CN113657586A - Spin orbit torque-based nerve component - Google Patents

Spin orbit torque-based nerve component Download PDF

Info

Publication number
CN113657586A
CN113657586A CN202110840235.1A CN202110840235A CN113657586A CN 113657586 A CN113657586 A CN 113657586A CN 202110840235 A CN202110840235 A CN 202110840235A CN 113657586 A CN113657586 A CN 113657586A
Authority
CN
China
Prior art keywords
layer
cofeb
magnetic anisotropy
orbit torque
spin orbit
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.)
Pending
Application number
CN202110840235.1A
Other languages
Chinese (zh)
Inventor
邢国忠
王迪
林淮
刘龙
刘明
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.)
Institute of Microelectronics of CAS
Original Assignee
Institute of Microelectronics of CAS
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 Institute of Microelectronics of CAS filed Critical Institute of Microelectronics of CAS
Priority to CN202110840235.1A priority Critical patent/CN113657586A/en
Publication of CN113657586A publication Critical patent/CN113657586A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/06Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • HELECTRICITY
    • H10SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10NELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N50/00Galvanomagnetic devices
    • H10N50/80Constructional details
    • H10N50/85Magnetic active materials
    • HELECTRICITY
    • H10SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10NELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N52/00Hall-effect devices

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Neurology (AREA)
  • Hall/Mr Elements (AREA)

Abstract

The present disclosure provides a spin orbit torque based neural component, including: an antiferromagnetic pinning layer, a first ferromagnetic layer and a spin orbit coupling layer sequentially formed on the substrate; a free layer formed on the spin orbit coupling layer and moving the magnetic domain wall according to the spin orbit torque; a tunneling layer formed on the free layer; left and right pinned layers formed at both sides of the free layer and having opposite magnetization directions; a reference layer formed on the tunneling layer; the free layer, the tunneling layer and the reference layer form a magnetic tunnel junction, and the magnetic tunnel junction is used for reading neuron signals. The disclosure also provides a preparation method of the spin orbit torque-based neural component.

Description

Spin orbit torque-based nerve component
Technical Field
The disclosure relates to the technical field of brain-like computation, in particular to a spin orbit torque-based nerve component.
Background
Brain inspired neuromorphic calculations are one of the potential trends in the "post-morgan" era to address memory problems in von neumann architectures and to address the termination of moore's law. The neuromorphic calculation is expected to further improve the calculation power of the chip and obviously reduce the power consumption of the chip.
In the process of developing neuromorphic calculations, it is first necessary to study neuromorphic devices (synapses, neurons, etc.) with certain biological properties. In recent years, researchers find that novel memories (magnetic memories, resistive random access memories, phase change memories, ferroelectric memories and the like) and some new principle devices (ion transistors and the like) and the like can better and more abundantly simulate some characteristics of human brain synapses and neurons than traditional CMOS neuron circuits with high power consumption and large area overhead. Among them, the high speed, high durability and low power consumption of the magnetic memory make it very competitive in the field of neuromorphic computing.
However, the current research on neuromorphic devices based on spintronic devices is more limited to the research on synapse devices and relatively less on nerve components. And only some of the stored nerve components simulate the accumulation characteristic of neurons based on the mode of driving the magnetic domain wall by the spin-orbit torque, but the speed of driving the magnetic domain wall by the spin-orbit torque is slower than that of driving the spin-orbit torque. And the leakage characteristic of the neuron is realized by means of shape anisotropy, bias field and anisotropic gradient.
Therefore, in order to be suitable for a high-speed and complex neural network, a neuron component for driving a magnetic domain wall by spin-orbit torque needs to be researched, and the leakage characteristic of a neuron is simulated through more mechanisms.
Disclosure of Invention
In order to solve the above problems in the prior art, the present disclosure provides a spin orbit torque-based neural device, which aims to solve the problem of achieving accumulation, leakage, and discharge characteristics of a simulated biological neuron based on a magnetic tunnel junction, and achieve pinning of a domain wall by depositing antiferromagnets with opposite local magnetization directions or a thicker local free layer on both sides of the free layer.
A first aspect of the present disclosure provides a spin orbit torque-based neural component, including: an antiferromagnetic pinning layer, a first ferromagnetic layer and a spin orbit coupling layer sequentially formed on the substrate; the spin orbit coupling layer is made of one or more materials of Ta, W and Mo; a free layer formed on the spin orbit coupling layer and moving the magnetic domain wall according to the spin orbit torque; a tunneling layer formed on the free layer; left and right pinned layers formed at both sides of the free layer and having opposite magnetization directions; a reference layer formed on the tunneling layer; the free layer, the tunneling layer and the reference layer form a magnetic tunnel junction, and the magnetic tunnel junction is used for reading neuron signals.
Further, the spin orbit coupling layer is composed of one or more materials of Ta, W and Mo.
Further, the free layer has perpendicular magnetic anisotropy, and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the first ferromagnetic layer has an in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe and Co; the antiferromagnetic pinning layer has a vertical exchange effect and is composed of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
Further, the free layer has perpendicular magnetic anisotropy, and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the first ferromagnetic layer has a tilted magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, Co, CoFeAl and CoFe; the antiferromagnetic pinning layer has a tilt exchange effect and is composed of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
Further, the free layer has a tilted magnetic anisotropy, which is composed of one or more materials of CoFeB, NiFe, Co, CoFeAl, and CoFe; the first ferromagnetic layer has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; antiferromagnetic pinningThe layer has a vertical exchange function and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
Further, the device further comprises: a second ferromagnetic layer formed between the antiferromagnetic pinning layer and the first ferromagnetic layer.
Further, the free layer has perpendicular magnetic anisotropy, and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the first ferromagnetic layer has an in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe and Co; the antiferromagnetic pinning layer has a vertical exchange effect and is composed of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO; the second ferromagnetic layer has perpendicular magnetic anisotropy and is composed of one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe.
Further, the device further comprises: a second ferromagnetic layer and an insulating layer formed in sequence on the substrate, wherein an antiferromagnetic pinning layer is located on the insulating layer.
Further, the free layer has perpendicular magnetic anisotropy, and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the first ferromagnetic layer has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the antiferromagnetic pinning layer has a vertical exchange effect and is composed of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO; the insulating layer is made of SiO2Forming; the second ferromagnetic layer has an in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, and Co.
Further, the device further comprises: a second ferromagnetic layer and a spacer layer formed in that order on the antiferromagnetic pinning layer, wherein the first ferromagnetic layer is on the spacer layer.
Further, the free layer has perpendicular magnetic anisotropy, and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the first ferromagnetic layer has an in-plane magnetic anisotropy and is composed of one or more materials selected from CoFeB, NiFe and Co(ii) a The spacing layer is made of one or more of Ru, Ta, W, V, Cr, Rh, Nd, Mo and Re; the second ferromagnetic layer has perpendicular magnetic anisotropy, and is composed of: one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe; the antiferromagnetic pinning layer has a vertical exchange effect and is composed of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
Furthermore, the spin orbit coupling layer is used for realizing the accumulation and leakage characteristics of the neurons, current is introduced into the spin orbit coupling layer, spin current in the vertical direction is generated based on the spin Hall effect, the motion of a domain wall is realized under the action of spin orbit torque, and the accumulation characteristics of the biological neurons are simulated; under the condition of no current, the free layer and the first ferromagnetic layer are coupled through the spin orbit coupling layer through the RKKY effect ferromagnetism or antiferromagnetism, so that a domain wall in the free layer has a movement trend opposite to the current driving direction, and the leakage characteristic of the biological neuron is realized; when the moving position of the domain wall in the free layer exceeds a threshold value area, the magnetic tunnel junction is switched from an anti-parallel state to a parallel state, and an external circuit is adopted to output a spike pulse to simulate the discharge characteristic of a biological neuron.
Furthermore, the thickness of the free layer is 0.8-2 nm.
Further, the RKKY effect is related to the layer thickness of the spin-orbit coupling layer, the material and the magnitude of the modulated injection current density.
Further, by modulating the DMI antisymmetric effect and the damping coefficient with proper size and different symbols, the speed of the neuron accumulation process is adjusted.
Further, the device further comprises: the spin-orbit coupling device comprises a left electrode, a right electrode and a top electrode, wherein the left electrode and the right electrode are respectively oppositely arranged on two sides of a spin-orbit coupling layer, which are not provided with a free layer, and the top electrode is positioned on a reference layer.
Furthermore, the layer thickness of the reference layer is larger than that of the free layer, and the layer thickness is preferably 0.8-2 nm.
Furthermore, the thickness of the tunneling layer is preferably 0.5-4 nm.
Furthermore, the thickness of the left electrode, the right electrode and the top electrode is 50-200 nm.
Further, the inclined magnetization direction or the inclined magnetic anisotropy of the first ferromagnetic layer or the free layer is used for generating an equivalent field in the x direction on the free layer so as to realize the unchangeable chirality of a magnetic domain wall and realize the high-speed motion of the domain wall, and the uncertainty of the motion direction of the next accumulation process caused by the chirality change of a precession induced domain wall in the leakage process is avoided.
A second aspect of the present disclosure provides a method for manufacturing a spin orbit torque-based neural device, including: sequentially growing an antiferromagnetic pinning layer, a first ferromagnetic layer, a spin orbit coupling layer and a free layer on a substrate; forming a left electrode and a right electrode on two sides of the spin orbit coupling layer respectively; and sequentially forming a tunneling layer, a reference layer and a top electrode on the free layer, wherein the free layer, the tunneling layer and the reference layer form a magnetic tunnel junction, and the magnetic tunnel junction is used for reading neuron signals.
Compared with the prior art, the method has the following beneficial effects:
(1) the nerve component based on the magnetic tunnel junction can accumulate current pulses from synapses under the condition of a full electric field, has high energy efficiency and a high reliability self-leakage function, and when the accumulated current pulses drive a magnetic domain wall to move and exceed a threshold region, neurons can activate and emit a spike signal to simulate the function of human brain neurons.
(2) The neural component realizes the spin orbit torque driving magnetic domain wall.
(3) And the chirality of the magnetic domain wall is ensured to be unchanged by modulating the inclined magnetization direction or the inclined magnetic anisotropy of the first ferromagnetic layer or the free layer, so that the high-speed nerve component is realized.
Drawings
For a more complete understanding of the present disclosure and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:
fig. 1 schematically illustrates a front view of spin orbit torque based neural components according to first to third embodiments of the present disclosure;
FIG. 2 schematically illustrates a front view of a spin orbit torque based neural component, according to a fourth embodiment of the present disclosure;
fig. 3 schematically illustrates a front view of a spin orbit torque based neural component, according to a fifth embodiment of the present disclosure;
FIG. 4 schematically illustrates a front view of a spin orbit torque based neural component, according to a sixth embodiment of the present disclosure;
FIG. 5 is a schematic diagram illustrating leakage-accumulation-discharge characteristics of spin-orbit torque-based neural components, according to an embodiment of the present disclosure;
FIG. 6 schematically illustrates a schematic diagram of the accumulation characteristics of spin-orbit torque based neural components and DM antisymmetric exchange effects, according to an embodiment of the present disclosure;
FIG. 7 is a schematic diagram illustrating an accumulation characteristic of spin-orbit torque based neural components versus damping coefficient according to an embodiment of the present disclosure;
fig. 8 schematically illustrates a flow chart of a method for fabricating a spin-orbit torque-based neural component, according to an embodiment of the present disclosure.
Detailed Description
Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is illustrative only and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present disclosure.
It will be understood that when an element such as a layer, film, region, or substrate is referred to as being "on" another element, it can be directly on the other element or intervening elements may also be present. Also, in the specification and claims, when an element is described as being "connected" to another element, the element may be "directly connected" to the other element or "connected" to the other element through a third element.
The present disclosure provides a spin orbit torque based neural component, including: an antiferromagnetic pinning layer, a first ferromagnetic layer and a spin orbit coupling layer sequentially formed on the substrate; a free layer formed on the spin orbit coupling layer and moving the magnetic domain wall according to the spin orbit torque; a tunneling layer formed on the free layer; left and right pinned layers formed at both sides of the free layer and having opposite magnetization directions; a reference layer formed on the tunneling layer; the free layer, the tunneling layer and the reference layer form a magnetic tunnel junction, and the magnetic tunnel junction is used for reading neuron signals.
The spin orbit torque-based nerve component provided by the embodiment of the disclosure achieves the following technical effects: first, pinning of the domain wall is achieved by depositing antiferromagnets with opposite magnetization directions on either side of the free layer or retaining/depositing a thicker local free layer. Secondly, the accumulation and leakage characteristics of the neurons are realized by multiplexing the spin orbit coupling layer, on one hand, current is introduced into the spin orbit coupling layer, and the spin hall effect generates spin current in the vertical direction, so that the motion of a domain wall is realized under the action of spin orbit torque, and the accumulation characteristics of the neurons are simulated; on the other hand, under the condition of no current, the free layer and the second ferromagnetic layer are coupled through the spin orbit coupling layer through the RKKY action ferromagnetism or antiferromagnetism, so that the domain wall in the free layer has the movement tendency opposite to the current driving direction, and the leakage function is realized. Thirdly, the inclined magnetization direction or the inclined magnetic anisotropy of the second ferromagnetic layer or the free layer is realized by depositing stray fields of ferromagnetic layers with magnetic anisotropy in the surface or by means of exchange bias, interlayer exchange coupling or annealing in an inclined magnetic field, so that an equivalent field in the x direction is generated on the free layer, the chirality of a magnetic domain wall is ensured to be unchanged, high-speed motion of the domain wall can be realized, and uncertainty of the motion direction of the next accumulation process caused by chirality change of a precession induced domain wall in the leakage process is avoided. Finally, according to the amplitude, pulse width and number of current pulses from synapses, magnetic domain walls are driven to move, and the function of accumulation of neurons is achieved; when no current pulse exists, the domain wall can move to the opposite direction under the RKKY action of the lower ferromagnetic layer, and the leakage function of the neuron is realized; when the domain wall moves to the threshold region, namely the magnetization direction of the corresponding local free layer below the reference layer is inverted, the MTJ combines with the peripheral circuit to output a spike signal, thereby realizing the discharge function of the neuron.
The following describes the technical solution of the present disclosure in detail with reference to the structure of a neural device in some specific embodiments of the present disclosure. It should be understood that the material layers, shapes and structures of the parts of the spin orbit torque based neural component shown in fig. 1 to 6 are only exemplary to help those skilled in the art understand the technical solution of the present disclosure, and are not intended to limit the scope of the present disclosure.
Example 1
Fig. 1 schematically shows a front view of a spin orbit torque based neuron device according to a first embodiment of the present disclosure, which is a neuromorphic device that implements a characteristic simulating leakage-accumulation-discharge similar to that of a biological neuron by means of spin orbit torque-induced domain wall movement.
As shown in fig. 1, a spin orbit torque based neural component according to an embodiment of the present disclosure includes: a substrate 115, an antiferromagnetic pinning layer 111, a first ferromagnetic layer 110, a spin-orbit coupling layer 109, a right electrode 108, a free layer 107, a right pinning layer 106, a top electrode 105, a reference layer 104, a tunneling layer 103, a left pinning layer 102, and a left electrode 101.
The base 115 may be a sapphire substrate, a silicon substrate, a quartz substrate, or the like with different crystal planes.
An antiferromagnetic pinning layer 111 formed on the upper surface of the substrate 115.
A first ferromagnetic layer 110 formed on the upper surface of the antiferromagnetic pinning layer 111.
A spin-orbit coupling layer 109 formed on an upper surface of the first ferromagnetic layer 110, the spin-orbit coupling layer 109 having a lateral length larger than that of the first ferromagnetic layer 110. In the embodiment of the present disclosure, the left electrode 101 and the right electrode 108 are provided on both sides of the upper surface of the spin orbit coupling layer 109 on which the free layer 107 is not provided.
A free layer 107 formed on the upper surface of the spin-orbit coupling layer 109, the free layer 107 having a smaller lateral length than the spin-orbit coupling layer 109. In the embodiment of the present disclosure, the left electrode 101 and the right electrode 108 are provided on both sides of the upper surface of the spin orbit coupling layer 109 on which the free layer 107 is not provided.
And a tunneling layer 103 formed on the upper surface of the free layer 107, wherein the tunneling layer 103 has a smaller lateral length than the free layer 107, and a threshold region is formed in a region of the tunneling layer 103 directly opposite to the free layer 107. In the embodiment of the present disclosure, the left pinned layer 102 and the right pinned layer 106 having opposite magnetization directions are disposed on both sides of the upper surface of the free layer 107 on which the tunneling layer 103 is not disposed.
And a reference layer 104 formed on the upper surface of the tunneling layer 103.
And a top electrode 105 formed on an upper surface of the reference layer 104.
Specifically, the free layer 107, the tunneling layer 103, and the reference layer 104 form a Magnetic Tunnel Junction (MTJ) for reading, the free layer 107, the spin-orbit coupling layer 109, and the first ferromagnetic layer 110 are coupled by RKKY exchange of the spin-orbit coupling layer 109, specifically, according to a change in the layer thickness of the spin-orbit coupling layer 109, a ferromagnetic coupling or an antiferromagnetic coupling can be achieved, generally, ferromagnetic and antiferromagnetic couplings exhibit a change in the oscillation property with an increase in the layer thickness of the spin-orbit coupling layer 109, and the oscillation period is about 1nm, for example, when the thickness of the spin-orbit coupling layer 109 made of a W material is less than 0.43nm, ferromagnetic coupling is achieved; when the thickness of the spin orbit coupling layer 109 composed of the W material is more than 0.43nm and less than 0.75nm, antiferromagnetic coupling is achieved.
In this embodiment, taking the antiferromagnetic coupling effect as an example, the reference layer 104 and the free layer 107 have perpendicular magnetic anisotropy, and are composed of one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe, and the thickness of the reference layer 104 is preferably 0.8-2 nm, and the thickness of the free layer 107 is preferably 0.8-2 nm. The tunneling layer 103 is made of MgO and Al2O3And the thickness of the layer is preferably 0.5 to 4 nm. The first ferromagnetic layer 110 has a thick in-plane magnetic anisotropy, and is made of one or more materials selected from CoFeB, NiFe, and Co, and the layer thickness thereof may satisfy the in-plane anisotropy. Spin orbit coupling layer109 is made of one or more of Ta, W, Mo and other metals, and is characterized in that the ferromagnetic layers on two sides can generate RKKY exchange effect, and meanwhile, the ferromagnetic layers have strong spin-orbit coupling effect and obvious spin Hall effect. The left pinned layer 102 and the right pinned layer 106 may be made of an antiferromagnetic material such as IrMn, PtMn, or the like, or may be made of the same material as the free layer 107. The antiferromagnetic pinning layer 111 has a perpendicular antiferromagnetic exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO. The left electrode 101, the top electrode 105 and the right electrode 108 are made of metal or alloy such as Ti, Au, Ti/Pt, Cr/Au, Ta/CuN, etc., and the thickness of each electrode layer is preferably 50 to 200 nm.
In this embodiment, the magnetization direction of the reference layer 104 is along the-z direction; the left pinned layer 102 has a magnetization direction in the-z direction and the right pinned layer 106 has a magnetization direction in the + z direction, so that a domain wall can move in a range sandwiched between the left and right pinned regions without annihilation; the magnetization direction of the first ferromagnetic layer 110 is along the x direction, the bottom antiferromagnetic pinning layer 111 is along the + z direction, so that the magnetization direction of the first ferromagnetic layer 110 can be pinned in the-z and x directions, a strong exchange bias field exists at the interface of the first ferromagnetic layer 100 and the bottom antiferromagnetic pinning layer 111 which is exchanged with the perpendicular, so that the magnetization direction of the first ferromagnetic layer 110 is inclined (in the x-z plane), equivalent fields in the + z direction and the x direction can be generated on the free layer 107 through the RKKY action of the spacer layer, the chirality of a domain wall in the free layer 107 can be kept unchanged, the high-speed motion of the domain wall can be realized, and the uncertainty of the motion direction of the next accumulation process caused by the chirality change of a precession induced domain wall in the leakage process is avoided.
Example 2
The spin orbit torque-based neural device structure in the present embodiment is shown in fig. 1, and the present embodiment differs from embodiment 1 in that:
in the present embodiment, taking the antiferromagnetic coupling effect as an example, the first ferromagnetic layer 110 has a first ferromagnetic layer with a tilted magnetic anisotropy (in the x-z plane), which is composed of one or more of CoFeB, NiFe, Co, CoFeAl, CoFe, and which can be specifically formed by annealing in a tilted magnetic field, sputtering at an inclined angle, and the likeAnd (5) realizing. The antiferromagnetic pinning layer 111 has a tilted antiferromagnetic exchange effect consisting of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO. It should be noted that, in this embodiment, other material layers are the same as those in embodiment 1, and are not described herein again.
Specifically, the magnetization direction of the reference layer 104 is in the-z direction, the magnetization direction of the left pinned layer 102 is in the-z direction, and the magnetization direction of the right pinned layer 106 is in the + z direction, so that a magnetic domain wall can move within a range sandwiched by the left and right pinned regions without annihilation; the bottom antiferromagnetic pinning layer 111 is along the + z and x directions so that the magnetization direction of the first ferromagnetic layer 110 can be pinned in the-z and x directions, the magnetization direction of the first ferromagnetic layer 110 with tilted magnetic anisotropy is tilted (in the x-z plane), and equivalent fields in the + z direction and the x direction can be generated on the free layer 107 by the RKKY action of the spacer layer, so that the chirality of the domain wall in the free layer 107 can be kept unchanged, and the high-speed motion of the domain wall can be realized and the uncertainty of the motion direction of the next accumulation process caused by the chirality change of the precession induced domain wall in the leakage process can be avoided.
Example 3
The spin orbit torque-based neural device structure in the present embodiment is shown in fig. 1, and the present embodiment differs from embodiment 1 in that:
in the present embodiment, taking the antiferromagnetic coupling effect as an example, the free layer 107 has a tilted magnetic anisotropy (in the x-z plane), which is made of one or more materials of CoFeB, NiFe, Co, CoFeAl, and CoFe, and which can be specifically realized by annealing in a tilted magnetic field. The first ferromagnetic layer 110 has perpendicular magnetic anisotropy and is composed of one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe.
Specifically, the magnetization direction of the left pinned layer 102 is in the-z direction, and the magnetization direction of the right pinned layer 106 is in the + z direction, so that a magnetic domain wall can move within a range sandwiched between the left and right pinned regions without annihilation; the first ferromagnetic layer 110 has a magnetization direction along the x-direction and the bottom antiferromagnetic pinning layer 111 along the + z-direction, so that the first ferromagnetic layer 110 magnetization direction can be pinned in the-z-direction and the reference layer 104 magnetization direction along the-z-direction; the magnetization direction of the region of the ferromagnetic free layer 107 in which the domain wall can move freely is along the x and + z directions, which can keep the chirality of the domain wall in the free layer 107 unchanged, not only can realize the high-speed movement of the domain wall, but also can avoid the uncertainty of the movement direction of the next accumulation process caused by the chirality change of the precession induced domain wall in the leakage process.
Example 4
Fig. 2 schematically illustrates a front view of a spin orbit torque based neural component, according to a fourth embodiment of the present disclosure.
As shown in fig. 2, the spin orbit torque-based neuron element structure in the present embodiment is different from that in embodiment 1 in that:
in this embodiment, a second ferromagnetic layer 113 is formed between the antiferromagnetic pinning layer 111 and the first ferromagnetic layer 110, wherein the second ferromagnetic layer 113 has perpendicular magnetic anisotropy and is composed of one or more materials selected from CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe. The first ferromagnetic layer 110 has a facing magnetic anisotropy and is composed of one or more of Co, CoFeB, NiFe.
In this embodiment, the magnetization direction of the left pinned layer 102 is along the-z direction, and the magnetization direction of the right pinned layer 106 is along the + z direction, so that the domain wall can move in the range sandwiched by the left and right pinned regions without annihilation; the first ferromagnetic layer 110 has a magnetization direction along the x-direction and the bottom antiferromagnetic pinning layer 111 along the + z-direction, so that the second ferromagnetic layer 113 has a magnetization direction pinned in the-z-direction and the reference layer 104 has a magnetization direction along the-z-direction; the exchange bias field existing at the interface of the first ferromagnetic layer 110 and the second ferromagnetic layer 113 can make the magnetization direction of the first ferromagnetic layer 110 inclined (in the x-z plane), and through the RKKY action of the spacer layer, equivalent fields in the + z direction and the x direction can be generated on the free layer 107, which can keep the chirality of the domain wall in the free layer 107 unchanged, thereby not only realizing high-speed motion of the domain wall, but also avoiding uncertainty of the motion direction of the next accumulation process caused by chirality change of precession-induced domain wall in the leakage process.
Example 5
Fig. 3 schematically illustrates a front view of a spin orbit torque based neural component, according to a fifth embodiment of the present disclosure.
As shown in fig. 3, the spin orbit torque-based neuron element structure in the present embodiment is different from that in embodiment 1 in that:
a second ferromagnetic layer 113 and an insulating layer 112 are formed in this order on the substrate, wherein the antiferromagnetic pinning layer 111 is located on the insulating layer 112. The first ferromagnetic layer 110 has perpendicular magnetic anisotropy, and is made of one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe. The second ferromagnetic layer 113 has in-plane magnetic anisotropy, and is composed of one or more materials of CoFeB, NiFe, and Co. The insulating layer 112 is made of SiO2Etc. of an insulating material.
In this embodiment, the magnetization direction of the left pinned layer 102 is along the-z direction, and the magnetization direction of the right pinned layer 106 is along the + z direction, so that the domain wall can move in the range sandwiched by the left and right pinned regions without annihilation; the bottom antiferromagnetic pinning layer 111 is in the + z direction so that the first ferromagnetic layer 110 magnetization can be pinned in the-z direction, with the reference layer 104 magnetization in the-z direction; the magnetization direction of the second ferromagnetic layer 113 is along the x direction, and the stray field generated by the second ferromagnetic layer 113 can keep the chirality of the domain wall in the free layer 107 unchanged, so that the high-speed motion of the domain wall can be realized, and the uncertainty of the motion direction of the next accumulation process caused by the chirality change of the precession-induced domain wall in the leakage process can be avoided.
Example 6
Fig. 4 schematically illustrates a front view of a spin orbit torque based neural component, according to a sixth embodiment of the present disclosure.
As shown in fig. 4, the spin orbit torque-based neuron element structure in the present embodiment is different from that in embodiment 1 in that:
a second ferromagnetic layer 113 and a spacer layer 114 are sequentially formed on the antiferromagnetic pinning layer 111, wherein the first ferromagnetic layer 110 is located on the spacer layer 114. Wherein, the spacing layer 114 is composed of one or more of Ru, Ta, W, V, Cr, Rh, Nd, Mo, Re. The second ferromagnetic layer 113 has perpendicular magnetic anisotropy, which is formed by: CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe.
In this embodiment, the magnetization direction of the left pinned layer 102 is along the-z direction, and the magnetization direction of the right pinned layer 106 is along the + z direction, so that the domain wall can move in the range sandwiched by the left and right pinned regions without annihilation; the first ferromagnetic layer 110 has a magnetization direction along the x-direction and the bottom antiferromagnetic pinning layer 111 along the + z-direction, so that the second ferromagnetic layer 113 has a magnetization direction pinned in the-z-direction and the reference layer has a magnetization direction along the-z-direction; the first ferromagnetic layer 110 and the second ferromagnetic layer 113 are coupled by RKKY exchange action of the spin-orbit coupling layer 109, so that the magnetization direction of the first ferromagnetic layer 110 is tilted (in an x-z plane), and equivalent fields in the + z direction and the x direction can be generated on the free layer 107 by the RKKY action of the spin-orbit coupling layer 109, so that the chirality of a domain wall in the free layer 107 can be kept unchanged, high-speed motion of the domain wall can be realized, and uncertainty of the motion direction of the next accumulation process caused by chirality change of a precession induced domain wall in a leakage process can be avoided.
In embodiments 1 to 6 provided in the present disclosure, in the initial state, the z-direction magnetization component of the freely movable region of the domain wall in the ferromagnetic free layer is in the + z direction, i.e., the magnetic domain wall is located near the boundary of the left pinned layer region. Therefore, when current is injected between the left electrode and the right electrode, the current flows through the spin orbit coupling layer, the spin current in the vertical direction is generated under the action of the spin Hall effect, the generated spin orbit torque can drive the domain wall to move along the + x direction, and the accumulation process of the neuron is simulated. When no current is injected, the first ferromagnetic layer indirectly acts on the ferromagnetic free layer through the synthetic antiferromagnetic coupling layer in the middle, so that the magnetization component of the first ferromagnetic layer in the z direction tends to be in the + z direction, and the leakage process of a neuron is simulated even if the domain wall moves in the-x direction; after a series of accumulation and leakage processes, the domain wall moves to exceed the corresponding area of the ferromagnetic reference layer, at the moment, the magnetization direction of the free layer is turned from the + z direction to the-z direction, the tunneling magnetoresistance is changed from larger antiparallel state resistance to smaller parallel state resistance, and a spike pulse can be output by combining an external circuit to simulate the discharge process of a neuron.
Fig. 5 schematically illustrates a leakage-accumulation-discharge characteristic diagram of a spin orbit torque-based neural component, according to an embodiment of the present disclosure, wherein,the free layer size is 60 × 300nm2The size of the left pinning region and the right pinning region is 60 multiplied by 30nm2. 6 successive applications of amplitude 4X 10 were applied to the device7A/cm2Current pulses with a pulse width of 0.4ns and a period of 0.8ns, the amplitude of the current pulses being 4X 107A/cm2In the period, the magnetic domain wall moves along the + x direction and is continuously accumulated; when the current pulse amplitude is 0, the magnetic domain wall moves along the-x direction under the RKKY exchange effect, so that the leakage process is realized; after 6 continuous pulses, the magnetic domain wall reaches a threshold region, the neuron is activated, and at the moment, the output circuit outputs a spike signal; and then, the neuron enters a withdrawing process, and the magnetic domain wall moves to the initial position along the-x direction under the action of the RKKY antiferromagnetic coupling, so that the complete leakage-accumulation-discharge process of the biological neuron is realized.
Fig. 6 is a schematic diagram illustrating the accumulation characteristics of a spin orbit torque-based neural component and DM anti-symmetric exchange, according to an embodiment of the present disclosure, and it can be seen from fig. 6 that as the DMI anti-symmetric effect is enhanced, the moving speed of the magnetic domain wall is increased, so that the characteristics of the accumulation process can be appropriately adjusted by adjusting the strength of the DMI anti-symmetric effect as needed.
Fig. 7 is a schematic diagram illustrating a relation between an accumulation characteristic of a spin-orbit torque-based neural component and a damping coefficient according to an embodiment of the present disclosure, and it can be seen from fig. 7 that as the damping coefficient increases, a magnetic domain wall motion speed is reduced, so that a characteristic of an accumulation process can be appropriately adjusted by adjusting the damping coefficient as needed.
It should be noted that, the length and width of each semiconductor material layer in the above embodiments, and specific examples are only illustrative, and do not limit the embodiments of the present disclosure.
Fig. 8 is a flowchart schematically illustrating a method for manufacturing a spin orbit torque-based neural device according to an embodiment of the present disclosure, where the structures of the neural device manufactured by the steps of the method are shown in fig. 1 to 4.
As shown in fig. 8, the method for manufacturing a spin orbit torque-based neural device includes:
s801, an antiferromagnetic pinning layer, a first ferromagnetic layer, a spin orbit coupling layer and a free layer are grown on a substrate in sequence.
And S802, respectively forming a left electrode and a right electrode on two sides of the spin orbit coupling layer.
And S803, sequentially forming a tunneling layer, a reference layer and a top electrode on the free layer, wherein the free layer, the tunneling layer and the reference layer form a magnetic tunnel junction, and the magnetic tunnel junction is used for reading neuron signals.
It should be noted that, in the embodiment of the present disclosure, the structure of the neural device prepared through the above process is shown in fig. 1 to 4, and the specific material layers, layer thicknesses, and structures formed by the material layers are also shown in fig. 1 to 4, which are not described in detail herein.
It should be noted that, the process methods and materials used in the above steps in the embodiments of the present disclosure are only exemplary, for example, the above semiconductor layer may also obtain a high-quality epitaxial film by using conventional film growth and etching methods, such as PVD, MBE, ALD, IBE, RIE, ICP, and the like, and the present disclosure is not limited thereto.
From the above description, it can be seen that the above-described embodiments of the present disclosure achieve at least the following technical effects:
1) the nerve component based on the magnetic tunnel junction can accumulate current pulses from synapses under the condition of a full electric field, has high energy efficiency and a high reliability self-leakage function, and when the accumulated current pulses drive a magnetic domain wall to move and exceed a threshold region, neurons can activate and emit a spike signal to simulate the function of human brain neurons.
2) The neural component realizes the spin orbit torque driving magnetic domain wall.
3) And the chirality of the magnetic domain wall is ensured to be unchanged by modulating the inclined magnetization direction or the inclined magnetic anisotropy of the first ferromagnetic layer or the free layer, so that the high-speed nerve component is realized.
While the disclosure has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive.
Those skilled in the art will appreciate that various combinations and/or combinations of features recited in the various embodiments and/or claims of the disclosure can be made to the extent not expressly recited in the disclosure. In particular, various combinations and/or combinations of the features recited in the various embodiments and/or claims of the present disclosure may be made without departing from the spirit or teaching of the present disclosure. All such combinations and/or associations are within the scope of the present disclosure.
While the disclosure has been shown and described with reference to certain exemplary embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure as defined by the appended claims and their equivalents. Accordingly, the scope of the present disclosure should not be limited to the above-described embodiments, but should be defined not only by the appended claims, but also by equivalents thereof.

Claims (10)

1. A spin orbit torque based neural component, comprising:
an antiferromagnetic pinning layer (111), a first ferromagnetic layer (110), and a spin orbit coupling layer (109) formed on a substrate in this order;
a free layer (107) formed on the spin orbit coupling layer (109) and moving a magnetic domain wall according to a spin orbit torque;
a tunneling layer (103) formed on the free layer (107);
left and right pinned layers (102, 106) formed on both sides of the free layer (107) and having opposite magnetization directions;
a reference layer (104) formed on the tunneling layer (103); wherein the free layer (107), the tunneling layer (103), and the reference layer (104) form a magnetic tunnel junction for reading neuron signals.
2. The spin orbit torque-based neural component of claim 1,
the free layer (107) has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the first ferromagnetic layer (110) has in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, and Co;
the antiferromagnetic pinning layer (111) has a vertical exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
3. The spin orbit torque-based neural component of claim 1,
the free layer (107) has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the first ferromagnetic layer (110) has a first ferromagnetic layer with a tilted magnetic anisotropy composed of one or more materials of CoFeB, NiFe, Co, CoFeAl, CoFe;
the antiferromagnetic pinning layer (111) has a tilted exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
4. The spin orbit torque-based neural component of claim 1,
the free layer (107) has a tilted magnetic anisotropy, and is composed of one or more materials of CoFeB, NiFe, Co, CoFeAl and CoFe;
the first ferromagnetic layer (110) has perpendicular magnetic anisotropy and is composed of one or more materials selected from CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe;
the antiferromagnetic pinning layer (111) has a vertical exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
5. The spin orbit torque-based neural component of claim 1, further comprising:
a second ferromagnetic layer (113) formed between the antiferromagnetic pinning layer (111) and the first ferromagnetic layer (110).
6. A spin-orbit torque-based neural component as claimed in claim 5,
the free layer (107) has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the first ferromagnetic layer (110) has in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, and Co;
the antiferromagnetic pinning layer (111) has a vertical exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO;
the second ferromagnetic layer (113) has perpendicular magnetic anisotropy and is composed of one or more materials selected from CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe.
7. The spin orbit torque-based neural component of claim 1, further comprising:
a second ferromagnetic layer (113) and an insulating layer (112) formed in that order on the substrate, wherein an antiferromagnetic pinning layer (111) is located on the insulating layer (112).
8. A spin-orbit torque-based neural component as claimed in claim 7,
the free layer (107) has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the first ferromagnetic layer (110) has perpendicular magnetic anisotropy and is composed of one or more materials selected from CoFeB, Co/Pt, CoFeAl, Co/Pd, and CoFe;
the antiferromagnetic pinning layer (111) has a perpendicularExchange by IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO;
the insulating layer (112) is made of SiO2Forming;
the second ferromagnetic layer (113) has in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, and Co.
9. The spin orbit torque-based neural component of claim 1, further comprising:
a second ferromagnetic layer (113) and a spacer layer (114) formed in that order on the antiferromagnetic pinning layer (111), wherein the first ferromagnetic layer (110) is located on the spacer layer (114).
10. A spin-orbit torque-based neural component as claimed in claim 9,
the free layer (107) has perpendicular magnetic anisotropy and is composed of one or more materials of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the first ferromagnetic layer (110) has in-plane magnetic anisotropy and is composed of one or more materials of CoFeB, NiFe, and Co;
the spacing layer (114) is made of one or more materials of Ru, Ta, W, V, Cr, Rh, Nd, Mo and Re;
the second ferromagnetic layer (113) has perpendicular magnetic anisotropy, and is formed from: one or more of CoFeB, Co/Pt, CoFeAl, Co/Pd and CoFe;
the antiferromagnetic pinning layer (111) has a vertical exchange effect and is made of IrMn, FeMn, NiMn, CoMn, PtMn, Mn2One or more of Au, NiO and MnO.
CN202110840235.1A 2021-07-21 2021-07-21 Spin orbit torque-based nerve component Pending CN113657586A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110840235.1A CN113657586A (en) 2021-07-21 2021-07-21 Spin orbit torque-based nerve component

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110840235.1A CN113657586A (en) 2021-07-21 2021-07-21 Spin orbit torque-based nerve component

Publications (1)

Publication Number Publication Date
CN113657586A true CN113657586A (en) 2021-11-16

Family

ID=78490077

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110840235.1A Pending CN113657586A (en) 2021-07-21 2021-07-21 Spin orbit torque-based nerve component

Country Status (1)

Country Link
CN (1) CN113657586A (en)

Similar Documents

Publication Publication Date Title
Guo et al. Spintronics for energy-efficient computing: An overview and outlook
JP6778866B2 (en) Magnetoresistive element, magnetic memory device, manufacturing method, operating method, and integrated circuit
Xiong et al. Antiferromagnetic spintronics: An overview and outlook
Ikeda et al. Magnetic tunnel junctions for spintronic memories and beyond
Slaughter Materials for magnetoresistive random access memory
US9300301B2 (en) Nonvolatile magnetic logic device
CN109952662B (en) Spin device, method of operating the same, and method of manufacturing the same
Bromberg et al. Novel STT-MTJ device enabling all-metallic logic circuits
US20180374526A1 (en) Systems and Methods for Reducing Write Error Rate in Magnetoelectric Random Access Memory Through Pulse Sharpening and Reverse Pulse Schemes
US9070455B2 (en) Memristor device with resistance adjustable by moving a magnetic wall by spin transfer and use of said memristor in a neural network
US20080049488A1 (en) Spin-transfer based MRAM with reduced critical current density
US8198919B1 (en) Spin transfer torque triad for non-volatile logic gates
JP7168231B2 (en) Magnetoresistive element, circuit device and circuit unit
CN110061127B (en) Magnetic tunnel junction forming method and magneto-resistive random access memory
JP2005522044A (en) Synthetic ferrimagnetic sense layer for high density MRAM applications
Bhattacharya et al. Skyrmion-mediated voltage-controlled switching of ferromagnets for reliable and energy-efficient two-terminal memory
Dao et al. Chiral domain wall injector driven by spin–orbit torques
WO2020018624A1 (en) Electric-field-induced switching of antiferromagnetic memory devices
Zheng et al. Paradigm of magnetic domain wall-based in-memory computing
Rana et al. Applications of nanomagnets as dynamical systems: I
Li et al. Field-free magnetization switching induced by bulk spin–orbit torque in a (111)-oriented CoPt single layer with in-plane remanent magnetization
Tey et al. Chiral spin textures for next-generation memory and unconventional computing
CN112864314B (en) Magneto-resistor device, method for changing resistance state of magneto-resistor device and synapse learning module
CN113657586A (en) Spin orbit torque-based nerve component
Sampaio et al. Domain wall motion in nanostructures

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