KR102145271B1 - Method and system for providing dual magnetic tunneling junctions using spin-orbit interaction-based switching and memories utilizing the dual magnetic tunneling junctions - Google Patents

Abstract

Magnetic memory is described. The magnetic memory includes double magnetic junctions and a spin orbit interaction (SO) active layer(s). Each of the double magnetic junctions includes a first reference layer, a first nonmagnetic spacer layer, a free layer, a second nonmagnetic spacer layer, and a second reference layer. The free layer is magnetic and is between the nonmagnetic spacer layers. The nonmagnetic spacer layers are between the corresponding reference layers and the free layer. The SO active layer(s) is adjacent to the first reference layer of each double magnetic junction. The SO active layer(s) applies SO torque to the first reference layer due to a current flowing through the SO active layer(s) in a direction perpendicular to the direction between the SO active layer(s) and the first reference layer. The first reference layer has a magnetic moment that can be changed by at least SO torque. The free layer can be switched using a spin transfer write current driven through the double magnetic junction.

Description

TECHNICAL FIELD A method and system for providing dual magnetic tunneling junctions using spin-orbit interaction-based switching and memories utilizing the dual magnetic tunneling junctions}

The present invention relates to a magnetic junction and a magnetic memory, and more particularly, to a double magnetic junction using switching based on a spin orbit interaction, and a magnetic memory using the same.

Magnetic memories, especially Magnetic Random Access Memories (MRAMs), are getting more and more attention due to their potential for high read/write speed, excellent durability, non-volatile and low power consumption during operation. The magnetic memory may store information by using magnetic materials as an information storage medium. One type of magnetic memory is STT-RAM (Spin Transfer Torque Random Access Memory). STT-RAM uses a magnetic junction in which at least a portion is recorded by the current passing through the magnetic junction. The spin polarized current through the magnetic junction exerts a spin torque on the magnetic moment in the magnetic junction. Thus, the layer(s) having a magnetic moment responsive to the spin torque can be switched to a desired state.

As an example, FIG. 1 shows a general magnetic tunneling junction (MTJ) 10 that can be used in a general STT-RAM. A typical double MTJ 10 is generally provided on the lower contact 11 and uses a seed layer(s) 12. The general double MTJ 10 includes a first general antiferromagnetic layer (AFM) 14, a first general pinned layer 16 (or a reference layer), and a first general tunneling barrier layer 18. ), a general free layer 20, a second general tunneling barrier layer 22, a second general pinned layer 24, a second general AFM layer 26, and a general capping layer 28. do. Also, an upper contact 30 is shown.

The general contacts 11 and 30 are used to drive a current in a current-perpendicular-to-plane (CPP) direction or a z-axis shown in FIG. 1. The general seed layer(s) 12 is generally utilized to aid in the growth of subsequent layers having the desired crystal structure, such as the AFM layer 14. Typical tunneling barrier layers 18 and 22 are non-magnetic, and may be, for example, a thin insulator such as MgO.

The general pinned layers 16 and 24 and the general free layer 20 are magnetic. The magnetization (17, 25) of the general reference layers (16, 24) is generally fixed in a specific direction by an exchange-bias interaction with the corresponding AFM layers (14, 26) or , Pinned. Although illustrated as a single layer, the general fixed layers 16 and 24 may include a plurality of layers. For example, the general pinned layers 16 and/or 24 are synthetic antiferromagnetic layers (SAF layers) including magnetic layers that are antiferromagnetically coupled through a thin conductive layer such as ruthenium (Ru). Can be A plurality of magnetic layers in which a ruthenium (Ru) thin film is inserted may be used for the SAF layer. In another embodiment, the coupling across the Ru films may be ferromagnetic. Although a single reference layer 16 or 24 and a single tunneling barrier layer 18 or 26 may be used, if the reference layers 16 and 24 are in a dual state (the magnetic moments 17 and 25 of the reference layers 16 and 24) are used. ) Is fixed to antiparallel), the dual MTJ 10 can have the advantages of improved spin torque. However, the dual MTJ 10 in the dual state may have a reduced magnetoresistance. On the other hand, if the reference layers 16 and 24 are fixed in an antidual state (the magnetic moments 17 and 25 of the reference layers 16 and 24 are parallel), the double MTJ 10 is improved. It can have magnetoresistive. In addition, in a semi-duplex arrangement, the contribution of the spin transfer torque from the two reference layers 16, 24 counteracts each other. As a result, the amplitude of the spin transfer torque for the free layer in the half-duplex state can be substantially reduced compared to that of a similar cell with a double state or even a single barrier (the amplitude of the spin transfer torque for the free layer). . Therefore, the read error rate (this is the probability of unintentionally switching the free layer during a read operation) can be significantly reduced. This is a significant difference in the sensing margin that the read currents approach the write currents in a dual arrangement (this is the difference between the minimum read current acceptable by the sense amplifier and the current causing unacceptable read errors). Can make an increase possible. It can also alleviate the requirements for MTJ cell parameters, in particular for the thermal stability of the MTJ cell. This is because the read error rate depends on the thermal stability of the cell. However, this also means that switching based on spin transfer may require a larger write current.

The general free layer 20 has a changeable magnetization 21. Although shown as a single layer, the general free layer 20 may also include multiple layers. For example, the free layer 20 may be a composite layer including magnetic layers that are antiferromagnetically or ferromagnetically coupled through thin conductive layers such as ruthenium (Ru). Although illustrated in-plane, the magnetization 21 of the general free layer 20 may have a perpendicular anisotropic property. Similarly, the magnetization 17 of the general pinned layer 16 may be perpendicular to the plane.

In order to switch the magnetization 21 of the general free layer 20, a current is driven in a direction perpendicular to the plane (Z direction). Current carriers are spin polarized to apply a torque to the magnetization 21 of the general free layer. In a typical double MTJ 10, the spin torque from the reference layers 16, 24 can be added when these layers are in a semi-double state (magnetic moments 17, 25 are antiparallel). There will be. The spin transfer torque with respect to the magnetic moment 21 of the general free layer 20 is initially small when the magnetic moment 21 is parallel to the easy axis (stable state). In this respect, the stable state of the magnetic moment 21 also corresponds to a stagnation point in the switching. Due to the thermal fluctuations, the magnetic moment 21 can rotate out of alignment with the easy magnetization axis of the general free layer 20. Then, the spin transfer torque can act more effectively, and the magnetic moment of the free layer 20 can be switched. When sufficient current flows from the upper contact 30 to the lower contact 11, the magnetization 21 of the general free layer 20 can be switched parallel to the magnetization 17 of the general reference layer 16. When sufficient current flows from the lower contact 11 to the upper contact 24, the magnetization 21 of the free layer can be switched antiparallel to the magnetization 17 of the reference layer 16. The differences in magnetic configurations correspond to different magnetoresistances and thus to different logic states (eg, logic 0 and logic 1) of the general MTJ 10.

When used in STT-RAM applications, the free layer 20 of the general MTJ 10 prevents damage to the general magnetic junction 10, reduces the size of the transistor supplying current (not shown), and reduces memory operation. It is required to be switched at a relatively relatively low current to reduce the energy consumption for this. In addition, short current pulses are required for use in programming the general magnetic element 10 at high data rates. For example, in order to switch the magnetization of the general free layer 20 more quickly, current pulses of about 5-30 ns or less are required.

A typical MTJ 10 can be written using spin transfer and can be used for STT-RAM, but there are problems with this. As an example, write error rates (WER) may be higher than that required for memories with an acceptable pulse width. The write error rate WER is the possibility that the cell (ie, the magnetization 21 of the free layer 20 of the conventional magnetic junction) is not switched when a current that is at least the same as the general switching current is applied. WER is required to be 10 ^-9 or less. However, very high currents may be required to achieve switching the general free layer 20 at this WER value. Also, in shorter write current pulses, improving the write error rate has proven to be challenging. As an example, FIG. 2 is a graph 50 showing the trend of WERs for pulses of different widths. Note that the actual data of the graph 50 are not written. Instead, graph 50 represents trends. The pulse width from the longest to the shortest draws curves 52,54,56,58. As can be seen in graph 50, for larger pulse widths, the voltage applied to the WER versus junction 10 has a larger slope. Therefore, application of a larger write voltage at the same pulse width can result in a significant reduction in WER. However, as the pulse widths in curves 54, 56 and 58 become shorter, the slopes of curves 54, 56 and 58 decrease. For decreasing pulse widths, it is less likely that an increase in voltage and/or current will result in a decrease in WER. At sufficiently short pulses, even high current/voltage does not result in a lower error rate. As a result, memories using the typical MTJ 10 may have an unacceptable large WER, which may not be healed by an increase in voltage. In addition, in order to obtain a high spin transfer torque, the reference layers 16 and 24 have magnetic moments 17 and 25 in a double state (fixed in opposite directions to each other). In this state, magneto-resistance is canceled during the read operation, which reduces the read signal. This reduction in signal is undesirable.

For that reason, there is a need for a method and system that can improve the performance of memories based on spin transfer torque. The methods and systems described herein address this need.

An object of the present invention is to provide a method and system capable of improving the performance of memories based on spin transfer torque.

Magnetic memory is described. The magnetic memory includes double magnetic junctions and a spin orbit interaction (SO) active layer(s). Each of the double magnetic junctions includes a first reference layer, a first nonmagnetic spacer layer, a free layer, a second nonmagnetic spacer layer, and a second reference layer. The free layer has magnetic properties, and is between the first nonmagnetic spacer layer and the second nonmagnetic spacer layer. The first nonmagnetic spacer layer is between the first reference layer and the free layer. The second nonmagnetic spacer layer is between the second reference layer and the free layer. The SO active layer(s) is adjacent to the first reference layer of each of the double magnetic junctions. The SO active layer(s) applies SO torque to the first reference layer due to a current flowing through the at least one SO active layer in a direction substantially perpendicular to a direction between the at least one SO active layer and the first reference layer. Is configured to apply. The first reference layer has a magnetic moment configured to be changed by at least the SO torque. The free layer is configured to be switched using a spin transfer write current driven through the double magnetic junction.

According to embodiments of the present invention, performance of memories based on spin transfer torque may be improved.

1 shows a typical magnetic junction.
2 shows the write error rate versus voltage of a typical spin transfer torque RAM.
3 shows an exemplary embodiment of a magnetic junction that is switched using spin orbit interaction.
4 shows an exemplary embodiment of a portion of a magnetic memory comprising a double magnetic junction that is switched using spin orbit interaction.
5A-5D illustrate another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin orbit interaction in switching.
6A-6D illustrate another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin orbit interaction in switching.
7A and 7B illustrate another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin orbit interaction in switching.
8A and 8B illustrate another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin orbit interaction in switching.
9 shows another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin orbit interaction.
10 shows another exemplary embodiment of a portion of a magnetic memory including a double magnetic junction that is switched using spin trajectory interaction.
11 shows another exemplary embodiment of a portion of a magnetic memory comprising a double magnetic junction that is switched using spin orbit interaction.
12 shows another exemplary embodiment of a portion of a magnetic memory comprising a double magnetic junction that is switched using spin orbit interaction.
13 shows an exemplary embodiment of a portion of a memory including a double magnetic junction that is switched using spin orbit interaction.
14 shows another exemplary embodiment of a portion of a memory including a double magnetic junction that is switched using spin orbit interaction.
15 shows another exemplary embodiment of a portion of a memory including a double magnetic junction that is switched using spin orbit interaction.
16 illustrates another exemplary embodiment of a portion of a memory including a double magnetic junction that is switched using spin orbit interaction.
17 is a flow chart illustrating one exemplary embodiment of a method of providing switched magnetic junction(s) using spin orbit interaction.
18 is a flow chart illustrating an exemplary embodiment of a method of programming a switched magnetic junction(s) using spin orbit interaction.

Exemplary embodiments relate to a magnetic junction that can be used in a magnetic device, such as magnetic memories, and to devices using such magnetic junctions. Hereinafter, the description has been provided so that those of ordinary skill in the technical field of the present invention can practice the present invention, and is provided as part of a patent application and its requirements. Various modifications of the exemplary embodiments and principles and forms thereof described in the present specification may be apparent to those of ordinary skill in the art to which the present invention pertains. Although the exemplary embodiments have been mainly described with specific methods and systems provided in a specific embodiment, the methods and systems may operate effectively in other implementations. Phrases such as “exemplary embodiment”, “one embodiment”, and “another embodiment” may refer to a plurality of embodiments as well as the same or different embodiments. The embodiments will be described with respect to systems and/or devices having certain configurations, but systems and/or devices may include more or less configurations than the illustrated configurations, and variations in the form of arrangement and configurations. Can be made within the scope of the present invention. In addition, exemplary embodiments may be described in the context of specific methods having certain steps, but such a method and system may have different and/or additional steps or other methods having steps out of order than the exemplary embodiments. Will work in valid. Accordingly, the present invention is not intended to be limited to the illustrated embodiments, and is to be accorded the widest scope not contradictory to the principles and forms described herein.

Exemplary embodiments describe a method and systems for providing a magnetic memory. The magnetic memory includes double magnetic junctions and at least one spin-orbit interaction (SO) active layer. Each of the double magnetic junctions includes a first reference layer, a first nonmagnetic spacer layer, a free layer, a second nonmagnetic spacer layer, and a second reference layer. The free layer is magnetic and is between the first nonmagnetic spacer layer and the second nonmagnetic spacer layer. The first nonmagnetic spacer layer is between the first reference layer and the free layer. The second nonmagnetic spacer layer is between the second reference layer and the free layer. The SO active layer(s) is adjacent to the first reference layer of each of the double magnetic junctions. The SO active layer(s) is due to a current flowing through the at least one SO active layer in a direction substantially perpendicular to the direction between the at least one SO active layer and the first reference layer, and thus a spin orbital interaction (SO) in the first reference layer. ) It is configured to apply torque. The first reference layer has a magnetic moment configured to be changeable by at least SO torque. The free layer is configured to be switched using a spin transfer write current driven through a double magnetic junction.

Exemplary embodiments are described within the context of magnetic memories having specific magnetic junctions, and specific components. Those of ordinary skill in the art to which the present invention pertains will readily appreciate that the present invention is consistent with the use of magnetic junctions and magnetic memories having different and/or additional configurations and/or other features not contradicting the present invention. will be. The methods and systems are also described within the context of the current understanding of spin orbit interactions, spin transfer phenomena, magnetic anisotropy and other physical phenomena. Therefore, those of ordinary skill in the art to which the present invention pertains will appreciate that theoretical explanations for the operation of the method and system are made based on this current understanding of spin orbital interaction, spin transfer, magnetic anisotropy, and other physical phenomena. You will know easily. However, the methods and systems described herein do not rely on any particular physical description. Those of ordinary skill in the art will also readily appreciate that the methods and systems are described within the context of structures having a specific relationship to the substrate. However, one of ordinary skill in the art will readily appreciate that the method and system are consistent in other structures. In addition, the methods and systems are described within the context of synthetic and/or single specific layers. However, one of ordinary skill in the art to which the present invention pertains will readily appreciate that the layers may have different structures. Furthermore, the method and system is described within the context of magnetic junctions, spin orbit active layers, and/or other structures having specific layers. However, those of ordinary skill in the art to which the present invention pertains may also use magnetic junctions, spin orbit active layers, and/or other structures having additional and/or other layers not contradicting the method and system. You will know easily. In addition, certain configurations are described as magnetic, ferromagnetic and ferrimagnetic. As used herein, the term magnetic may include ferromagnetic, ferrimagnetic, or similar structures. Thus, as used herein, the term “magnetic” or “ferromagnetic” includes, but is not limited to, ferromagnetic materials and ferrimagnetic materials. The method and system are also described within the context of single magnetic junctions. However, one of ordinary skill in the art will readily appreciate that the method and system are consistent with the use of magnetic memories having a plurality of magnetic junctions. Furthermore, as used herein, "in-plane" is substantially in-plane or parallel to one or more layers of a magnetic junction. Conversely, "perpendicular" corresponds to a direction substantially perpendicular to one or more layers of the magnetic junction.

3 shows an exemplary embodiment of a portion of a previously developed magnetic memory 100 that utilizes spin trajectory interactions in switching. 3 is not a ratio of the actual size, it is for the purpose of understanding. Also, some of the magnetic memory 100 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. Magnetic memory 100 includes magnetic storage cells 102. The magnetic storage cell 102 may be one of a plurality of magnetic storage cells arranged in an array. Each of the magnetic storage cells may include a selection element 104 and a magnetic junction 110. In some embodiments, a plurality of magnetic junctions 110 and/or a plurality of selection elements 104 may be used in a single cell. Also shown is a bus 120 comprising a spin orbit interaction (SO) active layer 122. The common bus 120 is spread over a plurality of storage cells, one of which is shown in FIG. 3. In the illustrated embodiment, the material(s) forming the SO active layer 122 are only in the vicinity of the storage cell 102. Thus, other materials (including, but not limited to, higher conductive materials and/or non-magnetic materials) may be used between cells 102. However, in other embodiments, the common bus 120 may be composed of an SO active layer 122. In still other embodiments, the SO active layers 122 may be separated from the common bus 120. For example, the SO active layer 122 may be between the magnetic junction 110 and the common bus 120. In other embodiments, SO active layer 122 may be included as part of storage cell 102 and common bus 120 may be omitted.

In the illustrated embodiment, the magnetic junction 110 includes a data storage layer (or free layer) 112, a nonmagnetic spacer layer 114, and a reference layer 116. The spacer layer 114 is nonmagnetic. In some embodiments, the spacer layer 114 is an insulator (eg, tunneling barrier). In such embodiments, each spacer layer 114 contains crystalline MgO, which can improve the tunneling magnetoresistance (TMR) of the magnetic junction 110, and spin transfer efficiency and/or spin orbital interaction. Can include. In other embodiments, the spacer layer 114 may be a conductor such as Cu. In other alternative embodiments, the spacer layer 114 may be a different structure, for example a granular layer comprising conductive channels in an insulating matrix.

The free layer 112 is a free layer 112 having a switchable magnetic moment (not shown). When the magnetic junction 110 is in the standby state (unswitched), the magnetic moment of the free layer 112 follows the easy axis of magnetization of the free layer 112. The magnetic moment of the reference layer 116 is required to be substantially fixed while the magnetic memory 100 is operating. The reference layer 116 is shown as a single layer. However, in other embodiments, the reference layer 116 may be a multilayer including (but not limited to) a synthetic antiferromagnetic material having ferromagnetic layers separated by a nonmagnetic layer(s). The nonmagnetic layer(s) may be ruthenium (Ru). In some embodiments, the magnetic junction 110 also includes a pinning layer, such as an antiferromagnetic layer (not shown) that fixes the magnetic moment of the reference layer 116. In other embodiments, the magnetic moment of the reference layer 116 is fixed in a different way. The free layer 112 and the reference layer 116 are ferromagnetic, and thus may include one or more of Fe, Ni, and Co. In the illustrated embodiment, the magnetic moment of the layers 112 and 116 may be perpendicular to the plane in some embodiments. Thus, each of the layers 112 and 116 may have a vertical anisotropy field exceeding an out-of-plane demagnetization filed (typically a significant portion of 4πM _s ). In other embodiments, the magnetic moments are in-plane.

The magnetic moment of the free layer 112 is switched using the spin orbital interaction effect described below. In some embodiments, the free layer 112 can be switched using a combination of effects. For example, the magnetic moment of the free layer 112 may be switched using a spin transfer torque that can be assisted by a torque induced by a spin trajectory interaction as a main effect. However, in other embodiments, the main switching mechanism is the torque induced by the spin trajectory interaction. In some such embodiments, other effects (including but not limited to spin transmission torque) may aid in switching and/or selection of magnetic junction 110. In still other embodiments, the magnetic moment of the free layer 112 is switched using only the spin orbital interaction effect.

SO active layer 122 is a layer that has a strong spin orbital interaction and can be used to switch the magnetic moment of the free layer 112. SO active layer 122 may be used to produce spin-orbit field (spin-orbit field, H _SO). More specifically, the current passes through the SO active layer 122 and is driven in the plane. This can be achieved by driving a current (eg, having a current density J _SO ) through the common bus 120. Due to the spin orbital interaction in the SO active layer 122, a current flowing through this layer may generate a spin orbital field H _SO that is proportional to the current density J _SO . In some embodiments, the spin trajectory field H _SO is parallel to the vector p _SO . It is determined by the material parameters and the geometry of the SO active layer 122, and by the direction of the current density J _SO . For some other embodiments, the spin trajectory field H _SO is parallel to the vectors [ M x p _SO ], p _SO . Here, M is the vector of the magnetic moment 115. For some other embodiments, the spin trajectory field H _SO is proportional to the linear combination of the vectors [ M x p _SO ], p _SO . This spin orbital field H _SO is equal to the spin orbital torque T _SO for the magnetic moment 115. The spin orbital torque for the free layer 112 is given by T _SO = -γ[ M x H _SO ], where M is the vector of the magnetic moment 115. This mutually correlated torque and field is referred to herein interchangeably as the spin orbit torque and the spin orbit field. This reflects the fact that the spin orbit interaction is the source of the spin orbit torque and the spin orbit field. This terminology also distinguishes between the more general spin transfer torque (STT) and spin trajectory (SO) torque. The spin orbital torque is generated by a spin orbital interaction and a current driven in the plane of the SO active layer 122. For example, in the illustrated embodiment, the spin trajectory torque is generated by the current density J _SO . In contrast, the spin transfer torque is due to a perpendicular-to-plane current flowing through the free layer 112, the spacer layer 114, and the reference layer 116. This plane perpendicular current injects spin-polarized charge carriers into the free layer 112. In the illustrated embodiment, the spin transfer torque is due to the current density (J _STT ). The spin trajectory torque T _SO can rapidly change the direction of the magnetic moment of the free layer 112 from an equilibrium state parallel to the axis of easy magnetization. Since the current is in-plane, the current flowing through the SO active layer 122 has a very large current density (up to approximately 10 ⁸ A/cm ^{2 ).} Degree). This current density for SO active layer 122 is much greater than the current density flowing through the barrier of the MTJ cell. This is because the current density of the latter is limited by the size of the cell transistor and the breakdown voltage of the MTJ. Thus, the plane perpendicular current through the magnetic junction 110 usually does not exceed several MA/cm ² . Therefore, the spin trajectory torque T _SO generated by J _SO may be significantly greater than the maximum spin transfer torque STT generated by the current flowing through the MTJ cell. As a result, the spin trajectory torque T _SO can tilt the magnetization of the free layer considerably faster than the general spin transfer torque STT. The spin trajectory torque T _SO can tilt the magnetization of the free layer considerably faster than the typical STT torque of similar maximum amplitude. In some embodiments, other mechanisms, such as spin transfer, may be used to complete the switching. In other embodiments, the switching can be accomplished using spin trajectory torque. In this way, the generated spin orbit field/spin orbit torque can be used to switch the magnetic moment of the free layer 112.

In some embodiments, the spin trajectory interaction may include a combination of two effects such as a spin Hall effect and a Rashba effect. In many spin orbital active layers, the spin orbit interaction includes both the spin Hall effect and the Lashva effect, but one of the two dominates. Other spin orbit effects can also be applied. The spin hall effect is generally considered a bulk effect. In general for the spin Hall effect, the vector p _SO at a given surface of the spin orbital active line 122 is oriented perpendicular to the direction of the current and a normal vector for that surface. Materials that often exhibit the spin Hall effect include heavy metals or heavy metal doped materials. As an example, such materials may be selected from at least one of A and M (B doped). A is Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Cd, In, Sb, Te, Hf, Ta (including highly resistant amorphous β-Ta), W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, and/or combinations thereof; M includes at least one of Al, Ti, V, Cr, Mn, Cu, Zn, Ag, Hf, Ta, W, Re, Pt, Au, Hg, Pb, Si, Ga, GaMn or GaAs, and B is V, Cr, Mn, Fe, Co, Ni, P, S, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, InSb, Te, I, Lu Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy , Ho, Er, Tm and Yb. In some embodiments, the SO active layer 122 may be formed of or include Ir-doped Cu and/or Bi-doped Cu. Doping generally ranges from 0.1 to 10 atomic percent. In other embodiments, other materials may be used.

Another source of the spin orbital field H _{SO in the} SO active layer 122 may be related to the interaction at the interfaces. In this case, the strength of the spin orbit field is often related to the strength of the crystal field, and is often large at the interface. Because of the lattice constant mismatch with adjacent layers, the presence of heavy metal at the interface, and other effects, the spin orbital interaction can be quite large at some interfaces. The strong spin orbit effect at the interface is related to the gradient of the crystal field in the direction perpendicular to the interface direction, and is often referred to as the Lashva effect. However, as used herein, regardless of its origin and orientation, the Rashva effect refers to the spin orbital interactions at the interface. In at least some embodiments, the interfaces for the SO active layer 122 must be different in order to obtain a significant size of the Lashva effect. For example, the SO active layer 122 adjacent to the magnetic junction 110 is a Pt layer, the free layer 112 is a Co layer, and the nonmagnetic spacer layer 114 is an aluminum oxide or MgO nonmagnetic layer 114 When is, such a Rashba effect may occur in the SO active layer 122. In some embodiments, other materials may be used.

The unit vector of spin polarization (P _SO ) for the Lashva effect is generally perpendicular to the crystal field and current direction. Many of the SO active layers 122 have a crystal field perpendicular to the plane of the layer 120. As such, the spin orbital polarization may be in-plane (eg, in the direction of _HSO in FIG. 3). Instead, the SO active layer 122 may have an in-plane or inclined crystal field. As such, the SO active layer 122 has spin orbital polarization perpendicular to the plane (not shown in FIG. 3) or relatively inclined toward the plane (not shown in FIG. 3). In some such embodiments, the SO active layer 122 may be a surface alloy. For example, the SO active layer 122 is Cu, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Te, I , Lu, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy , Ho, Er, Tm, Yb, and/or combinations thereof. In other embodiments, the SO active layer 122 includes surface alloys of A/B (e.g., atoms of A disposed on the (111) surface of B, which is the main material), of the upper atomic layers. It is a mixture of A and B. A is Cu, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Te, I, Lu, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, It includes at least one of Yb, and B includes at least one of Si, Zn, Cu, Ag, Au, W, Zn, Cr, Pt, and Pd. In many embodiments, A includes two or three different materials. In some embodiments, at least 0.1 to three monolayers of A are deposited. In some such embodiments, approximately one third of a single layer of A is deposited. In some embodiments, it is substituted Bi/Ag, substituted Pb/Ag, substituted Sb/Ag, substituted Bi/Si, substituted Ag/Pt, substituted Pb/Ge, substituted Bi/Cu, and It may be one or more of a double layer comprising a layer disposed on the (111) surface of Au, Ag, Cu or Si. In other embodiments, SO active layer 122 may include compounds such as InGaAs, HgCdTe, or the double layer of _{LaAlO 3 / SrTiO 3, LaTiO 3} / SrTiO 3. In other embodiments, other materials may be used. In some embodiments, the Lashva effect will cause a spin orbit torque T _SO and an associated spin orbit field H _SO in the free layer 112.

Accordingly, the magnetic memory 100 can use the spin orbital interaction and the spin orbital field generated by the SO active layer 122 to switch the magnetic moment of the free layer 112. In some embodiments, the SO active layer 122 may generate a spin trajectory field H _SO depending on one or both of the spin Hall effect and the Rashba effect. Consequently, as used herein, conditions such as “spin orbital effect”, “spin orbital field” and/or “spin orbital interactions” include the Rashva effect, the spin Hall effect, some combination of the two effects, and / Or may include spin orbit coupling through some other spin orbit interaction or effect similar to a spin orbit interaction. The spin trajectory fields torque the magnetic moment of the data storage layer/free layer 112. This spin trajectory torque can be used to switch the magnetic moment of the free layer 112. In some embodiments, the spin trajectory field assists in switching the magnetic moment of the free layer 112. Other mechanisms, such as spin transfer torque, are the dominant switching mechanisms. In other embodiments, the spin trajectory torque is the primary switching mechanism for the magnetic moment of the free layer 112. However, in some such embodiments, the spin trajectory torque may be assisted by another mechanism such as spin transfer torque. The assistance can be in switching the magnetic moment of the free layer 112 and/or selecting a switched magnetic junction.

Since the spin trajectory torque can be used to switch the magnetic moment of the free layer 112, the performance of the memory 100 can be improved. As described above, the spin trajectory torque generated by the SO active layer 122 may reduce the switching time of the magnetic junction 110. The spin trajectory torque typically has a high efficiency (P _SO ) and is proportional to the current (J _SO ). Since this current density is in-plane and does not flow through the spacer layer 114, this spin orbital current can be increased without damaging the magnetic junction 110. As a result, the spin orbit field and the spin orbit torque can be increased. Therefore, a write time can be reduced and a write error rate can be improved.

Even if the previously developed memory 100 functions properly, a person of ordinary skill in the art to which the present invention pertains will readily find that more improvements are required. For example, in the previously developed memory 100, the SO active layer 122 applies a spin trajectory torque to the free layer 112 to assist in switching. To do so, the SO active layer 122 is close to the free layer 112. For example, the SO active layer 122 is adjacent to the free layer 112 or separated from the free layer 112 only by an optional spacer layer. In either case, the spin orbital torque acts on the free layer 112. Thus, the magnetic junction 110 is a single magnetic junction having one free layer 112 and one reference layer 116. The advantages of other configurations (such as a double magnetic junction with two reference layers and two nonmagnetic spacer layers) may not be achieved.

4 shows an exemplary embodiment of a portion of a magnetic memory 200 including a double magnetic junction 210 that is switched using spin orbit interaction. 4 is not a ratio of the actual size, and is for better understanding. Also, some of the magnetic memory 200 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. The magnetic memory 200 includes a magnetic storage cell having at least one magnetic junction 210. In some embodiments, the magnetic storage cell may include other components including other magnetic junctions and one or more selection devices, but is not limited thereto. This selection device may be a transistor. The magnetic storage cell may be one of a plurality of magnetic storage cells arranged in an array.

In the illustrated embodiment, the magnetic junction 210 includes a first reference layer 212, a first nonmagnetic spacer layer 214, a data storage layer/free layer 216, a second nonmagnetic spacer layer 218, and It includes a second reference layer 220. The spacer layers 214 and 218 are non-magnetic. In some embodiments, one or both of the spacer layers 214 and 218 are insulators (eg, tunneling barriers). In such embodiments, each of the spacer layers 214 and 218 may include crystalline MgO that may improve the TMR and spin transfer efficiency and/or spin orbital interaction of the magnetic junction 210. In other embodiments, the spacer layers 214 and 218 may be a conductor such as Cu. In other alternative embodiments, the spacer layers 214 and 218 may be a different structure, for example a granular layer comprising conductive channels in an insulating matrix.

The free layer 216 is a free layer 216 having a switchable magnetic moment (not shown). When the magnetic junction 210 is in the standby state (unswitched), the magnetic moment of the free layer 216 follows the easy axis of magnetization of the free layer 216. In some embodiments, the free layer 216 is a single layer comprising a ferromagnetic material and/or an alloy. In other embodiments, the free layer 216 may be multiple layers. The multilayer may be formed of ferromagnetic layers or a mixed layer of ferromagnetic and nonferromagnetic layer(s). For example, the free layer 216 may be a synthetic antiferromagnetic material (SAF) including magnetic layers into which nonmagnetic layer(s) such as ruthenium (Ru) are inserted. The free layer 216 may also be a multilayer ferromagnetic.

The magnetic junction 210 also includes reference layers 212 and 220. The reference layer(s) 212 and/or 220 may be a single layer or multiple layers composed of a ferromagnetic material. In some embodiments, the reference layer(s) 212 and/or 220 may include ferromagnetic layers and nonmagnetic layers. In some such embodiments, the reference layer(s) 212 and/or 220 may be SAF.

The magnetic moment (not shown) of the reference layer 220 is required to be fixed. Accordingly, in some embodiments, the double magnetic junction 210 may include a pinned layer that fixes the magnetic moment of the reference layer. For example, the pinned layer may be an antiferromagnetic layer (AFM layer) adjacent to the reference layer 220. In other embodiments, the magnetic moment of the reference layer 220 is fixed in a different way. The free layer 216 and the reference layers 216 and 220 are ferromagnetic, and thus may include one or more of Fe, Ni, and Co. Although magnetic moments are not shown, the magnetic moments of layers 212, 216, 220 may be perpendicular to the plane in some embodiments. Thus, each of the layers 212, 216 and/or 220 will have a perpendicular anisotropy field exceeding an out-of-plane demagnetization filed (typically a significant portion of 4πM _s ). I can. In other embodiments, the magnetic moments are in-plane.

SO active layer 230 has a strong spin-orbit interaction can be used to generate the spin-orbit field (spin-orbit field, H _SO). Thus, the SO active layer 230 is similar to the SO active layer 122. The spin trajectory interaction may be due to the spin Hall effect, the Lashva effect, other effects, or a combination thereof. In the illustrated embodiment, the SO active layer 230 may be an entire line. Accordingly, the SO active layer 230 may extend to a plurality of magnetic junctions 210. In some other embodiments, the SO active layer 230 may simply be within the region of the magnetic junction 210. This is indicated by a dotted line on layer 230. In still other embodiments, the SO active layer may be between the magnetic junction 210 (hence the reference layer 212) and the word line carrying the in-plane current J _SO . For simplicity of explanation, reference numeral 230 will be used for both the line and the portions of the line forming the SO active layer.

During operation, the current J _SO passes through the SO active layer 230 and is driven in the plane. The current through the SO active layer 230 is associated with a spin orbital interaction that can cause a spin orbital field ( _HSO ). The spin orbit field (H _SO ) is similar to the spin orbit torque (T _SO ) described earlier. Accordingly, the SO active layer 230 may include materials selected from A and M (B doped). A is Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Cd, In, Sb, Te, Hf, Ta (including highly resistant amorphous β-Ta), W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, and/or combinations thereof; M includes at least one of Al, Ti, V, Cr, Mn, Cu, Zn, Ag, Hf, Ta, W, Re, Pt, Au, Hg, Pb, Si, Ga, GaMn or GaAs, and B is V, Cr, Mn, Fe, Co, Ni, P, S, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, InSb, Te, I, Lu Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy , Ho, Er, Tm and Yb. In some embodiments, the SO active layer 230 may be formed of or include Ir-doped Cu and/or Bi-doped Cu. Doping generally ranges from 0.1 to 10 atomic percent. In other embodiments, other materials may be used. In other embodiments, SO active layer 230 may also have a spin orbit (SO) interaction related to the spin orbit interaction at the interfaces. In this case, the strength of the spin orbit field is often related to the strength of the crystal field, and is often large at the interface. Because of the lattice constant mismatch with adjacent layers, the presence of heavy metal at the interface, and other effects, the spin orbital interaction can be quite large at some interfaces. For example, the SO active layer 230 may be a Pt layer adjacent to the magnetic junction 210, the reference layer 212 may be a Co layer, and the nonmagnetic spacer layer may be an aluminum oxide or MgO layer. In some embodiments, other materials may be used.

In some embodiments, the SO active layer 230 may be a surface alloy. For example, the SO active layer 230 is Cu, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Te, I , Lu, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy , Ho, Er, Tm, Yb, and/or combinations thereof. In other embodiments, the SO active layer 230 includes surface alloys of A/B (e.g., atoms of A disposed on the (111) surface of B, which is the main material), of the upper atomic layers. It is a mixture of A and B. A is Cu, Zn, Ga, Ge, As, Se, Y, Zr, Nb, Mo, Tc, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Te, I, Lu, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Po, At, La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, It includes at least one of Yb, and B includes at least one of Si, Zn, Cu, Ag, Au, W, Zn, Cr, Pt, and Pd. In many embodiments, A includes two or three different materials. In some embodiments, at least 0.1 to three monolayers of A are deposited. In some such embodiments, approximately one third of a single layer of A is deposited. In some embodiments, it is substituted Bi/Ag, substituted Pb/Ag, substituted Sb/Ag, substituted Bi/Si, substituted Ag/Pt, substituted Pb/Ge, substituted Bi/Cu, and It may be one or more of a double layer comprising a layer disposed on the (111) surface of Au, Ag, Cu or Si. In other embodiments, SO active layer 230 may include compounds such as InGaAs, HgCdTe, or the double layer of _{LaAlO 3 / SrTiO 3, LaTiO 3} / SrTiO 3. In other embodiments, other materials may be used. In some embodiments, the Lashva effect will cause a spin trajectory torque T _SO and an associated spin trajectory field H _SO in the reference layer 212.

In some embodiments, the spin diffusion layer (not shown in FIG. 4) may be between the reference layer 212 and the SO active layer 230. In some embodiments the selective spin diffusion layer is a metal. However, in other embodiments, this layer may be a thin insulating material, for example crystalline MgO or other oxide or other insulating layer. The resistance-area (RA) of these layers should be small. For example, it is less than 2 Ohm-μm ² . In other embodiments, the selective spin diffusion layer may be a multilayer comprising two or more layers of different materials. If the selective spin diffusion insert layer is required to make a major contribution to the switching of the magnetic junction 210, the selective spin diffusion insert layer can be used to reduce the contribution of the spin orbit field or/and to enhance the contribution of the spin orbit field. In addition, the selective spin diffusion intercalation layer may be used to provide an improved seed layer for the reference layer 212 and/or to reduce the damping of the reference layer associated with proximity to the SO active layer 230. However, in other embodiments, the spin diffusion layer may be omitted as shown in FIG. 4.

The magnetic moment of the reference layer 212 is configured to be changed using the spin orbital interaction from the SO active layer 230. However, the magnetic moment of the reference layer 212 is fixed with respect to the read current and the spin transfer current driven through the magnetic junction 210. In some embodiments, the magnetic moment of the reference layer 212 may be disturbed from the equilibrium position by a spin trajectory (SO) torque from the SO active layer 230. In other embodiments, the magnetic moment of the reference layer 212 may be switched between equilibrium positions using the SO torque from the SO active layer 230. For example, the magnetic moment of the reference layer 212 is switched between states, so that the magnetic moments of the reference layers 212 and 220 may be double and half-duplex states for writing and reading, respectively. In these embodiments, the magnetic moment of the reference layer 212 may be stable in double and half-duplex states, or may require SO torque for stability. In still other embodiments, a combination of effects may be used. For example, the magnetic moment of the reference layer may be switched to a dual state, and parallel/antiparallel for writing may be disturbed by the magnetic moment of the free layer. For reading, the magnetic moment of the reference layer 212 can be switched to a half-duplex state.

The free layer 216 may be recorded using a combination of a spin orbit SO torque and a spin transfer torque. More specifically, the current J _SO passes through the SO active layer 230 and is driven in the plane. The current through the SO active layer 230 is associated with a spin orbital interaction that can cause a spin orbital field ( _HSO ). This spin orbital field ( _HSO ) is equivalent to the spin orbital torque ( _TSO ) with respect to the magnetic moment. The spin orbital torque for the reference layer 212 is given by T _SO =-γ [ M x H _SO ], where M is the magnitude of the magnetic moment of the reference layer 212. The SO torque with respect to the reference layer 212 disturbs the reference layer from an equilibrium state that is aligned with or opposite to the magnetic moment of the free layer 216. The magnetic moment of the reference layer 212 applies a field to the free layer 216 that tilts the free layer magnetic moment from the equilibrium state. In other words, the free layer magnetic moment is inclined from the point of stagnation. The spin transfer current J _STT may pass through the magnetic junction 210 and be driven perpendicular to the plane. Spin transfer torque can be applied to the free layer magnetic moment. Therefore, the magnetic moment of the free layer 216 is switched using STT.

In other embodiments, the current J _SO is driven in-plane through the SO active layer 230. This torque puts the reference layers 212 and 220 in a dual state. Thereafter, the free layer 216 may be recorded using a spin transfer torque. The magnetic moment of the free layer 216 may be recorded using spin transfer torque. When data is read from the magnetic junction 210, the SO current is driven in the opposite direction. Then, the reference layers 212 and 260 are in a semi-duplex state. After that, the magneto-resistance of the magnetic junction 210 can be read.

The magnetic junction 210 may undergo faster switching. Since the SO torque disturbs the magnetic moment of the reference layer 212 and the generated stray field moves the free layer magnetic moment away from the stagnation point, switching using the STT torque may be faster. In addition, since the double magnetic junction 210 is used, the spin transfer torque for the free layer 216 may be higher in the reference layers 212 and 220 in the double state. Thereafter, a lower switching current may be driven through the magnetic junction 210. If the reference layers 212 and 220 are in a half-duplex state, the magnetic resistance of the magnetic junction 210 may be higher. Thus, a higher signal can be obtained. Furthermore, in the half-duplex state, the read error rate can be significantly reduced due to the reduced STT amplitude at a given current density. Therefore, the sensing margin can be increased, and the requirements for thermal stability of the cells can be relaxed.

5A-5D illustrate an exemplary embodiment of a portion of a magnetic memory 300 including a double magnetic junction 310 that is switched using a spin trajectory interaction. 5A to 5D are not ratios of actual sizes, and are for better understanding. Also, some of the magnetic memory 200 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. The magnetic memory 300 is similar to the magnetic memory 200. Thus, the magnetic memory 300 includes a magnetic storage cell 310 and a line 330 including the SO active layer 330, similar to the line 230 including the magnetic junction 210 and the SO active layer 230 do. The magnetic storage cell 310 may be one of a plurality of magnetic storage cells arranged in an array.

The magnetic junction 310 includes a first reference layer 212, a first nonmagnetic spacer layer 214, a data storage layer/free layer 216, a second nonmagnetic spacer layer 218, and a second reference layer 220, respectively. ), a first reference layer 312, a first nonmagnetic spacer layer 314, a data storage layer/free layer 316, a second nonmagnetic spacer layer 318, and a second reference layer 320. . The spacer layers 314 and 318 are non-magnetic. In some embodiments, one or both of the spacer layers 314 and 318 are an insulating tunneling barrier such as crystalline MgO. In other embodiments, the spacer layer 314 and/or 318 may be a conductor. In other alternative embodiments, the spacer layers 314 and/or 318 may have other structures. Layers 312, 316, 320 are ferromagnetic and thus may include materials such as Co, Ni and/or Fe.

The free layer 316 has a switchable magnetic moment 317. When the magnetic junction 310 is in the standby state (unswitched), the magnetic moment 317 of the free layer 316 follows the easy axis of magnetization of the free layer 316. 5A-5D, the easy axis of magnetization is in the plane. However, in other embodiments, the easy axis of magnetization may be in another direction (including, but not limited to, a direction perpendicular to the plane). In some embodiments, the free layer 316 is a single layer comprising a ferromagnetic material and/or an alloy. In other embodiments, the free layer 316 may be a multilayer including SAF or other structures, but is not limited thereto.

Magnetic junction 310 also includes reference layers 312 and 320 having magnetic moments 313 and 321 respectively. Each of the reference layer(s) 312 and/or 320 may be a single layer made of a ferromagnetic material, or a multilayer such as SAF. The magnetic moment 321 of the reference layer 320 is required to be fixed or pinned. Thus, in some embodiments, the double magnetic junction 310 may include a pinned layer (not shown in FIGS. 5A-5D ), such as an AFM layer, that fixes the magnetic moment 321. The magnetic moment 313 of the reference layer 312 is fixed with respect to the read current and the spin transfer current driven through the magnetic junction 310. In other embodiments, the magnetic moment 321 may be fixed in different ways. In the illustrated embodiment, magnetic moments 313 and 321 are in plane. However, in some embodiments, the magnetic moment(s) 313 and/or 321 may be in other directions (including, but not limited to, a direction perpendicular to a plane).

SO active layer 330 has a spin-orbit field strong spin-orbit (SO) interaction, which may be used to produce (spin-orbit field, H _SO). The SO interaction may be due to the spin Hall effect, the Lashva effect, other effects, or a combination thereof. In the illustrated embodiment, the SO active layer 330 may be an entire line. Accordingly, the SO active layer 330 may extend to a plurality of magnetic junctions 310. In other embodiments, the SO active layer 330 may simply be within the region of the magnetic junction 310. In yet other embodiments, the SO active layer may be between the magnetic junction 310 (hence the reference layer 312) and the word line carrying the in-plane current J _SO . For simplicity of description, reference numeral 330 will be used for both the line and the portions of the line forming the SO active layer. Accordingly, the SO active layer 330 may include materials similar to those of the SO active layers 122 and 230 described above.

In some embodiments, an optional spin diffusion layer (not shown in FIGS. 5A-5D) may be between the reference layer 312 and the SO active layer 330. However, in other embodiments, the spin diffusion layer may be omitted as shown in FIGS. 5A to 5D.

In general, it is not caused by the spin transfer current and the read current driven through the magnetic junction 310, but the magnetic moment 313 of the reference layer 312 is changed using the spin orbital interaction from the SO active layer 330 It is structured to be. This change in magnetic moment occurs during a write operation of the free layer 316, as shown in FIGS. 5A to 5D. As shown in FIG. 5A, the in-plane current J _SO may be driven through the SO active layer 330. This creates a spin orbit field ( _HSO ). The SO field disturbs the position of the magnetic moment 313. The results are shown in Fig. 5B. The magnetic moment 313a is inclined from its previous position. Accordingly, the magnetic moment 313a of the reference layer 312 is no longer (inverted) aligned with the equilibrium position of the magnetic moment 317 of the free layer 316. As a result, the reference layer 312 _{applies a} magnetic field H _stray to the free layer magnetic moment 317. The free layer magnetic moment 317 is inclined from the equilibrium position by the magnetic field H _stray . This is shown in Figure 5c. Accordingly, the free layer magnetic moment 317a is inclined from the axis of easy magnetization shown by the dotted line in the free layer 316. The free layer magnetic moment 317a is inclined from the point of stagnation. The spin transfer current J _STT ,) is driven in a direction perpendicular to the plane through the magnetic junction 310, as shown in FIG. 5C. The spin transfer torque may be applied to the free layer magnetic moment 317a. In addition, the in-plane current J _SO can be removed. Accordingly, the magnetic moment of the free layer 316 is switched using STT, and the reference layer 312 returns to an equilibrium state. This is shown in Figure 5D. Thus, the free layer magnetic moment 317b is aligned antiparallel with the magnetic moment 313b. This is because the spin transfer torque current is driven away from the line 330 in FIGS. 5A to 5D. If the spin transfer torque current is driven in the opposite direction, the free layer magnetic moment 317b may be parallel to the magnetic moment 313b. In other embodiments, the switching direction of the free layer for a given spin transfer torque current may be opposite to that shown in FIGS. 5A-5D.

The magnetic junction 310 can undergo faster switching. Since the SO torque disturbs the magnetic moment 313a of the reference layer 312, the field Hs _tray moves the free layer magnetic moment away from the stagnation point. As a result, switching using STT torque can be faster with a reduced write error rate (WER). In addition, since the double magnetic junction 310 is used, the spin transfer torque for the free layer 316 is further reduced at the magnetic moments 313, 313a, 313b, 321 of the reference layers 312, 320 in the double state. It can be high. Thereafter, a lower spin transfer switching current J _STT may be driven through the magnetic junction 310. If the reference layers 312 and 320 are in a half-duplex state, the magnetic resistance of the magnetic junction 310 may be higher. Thus, a higher signal can be obtained. Furthermore, in the half-duplex state, the read error rate can be significantly reduced due to the reduced STT amplitude at a given current density. Therefore, the sensing margin can be increased, and the requirements for thermal stability of the cells can be relaxed. Switching between dual and semi-duplex states is described in the example with respect to FIGS. 8A and 8B below.

6A-6D illustrate another exemplary embodiment of a portion of a magnetic memory 400 including a double magnetic junction 410 that is switched using spin orbit interaction. 6A to 6D are not ratios of actual sizes, and are for better understanding. Also, some of the magnetic memory 200 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. The magnetic memory 400 is similar to the magnetic memories 200 and 300. Thus, the magnetic memory 400 includes magnetic storage cells 410 and SO active layers 430, similar to lines 230 and 330 including magnetic junctions 210 and 310 and SO active layers 230 and 330. Includes line 430. The magnetic storage cell 410 may be one of a plurality of magnetic storage cells arranged in an array.

The magnetic junction 410 includes a first reference layer 212 and 312, a first nonmagnetic spacer layer 214 and 314, a data storage layer/free layer 216 and 316, and a second nonmagnetic spacer layer 218 and 318, respectively. ), and a first reference layer 412 similar to the second reference layers 220 and 320, a first nonmagnetic spacer layer 414, a data storage layer/free layer 416, a second nonmagnetic spacer layer 418, And a second reference layer 420. The spacer layers 414 and 418 are each non-magnetic. In some embodiments, one or both of the spacer layers 414 and 418 are insulating tunneling barriers such as crystalline MgO. In other embodiments, the spacer layer 414 and/or 418 may be a conductor. In other alternative embodiments, the spacer layers 414 and/or 418 may have other structures. Layers 412, 416, 420 are ferromagnetic and thus may include materials such as Co, Ni and/or Fe.

The free layer 416 has a switchable magnetic moment 417. When the magnetic junction 410 is in the standby state (unswitched), the magnetic moment 417 of the free layer 416 follows the easy axis of magnetization of the free layer 416. In the embodiment shown in Figs. 6A to 6D, the easy axis of magnetization is perpendicular to the plane. Thus, the free layer 416 may have a vertical anisotropy field exceeding an out-of-plane demagnetization filed (typically a significant portion of 4πM _s ). However, in other embodiments, the easy axis of magnetization may be a different direction (including, but not limited to, an in-plane direction). In some embodiments, free layer 416 is a single layer comprising a ferromagnetic material and/or alloy. In other embodiments, the free layer 416 may be a multilayer including SAF or other structures, but is not limited thereto.

The magnetic junction 410 also includes reference layers 412 and 420 having magnetic moments 413 and 421 respectively. Each of the reference layer(s) 412 and/or 420 may be a single layer composed of a ferromagnetic material, or a multilayer such as SAF. The magnetic moment 421 of the reference layer 420 is required to be fixed or pinned. The magnetic moment 413 of the reference layer 412 is fixed with respect to the read current and the spin transfer current driven through the magnetic junction 410. In the illustrated embodiment, the magnetic moments 413 and 421 are perpendicular to the plane. Accordingly, each of the reference layers 412 and 420 may have a vertical anisotropy field exceeding an off-plane half-magnetization (generally a significant portion of 4πM _s ). However, in some embodiments, the magnetic moment(s) 413 and/or 421 may be in other directions (including, but not limited to, in-plane directions).

SO active layer 430 has a spin-orbit field strong spin-orbit (SO) interaction, which may be used to produce (spin-orbit field, H _SO). The SO interaction may be due to the spin Hall effect, the Lashva effect, other effects, or a combination thereof. In the illustrated embodiment, the SO active layer 430 may be an entire line. Accordingly, the SO active layer 430 may extend to a plurality of magnetic junctions 410. In other embodiments, the SO active layer 430 may simply be within the region of the magnetic junction 410. This is indicated by a dotted line on layer 430. In still other embodiments, the SO active layer may be between the magnetic junction 410 (hence the reference layer 412) and the word line carrying the in-plane current J _SO . For simplicity of description, reference numeral 430 will be used for both the line and the portions of the line forming the SO active layer. Accordingly, the SO active layer 430 may include materials similar to those of the SO active layers 122, 230, and 330 described above.

In some embodiments, an optional spin diffusion layer (not shown in FIGS. 6A-6D) may be between the reference layer 412 and the SO active layer 430. However, in other embodiments, the spin diffusion layer may be omitted as shown in FIGS. 6A to 6D.

The magnetic moment 413 of the reference layer 412 is configured to be changed using the spin orbital interaction from the SO active layer 430. However, as described above, the magnetic moment 413 may be stabilized with respect to a read current or a spin transfer current driven through the magnetic junction 410 in a direction perpendicular to the plane. The change of the magnetic moment 413 occurs during a write operation of the free layer 416, as shown in FIGS. 6A to 6D. As shown in FIG. 6A, the in-plane current J _SO may be driven through the SO active layer 430. This creates a spin orbit field ( _HSO ). The SO field disturbs the position of the magnetic moment 413. The results are shown in Fig. 6B. The magnetic moment 413a is inclined from its previous position. Now, the magnetic moment 413a has an in-plane component. Accordingly, the magnetic moment 413a of the reference layer 412 is no longer (inverted) aligned with the equilibrium position of the magnetic moment 417 of the free layer 416. As a result, the reference layer 412 _{applies a} magnetic field H _stray to the free layer magnetic moment 417. The free layer magnetic moment 417 is tilted from the equilibrium position by the magnetic field H _stray . This is shown in Figure 6c. Accordingly, the free layer magnetic moment 417a is inclined from the easy magnetization axis (shown in FIG. 6C) shown by a dotted line in the free layer 416. The free layer magnetic moment 417a has an in-plane component. In addition, the free layer magnetic moment 417a is inclined from the point of stagnation. The spin transfer current (J _STT ,) is driven in a direction perpendicular to the plane through the magnetic junction 410, as shown in FIG. 6C. Spin transfer torque may be applied to the free layer magnetic moment 417a. In addition, the in-plane current J _SO can be removed. Thus, the magnetic moment of the free layer 416 is switched using STT, and the reference layer 412 returns to an equilibrium state. This is shown in Fig. 6D. Thus, the free layer magnetic moment 417b is aligned parallel to the magnetic moment 413b. This is because the spin transfer torque current is driven toward the line 430. If the spin transfer torque current is driven in the opposite direction, the free layer magnetic moment 417b may become antiparallel to the magnetic moment 413b. In other embodiments, the switching direction of the free layer for a given spin transfer torque current may be opposite to that shown in FIGS. 6A-6D.

The magnetic junction 410 may undergo faster switching. Since the SO torque disturbs the magnetic moment 413a of the reference layer 412, the field Hs _tray moves the free layer magnetic moment away from the stagnation point. As a result, switching using STT torque can be faster with a reduced write error rate (WER). In addition, since the double magnetic junction 410 is used, the spin transfer torque for the free layer 416 is further at the magnetic moments 413, 413a, 413b, 421 of the reference layers 412, 420 in the double state. It can be high. Thereafter, a lower spin transfer switching current J _STT may be driven through the magnetic junction 410. If the reference layers 412 and 420 are in a half-duplex state, the magnetic resistance of the magnetic junction 410 may be higher. Thus, a higher signal can be obtained. Furthermore, in the half-duplex state, the read error rate can be significantly reduced due to the reduced STT amplitude at a given current density. Therefore, the sensing margin can be increased, and the requirements for thermal stability of the cells can be relaxed. Switching between dual and semi-duplex states is described in the example with respect to FIGS. 8A and 8B below.

7A and 7B illustrate another exemplary embodiment of a portion of a magnetic memory 400A including a double magnetic junction 410a that is switched using spin orbit interaction. 7A and 7B are not ratios of actual sizes, and are for better understanding. Also, some of the magnetic memory 200 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. The magnetic memory 400A is similar to the magnetic memories 200, 300 and 400. Accordingly, the magnetic memory 400A has a magnetic storage cell 410a and an SO active layer, similar to the lines 230, 330, and 430 including the magnetic junctions 210, 310, and 410 and the SO active layers 230, 330, and 430. It includes line 430a that includes 430a. The magnetic storage cell 410a may be one of a plurality of magnetic storage cells arranged in an array. The SO active layer may constitute an entire bus, or may exist only in the vicinity of the magnetic junction 410a. This is indicated by the dotted line of the common line/SO active layer 430a. For simplicity of description, reference numeral 430a will be used for both the line and the portions of the line forming the SO active layer.

The magnetic junction 410a includes a first reference layer 212, 312, 412, a first nonmagnetic spacer layer 214, 314, 414, a data storage layer/free layer 216, 316, 416, and a second nonmagnetic, respectively. A first reference layer 412a similar to the spacer layers 218, 318, 418, and the second reference layers 220, 320, 420, a first nonmagnetic spacer layer 414a, a data storage layer/free layer 416a, And a second nonmagnetic spacer layer 418a and a second reference layer 420a. Although layers 414a and 418a are shown as simple layers, one or both of layers 414a and 418a may be multi-layered (including, but not limited to, SAF).

In the illustrated embodiment, the reference layer 412a is a SAF comprising ferromagnetic layers 442 and 446 separated by a nonmagnetic layer 444 such as Ru. Each of the ferromagnetic layers 442 and 446 has magnetic moments 443 and 447. Layers 412a, 416a, 420a are shown having magnetic moments perpendicular to the plane. However, in other embodiments, the magnetic moment(s) may be in plane. In some embodiments, an optional spin diffusion layer (not shown in FIGS. 7A and 7B) may be between the reference layer 412a and the SO active layer 430a. However, in other embodiments, the spin diffusion layer may be omitted as shown in FIGS. 7A and 7B.

Although not caused by a read current and a spin transfer current driven through the magnetic junction 410a in a direction perpendicular to the plane, the magnetic moments 443 and 447 of the reference layer 412a are spin from the SO active layer 430a. It is constructed so that it can be changed using orbital interactions. As shown in FIG. 7A, the in-plane current J _SO may be driven through the SO active layer 430a. This creates a spin orbit field ( _HSO ). The SO field disturbs the position of the magnetic moment 443. Since it is antiferromagnetically coupled, the magnetic moment 447 is also inclined from a direction perpendicular to the plane, as shown in Fig. 7A. The magnetic moments 443 and 447 of the reference layer 412a are no longer aligned or aligned with the equilibrium position of the magnetic moment 417c of the free layer 416a, respectively. As a result, the reference layer 412a applies a magnetic field H _stray to the free layer magnetic moment 417c. In some embodiments, the magnitudes of the magnetic moments 443 and 447 are required to be unbalanced to increase the strength of the field H _stray . The free layer magnetic moment 417c will be tilted from the equilibrium position by the magnetic field H _stray . Thus, the free layer magnetic moment 417c will be tilted from the stagnation point. The spin transfer current J _STT is driven in a direction perpendicular to the plane through the magnetic junction 410a, as shown in FIG. 7A. Spin transfer torque may be applied to the free layer magnetic moment 417c. In addition, the in-plane current J _SO can be removed. Accordingly, the magnetic moment of the free layer 416a is switched using STT, and the reference layer 412a returns to an equilibrium state. This is shown in Figure 7b. Thus, the free layer magnetic moment 417d shown in FIG. 7B is aligned parallel to the magnetic moment 447a. This is because the spin transfer torque current is driven toward the line 430a. If the spin transfer torque current is driven in the opposite direction, the free layer magnetic moment 417d may become antiparallel to the magnetic moment 447a. In other embodiments, the switching direction of the free layer for a given spin transfer torque current may be opposite to that shown in FIGS. 7A and 7B.

The magnetic junction 410a may undergo faster switching. Since the SO torque disturbs the magnetic moments 443 and 447 of the reference layer 412a, the field Hs _tray moves the free layer magnetic moment 417c away from the stagnation point. As a result, switching using STT torque can be faster. In addition, since the double magnetic junction 410a is used, the spin transfer torque for the free layer 416a may be higher in the reference layers 412a, 420a in the double state. Thereafter, a lower spin transfer switching current J _STT may be driven through the magnetic junction 410a. If the reference layers 412a and 420a are in a half-duplex state, the magnetic resistance of the magnetic junction 410a may be higher. Thus, a higher signal can be obtained. Furthermore, in the half-duplex state, the read error rate can be significantly reduced due to the reduced STT amplitude at a given current density. Therefore, the sensing margin can be increased, and the requirements for thermal stability of the cells can be relaxed. Switching between dual and semi-duplex states is described in the example with respect to FIGS. 8A and 8B below.

8A and 8B illustrate another exemplary embodiment of a portion of a magnetic memory 500 including a double magnetic junction 510 that is switched using spin trajectory interaction while being switched. 8A and 8B are not ratios of actual sizes, and are for better understanding. Also, some of the magnetic memory 500 such as bit lines, word lines, row selectors, and column selectors are not shown or not shown. The magnetic memory 500 is similar to the magnetic memories 200, 300, 400, 400A. Thus, the magnetic memory 500 is a magnetic storage cell similar to the lines 230, 330, 430, and 430a including magnetic junctions 210, 310, 410, 410a and SO active layers 230, 330, 430, 430a 510 and a line 530 comprising an SO active layer 530. The magnetic storage cell 510 may be one of a plurality of magnetic storage cells arranged in an array. For simplicity of explanation, reference numeral 530 will be used for both the line and the portions of the line forming the SO active layer.

The magnetic junction 510 includes a first reference layer 212, 312, 412, 412a, a first nonmagnetic spacer layer 214, 314, 414, 414a, and a data storage layer/free layer 216, 316, 416, 416a, respectively. ), a second nonmagnetic spacer layer 218, 318, 418, 418a, and a first reference layer 512 similar to the second reference layer 220, 320, 420, 420a, a first nonmagnetic spacer layer 514, A data storage layer/free layer 516, a second nonmagnetic spacer layer 518, and a second reference layer 520 are included. The spacer layers 514 and 518 are each non-magnetic. In some embodiments, one or both of the spacer layers 514 and 518 are an insulating tunneling barrier such as crystalline MgO. In other embodiments, the spacer layer 514 and/or 518 may be a conductor. In other alternative embodiments, the spacer layers 514 and/or 518 may have other structures. Layers 512, 516, 520 are ferromagnetic and thus may include materials such as Co, Ni and/or Fe.

The free layer 516 has a switchable magnetic moment 517. When the magnetic junction 510 is in the standby state (unswitched), the magnetic moment 517 of the free layer 516 follows the easy axis of magnetization of the free layer 516. In the embodiment shown in Figs. 8A and 8B, the easy axis of magnetization is perpendicular to the plane. Thus, the free layer 516 may have a vertical anisotropic growth that exceeds an off-plane diamagnetic (typically a significant portion of 4πM _s ). However, in other embodiments, the easy axis of magnetization may be a different direction (including, but not limited to, an in-plane direction). In some embodiments, the free layer 516 is a single layer comprising a ferromagnetic material and/or an alloy. In other embodiments, the free layer 516 may be a multilayer including SAF or other structures, but is not limited thereto.

SO active layer 530 has a spin-orbit field strong spin-orbit (SO) interaction, which may be used to produce (spin-orbit field, H _SO). The SO interaction may be due to the spin Hall effect, the Lashva effect, other effects, or a combination thereof. In the illustrated embodiment, the SO active layer 530 may be an entire line. Accordingly, the SO active layer 530 may extend to a plurality of magnetic junctions 510. In other embodiments, the SO active layer 530 may simply be within the region of the magnetic junction 510. This is indicated by a dotted line on layer 530. In still other embodiments, the SO active layer may be between the magnetic junction 510 (hence the reference layer 512) and the word line carrying the in-plane current J _SO . Accordingly, the SO active layer 530 may include materials similar to those of the SO active layers 122, 230, 330, 430, 430a described above.

In some embodiments, an optional spin diffusion layer (not shown in FIGS. 8A and 8B) may be between the reference layer 512 and the SO active layer 530. However, in other embodiments, the spin diffusion layer may be omitted as shown in FIGS. 8A and 8B.

The magnetic junction 510 also includes reference layers 512 and 520. The reference layer 520 includes ferromagnetic layers 522 and 526 separated by a nonmagnetic layer 524. The ferromagnetic layers 522 and 526 have magnetic moments 523 and 527, respectively. Thus, the reference layer 520 may be SAF. However, the reference layer 512 is a simple layer having a magnetic moment 513. In other embodiments, the reference layer may be multiple layers. The magnetic moments 523 and 527 of the reference layer 520 are required to be fixed or pinned. However, the magnetic moment 513 of the reference layer 512 is required to be variable. In particular, the reference layer 512 is required to be switchable between the half-duplex state shown in FIG. 8A and the dual state shown in FIG. 8B. The half-duplex state is used for reading, while the dual state is used for writing.

The magnetic moment 513 of the reference layer 512 is configured to be changed using the spin orbital interaction from the SO active layer 530. The change of the magnetic moment 513 occurs during a read or write operation of the free layer 516. As shown in FIG. 8A, the in-plane current J _SO may be driven through the SO active layer 530. This creates a spin orbit field ( _HSO ). In order to switch the magnetic moment 513 of the reference layer 512 perpendicular to the plane, the SO field is also basically perpendicular to the plane. However, if the magnetic moment 513 is in-plane, the SO field is essentially in-plane. The SO field disturbs the position of the magnetic moment 513. For reading, the in-plane current J _SO1 is driven in the direction of generating the SO field for the double/half-duplex state shown in FIG. 8A. Thus, the magnetic moment 513 is switched to be in a half-duplex state. After that, the magnetic junction 510 can be read. For writing, the in-plane current J _SO2 is driven in the opposite direction, creating an SO field in the opposite direction. As a result, the magnetic moment 513a is switched to become a double state. Thereafter, the magnetic junction 510 may be written using a spin transfer torque.

In some embodiments, the magnetic moment 513 of the reference layer 512 is stable in dual and semi-duplex states in the absence of SO field/SO torque. In other words, the magnetic moment 513 is stably aligned parallel or antiparallel to the easy axis of magnetization of the free layer 516. Therefore, the reference layer 512 has anisotropy perpendicular to the plane. In such embodiments, the SO current may be reduced or eliminated during read or write. In other embodiments, the magnetic moment 513 of the reference layer 512 is not stable in the absence of SO field/SO torque. In these embodiments, the SO current remains during read and write.

In the magnetic junction 510, writing may be performed using a spin transfer torque assisted by a stray field from the reference layer 512. In some embodiments, when the magnetic layer 512 switches from a semi-duplex state to a dual state (or vice versa), the magnetic layer 512 applies a sufficiently strong stray field to the free layer 516 to magnetic The moment 517 can be moved away from the point of congestion. In some embodiments, the magnetic moment 513 may be tilted from parallel/antiparallel to the magnetic moment 517 of the free layer 516. For example, if the magnetic moments 513/513a are only stable when the SO torque is applied, the SO current may be reduced to tilt the magnetic moments 513/513a from the positions shown in FIGS. 8A and 8B. In such embodiments, the tilted magnetic moment 513/513a may also cause a stray field in the free layer, moving the magnetic moment 517 away from the point of stagnation. The magnetic junction 510 may be recorded in a similar manner to the magnetic junctions 210, 310, 410, and 410a. In these embodiments, the magnetic moment 513/513a of the reference layer 512 should be sufficiently large. For example, in these embodiments, the average saturation magnetization of the reference layer 512 may be 700-1200 emu/cm ³ . However, in other embodiments, the magnetic junction 510 may be written in a different way. For example, the magnetic junction 510 may be recorded using only a general spin transfer torque. In such embodiments, the magnetic moment 513/513a of the reference layer 512 may be switched first before the free layer magnetic moment 517 is switched using, for example, spin transfer torque. In these embodiments, since the static and dynamic stray field for the free layer 516 applied by the magnetic moment 513 can reduce thermal stability, reducing the magnetic moment of the reference layer 512 desirable. Therefore, in some embodiments, the average saturation magnetization may be between 0 and 500 emu/cm ³ . In some embodiments, to achieve this low magnetic moment, the reference layer 512 may be formed of SAF in which the magnetic moments of the two magnetic layers are completely or partially compensated. Furthermore, in some embodiments, to reduce the static stray field for the free layer 516 exerted by the magnetic moment 513, the magnetic moment 513 is It may be formed to be perpendicular to the easy magnetization axis of the magnetic moment 517. For example, if the magnetic moment 517 is perpendicular to the plane as shown in FIGS. 8A and 8B, the magnetic moment 513 of the reference layer 512 may be formed in the plane (not shown). If the magnetic moment 517 is in the plane, the magnetic moment 513 may be formed perpendicular to the plane, or may be formed in the plane with an easy magnetization axis perpendicular to the easy magnetization axis of the free layer 516 (not shown). ).

In embodiments where SO torque is used to tilt the magnetic moment 513/513a from being parallel/antiparallel to the magnetic moment 517 of the free layer 516, the magnetic junction 510 You can enjoy the benefits of (210, 310, 410 and/or 410a). For example, the magnetic junction 510 may undergo faster switching. Magnetic junction 510 may also be in a double or semi-duplex state as desired. Since the double magnetic junction 510 is used in the dual state for writing, the spin transfer torque for the free layer 516 may be higher. Thereafter, a lower spin transfer switching current J _STT may be driven through the magnetic junction 510. Since the reference layers 512 and 520 are in a half-duplex state for reading, the magnetoresistance of the magnetic junction 510 may be higher. Thus, a higher signal can be obtained. Furthermore, in the half-duplex state, the read error rate can be significantly reduced due to the reduced STT amplitude at a given current density. Therefore, the sensing margin can be increased, and the requirements for thermal stability of the cells can be relaxed. Although the switching between the double and semi-duplex states is shown for magnetic junction 510 with magnetic moments perpendicular to the plane, this switching is in different arrangements (including but not limited to magnetic moments in the plane). Can also be achieved. Therefore, the performance of the magnetic junction can be improved.

9 shows another exemplary embodiment of a magnetic memory 550 using a magnetic junction 562 that uses spin orbit interaction to change the magnetic moment of the reference layer. 9 is not a ratio of the actual size, and is for better understanding. The magnetic memory 550 is similar to the magnetic memories 200, 300, 400, 400A and/or 500. Accordingly, the magnetic memory 550 includes a magnetic junction 562 and an SO active layer 572 similar to the magnetic junctions 210, 310, 410, 410a, 510 and SO active layers 230, 330, 430, 430a, 530, respectively. Include. Thus, the magnetic junction 562 is a double junction. Therefore, the structure and function of the components 562 and 572 are similar to those of the previously described configurations 210/310/410/410a/510 and 230/330/430/430a/530, respectively. As an example, the magnetic layers may have an easy axis of magnetization perpendicular to or within a plane. The reference layers may also have magnetic moments in-plane or perpendicular to the plane. As described above, the magnetic junction 562 may be switched using a spin trajectory interaction for controlling a moment of the reference layer as well as a spin transfer torque for switching the free layer.

In addition to the magnetic junction 562, the magnetic memory 550 includes a selection device 564 corresponding to each magnetic junction 562 of the storage cell 560. In the illustrated embodiment, the memory cell includes a magnetic junction 562 and a selection device 564. The selection device 564 is a transistor and may be coupled to a bit line. In the illustrated embodiment, the magnetic memory 550 may also include an optional spin diffusion insert layer 566. The selective spin diffusion layer 566 in some embodiments may be a metal. However, in other embodiments, this layer may be a thin insulating material, such as crystalline MgO or other oxide, or another insulating layer. The resistance-area (RA) of this layer should be small. For example, it is less than 2 Ohm-μm ² . In other embodiments, the selective spin diffusion layer 566 may be a multilayer comprising two or more layers of different materials. The selective spin diffusion intercalation layer 566 may be used to reduce the contribution of the spin orbital field similar to the Lashva effect or/and to enhance the contribution of the spin orbital field similar to the spin Hall effect. Also, the optional spin diffusion intercalation layer 566 can be used to provide an improved seed layer for the reference layer.

Although only one magnetic junction 562 is shown in FIG. 9, the SO active layer 572 may extend to a plurality of magnetic junctions. Thus, SO active layer 572 can also function as word line 570. Further, the SO active layer 572 is shown to have a substantially constant thickness (dimension in the z direction) and width (dimension in the y direction). In some embodiments, the thickness and/or width of the SO active layer is reduced at least below the magnetic junction 562. In these embodiments, the spin orbit current density is increased within the region of the magnetic junction 562. Thus, switching using spin trajectory interaction can be improved. In some embodiments, SO active layer 572 may be part of a transistor including an optional source 574 and an optional drain 576. However, in other embodiments, such structures may be omitted.

The magnetic memory 550 may share the advantages of the magnetic memories 200, 300, 400, 400A and/or 500. Since the spin trajectory torque is used to switch the magnetic moment of the free layer, the performance of the magnetic memory 550 can be improved. Since the SO current for the SO active layer 572 is in-plane, the current density J _SO may be large. In addition, magnetic junction 562 may be a double magnetic junction. Thus, greater magnetoresistance and/or higher spin transfer torque can be achieved. Accordingly, the performance of the magnetic memory 550 may be improved.

10 shows another exemplary embodiment of a magnetic memory 550A using a magnetic junction 562a that utilizes spin orbit interaction to change the magnetic moment of the reference layer. 10 is not a ratio of the actual size, and is for better understanding. The magnetic memory 550A is similar to the magnetic memories 200, 300, 400, 400A and/or 500. Accordingly, the magnetic memory 550A includes a magnetic junction 562a and an SO active layer 572a similar to the magnetic junction 210/310/410/410a/510 and the SO active layers 230/330/430/430a/530, respectively. Include. Therefore, the structure and function of the components 562a and 572a are similar to those of the previously described configurations 210/310/410/410a/510 and 230/330/430/430a/530, respectively. For example, the magnetic layers may have magnetic moment(s) perpendicular to or in a plane. The reference layer closest to the SO active layer 572a may also have a magnetic moment that may change due to the spin orbit field. As described above, the magnetic junction 562a is switched using the spin trajectory interaction and the spin transfer torque, so that the moment of the reference layer can be changed and the free layer can be written. Magnetic memory 550A may also include a selective spin diffusion insert layer 566a similar to selective spin diffusion insert layer 566. For simplicity of explanation, the SO field is shown in the y direction. However, the SO field may be in another direction (including, but not limited to, a direction perpendicular to the plane, for example, a positive or negative z direction).

Magnetic memory 550A also includes word lines 570a. The word line 570a extends to a plurality of magnetic junctions 562a, and thus, extends to a plurality of memory cells. The SO active layer 572a is electrically coupled to the word line, but is limited to the area of the single magnetic junction 562a. Thus, in the illustrated embodiment, each SO active layer 572a corresponds to a magnetic junction 562a. In the illustrated embodiment, the SO active layer 572a extends to the word line 570a. However, in other embodiments, the upper end of the SO active layer 572a may be at a different position including substantially the same height as the upper end of the word line 570a, but is not limited thereto. In the illustrated embodiment, the lower end of the active layer 572a is within the word line 570a. Accordingly, the SO active layer 572a may be disposed in a recess in the word line 570a. However, in other embodiments, the lower end of the SO active layer 572a may be at a different location. Instead, the SO active layer 572a may have a thickness equal to or less than the thickness of the word line 570a, and may be in an aperture of the word line. In such embodiments, the current density through the SO active layer 572a may be greater than that in the surrounding word line 570a. SO active layer 572a is also shown extending to the edges of magnetic junction 562a. However, in other embodiments, the SO active layer 572a may extend farther than the magnetic junction 562a in the x-y plane.

The magnetic memory 550A may share the advantages of the magnetic memories 200, 300, 400, 400A, 500, and 550. Since the spin orbit torque is used to change the magnetic moment of the reference layer closest to the SO active layer 572a, the performance of the magnetic memory 550A can be improved. Since the SO current to the SO active layer 572a is in the plane, the current density J _SO may be large. In addition, the magnetic junction 562a may be a double junction with improved spin transfer torque and/or magnetoresistance. Accordingly, the performance of the magnetic memory 550A can be improved.

11 shows another exemplary embodiment of a magnetic memory 550B using a magnetic junction 562b utilizing spin orbit interaction. 11 is not a ratio of the actual size, and is for better understanding. The magnetic memory 550B is similar to the magnetic memories 200, 300, 400, 400A, 500, 550 and/or 550A. Accordingly, the magnetic memory 550B includes a magnetic junction 562b and an SO active layer 572b similar to the magnetic junction 210/310/410/410a/510 and the SO active layer 230/330/430/430a/530, respectively. Include. Therefore, the structure and function of the components 562b and 572b are similar to those of the previously described configurations 210/310/410/410a/510 and 230/330/430/430a/530, respectively. For example, the magnetic layers may have magnetic moments within or perpendicular to the plane. The reference layer closest to the SO active layer 572b may also have a magnetic moment that is changed by SO torque. Magnetic memory 550B may also include an optional spin diffusion insert layer 566b.

Magnetic memory 550B also includes a word line 570b similar to word line 570a. The word line 570b extends to a plurality of magnetic junctions 562b, and thus, to a plurality of memory cells. The SO active layer 572b is electrically coupled to the word line, but is limited to the area of a single magnetic junction 562b. Thus, in the illustrated embodiment, each SO active layer 572b corresponds to a magnetic junction 562b. In the illustrated embodiment, the SO active layer 572b extends above and below the word line 570b. In the illustrated embodiment, the SO active layer 572b may be in an aperture of the word line 570b. However, in other embodiments, the top and/or bottom of the SO active layer 572b may be at different locations. SO active layer 572b is also shown extending to the edges of magnetic junction 562b. However, in other embodiments, the SO active layer 572b may extend farther than the magnetic junction 562b in the x-y plane.

The magnetic memory 550B may share the advantages of the magnetic memories 200, 300, 400, 400A, 500, 550 and/or 550A. Since the spin orbit torque is used to change the magnetic moment of the reference layer closest to the SO active layer 572b, the performance of the magnetic memory 550B can be improved. Since the SO current to the SO active layer 572b is in-plane, the current density J _SO may be large. In addition, the magnetic junction 562b may be a double junction with improved spin transfer torque and/or magnetoresistance. Accordingly, the performance of the magnetic memory 550B can be improved.

12 illustrates another exemplary embodiment of a magnetic memory 550C using a magnetic junction 562c having a magnetic moment of a reference layer that is switched using spin orbit interaction. 12 is not a ratio of the actual size, it is for better understanding. The magnetic memory 550C is similar to the magnetic memories 200, 300, 400, 400A, 500, 550, 550A and/or 550B. Accordingly, the magnetic memory 550C includes a magnetic junction 562c and an SO active layer 572c similar to the magnetic junction 210/310/410/410a/510 and the SO active layers 230/330/430/430a/530, respectively. Include. The structure and function of the components 562c and 572c are similar to those of the previously described configurations 210/310/410/410a/510 and 230/330/430/430a/530, respectively. For example, the magnetic layers may have magnetic moments within or perpendicular to the plane. The reference layer closest to the SO active layer 572c may also have a magnetic moment that is changed by SO torque. Magnetic memory 550C may also include an optional spin diffusion insert layer 566c.

Magnetic memory 550C also includes word lines 570c similar to word lines 230/330/430/430a/530. The word line 570c extends to a plurality of magnetic junctions 562c, and thus, extends to a plurality of memory cells. The SO active layer 572c is electrically coupled to the word line, but is limited to the area of a single magnetic junction 562c. In the illustrated embodiment, the SO active layer 572c is adjacent to the magnetic junction 562c. The SO active layer 572c is not underneath the magnetic junction 562c. Instead, another portion of word line 570c is under magnetic junction 562c. The SO active layer 572c may be somewhat separated from the magnetic junction 562c. This separation is not very large and is generally smaller than the width of the MTJ. However, in some embodiments, it may be larger than that, and may be up to 100 nm.

The magnetic memory 550C may share the advantages of the magnetic memories 200, 300, 400, 400A, and 500. Since the spin orbit torque is used to change the magnetic moment of the reference layer, the performance of the magnetic memory 550C can be improved. Since the SO current for the SO active layer 572c is in-plane, the current density J _SO may be large. In addition, the magnetic junction 562c may be a double magnetic junction capable of improving spin transfer torque and/or increasing magnetoresistance. Accordingly, the performance of the magnetic memory 550C can be improved.

13 shows an exemplary embodiment of a magnetic memory 600 using a double magnetic junction 260 having a reference layer magnetic moment that can be varied using features that mimic spin orbit interactions. 13 is not a ratio of the actual size, and is for better understanding. The magnetic memory 600 is similar to the magnetic memories 200, 300, 400, 400A and/or 500. As a result, similar components have similar reference numbers. Accordingly, the magnetic memory 600 includes a magnetic junction 610 and a structure 620 similar to the magnetic junction 210/310/410/410a/510 and the SO active layer 230/330/430/430a/530, respectively. do. Accordingly, the structures and functions of the components 610 and 620 are similar to those of the components 210/310/410/410a/510 and 230/330/430/430a/530 described above, respectively. The magnetic memory 600 may also include a selective spin diffusion insert layer 614 similar to the selective spin diffusion insert layer 566.

In the context of this specification, since a structure 620 similar to the SO active layer is used, the magnetic memory 600 is configured to use the spin trajectory interaction in switching the magnetic junctions 610. More specifically, the outer structure 620 of the magnetic junctions 610 provides a spin polarized in-plane current that is used to change the magnetic moment of the magnetic junction 610. More clearly, the magnetic moment closest to the word line 624 can be varied. Thus, the switching mechanism of memory 600 mimics the spin trajectory interaction.

In the magnetic memory 600, a structure 620 similar to an SO active layer is formed from a combination of a highly conductive word line 624 and at least one spin polarized current injector 622. In the embodiment shown in Fig. 13, only one spin polarization current injector 622 is used. However, in other embodiments, a plurality of spin polarization current injectors may be used. As an example, two injectors with opposite spin polarizations may be used. Otherwise, one spin polarization current injector 622 could be used. The spin polarization current injector 622 polarizes the spindle of charge carriers with respect to the current driven through the spin polarization current injector 622. For example, the spin polarization current injector 622 may be a magnetic layer. In addition, one spin polarization injector 622 is required to provide a polarized spindle for a plurality of magnetic junctions 610. Accordingly, the highly conductive word line 624 is at least one conductive layer having a long spin diffusion length. As an example, in some embodiments, the spin diffusion length is at least 100 nanometers. In some such embodiments, the spin diffusion length is at least 1 micrometer. For example, in one embodiment, the highly conductive word line 624 may be a graphene line. The long spin diffusion length is such that the spin polarized charge carriers from the injector 622 can reach at least one magnetic junction 610 across the word line 624 without undergoing severe scattering that destroys their spin information. Required.

Since the current is polarized by the injector 622 and maintains the spin information as it flows through the highly conductive word line 624, the polarized current is the method of spin polarization for the spin Hall and Rashbar effects described above. Behave in a similar way to Thus, the combination of the injector 622 and the highly conductive word line 624 functions in a similar manner to the SO active layers 230/330/430/430a/530. In other words, the spin polarization current can provide a field and torque similar to the spin orbit field and the spin orbit torque.

The magnetic memory 600 shares the advantages of the magnetic memories 200, 300, 400, 400A, 500. Since the spin trajectory torque is used to change the magnetic moment of the reference layer, the performance of the memory 600 can be improved. In addition, the magnetic junction 610 may be a double magnetic junction capable of improving spin transfer torque and/or increasing magnetoresistance. Accordingly, the performance of the memory 600 may be improved. Thus, the memories 550, 550A, 550B, 550C, and 600 illustrate various arrangements of SO active layers 572, 572a, 572b, 572c, and 620, respectively. Using one or more of these arrangements, the performance of the magnetic memory can be improved.

14 shows an exemplary embodiment of a magnetic memory 700 using magnetic junctions 710 that are switched primarily using spin orbit interaction. 14 is not a ratio of the actual size, and is for better understanding. The magnetic memory 700 is similar to the magnetic memories 200, 300, 400, 400A, 500). As a result, similar components have similar reference numerals. Accordingly, the magnetic memory 700 includes magnetic junctions, selection elements, spin diffusion insertion layers, and magnetic junctions 710 similar to SO active layers, selection elements 718, and a selective spin diffusion insertion layer ( 730a) and an SO active layer 720a. Although not shown, the magnetic junction 710 includes a data storage layer/free layer, nonmagnetic spacer layers, and reference layers similar to those described above. Accordingly, the structures and functions of the components 710 and 720a are similar to those described above. Although the SO active layer 720a is shown as a word line, in other embodiments, other arrangements may be used. Each of the magnetic layers of the magnetic junctions 710 may have an easy axis of magnetization perpendicular to or in a plane. Although both H _SO1 and H _SO2 are shown to be in the xy plane, in other embodiments, the fields H _SO1 , and H _SO2 may be in other directions including a direction perpendicular to the plane.

In the memory 700, spin trajectory interaction switching is used to change the magnetic moment of the reference layer assisted by resistance adjustment. In the illustrated embodiment, the resistance of resistor 735 is regulated through resistance select transistors 736. Thus, the resistors 735 are variable resistance elements. The resistance R1 is a relatively high resistance compared to the resistance of the SO active layer 720a. Thus, the current in the SO active layer 720a does not change direction to flow through the resistor 735 (ie, it is not classified). As such, the spin trajectory torque generated by H _SO1 is still sufficient to switch the reference layer magnetic moment of the magnetic element 710. For the resistor R2, which is lower than the SO active layer 720a, the spin orbit current J _SO flows through the resistor R2 (ie, it is shunted). The accumulation of charge carriers on the top surface of layer 720a is reduced. The spin orbit field ( _HSO2 ) is also reduced. The spin orbital field generated over R2 is not sufficient to change the reference layer magnetic moment of the magnetic junction 710. Accordingly, the magnetic memory 700 selects the magnetic junction 710 in which the reference layer magnetic moment is to be changed by utilizing the resistance changes to the magnetic junction. Thus, the magnetic memory 700 may be considered using resistance variations to select a double magnetic junction that can be written and/or read.

Magnetic memory 700 shares the advantages of magnetic memories 200, 300, 400, 400A, 500. Since the spin trajectory torque is used to change the magnetic moment of the reference layer, the performance of the memory 700 can be improved. In addition, the desired magnetic junction 710 to be programmed can be selected using resistance changes. Accordingly, the performance of the memory 700 can be improved.

FIG. 15 shows an exemplary embodiment of a magnetic memory 700A using magnetic junctions 710a that use spin orbit interaction to change the magnetic moment of the reference layer. 15 is not a ratio of the actual size, and is for better understanding. The magnetic memory 700A is similar to the magnetic memories 200, 300, 400, 400A, 500). As a result, similar components have similar reference numerals. Accordingly, the magnetic memory 700A includes magnetic junctions, selection elements, spin diffusion insertion layers, and magnetic junctions 710a similar to SO active layers, selection elements 718a, and a selective spin diffusion insertion layer ( 730a) and an SO active layer 720a. Although not shown, the magnetic junction 710a includes a data storage layer/free layer, nonmagnetic spacer layers, and reference layers similar to those described above. Accordingly, the structures and functions of the components 710a and 720a are similar to those described above. Although the SO active layer 720a is shown as a word line, in other embodiments, other arrangements may be used. Each of the magnetic layers of the magnetic junctions 710a may have an easy axis of magnetization perpendicular to or in a plane. Although both H _SO1 and H _SO2 are shown to be in the xy plane, in other embodiments, the fields H _SO1 , and H _SO2 may be in other directions including a direction perpendicular to the plane.

In memory 700A, spin trajectory interaction switching is assisted by heating the SO active layer 720a using heaters 740. The heaters 740 are controlled through heater selection transistors 742. When a heating element such as heater 1 is in the air, the SO active layer 720a may generate a desired spin orbit field H _so1 for switching the magnetic moment of the reference layer of the magnetic junction 710a. However, heater 2 can be driven. The SO active layer 720a is heated to increase the relaxation of the spin accumulations induced by SO, and thus reduce the spin orbit field H _so2 . The generated spin orbit field is not sufficient to change the magnetic moment of the reference layer of the magnetic junction 710a. Accordingly, the magnetic memory 700A utilizes heating of the SO active layer 720a to select the magnetic junction 710a in which the reference layer magnetic moment is changed. The magnetic junction 710a may have a reference layer magnetic moment that is changed using heating.

It is also noted that the switching of the magnetic moments of the free layer and/or the reference layer of the magnetic junction 710a can be improved through heating. The heating current driven through the heater 740 and/or the magnetic junctions 710a may heat the magnetic junctions 710a. As a result, the magnetic moments of the free layer and/or the reference layer become more thermally unstable and thus can be switched more easily. Thus, the switching and changes in the reference layer magnetic moment can be assisted by heating.

Magnetic memory 700A shares the advantages of magnetic memories 200, 300, 400, 400A, 500. Since the spin trajectory torque is used to change the magnetic moment of the reference layer, the performance of the memory 700A can be improved. In addition, the desired magnetic junction 710a to be programmed can be selected using heating. Accordingly, the performance of the memory 700A can be improved.

FIG. 16 shows an exemplary embodiment of a magnetic memory 700B using magnetic junctions 710b that use spin orbit interaction to change the magnetic moment of the reference layer. 16 is not a ratio of the actual size, and is for better understanding. The magnetic memory 700B is similar to the magnetic memories 200, 300, 400, 400A, 500. As a result, similar components have similar reference numerals. Accordingly, the magnetic memory 700B includes magnetic junctions, selection elements, spin diffusion insertion layers, and magnetic junctions 710b similar to SO active layers, selection elements 718b, and a selective spin diffusion insertion layer ( 730b) and an SO active layer 720b. Although not shown, the magnetic junction 710b includes a data storage layer/free layer, nonmagnetic spacer layers, and reference layers similar to those described above. Accordingly, the structures and functions of the components 710b and 720b are similar to those described above. Although the SO active layer 720b is shown as a word line, in other embodiments, other arrangements may be used. Each of the magnetic layers of the magnetic junctions 710b may have an easy axis of magnetization perpendicular to or in a plane. Although both H _SO1 and H _SO2 are shown to be in the xy plane, in other embodiments, the fields H _SO1 , and H _SO2 may be in other directions including a direction perpendicular to the plane.

In memory 700B, the change in the spin orbital interaction in the magnetic moment is assisted by the concentration of the spin orbital current in the region of the magnetic junction 710b. In the illustrated embodiment, the thickness (in the z direction) of the SO active layer 720b is limited in the region of the magnetic junction 710b. In other embodiments, the width (in the y direction), or both the thickness and the width, may be limited such that the cross-sectional area of the SO active layer 720b is reduced in the area of the magnetic junctions 710b. As a result, the spin orbital current may be concentrated in these regions, and thus it may be possible to apply a greater spin orbital torque at a given current.

Magnetic memory 700B shares the advantages of magnetic memories 200, 300, 400, 400A, 500. Since the spin trajectory torque is used to change the magnetic moment of the reference layer, the performance of the memory 700B can be improved. In addition, the desired magnetic junction 710b to be programmed can be selected using heating. Accordingly, the performance of the memory 700B can be improved.

Note that in the magnetic memories 700, 700A, 700B, the anisotropy of the reference layer closest to the SO active layers 720, 720a, 720b may be changed, respectively. For example, the anisotropy of the reference layer may be changed by a voltage applied to the reference layer. In these embodiments, the magnetic junction 710/710a/710b may be selected for a change in the reference layer magnetic moment by applying a control voltage to the magnetic junction 710/710a/710b. In addition, this control voltage may be used for other memories including the memories 200, 300, 400, 400A, 500, 550, 550A, 550B, 550C and/or 600, but is not limited thereto.

FIG. 17 is a flow chart illustrating an exemplary embodiment of a method 800 of providing a magnetic memory having magnetic junction(s) switched using spin orbit interaction. For simplicity of description, some steps may be omitted, combined, and/or inserted. Method 800 is described in the context of magnetic memory 200. However, the method 800 includes (but is not limited to) magnetic memories 300, 400, 400A, 500, 550, 550A, 550B, 550C, 600, 700, 700A and/or 700B). Can be used to provide memories.

By step 802, an SO active layer 230 is provided. In some embodiments, step 802 includes providing a layer suitable for the spin Hall effect. In other embodiments, a layer suitable for the Rashba effect is provided. In still other embodiments, the provided SO active layer 230 may use a combination of a spin Hall effect and a Lashva effect. Other spin orbital interaction mechanisms may also be provided. Additionally, step 802 may include patterning the SO active layer. By step 804, a spin diffusion layer (not shown in the magnetic memory 200) may be selectively provided. The spin diffusion layer, if provided, is between the SO active layer 230 and the magnetic junctions 210.

By step 806, double magnetic junctions 210 are provided. In some embodiments, step 806 includes a first reference layer 212, a first nonmagnetic spacer layer 214, a free layer 216, a second nonmagnetic spacer layer 218 such as a tunneling barrier layer, and It includes providing a second reference layer 220. Thereafter, manufacturing of the magnetic memory 200 may be completed. Thus, using method 800, the advantages of one or more of the magnetic memories 200, 300, 400, 400A and/or 500 may be achieved.

18 is a flow chart illustrating an exemplary embodiment of a method 850 of programming a switched magnetic junction(s) using spin orbit interaction. Method 850 may be used in one or more of memories 200, 300, 400, 400A and/or 500. For simplicity of description, some steps may be omitted, combined, and/or inserted. Method 850 is described within the context of magnetic memory 200. However, method 850 includes (but is not limited to) magnetic memories 300, 400, 400A, 500, 550, 550A, 550B, 550C, 600, 700, 700A, and/or 700B. Can be used for memories.

By step 852, an in-plane spin trajectory write current is applied. The spin trajectory write current can be applied as a pulse. If the reference layers 212 and 220 are required to be switched between the double and half-duplex states, the magnitude and duration of the pulse will be sufficient to switch the direction of the magnetic moment (not shown) of the reference layer 212. I can. For example, such a pulse may be used in the memory 500. In other embodiments, the magnitude and duration of the pulse is sufficient to tilt the magnetic moment (not shown) of the reference layer 212, and the generated stray field is magnetic of the free layer 216 from the stagnation point. It makes it possible to disturb the moment (not shown).

By step 854, the spin transfer torque write current is driven through the magnetic junction 210. The current in step 854 can also be applied in pulses, as described above. The current pulse applied in step 854 may be required to be timed so that the free layer 216 is not at a stagnation point on or before the start of the pulse for the spin trajectory current driven in step 852. In other embodiments, the timing may be different. Thus, writing of the cells can be completed using steps 852 and 854.

In addition, by step 856, the magnetic junctions 210 to be written may be selected. For example, spin transfer torque, heating of the magnetic junction 210, controlling the resistance of the SO active layer 230, heating the SO active layer 230, a combination of these and/or other mechanisms can be used to select cells to be recorded. have. Further, step 856 may be performed substantially simultaneously with step 852. Accordingly, desired magnetic junctions 210 of the magnetic memory 200 can be programmed. The magnetic junctions 210 may be read by driving a read current passing through the magnetic junctions 210 and determining whether the magnetic junctions 210 are in a high resistance state or a low resistance state. As part of the reading, the magnetic junctions can be switched from the double state of the reference layer to the half-duplex state.

Thus, using method 850, magnetic memories 200, 300, 400, 400A and/or 500 can be programmed. Thus, the advantages of magnetic memories 200, 300, 400, 400A and/or 500 can be achieved.

A method and system for providing a double magnetic junction and a memory fabricated using such double magnetic junction has been described. Various combinations of features of the magnetic memories 200, 300, 400, 400A, 500, 550, 550A, 550B, 550C, 600, 700, 700A, and/or 700B may be combined. The method and system have been described in accordance with the illustrated exemplary embodiments, and those of ordinary skill in the art to which the present invention pertains may have variations in the embodiments, and any variations may be applied to the concept of the method and system. And it will be readily appreciated that it must be within a range. For that reason, many changes can be made by those of ordinary skill in the art to which the present invention pertains without departing from the spirit and scope of the following appended claims.

Claims (30)

A plurality of double magnetic junctions, each of the plurality of double magnetic junctions includes a first reference layer, a first nonmagnetic spacer layer, a free layer, a second nonmagnetic spacer layer, and a second reference layer, wherein the free layer is magnetic And is between the first nonmagnetic spacer layer and the second nonmagnetic spacer layer, the first nonmagnetic spacer layer is between the first reference layer and the free layer, and the second nonmagnetic spacer layer is Between the second reference layer and the free layer; And
At least one spin orbit interaction (SO) active layer adjacent to the first reference layer of each of the plurality of double magnetic junctions,
The at least one SO active layer applies SO torque to the first reference layer due to a current flowing through the at least one SO active layer in a direction perpendicular to a direction between the at least one SO active layer and the first reference layer. Composed,
The free layer is configured to be switched using a spin transfer write current driven through the double magnetic junction,
The first reference layer has a magnetic moment configured to be changed by at least the SO torque. The method of claim 1,
The free layer has a free layer magnetic moment having a plurality of stable states along an easy magnetization axis,
The magnetic moment of the first reference layer is at least inclined at an angle other than zero from the easy magnetization axis by the SO torque. The method of claim 1,
The second reference layer has an additional reference layer magnetic moment,
The magnetic moment of the first reference layer is configured to be changed to a double state by the SO torque for a write operation and to a half-double state by the SO torque for a read operation. The method of claim 3,
The free layer has a free layer magnetic moment having a plurality of stable states along an easy magnetization axis,
The magnetic moment of the first reference layer is stabilized along the axis of easy magnetization in the absence of the SO torque. The method of claim 3,
The free layer has a free layer magnetic moment having a plurality of stable states along an easy magnetization axis,
The magnetic moment of the first reference layer is stable along the axis of easy magnetization in the presence of the SO torque, and is unstable with respect to the axis of easy magnetization in the absence of the SO torque. The method of claim 1,
At least one of the first reference layer and the second reference layer is a composite layer including a plurality of ferromagnetic layers. The method of claim 1,
At least one of the first nonmagnetic spacer layer and the second nonmagnetic spacer layer is a tunneling barrier layer. The method of claim 1,
Further comprising a spin diffusion insertion layer for each of the at least one SO active layer,
The spin diffusion insertion layer is a magnetic memory between the first reference layer and the at least one SO active layer. The method of claim 1,
The SO active layer is a magnetic memory adjacent to the first reference layer. The method of claim 1,
The at least one SO active layer is a magnetic memory extending to at least two of the plurality of double magnetic junctions.
delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete delete

Images