High magnetoresistance tunnel junctions with Mg–B–O barriersand Ni–Fe–B free electrodes

J. C. Read,1,a� Judy J. Cha,1 William F. Egelhoff, Jr.,2 H. W. Tseng,1 P. Y. Huang,1 Y. Li,1

David A. Muller,1 and R. A. Buhrman1

1School of Applied and Engineering Physics and Center for Materials Research, Cornell University,Ithaca, New York 14853, USA2Magnetic Materials Group, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA

�Received 7 January 2009; accepted 16 February 2009; published online 18 March 2009�

The use of boron-alloyed electrodes with the radio frequency �rf� sputter deposition of MgO yieldsmagnetic tunnel junctions �MTJs� with Mg–B–O tunnel barriers. After annealing, such MTJs canexhibit very high tunneling magnetoresistance �TMR� in the thin ��1.0 nm� barrier regime.Scanning tunneling spectroscopy of Mg–B–O layers reveals a better defined, but smaller band gapin comparison to that of thin MgO. We produced Fe60Co20B20 /Mg–B–O /Ni65Fe15B20 MTJs whereafter a 350 °C annealing the Ni–Fe–B free electrode crystallizes into a highly textured �001�-normalbody centered cubic �bcc� crystal structure and the MTJs achieve 155% TMR. © 2009 AmericanInstitute of Physics. �DOI: 10.1063/1.3095595�

The development of high tunneling magnetoresistance�TMR� MgO magnetic tunnel junctions1–3 �MTJs� is key tothe realization of next generation magnetic random accessmemory �MRAM� and high-performance sensors for high-density data storage, biomedical, and security applications.The understanding of high TMR in MgO MTJs is based oncoherent spin-filtered tunneling due to properly orientedcrystalline electrodes and an epitaxial barrier layer.4,5

However, sputter deposition of MgO between amorphousCo–Fe–B electrodes that crystallize during annealing6

achieves the best results in the thin barrier ��1.5 nm�, lowresistance-area �RA� regime required for many important ap-plications. We find this deposition process partially oxidizesthe surface of the base electrode, generally resulting in theformation of a Mg–B–O tunnel barrier where B trigonallycoordinated with O �BO3� composes �12% of the oxidecations.7–9

In high TMR MgO MTJs, CoFe-based alloys are oftenused as the “free” electrode, but the magnetic properties ofthis material are not optimal in comparison to Permalloy�Py�, Ni81Fe19.

10 Py has essentially no magnetostriction incomparison to larger magnetostriction ��10−5� in Co80Fe20

�Ref. 11� and low microcrystalline anisotropy ��102 J /m3�relative to Co70Fe30 ��104 J /m3�.11 The saturation magneti-zation �MS� of Py is about half of that of Co80Fe20,

10 which isbeneficial12 for spin-torque �ST� MRAM,13 where the spinpolarized tunnel current reverses the free layer magnetiza-tion. ST-MRAM promises to substantially enhance circuitdensity and speed, in comparison to magnetic field switch-ing, and could become a universal memory solution. Untilnow the TMR level realized in MTJs incorporating Py freeelectrodes is considerably less ��85%,10 �100%,14 in thehigher RA regime� than the 150% level that essentially maxi-mizes the efficiency of the ST process15 that we report herein low RA MTJs.

In this letter we confirm that Mg–B–O barriers have spinfiltering properties in the low RA junction regime similar to

those of pure MgO layers, and we examine the electronicstructure of the Mg–B–O material using ultrahigh-vacuum�UHV� scanning tunneling spectroscopy �STS�. We alsodemonstrate that annealed Co–Fe–B/Mg–B–O/Ni–Fe–BMTJs have body centered cubic �bcc� textured Py free elec-trodes, and that these MTJs achieve TMR values in the lowRA, �20 � �m2, regime that are comparable to those ob-tained with less desirable CoFe free electrodes, provided thePyB alloy is sufficiently B rich. This approach simulta-neously achieves high TMR in low RA MTJs with in-planemagnetized free electrode layers with low MS, magnetostric-tion, and microcrystalline anisotropy, that should lead to im-proved ST-MRAM structures.

We grew the thin film stacks on 3 in. thermally oxidizedSi wafers in a vacuum system �Pbase�3�10−9 Torr�containing multiple magnetron sputtering sources. Thefilm structure is 5 nm Ta / �20 nm Cu�N� /3 nm Ta�x4 /15 nm IrMn /4 nm base electrode/MgO �1.1–1.7 nm�/2.5 nm top electrode/8 nm Ta/7 nm Ru. Cu�N� indicates Cureactively sputtered in an Ar /N2 atmosphere and theCu�N�/Ta multilayer stack produces a smooth but highly con-ducting base layer, necessary for current-in-plane tunneling16

�CIPT� measurement of TMR and RA. Atomic force micros-copy measurements of the base electrodes indicate a rmsroughness of 0.3–0.4 nm over a 9 �m2 area. We useCo60Fe20B20 �CFB� and Fe60Co20B20 �FCB� alloys for thereference or pinned electrode. We compare the results ob-tained with two different Permalloy-boron �PyB� alloy freelayer electrodes, Ni77Fe18B5 �Py95B5� and Ni65Fe15B20

�Py80B20�, with those obtained with symmetric junctions us-ing either CFB or FCB as both the reference and free layerelectrodes. The tunnel barrier material is formed from rfsputtering of a sintered MgO target. After deposition, we cutthe centers of the wafers into small �1 cm2 chips and an-nealed some in modest vacuum ��3�10−6 Torr� to 250,300, or 350 °C, for up to 2 h, prior to study with CIPT andscanning transmission electron microscopy utilizing electronenergy-loss spectroscopy �EELS�. For the in situ STS stud-ies, we grew CFB/MgO samples with sputtering and electronbeam deposition �EBD� in a separate vacuum system �Pbase

a�Electronic mail: jcr33@cornell.edu.

APPLIED PHYSICS LETTERS 94, 112504 �2009�

0003-6951/2009/94�11�/112504/3/$25.00 © 2009 American Institute of Physics94, 112504-1

Downloaded 08 Dec 2010 to 128.253.10.141. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

�2�10−9 Torr� that is connected to a UHV �Pbase�5�10−10 Torr� scanning tunneling microscope.

The CIPT data shown in Fig. 1 confirm that Mg–B–Obarrier layers can yield high values of TMR. Upon anneal-ing, the TMR of both symmetric CFB �Fig. 1�a�� and FCB�Fig. 1�b�� junctions increases dramatically, reaching the160%–190% range after a 350 °C annealing. There is onlyslight variation in TMR with thickness over the 1.1–1.7 nmrange. This behavior is not predicted by the theory of idealMgO MTJs, but the TMR of our thicker junctions is alsobelow these predictions.4,5 In considering the effect of an-nealing, it is useful to note that in the context of the Jullieremodel,17 increasing TMR from 25% to 160% indicates a100% increase in the tunnel current polarization �Pt� from�33% to �67%. An additional increase in TMR to 600%,the current 300 K record18 for high RA MgO MJTs, repre-sents another 28% improvement in polarization �Pt�86%�.Thus, the improvement of the spin filtering properties ofthese Mg–B–O barriers that occurs with annealing is alreadyvery good at the �160% TMR level.

Comparative in situ STS measurements of the electronicstructure of thin MgO layers grown on CFB electrodes pro-vide insight into the beneficial effect of mixing BO3 into theMgO barrier. Figure 2 compares STS results obtained from a2 nm MgO layer formed by EBD with those taken from a rfsputter deposited 2 nm Mg–B–O layer, before and afterin situ annealing. Each trace represents averaged STS datataken at multiple spots on the sample. In the EBD MgO case�Fig. 2�a��, wherein oxidation of the base electrode is mini-mal, the STS data indicate a band gap of �3 eV with noband offset, but significant band tailing to nearly the Fermilevel �0 V�, and there is only a small change, primarily in theband offset, upon annealing. Similar STS results were previ-ously reported for EBD MgO and dc-reactively sputteredMg/MgO bilayers deposited on �001� Fe, and likewise weattribute the small band gap and low energy band-tail statesof EBD MgO on CFB to lattice distortion and atomic defects

in the crystalline MgO arising from strain at the oxide-electrode interface.19

The STS data of the as-sputtered layer �Fig. 2�b�� show arelatively wide band gap ��4 eV� and greatly reduced bandtailing in comparison to EBD MgO, which we attribute to thestrain reducing effect of forming a mixed oxide with a sub-stantial BO3 component at the electrode-MgO interface. Notethat the conduction band edge of the mixed oxide is shiftedsignificantly toward the Fermi level, consistent with ourprevious x-ray photoemission spectroscopy �XPS�measurements,7 indicating that Mg–B–O has a higher workfunction than MgO. After annealing, the Mg–B–O band gapshrinks substantially to �2.5 eV, smaller than EBD MgO,and the conduction band edge becomes more abrupt, sug-gesting a more ideal barrier material forms as the mixedoxide becomes more uniform and ordered, as also indicatedby XPS �Ref. 7� and EELS �Ref. 9� studies.

The beneficial effect of incorporating BO3 into the MgObarrier promotes consideration of PyB alloys as the free elec-trode in order to enhance the formation of Mg–B–O at thetop of the barrier layer through reaction with surface oxygenon the previously deposited MgO and to possibly achievepreferred crystallization of the Py after the annealing pro-cess. As indicated by Fig. 1�c�, the use of a Py95B5 freeelectrode results in a roughly thickness independent TMR of�40% �Pt�40%� upon moderate �300 °C� annealing, buthigher temperature annealing deteriorates TMR. Junctionswith a Py80B20 free electrode �Fig. 1�d�� show a steadilyincreasing TMR with annealing and little dependence ofTMR on barrier thickness. FCB/1.1 nm Mg–B–O /Py80B20MTJs annealed to 350 °C achieve �155% TMR �Pt

�66%� and low RA ��15 � �m2�.TEM images of the two types of PyB junctions are

shown in Fig. 3. The �1.1 nm Mg–B–O barriers are poly-crystalline in both cases, but the Py95B5 and Py80B20 elec-trodes exhibit quite different crystal structures. The polycrys-talline Py95B5 electrode has some texturing in the as-growncase �Fig. 3�a�� but after annealing to 350 °C becomes lesstextured �Fig. 3�c�� as is clearly indicated by the nanometerspot size convergent beam electron diffraction �CBED� pat-tern �Fig. 3�c�, inset�. In contrast, the as-deposited Py80B20electrode is amorphous �Fig. 3�b�� as expected for such ahigh concentration of the glass-forming B component,20 butafter annealing to 350 °C, the TEM and CBED measure-

FIG. 1. �Color online� CIPT measurements of MTJ structures showingdependence of TMR vs RA on annealing for �a� all-Co60Fe20B20

electrodes, �b� all-Fe60Co20B20 electrodes, �c� Co60Fe20B20 /Mg–B–O /Py95B5 structures, and �d� Fe60Co20B20 /Mg–B–O /Py80B20 structures. Eachsample has been measured before �squares� and after annealing to 250�circles�, 300 �up triangles�, and 350 °C �down triangles� as labeled.

FIG. 2. �Color online� STS measurements of MgO and Mg–B–O layersdeposited on Co60Fe20B20 films, before �black solid lines� and after �reddotted lines� annealing to 375 °C. The EBD MgO layer �a� shows the pres-ence of low energy states down to the Fermi level �0 V� and little change inthe band gap after annealing. The rf-sputtered Mg–B–O layer �b� shows fewlow energy states before annealing with a conduction band offset of�0.8 eV, and a dramatic decrease in the band gap after annealing, with aconduction band offset of �0.6 eV.

112504-2 Read et al. Appl. Phys. Lett. 94, 112504 �2009�

Downloaded 08 Dec 2010 to 128.253.10.141. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

ments �Fig. 3�d�, and inset� reveal that it has a highly tex-tured �001� bcc crystal structure, which is optimal accordingto the coherent spin-dependent tunneling model.4,5 EELSshows that this crystallization is accomplished by any glass-forming B that is not incorporated into the oxide barrier dif-fusing out of the Py80B20 into the capping layer.9

Magnetization measurements of the various free layersindicate that, as expected from the literature,20 the MS ofCFB and FCB increases as B diffuses out of the electrodesduring annealing. However, the MS of the Py80B20 layer is, towithin measurement accuracy, the same as Py �4�MS

�9600 G� and does not change significantly with annealing.The magnetic coupling between the fixed and free electrodes,measured by the shift �Hd� from zero of the centroid of thefree layer magnetization loops of unpatterned samples, is animportant parameter in MTJs.21 This coupling is typicallyferromagnetic and attributable to thin film roughness.22 Wefind that as-deposited MTJs with FCB base electrodes haveHd values that are roughly twice those of MTJs with CFBbase electrodes. This is consistent with the former beingmagnetically rougher due to greater oxidation of the Fe richelectrode. Upon 350 °C annealing, Hd decreases in allsamples, with the most pronounced decrease apparent inFCB /Mg–B–O /Py80B20 MTJs. After annealing, these MTJswith barrier thicknesses �1.1 nm exhibit Hd�2.5 Oe, in-dicative of very magnetically smooth junctions. However, Hdis higher ��10 Oe� for the 1.1 nm barrier case, which wetentatively attribute to the onset of significant ferromagneticinterlayer exchange coupling.23

In summary, we obtain high TMR MTJs with a rf plasmaMgO deposition process that naturally utilizes the reactivityof the B component of the base electrode with adventitious Oin the process chamber to incorporate B into the Mg–B–Otunnel barrier. Consistent with the TMR results, the elec-tronic properties of Mg–B–O appear quite favorable foryielding low leakage junctions, which indicates why the rfsputter deposition of MgO onto B-alloyed base electrodes iseffective in the thin barrier, low RA regime. Using PyB alloyelectrodes, this technique yields high TMR �155%, Pt

�66%�, low RA ��15 � �m2� MTJs with Py-like freeelectrodes whose magnetic properties could to be advanta-geous for magnetic sensing and ST-MRAM applications.

The authors thank P. J. Chen and Audie Castillo of NISTfor assistance with CIPT measurements, Derek Stewart ofthe Cornell Nanoscale Facility for helpful discussions, andDaniel Worledge and Eileen Galligan of IBM, YorktownHeights and Phil Mather and Jon Slaughter of Everspin forhelpful discussions regarding CIPT measurements and formeasurements of initial MTJ structures. This research wassupported by the Cornell Center for Materials Research, aNational Science Foundation �NSF� Materials Research Sci-ence and Engineering Center. Support was also provided bythe Center for Nanoscale Systems, which is a NSF NanoscaleScience and Engineering Center, the Semiconductor Re-search Corporation, and the Office of Naval Research.

1T. Linn and D. Mauri, US Patent No. 6841395 �January 11, 2005�.2S. S. P. Parkin, C. Kaiser, A. Panchula, P. M. Rice, B. Hughes, M. Samant,and S.-H. Yang, Nature Mater. 3, 862 �2004�.

3S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, NatureMater. 3, 868 �2004�.

4W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Phys.Rev. B 63, 054416 �2001�.

5J. Mathon and A. Umerski, Phys. Rev. B 63, 220403�R� �2001�.6S. Yuasa, Y. Suzuki, T. Katayama, and K. Ando, Appl. Phys. Lett. 87,242503 �2005�.

7J. C. Read, P. G. Mather, and R. A. Buhrman, Appl. Phys. Lett. 90,132503 �2007�.

8J. J. Cha, J. C. Read, R. A. Buhrman, and D. A. Muller, Appl. Phys. Lett.91, 062516 �2007�.

9The barrier composition is Mg0.38B0.05O0.57 and the cation ratio isB/�Mg�B�0.05/�0.05�0.38��0.12, J. J. Cha, J. C. Read, R. A.Buhrman, and D. A. Muller �unpublished�.

10R. S. Dave, G. Steiner, J. M. Slaughter, J. J. Sun, B. Craigo, S. Pietam-baram, K. Smith, G. Grynkewich, M. DeHerrera, J. Akerman and S. Te-hrani, IEEE Trans. Magn. 42, 1935 �2006�.

11R. M. Bozorth, Ferromagnetism �IEEE, Piscataway, NJ, 1951�.12P. M. Braganca, I. N. Krivorotov, O. Ozatay, A. G. F. Garcia, N. C. Emley,

J. C. Sankey, D. C. Ralph, and R. A. Buhrman, Appl. Phys. Lett. 87,112507 �2005�.

13J. Z. Sun and D. C. Ralph, J. Magn. Magn. Mater. 320, 1227 �2008�.14D. Mazumdar, W. Shen, X. Liu, B. D. Schrag, M. Carter, and G. Xiao, J.

Appl. Phys. 103, 113911 �2008�.15J. C. Sankey, Y.-T. Cui, J. Z. Sun, J. C. Slonczewski, R. A. Buhrman, and

D. C. Ralph, Nat. Phys. 4, 67 �2008�.16D. C. Worledge and P. L. Trouilloud, Appl. Phys. Lett. 83, 84 �2003�.17M. Julliere, Phys. Lett. 54A, 225 �1975�.18S. Ikeda, J. Hayakawa, Y. Ashizawa, Y. M. Lee, K. Miura, H. Hasegawa,

M. Tsunoda, F. Matsukura, and H. Ohno, Appl. Phys. Lett. 93, 082508�2008�.

19P. G. Mather, J. C. Read, and R. A. Buhrman, Phys. Rev. B 73, 205412�2006�.

20C. D. Graham, Jr., and T. Egami, Annu. Rev. Mater. Sci. 8, 423 �1978�.21S. Tegen, I. Mönch, J. Schumann, H. Vinzelberg, and C. M. Schneider, J.

Appl. Phys. 89, 8169 �2001�.22L. Néel, C. R. Acad. Sci 255, 1545 �1962�.23T. Katayama, S. Yuasa, J. Velev, M. Y. Zhuravlev, S. S. Jaswal, and E. Y.

Tsymbal, Appl. Phys. Lett. 89, 112503 �2006�.

FIG. 3. Cross-sectional TEM images of as-grown CFB /Mg–B–O /Py95B5

�a� and FCB /Mg–B–O /Py80B20 MTJs �b�, and after annealing to 350 °C��c� and �d��. Insets: CBED patterns from the Py95B5 electrode �inset in �c��and from the Py80B20 electrode �inset in �d�� after annealing.

112504-3 Read et al. Appl. Phys. Lett. 94, 112504 �2009�

Downloaded 08 Dec 2010 to 128.253.10.141. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions