MTJ element for STT MRAM

The disclosed device relates to the use of Perpendicular Magnetic Anisotropy (PMA) in Magnetic Tunnel Junction (MTJ) elements in a Spin Transfer Torque (STT) Magnetic Random Access Memory (MRAM).

In current practice PMA STT MRAMs, magnetizations of both the MTJ Free Layer (FL) and the Reference Layer (RL) are perpendicular to the plane of MTJ layers and switch between low-resistivity parallel (Rp) and high-resistivity antiparallel (Rap) states which represent logical “0” and “1,” respectively. To effect proper writing operations, the PMA of the FL should be high enough to provide thermal stability for data retention while also being low enough to permit STT switching, and the PMA of the RL should be higher than that of FL—high enough to keep the magnetization of RL fixed and protected from STT disturbance or switching.

For exercising a proper reading operation, the resistance difference between the two memory states, expressed as the MTJ Magneto-resistance Ratio (MR=(Rap−Rp)/Rp), should be high enough to provide sufficient readback signal. It is therefore important that MTJ elements combine strong well-controlled PMA with high MR.

The MTJ elements disclosed to date have utilized a (100) oriented MgO layer as the tunnel barrier, since (100) is the orientation that MgO naturally assumes during deposition/oxidation. This MgO layer is then in contact, at both top and bottom, with ferromagnetic layers that, during the subsequent annealing step, will adopt the crystalline orientation of the MgO layer.

A commonly used configuration is [amorphous CoFeB]/MgO/[amorphous CoFeB]. Upon annealing, crystallization progresses from the MgO interfaces, transforming the structure to crystalline [bcc (100) CoFeB]/(100) MgO/[bcc (100) CoFeB] which is then used to effect the MTJ MR device.

Although such a system may possess some PMA due to interfacial anisotropy between MgO and CoFeB, it is necessary to enhance such PMA by adding some bulk PMA materials and/or multilayers. However, the best choices for enhancing the PMA (such as [Pt/Co(Fe)], [Pd/Co(Fe)], [Ni/Co] magnetic layers) “prefer” a fcc (111) crystal orientation which, unfortunately, is incompatible with the bcc (100) orientation induced by the MgO.

As a result, although their detailed materials selections may vary, state of the art MTJ elements are variations of the following two general structures (in which the FL can be either above or below the MgO layer) which are illustrated in FIGS. 1 a and 1 b:

Shown in FIG. 1 a is: Bottom layer(s) 10 a/[ fcc (11 1 ) PMA layer 11 a]/[ bcc (100) MR layer 13 a]/ (100) MgO/[bcc (100) MR layer 14 a]/[ fcc (111) PMA layer 16 a]/ Cap layer 17 . Note that an optional transitional layer between layers 14 a and 16 a is not shown. Shown in FIG. 1 b is: Bottom layer(s) 10 b/[ fcc (111) PMA layer 11 b]/[ bcc (100) MR layer 13 b]/ (110) MgO/[bcc (111) MR layer 14 b]/[ fcc (111) PMA layer 16 b]/ Cap layer 17 . Note that an optional transitional layer between layers 14 b and 16 b is not shown.

In that, in these structures as a matter of design choice, the FL can be either above or below the MgO layer, the FL and RL can be either simple or Synthetic Antiferromagnetic (SAF) layers, and an additional magnetic Dipole Layer (DL) could be added above or below that stack to improve the symmetry of FL switching.

Some of the layers in FIGS. 1 a or 1 b may be present, or not, and some materials selections may be different (e.g., L10 ordered phase PMA material instead of fcc (111) PMA), but one important feature remains common:

In at least in one place in the structure, below or above the MgO or both, there is a transition between two different and incompatible crystalline orientations. This mismatch leads to several drawbacks in current practice MTJs as follows:

(a) if the “fcc (111) PMA layer” to “bcc (100) MR layer” transition is part of the bottom layers (i.e. below the MgO as shown in FIG. 1 a ), neither the (111) nor (100) textures can fully develop at their interface, and the PMA of the PMA layer and the MR of the MR layer are both weakened;

(b) introduction of a transitional layer to mitigate drawback (a) is of only limited help, since the transitional layer decouples the PMA layer from the MR layer making it more difficult for the MR layer to remain in PMA mode; and (c) if the “fcc (111) PMA layer” to “bcc (100) MR layer” transition appears in the top layers (i.e., above the MgO as also shown in FIG. 1 a ), the situation is even worse due to lack of proper seeding for the “fcc (111) PMA layer” (the role played by “underlayer” in the bottom layers' case) makes it very difficult to grow the PMA layer on top of the MR layer. This is in addition to drawbacks (a) and (b) above, which still apply.

All these drawbacks are absent from the device that will be disclosed below.

It has been an object of at least one embodiment of the present disclosed device to eliminate all transitions between two different and incompatible crystalline orientations at all internal interfaces of a MTJ structure.

Another object of at least one embodiment of the present disclosed device has been to achieve the foregoing object without the introduction of transitional layers to mitigate the negative effects of a transition between two different crystalline orientations.

Still another object of at least one embodiment of the present disclosed device has been to ensure that the layer on which the magnesia tunnel barrier layer is deposited has the correct crystal orientation for it to serve as a proper seed.

A further object of at least one embodiment of the present disclosed device has been to provide a process for achieving the foregoing objects.

These objects have been achieved by the utilization of a single (111) crystalline orientation for all active layers i.e. the PMA layers, MR layers, and the MgO tunnel barrier layer. This is as opposed to current practice MTJs in which the (111) PMA layers are combined with the (100) MR layers and the (100) MgO tunnel barrier, which results in deteriorated crystalline growth and magnetic performance of both the (111) and (100) layers.

FIG. 1 a shows an MTJ of the current practice in which the MgO layer has a 100 orientation. Note that an optional transitional layer between layers 14 a and 16 a is not shown.

FIG. 1 b shows an MTJ of the current practice in which the MgO layer has a 110 orientation. Note that an optional transitional layer between layers 14 a and 16 a is not shown.

FIG. 2 shows the general layer structure of the disclosed device.

FIG. 3 shows of the detailed layer structure of one implementation of the disclosed device.

FIG. 4 shows general magnetization curves for the disclosed structure.

The structure that is disclosed below is an all (111) MTJ stack, in which there are no transitions between different crystalline orientations since all PMA layers, MR layers, and the MgO tunnel barrier have the same (111) crystalline orientation. This is accomplished by providing strongly (111)-textured layers immediately below the MgO layer which induce a (111) orientation therein. This is in contrast to the current methodology of growing the MgO on an amorphous template and allowing the MgO's (100) orientation to progress from there into the MR layers.

The general layer structure of the disclosed device is illustrated in FIG. 2 as follows:

(111) seed layer(s) 21/[fcc (111) PMA layer 22]/[fcc (111) MR layer 23]/(111) MgO/[fcc (111) MR layer 25]/[fcc (111) PMA layer 26]/Cap layer 27

In this structure, as a matter of design choice, the FL can be either above or below the MgO layer, the FL and RL can either be simple or Synthetic Antiferromagnetic (SAF) structures. An additional ferromagnetic layer to serve as a Dipole Layer (DL) may be added above or below the MTJ stack to improve the symmetry of FL switching. Depending on the particular implementation, some of the layers shown in FIG. 2 may or may not be present, and/or a single layer that serves both PMA and MR purposes may replace separate PMA and MR layers.

The disclosed MTJ eliminates the drawbacks of current practice MTJ devices listed earlier. In particular, it enables the growth of fully crystallized and properly oriented PMA and MR layers both below and above the MgO tunnel barrier, thereby benefiting both the magnetic design as well as writing and reading performance of the PMA STT MRAM devices.

Examples of Some Specific Structures:

Referring now to FIG. 3 , the following is an example of the detailed layer structure of one implementation of the disclosed device (layers listed starting from the bottom).

Ta 31 50 Å/NiCr 32 50 Å/[Co 33 2.5 Å/Ni 34 6 Å]×7/Co 35 4 Å/Ru 36 4 Å/Co 37 4 Å/[Co 38 2.5 Å/Ni 39 6 Å]×6/Co 41 2. Å/NiFe 42 6 Å/MgO 11 Å/NiFe 43 6 Å/[Co 44 2.5 Å/Ni 45 6 Å]×6/NiCr 46 50 Å/Ru 47 50 Å

Additional Comments:

NiCr underlayer 32 induces a strong (111) texture in the [Co/Ni] PMA multilayer 33-35 that is grown on it. The entire [Co/Ni]/Ru/[Co/Ni] SAF structure (33-41) has a strong (111) orientation and functions as the (111) PMA layer for the RL. The fcc (111) NiFe layer on top of “the (111) PMA layer is the actual RL. This compound (111) RL induces a (111) texture in the MgO tunnel barrier. Fcc (111) NiFe 43, on top of the MgO, is the actual FL while the (111) [Co/Ni] PMA multilayer 44 is the “(111) PMA layer” of the FL.

The structure was annealed at 300° C. for 10 min. Magnetic properties of the annealed structure were measured with the VSM both in the “in-plane” and “out-of-plane” field configurations, as shown in FIG. 4 . Both the RL and the FL of the disclosed MTJ exhibit strong PMA properties.

It is important to note that if the (111) NiFe layers adjacent to MgO had been replaced by amorphous CoFeB layers (as in the current practice), the PMA in the top [Co/Ni] multilayer would have been lost. This is because the CoFeB/MgO/CoFeB part of the structure recrystallizes upon anneal into the (100) orientation, no longer providing a suitable template for the [Co/Ni] multilayer to grow on in the (111) orientation preferred for the PMA.

Magnetization curves for the structure detailed above are shown In FIG. 4 . Both the major loop (continuous line) and the minor loop (broken line) were measured. The major loop shows that the magnetization of the entire structure is in the perpendicular direction while the minor loop shows the switching of the free layer occurs only with full remanence in the perpendicular direction at zero applied field.

It should be noted that any of many possible variations in the design and materials of the MTJs could be used while still remaining within the scope and spirit of the disclosed device. Examples include, but are not limited to:

In the design, (a) the FL and RL can either be above or below the MgO, (b) the FL and/or RL can be either simple layer(s) or SAF layers, (c) an additional magnetic DL can be added below or above the stack, and (d) the dedicated MR layer(s) can be eliminated if the PMA layer(s) provide sufficient (111) texture and the desired MR. For the materials, (a) the [Co/Ni] multilayer can be replaced by [Co(Fe)/Ni(Fe)], [Co(Fe)/Pd], [Co(Fe)/Pt] or similar multilayers, (b) the fcc NiFe MR layer(s) can be replaced by fcc CoFe or CoNiFe that have the right compositions for the fcc phase, and (c) the NiCr underlayer and/or cap can be replaced by other (111) seeding materials such as, but not limited to, Pt, Pd, Cu, Rh, and Ir. In particular, for the foregoing structure some may prefer to replace the NiFe layers adjacent to the MgO with fcc CoFe (e.g., CoFe 10 ) because of the higher MR provided by CoFe relative to NiFe.

Additional Benefits Provided by the Disclosed Device Include:

(i) the FL thickness, PMA strength, and Eb for data retention can be easily adjusted/increased, since the FL PMA derives from the ‘bulk’ PMA distributed throughout the FL thickness (e.g., that of the [Co/Ni] multilayer) rather than the interfacial PMA generated only at the top and/or bottom FL interface(s);

(ii) similar benefits apply to the MTJ RL; and

(iii) a simplified set of materials can be used without degrading the MTJ properties, thereby making the structure more robust against high-temperature processing, including the 400° C. integration process.