Synthetic antiferromagnetic layer, magnetic tunnel junction and spintronic device using said synthetic antiferromagnetic layer

Abstract

A synthetic antiferromagnetic layer includes a first ferromagnetic layer containing an amorphizing element, the first ferromagnetic layer having a first structural symmetry; a second ferromagnetic layer having a second structural symmetry; wherein the first and the second ferromagnetic layers are antiferromagnetically coupled by a trifunctional non-magnetic multi-layered structure, the antiferromagnetic coupling being an Ruderman-Kittel-Kasuya-Yosida (RKKY) coupling, the non-magnetic multi-layered structure including at least two non-magnetic layers, the non-magnetic multilayered structure being at least partially nano-crystalline or amorphous in order to ensure a structural transition between the first ferromagnetic layer having the first structural symmetry and the second ferromagnetic layer having the second structural symmetry, the non-magnetic multilayered structure being adapted to absorb at least part of the amorphizing element out of the first ferromagnetic layer in contact with the non-magnetic multi-layered structure.

Claims

1. A synthetic antiferromagnetic layer comprising: a first ferromagnetic layer containing an amorphizing element, said first ferromagnetic layer having a first structural symmetry; a second ferromagnetic layer having a second structural symmetry; wherein the first and the second ferromagnetic layers are antiferromagnetically coupled by a trifunctional non-magnetic multi-layered structure, the antiferromagnetic coupling being an Ruderman-Kittel-Kasuya-Yosida (RKKY) coupling, said non-magnetic multi-layered structure comprising at least two non-magnetic layers, said non-magnetic multilayered structure being at least partially nano-crystalline or amorphous in order to ensure a structural transition between the first ferromagnetic layer having the first structural symmetry and the second ferromagnetic layer having the second structural symmetry, said non-magnetic multilayered structure being adapted to absorb at least part of the amorphizing element out of the first ferromagnetic layer in contact with the non-magnetic multi-layered structure.

2. The synthetic antiferromagnetic layer according to claim 1, wherein the non-magnetic multilayered structure comprises at least a first and a second non-magnetic layer, the first non-magnetic layer comprising one of the following elements Ru, Re, Rh, Cu, Ir, Os or an alloy thereof and/or the second non-magnetic layer comprising one of the following elements W, Mo, Nb, Cu, Ta, V, Cr or an alloy thereof.

3. The synthetic antiferromagnetic layer according to claim 1, wherein a thickness of the first non-magnetic layer is lower than 2 nm.

4. The synthetic antiferromagnetic layer according to claim 3, wherein the thickness of the first non-magnetic layer is comprised between 0.2 nm and 0.9 nm.

5. The synthetic antiferromagnetic layer according to claim 1, wherein a thickness of the second non-magnetic layer is lower than 1 nm.

6. The synthetic antiferromagnetic layer according to claim 5, wherein the thickness of the second non-magnetic layer is comprised between 0.1 nm and 0.6 nm.

7. The synthetic antiferromagnetic layer according to claim 1, wherein the non-magnetic multi-layered structure comprises n bilayers, each bilayer comprising a first and a second non-magnetic layers.

8. A magnetic tunnel junction comprising the synthetic antiferromagnetic layer according to claim 1, a first tunnel barrier layer, a storage layer and a cap layer and/or a second tunnel barrier layer added in contact with the storage layer on its interface opposite to the first tunnel barrier layer.

9. The magnetic tunnel junction according to claim 8, wherein the first and second tunnel barrier layers are MgO layers.

10. The magnetic tunnel junction according to claim 8, wherein the storage layer is a composite storage layer, said composite storage layer comprising a stack formed by at least a first ferromagnetic layer, a non-magnetic layer or multilayer and a second ferromagnetic layer, said first and second ferromagnetic layers comprising at least one of the following alloys CoFeB, FeB, CoFe and the non-magnetic layer or multilayer being adapted to absorb the amorphizing element and comprising at least one of the following elements Ta, W, Mo, Ru or an alloy thereof.

11. The magnetic tunnel junction according to claim 8, comprising an additional magnetic layer added in contact with the second tunnel barrier layer and/or an additional synthetic antiferromagnetic layer in contact with the second tunnel barrier layer.

12. The magnetic tunnel junction according to claim 11, wherein the additional synthetic antiferromagnetic layer comprises an additional trifunctional non-magnetic multi-layered structure.

13. The magnetic tunnel junction according to claim 11, wherein the storage layer is formed by a synthetic antiferromagnetic layer.

14. A spintronic device comprising a synthetic antiferromagnetic layer SAF according to claim 1.

Description

BRIEF DESCRIPTION OF THE FIGURES

(1) Other characteristics and benefits of the invention will be evident from the description thereof that is provided in the following, purely for non-limiting, exemplary purposes, with reference to the accompanying drawing, in which:

(2) FIG. 1 shows a schematic representation of a typical MTJ stack used in the prior art.

(3) FIG. 2 shows a schematic drawing of the layer configuration of a perpendicular synthetic anti-ferromagnetic SAF layer or pSAF using a trifunctional antiferromagnetic coupling TriAF layer according to an embodiment of the invention.

(4) FIG. 3 shows the magnetization M of a SAF according to the invention as a function of the applied field H. Only one branch of the hysteresis loop is shown here corresponding to the field decreasing from positive to negative saturation.

(5) FIG. 4 shows the RKKY coupling energy of the SAF versus thicknesses of the layers constituting the trifunctional antiferromagnetic coupling layer TriAF layer.

(6) FIGS. 5a and 5b show a schematic drawing of layer configuration of perpendicular magnetic tunnel junction MTJ or pMTJ stack with the proposed synthetic antiferromagnetic SAF layer using a trifunctional coupling layer TriAF according to an embodiment of the invention. Two configurations are shown: (a) the storage layer SL is on top and (b) opposite of first, i.e. SAF layer is on top. The arrows in the diagrams indicates magnetization configuration at zero applied field.

(7) FIGS. 6a and 6b show a second layer configuration of perpendicular magnetic tunnel junction pMTJ stack with a synthetic antiferromagnetic layer SAF according to an embodiment of the invention comprising a trifunctional non-magnetic layered structure TriAF according to the invention.

(8) FIG. 7 shows the magnetic hysteresis loop of a perpendicular MTJ stack according to FIG. 2 after annealing at 400 C.

(9) FIG. 8 shows a third layer configuration of a dual perpendicular magnetic tunnel junction DMTJ stack with two synthetic antiferromagnetic layers SAF according to an embodiment of the invention each comprising a non-magnetic layered structure TriAF and the storage layer being a composite storage layer CSL comprising two ferromagnetic layers ferromagneticaly coupled through a thin B absorbing layer.

(10) FIG. 9 shows a fourth layer configuration of dual perpendicular magnetic tunnel junction DMTJ stack with two synthetic antiferromagnetic layers SAF according to an embodiment of the invention each comprising a non-magnetic layered structure TriAF according to the invention and the storage layer SAF SL being a SAF comprising two ferromagnetic layers antiferromagneticaly coupled through a thin antiferromagnetic coupling layer.

DETAILED DESCRIPTION

(11) FIG. 2 shows a first example of a synthetic antiferromagnetic SAF layer comprising a trifunctional non-magnetic multi-layer structure TriAF according to an embodiment of the present invention.

(12) FIG. 2 shows that the synthetic antiferromagnetic layer SAF according to an embodiment of the invention comprises: a first ferromagnetic layer PL having a first crystalline structure after annealing; a non-magnetic multilayer structure TriAF; a second ferromagnetic layer HL having a second crystalline structure.

(13) The magnetic stack shown in FIG. 2 comprises the following layers (from bottom to top): a substrate layer SUB, a seed layer on top of the SUB layer, the synthetic antiferromagnetic layer SAF according to an embodiment of the invention, a tunnel barrier TB on the top of the SAF layer, a storage layer SL and a cap layer above the storage layer. This forms a magnetic tunnel junction MTJ.

(14) In the case where the layers constituting the synthetic antiferromagnetic layer SAF are magnetized out-of-plane, the structure is called a perpendicular synthetic antiferromagnet pSAF.

(15) The first ferromagnetic layer PL can act as a polarizer or reference layer. In the example of FIG. 2 the polarizing layer PL can be a CoFeB alloy layer or a CoFe based alloy layer or an Heusler alloy layer or a combination of such layers.

(16) In the case of out-of-plane magnetized MTJs, the hard layer HL comprises n bilayers of A/X nano-laminates which have intrinsic perpendicular magnetic anisotropy PMA originated from A/X interfaces via electron hybridization effect or interfacial strain. The material for A can be made of Co, Fe, CoFe, CoFeB or an alloy thereof, and X is Pt, Au, Ir, Pd, Ni, NiCo, NiFe or NiFeCo based alloys. Similar PMA can be obtained if A and X are interchanged to form X/A nano-laminates.

(17) Still in the case of out-of-plane magnetized MTJs, the hard layer HL may also be a material exhibiting bulk PMA, for example L1.sub.0 ordered alloy of FePt, FePd and CoPt; L1.sub.1 ordered alloy CoPt, CoPd, CoNiPt. It may also be an alloy of rare earth-transition metals in the form of RE-TM where RE stands for Tb, Dy, Er, Gd or an alloy thereof, and TM stands for Fe, Co or an alloy thereof. It can also be a Heusler alloy with perpendicular anisotropy such as Mn2.7Co0.3Ga.

(18) In the case of in-plane magnetized MTJs, as well-known by the man skilled in the art, the hard layer HL is, for example, an exchange biased layer i.e. a ferromagnetic layer essentially made of a CoFe alloy 1.5 to 3 nm thick which is exchange coupled to an antiferromagnetic layer most often made of IrMn typically between 5 nm and 8 nm thick or PtMn typically 10 nm to 20 nm thick.

(19) These hard layers HL are actually similar to those used in the prior art.

(20) The trifunctional coupling layer TriAF of the synthetic antiferromagnetic layer SAF according to an embodiment of the invention ensures three functions: the first is to provide antiferromagnetic coupling between the hard layer HL and the reference layer PL. The second function is to ensure a structural transition between two layers of different structural symmetry for instance here the fcc structure of the hard layer HL and the bcc structure of the reference or polarizer layer PL. The third function is absorption of boron from CoFeB polarizer layer PL upon annealing.

(21) The HL and the polarizer layer PL are antiferromagnetically coupled by the trifunctional anti-ferromagnetic TriAF coupling layer. According to an embodiment the TriAF coupling layer consists of a C/T bilayer where C is one of Ru, Re, Rh, Ir, Cu and Os or an alloy thereof. The thickness of this C layer is adjusted in the range 0.2 nm-0.9 nm together with that of the T layer to provide a strong RKKY interaction between HL and PL through the C/T composite layer. T can be one of W, Mo, Nb, V, Ta, Cr or an alloy thereof, which functions as a symmetry breaking layer allowing the structural transition from the HL layer.

(22) This is particularly useful because the hard layer HL and the polarizer layer have usually different crystal structures. For example, the hard layer HL can have a fcc crystal structure as for Co/Pt type multilayers or exchange biased IrMn/CoFe bilayers, or tetragonal as with FePt L1.sub.0. The polarizer layer PL can have a bcc (001) crystal structure like CoFeB.

(23) The T layer can also act as inter-diffusion barrier between the C layer and the PL layer and/or as a partial B getter layer. The thickness of the T layer is also adjusted in the range (0.1-0.6 nm) together with that of the C layer to provide strong antiferromagnetic coupling between HL and PL through the C/T composite layer. The TriAF coupling layer can be a multilayer formed with (C/T)n configuration with T on top. This means that the TriAF layer is formed by n bilayers of C/T bilayers. Since the C/T bilayers play the three functions of coupling, symmetry breaking and B absorption, this stack of layers acts as a trifunctional anti-ferromagnetic coupling layer (TriAF).

(24) The tunnel barrier can be one of MgO, AlOx, TaOx, TiO2, HfOx, ZrOx, GaOx, SrTiO3 or other metal oxide on top of pSAF layer which generate interface perpendicular magnetic anisotropy with PL and SL and provides large TMR amplitude.

(25) FIG. 3 shows the variation of the magnetization M of a synthetic antiferromagnetic layer SAF as a function of the applied magnetic field H varying from positive saturation to negative saturation. This represents the descending branch of the M-H hysteresis loop. For clarity, the ascending branch is not represented. The ascending branch would be symmetric of the descending one with respect to the origin of the graph. The y axis represents the measured magnetization expressed in electromagnetic unit and the x axis represents the applied magnetic field in Oe.

(26) The SAF/MgO/cap measured sample has the composition Pt/(Co/Pt)3/Co/Ru 4 /W 2 /CoFeB 10 /MgO/Cap and was annealed at 340 C. The M-H loop was measured by vibrating sample magnetometer VSM. Along this descending branch of hysteresis loop, the field is decreased from the positive saturation to the negative saturation. Three magnetization reversal steps are observed while decreasing the field from positive to negative saturation as described below. All the magnetization reversal steps with the sweeping of magnetic field are sharp with high squareness. This signifies that the SAF layer has perpendicular anisotropy. Coming from high saturation positive field, the first reversal around 4280 Oe is the characteristic field H.sub.0+ at which the RKKY coupling energy gain associated with bringing the HL and the polarizing 10 CoFeB layer PL magnetizations in antiparallel alignment balances the cost in Zeeman energy associated with reversing the PL magnetization in the direction opposite to the applied field. Below this characteristic field, the magnetic moments across the TriAF coupling layers are aligned antiparallel. When the field is reversed and increased negatively, another characteristic field around 1000 Oe corresponds to the simultaneous reversal of both PL and HL magnetization. This characteristic field can be viewed as the coercive field of the net moment formed by the PL and HL antiferromagnetically coupled sandwich. Finally, at H.sub.0=4450 Oe, the Zeeman energy exceeds the RKKY coupling energy so that the two magnetic moments again switch to parallel configuration. Both HL's and PL's moments are then completely saturated along the negative field direction. In macrospin approximation the field H.sub.0+ and H.sub.0 would be expected to be identical in absolute value. However, in reality, they slightly differ (4280 Oe vs 4450 Oe) because the switching process is different in the two transitions.

(27) At 4280 Oe, the switching of the CoFeB layer at 4280 Oe is driven by the RKKY antiferromagnetic coupling against the applied field whereas the switching at 4450 Oe is driven by the applied field against the antiferromagnetic RKKY coupling. A characteristic field Ho can then be defined as (H.sub.0+H.sub.0)/2 called inter-layer exchange coupling field or RKKY coupling field. It is equal here to 4365 Oe.

(28) The RKKY coupling energy (J.sub.RKKY) has been calculated by using the equation below,
J.sub.RKKY=H.sub.oM.sub.st
where M.sub.s and t are the saturation magnetization and thickness of the CoFeB PL. The coupling energy depends on the thickness of C and T of the TriAF coupling layer.

(29) FIG. 4 shows the RKKY coupling energy of the SAF versus thicknesses of the different layers of the TriAF layer. The material configuration of the SAF is Pt/(Co/Pt)3/Co/Ru (x)/W (y)/CoFeB 10 /MgO, where x and y respectively represent the thicknesses of Ru and W in Angstroms.

(30) In particular, FIG. 4 shows the variation of coupling energy as a function of Ru thickness for different thicknesses of W in Angstroms. The y axis of FIG. 4 represents the RKKY coupling energy expressed in erg/cm.sup.2. The x axis represents the thickness of the Ru layer. Each set of points in FIG. 4 represents a different thickness of the W layer.

(31) Depending on the thickness of the TriAF coupling layer, the coupling energy can be tuned over a broad range. For all W thicknesses, the peak in coupling energy appears with Ru thickness in the range of 0.4 to 0.45 nm which is the thickness region for the first peak of the oscillatory curve of RKKY coupling energy.

(32) Using the synthetic antiferromagnetic layer SAF with the non-magnetic multilayered structure TriAF, the MTJ stack (either in-plane or out-of-plane magnetized) can be configured as top free or bottom free layer configuration as shown for example in FIGS. 5a and b respectively in the case of out-of-plane magnetized MTJ.

(33) The top free layer configuration of the magnetic tunnel junction MTJ of FIG. 5a comprises the following layers from the bottom to the top of the stack: a substrate layer SUB, a seed layer, a perpendicular antiferromagnetic synthetic layer pSAF according to an embodiment of the invention, a MgO tunnel barrier, a storage layer SL and a cap layer.

(34) The pSAF according to an embodiment of the invention showed in FIG. 5a comprises the following layers from bottom to top: a hard layer HL on the top of the seed layer of the MTJ, a non-magnetic multi-layered structure TriAF on the top of the HL layer and a polarizing layer PL on the top of the TriAF layer, the PL layer being below the MgO tunnel barrier of the MTJ.

(35) In top free layer configuration showed in FIG. 5a, the pSAF layer is below the MgO tunnel barrier whereas the storage layer SL is above it. In the bottom, free layer configuration shown in FIG. 5b, the opposite is true i.e. the free layer SL is below the MgO barrier and the pSAF is above it.

(36) In particular the bottom free layer configuration of the magnetic tunnel junction MTJ of FIG. 5b comprises the following layers from the bottom of the stack to the top of the stack: a substrate layer SUB, a seed layer, a storage layer SL, a MgO tunnel barrier, a perpendicular antiferromagnetic synthetic layer pSAF according to an embodiment of the invention and a cap layer.

(37) The pSAF according to an embodiment of the invention shown in FIG. 5b comprises the following layers from bottom to top: a polarizing layer PL placed on top of the MgO layer of the MTJ, a non-magnetic multi-layered structure TriAF on top of the polarizing layer PL and a hard layer HL on top of the TriAF layer, the HL layer being below the Cap layer of the MTJ.

(38) This configuration with bottom storage layer SL is in particular adapted for SOT-MRAM since the free layer SL has to be in contact with a horizontal conducting line having high spin orbit constant. As known by the man skilled in the art, the MgO tunnel barrier can be prepared by RF deposition of MgO from an MgO target or by natural oxidation of a metallic Mg layer. The free layer SL can be made of CoFeB alloy or be a composite layer consisting of a stack of several layers of different (Co1xFex)1yBy compositions (x between 0 and 1 and y between 0 and 0.3). In perpendicular MTJ, as known by the man skilled in the art, provided this layer is thin enough (1.6 nm), its magnetization is pulled out-of-plane by the interfacial anisotropy which exists at the interface between the storage layer and the MgO barrier. But other materials exhibiting perpendicular anisotropy and low Gilbert damping (0.02) could be used as storage layer or free layer SL in particular among the Heusler alloys.

(39) In FIG. 5a (top free layer configuration), the cap layer Cap is in contact with the storage layer SL. It is a non-magnetic metal which has the ability to absorb the B from the storage layer SL upon annealing. It can be made of Ta but also Mo, W, Ru or other non-magnetic species having the ability to absorb B out of CoFeB upon annealing. Concerning the TriAF layer which is coupling the hard layer HF to the polarizer layer PL, it can be here of the form C/T with the Boron getter layer T above the C layer i.e. in contact with the Boron doped electrode.

(40) In the bottom free layer configuration (FIG. 5b), the cap layer is Cap in contact with the hard layer HL. In this case, since the hard layer HL usually does not contain B, the requirement for this cap layer Cap to be able to absorb B is relieved. The cap layer Cap can still be made of Ta, Ru, W, Mo with other existing possibilities such as Cu, Ti, Au, Pd, Pt, Hf. Concerning the TriAF layer which is coupling the hard layer HL to the polarizer layer PL, it can be here of the form T/C with the Boron getter layer T below the C layer i.e. in contact with the Boron doped electrode.

(41) Similar stacks with bottom free or top free layer can be considered with in-plane magnetized materials. In this case, the HL is an exchange biased layer as previously described and the storage layer SL is chosen with in-plane magnetization. This can be achieved by using also CoFeB layer but with a larger thickness than in the perpendicular anisotropy case so that the easy-plane demagnetizing energy exceeds the interfacial perpendicular MgO/CoFeB anisotropy resulting in an in-plane magnetization of the layer.

(42) FIGS. 6 (a) and (b) show another embodiment of the magnetic tunnel junction MTJ according to the invention.

(43) The magnetic tunnel junction MTJ of FIG. 6a comprises the following layers from bottom to top: a substrate layer SUB, a seed layer, a perpendicular synthetic antiferromagnetic layer pSAF according to an embodiment of the invention, a first MgO layer, a com posited storage layer CSL, a second MgO layer and a Cap layer.

(44) The pSAF layer of FIG. 6a comprises from bottom to top: a hard layer HL on top of the Seed layer of the MTJ, a trifunctional non-magnetic multi-layered structure TriAF and a polarizer layer PL below the first MgO layer.

(45) The composite storage layer (CSL) of FIG. 6a comprises two ferromagnetic FM layers ferromagnetically coupled through a thin B absorbing layer or NM layer.

(46) The magnetic tunnel junction of FIG. 6b comprises the following layers from bottom to top: a substrate layer Sub, a seed layer, a second MgO layer, a composite storage layer CSL, a first MgO layer, a perpendicular antiferromagnetic layer pSAF according to an embodiment of the invention and a cap layer.

(47) The pSAF layer of FIG. 6b comprises from bottom to top: a polarizer layer PL above the second MgO layer, a trifunctional non-magnetic multi-layered structure TriAF and a hard layer HL below the cap layer of the MTJ.

(48) A peculiarity of these stacks is that they also comprise a second MgO layer added in contact with the composite storage layer CSL on its interface opposite to the first MgO tunnel barrier. This type of stacks is especially interesting in the case of out-of-plane magnetized MTJs since the second MgO layer allows to further increase the perpendicular anisotropy of the composite storage layer CSL. This increases the thermal stability factor of the composite storage layer CSL and therefore retention of memory devices. This also allows to increase the CSL thickness (typically from 1.3 nm to 3 nm) resulting in an enhanced TMR amplitude. It is also interesting in the case of in-plane magnetized MTJ to reduce the effective demagnetizing field thanks to the doubled interfacial anisotropy at storage layer SL/MgO interfaces which thereby reduces the switching current. In these stacks where the storage layer is sandwiched between two MgO layers, it is then a composite storage layer CSL of composition of the form FM/NM/FM. The ferromagnetic FM layer can be comprised of CoFeB, FeB, CoFe, Fe or combination. NM is generally a non-magnetic layer. The NM layer is intended to absorb the B away from the MgO interfaces. NM can be made of B absorbing materials such as Ta, W, Ru, Mo, or an alloy or MLs thereof. It may also comprise some magnetic elements such as Co and Fe to increase the magnetic coupling across it. In that case, NM may be slightly magnetic. This layer is thick enough to be able to absorb a sufficient amount of B from the adjacent CoFeB layers but is thin enough to still ensure a good magnetic coupling (ferromagnetic or antiferromagnetic depending on its thickness) between the two magnetic layers on both sides of it. Its thickness is typically in the range 0.2 nm to 1 nm. Alternatively, as about the stacks described in FIG. 5, the storage layer can be made of Heusler alloys with perpendicular anisotropy and low damping.

(49) The cap layer Cap above the 2nd MgO layer in the stack of FIG. 6 a is non-magnetic. It can be made for instance of Ru, W, Mo, Ta. Similarly, the seed layer below the second MgO layer in FIG. 6b is a non-magnetic one. It can be made of Ta or another material favoring the bcc growth of MgO such as W or Cr.

(50) As an example, a top free perpendicular MTJ stack as described in FIG. 6(a) was deposited. The material configuration for the stack was Pt/(Co/Pt)3/Co/Ru 4 /W 2 /CoFeB 10 /MgO/Composite Storage layer/MgO/Cap. Its hysteresis loop was measured by vibrating sample magnetometry (VSM) after annealing at 400 C.

(51) The measured hysteresis loop is shown in FIG. 7. The y axis of FIG. 7 represents the measured magnetization of the stack, expressed in electromagnetic unit. The x axis of FIG. 7 represents the applied magnetic field expressed in Oe. Coming from high positive field, the polarizer layer PL magnetization switches first at 3120 Oe. Then the field is reversed and with a very small negative field (40 Oe) the free layer magnetization is oriented along the field. With the increase of negative field, there is a characteristic field (1000 Oe) where magnetization of both PL and HL switch simultaneously (coercive field of the antiferromagnetically coupled sandwich formed by the HL/TriAF/PL). Then finally the PL magnetization switches at 3350 Oe to align with the applied magnetic field. The minor loop of polarized layer across the dotted line at 3240 Oe shows a sharp vertical reversal with coercivity of 130 Oe. This means that at high annealing temperature (400 C.), the perpendicular magnetic anisotropy PMA of PL and its strong RKKY coupling to the HL are preserved by the TriAF coupling layer.

(52) FIG. 8 shows another embodiment of dual magnetic tunnel junction MTJ stack comprising two proposed thin SAF layers each using a trifunctional coupling layer TriAF. In this case a top perpendicular synthetic antiferromagnetic layer SAF and a bottom perpendicular synthetic antiferromagnetic layer SAF are used.

(53) The stack of FIG. 8 comprises the following layers from bottom to top: a substrate layer Sub, a seed layer, a bottom SAF, a first MgO layer, a composite storage layer CSL, a second MgO layer, a top SAF and a cap layer.

(54) The bottom SAF of FIG. 8 comprises the following layers from bottom to top: a hard layer HL, a trifunctional non-magnetic multi-layered structure TriAF and a polarizing layer PL, the polarizing layer being below the first MgO layer.

(55) The top SAF of FIG. 8 comprises an additional polarizing layer APL, a trifunctional non-magnetic multi-layered structure TriAF and a second hard layer HL, the second hard layer HL being below the cap layer.

(56) In contrast to the stacks described in FIG. 6, the layer above the second MgO barrier is here a magnetic additional polarizing layer APL whereas it is a non-magnetic layer in the embodiment depicted in FIG. 6. The purpose of this additional polarizing layer APL is to further increase the spin transfer efficiency by combining the spin transfer torque acting on the storage layer CSL magnetization originating from the first polarizing layer PL with the spin transfer torque originating from this additional polarizing layer APL. This APL layer can be CoFeB, or other magnetic material with high spin polarization. As described previously, the polarizer layer PL can be antiferromagnetically coupled to a hard layer HL through a TriAF layer to increase its pinning energy. Similarly, the additional polarizing layer APL can be antiferromagnetically coupled to a second hard layer HL through a TriAF layer to increase its pinning energy. The advantages of using such bottom and top SAF layers is to reduce the magnetostatic coupling that they may have between them as well as with the storage layer CSL.

(57) The storage layer CSL is composite and consists of a ferromagnetic/non-magnetic/ferromagnetic or FM/NM/FM sandwich where the magnetizations of the two ferromagnetic layers are ferromagnetically coupled though the non-magnetic spacer (the latter can be made of Ta, W, Mo typically 0.1 to 1 nm thick). When the storage layer is a composite ferromagnetic layer CSL as described here, the magnetization of the CoFeB polarizer layer and of the 2nd polarizer layer is set in antiparallel alignment to obtain the maximum STT efficiency. As well known by the man skilled in the art, setting these magnetizations in antiparallel alignment can be achieved by performing minor hysteresis loops taking advantage of the difference of coercivity of the two polarizer layers.

(58) In this embodiment, as known by the man skilled in the art, the two MgO barriers have different resistance x area RA products since the changes of resistance associated with CSL magnetization switching from the bottom MgO barrier and from the top MgO barrier are opposite in sign. Indeed, when there is parallel magnetic alignment across one barrier, the other is in antiparallel alignment and vice versa. Since these two tunnel barriers can be considered as connected in series, using MgO barriers of different RA allows not to reduce excessively the net tunnel magnetoresistance of the stack. Typically, a difference of RA product by at least a factor 3 is desirable.

(59) FIG. 9 shows another embodiment of the magnetic tunnel junction MTJ according to the invention.

(60) The stack of FIG. 9 comprises the following layers from bottom to top: a substrate layer Sub, a seed layer, a bottom synthetic antiferromagnetic layer SAF, a first MgO layer, a synthetic antiferromagnetic storage layer SAF SL, a second MgO layer, a top SAF and a cap layer.

(61) The bottom SAF of FIG. 9 comprises a hard layer HL placed above the seed layer, a trifunctional non-magnetic multilayered structure TriAF, a polarizing layer PL below the first MgO layer.

(62) The top SAF of FIG. 9 comprises from bottom to top an additional polarized layer APL placed above the second MgO layer, a trifunctional non-magnetic multilayered structure TriAF and a second hard layer HL below the cap layer.

(63) Here the storage layer SAF SL, instead of being made of a composite ferromagnetic layer, is made of a synthetic antiferromagnetic layer SAF of the form ferromagnetic/non-magnetic/ferromagnetic in which the two ferromagnetic layers are antiferromagnetically coupled through the non-magnetic layer. The later can be made for instance of TriAF, Ta, W, Mo or Ru with thickness chosen so as to get antiferromagnetic coupling.

(64) As well known by the man skilled in the art, when such SAF SL storage layer is used, the polarizer layer PL and the additional polarizing layer APL are then set in parallel magnetic configuration to get the maximum spin transfer efficiency.

(65) Another object of the invention is a spintronic device comprising a synthetic antiferromagnetic SAF layer comprising a trifunctional layer TriAF according to the invention.

(66) Such a device can be for example a Spin Transfer Torque Magnetic Random Access Memory STT-MRAM, a Spin Orbit Transfer Magnetic Random Access Memory SOT-MRAM, a Spin transfer oscillator, a spin-diode, a memristor or a magnetoresistive sensor.

(67) For example, the magnetic tunnel junction of FIG. 5b is particularly adapted to a SOT-MRAM due to the accessibility of the storage layer SL.

(68) The magnetic tunnel junctions described in FIGS. 5a, 6a, 6b, 8 and 9 are particularly adapted for STT-MRAM applications.