阅读说明：本技术 包含包括β相钨的复合自旋霍尔效应层的自旋轨道矩磁阻随机存取存储器 (Spin-orbit torque magnetoresistive random access memory including composite spin hall effect layer including β phase tungsten ) 是由 Y·崔 B·约克 N·史密斯 于 2018-05-30 设计创作，主要内容包括：自旋轨道矩磁阻随机存取存储器(SOT MRAM)单元包括：磁隧道结,该磁隧道结包含具有两个双稳磁化方向的自由层、具有固定磁化方向的参考磁性层、以及位于自由层与参考层之间的隧道阻挡层；以及非磁性自旋霍尔效应层。自旋霍尔效应层可包括β相钨层和贵金属非磁性除尘层的交替叠堆。另选地或除此之外,铪层可位于非磁性自旋霍尔效应层与自由层之间。(A spin-orbit torque magnetoresistive random access memory (SOT MRAM) cell includes a magnetic tunnel junction including a free layer having two bistable magnetization directions, a reference magnetic layer having a fixed magnetization direction, and a tunnel barrier layer between the free layer and the reference layer, and a nonmagnetic spin Hall effect layer.)

1. A spin-orbit torque magnetoresistive random access memory (SOT MRAM) cell, comprising:

the magnetic tunnel junction includes a free layer having two bistable magnetization directions, a reference magnetic layer having a fixed magnetization direction, and a tunnel barrier layer between the free layer and the reference layer, and a nonmagnetic spin Hall effect layer including an alternating stack of β phase tungsten layers and a noble metal nonmagnetic dusting layer.

2. The SOT MRAM cell of claim 1, wherein:

the free layer is positioned between the tunnel barrier layer and the nonmagnetic spin Hall effect layer; and the non-magnetic dusting layer comprises a ruthenium, platinum, iridium, palladium, rhodium, rhenium, or osmium layer.

3. The SOT MRAM cell of claim 2, wherein:

the tunnel barrier layer comprises a MgO layer;

each of the β -phase tungsten layers has a thickness in the range of 0.2nm to 1.2nm, and

each of the nonmagnetic dusting layers has a thickness in the range of 0.1nm to 0.3 nm.

4. The SOT MRAM cell of claim 2, wherein the alternating stack comprises N periodic repetitions of a pair of the β phase tungsten layers and the nonmagnetic dusting layer, and wherein N is an integer in the range of 5 to 30.

5. The SOT MRAM cell of claim 4, wherein a ratio of a thickness of the β phase tungsten layer to a thickness of the nonmagnetic dusting layer is in a range of 3 to 6.

6. The SOT MRAM cell of claim 1, further comprising a first hafnium layer located between the nonmagnetic spin hall effect layer and the free layer.

7. The SOT MRAM cell of claim 6, further comprising a nonmagnetic spacer layer located between the first hafnium layer and the nonmagnetic spin Hall effect layer and comprising a material selected from ruthenium and platinum, wherein the nonmagnetic spacer layer contacts the first hafnium layer.

8. The SOT MRAM cell of claim 6, further comprising a hafnium oxide layer located between the first hafnium layer and the free layer, and a second hafnium layer located between the hafnium oxide layer and the free layer.

9. The SOT MRAM cell of claim 1, wherein:

the SOT MRAM cell is a three terminal device including a first terminal electrically connected to a hard magnetic layer located on a reference layer side of the magnetic tunnel junction, a second terminal directly or indirectly electrically connected to a first end of the nonmagnetic spin Hall effect layer, and a third terminal directly or indirectly electrically connected to a second end of the nonmagnetic spin Hall effect layer;

the SOT MRAM cell is configured to generate a read current flowing between the first terminal and the third terminal through a tunnel junction during a sensing operation; and is

The SOT MRAM cell is configured to generate a write current flowing between the second terminal and the third terminal during a programming operation such that the write current does not flow through the tunnel junction, and such that the read current and the write current flow in different directions during respective sensing and programming operations.

10. The SOT MRAM cell of claim 9, further comprising:

a ferromagnetic bias layer configured to provide a spin-orbit torque via an abnormal Hall effect and a magnetic bias field on the free layer to achieve deterministic switching; and

an Antiferromagnetic (AFM) layer positioned adjacent to the ferromagnetic bias layer and configured to fix a magnetization direction of the ferromagnetic bias layer in a predetermined direction,

wherein the second terminal is electrically connected to a first end of the AFM layer and the third terminal is electrically connected to a second end of the AFM layer such that the second terminal and the third terminal are indirectly electrically connected to respective first and second ends of the nonmagnetic spin Hall effect layer.

11. The SOT MRAM cell of claim 9, further comprising at least one permanent magnet positioned adjacent to the magnetic tunnel junction, wherein the second and third terminals are directly electrically connected to respective first and second ends of the nonmagnetic spin hall effect layer.

12. A spin-orbit torque magnetoresistive random access memory (SOT MRAM) cell, comprising:

a magnetic tunnel junction comprising a free layer having two bistable magnetization directions, a reference magnetic layer having a fixed magnetization direction, and a tunnel barrier layer between the free layer and the reference layer;

a nonmagnetic spin hall effect layer; and a first hafnium layer between the nonmagnetic spin hall effect layer and the free layer.

13. The SOT MRAM cell of claim 12, wherein:

the free layer is positioned between the tunnel barrier layer and the nonmagnetic spin Hall effect layer;

the nonmagnetic spin hall effect layer comprises at least one β phase tungsten layer, and the tunnel barrier layer comprises a MgO layer.

14. The SOT MRAM cell of claim 12, wherein:

the nonmagnetic spin hall effect layer comprises an alternating stack of β phase tungsten layers and ruthenium, platinum, iridium, palladium, rhodium, rhenium, or osmium nonmagnetic dusting layers;

each of the β -phase tungsten layers has a thickness in a range of 0.2nm to 1.2nm, and each of the nonmagnetic dusting layers has a thickness in a range of 0.1nm to 0.3 nm.

15. The SOT MRAM cell of claim 14, wherein:

the alternating stack includes a pair of N periodic repetitions of the β -phase tungsten layers and the nonmagnetic dusting layer, and wherein N is an integer in the range of 5 to 30, and a ratio of a thickness of the β -phase tungsten layers to a thickness of the nonmagnetic dusting layer is in the range of 3 to 6.

16. The SOT MRAM cell of claim 12, further comprising a nonmagnetic spacer layer located between the first hafnium layer and the nonmagnetic spin hall effect layer and comprising a material selected from ruthenium and platinum, wherein the nonmagnetic spacer layer contacts the first hafnium layer.

17. The SOT MRAM cell of claim 16, further comprising a hafnium oxide layer located between the first hafnium layer and the free layer, and a second hafnium layer located between the hafnium oxide layer and the free layer.

18. The SOT MRAM cell of claim 12, wherein:

the SOT MRAM cell is a three terminal device including a first terminal electrically connected to a hard magnetic layer located on a reference layer side of the magnetic tunnel junction, a second terminal directly or indirectly electrically connected to a first end of the nonmagnetic spin Hall effect layer, and a third terminal directly or indirectly electrically connected to a second end of the nonmagnetic spin Hall effect layer;

the SOT MRAM cell is configured to generate a read current flowing between the first terminal and the third terminal through a tunnel junction during a sensing operation; and is

The SOT MRAM cell is configured to generate a write current flowing between the second terminal and the third terminal during a programming operation such that the write current does not flow through the tunnel junction, and such that the read current and the write current flow in different directions during respective sensing and programming operations.

19. The SOT MRAM cell of claim 18, further comprising:

a ferromagnetic bias layer configured to provide a spin-orbit torque via an abnormal Hall effect and a magnetic bias field on the free layer to achieve deterministic switching; and an Antiferromagnetic (AFM) layer positioned adjacent to the ferromagnetic bias layer and configured to fix a magnetization direction of the ferromagnetic bias layer in a predetermined direction,

wherein the second terminal is electrically connected to a first end of the AFM layer and the third terminal is electrically connected to a second end of the AFM layer such that the second terminal and the third terminal are indirectly electrically connected to respective first and second ends of the nonmagnetic spin Hall effect layer.

20. The SOT MRAM cell of claim 18, further comprising at least one permanent magnet positioned adjacent to the magnetic tunnel junction, wherein the second and third terminals are directly electrically connected to respective first and second ends of the nonmagnetic spin hall effect layer.

Technical Field

The present disclosure relates generally to the field of magnetic memory elements, and in particular, to spin-orbit torque (SOT) magnetoresistive memory cells.

Background

Magnetoresistive Random Access Memory (MRAM) is a non-volatile random access memory technology. Unlike conventional Random Access Memory (RAM), data in MRAM devices is not stored as electrical charge or current. Instead, data is stored by magnetic storage elements. MRAM devices include cells or elements having a magnetic hard layer (i.e., a "reference" layer) and a magnetic soft layer (i.e., a "free" layer). Writing to MRAM is performed by passing a current through current leads formed on either side of each memory element in order to generate a local induced magnetic field that sets the direction of magnetization of the soft layer. A number of problems arise when scaling these devices to high densities. In particular, the current required to generate sufficient field to switch the free layer becomes prohibitively large and interference with adjacent cells or elements can occur during writing, which in turn can cause adjacent cells to be written incorrectly.

Spin Transfer Torque (STT) MRAM devices are similar to conventional MRAM devices except that the write current path passes through the magnetic layer of each memory element. The free layer is set by the spin transfer torque from the spin-polarized current through the reference magnetic layer. Spin-orbit torque (SOT) MRAM devices are similar to spin-transfer torque (STT) MRAM devices, except that the read and write paths are independent. SOT MRAM devices may have better endurance because the write current does not pass through the thin tunnel barrier layer. The most recent design of SOTMRAM devices is disclosed, for example, in U.S. patent application publication No. 2017/0125078.

Disclosure of Invention

According to one aspect of the present disclosure, a spin-orbit-torque magnetoresistive random access memory (SOT MRAM) cell includes a magnetic tunnel junction including a free layer having two bi-stable magnetization directions, a reference magnetic layer having a fixed magnetization direction, and a tunnel barrier layer between the free layer and the reference layer, and a nonmagnetic spin Hall effect layer.

Drawings

FIG. 1A illustrates a first exemplary structure including an SOT-MRAM device according to an embodiment of the disclosure.

FIG. 1B illustrates an alternative embodiment of a first exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 2 illustrates a second exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 3A illustrates a third exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 3B illustrates an alternative embodiment of a third exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 4 illustrates a fourth exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 5A illustrates a fifth exemplary structure including an SOT-MRAM device according to an embodiment of the disclosure.

FIG. 5B illustrates an alternative embodiment of a fifth exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

FIG. 6 illustrates a sixth exemplary structure including a SOT-MRAM device according to an embodiment of the disclosure.

Fig. 7A, 7B, and 7C illustrate seventh, eighth, and ninth exemplary structures, respectively, including an SOT-MRAM device, according to embodiments of the disclosure.

Fig. 8A, 8B, and 8C illustrate alternative embodiments of seventh, eighth, and ninth exemplary structures, respectively, including an SOT-MRAM device, according to embodiments of the present disclosure.

Figure 9A shows X-ray diffraction data from a 2 theta scan of three alternating stacks of tungsten layers and intervening layers, according to an embodiment of the present disclosure.

Fig. 9B is a graph of resistivity- β phase tungsten layer thickness showing resistivity measurement data before and after annealing for an alternating stack of tungsten layers and dusting layers according to an embodiment of the present disclosure.

FIG. 10 is a graph illustrating the effective magnetization of the free layer as a function of the thickness of the intervening hafnium layer, according to one embodiment of the present disclosure.

Fig. 11, 12, and 13 are graphs showing dependence of γ Δ H/2 as a function of frequency during a ferroresonance measurement according to an embodiment of the present disclosure.

Detailed Description

As discussed above, the present disclosure relates to spin-orbit torque (SOT) magnetoresistive memory cells, random access memory devices employing the same, and methods of fabricating the same, various aspects of which are described below. Embodiments of the present disclosure may be employed to form various semiconductor devices, such as three-dimensional memory array devices including MRAM devices. The figures are not drawn to scale. Where a single instance of an element is illustrated, multiple instances of the element may be repeated unless explicitly described or otherwise clearly indicated to be absent repetition of the element. Ordinal numbers such as "first," "second," and "third" are used merely to identify similar elements, and different ordinal numbers may be employed throughout the specification and claims of the present disclosure.

As used herein, "layer" refers to a portion of a material that includes a region having a thickness. A layer may extend over the entirety of an underlying or overlying structure, or may have a range that is less than the range of an underlying or overlying structure. In addition, a layer may be a region of uniform or non-uniform continuous structure having a thickness less than the thickness of the continuous structure. For example, a layer may be positioned between the top and bottom surfaces of the continuous structure or between any pair of horizontal planes at the top and bottom surfaces of the continuous structure. The layers may extend horizontally, vertically, and/or along a tapered surface. The substrate may be a layer, may include one or more layers therein, and/or may have one or more layers thereon, above, and/or below.

As used herein, "stack of layers" refers to a stack of layers. As used herein, "line" or "line structure" refers to a layer having a predominant direction of extension, i.e., having the direction in which the layer extends the most.

Referring to fig. 1A and 1B, an embodiment of a first exemplary spin-orbit torque (SOT) magnetoresistive memory cell 100 is shown, according to an embodiment of the disclosure. The first exemplary SOT magnetoresistive memory cell 100 includes a stacked stack of, from bottom to top, an optional Antiferromagnetic (AFM) layer 202, an optional ferromagnetic bias layer 206, a nonmagnetic spin hall effect layer (also referred to as SOT layer) 204, a free layer 104, a tunnel barrier layer 106 (e.g., MgO layer), a reference layer 108, an antiparallel coupling layer 110, and a hard magnetic layer 112. The SOT magnetoresistive memory cell is a three terminal device comprising a first terminal 1 electrically connected to the hard magnetic layer 112, a second terminal 2 electrically connected to one end of the AFM layer 202, and a third terminal 3 electrically connected to the other end of the AFM layer 202.

The magnetization of the hard magnetic layer 112 maintains a fixed orientation perpendicular to the plane or parallel to the plane. In one embodiment, the magnetization of the hard magnetic layer 112 may be in an "up" direction or a "down" direction due to the strong Perpendicular Magnetic Anisotropy (PMA) in the first configuration shown in FIG. 1A. In another embodiment, the magnetization of the hard magnetic layer 112 may be in a horizontal direction (i.e., parallel to the interface between the hard magnetic layer 112 and the antiparallel coupling layer 110), due to the strong in-plane magnetic anisotropy (IMA), as shown in the second configuration shown in FIG. 1B. The magnetization of the reference layer 108 couples to the magnetization of the hard magnetic layer 112 via the strong anti-parallel coupling (AP) provided by the anti-parallel coupling layer 110. Thus, the magnetization of the reference layer 108 is antiparallel to the magnetization of the hard magnetic layer 112. For example, if the hard magnetic layer 112 has an "upward" magnetization, the reference layer 108 may have a "downward" magnetization, and vice versa. If the hard magnetic layer 112 has a horizontal magnetization, the reference layer 108 may have a horizontal magnetization in the opposite direction. The antiparallel coupling layer 110 can comprise, for example, ruthenium.

The free layer 104 (e.g., a CoFeB and/or CoFe layer) has primarily two bi-stable magnetization directions. The free layer 104 may have a vertical easy axis due to a perpendicular magnetic anisotropy of sufficient strength or may have a horizontal easy axis due to an in-plane magnetic anisotropy of sufficient strength. The magnetization of the free layer 104 is nominally maintained along an easy axis (such as the z-axis or x-axis), but its direction may be responsive to a write current I passing through it_WRThe Spin Orbit Torque (SOT) produced by the nonmagnetic spin Hall effect layer 204 switches between two states: parallel or anti-parallel to the reference layer 108. The spin hall effect is a transport phenomenon in a nonmagnetic conductor, which is constituted by generating a spin current in a direction perpendicular to a plane defined by a current direction and a spin polarization direction. In an exemplary case, if the magnetization of the reference layer 108 is in the "up" direction, the magnetization of the free layer 104 is directed in the up z-direction in the parallel state and in the down z-direction in the anti-parallel state. If the magnetization of the reference layer 108 is in the positive x-axis direction, the magnetization of the free layer 104 points in the positive x-axis direction in the parallel state and points in the negative x-axis direction in the anti-parallel state. The two magnetization states of the free layer 104 may correspond to logic states 0 and 1. The stacked stack of the reference layer 108, the tunnel barrier layer 106, and the free layer 104 constitutes a magnetic tunnel junction 101. Various additional non-magnetic layers such as tantalum, tungsten and/or ruthenium may be formed as part of the magnetic tunnel junction.

The non-magnetic spin Hall effect layer 204 is located below the free layer 104 and may be in contact with the free layer 104 according to an embodiment of the present disclosure, the spin Hall effect layer 204 includes a stacked stack including a plurality of β phase tungsten layers such that a significant volume fraction (such as more than 50%, more than 80%, and/or more than 90%) of the β phaseTungsten is provided for the write current I_WRThe conductance of (2).

The ferromagnetic bias layer 206 (if present) is configured to provide a magnetic bias field on the free layer 104 to achieve deterministic switching. In the first configuration of FIG. 1A, the ferromagnetic bias layer 206 is configured to provide an in-plane magnetic bias field for the free layer 104 via its stray field (primarily along the x-axis) in order to achieve deterministic switching of the magnetization of the free layer 104. This can be achieved by partially milling the ferromagnetic bias layer 206 to a depth d relative to the spin hall effect layer 204. The milling depth d provides an additional variable for adjusting the strength of the stray magnetic field on the free layer 104. The milling depth can be anywhere between 0nm and tnm, where t is the thickness of the ferromagnetic bias layer 206. Because the ferromagnetic bias layer 206 is configured to generate a magnetic bias field, it eliminates the need for permanent magnets (described below with reference to fig. 7A-7C) and simplifies the chip design of the SOT-MRAM cell. The ferromagnetic bias layer 206 may be omitted in a second configuration shown in FIG. 1B, in which the free layer 104 and the reference layer 108 have horizontal magnetizations.

The AFM layer 202 (if present) is configured to fix the magnetization direction of the ferromagnetic bias layer 206 in a predetermined direction. For example, the AFM layer 202 may fix the magnetization direction of the ferromagnetic bias layer in the x-direction via exchange biasing. In one embodiment, the AFM layer 202 may be composed of an antiferromagnetic material. For example, the AFM layer 202 may comprise IrMn. The AFM layer 202 may be omitted in the second configuration shown in fig. 1B, wherein the free layer 104 and the reference layer 108 have horizontal magnetizations.

During a sensing (i.e., read) operation, the read current I_RDMay flow between terminal 1 and terminal 3 through tunnel junction 101. During a programming (i.e., write) operation, the write current I_WRCan flow between terminal 2 and terminal 3. Write current I_WRCan flow up (and down) into the nonmagnetic spin hall effect layer 204 to cause a magnetization transition of the free layer 104 when passing underneath. The write current does not flow through the tunnel junction 101 to the terminal 1. Thus, the read and write currents flow in different directions.

In the first configuration of fig. 1A, since the resistivity of most materials used for the AFM layer 202 (such as IrMn) is relatively high, joule heating due to write current in the extended regions of the AFM layer 202 between adjacent memory cells can be problematic when high current densities are flowed through the ferromagnetic bias layer 206. In addition, it is desirable to have as much of the write current as possible flow through the nonmagnetic spin hall effect layer 204 to maximize the switching current through the tunnel barrier layer 106. A patterned capping layer 210 can be deposited on the ferromagnetic bias layer 206 to cover most, if not all, of the ferromagnetic bias layer 206 in the extended regions between the memory cells, as close as possible to the sidewalls of the nonmagnetic spin hall effect layer 204 and the sidewalls of the ferromagnetic bias layer 206. Capping layer 210 may comprise a highly conductive material such as gold, copper, or silver.

β phase tungsten provides over 30% Spin Hall Angle (SHA) and is therefore one of the most promising materials to provide spin Hall effect that produces pure spin current under which to switch the magnetization of the free layer large spin Hall angle is an important factor in reducing the switching current because it produces more spin current for the current provided by the transistor to the non-magnetic spin Hall effect layer 204 however, β phase tungsten has thermal and thickness instability, for example, β phase tungsten can easily transform (e.g., recrystallize) to α phase tungsten upon annealing, additionally, β phase tungsten can spontaneously transform to α phase tungsten when the layer thickness exceeds 20nm under typical growth conditions, furthermore, reactive sputtering of tungsten lacks process reproducibility due to target poisoning although high pressure low power deposition processes of β phase tungsten are available, this process produces β phase tungsten with high surface roughness over 3.5nm, making the deposited material performance worse for thin film applications such as spin Hall layers of SOT magnetoresistive cells.

Critical switching current J required to induce a magnetization change in the free layer 104_cSOT has the following relationship:

wherein M is_sIs the saturation magnetization of the free layer 104, t_FLIs the thickness of the free layer 104, α is the damping constant, and

by

Is given in which H_k⊥Is the effective (interfacial) perpendicular anisotropy field of the free layer 104 in accordance with one aspect of the present disclosure, a configuration is provided for the combination of the β phase tungsten layer and the interfacial layer for reducing effects on the critical switching currentVarious parameters of (2).

In accordance with one aspect of the present disclosure, a nonmagnetic spin Hall effect layer 204 is provided that comprises an alternating stack of β phase tungsten layers 10 and nonmagnetic dusting layers 20 (i.e., thin nonmagnetic layers). in one embodiment, the nonmagnetic dusting layers 20 may comprise or consist of noble metal layers, such as noble metals having fcc or hcp unit cell structures, for example, ruthenium, platinum, iridium, palladium, rhodium, rhenium, and/or osmium layers.an optional β phase tungsten capping layer 12 may be formed on the alternating stack (10, 20) between the alternating stack and the free layer 104. the capping layer 12 may have a thickness in the range of 0.5nm to 1 nm.

In one embodiment, the alternating stack (10, 20) includes a pair of N periodic repetitions of β phase tungsten layers 10 and nonmagnetic dusting layers 20, and where N is an integer in the range of 5 to 30, such as 5 to 15, for example 8 to 10, in which case each β phase tungsten layer 10 may have the same first thickness, and each nonmagnetic dusting layer 20 may have the same second thickness.

In one embodiment, each β -phase tungsten layer 10 may have a thickness of no greater than 2nm, for example, no greater than 1.2nm, such as 0.2nm to 1.2nm, including 0.5nm to 1nm, to ensure that tungsten material deposited in β phase remains in β phase during the deposition process the thicknesses of β -phase tungsten layers 10 in the stack may be the same as or different from each other. each of the nonmagnetic dusting layers 20 may have a thickness in the range of 0.1nm to 0.3nm (such as 0.2nm to 0.25 nm). alternatively or in addition, the ratio of the thickness of each pair of inner β -phase tungsten layers 10 to the thickness of the pair of nonmagnetic dusting layers 20 may be in the range of 3 to 6, such as 4 to 5 (i.e., the ratio of a 1nm thick tungsten layer to a 0.2nm thick dusting layer is 5).

As described above, the laminated multilayer structure of β phase tungsten tends to change phase from β to α phase as the film thickness increases, the physical continuity of the alternating stack (10, 20) with the non-magnetic dusting layer 20 of a noble metal, such as ruthenium or platinum, in the z-direction perpendicular to the interface of β phase tungsten material 10 within the alternating stack (10, 20) is broken by the laminated multilayer structure of the alternating stack (10, 20). the non-magnetic dusting layer 20 provides the function of stabilizing the β phase adjacent to β 1 phase tungsten layer 10. the thickness of the β phase tungsten layer 10 does not exceed a critical thickness (such as 2nm or another suitable thickness depending on the process conditions) above which transition to α phase can occur during subsequent thermal processing.

Referring to FIG. 2, a second exemplary spin-orbit torque (SOT) magnetoresistive memory cell 200 is shown, which is derived from the first exemplary SOT magnetoresistive memory cell 100 shown in FIG. 1A, but with the ferromagnetic bias layer 206, the nonmagnetic spin Hall effect layer 204, and the patterned capping layer 210 modified, according to an embodiment of the present disclosure.

In particular, the ferromagnetic bias layer 206 is formed with tapered edges 302. The tapered edge 302 may be formed by adjusting the milling angle during patterning of the ferromagnetic bias layer 206. The slope of the tapered edge 302 provides an additional parameter to tune the strength of the stray field from the ferromagnetic bias layer 206 to the free layer 104.

The nonmagnetic spin hall effect layer 204 may also be formed with tapered edges 310. The tapered edge 310 can be formed by adjusting the milling angle during patterning of the nonmagnetic spin hall effect layer 204. The slope of the tapered edge 310 provides an additional parameter to amplify the fraction of the write current that flows through the nonmagnetic spin hall effect layer 204 by conducting directly through the interface of the nonmagnetic spin hall effect layer 204 and the patterned capping layer 210.

The capping layer 210 may directly contact portions of the top surface of the AFM layer 202. The tapered edge 302 of the ferromagnetic bias layer 206 helps to allow the capping layer 210 to be as close as possible to the nonmagnetic spin hall effect layer 204 while preventing deposition (or redeposition) on the sidewalls of the free layer 104.

Referring to fig. 3A and 3B, a configuration of a third exemplary spin-orbit torque (SOT) magnetoresistive memory cell 300 is shown, according to an embodiment of the disclosure. The configuration of the third exemplary SOT magnetoresistive memory cell 300 can be derived from the configuration of the first exemplary SOT magnetoresistive memory cell 100 shown in fig. 1A and 1B, but with the nonmagnetic spacer layer 40 and the hafnium layer 60 interposed on the nonmagnetic spin hall effect layer 204.

For example, nonmagnetic spacer layer 40 may be deposited directly on the top surface of β phase tungsten capping layer 12 (or directly on top of the alternating stack (10, 20) if capping layer 12 is omitted.) nonmagnetic spacer layer 40 is preferably a noble metal and may consist essentially of ruthenium or platinum and may have a thickness in the range of 0.1nm to 0.4nm hafnium layer 60 may be deposited directly on the top surface of nonmagnetic spacer layer 40 hafnium layer 60 may consist essentially of hafnium and may have a thickness in the range of 0.4nm to 1.5nm, such as in the range of 0.5nm to 1 nm.

Interposing a hafnium layer 60 between an β phase tungsten layer (e.g., β phase tungsten cap 12 or alternating stack (10, 20)) and the free layer 104 increases the effective (interfacial) perpendicular anisotropy field of the free layer 104, thereby reducing the effective magnetization of the free layer 104

And as a corollary, the critical switching current is reduced. β direct contact between the tungsten layer and the hafnium layer can initiate the conversion of the β phase tungsten material to another (e.g., α) phase of tungsten material.the nonmagnetic spacer layer 40 acts as a spacer layer that prevents the β phase tungsten capping layer 12 from converting to the α phase.

In an alternative embodiment, the alternating stack of β phase tungsten layers 10 and dusting layers 20 of memory cell 300 can be replaced by a single thicker β phase tungsten layer in this embodiment, the spin Hall effect layer 204 can be composed of a single β phase tungsten layer, and the nonmagnetic spacer layer 40 and hafnium layer 60 are formed over the β phase tungsten nonmagnetic spin Hall effect layer 204. thus, only one of the alternating stack (10, 20) or hafnium layer 60 can be included in the memory cell 300, or both can be included in combination in the memory cell 300.

Referring to FIG. 4, a fourth exemplary spin-orbit-torque (SOT) magnetoresistive memory cell 400 is shown, which originates from the second exemplary SOT magnetoresistive memory cell 200 shown in FIG. 2A, but with a nonmagnetic spacer layer 40 and a hafnium layer 60 interposed above the nonmagnetic spin Hall effect layer 204, according to an embodiment of the present disclosure. The nonmagnetic spacer layer 40 and the hafnium layer 60 may be the same as in the third exemplary SOT magnetoresistive memory cell 300 shown in fig. 3A and 3B and provide the same beneficial effects as in the third exemplary SOT magnetoresistive memory cell 300.

Referring to fig. 5A and 5B, a configuration of a fifth exemplary spin-orbit torque (SOT) magnetoresistive memory cell 500 according to an embodiment of the disclosure is shown. The configuration of the fifth exemplary SOT magnetoresistive memory cell 500 stems from the configuration of the third exemplary SOT magnetoresistive memory cell 300 shown in fig. 3A and 3B, but with the hafnium layer 60 replaced by a stacked stack comprising a first hafnium layer 61, a hafnium oxide layer 55, and a second hafnium layer 62. Alternatively, modifying the third exemplary Spin Orbit Torque (SOT) magnetoresistive memory cell 300 to the fifth exemplary Spin Orbit Torque (SOT) magnetoresistive memory cell 500 can be viewed as inserting an additional hafnium oxide layer 55 and a second hafnium layer 62 between the hafnium layer 60 (which corresponds to the first hafnium layer 61) and the free layer 104.

The hafnium oxide layer 55 may be formed by oxidizing a surface portion of the first hafnium layer 61 the hafnium oxide layer 55 adjacent to the free layer 104 may induce a reduction in the damping constant α in the free layer 104, thereby reducing the critical switching current for inducing a magnetization transition of the free layer 104 however, oxidation of the free layer 104 is undesirable the second hafnium oxide layer 62 may be interposed between the hafnium oxide layer 55 and the free layer 104 to prevent the free layer 104 from oxidizing the free layer 104 of the magnetic tunnel junction 101 may be formed over the second hafnium layer 62 and directly on the second hafnium layer 62 in one embodiment, the first hafnium layer 61 may have a thickness in the range of 0.3nm to 1nm, such as in the range of 0.5nm to 0.8nm, the hafnium oxide layer 55 may have a thickness in the range of 0.1nm to 0.5nm, and the second hafnium layer 62 may have a thickness in the range of 0.3nm to 1nm, such as in the range of 0.5nm to 0.8 nm.

In an alternative embodiment, the alternating stack of β phase tungsten layers 10 and dusting layers 20 of memory cell 500 can be replaced by a single thicker β phase tungsten layer in this embodiment, the spin Hall effect layer 204 can be composed of a single β phase tungsten layer, and the nonmagnetic spacer layer 40, first hafnium layer 61, hafnium oxide layer 55, and second hafnium layer 62 are formed over the β phase tungsten nonmagnetic spin Hall effect layer 204. thus, only one of the alternating stack (10, 20) or stack of layers (61, 55, 62) can be included in the memory cell 500, or both can be included in combination in the memory cell 500.

Referring to fig. 6, a sixth exemplary spin-orbit-torque (SOT) magnetoresistive memory cell 600 is shown, derived from the fourth exemplary SOT magnetoresistive memory cell 400 shown in fig. 4, but replacing the hafnium layer 60 with a stacked stack including a first hafnium layer 61, a hafnium oxide layer 55, and a second hafnium layer 62, according to an embodiment of the disclosure. The first hafnium layer 61, the hafnium oxide layer 55 and the second hafnium layer 62 may be the same as in the fifth exemplary SOT magnetoresistive memory cell 500 shown in fig. 5 and provide the same beneficial effects as in the fifth exemplary SOT magnetoresistive memory cell 500.

Referring to FIG. 7A, a seventh exemplary spin-orbit torque (SOT) magnetoresistive memory cell 700 according to an embodiment of the disclosure is shown, which originates from the first exemplary SOT magnetoresistive memory cell 100 shown in FIG. 1A, but replaces the combination of layers 202, 204, and 206 with any of the spin Hall effect layers (i.e., SOT layers) 204 described in the above embodiments. Thus, in this embodiment, the AFM layer 202 and ferromagnetic bias layer 206 may be omitted. The second terminal 2 and the third terminal 3 may be electrically connected directly to the spin hall effect layer 204, rather than to the AFM layer 202 of the previous embodiments. The permanent magnet 120 may surround the memory cell 700Positioned to provide an additional magnetic bias field H_B. In-plane oriented magnetic bias field and write current I_wrThe spin-orbit torque 118 in the y-direction results from combining and being parallel to the write current. The spin orbit torque 118 exerts a torque on the initial magnetization of the free layer 104 such that the magnetization can change from a parallel state to an anti-parallel state, for example. The free layer 104 and the reference layer 108 have perpendicular magnetic anisotropy and thus have a vertical easy axis.

Referring to FIG. 8A, an alternative embodiment of a seventh SOT magnetoresistive memory cell 700 is shown, which originates from the seventh exemplary SOT magnetoresistive memory cell 700 shown in FIG. 7A, wherein the free layer 104 and the reference layer 108 have in-plane magnetic anisotropy, and thus an in-plane easy axis of magnetization. In contrast to the structure shown in fig. 7A, the permanent magnet 120 is not necessary.

Referring to FIG. 7B, an eighth exemplary spin-orbit torque (SOT) magnetoresistive memory cell 800 is shown that originates from the seventh exemplary SOT magnetoresistive memory cell 700 shown in FIG. 7A but with a nonmagnetic spacer layer 40 and a hafnium layer 60 interposed above the nonmagnetic spin Hall effect layer 204, according to an embodiment of the present disclosure. The nonmagnetic spacer layer 40 and the hafnium layer 60 may be the same as in the third exemplary SOT magnetoresistive memory cell 300 shown in fig. 3A and provide the same beneficial effects as in the third exemplary SOT magnetoresistive memory cell 300.

In an alternative embodiment, the alternating stack of β phase tungsten layers 10 and dusting layers 20 of memory cell 800 can be replaced by a single thicker β phase tungsten layer in this embodiment, the spin Hall effect layer 204 can be composed of a single β phase tungsten layer, and the nonmagnetic spacer layer 40 and hafnium layer 60 are formed over the β phase tungsten nonmagnetic spin Hall effect layer 204. thus, only one of the alternating stack (10, 20) or hafnium layer 60 can be included in the memory cell 800, or both can be included in combination in the memory cell 800. the free layer 104 and reference layer 108 have perpendicular magnetic anisotropy, and thus have a vertical easy axis of magnetization.

Referring to FIG. 8B, an alternative embodiment of an eighth SOT magnetoresistive memory cell 800 is shown, which originates from the eighth exemplary SOT magnetoresistive memory cell 800 shown in FIG. 7B, wherein the free layer 104 and the reference layer 108 have in-plane magnetic anisotropy, and thus an in-plane easy axis of magnetization. In contrast to the structure of fig. 7B, the permanent magnet 120 is not necessary.

Referring to fig. 7C, a ninth exemplary spin-orbit-torque (SOT) magnetoresistive memory cell 900 is shown, which originates from the eighth exemplary SOT magnetoresistive memory cell 800 shown in fig. 7B but replaces the hafnium layer 60 with a stacked stack comprising a first hafnium layer 61, a hafnium oxide layer 55, and a second hafnium layer 62, according to an embodiment of the present disclosure. The first hafnium layer 61, the hafnium oxide layer 55 and the second hafnium layer 62 may be the same as in the fifth exemplary SOT magnetoresistive memory cell 500 shown in fig. 5A and provide the same beneficial effects as in the fifth exemplary SOT magnetoresistive memory cell 500.

In an alternative embodiment, the alternating stack of β phase tungsten layers 10 and dusting layers 20 of memory cell 900 can be replaced by a single thicker β phase tungsten layer in this embodiment, the spin Hall effect layer 204 can be composed of a single β phase tungsten layer, and the nonmagnetic spacer layer 40, first hafnium layer 61, hafnium oxide layer 55, and second hafnium layer 62 are formed over the β phase tungsten nonmagnetic spin Hall effect layer 204. thus, only one of the alternating stack (10, 20) or stack of layers (61, 55, 62) can be included in the memory cell 900, or both can be included in combination in the memory cell 900. the free layer 104 and reference layer 108 have perpendicular magnetic anisotropy, and thus have a vertical easy axis.

Referring to FIG. 8C, an alternative embodiment of a ninth SOT magnetoresistive memory cell 900 is shown, which originates from the ninth exemplary SOT magnetoresistive memory cell 900 shown in FIG. 7C, wherein the free layer 104 and the reference layer 108 have in-plane magnetic anisotropy, and thus an in-plane easy axis of magnetization. In contrast to the structure of fig. 7C, the permanent magnet 120 is not necessary.

Figure 9A shows X-ray diffraction data from a 2 theta scan of three alternating stacks of tungsten layers 10 and intervening layers (i.e., nonmagnetic dusting layers) 20 covered by capping layers 12. curve 710 shows the intensity of a scattered X-ray beam as a function of 2 theta delta for a first sample of 10 repetitions of a 1nm thick β -phase tungsten layer 10 and a 0.2nm thick ruthenium dusting layer 20 capped with a 1nm thick β -phase tungsten layer 12 according to an embodiment of the present disclosure. curve 720 shows the intensity of a scattered X-ray beam as a function of 2 theta delta for a second sample of 10 repetitions of 1nm thick β -phase tungsten layer 12 including a 1nm thick β -phase tungsten layer 10 and a 0.2nm thick platinum dusting layer 20 capped with a 1nm thick β -phase tungsten layer 12 according to an embodiment of the present disclosure. curve 730 shows the intensity of a scattered X-ray beam as a function of 2 theta delta for a 10 sample of 1nm thick β -phase tungsten layer and a 0.2nm thick hafnium dusting layer 730 shows the intensity of a scattered X-ray beam as a function of a plot of a spacing between a spacing of a spacing between the alternating peaks of a tungsten layer 30 nm thick β -phase tungsten layer as a comparison of a plot of a visible tungsten layer 20, a plot 730, a plot of a tungsten layer spacing curve of a tungsten layer spacing, a plot of a visible tungsten layer spacing, a plot of a spacing, a plot of a platinum spacing, a plot of a spacing, a plot of a spacing between a plot of a plot.

FIG. 9B shows resistivity measurement data for an alternating stack of β -phase tungsten layers 10 and nonmagnetic dusting layers 20 (10, 20) comprising ruthenium or platinum, according to an embodiment of the disclosure, before and after annealing. β -phase tungsten has a higher resistivity than α -phase tungsten.Curve 810 corresponds to a set of resistivity measurements after ten repeated first sample film depositions on a 1nm thick β -phase tungsten capping layer comprising β -phase tungsten layers 10 and 0.2nm thick ruthenium dusting layers 20 having thicknesses represented by the x-axis (1 nm, 1.5nm, and 2nm, respectively). Curve 820 corresponds to a set of resistivity measurements on a first sample after 5 hours of 280 degree annealing.Curve 830 corresponds to a set of resistivity measurements on a second sample 840 after ten repeated dust depositions of a 1nm thick β -phase tungsten capping layer comprising β -phase tungsten layers 10 and 0.2nm thick platinum layers 20 having thicknesses represented by the x-axis (1 nm, 1.5nm, and 2nm, respectively.) the second set of resistivity measurements after 280 degree centigrade samples.

Such resistivity changes in samples having a β tungsten layer thickness of 1.5nm or 2nm are believed to reflect a phase change of the tungsten material from the β phase to the α phase caused by annealing, therefore, limiting the thickness of the β phase tungsten layer 10 has the beneficial effect of preventing the β phase tungsten material phase from changing to the α phase tungsten material under annealing conditions.

FIG. 10 illustrates the effective magnetization of the free layer 104 as a function of the thickness of the intervening hafnium layer 60, according to one embodiment of the present disclosure

Curve 910 shows the effective in-plane magnetic stiffness of the free layers 104 for a sample comprising an alternating stack (10, 20)

The alternating stack includes eight repetitions of a pair of a 1nm thick β phase tungsten layer 10 and a 2nm thick ruthenium dusting layer 20 capped with a 1nm thick β phase tungsten capping layer 12, then a 0.2nm thick ruthenium layer 40, and a thickness corresponding to a x-coordinate (i.e., 0.6nm, 0.8nm, and 1nm, respectively) hafnium layer 60, curve 920 shows the effective magnetization of the free layer 104 of a sample comprising the alternating stack (10, 20)

The alternating stack includes eight repetitions of a pair of a 1nm thick β phase tungsten layer 10 and a 2nm thick platinum dusting layer 20 capped with a 1nm thick β phase tungsten cap layer 12, then a 0.2nm thick platinum layer 40, and a thickness corresponding to a x-coordinate (i.e., 0.6nm, 0.8nm, and 1nm, respectively) hafnium layer 60, a free layer 104The values decrease with the thickness of the hafnium layer 60, showing the effectiveness of the hafnium layer in reducing the critical switching current.

11-13 show the dependence of γ Δ H/2 as a function of frequency during ferromagnetic resonance measurements on first, second, and third exemplary films, respectively.A first exemplary film characterized in FIG. 11 includes 8 repetitions of a unit having β phase tungsten layer 1nm thick and ruthenium layer 0.2nm thick, and β phase tungsten capping layer 1nm thick according to an embodiment of the disclosure.A second exemplary film characterized in FIG. 12 includes 5 repetitions of a unit having β phase tungsten layer 1.5nm thick and ruthenium layer 0.2nm thick, tungsten capping layer 1nm thick β phase tungsten capping layer 0.2nm thick, and ruthenium layer 1nm thick according to an embodiment of the disclosure.A third exemplary film characterized in FIG. 13 includes 5 repetitions of a unit having ruthenium layer 1.5nm thick and ruthenium layer 0.2nm thick, tungsten capping layer 1nm thick β, ruthenium layer 0.2nm thick, and a fifth exemplary film characterized in FIG. 13 includes 5 repetitions of a unit having ruthenium layer β nm thick and ruthenium layer 0.2nm thick, tungsten capping layer 1nm thick β nm thick, hafnium layer 0.2nm thick, a fifth exemplary film including a magnetoresistive sample layer (SOT) and a seventh configuration representing a fourth exemplary memory cell (200 nm, 300 nm, 200, 300 nm, 300, and 300 nm, 300, and 300, and 12, 300.

In fig. 11-13, the slope of the fitted line represents damping constant α for the first, second and third samples, damping constant α has values of 0.013, 0.0107 and 0.007 respectively although damping constant α provides an improvement over prior art devices, a reduction in damping constant α in the second sample relative to damping constant α in the first sample represents an additional improvement in reducing the critical switching current, in addition, a reduction in damping constant α in the third sample relative to damping constant α in the second sample represents a further improvement in reducing the critical switching current.

According to one embodiment of the present disclosure, a spin-orbit-torque magnetoresistive random access memory (SOT MRAM) cell is provided that includes a magnetic tunnel junction 101 including a free layer 104 having primarily two bi-stable magnetization directions, a reference layer 108 having a fixed magnetization direction, and a tunnel barrier layer 106 located between the free layer 104 and the reference layer 108. the SOT MRAM cell also includes a nonmagnetic spin Hall effect layer 204. in one embodiment, the spin Hall effect layer 204 includes an alternating stack of β phase tungsten layers 10 and noble metal nonmagnetic dust layers 20. in another embodiment, a hafnium layer (60 or 61) is located between the nonmagnetic spin Hall effect layer 204 and the free layer 104 in addition to or in place of the alternating stack (10, 20).

In one embodiment, the free layer 104 is located between the tunnel barrier layer 106 and the nonmagnetic spin hall effect layer 204. In one embodiment, the nonmagnetic dusting layer 20 comprises a ruthenium or platinum layer and the tunnel barrier layer 106 comprises a MgO layer.

In one embodiment, each β -phase tungsten layer 10 has a thickness in the range of 0.2nm to 1.2nm, and each nonmagnetic dusting layer 20 has a thickness in the range of 0.1nm to 0.3nm in one embodiment, the alternating stack includes N periodic repetitions of a pair β -phase tungsten layers 10 and nonmagnetic dusting layers 20, and N is an integer in the range of 5 to 30.

In one embodiment, the SOT MRAM cell may further include a first hafnium layer (60 or 61) located between the free layer 104 and the nonmagnetic spin hall effect layer 204 of the magnetic tunnel junction 101. In one embodiment, the SOT MRAM cell further includes a nonmagnetic spacer layer 40 located between the first hafnium layer (60 or 61) and the nonmagnetic spin Hall effect layer 204. The nonmagnetic spacer layer 40 may contact the first hafnium layer (60 or 61). In one embodiment, the first hafnium layer (60 or 61) has a thickness in the range of 0.4nm to 1.5nm, and the nonmagnetic spacer layer 40 has a thickness in the range of 0.1nm to 0.4 nm.

In one embodiment, the SOT MRAM cell further comprises a hafnium oxide layer 55 located between the free layer 104 and the first hafnium layer 61 of the magnetic tunnel junction 101. In one embodiment, the SOT MRAM cell further comprises a second hafnium layer 62 located between the free layer 104 and the hafnium oxide layer 55 of the magnetic tunnel junction 101. In one embodiment, the first layer of hafnium 61 has a thickness in the range of 0.3nm to 1 nm; the hafnium oxide layer 55 has a thickness in the range of 0.1nm to 0.5 nm; and the second hafnium layer 62 has a thickness in the range of 0.3nm to 1 nm.

As described above, each SOT MRAM cell (100-900) is a three terminal device that includes a first terminal 1 electrically connected to the hard magnetic layer 112 on the reference layer side 108 of the magnetic tunnel junction 101, a second terminal 2 electrically connected directly or indirectly to a first end of the nonmagnetic spin Hall effect layer 204, and a third terminal 3 electrically connected directly or indirectly to a second end of the nonmagnetic spin Hall effect layer 204. The SOT MRAM cell is configured to generate a read current flowing between the first terminal 1 and the third terminal 3 through the tunnel junction 101 during a sensing operation. The SOT MRAM cell is also configured to generate a write current flowing between the second terminal 2 and the third terminal 3 during a programming operation such that the write current does not flow through the tunnel junction 101 and such that the read and write currents flow in different directions during respective sensing and programming operations.

In one embodiment, the SOT MRAM cell (100, 200, 300, 400, 500, 600) further includes a ferromagnetic bias layer 206 configured to provide a spin-orbit torque via an abnormal hall effect and a magnetic bias field on the free layer 104 to achieve deterministic switching, and an Antiferromagnetic (AFM) layer 202 positioned adjacent the ferromagnetic bias layer 206 and configured to fix a magnetization direction of the ferromagnetic bias layer 206 in a predetermined direction. In these embodiments, the second terminal 2 is electrically connected to a first end of the AFM layer 202 and the third terminal 3 is electrically connected to a second end of the AFM layer 202 such that the second and third terminals (2, 3) are indirectly electrically connected to respective first and second ends of the nonmagnetic spin hall effect layer 204.

In another embodiment, the SOT MRAM cell (700, 800, 900) includes at least one permanent magnet 120 positioned adjacent to the tunnel junction 101 and omits layers 202 and 206. In these embodiments, the second and third terminals (2, 3) are directly electrically connected to respective first and second ends of the nonmagnetic spin hall effect layer 204.

The various SOT MRAM cells of the present disclosure provide non-limiting advantages through the alternating stack of nonmagnetic spin Hall effect layer 204 and/or intervening layers between nonmagnetic spin Hall effect layer 204 and free layer 104. a laminated multilayer structure with tight control of the thickness ratio between β phase tungsten layer 10 and nonmagnetic dusting layer 20 ensures a thermally stable β phase tungsten material with a high spin Hall angle, which is three times the spin Hall angle of platinum.

While the foregoing refers to certain preferred embodiments, it is to be understood that the disclosure is not so limited. Various modifications to the disclosed embodiments will be apparent to those skilled in the art, and such modifications are intended to be within the scope of the present disclosure. Embodiments employing specific structures and/or configurations are shown in the present disclosure, it being understood that the present disclosure may be practiced in any other compatible structures and/or configurations that are functionally equivalent, provided that such substitutions are not explicitly prohibited or otherwise considered to be impossible by one of ordinary skill in the art. All publications, patent applications, and patents cited herein are incorporated by reference in their entirety.