein the action time of each current pulse can be 1ns, and the time interval between two current pulses can be 3 ns. As can be seen from fig. 5, when the driving current J2 passes through the tunnel junction formed by the current injection layer 120 and the ferromagnetic nanodisk 110, the formed periodic uniform spin-polarized current can drive the magnetic skullet SK and also can simulate the function of a biological neuron.
Reference is now made to FIG. 6, which is a schematic diagram illustrating the behavior of the magnetic skamming molecule under the action of a periodic non-uniform spin-polarized current. In the figure, d represents the distance of the magnetic skminger SK from the center of the ferromagnetic nanodisk 110. J2 denotes drive current in units of A/m2. Wherein the action time of each current pulse can be 1ns, and the two currents areThe time interval between stream pulses may be 2 ns. As can be seen from fig. 6, when the driving current J2 passes through the tunnel junction formed by the current injection layer 120 and the ferromagnetic nanodisk 110, the periodic non-uniform spin-polarized current can also drive the magnetic skullet SK and can also mimic the function of the biological neuron.
Reference is now made to fig. 7, which is a schematic diagram of the motion behavior of the magnetic skamming molecule under the action of a sinusoidal spin-polarized current. In the figure, d represents the distance of the magnetic skminger SK from the center of the ferromagnetic nanodisk 110. J2 denotes drive current in units of A/m2. Wherein, the action time of each current pulse can be 2ns, and the time interval between two current pulses can be 2 ns. As can be seen from fig. 7, when the driving current J2 passes through the tunnel junction formed by the current injection layer 120 and the ferromagnetic nanodisk 110, the resulting sinusoidal spin-polarized current can also drive the magnetic skullet SK and can also mimic the function of a biological neuron.
In one embodiment, the generation current J1 and the driving current J2 can be injected by a current injection device (not shown) electrically connected to the current injection layer 120. In this embodiment, the current injection device may be an electrode for injecting the generation current J1 or the driving current J2. In other embodiments, the current injection device may be any other device capable of generating spin-polarized current, for example, a metal with a strong spin-polarized energy band structure such as Pd can be used to generate spin-polarized current after being loaded with a normal current.
In order to make the principle of the present application clearer, the working principle of the artificial neuron of the present application is described in detail by taking the magnetic nano-film 110 as a ferromagnetic nano-disc, taking the surface of the current injection layer 120 contacting the magnetic nano-film 110 as a nano-point contact, and taking the annular detection device 20 as an example of a circular ring detector.
First, through the current injection device, the current J1 is generated by injecting in the direction perpendicular to the direction of the nano-point contact at the nano-point contact to break the original topological potential barrier of the ferromagnetic state, and after the potential barrier is stabilized, a complete magnetic skgmuim SK can be obtained at the center of the ferromagnetic nanodisk 110. At this time, the driving current J2 is injected in the same direction by the current injection device, and the magnetic sigecum SK gradually moves away from the center of the ferromagnetic nanodisk 110 to start circular motion. With the continuous injection of the driving current J2, the distance between the magnetic skmins SK and the center of the ferromagnetic nanodisk 110 will reach a critical value, and at this time, the annular detector 20 will sense the magnetic skmins SK according to the magnetic resistance variation in the detection region, so as to emit an electrical signal under the control of the peripheral circuit.
In summary, the electronic neuron based on the magnetic siganus provided by the application can realize the basic functions of the original biological neuron, and simultaneously, the size and the cost of the electronic neuron can be reduced compared with the size and the cost of the existing neuron based on the CMOS process.
Based on the same inventive concept, the application also provides an artificial neural network, which comprises a plurality of interconnected neurons, wherein the neurons can be any one of the electronic neurons based on the magnetic sigramins. It can be understood that, since the artificial neural network in this embodiment includes any one of the foregoing electronic neurons based on magnetic segmentum, for the same beneficial effects as those of the electronic neurons based on magnetic segmentum, reference may be made to the foregoing embodiments, and further description is not repeated herein.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (10)

1. An electronic neuron based on magnetic siganus, which is characterized by comprising a siganus generating and driving device and an annular detecting device; wherein, the Sjgren seed generating and driving device comprises a magnetic nano-film and a current injection layer;
the current injection layer and the annular detection device are both arranged on the magnetic nano film, and the current injection layer is arranged in the center of the annular detection device;
the current injection layer is arranged to generate a magnetic skyrmion through the magnetic nano-film under the action of a generated current, and the magnetic skyrmion is driven to move in a direction far away from the current injection layer under the action of a driving current; wherein the generation current and the driving current are both injected into the current injection layer in a direction perpendicular to the magnetic nano thin film;
the annular detection device is arranged to detect magnetic skamming that moves into its detection range.
2. The magnetic sigramin-based electronic neuron of claim 1, wherein the magnetic nano-film is made of a magnetic nanomaterial that has DM interaction and is capable of stabilizing the magnetic sigramin.
3. The magnetic skyrmion-based electronic neuron of claim 2, wherein the ferromagnetic nano-thin film is in the shape of a disk, and the diameter of the disk-shaped ferromagnetic nano-thin film ranges from 80nm to 120 nm.
4. The magnetic skyrmion-based electronic neuron of claim 1, wherein the current injection layer comprises a pinned layer and a spacer layer;
the spacing layer is arranged on the magnetic nano film;
the pinning layer is disposed on the spacer layer.
5. The magnetic skyrmion-based electronic neuron of claim 1, wherein a current density of the generated current is greater than a current density of the drive current.
6. The magnetic skyrmion-based electronic neuron of claim 1, wherein the surface of the current injection layer in contact with the magnetic nanofilm forms a nanolinked contact having a radius r1And the radius r of the magnetic skynolve2Satisfies the following formula:
r1＝r2±C
wherein the value range of C is between 2nm and 3 nm.
7. The magnetic sigramin-based electronic neuron of claim 1, characterized in that the ring-shaped detection device is a circular ring-shaped tunnel junction, the inner diameter of which is smaller than or equal to the maximum radius of motion of the magnetic sigramin.
8. The magnetic skyrmion-based electronic neuron of any one of claims 1-7, further comprising:
a current injection device connected to the current injection layer for injecting the generation current or the driving current into the current injection layer.
9. The magnetic skyrmion-based electronic neuron of any one of claims 1-7, further comprising:
and the peripheral circuit is electrically connected with the annular detection device and is used for outputting an electric signal when the annular detection device detects the magnetic sigramins.
10. An artificial neural network comprising a plurality of interconnected neurons, wherein said neurons are electronic neurons based on meglumine according to any one of claims 1 to 9.
CN202010165391.8A 2020-03-11 2020-03-11 Electronic neuron and artificial neural network based on siganmin Active CN111415001B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010165391.8A CN111415001B (en) 2020-03-11 2020-03-11 Electronic neuron and artificial neural network based on siganmin

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010165391.8A CN111415001B (en) 2020-03-11 2020-03-11 Electronic neuron and artificial neural network based on siganmin

Publications (2)

Publication Number Publication Date
CN111415001A true CN111415001A (en) 2020-07-14
CN111415001B CN111415001B (en) 2023-03-21

Family

ID=71491009

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010165391.8A Active CN111415001B (en) 2020-03-11 2020-03-11 Electronic neuron and artificial neural network based on siganmin

Country Status (1)

Country Link
CN (1) CN111415001B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112768605A (en) * 2021-01-07 2021-05-07 香港中文大学（深圳） Periodic signal detection device and periodic signal detection method
CN113161476A (en) * 2021-03-08 2021-07-23 湖北大学 Storage device of neuron synapse based on siganus
CN115271058A (en) * 2022-09-29 2022-11-01 北京芯可鉴科技有限公司 Brain-like computing chip and data processing terminal
WO2023178783A1 (en) * 2022-03-24 2023-09-28 中国科学院微电子研究所 Skyrmion transistor and skyrmion transistor control method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104347226A (en) * 2013-07-23 2015-02-11 中国科学院物理研究所 Magnetic multilayer film based on magnetic skyrmion layer
CN107332554A (en) * 2017-06-09 2017-11-07 香港中文大学（深圳） OR gate and its control and methods for using them based on magnetic Skyrmion
CN109065703A (en) * 2018-07-31 2018-12-21 电子科技大学 A method of driving magnetic Skyrmion movement

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104347226A (en) * 2013-07-23 2015-02-11 中国科学院物理研究所 Magnetic multilayer film based on magnetic skyrmion layer
CN107332554A (en) * 2017-06-09 2017-11-07 香港中文大学（深圳） OR gate and its control and methods for using them based on magnetic Skyrmion
CN109065703A (en) * 2018-07-31 2018-12-21 电子科技大学 A method of driving magnetic Skyrmion movement

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112768605A (en) * 2021-01-07 2021-05-07 香港中文大学（深圳） Periodic signal detection device and periodic signal detection method
CN113161476A (en) * 2021-03-08 2021-07-23 湖北大学 Storage device of neuron synapse based on siganus
WO2023178783A1 (en) * 2022-03-24 2023-09-28 中国科学院微电子研究所 Skyrmion transistor and skyrmion transistor control method
CN115271058A (en) * 2022-09-29 2022-11-01 北京芯可鉴科技有限公司 Brain-like computing chip and data processing terminal
CN115271058B (en) * 2022-09-29 2023-01-17 北京芯可鉴科技有限公司 Brain-like computing chip and data processing terminal

Also Published As

Publication number Publication date
CN111415001B (en) 2023-03-21

Similar Documents

Publication Publication Date Title
CN111415001B (en) Electronic neuron and artificial neural network based on siganmin
Bian et al. Stimuli‐responsive memristive materials for artificial synapses and neuromorphic computing
US10497866B1 (en) Ionic floating-gate memory device
US9620562B2 (en) Voltage-controlled magnetic anisotropy switching device using an external ferromagnetic biasing film
US9087593B2 (en) Random bit generator that applies alternating current (AC) to magnetic tunnel junction to generate a random bit
Zheng et al. Programmable Negative Differential Resistance Effects Based on Self‐Assembled Au@ PPy Core–Shell Nanoparticle Arrays
Kikuchi et al. Vertical bistable switching of spin vortex in a circular magnetic dot
Fink et al. High energy ion beam irradiation of polymers for electronic applications
KR101559216B1 (en) Bipolar spin-transfer switching
KR20040030495A (en) Conductive polymer materials and methods for their manufacture and use
Portal et al. An overview of non-volatile flip-flops based on emerging memory technologies
Hota et al. Electrochemical Thin‐Film Transistors using Covalent Organic Framework Channel
Li et al. Emerging neuromorphic computing paradigms exploring magnetic skyrmions
Zhao et al. Magnetoresistance behavior of conducting filaments in resistive-switching NiO with different resistance states
US20110069541A1 (en) Ferromagnetic thin wire element
EP3026721A1 (en) Magnetic device with spin polarisation
Wang et al. Neurotransmitter‐Mediated Plasticity in 2D Perovskite Memristor for Reinforcement Learning
US8331129B2 (en) Memory array with write feedback
Amer et al. Design techniques for in-field memristor forming circuits
de h-Óra et al. Highly cyclable voltage control of magnetism in cobalt ferrite nanopillars for memory and neuromorphic applications
Peng et al. Voltage control of perpendicular magnetic anisotropy in multiferroic composite thin films under strong electric fields
TWI774765B (en) Bit-line sensing for correlated electron switch elements
Fink et al. Swift heavy ion irradiation as a tool for creating novel nanoelectronic structures
JP2018081735A (en) Manufacturing method of compound element
Bilgaye et al. Memristor Materials: Working Conditions And Properties

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant