Embodiments described herein relate generally to magnetic memories.
Recently, research and development of magnetic memories employing write methods using spin orbit torque or spin Hall effect is being actively performed. The spin Hall effect is a phenomenon caused by a current flowing through a nonmagnetic layer. Due to the influence of the current, electrons having a spin angular momentum (“spin”) are diffused in one direction and electrons having a spin angular momentum in a direction opposite to the one direction are diffused in the opposite direction to cause a spin current Is that flows in a direction perpendicular to the direction in which the current flowing through the nonmagnetic layer. As a result, opposite spins are accumulated around opposite interfaces of the nonmagnetic layer.
A magnetic tunnel junction (MTJ) element includes a first magnetic layer (“reference layer”) in which the magnetization direction is fixed, a second magnetic layer (“storage layer”) in which the magnetization direction is changeable, and a nonmagnetic insulating layer disposed between the first magnetic layer and the second magnetic layer. If the second magnetic layer (storage layer) of the MTJ element is disposed on the aforementioned nonmagnetic layer, and a current is caused to flow through the nonmagnetic layer to generate a spin current in the nonmagnetic layer, the magnetization direction of the storage layer may be switched by the spin orbit torque (SOT) applied to the storage layer by means of the spin current generated in the nonmagnetic layer and the electrons with a spin accumulated near the MTJ element. A magnetic random access memory (MRAM) to which data is written by using the spin orbit torque or spin Hall effect is called “SOT-MRAM.” Data is read from the SOT-MRAM using a magnetoresistive effect (MR effect) of the MTJ element, by causing a read current to flow between the reference layer and the nonmagnetic layer.
An MRAM called STT-MRAM is also known, to which data is written by causing a write current to flow between the storage layer and the reference layer of the MTJ element to apply a spin transfer torque (STT) to the storage layer. Data is read from the STT-MRAM in the same manner as in the write operation, by causing a read current to flow between the storage layer and the reference layer. Thus, the read current path and the write current path are the same in the STT-MRAM. This increases the variation in device characteristics as the device size is decreased. Therefore, it is difficult to secure the margin in each of the read current, the write current, the current flowing through the transistor connected to the MTJ element, and the breakdown current of the nonmagnetic insulating layer of the MTJ element by suppressing the variation in each current.
In contrast, the margin with respect to the variation of each current is greater in the SOT-MRAM since the read current path is different from the write current path. Therefore, the variation in each of the read current, the transistor current, and the breakdown current of the nonmagnetic insulating layer of the MTJ element may be controlled in a manner from the control of the variation in each the write current, the transistor current, and the electromigration current to the nonmagnetic layer. Thus, if the MTJ elements acting as the memory elements are miniaturized (to increase the capacity), the margin with respect to the variation in each current is considerably greater than that of the STT-MRAM. However, at present, the write efficiency of the SOT-MRAM is inferior to that of the STT-MRAM.
A magnetic memory according to an embodiment includes: a first terminal, a second terminal, and a third terminal; a first nonmagnetic layer, which is conductive, including a first portion, a second portion, and a third portion, the first portion being disposed between the second portion and the third portion, the second portion being electrically connected to the first terminal, and the third portion being electrically connected to the second terminal; a first magnetoresistive element including a first magnetic layer electrically connected to the third terminal, a second magnetic layer disposed between the first magnetic layer and the first portion, and a second nonmagnetic layer disposed between the first magnetic layer and the second magnetic layer; and a first layer at least disposed between the first portion and the second magnetic layer, the first layer including at least one of Mg, Al, Si, Hf, or a rare earth element, and the first layer further including at least one of oxygen or nitrogen.
Before embodiments are described, how the inventors have reached the present invention will be described.
The MTJ element 20 includes a storage layer 21 in which the magnetization direction is changeable, a reference layer 23 in which the magnetization direction is fixed, and a nonmagnetic insulating layer 22 disposed between the storage layer 21 and the reference layer 23. The feature “magnetization direction is changeable” means that the magnetization direction may be changed after a write operation, and the feature “magnetization direction is fixed” means that the magnetization direction is not changed after a write operation. The storage layer 21 is connected to the conductive layer 12a, and the reference layer 23 is connected to the wiring line 40. One (“terminal”) of the source and the drain of the transistor 30 is connected to the terminal 13a of the conductive layer 12a. The other (“terminal”) of the source and the drain of the transistor 30 and the gate (“control terminal”) are connected to a control circuit that is not shown. The terminal 13b of the conductive layer 12b is grounded as shown in
In this SOT-MRAM, a write operation is performed by causing a write current Iw to flow through the conductive layers 12a and 12b between the terminal 13a and the terminal 13b, by means of the transistor 30, and a read operation is performed by causing a read current Ir to flow through the terminal 13a, the conductive layer 12a, the MTJ element 20, and the wiring line 40, by means of the transistor 30. Thus, as described above, the write current path and the read current path are different from each other.
STT-MRAM. The memory cell incudes a wiring line 16, an MTJ element 20, and a wiring line 40. The MTJ element 20 is disposed between the wiring line 16 and the wiring line 40, and includes a storage layer 21, a reference layer 23, and a nonmagnetic insulating layer 22 disposed between the storage layer 21 and the reference layer 23. One of the storage layer 21 and the reference layer 23 is connected to the wiring line 16, and the other is connected to the wiring line 40. In
As described above, the write efficiency of the SOT-MRAM is inferior to that of the STT-MRAM. Therefore, the write efficiency needs to be improved. The write efficiency is expressed by dividing Δ(=KV/(kBT)), which is the index of the thermal stability, by Ic, i.e., Δ/Ic, where K is the uniaxial magnetic anisotropy of the storage layer, V is the volume of the storage layer, kB is the Boltzmann constant, T is the absolute temperature of the storage layer, and KV is the height of the energy barrier in cases where the spin in the storage layer and the spin in the reference layer are in the parallel state and the antiparallel state. Assuming that the write current needed to change the magnetization direction of the storage layer from parallel to antiparallel relative to the magnetization direction of the reference layer is Ip, and the write current needed to change the magnetization direction of the storage layer from antiparallel to parallel relative to the magnetization direction of the reference layer is Iap, Ic is the mean value of Ip and Iap, i.e., Ic=(Ip+Iap)2.
An MTJ element including a CoFeB storage layer is generally formed on an amorphous layer. Therefore, the CoFeB layer grows as an amorphous layer. A nonmagnetic insulating layer of MgO is formed on the CoFeB layer to be (100)-oriented. Thereafter, due to the post annealing, CoFeB is uniformly oriented on the MgO(100) crystal surface. Therefore, the variation in the coercive force Hc is subtle.
However, the conductive layer underneath an MTJ element of a SOT-MRAM is a crystalline layer of, for example, β-Ta having a crystalline structure with great spin orbit torque in order to reduce the write current. Therefore, the CoFeB layer formed on the conductive layer is not a complete amorphous layer and grows in various directions. This leads to the variation in coercive force Hc. Another reason for the variation in coercive force Hc is a large absolute value of the magnetization, i.e., saturation magnetization Ms, of CoFeB, which is approximately equal to 1600 emu/cc even after the annealing at a temperature of 300° C. This causes B in CoFeB to be absorbed by β-Ta and diffused to the conductive layer.
In order to reduce the write current, a material having a large spin Hall angle ΘSH is preferably used to form the conductive layer, as described above. Known materials having a large spin Hall angle ΘSH include a metal such as Ta, W, Re, Os, Ir, Pt, Au, or Ag, an alloy containing at least one of the above elements, and an alloy of Cu and a material having 5d electrons that cause great spin orbit scattering such as Cu-Bi.
It is reported that if a β-W layer is formed in an atmosphere including the noble gas Ar and oxygen, a maximum value at the present stage of the spin Hall angle ΘsH is −0.5 (Nature Comm. DOI:10.1038/ncomms10644).
Next, a problem of the material of the conductive layer will be described. If a CoFeB monolayer film is formed on a β-W layer, and the spin Hall angle ΘSH of this layer is evaluated by the ferromagnetic resonance method, ΘSH of −0.5 may be obtained, as described above (Nature Comm. DOI:10.1038/ncomms10644). However, the characteristics of an MTJ element including a storage layer of CoFeB formed on the β-W layer are considerably degraded, and the MR characteristics of the CoFeB layer are also considerably degraded due to the generation of a nonmagnetic layer (dead layer) in the CoFeB layer after the annealing at a temperature of 300° C. In contrast, the characteristics of an MTJ element formed on a β-Ta layer have no problem. It has become apparent that the thickness of the nonmagnetic layer in the CoFeB layer is from 0.2 nm to 0.3 nm or more, and that the MR ratio of the CoFeB layer is reduced from about 200% to less than 50%. This is a great problem to be solved in achieving a large-capacity MRAM.
The inventors of the present invention have studied hard to obtain SOT-MRAMs that are capable of solving the above problem. Such SOT-MRAMs will be described in the descriptions of embodiments.
A magnetic memory according to a first embodiment will be described with reference to
The layer 15 is formed of an oxide or nitride containing at least one of Mg, Al, Si, Hf or a rare earth element. In other words, the layer 15 may be formed of an oxide or nitride of an alloy containing at least one of the aforementioned elements. As used herein, a phrase referring to “at least one of” a list of items refers to any combination of those items, including a single member. As an example, “at least one of: a, b, or c” is intended to cover a, b, c, a-b, a-c, b-c, and a-b-c.”
The MTJ element 20 includes a storage layer 21 in which the magnetization direction is changeable, a reference layer 23 in which the magnetization direction is fixed, and a nonmagnetic insulating layer 22 disposed between the storage layer 21 and the reference layer 23. The storage layer 21 is connected to the conductive layer 12a via the layer 15, and the reference layer 23 is connected to one of the source and the drain (“terminal”) of the transistor 25. The other of the source and the drain (“terminal”) of the transistor 25 is connected to a control circuit (not shown) via a third terminal 26, and the gate (“control terminal”) is also connected to the control circuit. The transistor 25 may be eliminated. In such a case, the control circuit controls the voltage applied to the reference layer 23 of the MTJ element 20 via the third terminal 26. The third terminal 26 is used to apply a voltage to and cause a current to flow through the MTJ element 20.
One of the source and the drain (“terminal”) of the transistor 30 is connected to the terminal 13a of the conductive layer 12a. The other of the source and the drain (“terminal”) and the gate (“control terminal”) of the transistor 30 are connected to a control circuit (not shown). The terminal 13b of the conductive layer 12b is grounded as shown in
In the SOT-MRAM, a write operation is performed by applying a voltage to the reference layer 23 of the MTJ element 20 by means of the transistor 25 and causing a write current Iw to flow through the conductive layers 12a and 12b between the terminal 13a and the terminal 13b by means of the transistor 30. When the write current Iw flows through the conductive layer 12a, electrons 14a that are spin-polarized in one of the up spin direction and the down spin direction flow on the top surface side of the conductive layer 12a, and electrons 14b that are spin-polarized in the other of the up spin direction and the down spin direction flow on the lower surface side of the conductive layer 12a. This causes a spin current, which applies a spin torque to the storage layer 21 of the MTJ element 20, resulting in the switching of the magnetization direction of the storage layer 21. In the write operation, the voltage may be applied to the reference layer 23 of the MTJ element 20 by means of the transistor 25. The applied voltage changes the uniaxial magnetic anisotropy in the storage layer 21 of the MTJ element 20. This may facilitate the switching of the magnetization direction in the storage layer 21. The transistor 25 may be eliminated and the reference layer 23 of the MTJ element 20 may be electrically connected to a bit line (not shown) via the third terminal 26, as a first modification of the first embodiment shown in
A read operation is performed by causing a read current Ir (not shown) to flow through the terminal 13a, the conductive layer 12a, the MTJ element 20, and the transistor 25 or the aforementioned bit line by means of the transistor 30. The control circuit includes a write circuit for performing the write operation and a readout circuit for performing the read operation.
In the first embodiment, the layer 15 is disposed immediately below the MTJ element 20 on the conductive layer 12a. If the layer 15 and the MTJ element 20 are projected upon the conductive layer 12a, the area of the layer 15 is greater than the area of the storage layer 21 of the MTJ element 20. Thus, the area of the surface of the layer 15 facing the conductive layer 12a is greater than the area of the surface of the storage layer 21 facing the layer 15. A distance do between the side surface the layer 15 and the side surface of the storage layer 21 crossing the direction in which the write current Iw flows is preferably longer than the spin diffusion length. The spin diffusion length of heavy metals is short, from 0.5 nm to several nm, although the actual length may differ for each material. With the above-described structure, a more amount of spin may be absorbed from the conductive layer 12a to the storage layer 21.
In the magnetic memory according to the first embodiment containing the above-described structure including the layer 15 of an oxide or nitride disposed between the conductive layer 12a and the storage layer 21 of the MTJ element 20, the element diffusion between the storage layer 21 and the conductive layer 12a may be prevented. For example, if the storage layer 21 contains boron (B), the boron may be prevented from being diffused into and absorbed by the conductive layer 12a. This prevents the generation of a nonmagnetic layer that may eliminate the magnetization in the storage layer 21. Since the generation of the nonmagnetic layer may be prevented, the value of the write current may be reduced, and the variation in coercive force Hc may also be reduced. On the other hand, B (boron) needs to be eliminated from CoFeB to improve the magnetoresistance (MR). From this viewpoint, the storage layer may preferably have a multilayer structure including a nonmagnetic layer, by stacking a ferromagnetic layer, a nonmagnetic layer, and a ferromagnetic layer in this order.
An increase in the thickness of the layer 15 leads to a steep increase in the value of the write current. Therefore, the thickness of the layer 15 is preferably 1 nm or less, and more preferably 0.9 nm or less. The material of the layer 15 is preferably an oxide that may prevent the spin-polarized electrons in the conductive layer 12a of such materials as Ta, W, and Pt. Rare earth elements include magnetic elements with f electrons, which do not have an energy band on the Fermi surface, and thus have less spin scattering from the electrical viewpoint. Therefore, a preferable result may be obtained if the layer 15 includes an oxide or nitride of a rare earth element. On the contrary, it has been revealed that the use of a material of the conductive layer 12a such as an oxide or nitride of Ta or W in the layer 15 may lead to an unfavorable result.
The layer 15 acts as an etching stopper when the MTJ element 20 is microfabricated. The layer 15 may be left on the conductive layer 12a as in a magnetic memory according to a second modification of the first embodiment shown in
Even if the layer 15 is used as an etching stopper, the thickness of a region of the conductive layer 12a that is not covered by the layer 15 may be reduced as compared to the thickness of the other region that is covered by the layer 15 due to the etching or oxidation. In order to prevent an increase in resistance of the conductive layer 12a caused by the decrease in thickness, the difference between the thickness of the region of the conductive layer 12a covered by the layer 15 and the thickness of the region not covered by the layer 15 is preferably 2 nm or less, and more preferably 1 nm or less. Thus, the difference between the thickness of the conductive layer 12a immediately below the layer 15 and the thickness in the other region is preferably 2 nm or less, and more preferably 1 nm or less.
In the first embodiment, the layer 15 is disposed in a region of the conductive layer 12a including the region immediately below the MTJ element 20. Therefore, the conductive layer 12a in the first embodiment may be reduced as in the case of the second modification to reduce the value of the write current Ic, thereby improving the write efficiency. While a current is flowing through the conductive layer 12a, electrons with the up spin and electrons with the down spin are separated to the top surface side and the lower surface side of the conductive layer 12a due to the spin Hall effect. The spin of the electrons on the storage layer 21 side is absorbed by the storage layer 21, and thus the magnetization switching is achieved. The spin is absorbed by the storage layer 21 from not only the region immediately below the MTJ element 20 but also from the region around the MTJ element 20 in which spin is accumulated. Therefore, the state shown in
As described above, the first embodiment and its modifications are capable of improving the current density of the write current flowing through the conductive layer 12a, thereby improving the write efficiency. Furthermore, the first embodiment and its modifications are also capable of reducing the variation in coercive force Hc. Since the layer 15 acts as an etching stopper of the conductive layer 12a, a magnetic memory with a thin conductive layer may be provided.
The magnetic material of the storage layer 21 and the reference layer 23 of the first embodiment is not limited, and may be a Ni-Fe alloy, a Co-Fe alloy or a Co-Fe-Ni alloy. An amorphous material such as (Co, Fe)—(B), (Co, Fe, Ni)—(B), (Co,
Fe, Ni)—(B)—(P, Al, Mo, Nb, Mn), or Co—(Zr, Hf, Nb, Ta, Ti) may also be used. For example, (Co, Fe, Ni) means that at least one of Co, Fe, or Ni is included in the material. Furthermore, (B) means that B may be included or not included. The magnetic material of the storage layer 21 and the reference layer 23 may also be a Heusler material such as Co—Fe—Al, Co—Fe—Si, Co—Fe—Al—Si, Co—Mn—Si, or Co—Mn—Fe—Si. These layers preferably have a multilayer structure in which a plurality of magnetic layers are stacked, instead of a monolayer structure. In this case, for example, a nonmagnetic layer 19 is disposed between magnetic layers 17 and 18 as shown in
In particular, the storage layer 21 preferably has a multilayer structure. If the magnetization direction (spin) is in parallel with the film plane, the preferable combinations of the multilayer structure include CoFe(B)/Cu/CoFe(B), Fe(CoB)/Cr/Fe(CoB), Mn-based Heusler/MgO/Mn-based Heusler, or a face-centered cubic (fcc) magnetic material/Ru/fcc magnetic material/(Ta, W, Mo)/CoFeB, CoFe/Cr/CoFe/(Ta, N, Mo)/CoFeB, CoFe/Cu/CoFe/(Ta, N, Mo)/CoFeB.
If the spin is perpendicular to the film plane, preferable combinations include Co(Fe)(B)/Pt/Co(Fe)(B), Co(Fe)(B)/Pd/Co(Fe)(B), Co(Fe)(B)/Ni/Co(Fe)(B), and fcc magnetic material (multilayer film) such as (Co/Pt)n/Ru/(Co/Pt)m/Ru/fcc magnetic material (multilayer film)/(Ta, W, Mo)/CoFeB. In the above multilayer film, m and n represent the number of stacked layers. For example, (Co/Pt)n means that Co/Pt are stacked n times. Instead of Pt, Pd may be used. If the fcc magnetic material (multilayer film) is used, an ultrathin (Ta, W, Mo)/CoFeB film is preferably disposed at the interface with the nonmagnetic insulating layer 22.
In a magnetic memory including multi-bit memory cells each including a plurality of MTJ elements like a magnetic memory according to a second embodiment that will be described later, the margin of the voltage applied to each MTJ element to cause a current to flow through the conductive layer to switch the spin of the storage layer of the MTJ element may be increased. If the polarity of a voltage applied to a plurality of MTJ elements is set to be different from that of a voltage applied to the other MTJ elements in the second embodiment, for example, if a voltage +V is applied to the former and a voltage −V is applied to the latter, and the spin of the storage layers included in the MTJ elements to which the voltage −V is applied is reversed, the margin may further be increased. The effect of increasing the margin is obtained by either or both of the change in magnetic anisotropy and the spin transfer torque magnetization switching assisted by the voltage applied to the MTJ element. From the viewpoint of power consumption, the change in magnetic anisotropy caused by increasing the resistance of the MTJ element when the voltage is applied is preferable. However, this also has a disadvantage that the read speed is decreased.
On the other hand, if the resistance of the MTJ element is reduced, the contribution of the voltage to the spin transfer torque magnetization switching increases to improve the read speed. However, the power consumption is increased as compared to the case where the magnetic anisotropy is changed by applying the voltage. Which assistance effect of the voltage, the change in magnetic anisotropy and the spin transfer torque magnetization switching, is used may be selected depending on the memory design, and at which value the resistance of the MTJ element needs to be set. The margin can be increased further if the storage layer of each MTJ element has a multilayer structure in the magnetic memory according to the second embodiment.
The reference layer 23 preferably has one-directional anisotropy, and the storage layer 21 preferably has uniaxial anisotropy. The thickness of these layers is preferably from 0.1 nm to 100 nm. Since these magnetic layers should not be superparamagnetic, the thickness is more preferably 0.4 nm or more.
A nonmagnetic element such as Ag (silver), Cu (copper),
Au (gold), Al (aluminum), Mg (magnesium), Si (silicon), Bi (bismuth), Ta (tantalum), B (boron), C (carbon), O (oxygen), N (nitrogen), Pd (palladium), Pt (platinum), Zr (zirconium), Ir (iridium), W (tungsten), Mo (molybdenum), or Nb (niobium) may be added to the magnetic material of these layers to adjust the magnetic characteristics, the crystallinity, the mechanical characteristics, and the chemical characteristics.
The magnetic layer that is close to the nonmagnetic insulating layer 22 is preferably formed of such materials as Co-Fe, Co-Fe-Ni, Fe-rich Ni-Fe which have a large MR (magnetoresistance), and the magnetic layer that is not in contact with the nonmagnetic insulating layer 22 is preferably formed of Ni-rich Ni-Fe or Ni-rich Ni-Fe-Co to adjust the switching magnetic field with the large MR being maintained.
The material of the nonmagnetic insulating layer 22 is preferably an oxide such as AIOx, MgO, and Mg-AlOx.
The material of the conductive layer 12a is preferably a metal including a nonmagnetic heavy metal element with one or more outer shell electrons that are 5d or greater electrons. For example, the material is preferably a metal selected from Ta, W, Re, Os, Ir, Pt, Au, and Ag, an alloy containing at least one of the above metals, or Cu-Bi.
The conductive layer 12a may have a multilayer structure including two or more layers. In this case, the electric resistance of a layer that is close to the storage layer is preferably low. Since the low electric resistance increases the amount of current flowing immediately below the MTJ element, the write current may become lower than that in the case where the electric resistance of the layer close to the storage layer is high. If the conductive layer 12a includes two layers, the layer that is more distant from the storage layer may include at least one of Hf, Al, Mg, or Ti, and B besides the above elements. The layer that is closer to the storage layer preferably includes a metal selected from Ta, W, Re, Os, Ir, Pt, Au, and Ag, an alloy containing at least one of the above metals, or Cu-Bi.
The material of the layer 15 is preferably selected from Mg, Al, Si, and Hf, or a rare earth element, or an oxide or nitride of an alloy of the above elements. More specifically, the layer 15 is preferably formed of a material such as magnesium oxide (MgO), aluminum nitride (AIN), aluminum oxide (AlOx), silicon nitride (SiN), silicon oxide (SiOx), hafnium oxide (HfOx), and an oxide or nitride of La, Ce, Pr, Nd, Pm, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, and Yb. In the above chemical formula, “x” represents the composition ratio. The compositions of the above materials do not need to be completely accurate from the stoichiometric point of view, but may lack or additionally include, for example, oxygen or nitrogen. Thus, the layer 15 includes at least one of Mg, Al, Si, Hf, or a rare earth element, and at least one of oxygen or nitrogen.
The thickness of the nonmagnetic insulating layer 22 is preferably thin enough to allow a tunneling current to flow. However, if the coercive force (i.e., the magnetic anisotropy) of the storage layer of the MTJ element needs to be changed by means of the voltage as in the second embodiment that will be described later, the sheet resistance RA should not be too low, and is preferably a few tens Ωμm2 to a few thousands KΩμm2. In this case, if the sheet resistance is a few thousands KΩμm2, the magnetization switching in the storage layer is mainly caused by the voltage control and the write operation through the conductive layer, and if the sheet resistance is a few tens Ωμm2, the magnetization switching of the storage layer is mainly caused by a combination of the voltage control, the SOT write operation and the STT write operation.
The material of the reference layer 23 is not particularly limited, as long as the magnetization of this layer is stably fixed in one direction. In order to fix the magnetization of the magnetic layer in one direction, a multilayer structure including a plurality of stacked magnetic layers is used. More specifically, multilayer structures such as Co(Co-Fe)/Ru (ruthenium)/Co(Co—Fe), Co(Co-Fe)/Rh (rhodium)/Co(Co—Fe), Co(Co—Fe)/Ir (iridium)/Co(Co—Fe), Co(Co—Fe)/Os (osmium)/Co(Co—Fe), Co(Co—Fe)/Re (rhenium)/Co(Co—Fe), amorphous material such as Co—Fe—B/Ru (ruthenium)/Co—Fe, amorphous material such as Co—Fe—B/Ir (iridium)/Co—Fe, amorphous material such as Co—Fe—B/Os (osmium)/ Co—Fe, or amorphous material such as Co—Fe—B/Re (rhenium)/Co—Fe are used.
Furthermore, a three-layer structure in which three magnetic layers are stacked may also be used, such as (Co/Pt)n/Ru/(Co/Pt)m/(Ta, W, Mo)/CoFeB, (Co/Pt)n/Ir/(Co/Pt)m/(Ta, W, Mo)/CoFeB, (Co/Pt)n/Re/(Co/Pt)m/(Ta , W, Mo)/CoFeB, or (Co/Pt)n/Rh/(Co/Pt)m/(Ta, W, Mo)/CoFeB. In the above three-layer combinations, m and n represent the number of stacked layers. For example, (Co/Pt)n means that Co/Pt are stacked n times. Instead of Pt, Pd may be used.
An antiferromagnetic layer may further be disposed to be adjacent to the reference layer having the multilayer structure.
The material of the antiferromagnetic layer may be Fe-Mn, Pt-Mn, Pt-Cr-Mn, Ni-Mn, Ir-Mn, NiO, and Fe2O3. The structure with an antiferromagnetic layer may prevent the magnetization of the reference layer from being influenced by a current magnetic field from a bit line or word line. Therefore, the magnetization of the reference layer is securely fixed. Furthermore, a stray field from the reference layer may be reduced, and the magnetization shift of the storage layer may be adjusted by changing the thicknesses of the two magnetic layers of the reference layer. A preferable thickness of each magnetic layer is 0.4 nm or more, and not be the thickness at which the magnetic layer becomes superparamagnetic.
A magnetic memory according to a second embodiment will be described with reference to
The conductive layer 12a has terminals 13a and 13b. The n MTJ elements 201 to 20n are disposed to be separate from each other on a region of the conductive layer 12a between the terminal 13a and the terminal 13b. Each of the MTJ elements 201 to 20n includes a reference layer 23 disposed above the conductive layer 12a, a storage layer 21 disposed between the reference layer 23 and the conductive layer 12a, and a nonmagnetic insulating layer 22 disposed between the storage layer 21 and the reference layer 23. Each MTJ element 20i (i=1, . . . , n) is a memory element for storing one bit, and each memory cell is a 1-byte cell including n bits. The material of each of the constituent elements of the second embodiment is the same as that of the first embodiment. The memory cell may include a dummy memory element (for example an MTJ element) that is not used as a memory element.
One of the source and the drain of the transistor 25i is connected to the reference layer 23 of the MTJ element 20i (1=1, . . . , n), and the other is connected to a third terminal 26. One of the source and the drain of the transistor 30 is connected to the terminal 13a, and the other is connected to a control circuit (not shown). The transistor 25i connected to the reference layer 23 of the MTJ element 20i (i=1, . . . , n) may be omitted, as in the first modification of the first embodiment shown in
A layer 15 is disposed between the storage layer 21 of each of the MTJ elements 201 to 20n and the conductive layer 12a in the second embodiment, like the first embodiment shown in
As in the first embodiment, the layer 15 of the second embodiment is disposed on a region of the conductive layer 12a including a region immediately below each MTJ element 20i, (i=1, . . . , n). When viewed from above, the plane area of the layer 15 is greater than the plane area of the storage layer 21 of the MTJ element 20. The distance do between the side surface of the layer 15 and the side surface of the storage layer 21 that cross the direction in which a write current Iw flows is preferably shorter than the spin diffusion length.
The layer 15 may be disposed to cover the top surface of the conductive layer 12a, as in a modification of the second embodiment shown in
A first write method used for the memory cell 10 will be described below. In this embodiment, a write operation for the memory cell 10 is performed in two stages. A write operation for writing 1-byte data (0, 1, 0, 0, . . . , 0, 1) to the memory cell 10 is taken as an example. In this write operation, data “1” is written to the MTJ elements 202 and 20n, and data “0” is written to the other MTJ elements.
First, the transistor 30 and the transistors 251 to 25n, are turned on by means of the control circuit that is not shown to apply a first potential (for example a positive potential) the reference layers 23 of the MTJ elements 201 to 20n and to cause a write current Iw to flow between the terminal 13a and the terminal 13b of the conductive layer 12a. At this time, the magnetization stability (uniaxial magnetic anisotropy) of the storage layers 21 of all the MTJ elements 201 to 20n, is weakened, and the threshold current of the storage layers changes from Ic to Ich. For example, the threshold current Ich is selected to be Ic/2, by applying a voltage to the reference layer of the MTJ element to lower the uniaxial magnetic anisotropy. In this state, a write current Iw0 (Iw>Iw0>Ich) is caused to flow through the conductive layer 12a to write data “0” to all of the MTJ elements 201 to 20n, (0, 0, 0, 0, . . . , 0, 0). Generally, a write error rate of about 10-11 may be obtained if a write current with a value about 1.5 times the value of the threshold current Ich is caused to flow. Therefore, the write current Iw0 is approximately equal to 1.5 times the threshold current Ich.
Next, the transistors of the bits that should be “1”, for example the transistors 252 and 25n, are turned on by means of the control circuit that is not shown to apply a second potential (for example a positive potential) to the reference layers 23 of the MTJ elements 202 and 20n. At this time, the transistor 30 is also turned on by means of the control circuit that is not shown to cause a write current Iw1 (Ic>Iw1>Ich) to flow through the conductive layer 12a in a direction that is opposite to the direction for writing data “0”. As a result, data “1” is written to the storage layers 21 of the MTJ elements 202 and 258. The write current Iw1 is approximately equal to 1.5 times the threshold current Ich, like the aforementioned case. Thus, 1-byte data (0, 1, 0, 0, . . . , 0, 1) can be written by the two-stage write operation. the two-stage write operation is performed by the control circuit that is not shown, which includes a first write circuit for performing the first-stage write operation and a second write circuit for performing the second-stage write operation.
The above-described first write method is performed by applying a first potential (for example a positive potential) to the reference layers 23 of the MTJ elements 201 to 20n and causing a first write current to flow between the terminal 13a and the terminal 13b of the conductive layer 12a, and then by applying a second potential to the reference layers of some of the MTJ elements among the MTJ elements 201 to 20n, to which data is written, and by causing a second write current to flow in a direction that is opposite to the direction of the first write current between the terminal 13a and the terminal 13b of the conductive layer 12a.
A second write method, which is different from the first write method, may also be used. Like the first write method, the second write method is performed in two stages. First, two types of potentials are applied to the MTJ elements 201 to 20n to make easy-to-write bits and difficult-to-write bits. For example, a positive potential Va is applied to activation bits (MTJ elements) 202 to 20n via the corresponding transistors 252 to 25n, and a negative potential Vp is applied to an inactivation bit (MTJ element) 201 via the corresponding transistor 251. A write current is caused to flow through the conductive layer 12a from the first terminal 13a to the second terminal 13b, for example. As a result, data “0” is written to the activation bits (MTJ elements) 202 to 20n. Thereafter, a positive potential Va is applied to the MTJ element 201 via the transistor 251, and a negative potential Vp is applied to the MTJ elements 202 to 20n via the transistors 252 to 25n, and a write current is caused to flow from the second terminal 13b to the first terminal 13a of the conductive layer 12a. As a result, data “1” is written to the MTJ element 201.
The second write method is performed by applying a first potential to the reference layers of the magnetoresistive elements in a first group in the magnetoresistive elements 201 to 20n and a second potential that is different from the first potential to the reference layers of the magnetoresistive elements in a second group that is different from the first group in the magnetoresistive elements 201 to 20n, causing a first write current to flow between the first terminal 13a and the second terminal 13b, and applying the second potential to the reference layers of the magnetoresistive elements in the first group and the first potential to the reference layers of the magnetoresistive elements in the second group and causing a second write current to flow in a direction opposite to the direction of the first write current between the first terminal 13a and the second terminal 13b.
An operation for reading data from the memory cell 10 is performed by turning on the transistor 30 and the transistors 251 to 25n and measuring the resistance of a selected bit by means of a current flowing through the transistors 251 to 25n, thereby determining the contents of data.
The MTJ element may be selected to write data to it easily. On the contrary, the MTJ element may be selected to increase the uniaxial magnetic anisotropy to make it difficult to write data to it. For example, a negative potential is applied to the reference layer 23 of the selected MTJ element to make it difficult to write data to it. In this case, data is written only to the non-selected MTJ elements.
The presence of the layer 15 disposed between the MTJ element and the conductive layer 12a in the second embodiment improves the current density of the write current, thereby improving the write efficiency as in the first embodiment. Furthermore, the variation in coercive force Hc is reduced. Since the layer 15 acts as an etching stopper of the conductive layer 12a, a magnetic memory with a thin conductive layer may be provided.
In the first and second embodiments and their modifications, the longitudinal direction of the MTJ elements is substantially perpendicular to the direction of the current flowing through the conductive layer 12a. If the magnetization direction in the storage layer or the reference layer is the vertical direction, the aspect ratio of the MTJ element does not need to be changed. If the magnetization direction is parallel to the plane, the longitudinal direction of the MTJ element may be inclined relative to the direction of the current flowing through the conductive layer 12a. If the inclined angle A is more than 30 degrees and less than 90 degrees, the write current may be reduced, which is an advantageous effect. If the inclined angle A is more than 0 degree and less than 30, the write speed may be improved although the write current may not be reduced considerably. Therefore, in any case, the power consumption may be reduced.
Assuming that the minimum feature size is represented by “F” in the first embodiment and its modifications, the size of a memory cell is represented by “12F2.” In the second embodiment and its modification, however, the size of a memory cell may be reduced to 6F2. Thus, the area occupied by the memory cells may be reduced as compared with that of the first embodiment and its modifications.
Although an MTJ element is used as the memory element in the first and second embodiments and their modifications, a magnetoresistive element in which the nonmagnetic insulating layer 22 is a nonmagnetic metal layer may also be used.
Hereinafter, the embodiments will be described further with reference to some examples.
Samples 1 to 14, which are memory cells according to the first embodiment shown in
Sample 1 includes a β-Ta conductive layer (SO layer) 12a with a thickness of 6.0 nm. No layer 15 is provided to Sample 1. Sample 2 includes a W conductive layer 12a having a thickness of 6.0 nm. No layer 15 is provided to Sample 2.
Sample 3 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of MgOx with a thickness of 0.95 nm is provided to Sample 3.
Sample 4 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of AIOx with a thickness of 0.9 nm is provided to Sample 4.
Sample 5 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of SiN with a thickness of 0.95 nm is provided to Sample 5.
Sample 6 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of HfOx with a thickness of 0.98 nm is provided to Sample 6.
Sample 7 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of GdOx with a thickness of 0.95 nm is provided to Sample 7.
Sample 8 includes a β-Ta conductive layer 12a with a thickness of 6.0 nm. A layer 15 of ErOx with a thickness of 0.98 nm is provided to Sample 8.
Sample 9 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of MgOx with a thickness of 0.9 nm is provided to Sample 9.
Sample 10 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of AIOx with a thickness of 0.93 nm is provided to Sample 10.
Sample 11 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of SiN with a thickness of 0.9 nm is provided to Sample 11.
Sample 12 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of HfOx with a thickness of 0.92 nm is provided to Sample 12.
Sample 13 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of GdOx with a thickness of 0.96 nm is provided to Sample 13.
Sample 14 includes a β-W conductive layer 12a with a thickness of 6.0 nm. A layer 15 of ErOx with a thickness of 0.96 nm is provided to Sample 14.
A second example will be described below. MTJ elements are prepared, which are the same as Samples 1 to 14 of the first example except for the storage layer of CoFeB that has a thickness of 1.2 nm. A write operation is performed on each MTJ element with a current caused to flow through the conductive layer (SO layer).
As can be understood from
A third example will be described. MTJ elements are prepared, which are the same as Samples 3, 4, 10, 11, and 13 of the first example except for the storage layer of CoFeB that has a thickness of 1.2 nm and the layer 15 that has various thickness. A write operation test is performed on each MTJ element with a current caused to flow through the conductive layer (SO layer).
As can be understood from
A magnetic memory according to a fourth example is prepared, which includes memory cells according to the second embodiment shown in
The region represented by “P” in
As can be understood from
Similar good characteristics may be obtained for the other types of memory cells in which the storage layer has a multilayer structure, like a CoFeB(1.2)/Cu/CoFeB(1.2) structure and a NiFe(1.2)/Ru/NiFe(0.8)/Ta(0.3)/CoFeB(0.8) structure.
In a memory cell including MTJ elements including a storage layer having a multilayer structure, the voltage applied to an MTJ element to switch the magnetization direction of the storage layer has the same absolute value and the opposite polarity to the voltage applied to another MTJ element not to switch the magnetization direction of the storage layer. For example, a negative voltage −V is applied to the reference layer not to switch the magnetization direction of the storage layer of an MTJ element, and a positive voltage +V is applied to the reference layer to switch the magnetization direction of the storage layer of an MTJ element. It is found that this increases the margin further.
An MTJ element having a perpendicular magnetization is formed. A memory cell including an MTJ element 20 with a monolayer storage layer 21 having perpendicular magnetization, and memory cells each including an MTJ element 20 with a multilayer storage layer 21 having perpendicular magnetization are prepared. The monolayer storage layer 21 is formed of CoFeB. Five types of monolayer storage layer 21 having a multilayer structure are formed. For example, a first multilayer structure is Co(Fe)(B)/Pt/Co(Fe)(B), a second multilayer structure is Co(Fe)(B)/Pd/Co(Fe)(B), a third multilayer structure is Co(Fe)(B)/Ni/Co(Fe)(B), a fourth multilayer structure is Co(Fe)(B)/Ni/Co(Fe)(B), and a fifth multilayer structure is CoPt/Ru/CoPt multilayer/(Ta, W, Mo)/CoFeB. The same tendencies as in the case of the memory cells including MTJ elements with in-plane magnetization shown in
The first and second embodiments and their specific examples have been described. However, the present invention is not limited to these specific examples. For example, the scope of the present invention includes MTJ elements and SO layers for which those skilled in the art suitably select a specific material, a specific thickness, a specific shape, a specific size, etc. to obtain the same effect as the present invention.
A magnetic memory according to a third embodiment will be described with reference to
Each memory cell MC corresponds to the memory cell 10 of the magnetic memory according to the first embodiment shown in
A first terminal of the magnetoresistive element 20 is connected to the conductive layer 12a via a layer 15, and a second terminal is connected to one of the source and the drain of the transistor 25. The other of the source and the drain of the transistor 25 is connected to the bit line BL1, and the gate is connected to the word line WL1. A first terminal (terminal 13a in
A method of writing data to a memory cell will be described below. First, the word line selection circuit 110 applies a high-level potential to the word line WL2 connected to the gate of the transistor 30 of the memory cell MC to which data is to be written, to turn on the transistor 30. At this time, the transistors 30 of other memory cells MC in the same column as the above memory cell MC are also turned on. However, a low-level potential is applied to the word line WL1 connected to the gates of the transistors 30 of the other memory cells MC in the same column as the above memory cell MC and the word lines WL1 and WL2 corresponding to the other columns.
Thereafter, the bit line selection circuits 120a and 120b select the bit lines BL2 and BL3 connected to the memory cell MC to which data is to be written. The write circuits 130a and 130b cause a write current to flow through the selected bit lines BL2 and BL3 from one of the bit line selection circuit 120a and the bit line selection circuit 120b to the other. The write current causes the magnetization direction of the storage layer 21 (
Next, a method of reading data from a memory cell will be described below. First, a high-level potential is applied to the word line WL1 connected to a memory cell MC from which data is to be read, to turn on the transistor 25 of the memory cell MC. At this time, the transistors 25 of the other memory cells MC in the same column as the memory cell MC from which data is to be read are also turned on. However, a low-level potential is applied to the word line WL2 connected to the gate of the transistor 30 of the memory cell MC from which data is to be read and the word lines WL1 and WL2 corresponding to the other columns.
Thereafter, the bit line selection circuits 120a and 120b select the bit lines BL1 and BL3 connected to the memory cell MC from which data is to be read. The readout circuits 140a and 140b cause a read current to flow through the selected bit lines BL1 and BL3 in a direction from one of the bit line selection circuit 120a and the bit line selection circuit 120b to the other. At this time, whether the magnetization direction of the storage layer 21 (
The word line selection circuit 110, the bit line selection circuits 120a and 120b, the write circuits 130a and 130b, and the readout circuits 140a and 140b are included in the control circuit described in the descriptions of the first and second embodiments.
Like the first embodiment, the current density of the write current flowing through the conductive layer 12a in the third embodiment is improved. As a result, the write efficiency may be improved. Furthermore, the variation in coercive force Hc is reduced. Since the layer 15 acts as an etching stopper of the conductive layer 12a, a magnetic memory with a thin conductive layer may be provided.
While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel methods and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the methods and systems described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.
Number | Date | Country | Kind |
---|---|---|---|
2016-153898 | Aug 2016 | JP | national |
This application is a divisional application of U.S. application Ser. No. 15/445,475, filed on Feb. 28, 2017, which is based upon and claims the benefit of priority from prior Japanese Patent Application No. 2016-153898 filed on Aug. 04, 2016 in Japan, the entire contents of which are incorporated herein by reference.
Number | Date | Country | |
---|---|---|---|
Parent | 15445475 | Feb 2017 | US |
Child | 15875549 | US |